Citation
The development of new preparation methods for surface enhanced raman active substrates

Material Information

Title:
The development of new preparation methods for surface enhanced raman active substrates
Creator:
Sutherland, William Scott, 1962-
Publication Date:
Language:
English
Physical Description:
vii, 201 leaves : ill., photos ; 29 cm.

Subjects

Subjects / Keywords:
Colloids ( jstor )
Filtration ( jstor )
Laser beams ( jstor )
Lasers ( jstor )
Molecules ( jstor )
Raman scattering ( jstor )
Raman spectroscopy ( jstor )
Signals ( jstor )
Silver ( jstor )
Wavelengths ( jstor )
Chemistry thesis Ph. D
Dissertations, Academic -- Chemistry -- UF
City of Gainesville ( local )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1990.
Bibliography:
Includes bibliographical references (leaves 190-199).
Additional Physical Form:
Also available online.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by William Scott Sutherland.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright William Scott Sutherland. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
026076001 ( ALEPH )
24887664 ( OCLC )

Downloads

This item has the following downloads:


Full Text












THE DEVELOPMENT OF NEW PREPARATION METHODS FOR SURFACE
ENHANCED RAMAN ACTIVE SUBSTRATES



















By

WILLIAM SCOTT SUTHERLAND


















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

UNIVERSITY OF FLORIDA

1990















ACKNOWLEDGEMENTS


A number of people have played major roles in my

undergraduate and graduate development as a scientist and as

a person. From my undergraduate career at the University of

Tennessee at Chattanooga, Dr. Randolph Peterson, my honors

research advisor, and Dr. Grayson Walker, my physics advisor,

were two of the primary forces who helped me develop my

interest in the sciences. During my summers in Gainesville

working on my honors thesis (three years in the making) I had

the privilege of working closely with two University of

Florida professors: Dr. Carl Stoufer and Dr. Martin Vala. I

appreciate the help and lab space that both of these men gave

to me during those long summers, and the advice I received

from them on science and chemistry in general. I also had the

distinct honor of working with one of the most interesting and

intelligent computer programmers I have ever encountered,

George Purvis (now pioneering the continuing development of

the CAChe molecular modelling system at Tektronix), during the

summer of 1983. My professional relationship with George

continued, culminating in 1987-1988 with a year-long project

simulating the scanning tunneling microscope on a SUN work

station and the presentation of the results at the Sanibel

Symposium. This experience has helped nurture my interests in

ii









both computer programming and computer graphics. I must not

forget to acknowledge the immense influence that both my Ph.D.

advisors have had on my focus as a scientist. Although my

direct interaction with Dr. Michael Zerner has been limited by

the nature of my research, he continues to inspire me with his

enthusiasm for theoretical research and the types of problems

that can be tackled. Although it is not possible to describe

in a few words the influence that Dr. Jim Winefordner has had

on the way that I approach research, what I can say is that

his general enthusiasm for life, science, research, and people

is an inspiration to all those who cross his path, including

myself. My whole philosophy for approaching a problem, be it

in the lab or in my life, has been irreversibly altered for

the better by this man. I cannot thank him enough. Finally,

I must thank my emotional rock, Suzan Oberle, my girl friend

for the past nine years. She has seen me at my best and at my

worst, but, despite the hard times we sometimes have, she has

stuck by my side and given me support. I thank you all.

















iii
















TABLE OF CONTENTS



ACKNOWLEDGEMENTS . ... . . .. ii

ABSTRACT . ... . . . vi

CHAPTER

1 INTRODUCTION . . .. .. . .. 1

2 RAMAN SPECTROSCOPY . . . ..... 3

Raman Scattering . . . . 3
Classical Description . . . 3
Quantum-mechanical Description . 6
Raman Spectroscopy . . . . 9
Advantages of Raman Spectroscopy . .. 10
Disadvantages of Raman Spectroscopy ... 11
Enhanced Raman Spectroscopy . . ... 12
Resonance Raman Spectroscopy . .. 14
Ultraviolet Resonance Raman Spectroscopy 17
Surface-enhanced Raman Spectroscopy . 19
Surface-enhanced Resonance Raman
Spectroscopy . . .... 22
Raman Microprobe Spectroscopy . ... 23
Theoretical Considerations of SERS . ... 25
Chemical Theories . . ... 26
Quantum Mechanical Theories . .. 27
Electromagnetic Theories . ... .... 28
Corrections to the Electromagnetic Theory 32
A Working Electrodynamic Theory . ... 36
Applications of the ED theory to SERS
Experiments . . ... 42

3 SERS ANALYSIS OF SULFONAMIDES ON COLLOIDAL SILVER 49

Introduction . . . . ... 49
Background . . . .... 50
Experimental Section . . . ... 54
Instrumentation . . . ... 54
Chemicals and Procedure . . .. 55
Results and Discussion . . ... 56
Quantitative Study . . ... 63
Conclusions . . . . ... .77

iv










4 MORPHOLOGY/ACTIVITY STUDIES OF NEW SERS-ACTIVE
SUBSTRATES . .. . . 79

Introduction . . ... . 79
Experimental . . ... . . 82
Instrumentation. . ... .. 82
Chemicals and Procedure . . .. 83
Results and Discussion . . ... 84
Raman Microprobe Studies . . .. 84
Raman Macroscopic Studies . .. 100
Conclusions . . . . 114

5 SPATIAL DISTRIBUTION STUDY OF SERS ACTIVE SUBSTRATES 116

Introduction . . . .. 116
Experimental ................. 117
Instrumentation . . .... 117
Chemicals and Procedure .. . .. 117
Results and Discussion . . .. 118
Spatial Distribution Study . .. 121
Laser Spot Size Study . . .. 130
Conclusions .. . . . 140

6 COLLOID FILTRATION . .. . ... 142

Introduction . . . ... 142
Experimental . . . . 147
Instrumentation . . ... 147
Chemicals and Procedure . . 148
Results and Discussion . . . 150
Polycarbonate Track-Etched Membranes 150
Anotec Anopore Alumina Membranes . 154
Conclusions . . . . 179

7 CONCLUSIONS AND FUTURE WORK . . .. 181

REFERENCE LIST . . . . .. 190

BIOGRAPHICAL SKETCH . . . ... 200















v















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

THE DEVELOPMENT OF NEW PREPARATION METHODS FOR SURFACE
ENHANCED RAMAN ACTIVE SUBSTRATES

By

WILLIAM SCOTT SUTHERLAND

DECEMBER, 1990


Chairperson: Dr. James D. Winefordner
Major Department: Chemistry

A wide variety of surfaces have been used for surface-

enhanced Raman spectroscopy (SERS). The methods developed for

preparing these surfaces are either motivated by analytical

factors, including low cost, speed, simplicity, and

reproducibility, or by physical factors, including control of

the surface morphology and the uniformity of the surface

features. Analytical methods are focused on the application

of SERS for routine analysis, while physical methods are

focused on comparisons of experimental results to theoretical

calculations, the latter being limited to sphere and spheroid

geometries. The ideal SERS substrate would satisfy both the

analytical and physical criteria. This work details a series

of experiments which evaluate a variety of surfaces with

respect to both the analytical and physical guidelines.

Silver colloid solutions, which appear to be an ideal

vi









substrate, are used to study the ability of SERS to

differentiate between compounds with similar structures,

specifically the three primary sulfapyrimidines. These

experiments graphically illustrate that, while SERS can be

used for trace detection and speciation, the flocculation of

metal colloids upon analyte addition renders them useless for

physical studies. The next step taken was to modify the two

most common techniques for SERS substrate preparation, metal

vapor deposition and direct chemical reduction of silver

salts, by using small pore size and fiber size filter papers

(Millipore 0.025 and 0.22 gm filters) and other surfaces

(frosted slides) to provide a more uniform support for the

formation of silver features. The results are that these

substrates provide more uniform silver surface features than

other supports used previously, but that the surface features

do not satisfy the physical factors. Also, nonuniform analyte

distribution, observed for spots resulting from solutions

applied to these membranes using a syringe, contribute to

reproducibility problems, hindering the utility of these

surfaces for analytical experiments. Syringe filtration of

metal colloids is shown to be an alternative to these other

methods, providing uniform surfaces with controlled morphology

and minimizing the analyte distribution problem. This method

is also inexpensive, fast, and simple, satisfying the

analytical requirements.



vii















CHAPTER 1
INTRODUCTION



The phenomenon of surface-enhanced Raman spectroscopy

(SERS) has been known for over 15 years, and, while use of

this technique and the diversity of systems studied continue

to increase (Gerrard and Bowley, 1988), this field is still

wide open in terms of opportunities for new and unique

applications as well as extensions of existing techniques.

Only in the past few years has the emphasis of SERS research

begun to shift from the elucidation of the physical processes

underlying the phenomenon to the development of the technique

for routine analysis, including surface preparation methods,

analysis of the analytical figures of merit of these surfaces,

and the utility of SERS as a detector for separation

techniques like liquid chromatography (LC), high performance

liquid chromatography (HPLC), thin layer chromatography (TLC),

and high performance thin layer chromatography (HPTLC).

The studies described herein encompass a wide variety of

topics. Chapter 2 provides a background of Raman spectroscopy

in general, enhanced Raman spectroscopy, the theoretical

considerations involved in SERS, and a description of how

these theories can be used for optimizing metal surface

geometries. Chapter 3 explores the utility of SERS on silver

1









2

colloids for the detection of sulfonamides and describes some

of the experimental problems of colloidal SERS. Chapter 4

discusses the preparation of a number of new SERS-active

substrates using modifications of previously published

techniques. The surface morphology of the prepared substrates

is examined using the microscope from a Raman microprobe and

using scanning electron microscopy. The SERS activity of

these substrates is tested using both standard Raman and Raman

microprobe techniques. Chapter 5 is a detailed look at one of

the major contributing factors to problems in the

reproducibility of SERS signals which is observed for the

substrates in Chapter four. Chapter 6 describes a new method

developed for the preparation of SERS substrates which

potentially circumvents some of the major problems of using

colloids and the surfaces described in Chapter 4. Finally,

Chapter 7 looks at the conclusions drawn from this work and

provides a look into the future of substrate preparation

methods for SERS, outlining potential modifications of some of

the methods presented in here as well as some novel

preparation methods which could improve the control of surface

morphology, the cost of substrate preparation, and the

potential for automating the preparation method.















CHAPTER 2
RAMAN SPECTROSCOPY



Raman Scattering


Raman scattering, discovered by Sir C. V. Raman in 1928,

involves the gain or loss of energy from light scattered off

a molecule. An excellent description of scattering processes

and their relation to molecular vibrations and Raman

scattering is given by Wilson et al. (1980). In essence, when

a beam of light is passed through a transparent substance, a

small amount of the radiation is scattered. If a

monochromatic light source is used, the majority of the

scattered radiation will be at the same frequency as the

incident light (Rayleigh scattering), but a small amount will

occur at discrete frequencies both greater than and less than

the incident frequency. These scattered frequencies are

referred to as Raman scattering. There are two fundamental

ways to interpret this result: a classical description and a

quantum mechanical description.


Classical Description


The classical theory of Raman scattering revolves around

the concept of molecular polarizability. This theory is often


3









4

referred to as the wave theory, and a detailed derivation is

given by Banwell (1972) and by Skoog (1985). A brief overview

is given here. The induced dipole moment, M (C m), of a

molecule in an electric field is proportional to the

polarizability of the molecule, a (J-1 C2 m2), and to the

magnitude of the applied field, E (V m' ), as given in equation

2.1.

p-aE (2.1)



When a sample of molecules is subjected to an incident

electromagnetic field of frequency v (s ', the electric field

experienced by each molecule varies according to equation 2.2.

E-Eosin(2xv t) (2.2)



The induced dipole oscillates according to equation 2.3.

-L-aE0sin(2xvt) (2.3)



This dipole emits radiation at a frequency v, yielding

Rayleigh scattering. If the molecule undergoes a vibration,

with a frequency rvib, which induces a change in the

polarizability of the molecule, the expression for a is

modified to give:

a-aO+psin(2nviyt) (2.4)








5

where 8 is the rate of change of polarizibility of the

vibration. The oscillating dipole becomes a superposition of

these two processes.

p- (ao+ sin(2yvtibt) ) Esin(2nv t) (2.5)



Using trigonometric identities, equation 2.5 can be written in

a more revealing form.


P-aoEosin(2tv t)+ I+Eo{cos2n(v-vvib) t-cos2 (v+Vib) t (2.6)
2



From equation 2.6, two things are apparent. First, only

those vibrational modes which impose a change in the

polarizability (a nonzero value for B) are Raman active.

Second, Raman shifted bands occur at frequencies lower

(Stokes) and higher (anti-Stokes) than the incident frequency.

The classical description of Raman scattering is very useful

for developing an intuitive feel for which vibrations of a

molecule should be expected to be Raman active. Two useful

tools which can be used, at least for simple molecules, to

help determine which vibrational modes of a molecule are Raman

active are group theory (Cotton, 1971) and the polarizability

ellipsoid (Banwell, 1972). In general, symmetric vibrations

give rise to intense Raman lines and nonsymmetric vibrations

are usually weak and sometimes not observable (Banwell, 1972).

The classical description is not complete, however, since it









6

predicts that Stokes and anti-Stokes Raman peaks should be of

the same intensity, a result contradicted by experiment. In

order to rectify this inconsistency, one must turn to a

quantum description of Raman scattering.


Quantum Mechanical Description


The quantum theory of Raman scattering focuses on the

concept of quantized vibrational and electronic energy levels

in molecules. This theory is often referred to as the

particle theory and a detailed description can be found in

Wilson et al. (1980), Skoog (1985), and Banwell (1972). In

essence, the incident radiation field is considered to be a

stream of photons which are elastically (Rayleigh) and

inelastically (Raman) scattered by the molecules of the

medium. The incident radiation is chosen to have a frequency

far from any electronic resonances of the molecule of

interest. Thus, excitation is into a "virtual state" which is

not a normal electronic or vibrational state of the molecule

(Figure 2.1). It is immediately evident from Figure 2.1 why

different intensities are observed for Stokes and anti-Stokes

Raman lines. Stokes lines arise from excitations from the

ground vibrational state of the molecule, whereas anti-Stokes

lines arise from the first vibrationally excited or higher

excited states of the molecule. The population of these



























Figure 2.1. Energy-level diagram illustrating Raman scattering (a) and the resulting Raman
spectrum (b).











Stokes scattering Anti-Stokes scattering
-..---- / ---..-
v I

he h(ex ) = hvs hvex h(vex + Pu) = hv,


u= 1 v=l
uh
hv, hv,
=O0 v=0

(a)


Rayleigh
line





SStokes
o line


Anti-Stokes
line




Vex v Vex Vex + v
Frequency

(b)
0o









9

states follows a Boltzmann distribution and, at room

temperature, the majority of the molecules are in the ground

vibrational state. Thus, it would be expected that Stokes

line intensities would be much greater than anti-Stokes line

intensities. For example, the intensity difference between a

Stokes line and an anti-Stokes line (arising from the first

excited vibrational state) of a 1000 cm'1 Raman peak at room

temperature should be approximately a factor of 100. This

difference will diminish as the temperature of the system is

increased (Skoog, 1985).


Raman Spectroscopy


Raman scattering is differentiated from Raman

spectroscopy in that the former is the phenomenon observed and

the latter is the technique which makes use of this

phenomenon. A large number of papers on Raman scattering

appeared in the chemical literature shortly after its

discovery in 1928. This enthusiasm soon waned due to the

length of time (several minutes to hours) and the large amount

of sample required (typically 10-20 ml) to obtain a Raman

spectrum with conventional white light sources. The most

significant advance in Raman spectroscopy was the development

of high powered, continuous wave (cw) gas ion lasers (e.g.

Ar, Kr) in the late 1960's (Gerrard and Bowley, 1989). An

increase in the quality of monochromators during this time

period also improved the utility of Raman spectroscopy (Ingle









10

and Crouch, 1988). To attest to the increased popularity of

Raman spectroscopy, over 6000 papers were published during the

period between late 1985 and late 1987 dealing with many

applications of this technique (Gerrard and Bowley, 1988).


Advantages of Raman Spectroscopy


It is not surprising, during the resurgence of interest

in Raman spectroscopy, that comparisons were made with its

closest comparable technique, infrared absorption

spectroscopy, since both reveal information about the

vibrational modes, and therefore the structure, of the

molecule of interest. In a direct comparison, Raman

spectroscopy displays a number of distinct advantages over its

counterpart.

1. Any frequency of light can be used, especially visible
radiation, which allows the use of standard UV-visible
transducers (photomultiplier tubes, diode arrays, charge-
coupled devices) instead of thermal detectors.

2. Water is a useful solvent in Raman spectroscopy, whereas
it is usually a poor solvent for IR spectroscopy.

3. The properties of the laser sources used in Raman
spectroscopy make it relatively easy to probe a variety
of sample types, including micro-samples (using a Raman
microprobe), surfaces, films, powders, solutions, gases,
and samples of virtually any size or shape.

4. A single Raman spectrometer can cover the entire range of
vibrational frequencies, whereas several instruments are
required for this task using IR spectroscopy (or a change
of beam splitters or detectors with fourier transform
IR).

5. Raman spectra are usually much simpler than IR spectra
due to the lack of intense overtone and combination
bands.









11

6. Polarization data add extra structural information not
available using IR spectroscopy.


Disadvantages of Raman Spectroscopy


Despite all the advantages that Raman spectroscopy offers

over conventional IR spectroscopy, especially with respect to

the simplified optics and the ability to use aqueous solvents

(a must for most biological systems), IR spectroscopy is still

the technique of choice in most analytical laboratories.

There are a number of reasons for this.

1. IR spectroscopy is more sensitive to small structural
differences due to the intensity of overtone and
combination bands.

2. Extensive libraries of IR spectra have been compiled.
Similar libraries for Raman spectra exist, but are not
nearly as complete.

3. Infrared instruments are generally less expensive than
Raman instruments. The monochromators used in Raman
spectroscopy must be of a higher quality than those used
in IR spectroscopy.

4. Because Raman spectra are highly dependent on laser
power, cell geometry, and instrument characteristics, it
is much more difficult to compare Raman intensities from
one instrument to another than it is for IR spectra.

5. IR cross sections are much larger than those for Raman
spectroscopy. Thus, IR detection limits are typically
much lower than for conventional Raman spectroscopy.

6. The use of visible excitation radiation in Raman
spectroscopy can often lead to background fluorescence
from the sample or impurities, completely swamping the
Raman signal.









12




Enhanced Raman Spectroscopy


If a single reason for the lack of interest in using

normal Raman spectroscopy (NRS) for routine analysis had to be

identified, it would be the weak intensity of the Raman

scattered light. Typical Raman cross sections are on the

order of 103 cm compared to typical IR absorption cross
-18 2
sections of 10 cm (Van Duyne, 1979). Raman scattering

cross sections are small due to the weak interaction of the

incident radiation with a virtual level of the molecule.

Infrared absorption cross sections are larger since the

process involves excitation into a real level, indicating a

strong interaction between the incident radiation and the

molecule. To determine possible methods for improving Raman

signals, it is helpful to examine a simple expression for the

Raman intensity.


PP ( )Q (2.7)



IR is the Raman intensity (photons s), P is the laser power

at the sample in photons per second, p is the density (cm-2)

of adsorbed scatterers, (aa/an) is the differential Raman

scattering cross section (cm2 sr1 ), and 0 is the solid angle

(in sr) over which scattered photons are collected. A second

expression of interest gives the frequency dependence of the

Raman scattered signal,









13

do(ws) 23 4I(a (2.8)
dQ 32c4 p) (2.8)


where ws is the frequency of the Stokes Raman shifted

scattered light (cm 1) and a is the polarizibility of the

molecule. Examination of these two equations reveals that

several options exist for increasing the Raman scattered

intensity:

1. Optimize the performance of the Raman spectrometer to
increase the solid angle of collection and to reduce
optical losses.

2. Increase the density of adsorbed Raman scatterers.

3. Increase the laser power at the sample.

4. Increase the Raman scattering cross section of the
molecule of interest.

Van Duyne (1979) gives several possibilities for

increasing the efficiency of the Raman spectrometer collection

optics, but realistic enhancements using these procedures are

only one order of magnitude. Another method for increasing

the collection efficiency is to use multichannel detection

(Campion, 1983), which can decrease the acquisition time for

Raman data by as much as 103 over conventional scanning

systems. This corresponds to a signal-to-noise enhancement of

32 if equal acquisition times are used in multichannel and

scanning instruments. This alone is a major improvement for

Raman spectroscopy. A detailed comparison of scanning and

multi-channel Raman systems is given by Campion (1983). The

density of adsorbed scatterers is typically increased by









14

adding surface area to the substrate using some sort of

roughening technique (e.g. oxidation-reduction cycles for

electrode surfaces). This type of approach will provide a

maximum additional enhancement of one to two orders of

magnitude (Van Duyne, 1979). Increasing the laser power is

limited by available laser hardware and by the tolerance of

the sample to high photon fluxes. Desorption, sample damage,

and substrate damage typically begin to occur for incident

laser powers of over 1 W (Campion, 1983). The final option

for enhancing the Raman scattering intensity suggested by

equations 2.7 and 2.8 is to increase the Raman cross section.

Three methods of achieving this result are to choose molecules

with large Raman cross sections, increase the frequency of the

laser excitation to take advantage of the w4 dependence of the

Raman scattering intensity, or use a frequency which is

coincident with an electronic transition of the molecule or

molecule/surface system. The molecule to be detected is

dictated by the problem to be solved, and thus is not usually

a variable. The remaining two choices have been employed to

produce the resonantly-enhanced Raman techniques.


Resonance Raman Spectroscopy


The resonance Raman effect, first described in detail by

Placzek (Bernstein, 1979), is an increase in the Raman

scattering cross section when the incident radiation is near

or coincides with an electronic transition of the molecule.









15

Enhancements as large as 106 are due to the much stronger

interaction of the incident electromagnetic field with a real

excited state of the molecule being studied rather than with

a virtual level excited in normal Raman scattering. An

excellent theoretical treatment of the resonance Raman effect

in general is given by Bernstein (1979) and a detailed

analysis of resonance Raman spectroscopy (RRS) on complex

molecules is given by Spiro and Stein (1977). RRS became

practical for analytical use with the advent of intense Ar

and Kr+ lasers and, more importantly, tunable dye lasers in

the late 1960's (Ingle and Crouch, 1988; Skoog, 1985). With

the discrete wavelength cw lasers, the molecule to be studied

must be chosen to have an electronic transition which overlaps

with the laser wavelength. This severely limits the systems

to which the technique is applicable. The use of a tunable

dye laser (cw or pulsed) allows the excitation radiation to be

"tuned" into an electronic resonance of a molecule under

study.

In a resonance Raman experiment, when the frequency of

the excitation radiation is tuned into an electronic

transition of the molecule of interest, only those vibrations

which are coupled to that electronic transition are enhanced,

so RR spectra are typically much less complex than NR spectra

(Mathies, 1979). Resonance Raman spectroscopy has been used

on a wide variety of molecular systems. Gerrard and Bowley

(1989) describe the use of RRS with cw and pulsed dye laser









16

systems for the detection of polycyclic aromatic hydrocarbons

(PAH). Spiro (1974) describes the use of RRS with a number of

molecules of biological interest, including amino acids,

nucleic acids, lipids, membranes, and a number of biologically

important chromophores, including heme proteins, iron-sulfur

proteins, and hemerythrin. RR detection and characterization

of visual pigments and the bacteriorhodopsin chromophore in

purple membranes is discussed by Mathies (1979), and a number

of papers have been published on the use of RRS with

chlorophyll and bacteriochlorophyll (Callahan and Cotton,

1987; Heald et al., 1988). Finally, several research groups

have focused their interests on the use of RRS for the study

of biochemical systems, especially for large biomolecules

containing chromophoric groups. These chromophoric regions

are quite often the areas of biological activity in these

molecules (Hughes, 1985). Use of an excitation frequency

which corresponds to an electronic transition of the

chromophoric group yields enhancements of the Raman peaks

associated only with the bonds of the chromophore. These

spectra usually have few peaks and are not subject to

interference by the Raman bands associated with the bulk of

the molecule (Mathies, 1979; Demtroder, 1982).

The major experimental problems with RRS are fluorescence

interference, either from impurities in the sample or from the

electronic transition being probed, photolysis, and local

heating of the sample from the high laser intensities used.









17

Impurities can often be removed, but eliminating fluorescence

from the molecule itself presents a more difficult problem.

One technique to minimize fluorescence interference from the

molecule being probed is to use time-resolved Raman

spectroscopy, taking advantage of the fact that the lifetime

of the fluorescence is much longer than that of the Raman

scatter. This requires the use of a pulsed laser and

discrimination electronics, adding cost and complexity to the

experiment. A second technique is to use an optical filter to

remove the fluorescence and to pass the Raman scatter, which

is typically 50-100 nm red-shifted from the excitation

wavelength. This method is successful only if the

fluorescence intensity in the region of the Raman peaks is

negligible. Techniques which have been employed to minimize

sample heating and photo decomposition include spinning the

sample under the laser beam, the use of a flowing streams, and

the use of low-temperature techniques (Mathies, 1979).


Ultraviolet Resonance Raman Spectroscopy


The use of cw ion lasers, both independently and coupled

with dye lasers, for RRS allows for excitation wavelengths to

be chosen over the entire visible wavelength range.

Unfortunately, this precludes the use of RRS for the majority

of organic compounds since their electronic transitions occur

in the ultraviolet (Gerrard and Bowley, 1989). Although

ultraviolet resonance Raman spectroscopy (UV RRS) is not, in









18

principle, different from visible RRS, the advantages realized

by using ultraviolet excitation frequencies are significant

enough that UV RRS can be considered a separate technique.

Among these advantages are that ultraviolet excitation greatly

expands the applicability of the RR technique (Hudson, 1986)

and that fluorescence interference is greatly reduced by using

excitation wavelengths below 250 nm (Asher and Johnson, 1984;

Asher, 1988).

High power, fixed UV wavelengths can be obtained using

pulsed excimer lasers, including ArF (193 nm), KrF (249 nm),

XeCl (308 nm), N2 (337 nm), and XeF (350 nm) (Hudson et al.,

1986). The feasibility of UV RRS using a KrF laser has been

demonstrated (Lin et al., 1987). A more practical system,

designed by Hudson et al. (1986), uses fixed wavelengths

generated from a Q-switched Nd:YAG laser with frequency

doubling, frequency mixing, and a Raman shifting. By far the

most flexible system to be developed for UV RRS is also based

on the Nd:YAG laser, but adds tunability with a dye laser.

Harmonics of the fundamental of the Nd:YAG laser (1064 nm) are

obtained using frequency doubling and tripling crystals.

These wavelengths are then used to pump a dye laser, the

output of which can be doubled and mixed with the output of

the Nd:YAG laser to obtain continuously tunable wavelength

selection between 265 and 800 nm. Deep UV wavelengths (190-

220 nm) are obtained with the use of a Raman Shifter (Asher et

al., 1983). This system has been used for the detection and









19

speciation of polycyclic aromatic hydrocarbons (Asher, 1984),

where detection limits as low as 20 ppb were observed, and for

the study of compounds which are models for the peptide bond

(Dudik et al., 1985). Analytical considerations of UV RRS are

discussed in detail by Jones et al. (1985), and a more recent

review of a variety of biophysical and analytical applications

of the technique is given by Asher (1988).


Surface-enhanced Raman Spectroscopy


When Fleischmann and co-workers (1974) first observed a

greatly enhanced Raman signal from pyridine on a silver

electrode, they attributed the effect to the increased surface

area of the electrode due to electrochemical roughening from

oxidation-reduction cycles performed in their experiment. It

was not until 1977 that two groups (Jeanmarie and Van Duyne,

1977; Albrecht and Creighton, 1977) independently determined

that the enhancement was a new physical phenomenon. This

realization touched off a flurry of investigations into this

so-called surface-enhanced Raman scattering (SERS). These

studies were divided into experiments to determine which

factors affected the enhancement and the development of

theoretical models to explain the variety of experimental

observations. The theoretical considerations will be

discussed in detail later, but the consensus is that there are

three basic mechanisms, grouped into two classifications

(chemical theories and electromagnetic theories), which









20

combine to give the overall enhancement of Raman scattered

light. The chemical theories include a charge-transfer

mechanism and an active-site mechanism. Both of these imply

that SERS is highly dependent on the molecule-surface

interaction. The electromagnetic theory is based on the

enhancement of the incident electric field by the metal

surface due to the excitation of surface plasmons in the metal

electrons. This effect is independent of molecular

interactions with the surface but highly dependent on the size

and shape of the surface features, the dielectric properties

of the surrounding medium, and the molecule-surface distance.

Many of the initial SERS experiments were performed on

electrode surfaces (Howard and Cooney, 1982; Chang and Laube,

1984), since the phenomenon was first discovered on a silver

electrode. An excellent review of the status of SERS on metal

electrodes is given by Chang (1987). When surface roughness

was determined to be a contributing factor to the enhancement,

many researchers began to use metal island films. These

substrates were prepared in vacuum, eliminating any

contamination effects and allowing exact amounts of the

molecule under study to be deposited onto the metal surface

(Pockrand, 1982b). These substrates also allowed for direct

comparisons with Raman scattering on single crystal surfaces,

providing detailed information on the role of surface

roughness in the Raman enhancement (Pockrand, 1982a). A

comparison of SERS with other standard surface science









21

techniques, such as Auger electron spectroscopy, electron

energy loss spectroscopy, and low energy electron diffraction

is given by Jha (1982).

Once the SERS effect was verified, researchers began to

try other types of surfaces, other metals, and molecules

other than pyridine to see how universal the effect really

was. The growth in the number of SERS studies can readily be

seen in the paper by Seki (1986), which documents over 500

experiments which had appeared in the literature by 1985.

These experiments involved many different surfaces, including

plasma-etched quartz posts, metal coated filter membranes, and

metal coated polystyrene spheres, on a variety of metals,

including gold, copper, lithium, potassium, and mercury.

Growth in the number and diversity of SERS experiments

continues, and publications numbered over 2800 by early 1989

(Garrell, 1989).

It is interesting to note that the enhancement of optical

processes at surfaces has been shown to be a general

phenomenon (Glass et al., 1980; Weitz, et al., 1983;

Moskovits, 1985). The electromagnetic theories of the SERS

effect have led to the postulation and discovery of the

surface enhancement of such processes as infrared absorption

(Olsen and Masel, 1988), second harmonic generation

(Moskovits, 1985), electronic absorption (Glass et al., 1980),

luminescence (Wang and Kerker, 1982; Wokaun et al., 1983; Das

and Metiu, 1985; Huang et al., 1986), and photochemistry









22

(Nitzan and Brus, 1981; Gersten and Nitzan, 1985). Since it

is the incident electric field which is enhanced upon

interaction with the roughened metal surfaces, the discovery

of these other techniques is not surprising. Many of these

techniques are still in their infancy and may eventually lead

to an arsenal of surface-enhanced optical processes which can

be used together for the trace detection and speciation of

compounds adsorbed to metal surfaces.


Surface-enhanced Resonance Raman Spectroscopy


When a molecule is chosen for a SERS experiment which has

an electronic absorption which overlaps spectrally, at least

in part, with the plasmon absorption profile of the metal

surface being used, the result is a coupling of the SER and RR

phenomena to give surface-enhanced resonance Raman

spectroscopy (SERRS). In an ideal situation, the enhancement

factors are multiplicative, since the processes of SERS and

RRS are essentially independent. Enhancements of twelve

orders of magnitude can be expected. In actual SERRS

experiments, however, it has been shown that the addition of

as little as a partial monolayer of the species to be studied

alters the index of refraction in the vicinity of the metal

surface feature and shifts the enhancement profile so that the

SERS part of the SERRS enhancement is no longer maximum and

sometimes nonexistent for the excitation wavelength being used

(Zeman et al. 1987; Kim et al., 1989). In practice, the









23

overall enhancement is usually larger than either the SERS or

RRS enhancements but less than what would be predicted by

simple multiplication of the individual enhancements.

The SERR technique has a number of advantages over either

of the separate techniques. The most obvious of these is

increased sensitivity due to the coupling of enhancements. A

second benefit of SERRS has to do with interferences in the

Raman spectrum. As mentioned earlier, one of the major

drawbacks to RRS is fluorescence interference. Adsorption of

molecules to the metal surfaces used in SERS experiments

usually results in fluorescence quenching by creating

nonradiative decay pathways for the electronic excited state

(Adams et al., 1980; Weitz et al., 1982; Pineda and Ronis,

1985). The fluorescence background observed in RRS is removed

in SERRS. Both advantages are realized in SERR experiments

for the detection of rhodamine 6G (Hildebrandt and

Stockburger, 1984), crystal violet (Chou et al., 1986), a

number of other dyes (Yamada et al., 1986), and other highly

fluorescent molecules (Chambers and Buck, 1984).


Raman Microprobe Spectroscopy


The first Raman microprobe was developed by Dhamelincourt

in 1975 (Dhamelincourt et al., 1979, Adar, 1988). The idea is

simple. The laser beam to be used in the Raman experiments is

focused through a microscope objective onto the sample. The

spatial extent of the beam at the sample is usually between









24

one and five micrometers and the incident laser power at the

sample can be as large as 105 W/cm2 (Van Duyne et al., 1986).

Problems with RMS include the potential for sample damage and

the need for high concentrations of the molecule to be

studied. The first problem can be addressed using the

techniques developed for RRS, including spinning of solid

samples or flowing of liquid samples, and by using other

techniques, such as cooling of the sample (for thermal damage

problems) and placing samples in vacuum or in an inert

environment such as nitrogen (for photo-oxidation problems).

RMS has several advantages over normal Raman scattering. The

small probe area (as little as 10 jm2) allows spectra of solid

samples to be obtained selectively from a crystal of the

compound of interest, thereby eliminating any fluorescence

interferences from impurities in the sample. Fluorescence

from the compound itself is still a potential interferant.

The small probe area also allows spectra of selective

compounds in a mixture to be obtained, assuming that the

crystals of each compound can be differentiated visually

(Huong, 1986). This ability to select a particular feature of

a sample has been used in the identification of impurities on

printed circuit boards and verification of the authenticity of

ancient Chinese vases (Dhamelincourt et al., 1979).

The idea of coupling the Raman microprobe with SERS was

presented by Van Duyne et al. (1986). Calculated detection

limits were less than one attomole in the probe beam. This









25

corresponds to as little as 105 molecules. Chapter 4 presents

some of the experimental results obtained in this dissertation

research using the Raman microprobe. Estimated mass limits of

detection are of the order of 100 ag or one attomole, in

excellent agreement with the predictions of Van Duyne et al.

(1986). Van Duyne et al. (1986) also hinted at the

possibility of even lower detection limits using RMS and

SERRS. Experimental verification of the potential of the

coupling of RMS and SERRS was recently reported by Taylor et

al. (1990), who found detection limits for crystal violet of

10-12 M, or approximately 4 x 10"19 g (600 molecules).


Theoretical Considerations of SERS


During the first few years following the discovery of the

SERS phenomenon, a wide variety of theories were published

which attempted to explain the large enhancements observed.

Furtak and Reyes (1980) presented a critical review of a

number of these theoretical models, many of which have been

proven incorrect by experiments designed to test them

(Campion, 1983). The process of experimental verification of

SERS theories has led to the conclusion that there are two

basic classifications of enhancement mechanisms which

contribute to the SERS effect. As mentioned previously, these

are the chemical and electromagnetic theories. Birke and

Lombardi (1982) review the influence of surface features on

the electromagnetic SERS enhancement, and Campion (1984)









26

discusses both mechanisms and their potential enhancements.

In general, both mechanisms are believed to be responsible for

the observed enhancement, but the relative contributions of

each appears to be dependent on the molecule-substrate system

under study.


Chemical Theories


Many of the early SERS experiments indicated that the

phenomenon was not a general effect, but that only certain

types of molecules gave rise to enhanced Raman spectra. These

results led to the conclusion that a significant molecule-

surface interaction was required (e.g. chemisorption or strong

physisorption) for SERS to be observed. Of the many theories

of this type to be proposed, two have withstood experimental

scrutiny: the active-site model and the charge-transfer model

(Furtak and Roy, 1985). Pettenkofer et al. (1985) give a

detailed description of SERS active sites and Sobocinski and

Pemberton (1988) show that these sites can be probed

experimentally using laser-induced thermal desorption. The

basic premise of this model is that certain locations on a

metal surface (e.g. defects) may allow for stronger adsorption

between a molecule and the metal surface. Adrian (1982) and

Lippitsch (1984) both give an excellent description of the

charge-transfer model. In this theory, it is assumed that the

enhancement arises from charge transfer between the molecule

and the electron conduction band of the metal. Although these









27

theories can only account for one to two orders of magnitude

of enhancement, they help to explain some of the experimental

observations which are not predicted by the purely

electromagnetic theories.


Quantum Mechanical Theories


A small number of papers have appeared in the literature

which detail a quantum-mechanical treatment of the SERS

effect. Initial studies attempted to calculate the

enhancement of H2 on a Li cluster (Pandey and Schatz, 1982;

Pandey and Schatz, 1984) using time-dependent Hartree-Fock

theory. A more general quantum theory was developed by Jha

(1985). In this paper, the "overall" enhancement is treated

as two separate components. The "long range" component, where

the distance to the surface is considered to be much greater

than atomic size (i.e. outside the first monolayer) is treated

using the electromagnetic theories, described below. The

"short range" component, where the distance to the surface is

on the order of the atomic size, is treated quantum-

mechanically, where the quantities of interest are the

molecular polarizability and its modulation upon adsorption to

the surface. The latter is centered around the calculation of

the single-particle electronic Green function for the

molecule-substrate complex, but will not be discussed further.

Pettinger (1987) presents a quantum-mechanical description for

SERRS which gives the overall enhancement in a single









28

expression which combines the chemical and electromagnetic

enhancements as well as the resonance Raman enhancement for

molecules adsorbed on a surface. Overall enhancements are

calculated to be between two and eleven orders of magnitude,

in agreement with experimental data, depending on the specific

metal-molecule system being studied.

Electromagnetic theories

The most successful and most thoroughly characterized of

the many theoretical treatments of the surface-enhanced Raman

effect are the electromagnetic theories. The first

manifestations of these theories were presented by Kerker et

al. (1980) for spherical particle shapes and by Gersten and

Nitzan (1980) for spheroidal particle shapes. These shapes

are the only ones which yield nontrivial solutions to the

computation of the polarizability (Van de Hulst, 1981). The

latter is more flexible, since it allows for the calculation

of SERS enhancement factors for nonspherical particles. The

basic premise behind these theories is the same, so they will

be described as a single model. The conduction electrons of

the metal of interest are treated as a free electron gas. The

electric field of the incident radiation excites a dipolar

surface plasmon in the metal particle, which is to say that it

polarizes the electrons in the particle. This has the effect

of concentrating the field just outside the particle, leading

to an enhancement of the electric field. The electromagnetic

theories solve for the magnitude of this enhancement subject









29

to the boundary conditions of the particle shape and size.

The important point of the solution to this problem is that

these particles show a dipolar plasmon resonance frequency.

The EM SERS effect is the direct result of the excitation of

this plasmon resonance. One important assumption made is that

the diameter of the particle is much smaller (typically less

than five percent) than the wavelength of light polarizing it.

Under these conditions, the radiation field incident on the

particle is homogeneous, Rayleigh scattering dominates, and

only the dipolar surface plasmon is excited. A detailed

description of the interactions of small dielectric particles

with a radiation field is given by Van de Hulst (1981).

A brief description of the model used by Gersten and

Nitzan (1980) is useful in visualizing the limitations imposed

by the assumptions made. Figure 2.2 illustrates the basic

assumptions of this model. The surface is described as a

prolate hemispheroid protruding from a flat plane. The

spheroid has a frequency dependent complex dielectric function

E(w) and the plane is considered to be perfectly conducting.

The molecule to be studied is approximated as a point dipole

along the symmetry axis perpendicular to the plane with the

molecular dipole parallel to the symmetry axis. The incident

electric field is taken to be along the symmetry axis. Under

these conditions, the molecule is in a position for maximum

field enhancement, and enhancement factors as large as 1011

have been calculated for silver prolate spheroids (Gersten and



























Figure 2.2. Geometry for the electromagnetic calculations. The semimajor axis is a and the
semiminor axis is b. The spheroid surface is =to, and the surface passing
through the molecule is ? = "..



























Oe-
\ 3
\/

\8 /


00









32

Nitzan, 1980), much larger than the maximum enhancements of

106 which had been observed experimentally at that time.

Despite this disagreement in the magnitude of the enhancement,

the EM theory is very useful in predicting many of the trends

observed experimentally. An excellent illustration of the EM

theory is given by Aroca and Martin (1985), who use the

Gersten and Nitzan (1980) model to calculate the enhancement

factors for several metals, including silver, gold, and

indium, and aluminum, for several particle geometries. The

results for silver and gold show excellent agreement for

trends in the enhancement versus Raman shift and the optimum

wavelength of the enhancement versus particle shape observed

experimentally. Table 2.1, which contains calculated data for

aluminum, illustrates that it is possible to use the EM theory

to qualitatively dictate how to "tune" the roughness of a

given metal surface to optimize the enhancement. The

information in Table 2.1 shows that, for a given incident

wavelength, the aspect ratio (the ratio of the major to minor

axes) which yields the optimum enhancement for a particular

metal can be calculated. This feature makes this theory a

powerful tool for the experimentalist.

Correction to the electromagnetic theory

It is evident that the model described above

overestimates the Raman enhancement factor by as much as five

orders of magnitude. By looking at the assumptions made in

the model, it is clear that a more realistic picture would be









33


Table 2.1 Enhancement factors for aluminum prolate
spheroids

Excitation Resonance Vibrational Enhancement
Wavelength Aspect Ratio Wavenumber Factor
(nm) (cm) (x 10)
488 8.24 500 1.75
8.37 1000 1.68
8.51 1500 1.56
8.64 2000 1.41
8.81 3000 1.07
514.4 8.82 500 1.80
8.94 1000 1.72
9.10 1500 1.58
9.20 2000 1.41
9.40 3000 1.05
590 10.42 500 1.66
10.62 1000 1.54
10.81 1500 1.39
10.90 2000 1.21
11.07 3000 0.867
640 11.56 500 1.37

11.71 1000 1.24
11.95 1500 1.10
12.04 2000 0.926
11.95 3000 0.595

Source: Aroca and Martin (1985).









34

provided by calculating the enhancement factors for molecules

at various positions around the spheroid and averaging these

to give an estimate of the overall enhancement (Barber et al.,

1983a, 1983b). There are other, more subtle corrections,

however, which need to be made to make the EM theory more

realistic. A number of these are described by Gersten and

Nitzan (1982), including the effects of neighboring bumps,

the surrounding medium, and the size of the particle, but none

of these were incorporated into their EM theory. Laor and

Schatz (1981; 1982) explore the role of neighboring bumps on

the local field enhancement and the effects of the surrounding

medium are addressed by Barber et al. (1983). Two additional

corrections to the Gersten and Nitzan (1980) EM theory are

surface scattering and the onset of multipolar plasmon

excitations, both of which are size dependent phenomena. When

the size of a metal particle becomes smaller than the mean

free path of the electrons of the metal, the plasmon resonance

of the particle is broadened, the conduction electrons

experience surface scattering, and the electric field

enhancement is decreased (Kraus and Schatz, 1983a, 1983b).

When the size of the particle increases beyond the Rayleigh

limit, the plasmon resonance broadens, shifts to longer

wavelengths, multipolar plasmon resonances appear, and the

field enhancement decreases (Barber et al., 1983, 1983a; Kraus

and Schatz, 1983). When these two effects are incorporated

into the Gersten and Nitzan (1980) EM theory, an optimum size









35

for maximum field enhancement for a given shape of particle

emerges since both large and small particle enhancements are

damped (Wokaun et al., 1982). This corrected EM theory is

often referred to as the electrodynamic (ED) SERS theory. The

result of these corrections is illustrated by Cline et al.

(1986). This study applies the same pictorial model for SERS

calculations as Gersten and Nitzan (1980), specifically by

ignoring any surface averaging. The overall maximum

enhancement factor for silver at the tip of a prolate spheroid

with a 2:1 aspect ratio is 10 smaller than for the pure EM

theory, and the magnitude of the enhancement factor is shown

to shift to the red and decrease with increasing particle

size. It is interesting to note that gold and some of the

other metals for which these calculations were performed

(platinum, palladium, rhodium, and iridium) actually have

increasing field enhancements as the particle size increases

from the Rayleigh limit. Thus, it is possible for metals

other than silver to yield competitive enhancements when

larger particles are used in SERS experiments, often the case

with the limits of the surface preparation techniques used.

A second important point is that the metals covered in this

study can provide enhancements of at least 104 over a spectral

range from 275 nm for rhodium to 700 nm for gold, with the

wavelengths in between available by changing either the metal

or the size of the particles of a given metal. This again









36

illustrates the potential power of these computations to the

experimental spectroscopist.

A working electrodynamic theory

The most complete implementation of the electrodynamic

SERS theory to appear in the literature is given by Zeman and

Schatz (1984). This theory incorporates the damping effects

for small and large particle sizes, as well as calculating all

enhancement factors as surface-averaged over the entire

particle. Both spherical and spheroidal shapes may be chosen,

as well as the dielectric constant of the surrounding medium

(Zeman and Schatz, 1987a). An additional feature of this

model is the incorporation of the effects of the monolayer or

submonolayer of adsorbed species on the dielectric constant of

the medium near the particle surface. As mentioned earlier,

this latter effect can alter the SERS enhancement by shifting

the plasmon resonance maximum out of resonance with the

incident radiation. As an exercise to test this model, Zeman

and Schatz (1987) calculated the size and shape dependence of

the incident electric field enhancement and the SERS

enhancement for ten metals, including silver, gold, and copper

(the metals most often used in SERS experiments), as well as

lithium, sodium, aluminum, gallium, indium, zinc and cadmium.

An example of the results is shown in Figure 2.3 and Figure

2.4 for silver and lithium, respectively. These metals are

chosen to illustrate the power of this model. One of the

major unanswered questions in SERS experiments with colloidal


























Figure 2.3. Optimum field enhancement (R) and wavelength versus the semimajor axis for
specified b/a values (right side) and optimum field enhancement (R) and SERS
enhancement (E) versus photon energy for specified b/a values with b = 25 nm
(left side) for silver.











5:1 5
250
4: 4:1
200 300
3:1 D
c 150 E 200 3:1

100- 2:1 00 1.




CDrg0 41 2 :1

C)
5:1
400 E


S300 4:1
200 500331
200 -
>< 3:1 2:1
C > 400-
2:1

0 300
2.0 3.0 20 30 40 50 60 70 80
Energy (eV) SemimaJor Rxis (nm)

























Figure 2.4. Optimum field enhancement (R) and wavelength versus the semimajor axis for
specified b/a values (right side) and optimum field enhancement (R) and SERS
enhancement (e) versus photon energy for specified b/a values with b = 25 nm
(right side) for lithium.










2:1
600- 6003
500- 500
4:
400 1 : 400
S5:1 E
300 300 \
O.
200 0 200 :

100 -100 1: 1

0 0
2:
1200 E 800-

C 900 -
+- 3: 1
4: P 600
x 600 5: 1
n500 2:1
300 400


2.0 3.0 20 40 60 80
Energy (eV) Semimajor Rxis (nm)









41

silver is the apparent lack of an enhancement for unaggregated

particles (Creighton et al., 1979). This has been a topic of

much debate, with some groups stating that they have observed

a SERS enhancement for monodisperse silver colloids (Garrell

et al., 1983) while others observe nothing (Blatchford et al.,

1982). From Figure 2.3 for silver, it is apparent from the ED

calculations that the optical properties of silver are such

that the spherical field enhancement is damped. The optimum

particle radius for silver spheres is shown to be 20 nm,

yielding an overall enhancement of around 2500. As the shape

of the particle is made more spheroidal (increasing the aspect

ratio), the optimum size for maximum enhancement increases and

the resonance shifts toward longer wavelengths. Enhancements

approaching 10 can be achieved for large (100 nm major axis),

eccentric (5:1 aspect ratio) particles. However, for sizes

and shapes commonly found in SERS experiments, enhancement

factors calculated with this ED theory are typically two

orders of magnitude lower than those measured. Whether this

result is an indication that chemical enhancements are

involved in these experiments is not certain, but it does

imply that their importance cannot be ignored.

The data for lithium (Figure 2.4) shows distinct

differences from the silver data. This is most evident in the

field enhancement versus semi-major axis plots, and most

dramatically for the 1:1 ratio spheroid. For silver, the

field enhancement is small and relatively flat as a function









42

of the particle radius. The same is true for gold and copper

(Zeman and Schatz, 1987). However, a lithium 1:1 spheroid has

a strongly size-dependent field enhancement, with a very

distinct maximum at around 12 nm. The intensity of this

maximum enhancement for spheres is nearly as large as for more

eccentric geometries. This sets lithium apart from the noble

metals, where the electrodynamic theory predicts that

eccentric particles are a necessity for large SERS

enhancements (Zeman and Schatz, 1987). The significance of

these ED results is made clear by also noticing that the

wavelength for optimum resonance shifts to the red as the

particle shape is made more eccentric. Thus, for metals such

as lithium, it should be possible to get strong enhancements

from the ultraviolet to the infrared by simply changing the

particle shape and making sure that the radius is within the

resonance peak predicted by the calculations. In fact, for

all the metals and geometries tested, lithium is calculated to

have the largest enhancement for all incident wavelengths when

the size and shape are optimized, and the magnitude of this

enhancement is constant from the ultraviolet to the infrared

(Figure 2.5).


Applications of the ED Theory to SERS Experiments


The power of the electrodynamic SERS theory, as

illustrated above, in identifying which metal and surface




























Figure 2.5. Raman enhancement (Ep) (optimized with respect to size and shape) versus photon
energy.









LI
5. 0
5 0 ...... .""'.."".. .""... .. ..".. . . *...... ........ ....... .........

4.5
4.5 ss> ----".------
a-- In
, 4.0 cJ --" -,YFq U u u r.. -----" -
.. Zn lo
0 3.5
Au/
S3. O
3 G"



2. 0
2.0 ......Energy .....eV .
2.0 2.25 2.5 2.75 3.0

Energy (eV)









45

morphology will yield a large Raman enhancement for a given

excitation wavelength has a number of ramifications. One of

the most significant of these relates to the coupling of SERS

and RRS to yield SERRS. As mentioned in an earlier section,

the most important technological advance for RRS was the

development of tunable cw and pulsed dye lasers. These lasers

allow the experimenter to tune the excitation wavelength into

resonance with the molecule under study. With the ED theory,

it is now possible to "tune" the SERS surface resonance

frequency into coincidence with the molecular resonance. The

metal plasmon resonance can be reasonably broad (e.g. FWHM of

0.5 eV), so a single surface can be used for obtaining the

SERRS spectra for classes of molecules with similar electronic

absorption maxima. A small range of particle shapes would

broaden the range of wavelengths which could be used for a

given surface. Another result of the ED theory relates to

catalysis. It should be possible to determine for catalytic

metals, such as the platinum group metals, if a particle size

and geometry can be found which is both strongly SERS active

and catalytically active. Such a surface would allow for in

situ study of catalytic reactions with structural information

on the species adsorbed on the metal surface. Finally, all

metals are expected to enhance to some degree (Moskovits,

1985). The ED theory can determine the optimum surface

geometry and magnitude of the expected enhancement for any

metal. The only data required are the frequency dependent









46

dielectric function of the metal, its bulk plasmon frequency,

and its plasmon width. These data appear throughout the

literature. Tables can be found for the noble metals

(Johnston and Christy, 1972), the transition metals (Johston

and Christy, 1974), indium (Theye and Devant, 1969), and

others. Finally, several compilations of complex dielectric

function data exist for many metals (Weaver et al., 1981a,

1981b; Bass et al., 1985; Hagemann et al., 1974).

Now that the ED theory can predict an optimum surface for

a given SERS experiment, what is required experimentally is

the ability to prepare surfaces with the particle morphology

indicated by the calculations. However, the size and shape of

the particles to be prepared are not the only important

factors for practical SERS-active surfaces. A list of

important factors in the preparation of these surfaces is

given below. Ultimately, all of these factors have equal

importance, but they are grouped here in two classifications;

factors important from a purely physical viewpoint and factors

important from a purely analytical viewpoint. These are

differentiated as follows. Most experiments which detail new

SERS substrate preparation methods are either interested in

making surfaces which closely mimic the theoretical models,

with little or no concern as to the speed or cost of the

preparation, or in making surfaces which are easy, fast, and

inexpensive, with little or no concern over the control of the

actual surface morphology. Both methodologies of surface









47

preparation are concerned with strong SERS activity, but the

former is classified here as physical and the latter as

analytical strictly for comparison with the current work

presented in the following chapters.

Factors of importance to the physical studies of the

preparation of SERS-active substrates include:

1. Control of particle size.

2. Control of particle shape.

3. Density and uniformity of particle coverage.

4. Uniformity of particle sizes and shapes.

5. Compatibility of the surface morphology with
theoretical models.

The first two items in the list are directly related to the

ability to prepare a substrate which conforms to the

morphology predicted by the ED theory for optimization of a

particular molecule-metal system. Control over the density

and uniformity of the particle coverage allows for a surface

to be prepared in which the maximum number of particles can be

irradiated while keeping the distance between particles

sufficient to ensure that they are noninteracting. Control

over the uniformity of particle sizes and shapes allows a

specific geometry of metal particles to be probed.

Compatibility with the ED theory implies that particle shapes

must resemble spheres or spheroids.

Factors of importance to the analytical studies of the

preparation of SERS-active substrates include:

1. Speed of the preparation method.









48

2. Cost of the preparation method.

3. Reproducibility of the preparation method.

4. Molecular interactions with the surface.

5. Surface stability to incident laser radiation.

The speed and cost of the preparation technique are important

factors for adapting SERS to routine analysis. The technique

must also consistently produce, within a chosen tolerance, the

same surface features, since the SERS enhancement is sensitive

to geometric changes in the particle sizes and shapes (Ni and

Cotton, 1986). Molecular interactions with the surface are

important since strong adsorption is needed to keep the

molecule close to the enhancing particles. As will be seen in

Chapter 5, chromatographic interactions between the molecule

of interest, the solvent, and the surface can also potentially

affect the reproducibility of the SERS signal. Finally, the

surface must be stable to the laser radiation used, since

photothermal damage and photo-oxidation can affect the

intensity of the SERS signal (Laserna et al., in press). An

example of this effect will be presented in Chapter 4.

The remainder of this dissertation is concerned with the

development of preparation methods for SERS active substrates

and the characterization of surfaces prepared using these

techniques with respect to the guidelines stated above.














CHAPTER 3
SERS ANALYSIS OF SULFONAMIDES ON COLLOIDAL SILVER



Introduction


Metal colloid solutions present an ideal surface for the

merging of the physical and analytical interests for SERS

substrates discussed in the last chapter (Moskovits, 1985).

These solutions consist of isolated, noninteracting particles

with essentially spherical shapes, ideal for theoretical

treatment. They are quickly and readily prepared for many

metals, stable, inexpensive, and the preparation methods can

be scaled up for large quantities, making them ideal for

routine analysis. The SERS activity of silver and gold

colloid particles was first demonstrated by Creighton et al.

(1979) for pyridine. Other early studies of colloidal SERS

involved small molecules like carbon monoxide (Abe et al.,

1981) so that adsorption modes and vibrational assignments

could be easily made. More practical applications have

involved p-aminobenzoic acid (Suh et al., 1983), proflavine

(Koglin and Sequaris, 1986), and nucleic acid components (Kim

et al., 1986), all on aqueous silver sols. SERS is not

limited to silver, having been detected on aqueous sols of

gold (Lee and Miesel, 1982), copper (Creighton et al., 1983),

49









50

platinum (Benner et al., 1983), rhodium (Parker et al., 1984),

as well as silver chloride sols (Gao et al., 1984) and gold-

platinum alloy particles (Takenaka and Eda, 1985). Colloidal

SERS is not, however, limited to aqueous media, as

demonstrated by Garrell and Schultz (1985), who studied the

SERS activity of silver colloids prepared in such solvents as

tetrahydrofuran, acetonitrile, and N,N -dimethylformamide.

This has implications for the application of SERS on metals

for which colloids cannot be prepared in a stable form in

aqueous solutions. SERS has also been detected on matrix

isolated silver (Reimer and Fischer, 1984), sodium (Rzepka et

al., 1981), and potassium (Schulze et al., 1984) particles.

This chapter presents the results of the use of colloidal

silver solutions in the detection of sulfonamides, also known

as sulfa drugs. Comparisons with "standard" methods for

sulfonamide detection are made, and the possibility of using

SERS for the speciation of structurally similar sulfonamides

is discussed. Finally, some problems associated with

colloidal SERS in general are described.


Background


The discoveries of sulfonamides (sulfa drugs) and

antibiotics constitute some of the most significant medical

achievements of this century. They have therapeutic uses in

both human and veterinary medicine and in disease prevention

in livestock. By 1985, over 100 million kilograms of these









51

drugs were used annually. Over 5000 sulfa drugs, which have

the general structure H2N-C6H4-SO2-NH-R, have been synthesized

and tested. Fewer than 30 of these have proven worthy of

sustained use (Bevill, 1984). Their biological activity is

based on the inhibition of the biosynthesis of folate

cofactors in bacteria by blocking a step in the formation of

dihydrofolic acid from p-aminobenzoic acid (Sigel and Woolley,

1979). The sulfapyrimidines, where the R group is a

pyrimidine ring, have been touted as the ultimate stage of

development of sulfa drugs. Important members of this group

are sulfadiazine, sulfamerazine and sulfamethazine (Sophian,

1952).

During the 1940s, these drugs enjoyed wide spread use in

combating bacterial infections in humans and in the treatment

of diseases affecting pet and food-producing animals.

Recently, the reduction of sulfa drug use in humans has been

triggered by increased bacterial resistance to the drugs and

the development of more effective antimicrobial agents.

However, the use of sulfa drugs in veterinary medicine has

persisted because the drugs are easily administered in feed

and water, are economical, and have proven to be effective for

the treatment of livestock diseases. The use of combinations

of two or more sulfa drugs was also shown to be of therapeutic

value. During treatment with sulfa drugs, the higher the

concentration of drug that can be maintained in the body, the

greater its effect. However, increasing the drug









52

concentration also increases the side effects, the most common

of which is crystalluria, a deposition of crystals in the

renal tubules due to the limited solubility of the sulfa drugs

in water. However, when used in combination, each drug acts

independently. Since all of the sulfapyrimidines have been

shown to have the same therapeutic effect, combination therapy

allows a higher total drug concentration to be maintained in

the blood while minimizing the effects of crystalluria.

The use of sulfa drugs to promote growth and treat

diseases of livestock animals has been a major cause of sulfa

drug residues in swine marketed for human consumption. In

1973, the U. S. Food and Drug Administration (FDA) set a

tolerance of 100 ng of sulfonamide per gram of edible tissue.

Random assays of tissues for drug and agricultural chemical

residues in slaughter animals and fowl are made monthly in all

packing plants butchering for commercial sale (Bevill, 1984).

At present, both the FDA and the Food Safety and Inspection

Service (FSIS) rely on modifications of the Bratton-Marshall

(B-M) procedure (Bratton and Marshall, 1939) for monitoring

residues. The Bratton-Marshall reaction involves

diazotization of the paraminobenzene group with nitrous acid

and the subsequent coupling of the diazotized moiety with an

amine to yield a stable, purple compound which absorbs

strongly at 545 nm. One major disadvantage to this technique

is that all sulfonamides form compounds with absorption maxima

at 545 nm. The drug treatment history of slaughtered animals









53

may not be well known, and determination of which sulfa drug

or drug combinations exist in the tissue could be important.

A second disadvantage of the B-M technique is that

interference is produced when primary aromatic amines other

than sulfonamides are diazotized and yield highly colored

products. Other methods which have been developed to overcome

these problems include TLC (Sigel and Woolley, 1979),

resonance Raman spectroscopy (Sato et al., 1980) and HPLC

(Johnson et al., 1975). A comprehensive review of all the

analytical methods used for sulfonamide detection was

presented by Horwitz (1981, 1981a). A major defect inherent

to all residue assay methods in current use is that the limit

of reliable measurement is of the same order of magnitude as

the tolerance they are intended to enforce (Bevill, 1984).

Since surface-enhanced Raman spectroscopy is a relatively

new technique for trace analysis and characterization, it is

of interest to ascertain its utility in the determination of

drugs. From an analytical point of view, the most outstanding

characteristic of SERS is that it provides spectral

fingerprint capability at trace concentration levels. In

addition, fluorescence of the analyte or concomitants is not

as serious a limiting factor in SERS as it is in regular Raman

spectroscopy. Thus, while normal Raman scattering is not

competitive with the previously tested methods for the

detection of sulfonamides, SERS has the potential to be as









54

sensitive as these other techniques while also providing

structural information.

In this chapter, the surface-enhanced Raman analysis of

the three primary sulfapyrimidines (sulfadiazine,

sulfamerazine, and sulfamethazine) is reported. Silver

colloidal dispersions prepared by simple borohydride reduction

of silver nitrate are used as substrates. The results

illustrate the potential of SERS for trace analysis and

characterization of analytes with close structural properties.


Experimental Section


Instrumentation

The spectra reported here were obtained using the 514.5

nm line of an argon ion laser (Spectra Physics, series 2000).

The nonlasing plasma lines were removed with a predispersing

grating. The power at the sample was typically 100 mW.

Spectra were obtained with a 0.85 m double grating

spectrometer (Spex Industries, model 1403). The spectrometer

resolution was set to 10 cm The detector was a cooled RCA,

model C31034, gallium arsenide photomultiplier tube coupled to

standard photon counting electronics. Data acquisition,

storage, processing, and plotting were under control of a

personal computer. All spectra reported represent single

scans and are provided without spectral smoothing. Right

angle geometry was used for Raman sampling. A JEOL JSM-35C

scanning electron microscope was used for Figure 3.1.









55

Chemicals and Procedure

All chemicals used were analytical reagent grade or

equivalent. Demineralized water was used throughout.

Sulfadiazine, sulfamerazine, and sulfamethazine were purchased

from Sigma and used without further purification. Stock

aqueous solutions (100 Ag mL" ) contained 3% methanol to help

dissolve the drugs. Silver hydrosols were prepared as

follows: a silver nitrate solution was prepared by adding 65

mL water to 35 mL of a 5.0 x 103 M aqueous AgNO3 solution.

This solution was ice-cooled and then added dropwise with

vigorous stirring to a sodium borohydride solution prepared by

adding 120 mL water to 180 mL of a 2.0 x 103 M aqueous NaBH4

solution. The NaBH4 solution was also ice-cooled. The

resulting silver hydrosol had an absorption maximum at 400 nm,

characteristic of silver particles with diameters between 1

and 50 nm (Creighton et al., 1979). This was experimentally

verified by filtering a small amount of the silver colloidal

solution onto a polycarbonate track-etched (PCTE) membrane and

imaging the resulting surface with scanning electron

microscopy (Figure 3.1). The silver colloid particles in this

photograph have diameters between 5 and 25 nm. This was

possible since the smooth surface of the PCTE membrane yields

high contrast and is ideal for SEM imaging of small objects

(Brock, 1983).

Each sample was prepared by adding 0.1 mL of the sulfa

drug solution to 0.9 mL of the Ag colloid at room temperature.









56

After mixing, 0.5 mL of the resulting sample was transferred

to the liquid cell and placed in the Spex model 1459

illuminator.


Results and Discussion


The surface-enhanced Raman spectra of sulfadiazine,

sulfamerazine and sulfamethazine are shown in Figure 3.2. The

numerical values of the band positions in cm'1 are listed in

Table 3.1. The extreme structural similarity of the three

drugs, which only differ in the presence of methyl group in

the pyrimidine ring of the molecule, should be noted. The

spectral fingerprinting capability of SERS at trace

concentration levels is illustrated by the following analysis

of the spectra.

The drugs have several common spectral features. There

is a narrow, very strong peak at 1114 cm'1 (sulfadiazine and

sulfamerazine),or at 1112 cm"1 (sulfamethazine), a strong band

at 1594 cm a broad band peaking at around 3390 cm with

shoulders at 3070 cm'1 and 3258 cm" and a peak at around 580
-1
cm is also present. Sulfadiazine and sulfamerazine have a

shoulder at 230 cm-1 on the sloping background of the Rayleigh

scatter. In the case of sulfamethazine, there is a well

defined band at 238 cm1.

Sulfamerazine and sulfamethazine have several spectral

singularities with respect to sulfadiazine. Both have a

common mode at around 2930 cm'1 not observed for sulfadiazine.




























Figure 3.1. Scanning electron micrograph of the silver colloid solution filtered onto
polycarbonate track-etched membrane with a pore size of 0.01 Am.






58















i:


























Figure 3.2. Surface-enhanced Raman spectra of sulfadiazine (top), sulfamerazine
(middle), and sulfamethazine (bottom) on colloidal silver.



















HCH,





0




I)
CL




nSO'NH-





Ramon Shift (cm-1)










61


Table 3.1. SERS peaks of sulfonamides on colloidal silver.
(Raman shift', cm' )

Sulfadiazine Sulfamerazine Sulfamethazine
230, sh 230, sh 238, a
294, sh 290, sh
364, vw 342, vw
416, vw 410, vw 396, vw
458, vw 462, vw 458, vw
506, vw 508, vw 516, vw

544, vw 554, w
574, m 582, m 590, m
634, w 626, vw 638, vw
684, w 674, w 674, w

762, a 738, m
818, s 830, m 834, m
974, vw 976, vw 968, bb

1002, vw
1068, sh 1066, sh 1054, w
1114, vs 1114, vs 1112, vs
1186, vw 1174, vw 1186, vw

1252, vw 1260, vw
1312, m 1302. vw
1340, m 1356, sh

1390, m 1394, m
1410, w

1454, w
1502, w 1502, w 1506, w
1594, a 1594, a 1594, a
1630, sh 1630, sh 1622, sh

2930, m 2922, m
3050, sh 3068, sh 3074, sh
3258, bb 3258, bb 3242, bb
3378, bb 3396, bb 3380, bb

a. v, very weak; w, weak; m, medium; s, strong; vs, very strong;
sh, shoulder; bb, broad band









62

The peaks at 738 cm (sulfamethazine) and 762 cm1

(sulfamerazine) do not appear in the case of sulfadiazine.

Sulfamerazine shows a weak mode at 1454 cm which is not

shown by the other two drugs. Sulfamethazine can be

distinguished from the other two drugs by a number of spectral

features. For instance, it shows a broad band at 968 cm ,

which in the other two drugs appear as very small peaks at

around 975 cm Also, the combination of a band at 1394 cm'

with a shoulder in the high energy side at 1356 cm"1 is unique

to sulfamethazine.

No attempts were made to provide a complete vibrational

analysis of the compounds. However, on the basis of group

frequencies, tentative assignments of the most prominent

spectral features can be made. The very strong 1112-1114 cm-1

band common to the three drugs could be an SO2 symmetric

stretch (Dollish et al., 1974; Baranska et al., 1987). The

1594 cm' mode in the three compounds could be a ring

stretching mode. The shoulder at around 1630 cm'1 could be an

NH2 bending mode. The peaks between 3050-3074 cm'I could be

aromatic C-H stretching modes, while those at 3258-3380 cm-

could be due to N-H stretching modes. The band at 2930 cm1

in sulfamerazine and that at 2922 cm1 in sulfamethazine are

presumably due to the methyl C-H stretching vibration, since

no band in this region appears in the spectrum of

sulfadiazine, which lacks a methyl group on the pyrimidine

ring.









63

The normal Raman spectra of the pure drugs (solid state)

were also recorded (Figure 3.3). Peak positions are listed in

Table 3.2. In general, the vibrational modes in the SERS

spectra are shifted with respect to the NRS. The most

prominent peak in the NR spectra occurs at 1150 cm

(sulfadiazine), at 1153 cm'1 (sulfamerazine), and at 1144 cm"1

(sulfamethazine), while the corresponding peak in the SERS
-1 -1 -1
spectra appears at 1114 cm 1114 cm and 1112 cm ,

respectively. There were smaller shifts for the second most

intense mode: 1598 cm1 (sulfadiazine) and 1596 cm

(sulfamerazine and sulfamethazine) in the NRS, and 1596 cm1

for the three drugs in the SERS spectrum. The bandwidths

(FWHM) of the solid state spectra are smaller than those in

SERS. For instance, the FWHM for the NRS band of sulfadiazine

at 1598 cm was 12 cm while the same vibrational mode in

the SERS spectrum (1594 cm ), has a FWHM of 22 cm It

should be noted, however, that the NRS are superimposed on a

strong background, presumably due to fluorescence of the drug

or of some impurity.

Quantitative Study

Experiments were conducted to study the dependence of

SERS intensity on the adsorbate concentration added to the

silver colloid. It was observed that for the drugs studied

there was a dynamic component in the measured SERS intensity.

Addition of the drug to the hydrosol did not result in an

immediate development of the intensity. Figure 3.4 shows the



























Figure 3.3. Normal Raman spectra of the sulfa drugs (crystalline, pure compounds). a)
sulfadiazine; b) sulfamethazine; c) sulfamerazine.


























0
--




c-e












200 400 600 800 1000 1200 1400 1600
Raman Shift (cm-1)










66


Table 3.2. Normal Raman peaks of sulfonamides on colloidal silver.
(Raman shift', cm )

Sulfadiazine Sulfamerazine Sulfamethazine
229, sh 237, w 248, vw
286, w 282, dbl, w 290, vw
319, vw 300, dbl, w 330, vw
339, vw 328, w 342, vw
373, vw 380, w 386, vw
402, vw 456, w 439, w
455, vw 547, m 526, vw
544, w 585, w 538, w
578, vw 634, w 567, vw
636, w 659, vw 579, w

661, vw 681, w 637, w
706, vw 715, vw 685, w
750, vw 743, m 716, vw
800, w 769, sh 829, dbl, m
827, dbl, m 822, dbl, m 842, dbl, m
849, dbl, m 840, dbl, m 873, vw
939, vw 892, w 999, w
995, m 999, mw 1092, w
1098, s 1094, s 1144, vs
1150, vs 1153, vs 1306, vw
1259, w 1191, w 1345, vw
1312, w 1241, vw 1385, vw
1338, w 1298, w 1416, vw
1407, vw 1333, w 1440, vw
1438, w 1371, vw 1466, vw
1505, w 1413, w 1478, vw
1581, sh 1506, m 1509, w
1598, vs 1560, w 1559, vw

1596, vs 1596, vs
1629, w 1640, w

a. v, very weak; w, weak; m, medium; s, strong; vs, very strong;
sh, shoulder; bb, broad band









67

time dependence of SERS intensity for sulfadiazine. There is

a rapid increase in intensity during the first minute after

mixing hydrosol and drug (colloid activation) followed by a

slow decrease in intensity. Activation of silver hydrosols

has been reported previously (Laserna et al., 1987a; Berthod

et al., 1987; Blatchford et al., 1982; Kerker et al., 1984).

All subsequent experiments were made with careful attention to

the timing of the measurement, with spectral scans starting 30

s after mixing hydrosol and drug.

Figure 3.5 shows the SERS spectra of sulfamerazine at

three different concentrations. As shown, at lower

concentrations, the SERS spectral features are poorly defined.

Several bands can be seen in the spectrum corresponding to 1

ng mL of sulfamerazine, including two apparent peaks between

600 cm' and 1100 cm'1 and a broad band between 1200 cm"1 and
-1
1700 cm These features remain constant for all

concentrations of the sulfa drugs used and are also present in

the blank. Possible sources are contaminants, luminescence

background of the metal substrate (Heritage et al., 1979), or

graphitic carbon (Cooney et al., 1983). These bands may

affect limit of detection of sulfa drugs by colloidal SERS,

although the 1112 cm-1 peak, which would be used for trace

detection of these drugs, does not overlap any of these

background features. Figures 3.6 and 3.7 show the

concentration dependence of sulfamethazine and sulfadiazine,

respectively.

























Figure 3.4. Time dependence of the SERS intensity of 1 ppm sulfadiazine on colloidal silver,
monitored at a Raman shift of 1112 cm .












SERS Intensity (Arbitrary)





0-





O
0-




0. -



0-
Ho
30









CD
C,





O"
O-








06
0-
O-





69


























Figure 3.5. Surface-enhanced Raman spectra of sulfamerazine obtained at different drug
concentrations. The relative intensities of the spectra represent true values.























S1 ppb




II1 1111 11 11111F|11 1 I IIIII 111111111 II III111 l III III I IT111111111 IIT
200 400 600 800 1000 1200 1400 1600
Raman Shift (cm-l)


























Figure 3.6. Surface-enhanced Raman spectra of sulfamethazine obtained at different drug
concentrations. The relative intensities of the spectra represent true values.


















>Q
S100 ppb





-4-
10 ppb

C:
-4-j









200 400 600 800 1000 1200 1400 1600
Raman Shift (cm-1)


























Figure 3.7. Surface-enhanced Raman spectra of sulfadiazine obtained at different drug
concentrations. The relative intensities of the spectra represent true values.















CI
^JL. 100 ppb







S10 ppb
._Q








)o 1 ppb
CY
Li.
,i



g ig hug i llg "11111 Ji 1 1111i1111 i li i11111111 ll 1f ill t1111111 111111ri
200 400 600 800 1000 1200 1400 1600
Raman Shift (cm-1)
U1









76

Table 3.3 summarizes the linear least squares fit of the

calibration curves for the sulfa drugs. The SERS

intensitieswere measured at the most prominent peak of each

drug, i.e., 1114 cm1 for sulfadiazine and sulfamerazine, and

1112 cm1 for sulfamethazine. Although other peaks could be

chosen to monitor the intensity changes with concentration,

this peak has the following advantages: (a) it provides the

largest dynamic range, 2 orders of magnitude for each drug;

(b) it allows the maximum sensitivity (slope), since the

concentration dependence of the intensity is the largest; and

(c) the limit of detection is improved since the selected peak

is superimposed on a much lower background than that for the

other peaks. For instance, Figure 3.5 shows that the peak at

1114 cm in sulfamerazine is over a lower background than the

second most intense peak, at 1594 cm Virtually the same

arguments can be used for the choice of the monitoring Raman

shifts for the other two drugs. Table 3.3 shows that the

sensitivity changes with the drug considered. The sensitivity

is largest for sulfamerazine and smallest for sulfamethazine.

The correlation coefficients were adequate in the three cases.

Limits of detection calculated for a signal-to-noise ratio of

3:1 are 1 ng mL1 for sulfadiazine and sulfamerazine and 10 ng

mL1 for sulfamethazine.









77

Table 3.3. SERS Calibration data for sulfapyrimidines on
colloidal silver.

Calibration r
equationa


Sulfadiazine y = 66.74 x + 679.8 0.997
Sulfamerazine y = 143.82 x + 771.2 0.999
Sulfamethazine y = 16.62 x + 751.3 0.999


a. Linear least squares best fit line. Y represents the
SERS intensity in counts per second at 1 14 cm
(sulfadiazine, sulfamerazine) or 1112 cm
(sulfamethazine). X represents the drug
concentration in ppb. The concentration range used
for all three least squares calculations was 1 ppb to
100 ppb.

b. Correlation coefficient.



Conclusions


These results demonstrate that silver hydrosols can be

developed as a practical substrate for the SERS detection of

sulfonamides. They are an attractive approach because of the

ease of production and manipulation. However, SERS

intensities in this substrate are determined by an activation

period which must be studied and optimized in each

experimental situation. Chapter 6 discusses some practical

solutions to the colloid instability problem. In addition,

extremely high sensitivity as well as quantitative information

can be obtained with careful attention to the timing of the

measurement process.









78

This study constitutes a laboratory test on pure sulfa

drugs. Although limits of detection for the colloidal SERS

technique are found to be below 10 ng mL the limit of

reliable measurement, used by the FDA and FSIS, may be

somewhat larger. Methods for obtaining the sulfa drugs from

real tissue samples that is compatible with the colloidal SERS

experiment need to be developed. Several of these methods are

described by Horwitz (1981, 1981a). Also, as is the case with

TLC, HPLC, and all methods besides the Bratton-Marshall

technique, collaborative studies among different laboratories

need to be performed in order to ascertain the ultimate

usefulness of colloidal SERS detection in the monitoring of

sulfa drugs in edible animal tissue.















CHAPTER 4
MORPHOLOGY/ACTIVITY STUDIES OF NEW SERS-ACTIVE SUBSTRATES



Introduction


When analytical chemists turned their attention toward

the possibility of using the technique of surface-enhanced

Raman spectroscopy for routine analysis, one of their major

concerns was the lack of preparation methods for SERS active

substrates. The most important factors for such methods were

that they be inexpensive, simple, and fast, with reproducible,

strong SERS activity, especially at the argon ion laser

wavelength of 514.5 nm, the most common excitation source for

routine Raman spectroscopy. Little concern was given to the

details of the surface morphology of the prepared substrates

and their ability to be modeled with the electrodynamic

theory. These analytically motivated surface preparation

methods include the coating of fumed silica with silver by

vacuum vapor deposition (Alak and Vo-Dinh, 1989), the coating

of silver onto frosted glass slides using a Tollen's reduction

of silver nitrate (Ni and Cotton, 1986), the coating of silver

onto smooth microscope slides by direct chemical reduction

using both the Rochelle salt process and the Brashear process

(Boo et al., 1985), and the coating of silver onto Whatman #1


79









80

cellulose filter paper (Berthod et al., 1988; Laserna et al.,

1988a), and Whatman #5 cellulose filter paper (Laserna et al.,

in press) by direct chemical reduction of an aqueous silver

nitrate solution with an aqueous sodium borohydride solution.

Another method of note involves evaporating silver onto

polystyrene spheres which have been coated onto Whatman #50

filter paper (Vo-Dinh et al., 1984) and glass slides (Moody et

al., 1987). This method allows for the control of particle

size by the choice of the polystyrene sphere size and control

of the particle density by control of the initial

concentration of spheres in solution. Experimental

optimization of the sphere size, density, and coating

thickness is given by Vo-Dinh et al. (1989a) and this surface

has been shown to have practical use in the detection of

organophosphorous chemical agents (Alak and Vo-Dinh, 1987) and

nitro polycyclic aromatic hydrocarbons (Enlow et al., 1986).

Two of the more unique surface preparation methods, mentioned

here for completeness, involve the observation of a spectra

sensitizing dye adsorbed to silver bromide crystals in a

photographic film (Brandt, 1988), and the detection of SERS on

silver nitrate photoreduced in solution (Ahern and Garrell,

1987).

A number of researchers have attempted to develop methods

to prepare metal substrates with control over the particle

shape and size. These physically motivated surface

preparation methods include the microlithographic production









81

of ordered arrays of quartz used as templates for evaporated

silver (Liao, 1982), stochastic arrays of quartz posts used as

templates for evaporated silver (Meier et al., 1985; Vo-Dinh

et al., 1986), and island films of metals such as silver and

indium (Jennings et al., 1984). Although these methods

produced some of the most strongly SERS-active substrates

prepared, the time involved in using these methods (e.g.

several hours for 1 cm2 for the ordered quartz post array) is

prohibitive for routine analysis. Two excellent reviews of

the relative merits of all of the substrates mentioned above

have been published (Vo-Dinh et al., 1988; Vo-Dinh, 1989).

The common element in the majority of the above surface

preparation methods is the evaporation of metallic silver or

the reduction of a silver salt onto a solid support. These two

techniques allow for the rapid and inexpensive deposition of

silver onto a surface. The most difficult aspect of these

methods is the control of the size, shape, uniformity, and

density of the silver surface features deposited. Choice of

the support to be used is one of the major determining factors

in obtaining this control.

This chapter describes the use of these two techniques on

a variety of new solid supports with the intent of improving

the uniformity and density of the silver particles deposited

onto the surface. The solid supports used include two small

pore size cellulose filter membranes from Millipore (0.025 Am

and 0.22 Mm), frosted glass slides, and a native silver filter









82

membrane. These substrates and deposition methods were chosen

since they are fast, inexpensive, and simple, making them

ideal candidates for routine use.


Experimental


Instrumentation.

The instrumentation used for these experiments was

described in Chapter 3. In addition, micro Raman analysis was

conducted with a Spex Micramate microprobe, consisting of a

modified Zeiss 20 research-grade microscope fitted with a 10X

objective, for coarse sample adjustment and location with the

X-Y microscope translation stage, and a 40X objective, for

fine tuning and analysis. The spatial resolution with the 40X

objective is 2 Am. The sample scatter is collected at 1800

and focused on the entrance slit of the spectrometer by

spatial filtering and a floating lens. The sample viewing

system consisted of a color television camera mounted to the

microscope and a 10-inch color monitor. The TV color system

can be adjusted to introduce artificial colors to enhance

features of poor-contrast samples. Switching from sample

viewing to spectral analysis was done with a rotating prism.

A Balzers, model Micro BA-3, high vacuum coating unit was used

for evaporating of silver on the supports. Photographs of the

substrates as they appear under the microscope were taken with

a 35 mm camera directly off the video camera monitor.









83

Scanning electron micrographs were taken using a JEOL JSM-35C

SEM.

Chemicals and Procedure

All chemicals used were analytical reagent grade or

equivalent. Demineralized water was used throughout. 9-

Aminoacridine hydrochloride monohydrate was purchased from

Aldrich and was used without further purification. Fisher

Finest frosted glass microscope slides were obtained from

Fisher. The Millipore filter papers (0.025 Am and 0.22 Am

pore size) and the silver membranes (0.45 pm pore size) were

obtained from Millipore.

Chemical reduction of silver salts on the substrates was

done using a published method (Laserna et al., 1988a). The

filter paper was immersed in a 0.1 M aqueous solution of

silver nitrate and sprayed with a 0.2 M aqueous solution of

sodium borohydride from a distance of about 20 cm for 30 s.

An atomic absorption nebulizer actuated by compressed air was

used for spraying. Silver evaporation was performed by

mounting the filter paper in the vacuum coating unit with

double-stick tape and resistively heating a small amount of

silver onto the surface of the paper or slide. Silver

thickness was determined by weighing the filter papers before

and after coating. The estimated thickness of the silver

layer using this method is between 250 nm and 1 Am.









84




Results and Discussion


Raman Microprobe Studies


All of the initial studies were performed using the Raman

microprobe. Since both vapor deposition and chemical

reduction have been shown to produce SERS-active substrates on

cellulose filter papers (Laserna et al., 1988a; Vo-Dinh et

al., 1987) and frosted glass slides (Ni and Cotton, 1986), the

important factor to be considered here is how these new

supports affect the silver surface morphology and the

uniformity of the silver coverage, as well as the SERS

activity. The Raman microprobe is used in these studies since

it probes a spatial area with a diameter of 2 gm, so it is

much more sensitive to submicron substrate uniformity than

normal Raman experiments which typically use a beam diameter

of between 0.5 mm and 1 mm. In fact, it has been shown that,

on a Whatman #5 filter paper substrate coated using chemical

reduction, where the silver particles have a wide range of

sizes and shapes (Laserna et al., 1988a), using a larger probe

beam actually averages the signal from a large number of these

particles. The result was a sample to sample reproducibility

of 15 percent (Laserna et al., 1989).

For all the substrates presented in this chapter, the

surface uniformity is described for two levels of detail. The









85

first is the uniformity with respect to what is seen through

the microscope and is shown through photographs taken of the

surfaces off the TV monitor output of the camera attached to

the microscope. The viewing area is approximately 110 pm by

90 Am, and the smallest discernible surface feature is on the

order of 1-2 pm. The second is the submicron uniformity of

the surface morphology and is shown in the scanning electron

micrographs of each surface. Finally, the relative SERS

activity of each substrate is tested with 9-aminoacridine

(AA), which has been shown to be an excellent SERS scatterer

(Laserna et al., 1989).

Vacuum vapor deposition

Millipore 0.025um filter membrane. This membrane is a

mixture of cellulose acetate and cellulose nitrate in a ratio

of 85:15. One side of the filter is much more reflective than

the other. This is a feature of how the membranes are made,

but the two surfaces have no significant physical differences

(Brock, 1983). For this study, the more reflective side was

coated with silver. After coating, the surface resembles a

mirror, indicating a thick, uniform coating. However, the

microscope monitor photograph of this surface seen in Figure

4.1 reveals that the surface coating is far from uniform.

Spotting of 2 AL of a 30 ppm aqueous solution of 9-

aminoacridine results in a final spot diameter on the surface

of 1 cm. Figure 4.1 also shows the SERS spectra of AA from

two different areas on this spot. The intensity difference



























Figure 4.1. Millipore cellulose filter membrane, 0.025 Am pore size, coated with silver by
vacuum vapor deposition.









Microscope Image: Coated SERS Spectra

10M0



Yi
c


2000


Roman Shift (cm-1)

SEM: Uncoated SEM: Coated





At









88

between the two spectra is an indication of the difficulty of

obtaining reproducible intensities for a given analyte spot.

The SEM images reveal that the surface is a combination of

flat surfaces and holes, with the flat surfaces often

extending over tens of microns. This may explain the peppered

appearance of the coated membrane under the microscope. The

SEM images also indicate that the silver coverage is uniform

on these flat surfaces but relatively featureless. The SERS

activity of this surface is smaller than for some of the other

substrates prepared, a result which is not surprising given

the lack of distinct surface roughness features.

Millipore 0.22 um filter membrane. This membrane is also

a mixture of cellulose acetate and cellulose nitrate (85:15)

and exhibits the same difference in the reflectivity of the

two sides as does the 0.025 jm membrane. Once again, the more

reflective side was coated and resembles a mirror. Figure 4.2

shows that this surface is very uniform when viewed by the

microscope. 2 AL of the 30 ppm AA solution was applied, with

a final spot diameter of 1 cm. Figure 4.2 also gives SERS

spectra from several places within the analyte spot. Once

again, the intensity differences observed in these spectra

indicate difficulties in the reproducibility of peak

intensities. The SEM images reveal the fibrous nature of the

membrane and show that the silver appears to be uniformly

coated over all the fibers. This surface is somewhat more

active than the 0.025 pm membrane, and this is not surprising



























Figure 4.2. Millipore cellulose filter membrane, 0.22 Am pore size, coated with silver by
vacuum vapor deposition.







Microscope Image: Coated SERS Spectra


S500000
50000
o.





Roman Shift (cm-1)
SEM: Uncoated SEM: Coated










Jlow



























Figure 4.3. The effects of laser damage on a millipore cellulose filter membrane, 0.22 pm
pore size, coated with silver by direct chemical reduction. Spectra a and b
correspond to data obtained before and after laser damage.









Microscope Image: New SERS Spectra
60000

50000

l-40000

CU
30M



10000


Raomon Shift (cm-1)
Microscope Image: After Scan




!!!





o









93

considering the increased submicron roughness features and

"cracking" of the metal coating on the fibers. One

experimental problem noted with this surface was that the high

power of the laser beam, which can exceed 105 W/cm2 (Van Duyne

et al., 1986), damaged the silver coating. Figure 4.3 shows

the spectra of a 2 AL spot of the 30 ppm AA solution for two

scans taken consecutively from the same spot. The laser

damage after one scan, as seen in Figure 4.3, results in

approximately a fifty percent decrease in the SERS intensity,

but the quality of the second spectrum is still quite good.

This indicates that the surface is still strongly SERS active.

It is interesting to note that the damaged area extends over

a diameter of 60-70 Am, even though the laser beam diameter at

the surface is only 2 Am. This damage is likely due to either

a local heating effect or some photo-oxidation process, but it

was not possible to confirm either of these.

Frosted glass slides. It was hoped that the roughness of

the frosted glass would induce silver surface roughness

features compatible with strong SERS activity, as was

apparently the case for Ni and Cotton (1986). Figure 4.4

shows the coated slide and graphically indicates one of the

major problems with this surface. The glass particles are

quite large, inhibiting the focusing of the microscope on a

particular area of the surface. Another practical problem

with this surface is that the 2 ML spot of 30 ppm AA applied

to the surface spread into an area with a diameter of over 3




Full Text

PAGE 1

THE DEVELOPMENT OF NEW PREPARATION METHODS FOR SURFACE ENHANCED RAMAN ACTIVE SUBSTRATES By WILLIAM SCOTT SUTHERLAND A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1990

PAGE 2

ACKNOWLEDGEMENTS A number of people have played major roles in my undergraduate and graduate development as a scientist and as a person. From my undergraduate career at the University of Tennessee at Chattanooga, Dr. Randolph Peterson, my honors research advisor, and Dr. Grayson Walker, my physics advisor, were two of the primary forces who helped me develop my interest in the sciences. During my summers in Gainesville working on my honors thesis (three years in the making) I had the privilege of working closely with two University of Florida professors: Dr. Carl Stoufer and Dr. Martin Vala. I appreciate the help and lab space that both of these men gave to me during those long summers, and the advice I received from them on science and chemistry in general. I also had the distinct honor of working with one of the most interesting and intelligent computer programmers I have ever encountered, George Purvis (now pioneering the continuing development of the CAChe molecular modelling system at Tektronix) during the summer of 1983. My professional relationship with George continued, culminating in 1987-1988 with a year-long project simulating the scanning tunneling microscope on a SUN work station and the presentation of the results at the Sanibel Symposium. This experience has helped nurture my interests in ii

PAGE 3

both computer programming and computer graphics. I must not forget to acknowledge the immense influence that both my Ph.D. advisors have had on my focus as a scientist. Although my direct interaction with Dr. Michael Zerner has been limited by the nature of my research, he continues to inspire me with his enthusiasm for theoretical research and the types of problems that can be tackled. Although it is not possible to describe in a few words the influence that Dr. Jim Winefordner has had on the way that I approach research, what I can say is that his general enthusiasm for life, science, research, and people is an inspiration to all those who cross his path, including myself. My whole philosophy for approaching a problem, be it in the lab or in my life, has been irreversibly altered for the better by this man. I cannot thank him enough. Finally, I must thank my emotional rock, Suzan Oberle, my girl friend for the past nine years. She has seen me at my best and at my worst, but, despite the hard times we sometimes have, she has stuck by my side and given me support. I thank you all. iii

PAGE 4

TABLE OF CONTENTS ACKNOWLEDGEMENTS ii ABSTRACT vi CHAPTER 1 INTRODUCTION 1 2 RAMAN SPECTROSCOPY 3 Raman Scattering 3 Classical Description 3 Quantum-mechanical Description 6 Raman Spectroscopy 9 Advantages of Raman Spectroscopy 10 Disadvantages of Raman Spectroscopy 11 Enhanced Raman Spectroscopy 12 Resonance Raman Spectroscopy 14 Ultraviolet Resonance Raman Spectroscopy 17 Surface-enhanced Raman Spectroscopy 19 Surface-enhanced Resonance Raman Spectroscopy 22 Raman Microprobe Spectroscopy 2 3 Theoretical Considerations of SERS 25 Chemical Theories 26 Quantum Mechanical Theories 27 Electromagnetic Theories 28 Corrections to the Electromagnetic Theory 32 A Working Electrodynamic Theory 3 6 Applications of the ED theory to SERS Experiments 42 3 SERS ANALYSIS OF SULFONAMIDES ON COLLOIDAL SILVER 49 Introduction 49 Background 50 Experimental Section 54 Instrumentation 54 Chemicals and Procedure 55 Results and Discussion 56 Quantitative Study 63 Conclusions 77 iv

PAGE 5

4 MORPHOLOGY/ACTIVITY STUDIES OF NEW SERS-ACTIVE SUBSTRATES 79 Introduction 79 Experimental 82 Instrumentation 82 Chemicals and Procedure 83 Results and Discussion 84 Raman Microprobe Studies 84 Raman Macroscopic Studies 100 Conclusions 114 5 SPATIAL DISTRIBUTION STUDY OF SERS ACTIVE SUBSTRATES 116 Introduction 116 Experimental 117 Instrumentation 117 Chemicals and Procedure 117 Results and Discussion 118 Spatial Distribution Study 121 Laser Spot Size Study 130 Conclusions 140 6 COLLOID FILTRATION 142 Introduction 142 Experimental 147 Instrumentation 147 Chemicals and Procedure 148 Results and Discussion 150 Polycarbonate Track-Etched Membranes 150 Anotec Anopore Alumina Membranes 154 Conclusions 179 7 CONCLUSIONS AND FUTURE WORK 181 REFERENCE LIST 190 BIOGRAPHICAL SKETCH 200 V

PAGE 6

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE DEVELOPMENT OF NEW PREPARATION METHODS FOR SURFACE ENHANCED RAMAN ACTIVE SUBSTRATES By WILLIAM SCOTT SUTHERLAND DECEMBER, 1990 Chairperson: Dr. James D. Winefordner Major Department: Chemistry A wide variety of surfaces have been used for surfaceenhanced Raman spectroscopy (SERS) The methods developed for preparing these surfaces are either motivated by analytical factors, including low cost, speed, simplicity, and reproducibility, or by physical factors, including control of the surface morphology and the uniformity of the surface features. Analytical methods are focused on the application of SERS for routine analysis, while physical methods are focused on comparisons of experimental results to theoretical calculations, the latter being limited to sphere and spheroid geometries. The ideal SERS substrate would satisfy both the analytical and physical criteria. This work details a series of experiments which evaluate a variety of surfaces with respect to both the analytical and physical guidelines. Silver colloid solutions, which appear to be an ideal vi

PAGE 7

substrate, are used to study the ability of SERS to differentiate between compounds with similar structures, specifically the three primary sulf apyrimidines. These experiments graphically illustrate that, while SERS can be used for trace detection and speciation, the flocculation of metal colloids upon analyte addition renders them useless for physical studies. The next step taken was to modify the two most common techniques for SERS substrate preparation, metal vapor deposition and direct chemical reduction of silver salts, by using small pore size and fiber size filter papers (Millipore 0.025 and 0.22 nm filters) and other surfaces (frosted slides) to provide a more uniform support for the formation of silver features. The results are that these substrates provide more uniform silver surface features than other supports used previously, but that the surface features do not satisfy the physical factors. Also, nonuniform analyte distribution, observed for spots resulting from solutions applied to these membranes using a syringe, contribute to reproducibility problems, hindering the utility of these surfaces for analytical experiments. Syringe filtration of metal colloids is shown to be an alternative to these other methods, providing uniform surfaces with controlled morphology and minimizing the analyte distribution problem. This method is also inexpensive, fast, and simple, satisfying the analytical requirements. vii

PAGE 8

CHAPTER 1 INTRODUCTION The phenomenon of surface-enhanced Raman spectroscopy (SERS) has been known for over 15 years, and, while use of this technique and the diversity of systems studied continue to increase (Gerrard and Bowley, 1988), this field is still wide open in terms of opportunities for new and unique applications as well as extensions of existing techniques. Only in the past few years has the emphasis of SERS research begun to shift from the elucidation of the physical processes underlying the phenomenon to the development of the technique for routine analysis, including surface preparation methods, analysis of the analytical figures of merit of these surfaces, and the utility of SERS as a detector for separation techniques like liquid chromatography (LC) high performance liquid chromatography (HPLC) thin layer chromatography (TLC) and high performance thin layer chromatography (HPTLC) The studies described herein encompass a wide variety of topics. Chapter 2 provides a background of Raman spectroscopy in general, enhanced Raman spectroscopy, the theoretical considerations involved in SERS, and a description of how these theories can be used for optimizing metal surface geometries. Chapter 3 explores the utility of SERS on silver 1

PAGE 9

2 colloids for the detection of sulfonamides and describes some of the experimental problems of colloidal SERS. Chapter 4 discusses the preparation of a number of new SERS-active substrates using modifications of previously published techniques. The surface morphology of the prepared substrates is examined using the microscope from a Raman microprobe and using scanning electron microscopy. The SERS activity of these substrates is tested using both standard Raman and Raman microprobe techniques. Chapter 5 is a detailed look at one of the major contributing factors to problems in the reproducibility of SERS signals which is observed for the substrates in Chapter four. Chapter 6 describes a new method developed for the preparation of SERS substrates which potentially circumvents some of the major problems of using colloids and the surfaces described in Chapter 4. Finally, Chapter 7 looks at the conclusions drawn from this work and provides a look into the future of substrate preparation methods for SERS, outlining potential modifications of some of the methods presented in here as well as some novel preparation methods which could improve the control of surface morphology, the cost of substrate preparation, and the potential for automating the preparation method.

PAGE 10

CHAPTER 2 RAMAN SPECTROSCOPY Raman Scattering Raman scattering, discovered by Sir C. V. Raman in 1928, involves the gain or loss of energy from light scattered off a molecule. An excellent description of scattering processes and their relation to molecular vibrations and Raman scattering is given by Wilson et al (1980) In essence, when a beam of light is passed through a transparent substance, a small amount of the radiation is scattered. If a monochromatic light source is used, the majority of the scattered radiation will be at the same frequency as the incident light (Rayleigh scattering) but a small amount will occur at discrete frequencies both greater than and less than the incident frequency. These scattered frequencies are referred to as Raman scattering. There are two fundamental ways to interpret this result: a classical description and a quantum mechanical description. Classical Description The classical theory of Raman scattering revolves around the concept of molecular polar izability. This theory is often 3

PAGE 11

4 referred to as the wave theory, and a detailed derivation is given by Banwell (1972) and by Skoog (1985). A brief overview is given here. The induced dipole moment, n (Cm), of a molecule in an electric field is proportional to the 1 2 2 polarizability of the molecule, a (J Cm), and to the magnitude of the applied field, E (V m"^) as given in equation 2.1. \i-aE (2.1) When a sample of molecules is subjected to an incident electromagnetic field of frequency f (s'^\ the electric field experienced by each molecule varies according to equation 2.2. £-£oSin(2iivt) (2.2) The induced dipole oscillates according to equation 2.3. H-aEgSin (27i:v t) (2.3) This dipole emits radiation at a frequency yielding Rayleigh scattering. If the molecule undergoes a vibration, with a frequency ''vib' which induces a change in the polarizability of the molecule, the expression for a is modified to give: a-ao+Psin(27iv^_ii,t) (2.4)

PAGE 12

5 where R is the rate of change of polarizibility of the vibration. The oscillating dipole becomes a superposition of these two processes. H(ao+psin(27iVy^t) ) E^sin Unv t) (2.5) Using trigonometric identities, equation 2.5 can be written in a more revealing form. [i-aQE^sin{2n\t) +^^Eq{cos2-k {v-v^.^) t-cos27i (v+v^,^^) t) (2.6) From equation 2.6, two things are apparent. First, only those vibrational modes which impose a change in the polarizability (a nonzero value for B) are Raman active. Second, Raman shifted bands occur at frequencies lower (Stokes) and higher (anti-Stokes) than the incident frequency. The classical description of Raman scattering is very useful for developing an intuitive feel for which vibrations of a molecule should be expected to be Raman active. Two useful tools which can be used, at least for simple molecules, to help determine which vibrational modes of a molecule are Raman active are group theory (Cotton, 1971) and the polarizability ellipsoid (Banwell, 1972) In general, symmetric vibrations give rise to intense Raman lines and nonsymmetric vibrations are usually weak and sometimes not observable (Banwell, 1972) The classical description is not complete, however, since it

PAGE 13

6 predicts that Stokes and anti-Stokes Raman peaks should be of the same intensity, a result contradicted by experiment. In order to rectify this inconsistency, one must turn to a quantum description of Raman scattering. Quantum Mechanical Description The quantum theory of Raman scattering focuses on the concept of quantized vibrational and electronic energy levels in molecules. This theory is often referred to as the particle theory and a detailed description can be found in Wilson et al (1980), Skoog (1985), and Banwell (1972). In essence, the incident radiation field is considered to be a stream of photons which are elastically (Rayleigh) and inelastically (Raman) scattered by the molecules of the medium. The incident radiation is chosen to have a frequency far from any electronic resonances of the molecule of interest. Thus, excitation is into a "virtual state" which is not a normal electronic or vibrational state of the molecule (Figure 2.1). It is immediately evident from Figure 2.1 why different intensities are observed for Stokes and anti-Stokes Raman lines. Stokes lines arise from excitations from the ground vibrational state of the molecule, whereas anti-Stokes lines arise from the first vibrationally excited or higher excited states of the molecule. The population of these

PAGE 14

c to e to a c 3 (0 (U u 0) -p T! C (0 (0 & C -H >^ 0) P +J (0 o u c (0 e to CP c -p (0 -p U} 3 to to (1) > ^ u u 0) 0) c o< 0) U ^> CP

PAGE 15

8 00 c CO L ^ \ u I o f CO I x c < II + >< ^ — ^ -s: — o II II 3 3 (/5 (l> M 2 00 C c < + 5, O c CT — 4) 00 C to o 0) o I K u to t: II 3 o II

PAGE 16

9 states follows a Boltzmann distribution and, at room temperature, the majority of the molecules are in the ground vibrational state. Thus, it would be expected that Stokes line intensities would be much greater than anti-Stokes line intensities. For example, the intensity difference between a Stokes line and an anti-Stokes line (arising from the first excited vibrational state) of a 1000 cm"^ Raman peak at room temperature should be approximately a factor of 100. This difference will diminish as the temperature of the system is increased (Skoog, 1985) ; Raman Spectroscopy Raman scattering is differentiated from Raman spectroscopy in that the former is the phenomenon observed and the latter is the technique which makes use of this phenomenon. A large number of papers on Raman scattering appeared in the chemical literature shortly after its discovery in 1928. This enthusiasm soon waned due to the length of time (several minutes to hours) and the large amount of sample required (typically 10-20 ml) to obtain a Raman spectrum with conventional white light sources. The most significant advance in Raman spectroscopy was the development of high powered, continuous wave (cw) gas ion lasers (e.g. Ar*, Kr*) in the late 1960's (Gerrard and Bowley, 1989), An increase in the quality of monochromators during this time period also improved the utility of Raman spectroscopy (Ingle

PAGE 17

10 and Crouch, 1988) To attest to the increased popularity of Raman spectroscopy, over 6000 papers were published during the period between late 1985 and late 1987 dealing with many applications of this technique (Gerrard and Bowley, 1988) Advantacfes of Raman Spectroscopy It is not surprising, during the resurgence of interest in Raman spectroscopy, that comparisons were made with its closest comparable technique, infrared absorption spectroscopy, since both reveal information about the vibrational modes, and therefore the structure, of the molecule of interest. In a direct comparison, Raman spectroscopy displays a number of distinct advantages over its counterpart. 1. Any frequency of light can be used, especially visible radiation, which allows the use of standard UV-visible transducers (photomultiplier tubes, diode arrays, chargecoupled devices) instead of thermal detectors. 2. Water is a useful solvent in Raman spectroscopy, whereas it is usually a poor solvent for IR spectroscopy. 3. The properties of the laser sources used in Raman spectroscopy make it relatively easy to probe a variety of sample types, including micro-samples (using a Raman microprobe) surfaces, films, powders, solutions, gases, and samples of virtually any size or shape. 4. A single Raman spectrometer can cover the entire range of vibrational frequencies, whereas several instruments are required for this task using IR spectroscopy (or a change of beam splitters or detectors with fourier transform IR) 5. Raman spectra are usually much simpler than IR spectra due to the lack of intense overtone and combination bands.

PAGE 18

11 6. Polarization data add extra structural information not available using IR spectroscopy. Disadvantages of Raman Spectroscopy Despite all the advantages that Raman spectroscopy offers over conventional IR spectroscopy, especially with respect to the simplified optics and the ability to use aqueous solvents (a must for most biological systems) IR spectroscopy is still the technique of choice in most analytical laboratories. There are a number of reasons for this. 1. IR spectroscopy is more sensitive to small structural differences due to the intensity of overtone and combination bands. 2. Extensive libraries of IR spectra have been compiled. Similar libraries for Raman spectra exist, but are not nearly as complete. 3. Infrared instruments are generally less expensive than Raman instruments The monochromators used in Raman spectroscopy must be of a higher quality than those used in IR spectroscopy. 4. Because Raman spectra are highly dependent on laser power, cell geometry, and instrument characteristics, it is much more difficult to compare Raman intensities from one instrument to another than it is for IR spectra. 5. IR cross sections are much larger than those for Raman spectroscopy. Thus, IR detection limits are typically much lower than for conventional Raman spectroscopy. 6. The use of visible excitation radiation in Raman spectroscopy can often lead to background fluorescence from the sample or impurities, completely swamping the Raman signal.

PAGE 19

12 Enhanced Raman Spectroscopy If a single reason for the lack of interest in using normal Raman spectroscopy (NRS) for routine analysis had to be identified, it would be the weak intensity of the Raman scattered light. Typical Raman cross sections are on the -30 2 order of 10 cm compared to typical IR absorption cross "18 2 sections of 10 cm (Van Duyne, 1979) Raman scattering cross sections are small due to the weak interaction of the incident radiation with a virtual level of the molecule. Infrared absorption cross sections are larger since the process involves excitation into a real level, indicating a strong interaction between the incident radiation and the molecule. To determine possible methods for improving Raman signals, it is helpful to examine a simple expression for the Raman intensity. I^-pPij^)Cl (2.7) Ip is the Raman intensity (photons s^\ P is the laser power at the sample in photons per second, p is the density (cm"^) of adsorbed scatterers, {da/dn) is the differential Raman scattering cross section (cm^ sr'^) and n is the solid angle (in sr) over which scattered photons are collected. A second expression of interest gives the frequency dependence of the Raman scattered signal.

PAGE 20

13 da (g) ) o^TT where is the frequency of the Stokes Raman shifted scattered light (cm'^) and a is the polarizibility of the molecule. Examination of these two equations reveals that several options exist for increasing the Raman scattered intensity: 1. Optimize the performance of the Raman spectrometer to increase the solid angle of collection and to reduce optical losses. 2. Increase the density of adsorbed Raman scatterers. 3. Increase the laser power at the sample. 4. Increase the Raman scattering cross section of the molecule of interest. Van Duyne (1979) gives several possibilities for increasing the efficiency of the Raman spectrometer collection optics, but realistic enhancements using these procedures are only one order of magnitude. Another method for increasing the collection efficiency is to use multichannel detection (Campion, 1983) which can decrease the acquisition time for Raman data by as much as 10^ over conventional scanning systems. This corresponds to a signal-to-noise enhancement of 3 2 if equal acquisition times are used in multichannel and scanning instruments. This alone is a major improvement for Raman spectroscopy. A detailed comparison of scanning and multi-channel Raman systems is given by Campion (1983) The density of adsorbed scatterers is typically increased by

PAGE 21

14 adding surface area to the substrate using some sort of roughening technique (e.g. oxidation-reduction cycles for electrode surfaces) This type of approach will provide a maximum additional enhancement of one to two orders of magnitude (Van Duyne, 1979) Increasing the laser power is limited by available laser hardware and by the tolerance of the sample to high photon fluxes. Desorption, sample damage, and substrate damage typically begin to occur for incident laser powers of over 1 W (Campion, 1983) The final option for enhancing the Raman scattering intensity suggested by equations 2.7 and 2.8 is to increase the Raman cross section. Three methods of achieving this result are to choose molecules with large Raman cross sections, increase the frequency of the laser excitation to take advantage of the w* dependence of the Raman scattering intensity, or use a frequency which is coincident with an electronic transition of the molecule or molecule/surface system. The molecule to be detected is dictated by the problem to be solved, and thus is not usually a variable. The remaining two choices have been employed to produce the resonantly-enhanced Raman techniques. Resonance Raman Spectroscopy The resonance Raman effect, first described in detail by Placzek (Bernstein, 1979), is an increase in the Raman scattering cross section when the incident radiation is near or coincides with an electronic transition of the molecule.

PAGE 22

Enhancements as large as 10^ are due to the much stronger interaction of the incident electromagnetic field with a real excited state of the molecule being studied rather than with a virtual level excited in normal Raman scattering. An excellent theoretical treatment of the resonance Raman effect in general is given by Bernstein (1979) and a detailed analysis of resonance Raman spectroscopy (RRS) on complex molecules is given by Spiro and Stein (1977) RRS became practical for analytical use with the advent of intense Ar* and Kr* lasers and, more importantly, tunable dye lasers in the late 1960's (Ingle and Crouch, 1988; Skoog, 1985). With the discrete wavelength cw lasers, the molecule to be studied must be chosen to have an electronic transition which overlaps with the laser wavelength. This severely limits the systems to which the technique is applicable. The use of a tunable dye laser (cw or pulsed) allows the excitation radiation to be "tuned" into an electronic resonance of a molecule under study. In a resonance Raman experiment, when the frequency of the excitation radiation is tuned into an electronic transition of the molecule of interest, only those vibrations which are coupled to that electronic transition are enhanced, so RR spectra are typically much less complex than NR spectra (Mathies, 1979) Resonance Raman spectroscopy has been used on a wide variety of molecular systems. Gerrard and Bowley (1989) describe the use of RRS with cw and pulsed dye laser

PAGE 23

16 systems for the detection of polycyclic aromatic hydrocarbons (PAH) Spiro (1974) describes the use of RRS with a number of molecules of biological interest, including amino acids, nucleic acids, lipids, membranes, and a number of biologically important chromophores including heme proteins, iron-sulfur proteins, and hemerythrin. RR detection and characterization of visual pigments and the bacteriorhodopsin chromophore in purple membranes is discussed by Mathies (1979) and a number of papers have been published on the use of RRS with chlorophyll and bacteriochlorophyll (Callahan and Cotton, 1987; Heald et al 1988). Finally, several research groups have focused their interests on the use of RRS for the study of biochemical systems, especially for large biomolecules containing chromophoric groups. These chromophoric regions are quite often the areas of biological activity in these molecules (Hughes, 1985) Use of an excitation frequency which corresponds to an electronic transition of the chromophoric group yields enhancements of the Raman peaks associated only with the bonds of the chromophore. These spectra usually have few peaks and are not subject to interference by the Raman bands associated with the bulk of the molecule (Mathies, 1979; Demtroder, 1982). The major experimental problems with RRS are fluorescence interference, either from impurities in the sample or from the electronic transition being probed, photolysis, and local heating of the sample from the high laser intensities used.

PAGE 24

17 Impurities can often be removed, but eliminating fluorescence from the molecule itself presents a more difficult problem. One technique to minimize fluorescence interference from the molecule being probed is to use time-resolved Raman spectroscopy, taking advantage of the fact that the lifetime of the fluorescence is much longer than that of the Raman scatter. This requires the use of a pulsed laser and discrimination electronics, adding cost and complexity to the experiment. A second technique is to use an optical filter to remove the fluorescence and to pass the Raman scatter, which is typically 50-100 nm red-shifted from the excitation wavelength. This method is successful only if the fluorescence intensity in the region of the Raman peaks is negligible. Techniques which have been employed to minimize sample heating and photo decomposition include spinning the sample under the laser beam, the use of a flowing streams, and the use of low-temperature techniques (Mathies, 1979) Ultraviolet Resonance Raman Spectroscopy The use of cw ion lasers, both independently and coupled with dye lasers, for RRS allows for excitation wavelengths to be chosen over the entire visible wavelength range. Unfortunately, this precludes the use of RRS for the majority of organic compounds since their electronic transitions occur in the ultraviolet (Gerrard and Bowley, 1989) Although ultraviolet resonance Raman spectroscopy (UV RRS) is not, in

PAGE 25

18 principle, different from visible RRS, the advantages realized by using ultraviolet excitation frequencies are significant enough that UV RRS can be considered a separate technique. Among these advantages are that ultraviolet excitation greatly expands the applicability of the RR technique (Hudson, 1986) and that fluorescence interference is greatly reduced by using excitation wavelengths below 250 nm (Asher and Johnson, 1984; Asher, 1988) High power, fixed UV wavelengths can be obtained using pulsed excimer lasers, including ArF (193 nm) KrF (249 nm) XeCl (308 nm) (337 nm) and XeF (350 nm) (Hudson et al 1986) The feasibility of UV RRS using a KrF laser has been demonstrated (Lin et al 1987). A more practical system, designed by Hudson et al (1986), uses fixed wavelengths generated from a Q-switched Nd:YAG laser with frequency doubling, frequency mixing, and a Raman shifting. By far the most flexible system to be developed for UV RRS is also based on the Nd:YAG laser, but adds tunability with a dye laser. Harmonics of the fundamental of the Nd:YAG laser (1064 nm) are obtained using frequency doubling and tripling crystals. These wavelengths are then used to pump a dye laser, the output of which can be doubled and mixed with the output of the Nd:YAG laser to obtain continuously tunable wavelength selection between 265 and 800 nm. Deep UV wavelengths (190220 nm) are obtained with the use of a Raman Shifter (Asher et al., 1983). This system has been used for the detection and

PAGE 26

19 speciation of polycyclic aromatic hydrocarbons (Asher, 1984) where detection limits as low as 2 0 ppb were observed, and for the study of compounds which are models for the peptide bond (Dudik et al 1985) Analytical considerations of XJV RRS are discussed in detail by Jones et al (1985), and a more recent review of a variety of biophysical and analytical applications of the technique is given by Asher (1988) Surface-enhanced Raman Spectroscopy When Fleischmann and co-workers (1974) first observed a greatly enhanced Raman signal from pyridine on a silver electrode, they attributed the effect to the increased surface area of the electrode due to electrochemical roughening from oxidation-reduction cycles performed in their experiment. It was not until 1977 that two groups (Jeanmarie and Van Duyne, 1977; Albrecht and Creighton, 1977) independently determined that the enhancement was a new physical phenomenon. This realization touched off a flurry of investigations into this so-called surface-enhanced Raman scattering (SERS) These studies were divided into experiments to determine which factors affected the enhancement and the development of theoretical models to explain the variety of experimental observations. The theoretical considerations will be discussed in detail later, but the consensus is that there are three basic mechanisms, grouped into two classifications (chemical theories and electromagnetic theories) which

PAGE 27

combine to give the overall enhancement of Raman scattered light. The chemical theories include a charge-transfer mechanism and an active-site mechanism. Both of these imply that SERS is highly dependent on the molecule-surface interaction. The electromagnetic theory is based on the enhancement of the incident electric field by the metal surface due to the excitation of surface plasmons in the metal electrons. This effect is independent of molecular interactions with the surface but highly dependent on the size and shape of the surface features, the dielectric properties of the surrounding medium, and the molecule-surface distance. Many of the initial SERS experiments were performed on electrode surfaces (Howard and Cooney, 1982; Chang and Laube, 1984) since the phenomenon was first discovered on a silver electrode. An excellent review of the status of SERS on metal electrodes is given by Chang (1987) When surface roughness was determined to be a contributing factor to the enhancement, many researchers began to use metal island films. These substrates were prepared in vacuum, eliminating any contamination effects and allowing exact amounts of the molecule under study to be deposited onto the metal surface (Pockrand, 1982b) These substrates also allowed for direct comparisons with Raman scattering on single crystal surfaces, providing detailed information on the role of surface roughness in the Raman enhancement (Pockrand, 1982a) A comparison of SERS with other standard surface science

PAGE 28

21 techniques, such as Auger electron spectroscopy, electron energy loss spectroscopy, and low energy electron diffraction is given by Jha (1982) Once the SERS effect was verified, researchers began to try other types of surfaces, other metals, and molecules other than pyridine to see how universal the effect really was. The growth in the number of SERS studies can readily be seen in the paper by Seki (1986) which documents over 500 experiments which had appeared in the literature by 1985. These experiments involved many different surfaces, including plasma-etched quartz posts, metal coated filter membranes, and metal coated polystyrene spheres, on a variety of metals, including gold, copper, lithium, potassium, and mercury. Growth in the number and diversity of SERS experiments continues, and publications numbered over 2800 by early 1989 (Garrell, 1989). It is interesting to note that the enhancement of optical processes at surfaces has been shown to be a general phenomenon (Glass et al., 1980; Weitz, at al 1983; Moskovits, 1985) The electromagnetic theories of the SERS effect have led to the postulation and discovery of the surface enhancement of such processes as infrared absorption (Olsen and Masel, 1988) second harmonic generation (Moskovits, 1985), electronic absorption (Glass et al 1980), luminescence (Wang and Kerker, 1982; Wokaun et al., 1983; Das and Metiu, 1985; Huang et al 1986), and photochemistry

PAGE 29

22 (Nitzan and Brus, 1981; Gersten and Nitzan, 1985) Since it is the incident electric field which is enhanced upon interaction with the roughened metal surfaces, the discovery of these other techniques is not surprising. Many of these techniques are still in their infancy and may eventually lead to an arsenal of surface-enhanced optical processes which can be used together for the trace detection and speciation of compounds adsorbed to metal surfaces. Surface-enhanced Resonance Raman Spectroscopy When a molecule is chosen for a SERS experiment which has an electronic absorption which overlaps spectrally, at least in part, with the plasmon absorption profile of the metal surface being used, the result is a coupling of the SER and RR phenomena to give surface-enhanced resonance Raman spectroscopy (SERRS) In an ideal situation, the enhancement factors are multiplicative, since the processes of SERS and RRS are essentially independent. Enhancements of twelve orders of magnitude can be expected. In actual SERRS experiments, however, it has been shown that the addition of as little as a partial monolayer of the species to be studied alters the index of refraction in the vicinity of the metal surface feature and shifts the enhancement profile so that the SERS part of the SERRS enhancement is no longer maximum and sometimes nonexistent for the excitation wavelength being used (Zeman at al 1987; Kim et al 1989). In practice, the

PAGE 30

23 overall enhancement is usually larger than either the SERS or RRS enhancements but less than what would be predicted by simple multiplication of the individual enhancements. The SERR technique has a number of advantages over either of the separate techniques. The most obvious of these is increased sensitivity due to the coupling of enhancements. A second benefit of SERRS has to do with interferences in the Raman spectrum. As mentioned earlier, one of the major drawbacks to RRS is fluorescence interference. Adsorption of molecules to the metal surfaces used in SERS experiments usually results in fluorescence quenching by creating nonradiative decay pathways for the electronic excited state (Adams et al 1980; Weitz at al 1982; Pineda and Ronis, 1985) The fluorescence background observed in RRS is removed in SERRS. Both advantages are realized in SERR experiments for the detection of rhodamine 6G (Hildebrandt and Stockburger, 1984), crystal violet (Chou et al 1986), a number of other dyes (Yamada et al 1986), and other highly fluorescent molecules (Chambers and Buck, 1984) Raman Microprobe Spectroscopy The first Raman microprobe was developed by Dhamelincourt in 1975 (Dhamelincourt et al 1979, Adar, 1988). The idea is simple. The laser beam to be used in the Raman experiments is focused through a microscope objective onto the sample. The spatial extent of the beam at the sample is usually between

PAGE 31

24 one and five micrometers and the incident laser power at the 5 2 sample can be as large as 10 W/cm (Van Duyne et al 1986). Problems with RMS include the potential for sample damage and the need for high concentrations of the molecule to be studied. The first problem can be addressed using the techniques developed for RRS, including spinning of solid samples or flowing of liquid samples, and by using other techniques, such as cooling of the sample (for thermal damage problems) and placing samples in vacuum or in an inert environment such as nitrogen (for photo-oxidation problems) RMS has several advantages over normal Raman scattering. The 2 small probe area (as little as 10 /xm ) allows spectra of solid samples to be obtained selectively from a crystal of the compound of interest, thereby eliminating any fluorescence interferences from impurities in the sample. Fluorescence from the compound itself is still a potential interferant. The small probe area also allows spectra of selective compounds in a mixture to be obtained, assuming that the crystals of each compound can be differentiated visually (Huong, 1986) This ability to select a particular feature of a sample has been used in the identification of impurities on printed circuit boards and verification of the authenticity of ancient Chinese vases (Dhamelincourt et al 1979). The idea of coupling the Raman microprobe with SERS was presented by Van Duyne et al (1986). Calculated detection limits were less than one attomole in the probe beam. This

PAGE 32

corresponds to as little as 10^ molecules. Chapter 4 presents some of the experimental results obtained in this dissertation research using the Raman microprobe. Estimated mass limits of detection are of the order of 100 ag or one attomole, in excellent agreement with the predictions of Van Duyne at al (1986). Van Duyne et al (1986) also hinted at the possibility of even lower detection limits using RMS and SERRS. Experimental verification of the potential of the coupling of RMS and SERRS was recently reported by Taylor et al (1990), who found detection limits for crystal violet of lO"^^ M, or approximately 4 x lO"^' g (600 molecules) Theoretical Considerations of SERS During the first few years following the discovery of the SERS phenomenon, a wide variety of theories were published which attempted to explain the large enhancements observed. Furtak and Reyes (1980) presented a critical review of a number of these theoretical models, many of which have been proven incorrect by experiments designed to test them (Campion, 1983) The process of experimental verification of SERS theories has led to the conclusion that there are two basic classifications of enhancement mechanisms which contribute to the SERS effect. As mentioned previously, these are the chemical and electromagnetic theories. Birke and Lombardi (1982) review the influence of surface features on the electromagnetic SERS enhancement, and Campion (1984)

PAGE 33

discusses both mechanisms and their potential enhancements. In general, both mechanisms are believed to be responsible for the observed enhancement, but the relative contributions of each appears to be dependent on the molecule-substrate system under study. Chemical Theories Many of the early SERS experiments indicated that the phenomenon was not a general effect, but that only certain types of molecules gave rise to enhanced Raman spectra. These results led to the conclusion that a significant moleculesurface interaction was required (e.g. chemisorption or strong physisorption) for SERS to be observed. Of the many theories of this type to be proposed, two have withstood experimental scrutiny: the active-site model and the charge-transfer model (Furtak and Roy, 1985). Pettenkofer et al. (1985) give a detailed description of SERS active sites and Sobocinski and Pemberton (1988) show that these sites can be probed experimentally using laser-induced thermal desorption. The basic premise of this model is that certain locations on a metal surface (e.g. defects) may allow for stronger adsorption between a molecule and the metal surface. Adrian (1982) and Lippitsch (1984) both give an excellent description of the charge-transfer model. In this theory, it is assumed that the enhancement arises from charge transfer between the molecule and the electron conduction band of the metal. Although these

PAGE 34

27 theories can only account for one to two orders of magnitude of enhancement, they help to explain some of the experimental observations which are not predicted by the purely electromagnetic theories. Quantum Mechanical Theories A small number of papers have appeared in the literature which detail a quantum-mechanical treatment of the SERS effect. Initial studies attempted to calculate the enhancement of H2 on a Li cluster (Pandey and Schatz, 1982; Pandey and Schatz, 1984) using time-dependent Hartree-Fock theory. A more general quantum theory was developed by Jha (1985) In this paper, the "overall" enhancement is treated as two separate components. The "long range" component, where the distance to the surface is considered to be much greater than atomic size (i.e. outside the first monolayer) is treated using the electromagnetic theories, described below. The "short range" component, where the distance to the surface is on the order of the atomic size, is treated quantummechanically where the quantities of interest are the molecular polarizability and its modulation upon adsorption to the surface. The latter is centered around the calculation of the single-particle electronic Green function for the molecule-substrate complex, but will not be discussed further. Pettinger (1987) presents a quantum-mechanical description for SERRS which gives the overall enhancement in a single

PAGE 35

I 28 expression which combines the chemical and electromagnetic enhancements as well as the resonance Raman enhancement for molecules adsorbed on a surface. Overall enhancements are calculated to be between two and eleven orders of magnitude, in agreement with experimental data, depending on the specific metal-molecule system being studied. Electromagnetic theories The most successful and most thoroughly characterized of the many theoretical treatments of the surface-enhanced Raman effect are the electromagnetic theories. The first manifestations of these theories were presented by Kerker et al (1980) for spherical particle shapes and by Gersten and Nitzan (1980) for spheroidal particle shapes. These shapes are the only ones which yield nontrivial solutions to the computation of the polarizability (Van de Hulst, 1981) The latter is more flexible, since it allows for the calculation of SERS enhancement factors for nonspherical particles. The basic premise behind these theories is the same, so they will be described as a single model. The conduction electrons of the metal of interest are treated as a free electron gas. The electric field of the incident radiation excites a dipolar surface plasmon in the metal particle, which is to say that it polarizes the electrons in the particle. This has the effect of concentrating the field just outside the particle, leading to an enhancement of the electric field. The electromagnetic theories solve for the magnitude of this enhancement subject

PAGE 36

29 to the boundary conditions of the particle shape and size. The important point of the solution to this problem is that these particles show a dipolar plasmon resonance frequency. The EM SERS effect is the direct result of the excitation of this plasmon resonance. One important assumption made is that the diameter of the particle is much smaller (typically less than five percent) than the wavelength of light polarizing it. Under these conditions, the radiation field incident on the particle is homogeneous, Rayleigh scattering dominates, and only the dipolar surface plasmon is excited. A detailed description of the interactions of small dielectric particles with a radiation field is given by Van de Hulst (1981) A brief description of the model used by Gersten and Nitzan (1980) is useful in visualizing the limitations imposed by the assumptions made. Figure 2.2 illustrates the basic assumptions of this model. The surface is described as a prolate hemispheroid protruding from a flat plane. The spheroid has a frequency dependent complex dielectric function e(a)) and the plane is considered to be perfectly conducting. The molecule to be studied is approximated as a point dipole along the symmetry axis perpendicular to the plane with the molecular dipole parallel to the symmetry axis. The incident electric field is taken to be along the symmetry axis. Under these conditions, the molecule is in a position for maximum field enhancement, and enhancement factors as large as lo" have been calculated for silver prolate spheroids (Gersten and

PAGE 37

0) TJ -rH C M (0 en (0 a 0) o m u (0 0) .si (0 -P c o <1> •H O 4J n) (0 <4-( H M o w (0 O -H O o -H -P c CP n] 0) ^ :m II o u -p o 0) 1 p e o 0) o 13 rH 3 U (U H o 0) X! -P X! 1T 13 O u X! -P 0) U

PAGE 38

31

PAGE 39

32 Nitzan, 1980) much larger than the maximum enhancements of 10^ which had been observed experimentally at that time. Despite this disagreement in the magnitude of the enhancement, the EM theory is very useful in predicting many of the trends observed experimentally. An excellent illustration of the EM theory is given by Aroca and Martin (1985) who use the Gersten and Nitzan (1980) model to calculate the enhancement factors for several metals, including silver, gold, and indium, and aluminum, for several particle geometries. The results for silver and gold show excellent agreement for trends in the enhancement versus Raman shift and the optimum wavelength of the enhancement versus particle shape observed experimentally. Table 2.1, which contains calculated data for aluminum, illustrates that it is possible to use the EM theory to qualitatively dictate how to "tune" the roughness of a given metal surface to optimize the enhancement. The information in Table 2.1 shows that, for a given incident wavelength, the aspect ratio (the ratio of the major to minor axes) which yields the optimum enhancement for a particular metal can be calculated. This feature makes this theory a powerful tool for the experimentalist. Correction to the electromagnetic theory It is evident that the model described above overestimates the Raman enhancement factor by as much as five orders of magnitude. By looking at the assumptions made in the model, it is clear that a more realistic picture would be

PAGE 40

33 Table 2 1 Enhancement factors for aluminum prolate spheroids Excitation Wavelength ( nm^ Resonance Aspect Ratio Vibrational Wavenumber (cm'^' Enhancement Factor (v 10 1 488 8.24 500 •J \J \J 1 7 5 8.37 1000 1 68 8.51 1500 1 56 8 64 2000 1 41 8.81 3 000 1 07 514 4 8.82 500 1 80 8 94 1000 1 72 9 10 1500 1 58 9.20 2000 1.41 9 .40 3000 1 05 590 10 42 500 1 66 10. 62 1000 1 54 10.81 1500 1.39 10.90 2000 1.21 11.07 3000 0.867 640 11.56 500 1. 37 11.71 1000 1.24 11.95 1500 1.10 12.04 2000 0.926 11.95 3000 0.595 Source: Aroca and Martin (1985)

PAGE 41

34 provided by calculating the enhancement factors for molecules at various positions around the spheroid and averaging these to give an estimate of the overall enhancement (Barber et al 1983a, 1983b) There are other, more subtle corrections, however, which need to be made to make the EM theory more realistic. A number of these are described by Gersten and Nitzan (1982) including the effects of neighboring bumps, the surrounding medium, and the size of the particle, but none of these were incorporated into their EM theory. Laor and Schatz (1981; 1982) explore the role of neighboring bumps on the local field enhancement and the effects of the surrounding medium are addressed by Barber et al (1983). Two additional corrections to the Gersten and Nitzan (1980) EM theory are surface scattering and the onset of multipolar plasmon excitations, both of which are size dependent phenomena. When the size of a metal particle becomes smaller than the mean free path of the electrons of the metal, the plasmon resonance of the particle is broadened, the conduction electrons experience surface scattering, and the electric field enhancement is decreased (Kraus and Schatz, 1983a, 1983b). When the size of the particle increases beyond the Rayleigh limit, the plasmon resonance broadens, shifts to longer wavelengths, multipolar plasmon resonances appear, and the field enhancement decreases (Barber et al 1983, 1983a; Kraus and Schatz, 1983). When these two effects are incorporated into the Gersten and Nitzan (1980) EM theory, an optimum size

PAGE 42

for maximum field enhancement for a given shape of particle emerges since both large and small particle enhancements are damped (Wokaun et al 1982). This corrected EM theory is often referred to as the electrodynamic (ED) SERS theory. The result of these corrections is illustrated by Cline et al (1986) This study applies the same pictorial model for SERS calculations as Gersten and Nitzan (1980) specifically by ignoring any surface averaging. The overall maximum enhancement factor for silver at the tip of a prolate spheroid with a 2:1 aspect ratio is lo', smaller than for the pure EM theory, and the magnitude of the enhancement factor is shown to shift to the red and decrease with increasing particle size. It is interesting to note that gold and some of the other metals for which these calculations were performed (platinum, palladium, rhodium, and iridium) actually have increasing field enhancements as the particle size increases from the Rayleigh limit. Thus, it is possible for metals other than silver to yield competitive enhancements when larger particles are used in SERS experiments, often the case with the limits of the surface preparation techniques used. A second important point is that the metals covered in this study can provide enhancements of at least 10 over a spectral range from 275 nm for rhodium to 700 nm for gold, with the wavelengths in between available by changing either the metal or the size of the particles of a given metal. This again

PAGE 43

illustrates the potential power of these computations to the experimental spectroscopist A working electrodynamic theory The most complete implementation of the electrodynamic SERS theory to appear in the literature is given by Zeman and Schatz (1984) This theory incorporates the damping effects for small and large particle sizes, as well as calculating all enhancement factors as surface-averaged over the entire particle. Both spherical and spheroidal shapes may be chosen, as well as the dielectric constant of the surrounding medium (Zeman and Schatz, 1987a) An additional feature of this model is the incorporation of the effects of the monolayer or submonolayer of adsorbed species on the dielectric constant of the medium near the particle surface. As mentioned earlier, this latter effect can alter the SERS enhancement by shifting the plasmon resonance maximum out of resonance with the incident radiation. As an exercise to test this model, Zeman and Schatz (1987) calculated the size and shape dependence of the incident electric field enhancement and the SERS enhancement for ten metals, including silver, gold, and copper (the metals most often used in SERS experiments) as well as lithium, sodium, aluminum, gallium, indium, zinc and cadmium. An example of the results is shown in Figure 2 3 and Figure 2.4 for silver and lithium, respectively. These metals are chosen to illustrate the power of this model. One of the major unanswered questions in SERS experiments with colloidal

PAGE 44

w e Pi c u w o w in m c II -rH (0 (0 ^ xi u —-p o -H (0 0) •H c 0) u 0) 0) O (0 0) X! XI Ul 0) 0) M -H -H 0) (|H MH > -H X! ;3 -p g C +J rH o o > -o o (fl (0 15 -O ^ Q) C 0) C (0 TJ OJ •rl x: o -p trx! C -H ft 0) u c o 0) o w w > >^ rH 0) 0) > C 0) (0 X! H C (0 Q) > ^ lu O "O nJ ^MH 0) •H 0) XI -P — ft ft o w o (0 <4H XI 0) C H 0) w n a> u §. H

PAGE 45

38 y ujnuJ!4.dQ (uuu) Lj4.6ua |^8ao//\ 1 1 1 • 1 I 1 • 1 1 1 T— y lo'o X 3 I

PAGE 46

w on o w in tH CM o (0 C II •H (0 X ^ (0 ^ x: U >-'-P O -H •o+J > (0 c g 0) Ul •H g Q) 0) • CO a) p c (0 c 0) 0) 0) H -H Q) <4-l g -p g C -P o o 0) a) o > -d (0 c >i > (0 D> u T3 Q) C 0) C (0 73 0) •H ^ W XJ O •nag c o 4J c g 0) 3 M -H :3 43 O W W +J c (0 X! c >^ -H > > ^ o 0) i3 -P 0) 0) g-H g-H •H O o w C T3 g w 0) O +J C X! (0 Cn X! -H c u

PAGE 47

y ujnuj I 4.do (luu) L|4.5ua ^ aAC>/y\

PAGE 48

silver is the apparent lack of an enhancement for unaggregated particles (Creighton et al 1979). This has been a topic of much debate, with some groups stating that they have observed a SERS enhancement for monodisperse silver colloids (Garrell et al 1983) while others observe nothing (Blatchford et al 1982). From Figure 2.3 for silver, it is apparent from the ED calculations that the optical properties of silver are such that the spherical field enhancement is damped. The optimum particle radius for silver spheres is shown to be 20 nm, yielding an overall enhancement of around 2500. As the shape of the particle is made more spheroidal (increasing the aspect ratio) the optimum size for maximum enhancement increases and the resonance shifts toward longer wavelengths. Enhancements approaching 10*^ can be achieved for large (100 nm major axis) eccentric (5:1 aspect ratio) particles. However, for sizes and shapes commonly found in SERS experiments, enhancement factors calculated with this ED theory are typically two orders of magnitude lower than those measured. Whether this result is an indication that chemical enhancements are involved in these experiments is not certain, but it does imply that their importance cannot be ignored. The data for lithium (Figure 2.4) shows distinct differences from the silver data. This is most evident in the field enhancement versus semi-major axis plots, and most dramatically for the 1:1 ratio spheroid. For silver, the field enhancement is small and relatively flat as a function

PAGE 49

of the particle radius. The same is true for gold and copper (Zeman and Schatz, 1987). However, a lithium 1:1 spheroid has a strongly size-dependent field enhancement, with a very distinct maximum at around 12 nm. The intensity of this maximum enhancement for spheres is nearly as large as for more eccentric geometries. This sets lithium apart from the noble metals, where the electrodynamic theory predicts that eccentric particles are a necessity for large SERS enhancements (Zeman and Schatz, 1987) The significance of these ED results is made clear by also noticing that the wavelength for optimum resonance shifts to the red as the particle shape is made more eccentric. Thus, for metals such as lithium, it should be possible to get strong enhancements from the ultraviolet to the infrared by simply changing the particle shape and making sure that the radius is within the resonance peak predicted by the calculations. In fact, for all the metals and geometries tested, lithium is calculated to have the largest enhancement for all incident wavelengths when the size and shape are optimized, and the magnitude of this enhancement is constant from the ultraviolet to the infrared (Figure 2.5) Applications of the ED Theory to SERS Experiments The power of the electrodynamic SERS theory, as illustrated above, in identifying which metal and surface

PAGE 50

m to u 0) > 0) (0 si ui 73 C rO (1) N •H Ul o +J +J o 0) T3 0) N -H •H -P o II) p c 0) o c Si ^ Jit in 0) •H

PAGE 51

t 44 >

PAGE 52

45 morphology will yield a large Raman enhancement for a given excitation wavelength has a number of ramifications. One of the most significant of these relates to the coupling of SERS and RRS to yield SERRS. As mentioned in an earlier section, the most important technological advance for RRS was the development of tunable cw and pulsed dye lasers. These lasers allow the experimenter to tune the excitation wavelength into resonance with the molecule under study. With the ED theory, it is now possible to "tune" the SERS surface resonance frequency into coincidence with the molecular resonance. The metal plasmon resonance can be reasonably broad (e.g. FWHM of 0.5 eV) so a single surface can be used for obtaining the SERRS spectra for classes of molecules with similar electronic absorption maxima. A small range of particle shapes would broaden the range of wavelengths which could be used for a given surface. Another result of the ED theory relates to catalysis. It should be possible to determine for catalytic metals, such as the platinum group metals, if a particle size and geometry can be found which is both strongly SERS active and catalytically active. Such a surface would allow for in situ study of catalytic reactions with structural information on the species adsorbed on the metal surface. Finally, all metals are expected to enhance to some degree (Moskovits, 1985) The ED theory can determine the optimum surface geometry and magnitude of the expected enhancement for any metal. The only data required are the frequency dependent

PAGE 53

dielectric function of the metal, its bulk plasmon frequency, and its plasmon width. These data appear throughout the literature. Tables can be found for the noble metals (Johnston and Christy, 1972) the transition metals (Johston and Christy, 1974) indium (Theye and Devant, 1969) and others. Finally, several compilations of complex dielectric function data exist for many metals (Weaver et al 1981a, 1981b; Bass et al 1985; Hagemann et al 1974). Now that the ED theory can predict an optimum surface for a given SERS experiment, what is required experimentally is the ability to prepare surfaces with the particle morphology indicated by the calculations. However, the size and shape of the particles to be prepared are not the only important factors for practical SERS-active surfaces. A list of important factors in the preparation of these surfaces is given below. Ultimately, all of these factors have equal importance, but they are grouped here in two classifications; factors important from a purely physical viewpoint and factors important from a purely analytical viewpoint. These are differentiated as follows. Most experiments which detail new SERS substrate preparation methods are either interested in making surfaces which closely mimic the theoretical models, with little or no concern as to the speed or cost of the preparation, or in making surfaces which are easy, fast, and inexpensive, with little or no concern over the control of the actual surface morphology. Both methodologies of surface

PAGE 54

47 preparation are concerned with strong SERS activity, but the former is classified here as physical and the latter as analytical strictly for comparison with the current work presented in the following chapters. Factors of importance to the physical studies of the preparation of SERS-active substrates include: 1. Control of particle size. 2. Control of particle shape. 3. Density and uniformity of particle coverage. 4. Uniformity of particle sizes and shapes. 5. Compatibility of the surface morphology with theoretical models. The first two items in the list are directly related to the ability to prepare a substrate which conforms to the morphology predicted by the ED theory for optimization of a particular molecule-metal system. Control over the density and uniformity of the particle coverage allows for a surface to be prepared in which the maximum number of particles can be irradiated while keeping the distance between particles sufficient to ensure that they are noninteracting. Control over the uniformity of particle sizes and shapes allows a specific geometry of metal particles to be probed. Compatibility with the ED theory implies that particle shapes must resemble spheres or spheroids. Factors of importance to the analytical studies of the preparation of SERS-active substrates include: 1. Speed of the preparation method.

PAGE 55

48 2. Cost of the preparation method. 3. Reproducibility of the preparation method. 4. Molecular interactions with the surface. 5. Surface stability to incident laser radiation. The speed and cost of the preparation technique are important factors for adapting SERS to routine analysis. The technique must also consistently produce, within a chosen tolerance, the same surface features, since the SERS enhancement is sensitive to geometric changes in the particle sizes and shapes (Ni and Cotton, 1986) Molecular interactions with the surface are important since strong adsorption is needed to keep the molecule close to the enhancing particles. As will be seen in Chapter 5, chromatographic interactions between the molecule of interest, the solvent, and the surface can also potentially affect the reproducibility of the SERS signal. Finally, the surface must be stable to the laser radiation used, since photothermal damage and photo-oxidation can affect the intensity of the SERS signal (Laserna et al in press). An example of this effect will be presented in Chapter 4. The remainder of this dissertation is concerned with the development of preparation methods for SERS active substrates and the characterization of surfaces prepared using these techniques with respect to the guidelines stated above.

PAGE 56

CHAPTER 3 SERS ANALYSIS OF SULFONAMIDES ON COLLOIDAL SILVER Introduction Metal colloid solutions present an ideal surface for the merging of the physical and analytical interests for SERS substrates discussed in the last chapter (Moskovits, 1985) These solutions consist of isolated, noninteracting particles with essentially spherical shapes, ideal for theoretical treatment. They are quickly and readily prepared for many metals, stable, inexpensive, and the preparation methods can be scaled up for large quantities, making them ideal for routine analysis. The SERS activity of silver and gold colloid particles was first demonstrated by Creighton et al. (1979) for pyridine. Other early studies of colloidal SERS involved small molecules like carbon monoxide (Abe et al 1981) so that adsorption modes and vibrational assignments could be easily made. More practical applications have involved p-aminobenzoic acid (Suh et al 1983), proflavine (Koglin and Sequaris, 1986), and nucleic acid components (Kim et al., 1986), all on aqueous silver sols. SERS is not limited to silver, having been detected on aqueous sols of gold (Lee and Miesel, 1982), copper (Creighton et al 1983), 49

PAGE 57

50 platinum (Benner et al 1983), rhodium (Parker et al 1984), as well as silver chloride sols (Gao et al 1984) and goldplatinum alloy particles (Takenaka and Eda, 1985) Colloidal SERS is not, however, limited to aqueous media, as demonstrated by Garrell and Schultz (1985) who studied the SERS activity of silver colloids prepared in such solvents as tetrahydrofuran, acetonitrile, and N,N'-dimethylf ormamide. This has implications for the application of SERS on metals for which colloids cannot be prepared in a stable form in aqueous solutions. SERS has also been detected on matrix isolated silver (Reimer and Fischer, 1984) sodium (Rzepka et al 1981), and potassium (Schulze et al 1984) particles. This chapter presents the results of the use of colloidal silver solutions in the detection of sulfonamides, also known as sulfa drugs. Comparisons with "standard" methods for sulfonamide detection are made, and the possibility of using SERS for the speciation of structurally similar sulfonamides is discussed. Finally, some problems associated with colloidal SERS in general are described. Background The discoveries of sulfonamides (sulfa drugs) and antibiotics constitute some of the most significant medical achievements of this century. They have therapeutic uses in both human and veterinary medicine and in disease prevention in livestock. By 1985, over 100 million kilograms of these

PAGE 58

51 drugs were used annually. Over 5000 sulfa drugs, which have the general structure H2N-C^H^-S02-NH-R, have been synthesized and tested. Fewer than 3 0 of these have proven worthy of sustained use (Bevill, 1984) Their biological activity is based on the inhibition of the biosynthesis of folate cofactors in bacteria by blocking a step in the formation of dihydrofolic acid from p-aminobenzoic acid (Sigel and Woolley, 1979) The sulf apyrimidines where the R group is a pyrimidine ring, have been touted as the ultimate stage of development of sulfa drugs. Important members of this group are sulfadiazine, sulf amerazine and sulfamethazine (Sophian, 1952). During the 1940s, these drugs enjoyed wide spread use in combating bacterial infections in humans and in the treatment of diseases affecting pet and food-producing animals. Recently, the reduction of sulfa drug use in humans has been triggered by increased bacterial resistance to the drugs and the development of more effective antimicrobial agents. However, the use of sulfa drugs in veterinary medicine has persisted because the drugs are easily administered in feed and water, are economical, and have proven to be effective for the treatment of livestock diseases. The use of combinations of two or more sulfa drugs was also shown to be of therapeutic value. During treatment with sulfa drugs, the higher the concentration of drug that can be maintained in the body, the greater its effect. However, increasing the drug

PAGE 59

52 concentration also increases the side effects, the most common of which is crystalluria, a deposition of crystals in the renal tubules due to the limited solubility of the sulfa drugs in water. However, when used in combination, each drug acts independently. Since all of the sulf apyrimidines have been shown to have the same therapeutic effect, combination therapy allows a higher total drug concentration to be maintained in the blood while minimizing the effects of crystalluria. The use of sulfa drugs to promote growth and treat diseases of livestock animals has been a major cause of sulfa drug residues in swine marketed for human consumption. In 1973, the U. S. Food and Drug Administration (FDA) set a tolerance of 100 ng of sulfonamide per gram of edible tissue. Random assays of tissues for drug and agricultural chemical residues in slaughter animals and fowl are made monthly in all packing plants butchering for commercial sale (Bevill, 1984). At present, both the FDA and the Food Safety and Inspection Service (FSIS) rely on modifications of the Bratton-Marshall (B-M) procedure (Bratton and Marshall, 1939) for monitoring residues. The Bratton-Marshall reaction involves diazotization of the paraminobenzene group with nitrous acid and the subseguent coupling of the diazotized moiety with an amine to yield a stable, purple compound which absorbs strongly at 545 nm. One major disadvantage to this technique is that all sulfonamides form compounds with absorption maxima at 545 nm. The drug treatment history of slaughtered animals

PAGE 60

'I 53 may not be well known, and determination of which sulfa drug or drug combinations exist in the tissue could be important. A second disadvantage of the B-M technique is that interference is produced when primary aromatic amines other than sulfonamides are diazotized and yield highly colored products. Other methods which have been developed to overcome these problems include TLC (Sigel and Woolley, 1979) resonance Raman spectroscopy (Sato et al., 1980) and HPLC (Johnson et al., 1975). A comprehensive review of all the analytical methods used for sulfonamide detection was presented by Horwitz (1981, 1981a) A major defect inherent to all residue assay methods in current use is that the limit of reliable measurement is of the same order of magnitude as the tolerance they are intended to enforce (Bevill, 1984). Since surface-enhanced Raman spectroscopy is a relatively new technique for trace analysis and characterization, it is of interest to ascertain its utility in the determination of drugs. From an analytical point of view, the most outstanding characteristic of SERS is that it provides spectral fingerprint capability at trace concentration levels. In addition, fluorescence of the analyte or concomitants is not as serious a limiting factor in SERS as it is in regular Raman spectroscopy. Thus, while normal Raman scattering is not competitive with the previously tested methods for the detection of sulfonamides, SERS has the potential to be as

PAGE 61

54 sensitive as these other techniques while also providing structural information. In this chapter, the surface-enhanced Raman analysis of the three primary sulf apyrimidines (sulfadiazine, sulf amerazine, and sulfamethazine) is reported. Silver colloidal dispersions prepared by simple borohydride reduction of silver nitrate are used as substrates. The results illustrate the potential of SERS for trace analysis and characterization of analytes with close structural properties. Experimental Section Instrumentation The spectra reported here were obtained using the 514.5 nm line of an argon ion laser (Spectra Physics, series 2000) The nonlasing plasma lines were removed with a predispersing grating. The power at the sample was typically 100 mW. Spectra were obtained with a 0.85 m double grating spectrometer (Spex Industries, model 1403) The spectrometer resolution was set to 10 cm'\ The detector was a cooled RCA, model C31034, gallium arsenide photomultiplier tube coupled to standard photon counting electronics. Data acquisition, storage, processing, and plotting were under control of a personal computer. All spectra reported represent single scans and are provided without spectral smoothing. Right angle geometry was used for Raman sampling. A JEOL JSM-35C scanning electron microscope was used for Figure 3.1.

PAGE 62

55 Chemicals and Procedure All chemicals used were analytical reagent grade or equivalent. Demineralized water was used throughout. Sulfadiazine, sulfamerazine, and sulfamethazine were purchased from Sigma and used without further purification. Stock -1 aqueous solutions (100 /xg mL ) contained 3% methanol to help dissolve the drugs. Silver hydrosols were prepared as follows: a silver nitrate solution was prepared by adding 65 mL water to 35 mL of a 5.0 x lO"^ M aqueous AgNOj solution. This solution was ice-cooled and then added dropwise with vigorous stirring to a sodium borohydride solution prepared by adding 120 mL water to 180 mL of a 2.0 x lO'^ M aqueous NaBH^ solution. The NaBH^ solution was also ice-cooled. The resulting silver hydrosol had an absorption maximum at 400 nm, characteristic of silver particles with diameters between 1 and 50 nm (Creighton et al., 1979). This was experimentally verified by filtering a small amount of the silver colloidal solution onto a polycarbonate track-etched (PCTE) membrane and imaging the resulting surface with scanning electron microscopy (Figure 3.1) The silver colloid particles in this photograph have diameters between 5 and 2 5 nm. This was possible since the smooth surface of the PCTE membrane yields high contrast and is ideal for SEM imaging of small objects (Brock, 1983) Each sample was prepared by adding 0.1 mL of the sulfa drug solution to 0.9 mL of the Ag colloid at room temperature.

PAGE 63

56 After mixing, 0.5 mL of the resulting sample was transferred to the liquid cell and placed in the Spex model 1459 illuminator. Results and Discussion The surface-enhanced Raman spectra of sulfadiazine, sulf amerazine and sulfamethazine are shown in Figure 3.2. The numerical values of the band positions in cm'^ are listed in Table 3.1. The extreme structural similarity of the three drugs, which only differ in the presence of methyl group in the pyrimidine ring of the molecule, should be noted. The spectral fingerprinting capability of SERS at trace concentration levels is illustrated by the following analysis of the spectra. The drugs have several common spectral features. There is a narrow, very strong peak at 1114 cm"^ (sulfadiazine and sulf amerazine) or at 1112 cm"^ (sulfamethazine), a strong band at 1594 cm \ a broad band peaking at around 3390 cm"\ with shoulders at 3070 cm"^ and 3258 cm"\ and a peak at around 580 cm ^ is also present. Sulfadiazine and sulf amerazine have a shoulder at 230 cm ^ on the sloping background of the Rayleigh scatter. In the case of sulfamethazine, there is a well defined band at 238 cm"\ Sulf amerazine and sulfamethazine have several spectral singularities with respect to sulfadiazine. Both have a common mode at around 2930 cm"^ not observed for sulfadiazine.

PAGE 64

73 0) U 0) 4J H •H a. C o •H P 4-1 O O 0) W N -H o (U o (d -P 52 M-t ^
PAGE 66

0) c •H N (0 u 0) (0 iH 0) c •H N (0 •H -o (0 4-( rH (4-1 o (0 p o 0) T3 •H o rH rH o o c o o p +J o XI 0) c 03 N (0 (0 n) 0) w -0 c (0 u c (13 x: c o (0 T3 -H 0) 3 CP -H

PAGE 67

(AjDj^iqjv) A:nsu9:^u| sd3S

PAGE 68

61 Table 3.1. SERS peaks of sulfonamides on colloidal silver. (Raman shift', cm' ) Sulfadiazine 230, sh 294, sh 364, vw 416, vw 458, vw 506, vw 574, m 634, w 684, w 818, s 974, vw 1068, sh 1114, vs 1186, vw 1340, m 1410, w 1502, w 1594, s 1630, sh 3050, sh 3258, bb 3378, bb Sulf amerazine 230, sh 290, sh 342 vw 410, vw 462 vw 508, vw 544, vw 582 m 626, vw 674, w 762, s 830, m 976, vw 1002, vw 1066, sh 1114, vs 1174, vw 1252, vw 1312, m 1390, m 1454, w 1502, w 1594, s 1630, sh 2930, m 3068, sh 3258, bb 3396, bb Sulfamethazine 238, s 396, vw 458, vw 516, vw 554, w 590, m 638, vw 674, w 738, m 834, m 968, bb 1054, w 1112, vs 1186, vw 1260, vw 1302. vw 1356, sh 1394, m 1506, w 1594, s 1622, sh 2922, m 3074, sh 3242, bb 3380, bb a. V, very weak; w, weak; m, medium; s, strong; vs, sh, shoulder; bb, broad band very strong;

PAGE 69

62 -1 -1 The peaks at 738 cm (sulfamethazine) and 762 cm (sulf amerazine) do not appear in the case of sulfadiazine. Sulf amerazine shows a weak mode at 1454 cm \ which is not shown by the other two drugs. Sulfamethazine can be distinguished from the other two drugs by a number of spectral features. For instance, it shows a broad band at 968 cm"\ which in the other two drugs appear as very small peaks at -1 -1 around 975 cm Also, the combination of a band at 1394 cm with a shoulder in the high energy side at 1356 cm^ is unique to sulfamethazine. No attempts were made to provide a complete vibrational analysis of the compounds. However, on the basis of group frequencies, tentative assignments of the most prominent spectral features can be made. The very strong 1112-1114 cm'^ band common to the three drugs could be an SOj symmetric stretch (Dollish et al., 1974; Baranska et al., 1987). The 1594 cm ^ mode in the three compounds could be a ring stretching mode. The shoulder at around 1630 cm'^ could be an NHj bending mode. The peaks between 3050-3074 cm'^ could be aromatic C-H stretching modes, while those at 3258-3380 cm"^ could be due to N-H stretching modes. The band at 2930 cm"^ in sulf amerazine and that at 2922 cm"^ in sulfamethazine are presumably due to the methyl C-H stretching vibration, since no band in this region appears in the spectrum of sulfadiazine, which lacks a methyl group on the pyrimidine ring.

PAGE 70

63 The normal Raman spectra of the pure drugs (solid state) were also recorded (Figure 3.3). Peak positions are listed in Table 3.2. In general, the vibrational modes in the SERS spectra are shifted with respect to the NRS. The most prominent peak in the NR spectra occurs at 1150 cm"^ (sulfadiazine) at 1153 cm'^ (sulf amerazine) and at 1144 cm'^ (sulfamethazine) while the corresponding peak in the SERS spectra appears at 1114 cm'\ 1114 cm'\ and 1112 cm'\ respectively. There were smaller shifts for the second most intense mode: 1598 cm"^ (sulfadiazine) and 1596 cm'^ (sulf amerazine and sulfamethazine) in the NRS, and 1596 cm"^ for the three drugs in the SERS spectrum. The bandwidths (FWHM) of the solid state spectra are smaller than those in SERS. For instance, the FWHM for the NRS band of sulfadiazine -1 -1 at 1598 cm was 12 cm while the same vibrational mode in the SERS spectrum (1594 cm'^) has a FWHM of 22 cm'\ It should be noted, however, that the NRS are superimposed on a strong background, presumably due to fluorescence of the drug or of some impurity. Quantitative Study Experiments were conducted to study the dependence of SERS intensity on the adsorbate concentration added to the silver colloid. It was observed that for the drugs studied there was a dynamic component in the measured SERS intensity. Addition of the drug to the hydrosol did not result in an immediate development of the intensity. Figure 3.4 shows the

PAGE 71

m c :3 o 04 o o 0) u 0) c •H (0 -p u 0) c H N (0 0) e n "J rH -o o (0 <-l OJ H C 3-H W N (0 Q) X! 43 -P P Q)
PAGE 72

65

PAGE 73

66 Table 3.2. Normal (Raman Raman t?h i f 1-' peaks of sulfonamides on cm' ) colloidal silver. Sulfadiazine Sulf amerazine Sulfamethazine 229, sh 237, w 248, vw 286, w 282, dbl. w 290, vw 319, vw 300, dbl. w 330, vw 339, vw 328, w 342, vw 373, vw 380, w 386, vw 402, vw 456, w 439, w 455, vw 547, m 526, vw 544, w 585, w 538, w 578, vw 634, w 567, vw 636, w 659, vw 579, w 661, vw 681, w 637, w 706, vw 715, vw 685, w 750, vw 743, m 716, vw 800, w 769, sh 829, dbl, m 827, dbl, m 822, dbl. m 842, dbl, m 849, dbl, m 840, dbl. m 873, vw 939, vw 892, w 999, w 995, m 999, mw 1092, w 1098, s 1094, s 1144, vs 1150, vs 1153, vs 1306, vw 1259, w 1191, w 1345, vw 1312, w 1241, vw 1385, vw 1338, w 1298, w 1416, vw 1407, vw 1333, w 1440, vw 1438, w 1371, vw 1466, vw 1505, w 1413, w 1478, vw 1581, sh 1506, m 1509, w 1598, vs 1560, w 1559, vw 1596, vs 1596, vs 1629, w 1640, w a. V, very weak; w, weak; ra, medium; s, strong; sh, shoulder; bb, broad band vs, very strong;

PAGE 74

67 time dependence of SERS intensity for sulfadiazine. There is a rapid increase in intensity during the first minute after mixing hydrosol and drug (colloid activation) followed by a slow decrease in intensity. Activation of silver hydrosols has been reported previously (Laserna et al., 1987a; Berthod et al., 1987; Blatchford et al., 1982; Kerker et al. 1984). All subsequent experiments were made with careful attention to the timing of the measurement, with spectral scans starting 30 s after mixing hydrosol and drug. Figure 3.5 shows the SERS spectra of sulf amerazine at three different concentrations. As shown, at lower concentrations, the SERS spectral features are poorly defined. Several bands can be seen in the spectrum corresponding to 1 ng mL of sulf amerazine, including two apparent peaks between 600 cm'^ and 1100 cm"^ and a broad band between 1200 cm'^ and 1700 cm \ These features remain constant for all concentrations of the sulfa drugs used and are also present in the blank. Possible sources are contaminants, luminescence background of the metal substrate (Heritage et al., 1979), or graphitic carbon (Cooney et al., 1983). These bands may affect limit of detection of sulfa drugs by colloidal SERS, although the 1112 cm"^ peak, which would be used for trace detection of these drugs, does not overlap any of these background features. Figures 3.6 and 3.7 show the concentration dependence of sulfamethazine and sulfadiazine, respectively.

PAGE 75

u I (Q H o 0 c o 0) c •H N
PAGE 77

u 0) (0 > 0) 4J c (U 0) <4-l <)-l -^^ 0) (C (1) U U c •H -p O P O 0) a) CO S5 o u 0) g (0 w (H a) c (U -p c •H 0) > o (0 u -p u Q).H &I4J W ro H C 0) TJ : C (0 c 0) (0 p c o If) n 0) U

PAGE 78

71 O — 1_Q CNj (AjDJiiqjv) A|isu9;u| sy3S

PAGE 79

CP U 0) rs 0 (0 > 0) -p -p c 0) 0) <*-( (U -P w (0 0) u -c ft 0) 0) c u (0 (0 o 0) 0) c •H Q) N X! (0 4J Q) O (0 :3 CO 0) > 0) -H +J -H
PAGE 80

73 CNJ (AjDj:^iqjV) A:^isu9:^u| sd3S

PAGE 81

U Q) rH 2§ 0) u -p p c 0) u (0 V4 4J (0 T3 0) c •H (0 o a) 0) -p o -0 H +J •H c 0) c -H (0 O 10 u Q) > "1^ c (0 (0 0) 43 Eh (U c (0 43 C 4J 0) n) I M o c 0) cn -H 1^

PAGE 82

75 O 1_Q CN (AjDJ^iqjv) /::nsu9:^u| sd3S

PAGE 83

76 Table 3.3 summarizes the linear least squares fit of the calibration curves for the sulfa drugs. The SERS intensitieswere measured at the most prominent peak of each drug, i.e., 1114 cm ^ for sulfadiazine and sulf amerazine, and 1112 cm ^ for sulfamethazine. Although other peaks could be chosen to monitor the intensity changes with concentration, this peak has the following advantages: (a) it provides the largest dynamic range, 2 orders of magnitude for each drug; (b) it allows the maximum sensitivity (slope) since the concentration dependence of the intensity is the largest; and (c) the limit of detection is improved since the selected peak is superimposed on a much lower background than that for the other peaks. For instance. Figure 3.5 shows that the peak at 1114 cm ^ in sulf amerazine is over a lower background than the second most intense peak, at 1594 cm'\ Virtually the same arguments can be used for the choice of the monitoring Raman shifts for the other two drugs. Table 3.3 shows that the sensitivity changes with the drug considered. The sensitivity is largest for sulf amerazine and smallest for sulfamethazine. The correlation coefficients were adequate in the three cases. Limits of detection calculated for a signal-to-noise ratio of 3:1 are 1 ng mL ^ for sulfadiazine and sulf amerazine and 10 ng -1 mL for sulfamethazine.

PAGE 84

77 Table 3.3. SERS Calibration data for sulf apyrimidines on colloidal silver. Calibration r'' equation^ Sulfadiazine y = 66.74 x + 679.8 0.997 Sulfaiiierazine y = 143.82 x + 771.2 0.999 Sulfamethazine y = 16.62 x + 751.3 0.999 a. Linear least squares best fit line. Y represents the SERS intensity in counts per second at 1114 cm (sulfadiazine, sulf amerazine) or 1112 cm' (sulfamethazine) X represents the drug concentration in ppb. The concentration range used for all three least squares calculations was 1 ppb to 100 ppb. b. Correlation coefficient. Conclusions These results demonstrate that silver hydrosols can be developed as a practical substrate for the SERS detection of sulfonamides. They are an attractive approach because of the ease of production and manipulation. However, SERS intensities in this substrate are determined by an activation period which must be studied and optimized in each experimental situation. Chapter 6 discusses some practical solutions to the colloid instability problem. In addition, extremely high sensitivity as well as quantitative information can be obtained with careful attention to the timing of the measurement process.

PAGE 85

78 This study constitutes a laboratory test on pure sulfa drugs. Although limits of detection for the colloidal SERS technique are found to be below 10 ng mL\ the limit of reliable measurement, used by the FDA and FSIS, may be somewhat larger. Methods for obtaining the sulfa drugs from real tissue samples that is compatible with the colloidal SERS experiment need to be developed. Several of these methods are described by Horwitz (1981, 1981a). Also, as is the case with TLC, HPLC, and all methods besides the Br atton-Mar shall technique, collaborative studies among different laboratories need to be performed in order to ascertain the ultimate usefulness of colloidal SERS detection in the monitoring of sulfa drugs in edible animal tissue.

PAGE 86

CHAPTER 4 MORPHOLOGY /ACTIVITY STUDIES OF NEW SERS-ACTIVE SUBSTRATES Introduction When analytical chemists turned their attention toward the possibility of using the technique of surface-enhanced Raman spectroscopy for routine analysis, one of their major concerns was the lack of preparation methods for SERS active substrates. The most important factors for such methods were that they be inexpensive, simple, and fast, with reproducible, strong SERS activity, especially at the argon ion laser wavelength of 514.5 nm, the most common excitation source for routine Raman spectroscopy. Little concern was given to the details of the surface morphology of the prepared substrates and their ability to be modeled with the electrodynamic theory. These analytically motivated surface preparation methods include the coating of fumed silica with silver by vacuum vapor deposition (Alak and Vo-Dinh, 1989) the coating of silver onto frosted glass slides using a Tollen's reduction of silver nitrate (Ni and Cotton, 1986) the coating of silver onto smooth microscope slides by direct chemical reduction using both the Rochelle salt process and the Brashear process (Boo et al., 1985), and the coating of silver onto Whatman #1 79

PAGE 87

80 cellulose filter paper (Berthod et al 1988; Laserna et al 1988a), and Whatman #5 cellulose filter paper (Laserna et al in press) by direct chemical reduction of an aqueous silver nitrate solution with an aqueous sodium borohydride solution. Another method of note involves evaporating silver onto polystyrene spheres which have been coated onto Whatman #50 filter paper (Vo-Dinh et al 1984) and glass slides (Moody et al., 1987). This method allows for the control of particle size by the choice of the polystyrene sphere size and control of the particle density by control of the initial concentration of spheres in solution. Experimental optimization of the sphere size, density, and coating thickness is given by Vo-Dinh et al (1989a) and this surface has been shown to have practical use in the detection of organophosphorous chemical agents (Alak and Vo-Dinh, 1987) and nitro polycyclic aromatic hydrocarbons (Enlow et al 1986). Two of the more unique surface preparation methods, mentioned here for completeness, involve the observation of a spectra sensitizing dye adsorbed to silver bromide crystals in a photographic film (Brandt, 1988), and the detection of SERS on silver nitrate photoreduced in solution (Ahem and Garrell, 1987) A number of researchers have attempted to develop methods to prepare metal substrates with control over the particle shape and size. These physically motivated surface preparation methods include the microlithographic production

PAGE 88

81 of ordered arrays of quartz used as templates for evaporated silver (Liao, 1982), stochastic arrays of quartz posts used as templates for evaporated silver (Meier et al 1985; Vo-Dinh et al 1986), and island films of metals such as silver and indium (Jennings et al 1984). Although these methods produced some of the most strongly SERS-active substrates prepared, the time involved in using these methods (e.g. several hours for 1 cm^ for the ordered quartz post array) is prohibitive for routine analysis. Two excellent reviews of the relative merits of all of the substrates mentioned above have been published (Vo-Dinh et al 1988; Vo-Dinh, 1989). The common element in the majority of the above surface preparation methods is the evaporation of metallic silver or the reduction of a silver salt onto a solid support. These two techniques allow for the rapid and inexpensive deposition of silver onto a surface. The most difficult aspect of these methods is the control of the size, shape, uniformity, and density of the silver surface features deposited. Choice of the support to be used is one of the major determining factors in obtaining this control. This chapter describes the use of these two techniques on a variety of new solid supports with the intent of improving the uniformity and density of the silver particles deposited onto the surface. The solid supports used include two small pore size cellulose filter membranes from Millipore (0.025 jum and 0.22 fMja) frosted glass slides, and a native silver filter

PAGE 89

82 membrane. These substrates and deposition methods were chosen since they are fast, inexpensive, and simple, making them ideal candidates for routine use. Experimental Instrumentation The instrumentation used for these experiments was described in Chapter 3. In addition, micro Raman analysis was conducted with a Spex Micramate microprobe, consisting of a modified Zeiss 20 research-grade microscope fitted with a lOX objective, for coarse sample adjustment and location with the X-Y microscope translation stage, and a 4 OX objective, for fine tuning and analysis. The spatial resolution with the 4 OX objective is 2 ura. The sample scatter is collected at 180 and focused on the entrance slit of the spectrometer by spatial filtering and a floating lens. The sample viewing system consisted of a color television camera mounted to the microscope and a 10-inch color monitor. The TV color system can be adjusted to introduce artificial colors to enhance features of poor-contrast samples. Switching from sample viewing to spectral analysis was done with a rotating prism. A Balzers, model Micro BA-3 high vacuum coating unit was used for evaporating of silver on the supports. Photographs of the substrates as they appear under the microscope were taken with a 35 mm camera directly off the video camera monitor.

PAGE 90

83 Scanning electron micrographs were taken using a JEOL JSM-35C SEM. Chemicals and Procedure All chemicals used were analytical reagent grade or equivalent. Demineralized water was used throughout. 9Aminoacridine hydrochloride monohydrate was purchased from Aldrich and was used without further purification. Fisher Finest frosted glass microscope slides were obtained from Fisher. The Millipore filter papers (0.025 /xm and 0.22 /xm pore size) and the silver membranes (0.45 nm pore size) were obtained from Millipore. i Chemical reduction of silver salts on the substrates was done using a published method (Laserna at al 1988a). The filter paper was immersed in a 0.1 M aqueous solution of silver nitrate and sprayed with a 0.2 M aqueous solution of sodium borohydride from a distance of about 2 0 cm for 30 s. An atomic absorption nebulizer actuated by compressed air was used for spraying. Silver evaporation was performed by mounting the filter paper in the vacuum coating unit with double-stick tape and resistively heating a small amount of silver onto the surface of the paper or slide. Silver thickness was determined by weighing the filter papers before and after coating. The estimated thickness of the silver layer using this method is between 250 nm and 1 /xm.

PAGE 91

84 Results and Discussion Raman Microprobe Studies All of the initial studies were performed using the Raman microprobe. Since both vapor deposition and chemical reduction have been shown to produce SERS-active substrates on cellulose filter papers (Laserna at ai 1988a; Vo-Dinh et al., 1987) and frosted glass slides (Ni and Cotton, 1986), the important factor to be considered here is how these new supports affect the silver surface morphology and the uniformity of the silver coverage, as well as the SERS activity. The Raman microprobe is used in these studies since it probes a spatial area with a diameter of 2 fim, so it is much more sensitive to submicron substrate uniformity than normal Raman experiments which typically use a beam diameter of between 0.5 mm and 1 mm. In fact, it has been shown that, on a Whatman #5 filter paper substrate coated using chemical reduction, where the silver particles have a wide range of sizes and shapes (Laserna et al 1988a) using a larger probe beam actually averages the signal from a large number of these particles. The result was a sample to sample reproducibility of 15 percent (Laserna et al 1989). For all the substrates presented in this chapter, the surface uniformity is described for two levels of detail. The

PAGE 92

85 first is the uniformity with respect to what is seen through the microscope and is shown through photographs taken of the surfaces off the TV monitor output of the camera attached to the microscope. The viewing area is approximately 110 /xm by 90 /xm, and the smallest discernible surface feature is on the order of 1-2 /xm. The second is the submicron uniformity of the surface morphology and is shown in the scanning electron micrographs of each surface. Finally, the relative SERS activity of each substrate is tested with 9-aminoacridine (AA) which has been shown to be an excellent SERS scatterer (Laserna et al 1989). Vacuum vapor deposition Millipore 0.025um filter membrane This membrane is a mixture of cellulose acetate and cellulose nitrate in a ratio of 85:15. One side of the filter is much more reflective than the other. This is a feature of how the membranes are made, but the two surfaces have no significant physical differences (Brock, 1983) For this study, the more reflective side was coated with silver. After coating, the surface resembles a mirror, indicating a thick, uniform coating. However, the microscope monitor photograph of this surface seen in Figure 4.1 reveals that the surface coating is far from uniform. Spotting of 2 juL of a 30 ppm aqueous solution of 9aminoacridine results in a final spot diameter on the surface of 1 cm. Figure 4.1 also shows the SERS spectra of AA from two different areas on this spot. The intensity difference

PAGE 93

>1 u (U > rH •H U) X! -P T3 0) P to O O N •rH U 0) O a a. IT) o •H rH ^> rH 0 •H (0 s > 0) cn •rH 1X4

PAGE 94

87

PAGE 95

88 between the two spectra is an indication of the difficulty of obtaining reproducible intensities for a given analyte spot. The SEM images reveal that the surface is a combination of flat surfaces and holes, with the flat surfaces often extending over tens of microns. This may explain the peppered appearance of the coated membrane under the microscope. The SEM images also indicate that the silver coverage is uniform on these flat surfaces but relatively featureless. The SERS activity of this surface is smaller than for some of the other substrates prepared, a result which is not surprising given the lack of distinct surface roughness features. Millipore 0.22 um filter membrane This membrane is also a mixture of cellulose acetate and cellulose nitrate (85:15) and exhibits the same difference in the reflectivity of the two sides as does the 0.02 5 /xm membrane. Once again, the more reflective side was coated and resembles a mirror. Figure 4.2 shows that this surface is very uniform when viewed by the microscope. 2 /xL of the 30 ppm AA solution was applied, with a final spot diameter of 1 cm. Figure 4.2 also gives SERS spectra from several places within the analyte spot. Once again, the intensity differences observed in these spectra indicate difficulties in the reproducibility of peak intensities. The SEM images reveal the fibrous nature of the membrane and show that the silver appears to be uniformly coated over all the fibers. This surface is somewhat more active than the 0.025 nm membrane, and this is not surprising

PAGE 96

0) > •H U) X3 -P T3 0) -P (0 O O 0) N •H U 0) o a. 0) C (0 ^ Si 0) 0) -p • •H O M-l -H P 0) -H (0 tn o o rH Q) V^ O O 0) n] U > O o > -H 0) 3 •H (X4

PAGE 98

a. T3 (N C eg as • O (0 n7 Q) o • rH 4J &> •H f\ 3 o rH Q) 0) (0 0) >'0 Q) C H (0 (0 T> +J rH 0) (0 +J T3 o o o O 4J (0 -P -T3 O 0) C Q) N O (WWW 0) 0) 0) ^ >H >H o o a o 0) •H

PAGE 99

92

PAGE 100

93 considering the increased submicron roughness features and "cracking" of the metal coating on the fibers. One experimental problem noted with this surface was that the high power of the laser beam, which can exceed 10^ W/cm^ (Van Duyne et al., 1986), damaged the silver coating. Figure 4.3 shows the spectra of a 2 /xL spot of the 30 ppm AA solution for two scans taken consecutively from the same spot. The laser damage after one scan, as seen in Figure 4.3, results in approximately a fifty percent decrease in the SERS intensity, but the quality of the second spectrum is still quite good. This indicates that the surface is still strongly SERS active. It is interesting to note that the damaged area extends over a diameter of 60-70 /xm, even though the laser beam diameter at the surface is only 2 iim. This damage is likely due to either a local heating effect or some photo-oxidation process, but it was not possible to confirm either of these. Frosted glass slides. It was hoped that the roughness of the frosted glass would induce silver surface roughness features compatible with strong SERS activity, as was apparently the case for Ni and Cotton (1986). Figure 4.4 shows the coated slide and graphically indicates one of the major problems with this surface. The glass particles are quite large, inhibiting the focusing of the microscope on a particular area of the surface. Another practical problem with this surface is that the 2 /xL spot of 30 ppm AA applied to the surface spread into an area with a diameter of over 3

PAGE 103

96 cm, greatly diluting the AA concentration probed by the laser beam. Figure 4.4 indicates not only the low SERS activity of this substrate but the difficulty of obtaining reproducible spectra at various spots on the surface. Spotting volumes as small as 0.2 /xL did little to alleviate this problem. The electron micrograph image of this surface is interesting since it reveals that the large glass particles of the slide are covered with a dense array of smaller particles, on the order of 0.1 nm in diameter, which resemble island films. This surface was not affected by the high laser power of the micro Raman experiment. Chemical reduction Millipore 0.02 Sum filter membrane The chemical reduction of silver onto this membrane was unsuccessful in producing a SERS-active substrate. The membrane was soaked in the silver nitrate solution and then sprayed with the sodium borohydride solution. The borohydride reduced the silver nitrate, but the resulting metallic silver remained mostly suspended in solution and deposited randomly over the surface as the solvent dried. No SERS activity was observed for 2 /xL of the 30 ppm AA solution. This experiment was repeated on the less reflective side of the membrane and, although the random deposition of silver over the surface was less than on the reflective side, no SERS activity was detected. Millipore 0.22 um filter membrane The difference in the deposition of the silver on the reflective and dull sides of

PAGE 104

97 the 0.025 fim membrane indicated that both sides of this filter be tested as well. Figure 4.5 shows the membrane under the microscope and indicates that the silver deposition is not uniform on this scale. The results of the SERS spectra from both sides indicate that both surfaces are strongly SERS active, with the dull side yielding the highest overall intensity. The average intensity is very similar for both sides. It was observed that the 2 /iL drop of 3 0 ppm AA took longer to absorb into the membrane on the reflective side than on the dull side. This may add to the differences in intensities observed. The spectra in Figure 4.5 are from the same analyte spot on the dull side of the membrane, indicating the difficulty in obtaining reproducible intensities. The SEM images show a startling interaction between this membrane and the reduced silver. The fibers are coated with highly dense, uniformly sized "islands" of silver. This morphology is consistent over the entire membrane, with some fibers having a larger particle number density, possibly causing the mottled appearance observed with the microscope. These silver features range in size from 50 to 100 nm with hemisphere or hemispheroidal shapes. Some clusters are also evident with sizes up to 200 nm. All of these particles are of an appropriate size and shape to be SERS active and, in fact, this surface provided the most intense AA peaks of all the substrates tested.

PAGE 105

u 0) > rH •H (/] £1 •H > T3 0) -P n3 O O 0) N •H tn Q) o a. a) c (0 e 0) e 0) p o •H •H +J M-l O ^1 0) T3 (0 0) O >^ rH 3 H rH (0 rH O O g Q) ^ O O a-P •H O rH (U rH >H •H -rl in (U u CP

PAGE 107

100 silver membrane filters In the case of silver membranes, no preparation was needed. SERS was first reported on silver membranes in a review by Vo-Dinh (Vo-Dinh et al., 1988), although no specific details of surface morphology were given. Figure 4.6 shows the membrane as observed in the TV monitor, and using an electron microscope. Figure 4.6 also shows SERS spectra taken from a 2 /LiL spot of 30 ppm AA. The final spot size had a diameter of close to 3 cm, similar to the result seen for the frosted glass slide. The lack of SERS activity observed is probably due to, in part, the low concentration of AA under the probe beam as a result of this spreading. However, the weak SERS signal is more likely caused by the absence of silver roughness features in the size range needed for SERS activity (1-100 nanometers). As observed in Figure 4.6, the silver structures on the membrane appear smooth. No smaller roughness features could be observed with magnifications as high as 100,000. Although the quality and intensity of the spectra obtained for AA are not comparable with the other substrates studied here, these membranes are commercially available, quality controlled, and show no surface damage with the high laser irradiances of the micro Raman system. Raman Macroscopic Studies One major problem which results from the microscopic studies described above but has not yet been addressed

PAGE 108

4) N •H n u o I in u 0) •p •H IH Q) c I > n 0) O VO 0)

PAGE 109

102

PAGE 110

103 pertains to the problem of obtaining reproducible SERS intensities from within a single analyte spot. The two most likely sources for this problem are the difference in activity of the microscopic regions probed or a difference in the concentration of the 9-aminoacridine probe molecule in these regions. It is possible to eliminate one of these effects by choosing a larger probe beam. The SERS signal will then be an average of a large number of particle sizes and shapes (Laserna et al 1989), so that each area probed will contain the same distribution of particles. Any difference in the intensities of SERS spectra taken from different areas of a given analyte spot will arise from AA concentration differences. The use of a larger beam will also determine if this variation of signal intensities observed with the Raman microprobe is a purely microscopic effect or if it affects standard macroscopic Raman experiments as well. For these experiments, a specially designed solid substrate sample holder was constructed (Figure 4.7). The sample holder (SH) consisted of a right circular cylinder of aluminum with a semicircular section removed for sample placement to keep the samples flat. The solid substrates were constrained by a cover plate (CP) A 1 cm diameter circular window on the cover plate limited the exposed substrate area. The sample holder was placed in a frame (F) which was mounted on an xyz translational stage with micrometer adjustments, allowing for precise positioning of the substrate under the

PAGE 111

U g 0) (0 T5 Q) O (1) (0 rH <0 0) & (0 P Ui (0 0) •H (TJ c o (0 (0 o --^ •H W O ?3 p •• u % • TJ Q) 0) .. € Ul El. O :i u V^ 0) U Q) 4J 0) TJ (0 a O 04 rM CO (U QJ rH > & 0 w S u "(^^ 2 ;J S 2 w CO CO O Eh u t7<

PAGE 112

105

PAGE 113

106 laser beam. The stage was on a kinematic mount for precise positioning within the sample chamber. The sample holder had free rotation on the frame to allow studies of the effect of the incidence angle of the laser beam on the SERS intensity. In Figure 4.7, the inset represents the incidence angle of the light on the substrate relative to the laser beam (L) and the transfer optics and spectrometer (TO,S) Right angle geometry between the laser and the collection optics was used. To minimize the difference between the sample spot size and the laser beam, the sample volume applied was consistently 0.2 /xL. Typical spot diameters were 1-2 mm and the laser probe beam at the surface is between 0.25 and 0.5 mm. Four serial dilutions of a 3 00 ppm aqueous solution of AA were used for each substrate. Spectra were taken at various places within each sample spot on the surface. The vacuum vapor deposition technique was chosen for this study since it provided the most uniform silver coating of the techniques described and could be used for all the solid supports. The membrane chosen for these macroscopic tests was the 0.025 fjim pore size cellulose membrane One difference immediately evident using the smaller volumes was that the spot size was smaller (1-2 mm in diameter) and, more importantly, that the spot does not appear homogeneous. The final spot shape consists of a grey center disk, similar in color to the metal-coated blank membrane, surrounded by a faint yellow ring. This structure was

PAGE 114

107 consistent for AA concentrations ranging from 30 ppb to 3 00 ppm and was the same for silver deposited on the reflective side and the dull side of the membrane. It is important to note that AA in aqueous solution is yellow and that this color is not readily visible to the eye for concentrations below 3 ppm. From a practical perspective, the main difference between the two sides of the membrane is that the analyte spot absorbs much more quickly for the dull coated side than for the reflective side. Spectroscopically these two sides are radically different. No SERS signal was obtained on the reflective coated side until the concentration of AA was 3 ppm. Only slightly stronger peaks were obtained for a 30 ppm solution and no further signal increase was obtained for a 300 ppm solution (Figure 4.8). In contrast, SERS peaks for AA were observed for concentrations as low as 30 ppb for membranes coated on the dull side, with a maximum signal for an AA concentration of 3 ppm. For concentrations above 3 ppm, the intensity remains essentially constant. It should be noted that the spectra represent experimentally optimized signals for each region. The maximum SERS intensity for each concentration for both sides of the membrane was obtained when the laser beam was placed directly over the boundary between the grey center disk and the yellow outer ring. Figure 4.9 shows the optimum SERS signal obtained for various concentrations of AA and Figure 4.10 shows the intensity

PAGE 115

0) e u d o 3 • ft O g •H (0 ft H > ft H •H >iO S XI O n •H > ^ rsi W ft o •r. (0 o 0) rH 00 0)

PAGE 116

109

PAGE 117

0) u u o o a •H > O (0 > •H 0) N •H (0 >i ?) u S > an •H e w s £ ft n 5 o o •* n ^•^'^ c o ft ft n O 0) o *^ ^ o W O o c ft O "O ft -H 0) ^ +j 4J O rH O <^ •H O fl) e ft "go o 3 — (0 -H -P O 0) 0) in ft o c o -p •H (0 o 0) T3

PAGE 118

Ill (sdo) X:nsu8:^u| sd3S

PAGE 119

o X) o o n (0 o -P O 0) p o aw w (0 c (0 +J o 0) c o ^ Q. 0) o 4J c u 0) (4-1 (H •H -p o 04 (/I 0) XI p o M-l 0-1 O 1) 04 0) o H 0) •rH

PAGE 120

113 (sdo) |Du5is Sd3S

PAGE 121

114 difference for scans taken at the center of a spot and at the optimized position for an AA concentration of 3 00 ppb. Conclusions It is clear that the SERS signal for both microscopic and macroscopic experiments is strongly dependent on the placement of the laser probe beam within the analyte spot. The two prime candidates for the cause of this effect are differences in the surface morphology, and thus SERS activity, within an analyte spot, and concentration differences within the analyte spot, caused by nonuniform deposition of the analyte molecules on the surface. The first of these two possible causes has been addressed here. SEM data indicate that the surface morphology on a submicron scale is constant. Thus, the average number and types of particles probed by the Raman microprobe for different areas within an analyte spot is the same. Secondly, the macroscopic study within an analyte spot indicates that, even by averaging thousands or millions of particles with varying sizes and shapes by using a large probe beam, the SERS signal is still dependent on the placement of the laser beam within the analyte spot. It is, therefore, difficult to obtain a reliable calibration curve for these surfaces, hindering their possible use for routine analysis. These results indicate that the concentration of analyte within a spot is not uniform. The next chapter describes the

PAGE 122

115 results of a more detailed analysis of analyte distributions within a spot on a SERS-active surface.

PAGE 123

CHAPTER 5 SPATIAL DISTRIBUTION STUDY OF SERS ACTIVE SUBSTRATES Introduction Many of the SERS-active substrates which have been prepared with the idea that they be used for routine analysis involve the deposition of a metal, usually silver, onto a solid support. A variety of these techniques and the supports prepared using them were described in the previous chapter. Some new solid supports were presented and analyzed with respect to their surface morphology and SERS activity. Both macroscopic and microscopic Raman configurations were employed. The majority of Raman data which is published in the literature involves the use of macroscopic systems, so this chapter will focus on SERS experiments on solid supports using this type of configuration. In many of the experiments using these substrates, the analyte to be studied is applied to the prepared surface in volumes ranging from 0.2 to 5.0 /xL with a microsyringe. The resultant spot size on the surface of the substrate is typically 2 to 10 mm in diameter. Many commercial Raman instruments, including the SPEX Raman system used for these experiments, focus the laser beam to a diameter of a few tenths of a millimeter at the sample. Thus, in 116

PAGE 124

117 typical solid substrate SERS experiment, the diameter of the laser beam is an order of magnitude less than the diameter of the analyte spot. From an analytical standpoint, it becomes important to determine the effect on the Raman signal of the placement of the laser beam in the analyte spot, as indicated by the results of the previous chapter. Another important factor to be considered in this type of experiment is the effect of the relative size of the laser beam at the substrate surface to the size of the analyte spot. In this work, we explore the spatial distribution of the analyte concentration of an aqueous solution of a test compound, 9-Aminoacridine (AA) on the surface of a cellulose filter paper coated with silver via a direct chemical reduction. We also examine the effect of the relative sizes of the analyte spot and the laser spot on the SERS signal. Experimental Instrumentation Spectral data were obtained using the SPEX Raman system described in Chapter 3 Typical laser power at the sample was 20 mW. Chemicals and Procedure All chemicals used were analytical reagent grade or equivalent. Demineralized water was used throughout. 9Aminoacridine hydrochloride monohydrate was purchased from Aldrich and was used without further purification. Whatman #5

PAGE 125

118 filter paper was used as a support for a SERS-active silver substrate. The procedure for coating the support with silver by chemical reduction was described in Chapter 4 The SERSactive substrate was allowed to dry. Square pieces (1.5 cm per side) of the substrate were placed in the sample holder (Figure 4.7) and a volume of 0.2 /xL of an aqueous solution of 9-Aminoacridine (3 mg/mL) containing 0.6 /xg of AA, was applied to the substrate with a microsyringe (Hamilton, #901) and allowed to dry before spectral data were obtained. The sample produced a roughly circular spot on the substrates with a diameter of 2 mm. The translational stage was then moved laterally (x direction) to perform the spatial distribution studies. To ascertain the effect of the relative sizes of the laser and analyte spots on the SERS spectra, the diameter of the laser beam at the substrate was changed to between 1 mm and 7 mm by defocusing a microscope objective in the 1459 Uvisir sample compartment. Results and Discussion Figure 5.1 shows the results of a spatial distribution study on an analyte spot of 2 of a 3 mg/mL aqueous solution of AA. The high concentration of AA ensures that all active sites on the surface are saturated. The SERS signal represents a signal-to-background ratio, using the intensity at the 1372 cm"^ SERS peak for the signal and the intensity at 1320 cm ^ as a measure of the background for each spectrum.

PAGE 126

X! O 0) e -P -P £! -P -H O (0 O ago)'"' P ^ (0 m s, e e Q) o 2 -P ^ ^ ^ n >^ Q) X: Q) s E-" c p o o o H in (1) vh •H

PAGE 128

121 The AA spot diameter is between 10.5 and 11 mm. The laser beam diameter at the sample was approximately 0.3 mm and the SERS intensity was recorded every 0.4 mm across the major diameter of the analyte spot. The analyte spatial distribution has three peaks approximately symmetric about the spot center. The two dimensional analog of this cross section is an approximately Gaussian center spot surrounded by a ring. Figure 5.2 shows the results of the laser irradiance study for the 2 mm diameter analyte spot. The laser beam diameter at the sample was altered by defocusing the beam in the sample chamber. The SERS intensity plotted in Figure 5.2a represents the background subtracted peak signal at 1372 cm"\ The SERS intensity plotted in Figure 5.2b corresponds to a signal to background ratio (S/B) the background taken at 1320 cm'\ From Figure 5.2b, it is immediately evident that the maximum S/B occurs where the laser beam and analyte spot diameters are equal. Spatial Distribution Study In order to understand the spatial distribution observed in these experiments, the chromatographic interactions between the analyte, solvent, and substrate during the initial spreading of the solution must be addressed. Analyte "spreading" is a commonly observed phenomenon in thin layer chromatography (TLC) high performance thin layer chromatography (HPTLC) and paper chromatography (PC)

PAGE 129

p n O (M (0 c > (0 CM >-i O *H ft w O 0) fM n3 n H M-l O -P 3 g -p c •H to c o •H ro i C o o 0) CN d O ^ N to tri-H A) fM to 0) d

PAGE 130

123 (AjDj:nqjV) A:^isu9^u| sd3S

PAGE 131

Id • in 0) u c •H Id o p 1 (J o u >i •P •H a c 0) (0 3 Id A I o 4J I H Id g. •H (0 43 in

PAGE 132

125 1-00

PAGE 133

126 Discussions of spot spreading in TLC are usually confined to the development stage of the technique, where interactions between the mobile phase, stationary phase, and analyte determine the final shape of the analyte spot. These interactions can be characterized by a variety of adsorption isotherms (Sewell and Clarke, 1987) (Hamilton and Hamilton, 1987) Most detection schemes for these techniques are for qualitative analysis and involve the determination of values, by looking at migration distances, and resolving power, by looking at spot separations (Hamilton and Hamilton, 1987) Colored spots can be observed directly. Techniques used to visualize noncolored spots include iodine staining, use of UV light, and acid charring. In many instances, analyte distributions after development are taken to be gaussian, but this information is used solely to determine the resolving power of the particular technique (Poole et al., 1986) Even in quantitative determinations, which often involve densitometry (Touchstone and Dobbins, 1978) absorption, or fluorescence (Ripphahn and Halpaap, 1975) the total amount of analyte is the quantity of interest. Densitometers actually scan over the spot and integrate to get the final analyte concentration, but the spatial distribution within the spot is ignored, since it is not significant to the trace analysis of compounds using these chromatographic techniques (Stahl, 1969) Finally, since the majority of spot spreading occurs during development of the chromatogram, and

PAGE 134

127 detection in TLC techniques occurs after the development, little attention has been focused on the analyte spreading during the initial spotting procedure. When a spot of solution is placed on an adsorbent surface, the solvent molecules compete with the sample molecules for active sites on the surface. If the analyte molecules have a stronger affinity for the solvent than the substrate, then the final analyte distribution will be a ring, with the analyte concentrated on the outer edge. If the analyte has a stronger affinity for the substrate than the solvent, the final analyte distribution will resemble a gaussian (Hamilton and Hamilton, 1987) This interaction is the basis for the technique of ring chromatography (Stahl, 1969). The results shown in Figure 5.1 imply that both processes are occurring for the Whatman #5 paper coated with chemically reduced silver and spotted with aqueous AA. Most books on TLC and HPTLC mention that initial spot sizes should be kept small, and guidelines are given as to how this can be accomplished (Poole et al., 1986; Stahl, 1969). These steps typically involve using repeat dosages of small volumes or using volatile solvents for analyte mixtures. This information is mainly used to help the experimenter avoid overlaps of spots applied to different lanes on the chromatographic plate, since the analyte spots expand as they elute up the plate. The smaller the initial spot size, the less likely overlap will occur during elution.

PAGE 135

128 In SERS on solid substrates, chromatographic spreading of analytes during the spotting procedure is the only distribution of analyte concentration that occurs, since no elution is performed after the initial spot application. For analytical studies, where intersample and intrasample reproducibility are important, any factors which affect the analyte concentration under the probe beam play a major role in determining the utility of the technique for routine analysis. Understanding the processes which influence the distribution of the analyte allows the chemist to control the type and extent of spreading which occurs. To this end, much of the information which is pertinent to characterizing solvents in TLC, HPTLC, and PC can be applied to the SERS case. As stated above, it is the chromatographic interaction between the analyte, solvent, and substrate which affects the distribution of the analyte during the initial spotting procedure. Solvent affinity for a given substrate is quantified by the Snyder solvent strength parameter, e", usually referred to as a measure of the relative eluting power of a series of solvents. These values are tabulated for a given type of substrate (e.g. alumina or silica) and referred to as an eluotropic series. These tables are used in TLC, HPTLC, and PC for determining an appropriate solvent or solvent mixture for separating a mixture of analytes on a given support (Hamilton and Hamilton, 1987) Deciding on a solvent for these techniques is not as simple as looking at

PAGE 136

129 the eluotropic series for a given substrate, however. The analyte molecules which are being separated interact with both the mobile and stationary phases during elution. Eluotropic series are not readily available for many compounds, and not available at all for solids. Thus, the experimenter must rely on his chemical intuition concerning the relative importance of such chemical interactions as dipole-dipole, hydrogen bonding, induced dipole, and London forces to decide if a given solvent will resolve all the compounds in a mixture. More often than not, several solvents are tested before an appropriate one is found, using the eluotropic series as a guide. The same will be true for the SERS case. However, no eluotropic data are available for the silver coated solid supports used in SERS, and each different method for preparing an SERS-active substrate generates a surface that is chromatographically different from those prepared by other methods. Thus, if an eluotropic series is prepared for cellulose filter papers coated by direct chemical reduction of silver, it cannot be used for silver coated stochastic quartz posts. Even different preparation methods for depositing silver onto the same solid support, such as direct chemical reduction of silver salts and vacuum vapor deposition of silver onto Millipore 0.22 /xm cellulose nitrate/cellulose acetate membranes, can lead to surfaces with different chromatographic properties. Scanning electron micrographs

PAGE 137

130 reveal that the surface of the membrane coated with reduced silver consists of bare cellulose fibers coated with small (20-100 nm) silver islands (Figure 4.5), whereas the membrane coated with silver by vapor deposition consists of cellulose fibers completely encased in silver (Figure 4.2). Chromatographically, these two surfaces will be quite different. Thus, it is important that, if SERS on solid substrates is to become a viable technique for routine analysis in chemical laboratories, an eluotropic series be obtained for each type of surface and each metal deposition method used, so that the experimenter has a guide to use for choosing an appropriate solvent to minimize analyte spot spreading. Laser Spot Size Study The data in Figure 5.2a shows that the SERS intensity increases rapidly as the laser diameter is decreased for a fixed analyte spot size. However, Van Duyne et al (1986) showed that the SERS intensity is independent of the sampling area. In support of this statement, they analyzed the expression for the SERS intensity for an infinitesimal area element on a surface (Van Duyne et al., 1986; Van Duyne, 1979) : <^^surf K^,Ql^^^{E^e{<^)-'TP^{r,e) rdrdQ (5.1)

PAGE 138

where Ig^^^ is the detected intensity at the Stokes shifted frequency Wg (photoelectron counts s'^) ; N^jj^^ is the adsorbate number density (molecules cm'^) ; fl is the solid angle of the collection optics (sr) ; (da/dn) is the Raman scattering cross section (cm^ molecule"^ sr"^) ; ^ = L(a)) ^L(a)g) ^ is the electromagnetic surface averaged enhancement factor for the excitation (w) and scattering (wg) frequencies; e (co) is the energy of the incident laser photon (J) ; T is the product of the photodetector quantum efficiency, transmittance of the collection optics, and throughput of the dispersion system; and P^^{r,e) is the laser beam irradiance (W cm ), dependent upon polar coordinates r, and e. P^^{r,e) can always be written as a peak irradiance, incorporating the total incident power, Pg, and some distance, w, characteristic of the laser spot size, multiplied by a normalized distribution function. For instance, the commonly used TEMgg Gaussian beam distribution is given by P,U,Q) -[^]exp[^] (5.2) Here, w is the radius from the laser beam center at which the irradiance has fallen to 1/e^ of its peak value. Then, according to Van Duyne (1986), by integrating Equation 5.1 with the laser distribution function in Equation 5.2, it is easily seen that Ig^^^^ is independent of w. The assumptions made are that T and n are roughly equivalent and independent

PAGE 139

132 of r and 6 for macroscopic and microscopic systems. More importantly, Ng^^^ is taken to be constant. To determine if this model predicts the results shown in Figure 5.2, we carried out the integration above and plotted the results that would be predicted with this theory. For the following equations, refers to the radius of the analyte spot on the solid support and w^^ refers to the radius of the laser beam at the sample surface and is analogous to the w from Equation 5.2. Since the purpose of these calculations is to study the effect of the relative sizes of the analyte spot and laser spot, the following portion of Equation 5.1 is taken to be constant. The integration of Equation 5.1 using Equation 5.2 over the interval 0 to some fixed value R then leads to the expression: R = w^, since a signal is only obtained where the laser irradiates the sample, leading to the expression: (5.3) (5.4) The two cases of interest are where < v^, which implies that I. surf (5.5)

PAGE 140

133 and where > ^s' iii'Plyirig that R = vr^, since a signal is only obtained where there is sample, leading to the expression: 1-e t\ (5.6) Figure 5.3 shows the qualitative results predicted by these expressions. Here, Igurf arbitrary units and FN^^j^^ is taken to be unity. The value of is set to 2 mm, corresponding to the 2 mm analyte spot in Figure 5.2a. This plot clearly shows that, as long as the laser beam diameter does not exceed the analyte boundaries, Ig^rf constant, as indicated by Van Duyne (1986) However, for the case where > Wg, Ig^j^f drops off rapidly, in accordance with Equation 5.6. These results clearly differ from our experimental data in Figure 5.2a, especially in the region where < ^s* This might be expected, since our analyte distribution is clearly not constant (Figure 5.1) In order to determine if the model for the SERS intensity expression given by Van Duyne (1979) can be used for our experiments, it is necessary to modify the expression in Equation 5.1 for a variable analyte distribution. This involves altering the expression for Ng^j.^. The analyte distribution shown in Figure 5.1 is difficult to model mathematically. However, if the symmetric shoulders are ignored, a good first approximation to this data would be a Gaussian distribution. This model is the easiest nonconstant distribution which resembles the experimental results and

PAGE 141

in c • 0) in M (0 (fl -P G o ui •H -H -P 3 Q) H 0) S (0 O -H ,in rtJ (0 Q) O C ^ 0) &< N
PAGE 142

135 roo o (/(jDJ|iqjv) A|isu9:^u| sd3S

PAGE 143

136 allows the equation for 1^^^^ to still be solved analytically. The following expression is used for Ng^^^: surf ( 1 \ 271 v| exp 2wt (5.7) Substituting this expression into Equation 5.1, applying Equation 5.3 and integrating using Equation 5.2 leads to an expression for Igurf* suxf 1 exp 2v/v/ (5.8) J/ The two regions of interest lead to two different expressions, For w. < w R = v, as before, and leads to the expression: ^surf exp 2wJ (5.9) For w, > w R = w as before, and leads to the expression: surf 1 exp 2v/ (5.10) Figure 5.4 shows the qualitative results predicted by these expressions. Here, 1^^^^,^ is again in arbitrary units and r is taken to be unity. The value for is set to 0.33 im. This implies that 99.7 percent of the analyte (3 standard deviations or 3 v^) is located within the 2 mm spot. The

PAGE 144

If) 10 c o -rH -P ta O u +J (0 iH ^1 u iH n3 O a) N •H o •H +J (0 u >1 c > c •H (0 +J c o u U) •rH 0) -P >1 t7>H (0 •H p V) •H 4J c 0) o vh 0) p o a u u 0) (/I (0 <4-l o -P (0 p (0 o 0) (0 c o I O -P s •P O c (0 (0 0) o -p (/] P ^(0 •rH ^ +J in c 0) -p c •H w Pi CO (0 •rH T3 (0 W CM •rH ,„<0 IT) c • (0 in c •H X!

PAGE 145

138 i

PAGE 146

139 important result from Figure 5.4 is that, by assuming a nonconstant analyte distribution, Ig^^^^ continues to increase with decreasing laser spot size in the region where v^ < w^, unlike the results shown in Figure 5.3. This increase more closely resembles the data in Figure 5.2a. The differences between Figures 5.4 and 5.2a are mainly due to the poor fit of the Gaussian distribution to the actual analyte spatial profile (Figure 5.1). However, despite this poor fit, the qualitative trends seen in Figure 5.2a, which cannot be accounted for from Equations 5.2a and 5.3, are reproduced. This indicates that, with the proper choice of a distribution function for i^suM' Equation 5.1 is still valid. An exact model for the distribution observed in Figure 5.1 would be complex and would involve numeric solutions to Equation 5.1. Finally, we were unable to determine a suitable model for I^^^^ for the background, since the background intensity changes upon addition of the analyte. This may be partially due to a small amount of aggregation of the silver on the membrane during the chromatographic spreading. Thus, it was not possible to calculate data for signal to background ratios to see if the result of Figure 5.2b could be theoretically predicted. Although the results obtained in Figures 5.2a and 5.2b were reproducible, it is not known if the result from Figure 5.2b is universal. The implications of these data, from an experimental point of view, merit further study to determine if this result is independent of both the support

PAGE 147

140 used and the analyte spatial distribution. If the maximum S/B occurs for = w^, the effects of analyte distribution could be effectively eliminated by matching the laser spot size at the surface to the analyte spot size to achieve the optimum signal-to-background ratio. Conclusions In this study, we have shown that the chromatographic spreading of analyte molecules applied to a SERS active substrate and the relative size of the laser probe beam to the analyte spot play important roles in the intensity of the SERS signal obtained experimentally. The SERS experimenter must take care to note the size of the analyte spot and the analyte concentration distribution for the substrate/ analyte/ solvent system being used. If SERS on solid substrates is to become a routine analytical technique, eluotropic series must be generated for each substrate to be used. These data will help the SERS experimenter choose an appropriate solvent for his work, much in the same fashion as eluotropic series for alumina, silica, and cellulose aid in the choice of appropriate solvents in TLC, HPTLC, and PC. This type of chromatographic information will become even more crucial for SERS using a Raman microprobe. As noted earlier, the diameter of the laser probe in standard SERS experiments is typically an order of magnitude smaller than the analyte spot. In surface-enhanced Raman microprobe spectroscopy, where the

PAGE 148

141 laser beam is focused to a spot between 1 and 5 /xm, the size difference can be as large as three orders of magnitude. For these experiments, detailed knowledge of the analyte spatial distribution on the SERS support is required if any degree of reproducibility is to be obtained. Also, detailed knowledge of the chromatographic properties of the support to be used will be necessary to determine optimum experimental conditions.

PAGE 149

CHAPTER 6 COLLOID FILTRATION Introduction Each of the substrates described in the previous chapters and from previous work can be evaluated according to the guidelines listed in Chapter 2. In general, each substrate exhibits some of the characteristics of the ideal SERS substrate from either a physical or an analytical point of view. Large fiber (Berthod et al 1988) and small fiber (Chapter 4) cellulose filter papers, and smooth and frosted glass slides (Boo et al 1985; Ni and Cotton, 1986) coated with silver by direct chemical reduction are easy to prepare and inexpensive. The same is true for large fiber (Vo-Dinh et al., 1987) and small fiber (Chapter 4) cellulose filter papers and frosted glass slides (Chapter 4) coated by vacuum vapor deposition. However, the metallic microstructures of these surfaces are not spherical (Laserna et al in press) and, therefore, not readily modelled theoretically. Plasma-etched quartz posts (Vo-Dinh et al 1986) coated with vacuum vapor deposited silver provide the most control over the particle size and, more importantly, particle shape of any SERS active substrate that has been produced. These substrates can be 142

PAGE 150

143 used to rigorously test the most important aspects of the electrodynamic theory. However, the method of preparing these surfaces is complex, slow (a 1 cm area takes several hours) and expensive. Silver membrane filters provide a quality controlled surface which requires no preparation, but their low SERS activity (Chapter 4) makes them unsuitable for routine analysis. The most promising substrate which has been developed for analytical SERS use is the deposition of silver onto polystyrene spheres which have been spin-coated onto glass slides (Vo-Dinh et al, 1984) This type of surface is easy to prepare, inexpensive, and has a surface morphology which can be controlled by varying the silver deposition thickness and the polystyrene sphere size and concentration. The major problem with this substrate is that the silver particles which are formed have a nonmetallic core and cannot be directly treated with the current electrodynamic theory. The optimization of the polystyrene sphere size, sphere density, and silver thickness must be performed experimentally. This is a time consuming process which must be done for each metal of interest and each support used. Three approaches to the development of new SERS substrates which adhere to the guidelines in Chapter 2 are to (1) improve the morphology control of the "analytical" preparation methods; (2) improve the speed and costeffectiveness of the "physical" preparation methods; and (3) develop new preparation methods. Examples of method one are

PAGE 151

144 detailed in Chapter 4, where new solid supports were used as better templates for the formation of more uniform silver particles than on previously used supports. Method two is not addressed here, but an example of this type of improvement is outlined in Chapter 7. An example of method three, involving metal colloid solutions, is the focus of this chapter. As mentioned in Chapter 3, metal colloids are ideally suited for both the analytical and physical aspects of a SERS substrate. However, as evident from the results of the three previous chapters, adherence to the guidelines of Chapter 2 is not enough to make a particular substrate useful for routine SERS analysis. Practical experimental problems present a major barrier to reaching this goal. For example, the chromatographic spreading detailed in Chapter 5 presents a major problem for metal coated solid supports in SERS experiments. In the case of metal colloid solutions, the difficulty arises since aggregation of the metal particles occurs upon addition of an analyte molecule to be probed, leading to a dynamic, time-dependent SERS signal (Blatchford et ai 1982). Several possible solutions to this problem have been presented in the literature, the most prominent of which involves the use of flow injection analysis (FIA) Berthod et al (1987) demonstrate the use of FIA to prepare silver hydrosols on line and to obtain precision of SERS signals of five percent. Conditions for the FIA system have been optimized (Laserna et al 1988) and this technique has

PAGE 152

145 been used for quantitative analysis (Laserna et al 1987; Berthod et al 1988). However, although FIA is a practical and useful solution to the problems associated with colloidal SERS, it is not an ideal solution. The idea behind the technique is to allow the flocculation of colloidal particles to occur and to control the mixing rate and mixing time of the analyte and colloid solutions. As the particles aggregate, they initially form strings (Ruckenstein, 1984) These strings have a split plasmon resonance, one at the sphere resonance and one shifting to longer wavelengths as the string length is increased. By controlling the mixing time and flow rates, FIA attempts to probe the mixture when the maximum number of particle strings are "in resonance" with the excitation radiation, typically the 514.5 nm line of an argon ion laser. Thus, the system is still dynamic, and changes in the solution temperature or analyte concentration, the presence of contaminants, and other external factors could affect the optimum mixing time. Another method which has been used to overcome the colloid flocculation problem is the application of colloid particles to chromatographic plates and papers. Tran (1984a) used a syringe to apply colloid and analyte solutions to a variety of filter papers to stabilize the particles against aggregation. Tran (1984) also used an flame photometric nebulizer to spray colloid solutions onto paper chomatograms after development to use SERS as the detection method. This

PAGE 153

146 technique has been adapted to HPLC as well (Sequaris and Koglin, 1987; Koglin, 1988). These methods are closer to an ideal solution to colloid instability, but they are still not ideal. The two deposition methods chosen, syringe application and nebulization, do not allow for a control of the distribution of particles on the surface. In fact, the spraying technique actually relies on a color change in the colloidal solution upon application to the surface, due to the long wavelength plasmon resonance of the colloid strings, in order to observe SERS signals. Once again, the technique is relying on the flocculation of the colloidal particles on the surface to "tune" the surface resonance into the excitation line. In order to obtain SERS spectra and maintain the ability to treat a colloidal surface theoretically so that optimum experimental conditions can be calculated, the particles must be deposited onto the surface in a controlled manner while eliminating any flocculation. The laser wavelength used can then be "tuned" to coincide with the particle resonance, rather than tuning the particle resonance into the laser wavelength by means of flocculation. Using a monodisperse surface allows the most control over both the experimental and theoretical aspects of the substrate and assures that all of the metal particles within the probe beam are in resonance with the excitation wavelength. This chapter describes the development and evaluation of methods for preparing a monodisperse layer of colloid

PAGE 154

147 particles by using filtration techniques. Several surfaces and types of filtration are tested and evaluated as to their ease of use and ability to eliminate the flocculation of the colloid particles on the surface. Experimental Instrumentation A Millipore Sterifil Aseptic System 47 vacuum filtration device was purchased from Millipore. Becton-Dickinson 10 mL disposable syringes and Gelman 25 mm Delrin syringe filter holders were obtained from Fisher. Anotec Anodiscs (0.02 jum pore size) were obtained from Alltech. Polycarbonate tracketched (PCTE) membranes (0.01 iim pore size) were obtained from Poretics. The Raman spectrometer used was the SPEX system described in Chapter 3. Laser power at the sample was 20 mW. Scanning electron micrographs of the PCTE membrane were obtained with a JEOL JSM-35C SEM (resolution 15 nm) while the micrographs for the Anopore membranes were obtained with a Hitachi S-4000 Field Emission SEM (resolution 1.5 nm) The PCTE images were obtained for membranes coated with 10 nm of a gold-palladium mixture to make them conductive. The images of the Anopore membranes required no pre-coating. It should be noted that the actual magnification of the images presented is different than the value printed on the image since the photograph used to prepare the image was either enlarged (for

PAGE 155

148 120 format, 6 x 7 mm negatives) or reduced (for 4x5 inch negatives) for purposes of reproduction. Electronic absorption spectra were obtained using an HP 8450 diode array spectrophotometer and a Perkin-Elmer Lambda-9 spectrophotometer with an integrating sphere. Chemicals and Procedure The silver colloids were produced using the procedure described in Chapter 3. The red gold colloidal solution was prepared as follows: 200 mL of water, filtered through a 0.2 /im filter to remove impurities, is brought to a boil. 0.5 mL of a 4% gold chloride solution is added, followed by 5 mL of a 1% trisodium citrate solution. The mixture is boiled for approximately five minutes. The resulting solution is raspberry red and consists of gold particles with a mean diameter of 15 nm. Vacuum filtration A 47 mm diameter PCTE membrane was placed into the Sterifil vacuum filtration unit. Initial tests used an aspirator to create the vacuum necessary for filtration. Subsequent tests used a mechanical pump to create a stronger vacuum. This experiment was also performed using a 47 mm diameter Anopore membrane. For this membrane, only the aspirator was used for creating the vacuum. For both experiments, 20 mL of the silver colloid solution were filtered. Syringe filtration A 25 mm diameter PCTE membrane was placed into the Gelman syringe filter holder. 2 mL of the

PAGE 156

149 silver hydrosol was filtered through the membrane using a syringe pump. This procedure was repeated using a 25 mm diameter Anopore membrane using both the silver and gold colloid solutions. For this membrane the pressure was applied using both a syringe pump and by hand. In both cases the membrane was removed and allowed to dry. SERS activity for the both the PCTE and Anopore membranes coated with silver were tested using 9-aminoacridine solutions ranging in concentration from 3 0 ppb to 300 ppm applied in volumes between 0.2 and 1 /iL with a Hamilton Microliter #901 syringe. The sample holder described in Chapter 4 was used for these studies. The Anopore membrane coated with gold colloid particles was not tested for SERS activity. Absorption spectra The Anopore membranes are somewhat transparent to visible light. It was possible to obtain absorption spectra of both the blank membrane and a coated membrane (silver colloid) for wavelengths as low as 400 nm by simply placing a blank membrane in the reference beam and a coated membrane in the sample beam of the HP spectrophotometer. To obtain data below 400 nm, the spectra were obtained while the membranes were wet. Filling the pores with a solution with an index of refraction closer to the membrane than air renders the membrane more transparent (Brock, 1983) Additional absorption data were obtained using the Perkin-Elmer instrument by first measuring a blank membrane placed over the front of the integrating sphere and

PAGE 157

150 then placing a coated membrane in the same place. The overall absorption spectra were obtained by background subtraction. Reagents. All chemicals used were analytical reagent grade or equivalent. Demineralized water was used throughout. 9-Aminoacridine hydrochloride monohydrate was purchased from Aldrich and was used without further purification. Results and Discussion Polycarbonate Track-Etched Membranes PCTE membranes are frequently used in biological laboratories for filtration of cells, bacteria, and other micron-sized particles for purposes of SEM imaging. These membranes are essentially flat on a micron scale and have a narrow pore size distribution. The smooth surface provides a high contrast between the sample and the particles on the surface, making it an ideal support for SEM images (Brock, 1983) They are available in pore sizes ranging from 10 nm to several microns, with the intermediate pore sizes used for biological purposes. Figure 6.1 shows the PCTE membrane coated with silver colloid particles using the vacuum filtration technique. The membrane is covered with thin black lines which conform to the pattern of the support mesh in the Sterifil unit (essentially concentric circles) This SEM image confirms that aggregation is prevalent using vacuum filtration on the PCTE membrane. No image of a clean membrane is given because the contrast

PAGE 158

P c o u O • >i< iH < O a u o Q) (M N (U -p (0 p o p c o T) 0) 4J >i P -H > •H -P O (0 w 4-1 W :3 o > CO 0) 0) -P c 5 0) •H ^ 0) u •H

PAGE 159

152

PAGE 160

153 between the colloid particles and the smooth PCTE surface, coupled with the small percent of the membrane surface area covered provides a good view of both the coated and uncoated membrane. The membrane produced using the syringe filtration technique has the same degree of aggregation. SERS activity was tested with 1 /liL of a 30 ppm AA solution using the Raman microprobe The result is also shown in Figure 6.1. Colloid filtration using the PCTE membrane presents a number of problems. Vacuum filtration using an aspirator did not provide enough vacuum for filtration. The mechanical pump created a sufficient vacuum to filter the colloid solution, but four hours were required to filter 20 mL of the hydrosol. These problems are not surprising considering the low pore density of the PCTE membrane, as seen in Figure 6.1. A second practical problem is that the 2 juL aqueous aliquots of AA typically used to probe the SERS activity do not readily absorb into the membrane. Over twenty minutes elapsed between the "spotting" of the analyte solution on the membrane complete evaporation of the solvent. It is likely that the solvent is evaporating rather than being taken into the membrane. Finally, the SERS activity to AA is low due to the excessive aggregation of the colloid particles on the surface. These problems imply that this membrane is not practical as a support for colloid filtration.

PAGE 161

154 Anotec Anopore Alumina Membranes The Anotec Anopore inorganic filter membranes make the syringe filtration method feasible and will be discussed in some detail. ANOPORE inorganic membranes were invented in the laboratories of Alcan International. Anotec Separations was formed by Alcan to further develop this technology. These membranes are composed of aluminum oxide and have a highly ordered, honeycomb structure with sieve-like capillary pores with diameters of 20, 100 and 200 nm (Furneaux et al 1989). Porosity exceeds 50 percent by volume, resulting in very high flow rates when compared to other filters with comparable pore sizes. The water flux rate at 25 C and 10 psi for a 100 nm pore size filter membrane with a 25 mm diameter is 8 mL min ^ cm A comparable flow rate for a polycarbonate track-etched (PCTE) membrane, which is also has sieve-like capillary pores, is less than 0.2 mL min ^ cm^, mainly due to the difference in pore density. For example, a 200 nm pore size PCTE membrane 8 "2 has 3 x 10 pores cm whereas a 2 00 nm pore size Anopore 9 -2 membrane has 3 x 10 pores cm A direct comparison of the Anopore and PCTE membranes is given by Jones et al (1989). Anopore membranes are also used in epif luorescence (Jones and Hoffman, 1989) and in general separations (Hoffman, 1989) Figure 6 2 shows a blank Anopore membrane at two different magnifications (15,000 and 35,000). Figure 6.3

PAGE 162

shows two images of this membrane coated with silver colloid particles using the syringe filtration technique. One image is split to show the results of backscattering detection. The backscattering is obtained using an optional attachment to the Hitachi S-4000 instrument and differentiates between the atomic masses of the objects visualized. The white particles on the left of the image are shown to consist of atoms with a different mass than aluminum or oxygen, the primary constituents of the membrane. These white particles are silver colloid. Since the concentration of the silver relative to the aluminum oxide of the membrane was very low, energy dispersive spectroscopy did not reveal a silver peak. Vacuum filtration was initially used with the Anopore membrane, but the syringe filt^tion technique outlined above was found to be much simpler and less expensive. It should be noted that the silver colloid solution prepared using the technique outlined in Chapter 3 is much too concentrated. Using the vacuum filtration technique and 20 mL of the colloid, the membrane appearance after filtering is solid black. The blank membrane is white. Using 2 mL of this solution with the syringe filtration technique results in a mirror-like silver surface on the membrane. The images in Figure 6.3 result from diluting the silver colloid preparation from Chapter 3 1:39 with water. Figure 6.4 shows a blank membrane and a membrane coated with gold colloid particles. Once again the concentration of the stock gold colloid

PAGE 164

157 E 3. e E S .2 o ^ •:*>j.v,-.:'.V.;'' A'I

PAGE 165

0) N •H (fl Q) U O (0 X3 +J •H > 0) c (0 0) g o o c < c (0 o (U o (0 (M (fl ^! -p o +J c o -o Q) 0) -p rH H <4-l o •H o 0) 0) 0) X! p m o 0) T) •H X o g c •H g iH (0 to p o o u 0) > -H CO c Q) 0) IS -p 0) X! -p •H T3 0) o 0) xi o -p •H > •H -p o (0 w a w w 0) 43 Eh 0) o (0 <)-l •H 3 vo 0) u g.

PAGE 166

159

PAGE 167

0) N •H (0 0) u o a. O X! P •H > 0) c (0 e 0) u o 04 o c < c to o +J c o -d 0) 0) -P -P •H M rH O O •H >1 (/I o D> CP c 0) to :3 V£) •H

PAGE 168

161 3 < C C T3 < U Si £ 3 O

PAGE 169

162 solution was too high. The coated image results from a dilution of the stock colloid solution by 1:79 with water. Figure 6.5 shows the UV-Visible spectra of the silver colloid in solution (a) and on the surface of the membrane (b) Both of the spectra show an absorption maximum at 400 nm, and both have the characteristic yellow color of silver colloid particles. The presence of a shoulder at longer wavelengths in the spectrum of the filtered colloid indicates the presence of particles on the surface with nonspherical geometries (Kerker et al 1982) most likely small aggregates of particles which formed during the filtration process, similar to those observed in the backscattering image in Figure 6.3. No attempt was made to optimize the colloid concentration or the filtration rate for this measurement. As noted in the procedure description above, it was necessary to keep the membrane wet during the collection of the absorption spectra so that light below 400 nm could be collected. Since the absorption maximum of aqueous colloidal silver is between 385 and 400 nm, the data below 400 nm is very important. The spectrum in Figure 6.5 was taken by placing a wet blank membrane in front of the reference beam and placing a wet, freshly prepared membrane in front of the sample beam of the HP spectrometer. As the data is collected over a period of ten seconds, the membranes begin to dry nonuniformly and some aggregation of the deposited colloid could occur. Thus it is uncertain if the "aggregation peaks"

PAGE 170

o o o o o m o 0) < N 4) O a T3 •rH o a. 8 0) 0) c 0) 0) M O O C •H 10 m o (0 +J T3 C O •H -P O 0) •H Q) If) vo 0) U •H

PAGE 171

164 (AjDj:^iqjV) 93UDqjosqv

PAGE 172

. 165 observed in Figure 6.5 are a result of the colloid filtration process or of the nonuniform drying of the wet membrane during the data collection period. Placing the wet membrane between two quartz windows would not ensure that post-filtration aggregation was eliminated. In order to look at this problem more carefully, and to study the effects of the initial colloid concentration on the absorption intensity, a PerkinElmer Lambda-9 spectrophotometer with an integrating sphere was used. This is a single beam instrument, so the background and sample spectra were collected separately and background subtraction was used to obtain the final spectra. Figure 6.6 shows a typical absorption spectrum of both a blank membrane and a membrane coated with silver colloid via syringe filtration. Both membranes were dry. The low intensity of the colloid absorption relative to the background is readily evident. Thus, the results of the background subtraction must be observed with the realization that errors can be significant when the small colloid absorption signal is retrieved off of the large background absorption of the blank membrane. Figure 6.7 shows the result of dilutions of the silver colloid prepared using the technique outlined in Chapter 3. The stock solution, referred to as the 3 0 mL solution, since 30 mL of the silver nitrate solution was used to prepare it, was diluted 1:9 with water. This "3 mL" solution is the highest concentration of silver used with the Perkin-Elmer instrument. The 3 mL solution was diluted 1:1,

PAGE 173

0) u (U 0) rH o o a o o c (0 CQ p c < c (0 o +J c o -p X! 0^ o (0 c o 0) u a) u -p a) p 0) o -p O t3 0) -iH a o <4-l •H <4-l C o -H P H O (0 (0 p (0 Q N H m 0 o •H a) (0 o o (U > o 0) (0 o n -P •rH (0 p u Q) CM (0 • 0) H A •H (0 0) c (0 -H +J (0 CP XI 0) g w 0) c (0 g 0) g 0) o (0 4-1 u d VD 0) •H

PAGE 175

. (0 ^ ^ o 0) <-l N O n as H-i o in c o •H -P iH •H T3 <*-l O (0 u -p o 0) c o 0) o U -H 0) -P a. c o CM 4-) O (0 • ^^ O &I w Q) p 0) c (0 £1 B 0) g 0) o ft o O Q) c o •H rH o rH O -P c ft (0 vh o Ul o -p c O r (0 -H 73 0) +J U 0) u o o n c o o (0 0) u 0) +J rH •H .u T3 •H o p o u -p o 0) ft Ul o •H p rH o Ul n o ,>H (0 4H Ul cr (0 (0 Q c o •H +J rH o Ul 0) CP

PAGE 176

169 (AjDj:^iqjV) 93UDqjosqv

PAGE 177

170 1:4, and 1:9 with water to yield 1.5 mL, 0.6 mL, and 0.3 iiiL solutions, respectively. Figure 6.8 shows the result of a Beer's Law plot for these spectra. These data represent a the background subtracted absorbance values versus concentration. The background was estimated by calculating a line between the minimum near 325 nm and the region for each spectrum to the right of the peak where the background was level (470 nm, 480 nm, 570 nm, and 585 nm for the 0.3 mL, 0.6 mL, 1.5 mL, and 3.0 mL solutions, respectively) These data indicate that Beer's law is obeyed, with deviations occurring at the lowest concentration. The line drawn through the data represent a linear least squares fit (with a correlation coefficient of 0.999) for the three largest concentrations. The deviation at the lowest concentration is not surprising, since the signal is the lowest there and is more likely to be affected by errors from the background subtraction. The SERS activity of an Anopore membrane coated with silver colloid particles using the syringe filtration technique has been tested with 9-aminoacridine (AA) The use of AA also allows the activity of this substrate to be compared to other substrates which have been prepared in this laboratory using different techniques (Laserna et al in press). Figure 6.9 shows a spectrum of a 200 nL volume of a 300 ppb aqueous solution of AA using the 514.5 nm line from an argon ion laser. The laser power at the sample was 2 0 mW. The total amount of AA in the analyte spot, which was

PAGE 178

0) u 0) Si o T3 C 3 O u W >! -P O C 0) Xi •rH -H Q) T3 C 0> O -P O to 1) 10 0) (0 (0 -p (0 Q O (1) o o (0 +J nj TJ 0) x: +j o 0) (0 -P -P >i o 43 T3 (U P (0 3 -P 0^ W 0) P nj <4-i 5 (0 +J (0 -0 0) p <4-l o p o H (0 -p a) 0) (0 a o o 0) CP-P o <-t (0 o XI 0) 0) T3 I/] (0 00 vo 0) •H

PAGE 180

173 approximately 0.75 mm in diameter, is 60 pg, of which approximately 25% is illuminated by the laser beam using the macroscopic sample chamber. Under ideal conditions, little or no SERS signal should be detected, since the 514.5 nm line is outside the absorption profile of the unaggregated silver hydrosol. However, as mentioned above and shown in Figure 6.5, there is some apparent aggregation on the membrane with the silver concentrations used for both the absorption spectrum and the SERS spectrum in Figure 6.9. Careful study of the absorption spectrum indicates that a maximum in the data occurs near 515 nm. The colloid aggregates on the surface which have long axis absorption resonances in this region are likely the source of the SERS signal observed. Little SERS intensity has been observed at 514 nm for membranes coated with more dilute silver solutions, where aggregation is less apparent (Figure 6.3). The data from the Perkin-Elmer instrument are not of sufficient quality to determine if the longer wavelength features disappear as the colloid concentration is decreased. The ability to prepare a surface of noninteracting spherical metal colloid particles allows for a direct comparison of experimental SERS data with theoretical calculations. Not only will this allow for theoretical optimization of parameters such as the particle size and metal to use for optimum SERS enhancement at a given wavelength, it may help to add new information to the "great colloid debate"

PAGE 181

174 in SERS which was mentioned in Chapter 2. When Creighton et al (1979) first studied SERS on silver colloids, they looked at the frequency dependence of the enhancement as a function of the amount of aggregation of particles in solution. They correlated this with the absorption spectra of the colloid solution as a function of time. As the particles aggregate, the single maximum splits into two maxima. One corresponds to an excitation perpendicular to the interparticle axis and occurs at the same wavelength as the isolated colloid particles. The other corresponds to an excitation along the interparticle axis and shifts to longer wavelengths as more particles are added to the string. Creighton observed that the SERS intensity as a function of excitation wavelength paralleled the long-axis absorption profile, but that no SERS could be observed for the short wavelength resonance (Blatchford et al 1982). This information is shown graphically in Figure 6.10. What has been inferred from these results is that aggregation is a necessary component for SERS activity in colloids. Initial theoretical treatments of this problem did not agree with the experimental data, predicting that, although the enhancement is larger and red-shifted for the long-axis excitation (modelling the string as a spheroid) and virtually nonexistent for the short-axis excitation, that strong SERS signals should still be observed for nonaggregated spheres. The difficulty in resolving this argument is in how

PAGE 182

c o 0) -P (0 o o 0) c V4 XI H 0) O hi w o C < n S nJ (0 c -I ^ o 3 ^ 3 Q. 6 O O U O
PAGE 183

176

PAGE 184

^0) >• • ffl 42 0) c o jd -H •H M o o O (0 c •H • in 73 (0 c •H O CP (0 o 0) c -0 IT) >i-H a o I u o c J3 0^ c O c o •H +J •H T3 T3 (0 o O 0) -P C7> C •H (0 (0 o u o •H -P (0 rO iH 0) -d c to u 0) o u 0) o 0) £1 -p W (0 0) u Si (0 ^ c (0 ft q c o -rH -P o c •H -p X o -rH -P (0 -p •H 0) o rH < 1* 5.-P O CO O H VO 0) 1.

PAGE 185

178 Oix ADuapijja 6uijajiP3s "ADuapijja uoijdjosqe

PAGE 186

179 to add an analyte to a metal colloid to obtain a SERS spectrum without any particle aggregation. Garrell at al (1983) attempted to circumvent this problem by using ultrafiltration to remove all the aggregated particles from a mixture of silver colloids and pyridine. They observe SERS for the unaggregated colloid particles which passed through the ultrafilter. The use of isolated colloid particles filtered onto an Anopore membrane is a much more practical and elegant solution to the problem of eliminating the aggregation problem and should aid in resolving this debate. Another consideration is that, as mentioned in Chapter 2, the lack of a strong SERS enhancement for isolated silver, gold, and copper spheres is predicted by the Zeman and Schatz (1987) electrodynamic theory. It is therefore possible that both sides of the argument are correct. Conclusions The use of colloid filtration in the preparation of SERSactive substrates can yield uniform, monodisperse layers of colloidal particles. Cellulose and other depth capture membranes are not appropriate supports for this technique due to their inefficient retention and the difficulties of probing the metal particles within the depth of the membrane. The PCTE surface capture membrane is also an inappropriate support for syringe filtration since the flow rates are too slow to permit smooth, uniform filtration in a reasonable amount of

PAGE 187

180 time and due to poor absorption of analyte aliquots. A number of characteristics of the ANOPORE membranes make them an ideal support for the preparation of SERS-active substrates by colloid filtration. The ability to capture virtually 100 percent of all particles greater than the rated pore size makes them highly efficient for isolating colloidal particles. The surface capture property of the ANOPORE membrane allows all the colloid particles trapped to be illuminated by the probe beam. The fast flow rate allows for the rapid preparation of SERS substrates. The chemical inertness of the membranes to many organic solvents also allows nonaqueous colloids to be used. The membranes are transparent to visible light with wavelengths longer than 400 nm. The thickness of the membranes is small enough, typically 60 /xm, that the electronic absorption of the colloids on the membrane can be measured directly by standard UV/VIS instruments. The membranes can be made transparent to wavelengths shorter than 400 nm by wetting them with water. The capillary nature of the membranes helps to minimize the radial spreading of analyte solutions dispensed by micro syringes which was observed in Chapter 4 and discussed in detail in Chapter 5. Finally, the ability to reproducibly prepare a SERS-active substrate in under five minutes with only a few dollars in equipment makes this a viable technique for routine analytical use.

PAGE 188

CHAPTER 7 CONCLUSIONS AND FUTURE WORK Surface-enhanced Raman spectroscopy is still a relatively new technique, especially in the regime of analytical chemistry. However, the scientist wanting to use the SERS technique has a wide variety of surfaces from which to choose. This work has looked at the utility of some of these types of surfaces and has attempted to improve upon a number of them. The colloid study in Chapter 3 has shown the analytical power of the SERS technique by exploiting not only the sensitivity of the method due to the enhancement, but the selectivity of Raman spectroscopy in general. Colloidal solutions are simple to prepare, stable over long periods of time, and provide an effective SERS surface. This method has been used in the automated detection of HPLC eluents via flow injection analysis, and thus has been shown to be somewhat unique among SERS surfaces in that it is mobile. This alone makes colloidal SERS attractive for certain applications. The results of Chapter 4 have shown that it is possible to improve the quality of a SERS substrate prepared by what are now considered the "standard" methods of vacuum vapor deposition and metal salt reduction by the appropriate choice of solid supports. More work in this area should reveal if 181

PAGE 189

182 certain compositions of membranes (PTFE, PVDF, cellulose, etc.) help to improve the uniformity and control of the formation of metal features on these supports. These methods, especially the reduction method, are very simple and is should be possible to automate them for routine or on-line analysis. The chromatographic spreading problems discussed in Chapter 5 are partly a result of the sample application method. Spraying an analyte onto a membrane could provide a more uniform distribution on the surface, and other, more appropriate deposition techniques may be found. An alternative is to conduct a rigorous series of tests to characterize each substrate chromatographically and to prepare an eluotropic series for each preparation method and substrate to be used in any routine analysis. This seems to be an extremely difficult and time consuming task, but is essentially the same procedure that was necessary to be able to use alumina, silica, and cellulose in TLC and HPTLC. Once this series is prepared for a given substrate, the ideal solvent or solvent mixture can be determined for a particular analyte or class of analytes much in the same way as is done in chromatography. Both colloidal solutions and the silver coated solid supports prepared via reduction and evaporation, with some exceptions, are adequate surfaces for "quick and dirty" SERS experiments, since they are simple to prepare and inexpensive, but lack control over the size and shape of the metal

PAGE 190

183 particles prepared. The exceptions mentioned above involve the quartz post and polystyrene sphere surfaces mentioned earlier. These surfaces allow for fine control over the particle size and density, but by sacrificing speed (quartz posts) or the ability to mimic theory (polystyrene spheres) The ability to prepare surfaces which closely fit the theoretical models is crucial to the ability to change metals in a surface preparation method with a minimum of experimental optimization. Colloids have controlled size and shape, but as shown in Chapter 3, the flocculation problems eliminate this advantage. The colloid filtration method discussed in Chapter 6 presents a solution to the aggregation problem and is a viable alternative to these methods for the preparation of SERS surfaces. Colloid filtration produces surfaces which closely resemble the theoretical models, while at the same time being a simple and inexpensive method. Since colloid solutions can be prepared for a variety of metals with some control over the particle size, the colloid filtration method solves all of the problems of finding an ideal SERS substrate, as outlined in Chapter 2, with the exception of control over the particle shape. A nice analogy to ideal SERS substrate, the importance of the control over particle shape, and where the colloid filtration method fits in, is fixed wavelength lasers versus tunable dye lasers. The use of metal colloids fixes the geometry which, as shown by the theoretical data in Chapter 2,

PAGE 191

184 allows only a small amount of tunability in choosing the optimum enhancement wavelength by altering the particle size. In order to make use of another wavelength, a different metal must be chosen. This is similar to having a number of fixed frequency lasers giving a limited number of lines (e.g. Ar* for 488 and 514 nm, HeCd for 325 nm, etc.). The ultimate flexibility in SERS substrates would allow for control of the particle shape as well. A method providing this type of control would solve the one problem not addressed by the filtered colloid substrate. The use of spheroids allows each metal chosen to be brought into resonance with a wide range of excitation wavelengths. Thus, each metal would be analogous to a laser dye, with its highest frequency limited by the sphere shape and its lowest frequency limited by damping effects and the physical stability of highly eccentric spheroids. To change to another frequency range, another metal would be chosen, but the number of metals required to span the electromagnetic spectrum from the UV to the near IR would be much smaller than for colloidal surfaces. The quartz post substrate is such a technique, but the time involved in preparing a surface renders it useless as an analytical method. What is required is a way to prepare a template similar to the one presented by the quartz post surface, but faster and easier. It is the quest for just such a surface, and the desire to improve other techniques, which motivates the future of the SERS research presented here.

PAGE 192

185 The following paragraphs discuss some of the possible avenues for continued research into finding new substrate preparation methods and improving on old ones. Many of these ideas have been submitted as a part of a molecular research proposal to the NIH by the Winefordner research group and to 3M as individual projects. They are mentioned here to show the diversity of possibilities and to perhaps inspire other ideas as well. Centrifuge filtration The colloid filtration technique from Chapter 6 represents one of the simplest, fastest, and least expensive preparation methods for SERS substrates which have been developed to date. It is possible, however, to improve dramatically on the amount of time required to prepare each surface by extending the method to use a centrifuge. Centrifuge filtration holders are available which allow a membrane to be inserted into them and then removed later, a necessity for using the Anopore membrane and for accessing the membrane for use in SERS experiments. The advantage lies in the ability to prepare as many as 3 6 surfaces simultaneously. Also, it may be possible to automate some of the process by using a robot arm to add colloid solution to the reservoir of the holders, place them into the centrifuge, and remove them when the filtration is finished. This technique could also help to improve the reproducibility of the filtration method. Empore membranes 3M has developed a new type of chromatographic extraction membrane which uses teflon to

PAGE 193

186 entrap a variety of particles (e.g. ion exchange resins, chelating agents, metal particles) into a thin flexible membrane, called Empore. Bio-rad, a large biochemical supply company, markets several of these membranes as an alternative to column chromatography. They are highly efficient at capturing whatever the particular membrane is designed for, usually within the first few percent of the membrane thickness. Thus, these membranes act much like a surface capture membrane. Preliminary studies in this lab using these surfaces to capture colloid particles is promising. The colloid is simply spotted onto the membrane. The diffusion of particles occurs mainly on the surface. Also, for dilute colloid concentrations, the particles appear to be relatively monodisperse (using UV/Visible absorption) This represents an improvement on the colloid spotting/ spraying work of Tran (1984, 1984a). Ultrafiltration The results of the theoretical calculations often show that the ideal particle size for SERS with a sphere geometry is less than 20 nm, the smallest size available for the Anopore membranes. Although PCTE membranes are available with pore sizes as low as 10 nm, the results shown in Chapter 6 indicate that they are not suitable for the colloid filtration method. To trap such small particles, ultrafiltration can be employed. Ultrafilters can capture particles with nominal molecular weights as small as 500. The comparable spherical particle size is less than 1 nm. Also,

PAGE 194

187 the use of a stirred cell ultrafiltration unit may help to improve on the uniformity and monodispersity of the colloid particles. The stirring action of such a unit is designed to prevent "concentration polarization" and flocculation at the surface, normally to prevent clogging of the membrane. This feature should help alleviate any colloid flocculation occurring at the surface. The smooth surface of an ultrafilter should also provide advantages in the imaging of particles on the surface, much like the PCTE membranes (Brock, 1983) Finally, Anotec is in the process of releasing an alumina ultrafilter analog to the Anopore membrane which could prove to be the ideal ultrafilter for this work. Spin coating The surfaces prepared using polystyrene spheres described in Chapter 4 involved the use of a spin coating device to place the spheres onto either glass slides or hardened filter papers. This coating technique resulted in highly uniform, submonolayer coverages of spheres on the surface of the membrane. Monodispersity was maintained for concentrations at least as high as a tenth of a monolayer. This is much higher than the concentrations of colloid particles which can be deposited by syringe filtration. It should be possible to prepare surfaces with higher densities of colloid particles using this method. One of the problems to be solved experimentally with regards to this technique is finding an ideal support. Aqueous colloid solutions are much less viscous than the polystyrene sphere solutions typically

PAGE 195

188 used, so absorption into the membrane is a major concern. Preliminary tests using Anopore membranes, PCTE membranes, and Empore membranes indicate that the capillary action of the Anopore membrane hinders its utility for this method, the lack of interaction between the PCTE membrane for water prevents any colloid deposition on this surface, and that the empore membrane shows promise as a substrate for this method. Preliminary UV/Visible studies of the deposition of a gold colloid in MEK onto an Empore membrane shows that the absorption maximum of the gold in solution (near 540 nm) is retained for the particles spin-coated onto the membrane surface, detected by reflection on a Perkin-Elmer Lambda 9 spectrometer with an integrating sphere. Experimental conditions to be optimized include not only the substrate to be used, but the spin rate, the deposition method, and the colloid solvent. This method also has the potential of automation. Etched polymer membranes Perhaps the most promising substrate on the horizon is the etched polymer membrane. The etching process is the same as that used for the stochastic quartz posts mentioned in Chapter 4. The membrane is coated with a metal (e.g. silver, cadmium) and then heated to cause the metal to bead into islands. The membrane is then etched away using a plasma etching technique. The stochastic posts are formed due to the differential in etching rates between the metal mask and the bare surface, often a factor of 80 or

PAGE 196

189 more. This etching process was developed to prepare antireflective coatings. Although many of the details are proprietary and cannot be disclosed here, images of the surface are remarkably similar to those of the quartz posts, the surface which has provided the closest approximation to the ideal SERS substrate of any mentioned in the literature. The polymer etching process removes the one barrier of the quartz post surfaces to routine use. The method has been automated and scaled up by 3M to prepare 90 ft^ min'^ of etched membrane, orders of magnitude higher than for the quartz surfaces. Another advantage of this polymer surface is its anti-reflective properties. The membranes are typically 92-96% transmittant to visible light, whereas a comparable number for the quartz surfaces is 10%. Scatter from the surface should be minimal for the polymer membrane, and the possibility of dual pass geometries exists.

PAGE 197

REFERENCE LIST Abe, H., K.-P. Charle, B. Tesche, and W. Schulze, 1982, Chem. Phys. 68, 137. Abe, H. K. Manzel, W. Schulze, M. Moskovits, and D. P. DiLella, 1981, J. Chem. Phys. 74, 792. Adams, A., R. W. Rendell, W. P. West, H. P. Broida, P. K. Hansma, and H. Metiu, 1980, Phys. Rev. B 21, 5565. Adar, F., 1988, Microchem. J. 38, 50. Adrian, F. J., 1982, J. Chem. Phys. 77, 5302. Ahem, A., and R. L. Garrell, 1987, Anal. Chem. 59, 2816. Alak, A. M. and T. Vo-Dinh, 1987, Anal. Chem. 59, 2149. Alak, A. M. and T. Vo-Dinh, 1989, Anal. Chem. 61, 656. Albrecht, M. G. and J. A. Creighton, 1977, J. Am. Chem. Soc. 99, 5215. Aroca, R. and F. Martin, 1985, J. Raman Spec. 16, 156. Asher, S. A., 1984, Anal. Chem. 56, 720. Asher, S. A., 1988, Ann. Rev. Phys. Chem. 39, 537. Asher, S. A., and C. R. Johnson, 1984, Science 225, 311. Asher, S. A., C. R. Johnson, and J. Murtaugh, 1983, Rev. Sci. Instrum. 54, 1657. Banwell, C. N. 1972, Fundamentals of Molecular Spectroscopy, 2^ ed. (McGraw-Hill, UK), Chap. 4. Baranska, H., A. Labudzinska, and J. Terpinski, 1987, Laser Raman Spectrometry: Analytical Applications (Ellis Horwood, Chichester), p. 136. Barber, P. W. R. K. Chang, and H. Massoudi, 1983, Phys. Rev. Lett. 50, 997. 190

PAGE 198

191 Barber, P. W. R. K. Chang, and H. Massoudi, 1983a, Phys. Rev. B 27, 7251. Bass, J., J. S. Dugdale, C. L. Foiles, and A. Meyers; 1985, Landolt-Bornstein, Vol. 15: Metals, Electronic Transport Phenomena, edited by K.-H. Hellwege and J. L. Olsen (Springer-Verlag, New York) Benner, R. E. K. U. Von Raben, K. C. Lee, J. F. Owen, R. K. Chang, and B. L. Laube, 1983, Chem. Phys. Lett. 96, 65. Bernstein, H. J., 1979, Phil. Trans. R. Soc. Lond A 293, 287. Berthod, A., J. J. Laserna, and J. D. Winefordner, 1987, Appl. Spec. 41, 1137. Berthod, A., J. J. Laserna, and J. D. Winefordner, 1988, J. Pharm. Biomed. Anal. 6, 599. Bevill, R. F. in CRC Handbook Series in Zoonoses; Volume 1, Section D: Anitbiotics, Sulfonamides, and Public Health, 1984, edited by J. H. Steele, G. W. Beran, T. H. Jukes, H. L. DuPont, and L. M. Crawford (CRC Press, Inc., Boca Raton) Birke, R. L. and J. R. Lombardi, 1982, in Advances in Laser Spectroscopy, Vol. 1, edited by B. A. Garetz and J. R. Lombardi (Heyden & Sons, London), Ch. 7. Blatchford, C. G. J. R. Campbell, and J. A. Creighton, 1982, Surf. Sci. 120, 435. Boo, D. W., W. S. Oh, M. S. Kim, K. Kim, and H. Lee, 1985, Chem. Phys. Lett. 120, 301. Brandt, E. S., 1988, Appl. Spec. 42, 882. Bratton, A. C. and E. K. Marshall, Jr., 1939, J. Biol. Chem. 128, 537. Brock, T. D., 1983, Membrane Filtration: A User's Guide and Reference Manual; (Science Tech, Inc., Madison). Campion, A., 1983, in Vibrational Spectroscopy of Molecules on Surfaces, edited by J. T. Yates, Jr. and T. E. Madey (Plenum Press, New York), Chap. 8. Campion, A., 1984, Comments Solid State Phys. 11, 107. Chang, R. K. 1987, Ber. Bunsenges. Phys. Chem. 91, 296.

PAGE 199

192 Chang, R. K. and B. L. Laube, 1984, CRC Crit. Rev. Solid State Mat. Sci. 12, 1. Chambers, J. A., and R. P. Buck, 1984, J. Raman. Spec. 15, 350. Chou, Y. C. N. T. Liang, and W. S. Tse, 1986, J. Raman Spec. 17, 481. Cline, M. P., P. W. Barber, and R. K. Chang, 1986, J. Opt. Soc. Am. B 3, 15. Cooney, R. P., T. P. Mernagh, M. R. Mahoney, and J. A. Spink, 1983, J. Phys. Chem. 87, 5314. Cotton^ F. A., 1971, Chemical Applications of Group Theory, 2 edition (John Wiley & Sons, New York) Creighton, J. A., M. J. Alvarez, D. A. Weltz, J. Garoff, and M. W. Kim, 1983, J. Phys. Chem. 87, 4793. Creighton, J. A., C. G. Blatchford, and M. G. Albrecht, 1979, J. Chem. Soc, Faraday, Trans. 75, 790. Das, P., and H. Metiu, 1985, J. Phys. Chem. 89, 4680. Demtroder, W. 1982, Laser Spectroscopy: Basic Concepts and Instrumentation (Springer-Verlag, New York), Chap. 14. Dhamelincourt, P. F. Wallart, M. Leclercq, A. T. N'Guyen, and D. O. Landon, 1979, Anal. Chem. 51, 414A. Dollish, F. R. W. G. Fateley, and F. F. Bentley, 1974, Characteristic Raman Frequencies of Orgainc Compounds (John Wiley & Sons, New York), p. 53. Dudik, J. M. C. R. Johnson, and S. A. Asher, 1985, J. Phys. Chem. 89, 3805. Enlow, P. D., M. Buncick, R. J. Warmack, and T. Vo-Dinh, 1986, Anal. Chem. 58, 1119. Fleschmann, M. P. J. Hendra, and A. J. McQuillan, 1974, Chem. Phys. Lett. 26, 123. Furneaux, R. C. W. R. Rigby, and A. P. Davidson, 1989, Nature 337, 147. Furtak, T. E. and J. Reyes, 1980, Surf. Sci. 93, 351. Furtak, T. E., and D. Roy, 1985, Surf. Sci. 158, 126. I

PAGE 200

193 • f Gao, X., C. Wan, T He, J. Li, H. Xin, and F. Liu, 1984, Chem. Phys. Lett. 112, 465. Garrell, R. L. 1989, Anal. Chem. 61, 401A. Garrell, R. L. and R. H. Schultz, 1985, J. Coll. Interface Sci. 105, 483. Garrell, R. L. K. D. Shaw, and S. Krimm, 1983, Surf. Sci. 124, 613. Gerrard, D. L. and H. J. Bowley, 1988, Anal. Chem. 60, 368R. Gerrard, D. L. and H. J. Bowley, 1989, in Chemical Analysis of Polycyclic Aromatic Compounds, edited by T. Vo-Dinh (John Wiley & Sons, New York), Chap. 14. Gersten, J., and A. Nitzan, 1980, J. Chem. Phys. 73, 3023. Gersten, J. I., and A. Nitzan, 1982, in Surf ace Enhanced Raman Scattering, edited by R. K. Chang and T. E. Furtak (Plenum, New York), 89. Gersten, J. I., and A. Nitzan, 1985, Surf. Sci. 158, 165. Glass, A. M. P. F. Liao, J. G. Bergmann, and D. H. Olson, 1980, Opt. Lett. 5, 368. Hagemann, H.-J., W. Gudat, and C. Kunz, 1974, DESY Report SR74/7, 1. Hamilton, R. and S. Hamilton, 1987, Thin Layer Chromatography: Analytical Chemistry by Open Learning (John Wiley & Sons, New York), p. 47. Heritage, J. P., H. G. Bergman, A. Pinczuk, and J. M. Worlock, 1979, Chem. Phys. Lett. 67, 229. Hildebrandt, P., and M. Stockburger, 1984, J. Phys. Chem. 88, 5935. Hoffman, J., 1989, Amer. Lab., August, 70, 189. Horwitz, W., 1981a, J. Assoc. Off. Anal. Chem. 64(1), 104. Horwitz, W., 1981b, J. Assoc. Off. Anal. Chem. 64(4), 814. Howard, M. W. and R. P. Cooney, 1982, Chem. Phys. Lett. 87, 299.

PAGE 201

194 Huang, X.-Y., K.-T. Lee, and T. F. George, 1986, J. Chem. Phys. 85, 567. Hudson, B., 1986, Spectroscopy 1, 22. i Hudson, B., P. B. Kelly, L. D. Zioegler, R. A. Desiderio, D. P. Gerrity, W. Hess, and R. Bates, 1986, in Advances in Laser Spectroscopy, Vol. 3, edited by B. A. Caret z and J. R. Lombardi (John Wiley & Sons, New York), Chap. 1. Hughes, M. N. 1985, The Inorganic Chemistry of Biological Processes, 2^ edition (John Wiley & Sons, New York) Huong, P. v., 1986, J. Pharm. Biomed. Anal. 4, 811. Ingle, J. D. Jr., and S. R. Crouch, 1988, Spectrochemical Analysis (Prentice-Hall, New Jersey), Chap. 16. Jeanmarie, D. J., and R. P. Van Duyne, 1977, J. Electroanal. Chem. 84, 1. Jennings, C, R. Aroca, A. Hor, and R. O. Loutfy, 1984, Anal. Chem. 56, 2033. Jha, S., 1982, Curr. Sci. 51, 1049. Jha, S. S., 1985, Surf. Sci. 158, 190. Johnson, P. B., and R. W. Christy, 1972, Phys. Rev. B. 6, 4370. Johnson, P. B., and R. W. Christy, 1974, Phys. Rev. B. 9, 5056. Johnson, K. L. D. T. Jeter, and R. C. Clairborne, 1975, J. Pharm. Sci. 64, 1657. Jones, C. M., T. A. Nairn, M. Ludwig, J. Murtaugh, P. L. Flaugh, J. M. Dudik, C. R. Johnson, and S. A. Asher, 1985, Trends in Anal. Chem. 4, 75. Jones, H. and J. Hoffman, 1989, Biomed. Products, December, 40. Jones, S. E., S. Ditner, C. Freeman, C. J. Whitaker, and M. A. Lock, 1989, Appl. Environ. Microbiol. 55, 529. Kerker, M. O. Siiman, and D.-S. Wang, 1984, J. Phys. Chem. 88, 3168.

PAGE 202

195 Kerker, M. D.-S. Wang, and H. Chew, 1980, Appl. Opt. 19, 4159. Kerker, M. D.-S. Wang, 0. Siiman, and L. A. Bumm, 1982, in Surface-Enhanced Raman Scattering edited by R. K. Chang, and T. E. Furtak (Plenum, New York), p. 109. Kim, S. K., T. M. Cotton, R. A. Uphaus, and D. Mobius, 1989, J. Raman Spec. 93, 3713. Kim, S. K., T. H. Joo, S. W. Suh, and M. S. Kim, 1986, J. Raman Spec. 17, 381. Koglin, E., 1988, J. Mol. Struct. 173, 369. Koglin, E. and J.-M. Sequaris, 1986, J. Mol. Struct. 141, 405. Kraus, W. A., and G. C. Schatz, 1983b, J. Chem. Phys. 79, 6130. Kraus, W. A., and G. C. Schatz, 1983, Chem. Phys. Lett. 99, 353. Laor, U. and G. C. Schatz, 1981, Chem. Phys. Lett. 82, 566. Laor, U., and G. C. Schatz, 1982, J. Chem. Phys. 76, 2888. Laserna, J. J., A. Berthod, and J. D. Winefordner, 1987, Talanta 34, 745. Laserna, J. J., A. Berthod, and J. D. Winefordner, 1988, Microchem. J. 38, 125. Laserna, J. J., A. D. Campligia, and J. D. Winefordner, 1988, Anal. Chim. Acta 208, 21. Laserna, J. J., A. D. Campliglia, and J. D. Winefordner, 1989, Anal. Chem. 61, 1697. Laserna, J. J., W. S. Sutherland, and J. D. Winefordner, Anal. Chim. Acta, in Press. Laserna, J. J., E. L. Torres, and J. D. Winefordner, 1987, Anal. Chim. Acta 200, 469. Lee, P. C, and D. Meisel, 1982, J. Phys. Chem. 86, 3391. Liao, P. F., 1982, in Surface Enhanced Raman Scattering, edited by R. K. Chang and T. E. Furtak (Plenum, New York) p. 379.

PAGE 203

196 Lin, L., C. K. Mann, and T. J. Vickers, 1987, Appl. Spec. 41, 422. Lippitsch, M. E., 1984, Phys. Rev. B 29, 3101. Mathies, R. 1979, in Chemical and Biochemical Applications of Lasers, Vol. 4, edited by C. B. Moore (Academic Press, New York), Chap. 4. Meier, M. A. Wokaun, and T. Vo-Dinh, 1985, J. Phys. Chem. 89, 1843. Moody, R. L. T. Vo-Dinh, and W. H. Fletcher, 1987, Appl. Spec. 41, 966. Moskovits, M. 1985, Rev. Mod. Phys. 57, 783. Murray, C. A., 1986, in Advances in Laser Spectroscopy, Vol. 3, edited by B. A. Garetz and J. R. Lombardi (John Wiley and Sons, New York), Chap. 3. Nabiev, I. R. R. G. Efremov, and G. D. Chumanov, 1988, Sov. Phys. Usp. 31, 241. Ni, F., and T. M. Cotton, 1986, Anal. Chem. 58, 3159. Nitzan, A., and L. E. Brus, 1981a, J. Chem. Phys. 74, 5321. Nitzan, A., and L. E. Brus, 1981b, J. Chem. Phys. 75, 2205. Olsen, C. W. and R. I. Masel, 1988, J. Vac. Sci. Technol. A 6, 845. Pandey, P. K. K. and G. C. Schatz 1982, Chem. Phys. Lett. 88, 193. Pandey, P. K. K. and G. C. Schatz, 1984, J. Chem. Phys. 80, 2959. Parker, W. L. R. M. Hexter, and A. R. Siedle, 1984, Chem. Phys. Lett. 107, 96. Pettenkofer, C. J. Eickmans, U. Erturk, and A. Otto, 1985, Surf. Sci. 151, 9. Pettinger, B., 1987, Ber. Bunsenges. Phys. Chem. 91, 308. Pineda, A. C, and D. Ronis, 1985, J. Chem. Phys. 83, 5330. Pockrand, I., 1982a, Chem. Phys. Lett. 92, 509.

PAGE 204

i 197 Pockrand, I., 1982b, Chem. Phys. Lett. 92, 514. Poole, C. F., S. Khatib, and T. A. Dean, 1986, Chrom. Forum, MayJune, 27. Reimer, H., and F. Fischer, 1984, Phys. Stat. Sol. B 124, 61. Ripphahn, J., and H. Halpaap, 1975, J. Chrom. 112, 81. Rothenhausler, B. and W. Knoll, 1988, J. Opt. Soc. Am. B 5, 1401. Ruckenstein, E., 1984, J. Coll. Interface Sci. 99, 270. Rzepka, E. L. Taurel, and S. Lefrant, 1981, Surf. Sci. 106, 345. Sato, S, S. Higuchi, and S. Tanaka, 1980, Anal. Chim. Acta 120, 209. Schulze, W. B. Breithaupt, K.-P. Charle, and U. Kloss, 1984, Ber. Bunsenges. Phys. Chem. 88, 3 08. Seki, H. 1986, J. Electron Spec. Rel. Phenom. 39, 289. Sequaris, J.-M. L. and E. Koglin, 1987, Anal. Chem. 59, 527. Sewell, P. A., and B. Clarke, 1987, Chromatographic Separations : Analytical Chemistry by Open Learning (John Wiley & Sons, New York) pp. 80-85. Sigel, C. W. and J. L. Woolley, Jr., in Densitometry in Thin Layer Chromatography, edited by J. C. Touchstone and J. Sherma (John Wiley & Sons, New York), p. 677. Skoog, D. A., 1985, Principles of Instrumental Analysis^ 3^ edition (Saunders College Publishing, New York) Chap. 12. Sobocinski, R. L. and J. E. Pemberton, 1988, Langmuir 4, 836. Sophian, L. H. D. L. Piper, and G. H. Schneller, 1952, The Sulfapyrimidines: Sulfadiazine, Sulfamerazine, Sulfamethazine (A. Colish, New York) Spiro, T. G., 1974, in Chemical and Biochemical Applications of Lasers, Vol. 1, edited by C. B. Moore (Academic Press) Chap. 2 Spiro, T. G., and P. Stein, 1977, Ann. Rev. Phys. Chem. 28, 501.

PAGE 205

198 Stahl, E., 1979, in Thin Layer Chromatography, 2"*^ edition, edited by E. Stahl (Springer-Verlag, New York), p. 38. Suh, J. S., D. P. DiLella, and M. Moskovits, 1983, J. Phys. Chem. 87, 1540. Takenaka, T., and K. Eda, 1985, J. Coll. Interface Sci. 105, 342. Taylor, G. T. S. K. Sharma, and K. Mohanan, 1990, Appl. Spec. 44, 635. Theye, M. L. and G. Devant, 1969, Thin Solid Films 4, 205. Torres, E. L. and J. D. Winefordner, 1987, Anal. Chem. 59, 1626. Touchstone, J. C. and M. F. Dobbins, 1978, Practice of Thin Layer Chromatography (John Wiley & Sons, New York) Chap. 2. Tran, C. D. 1984a, J. Chromat. 292, 432. Tran, C. D., 1984b, Anal. Chem. 56, 824. Van de Hulst, H. C. 1981, Light Scattering by Small Particles (Dover Publications, New York) Van Duyne, R. P., 1979, in Chemical and Biochemical Applications of Lasers, Vol. 4, edited by C. B. Moore (Academic Press, New York), Chap. 5. Van Duyne, R. P., K. L. Haller, and R. I. Altkorn, 1986, Chem. Phys. Lett. 126, 190. Vo-Dinh, T., 1989, in Chemical Analysis of Polycyclic Aromatic Compounds, edited by T. Vo-Dinh (John Wiley & Sons, New York) Chap. 15. Vo-Dinh, T., A. Alak, and R. L. Moody, 1988, Spectrochim. Acta 43B, 605. Vo-Dinh, T., M. Y. K. Hiromoto, G. M. Begun, and R. L. Moody, 1984, Anal. Chem. 56, 1667. Vo-Dinh, T. M Meier, and A. Wokaun, 1986, Anal. Chim. Acta 181, 139. Vo-Dinh, T., G. H. Miller, J. Bello, R. Johnson, R. L. Moody, A. Alak, and W. R. Fletcher, 1989, Talanta 36, 227.

PAGE 206

199 Vo-Dinh, T., M. Uziel, and A. L. Morrison, 1987, Appl. Spec. 41, 605. Wang, D.-S., and M. Kerker, 1982, Phys. Rev. B 25, 2433. Weaver, J. H. C. Krafka, D. W. Lynch, and E. E. Koch; 1981a. Optical Properties of Metals, Part 1 (Fachinf ormationszentrum, Karleruke) Weaver, J. H., C. Krafka, D. W. Lynch, and E. E. Koch; 1981b. Optical Properties of Metals, Part 2 (Fachinf ormationszentrum, Karleruke) Weitz, D. A., S. Garoff, J. I. Gersten, and A. Nitzan, 1983, J. Chem. Phys. 78, 5324. Weitz, D. A., S. Garoff, C. D. Hanson, T. J. Gramila, and J. I. Gersten, 1982, Optics Lett. 7, 89. Wilson, E. B., Jr., J. C. Decius, and P. C. Cross, 1980, Molecular Vibrations: The Theory of Infrared and Raman Vibrational Spectra (Dover, New York) Wokaun, A., J. P. Gordon, and P. F. Liao, 1982, Phys. Rev. Lett. 48, 957. Wokaun, A., H.-P. Lutz, A. P. King, U. P. Wild, and R. R. Ernst, 1983, J. Chem. Phys. 79, 509. Yamada, H. H. Nagata, and K. Kishibe, 1986, J. Phys. Chem. 90, 818. Zeman, E. J., K. T. Carron, G. C. Schatz, and R. P. Van Duyne, 1987, J. Chem. Phys. 87, 4189. Zeman, E. J., and G. C. Schatz, 1984, in Proceedings of the 17— Jerusalem Symposium, edited by B. Pullman, J. Jortner, B. Gerber, and A. Nitzan (D. Reidel, Dordrecht, Holland) 413 ^ Zeman, E. J., and G. C. Schatz, 1987, J. Phys. Chem. 91, 634. i

PAGE 207

BIOGRAPHICAL SKETCH Scott Sutherland was born in 1962 in Gainesville, Florida, the son of Dr. Alan D. Sutherland, a professor of electrical engineering at the University of Florida, and Joyce Pine. While attending Gainesville High School, he became a member of the varsity tennis team, lettering for three years and posting a 21-3 singles and doubles record in his senior season. After finishing eighth in his graduating class, he chose to accept a four year William E. Brock Scholarship to the University of Tennessee at Chattanooga (UTC) While at UTC, he developed an interest in volleyball and was a member of the varsity volleyball team for two years. It was at UTC that he also developed his interests in both chemistry and physics, receiving a double major and being graduated magna cum laude. As part of his scholarship, he conducted research on the effects of magnetic fields on the rates of chemical reactions. He then was accepted to Stanford University for graduate studies, A number of factors led to his transferring to the University of Florida (UF) after one year of graduate studies. Just prior to his transfer to UF, he was awarded a Hughes Aircraft Doctoral Fellowship, which he retained for one year, after which it was cancelled by mutual agreement between himself and Hughes Aircraft. During his second year at UF, he 200

PAGE 208

wrote a program to simulate the scanning tunneling microscope and presented the results and the Sanibel Symposium in 1988. After this project, he began work on his dissertation research, developing new substrate preparation methods for surface-enhanced Raman active substrates. He has published one paper, has two in press, and two more on the way. He is currently seeking employment in industry as a research chemist.

PAGE 209

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. kmes D. Winefos'dner, Chair Graduate Research Professor of Chemistry I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. ^^i'chael C. Professor of Chemistry I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Martin T. Vala Professor of Chemistry I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. 1^ David B. Tanner Professor of Physics I certify that I have read this study and that in my opinion It conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Timothy J. 'Anderson Professor of Chemical Engineering

PAGE 210

This dissertation was submitted to the Graduate Faculty of the Department of Chemistry in the College of Liberal Arts and Sciences and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December, 1990 Dean, Graduate School


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EQZZHKHPS_NF3G33 INGEST_TIME 2014-12-19T00:13:10Z PACKAGE AA00026524_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES