Surface-enhanced Raman spectrometry


Material Information

Surface-enhanced Raman spectrometry an evaluation of commercially available substrates for routine analytical applications
Physical Description:
xii, 202 leaves : ill. ; 29 cm.
Szabo, Nancy Joy, 1965-
Publication Date:


Subjects / Keywords:
Chemistry thesis, Ph. D
Dissertations, Academic -- Chemistry -- UF
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1996.
Includes bibliographical references (leaves 188-201).
Statement of Responsibility:
by Nancy Joy Szabo.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002100756
oclc - 35122414
notis - AKT9785
System ID:

Full Text







With great love and respect, I dedicate
this volume to my father, Dr. James I.
Szabo, for the research that he always
longed to do.


My heartfelt appreciation and gratitude goes first to James D. Winefordner, my

research director and the mentor I had hoped against hope for. He has treated me with

patience and fairness at all times. I admire both his incredible breadth of knowledge and

wisdom, as well as his humanity. He is one of the few people I have ever known to strike an

honest balance between the requirements of a disciplined career and the requirements of being


On the technical side of things, I gratefully acknowledge Jim's second in command,

Ben Smith, for his practical approaches to bothersome problems; Jason Miller who worked

with me for a year as an undergraduate research assistant; Steve Bourden of the Immunology

Department for the use of the bench-top sputterer; John Fijol of Dr. Paul Holloway's lab in

the Department of Material Science and Engineering for the use of the vapor deposition

chamber; Greg Erdos and company of ICBR for all of the SEM work; 3M scientists Steve St.

Mary (EMPOREM and TLC) and Don McClure (PET) for their support and assistance, along

with a supply of free substrate materials, and 3M for partial funding of this research.

I also want to thank my parents, my mother who has a constant supply of faith--come

what may--and my father whose example taught me tenacity and perseverance even during

the slow and painful times of life. And my back-up parents, Bev and Dave Vorbach for

always being there and always being supportive, even when I don't make sense.

Next to my family are four people who deeply affected my life, and they will always

hold my warmest regards: Mrs. Foster, my second grade teacher for refusing to let the label

of dyslexia limit a child's potential; she, in my mind, is what every elementary teacher should

be; Mr. Fred Dupre and Miss Kathy Gillespie for their encouragements and challenges in high

school math and chemistry; and Dr. Brian Vogt, who called me up in the middle of lab one

day to tell me that chemistry was where my heart was.


DEDICATION .................................. .................. ii

ACKNOWLEDGEMENTS ............................................ iii

LIST OF ABBREVIATIONS ......................................... viii

ABSTRACT ............ ..... ................................... x


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

2 BACKGROUND AND THEORY OF SERS .............................. 5

Raman Scattering and Spectroscopy ................................. 5
Quantum Mechanical Description ............................. 7
Classical Description ............. ...................... 10
Practical Considerations ................................... 12
Enhanced Raman Techniques ..................................... 15
Resonance Raman Spectroscopy ............................. 15
Surface-Enhanced Raman Spectroscopy ....................... 19
Surface-Enhanced Resonance Raman Spectroscopy .............. 29
Enhancement M echanisms ....................................... 31
Electromagnetic Theories ................................... 32
Spherical and Spheroidal Theories ........................... 32
Electrodynamic Theory .................... ............... 36
Corrected Electrodynamic Theory ........................... 39
Chemical Theories ......... ........ ................... 42
Selection Rules ....................................... 44

ACTIVE SUBSTRATES ............................................. 47

B background ............................ ..... ............... 47

SERS from TLC Materials ......... ...................... 47
Description of TLC Plates ................ .............. 50
TLC Application Study .................................... 51
General Application Study ................................. 56
Experimental .............. ................................ 56
Instrumentation .......... ......................... 56
Chemicals and M materials ............................... 57
Procedures .............. ... ...................... 58
Results and Discussion ........................................ 62
TLC Study ........................................... 62
General Applications Study ............................... 75
Conclusions ................................................. 95


Background ................. ............................... 97
Description of Empore .................................. 97
Purpose of Study ..................................... 98
Experim ental .......... .. ....... .. ....... ..... ........... 99
Chemicals and M materials ................................... 99
Procedures ............... ............... ............. 99
Results and Discussion .......... ......................... 101
Silver-Coated Empore Surfaces .......................... 101
TLC Study .................... ....................... 108
General Applications Study .............................. 118
Conclusions ................................. ............ 129


Background ........................... .................. 140
SERS from Posted Substrates .......................... 140
Description of Anti-Reflective Film .............. ........... 143
Experimental ............................................. 151
Chemicals and Materials ................................ 151
Procedures .......... .... ... ... ..... ..... ........... 151
Results and Discussion .......................... ........... 153
Chemical Reduction ......... ...... ................... 153
Thermal Evaporation .................................... 163
General Applications Study ............................... 167
Conclusions ............... .......... ...... ..... ......... .. 182

6 CONCLUSIONS AND FUTURE WORK .............................. 184

Conclusions .............................................. 184
Future Work ........................ .................... 186

LIST OF REFERENCES ........... ............................... 188

BIOGRAPHICAL SKETCH .......... ............................. 202


9AA 9-Aminoacridine
AOAC Association of Official Analytical Chemists
APYR 1-Aminopyrene
ARG Argentation Plate
BA Benzoic Acid
BAK Bakerflex Plate
BYP 2,2'-Bipyridine
CCD Charged Couple Device
CED Corrected Electrodynamic Theory
CRS Conventional Raman Spectrometry
CV Crystal Violet
EF Enhancement Factor
ED Electrodynamic Theory
EM Electromagnetic
EMP Empore
EPA United States Environmental Protection
FDA United States Food and Drug Administration
FOM Figure of Merit
FSIS Food Safety and Inspection Service
HPLC High Performance Liquid Chromatography
IR Infrared
LDR Linear Dynamic Range
LOD Limit of Detection
MNA m-Nitroaniline
NPYR 1-Nitropyrene
ONA o-Nitroaniline
ORC Oxidation-Reduction Cycle
PABA p-Aminobenzoic Acid
PET poly ethylene terephthalate
PMT Photomultiplier Tube
PNA p-Nitroaniline
PNBA p-Nitrobenzoic Acid

PTFE Polytetrafluoroethylene
PYR Pyrene
R2 Correlation Coefficient
Rf Retardation Factor
RRS Resonance Raman Spectrometry
RSD Relative Standard Deviation
SDZ Sulfadiazine
SEM Scanning Electron Microscopy
SMR Sulfamerazine
SMT Sulfamethazine
SERS Surface-Enhanced Raman Spectrometry
SERRS Surface-Enhanced Resonance Raman
SPE Solid-Phase Extraction
SPET PET coated by sputtering
TLC Thin Layer Chromatography
UV-vis Ultraviolet-visible
VPET PET coated by vapour deposition

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



Nancy Joy Szabo

May 1996

Chairperson: James D. Winefordner
Major Department: Chemistry

The most active substrates for Surface-Enhanced Raman Spectrometry (SERS) are

those designed for theoretical studies. Intense, reproducible signals arise from features

uniform in size and shape. The methods permitting the control necessary for producing

surfaces of this quality are not, however, practical for use in commercial laboratories, where

money and equipment are limited, as are time and employee labour. Less expensive surfaces

are formed by less precise methods which compromise signal intensity and reproducibility, and

limit the applicability of SERS as a routine method.

The ideal applications-oriented SER substrate would be inexpensive and require little

time, labour or equipment for preparation. A high percentage of the prepared surfaces would

respond (as few as 15% of some surfaces are Raman active), giving reasonably intense signals

with a high reproducibility between regions on a single substrate and between different

substrates of the same preparation method. The materials would also be rugged enough for

packaging, transport, and storage.

With this in mind, four commercially available materials were evaluated for use as

SERS substrates. Two different thin-layer chromatography (TLC) plates were studied with

specific regard to the analysis of sulfonamides. One plate was purchased pre-coated with

silver, the other was coated in-house via chemical reduction. Although the separation of the

sulfa drugs was not strongly affected by the silver, the SER response was disappointing.

These surfaces were moderately active.

The third was an unconventional membrane used for both TLC and solid-phase

extraction. The material, silica gel embedded in a Teflon mesh, was also coated in-house by

chemical reduction. The sulfa drugs again separated easily, although the signals obtained

were somewhat disappointing. Other compounds responded strongly, however, with an

improved reproducibility. In addition to being sturdy to the normal rigors of laboratory use,

this material could also be stored for at least six months with no change in performance.

The fourth material, an etched polymeric film, approximated the theoretical substrates.

Silver was applied by both chemical reduction, which produced only a weakly active surface,

and by thermal evaporation. The evaporated surface, strongly active (enhancement factor,

> 10), yet simple to prepare, showed the most promise for successful commercial application,

and should meet most of the criteria set forth by the analytical sector.

The figures of merit varied with the substrates. Overall, the signal vs concentration

curves were linear across one to three orders of magnitude with limits of detection in the low

ng range (total amount applied). Under routine sample preparation and data collection

procedures, reproducibility was within the range accepted for SER substrates (<30%) but

improved to 5-15% when greater care was applied. Signals were further enhanced when a

thin film of water was held on the substrate surface during analysis in a step that modified the

dielectric constant and improved the thermal conductivity.



Modem surface-enhanced Raman spectrometry (SERS) has overcome most of the

drawbacks associated with the conventional Raman method. The low cross section of the

scattering phenomenon has been improved by up to eight orders of magnitude, and

fluorescent interference have been virtually eliminated. At the same time, the integrity of the

spectral information has been maintained and the intensity enhanced.

Currently, SERS is used for the identification of molecules present in trace amounts,

the determination of structural components, the investigation of kinetics and mechanisms for

chemical or biochemical reactions, and the characterization ofinterfacial regions. The method

is both qualitative and quantitative by nature and can differentiate between isomers (Sarkar

et al, 1993). Samples may be studied in solution or in a dry state in amounts that vary from

a submono-layer to a thin film. Analytes have included polymers (Bulkin, 1991; Kudelski and

Bukowska, 1992), anti-tumour drug-target complexes ( Barton et al., 1992; Nabiev et al.,

1994), pollutants (Carrabba et al., 1987), pesticides (Alak and Vo-Dinh, 1987; Narayanan et

al, 1992, 1993a and b), pharmaceuticals (Sutherland et al., 1990; Ruperez et al., 1991), dyes,

(Brandt, 1989) petroleum products (Gerrard, 1991), and surfactants (Suga et al., 1993)

among others. SERS has even been applied successfully to remote sensing (Alarie et al.,


1992; Carron et al., 1992) and as a postcolumn detector for high-performance liquid

chromatography (Pothier and Force, 1992; Cabalin et al., 1993). Despite its proven abilities,

SERS is not seen as a generally applicable analytical technique, as much as a method for

special applications.

The major factor barring SERS from a wider analytical acceptance is the commercial

availability of analytically suitable substrates (Gerrard and Birnie, 1992). To summarize the

problem, substrates that are available are not suitable and those that are suitable are not

available. For instance, substrates that are inexpensive, easy to use, and sufficiently sturdy

for routine analysis (ie, silver-coated filter papers), do not give large signal enhancements, do

not have features readily described by theory, and frequently have a quantitative precision

around 30%--not acceptable for commercial application. On the other hand, substrates (ie,

plasma-etched silica posts) giving large theoretically-predicted signal enhancements with high

substrate-to-substrate reproducibility are an economic nightmare: (1) the production tends

to be time consuming and labour intensive; (2) the size of the individual substrate is limited

to a few millimeters; (3) the percentage of finished surfaces suitable for use is low; and, (4)

those surfaces not damaged during production, are often so fragile that damage by sample

application is not unusual.

Substrate research has been sharply divided between studies that are theory-oriented

and those that have a goal of general application. Theoretical research continues refining the

descriptions and predictions of SERS phenomena, searching for substrates with theoretically

proper features (size and shape) in uniform distributions. On the other side, applications-

oriented research concentrates on creating reliable substrates suitable for routine analysis:


economically feasible; simple to prepare, store, and use; sufficiently sturdy for shipping and

sample application; and reproducible from region-to-region and substrate-to-substrate.

The scope of these studies encompasses the behavior of four commercially available

materials that have been prepared and tested under similar conditions. The performances of

these materials as substrates suitable for routine SERS analysis have been evaluated, as have

the behaviors of three of these materials with specific regard to the use of SERS as a

detection method for thin-layer chromatography (TLC). Several suggestions have also been

made for the modification of these procedures to take advantage of the future potential of

these materials.

The scattering processes leading to the Raman effect are described in Chapter 2, along

with the need for signal-enhancement, and the methods used to increase the weak Raman

cross section. The relative merits and limitations of Raman and its enhancement techniques

are discussed. An overview of the major theories of surface-enhanced Raman is included with

an emphasis on those points pertinent to this study.

In Chapter 3, two commercially available TLC plates are studied. Both are composed

of silica gel; neither contain fluorescent indicator. One was purchased already embedded with

silver (argentation plate); the other was coated in-house by chemical reduction. Observations

regarding the coating processes and general behavior are discussed as a part of which the

signal-enhancing properties are evaluated. To determine the suitability of SERS as a detection

method for TLC, a mixture of sulfonamides was separated and detected directly from each

plate. Due to extremely weak signals, improvements in signal enhancement were sought

through modification of the dielectric constant by the addition of a thin film of water to the


substrate surface. Spectra of several analytes taken under these conditions are compared to

signals obtained from conventionally treated substrates.

The study of TLC materials is continued in Chapter 4 with the novel EMPOREM

material, a supportless plate composed of silica gel particles embedded in a Teflon webbing.

A comparison is made with the first two materials regarding coating, chromatographic, and

signal-enhancing behaviors.

Chapter 5 introduces an etched polymeric film with features approximating in size,

shape, and distribution the theoretically well-behaved substrates mentioned earlier. Silver was

applied by chemical reduction (for comparison with the TLC surfaces), by vapour deposition

and by sputtering (conventional deposition techniques for surfaces with features of this type).

Additional enhancements were obtained for the less active chemically reduced films by "back-

passing" the laser through the etched polymer film and silver layer to the analyte rather than

striking the analyte molecules directly. Observations regarding general behavior and analyte

response are discussed and the activities of all four surfaces are compared. This material was

not a part of the TLC study.

Chapter 6 contains the conclusions and recommendations for future studies, such as

the filtration of colloids onto the EMPORE material, the formation of needle-like metal

particles on posted polymeric materials, the substitution of polystyrene and/or collodion for

polyethylene terephthalate) to reduce the interfering background of the etched polymer

substrate, and the use of rolled silver-coated polymer strips for the on-line detection of high-

performance liquid chromatography.


Raman Scattering and Spectroscopy

Five years prior to the first observation of the Raman effect, the Compton effect was

described using quantum mechanics to explain a change in wavelength observed for X-rays

scattered from electrons. Later that same year, an Austrian physicist extended the theory,

predicting that radiation scattered from molecules would also consist of photons at discretely

altered energies in addition to radiation at the incident energy. Proposed by Smekal in 1923,

the theory found support two years later from Kramers and Heisenberg (1925) who reached

the same conclusions by classical means. The findings of Schrodinger in 1926 and Dirac in

1927 independently agreed with the earlier predictions.

The inelastic scattering of light from molecules was finally documented by Sir C. V.

Raman, an Indian physicist, in 1928, when "true scattering" (Raman, 1928; Raman and

Krishnan, 1928) was detected from a sample of liquid benzene. Light for the experiment was

obtained by passing a converged beam of sunlight through a blue-violet filter to the sample

cell. Detection was based on the principle of complementary filters--a yellow-green filter was

moved into and out of the observer's line of sight while the difference in visible light was

noted. The observation was confirmed by American scientist, R.W. Wood (1928), with

reports from Russia (Landsberg and Mandelstam, 1928) and France (Rocard, 1928;


Cabannes, 1928) soon following for samples in the solid and gaseous phases, respectively.

Raman received the Nobel Prize in Physics for his discovery in 1930.

Light incident upon a molecule may interact by either absorption-emission or

scattering. The most common absorption-emission techniques are ultra-violet, visible, and

infrared absorption spectroscopies. In ultra-violet and visible methods, a molecule absorbs

a photon in the ultra-violet or visible range and is excited to a higher electronic state. Infrared

absorption spectroscopy differs slightly from these methods in that the incident energy is

considerably lower, exciting the molecule only to a higher vibrational or rotational level

without change in the electronic state. Broadband excitation sources in the region of interest

are generally used to supply radiation corresponding to a difference in energies between two

stationary or real levels in the sample molecule. Fluorescence processes have lifetimes on the

order of 10-8 s (Barafiska, 1987) which allows absorption and emission to be observed as

separate events.

For scattering to be detected, the exciting radiation must be monochromatic with the

incident energy not matching energy differences between stationary states. For Raman

scattering, this energy is usually significantly less than that required to reach the first excited

electronic state. Scattering results from the weak interaction of the electric vector of the

electromagnetic wave with the electrons of the molecule. As the electrons shift with respect

to the positive center of the molecule, a periodic vibration in the electrons causes a temporary

polarization, during which the energy of the radiation is momentarily retained in a non-

stationary or virtual state. When the molecule ceases oscillating, it returns to one of several


vibrational levels by emission of a photon. The entire process takes 10`5 10-14 S (Campion,


If the molecule returns to its original level, the photon is elastically (Rayleigh)

scattered and contains the same energy as the incident radiation. If the molecule experiences

a vibration during the excitation process, it may relax to a vibrational state higher or lower

than its original level Inelastic (Raman) scattering then occurs and the energy of the emitted

photon differs from that of the incident photon. Unlike the absorption processes mentioned

earlier, scattering cannot, at this time, be studied as separate photon events (Barafiska, 1987).

Two theories are used to describe the conventional Raman scattering process. The

first is based on quantum mechanics, and, therefore, considers molecular energy levels to be

quantized and treats radiation according to particle theory. The counterpart to quantum

theory is classical theory which regards radiation as waves and describes Raman scattering

in terms of the polarizability of the molecular electron cloud. Brief descriptions of each of

these theories are included. For greater detail, thorough discussions of each have been

published by Banwell (1972) and Skoog (1985).

Quantum Mechanical Description

The Raman spectrum is essentially an emission spectrum based on energy differences

depicted as frequency shifts, using the relationship

AE =h Av, (2.1)

where the energy (E) and frequency (v) differences of photons or molecular transitions are

related by the Planck constant, h. At room temperature, most molecules are found in the

electronic ground state. Raman Stokes scattering (hvst) occurs after excitation as the


molecule relaxes into a vibrational level above the ground state (Figure 2. la) with a red-shift

in energy corresponding to

hvs, = hvo hv, (2.2)

the difference between the incident photon frequency (v.) and frequency of the molecular

vibration, vy. When, however, the molecule is in some elevated state prior to excitation,

relaxation to the ground state releases a photon of higher energy (anti-Stokes scattering,

hvs,,) than the incident radiation and corresponds to a blue-shift in the frequency:

hvast = hvo + hv, (2.3)

Taken together, the Stokes and anti-Stokes shifts create a symmetrical arrangement of bands

(Figure 2. lb) on either side of the Rayleigh frequency.

As seen by rearrangement of equations (2.2) and (2.3),

hvo hvst = hv, (2.4)

hvas, hvo = hv, (2.5)

absolute differences between the frequencies of the incident photon and both scattered

photons are the same and equal to the frequency of the molecular vibration. Thus, frequency

shifts are shown to be independent of the frequency of the exciting energy, depending only

on the vibrational character of the molecular species.

Out of all scattering events, only a small portion--106 or less (Campion, 1983)--are

inelastic with the majority contributing to the strong Rayleigh line. Compared to the Rayleigh

intensity, the intensities of Stokes bands are about 10' times weaker (Figure 1. Ib). Anti-

Stokes bands are weaker yet--by about one hundred times. This discrepancy is a function of

the Boltzmann distribution. At room temperature, most molecules are in the ground

v-2 ___________________
v-1 __I_________

First Excited State

Virtual States

Vibrational Statev=2

Ground State

hv -hv
O v

_ 1________
--______ ------



} hv

hv +hv
o V








0 V 0 OV


Figure 2.1 (a) Energy level diagram showing the major transitions for infrared absorption
and Stokes, Rayleigh, and anti-Stokes scattering. (b) Symmetrical spectral bands for Raman
interactions of(a) with average relative intensities.


electronic state with very few in vibrational levels that could produce an anti-Stokes event.

Using the conclusions of Placzek's polarizability theory (Barafiska, 1987; Bulkin, 1991), the

observed relative intensities of Stokes (Ist) to anti-Stokes (Lts) lines can be described as:

Ist (Vo-V)4 hv (2.6)
I, (v+v)4 kT

where k is the Boltzmann constant and the intensity dependence on the absolute temperature

(T) is apparent. Due to the difference in intensity, most analytical methods scan only the

Stokes side of the spectrum.

Classical Description

When a molecule interacts with an electric field, the electrons distort, shifting with

respect to the positive centre, and create an induced dipole. The moment (Cm) of this dipole

is proportional to the intensity of the electric field (E, Vm') according to the molecular

polarizability, a,

i = aE. (2.7)

Polarizability (j'C2m) is a property specific to the molecule, determined by its size and shape

and indicating the ease with which the electrons may be shifted with respect to the molecular

spatial coordinate system. Unlike the dipole moment which is a vector quantity, polarizability

is a tensor quantity and requires a nine-member matrix to describe the shape-volume shift.

Depending on the molecular bond axes, the values may vary in all three spatial directions.

When the electric field of the incident radiation varies periodically in intensity, the field

experienced by each molecule is described by

E = Eocos(2nv ot) (2.8)

where Eo is the intensity of the electric vector and t is the time (s). Upon interaction, an

oscillating dipole moment is induced in the molecule:

p = aE.cos(2nitvt). (2.9)

Since periodic dipoles emit radiation of the same frequency as the vibration, the molecule

scatters light of vo, the Rayleigh frequency. When the molecule vibrates or rotates in such

a way that the molecular polarizability is affected, the changing polarizability is described as

a = ao + Bcos(2nvvt), (2.10)

where ao is the polarizabilty in the ground state and B is the rate of change of the

polarizability with respect to the normal coordinate system of the molecule. The resulting

induced dipole can then be better defined by combining expressions (2.9) and (2.10):

V = a0oEcos(2iTvot) + E0Bcos(21ivot)cos(2invt). (2.11)

Using the trigonometric identity,

cosacosB = /2cos(a-B) + '/2cos (a+B), (2.12)

scatter from the vibrating dipole is described in this fashion:

p = aoEocos(2niVot) + V2BE0{cos[2n(vo-Vj)t] + cos[2n(v.+v,)t]}. (2.13)

From this expression, the source of both the Rayleigh line and Raman bands are described,

as is the symmetry of the Raman spectrum There is, however, no indication of the intensities

of the lines. From a strictly classical description, the intensities are expected to be the same.

Observed variations can be explained by the Placzek theory or quantum mechanical treatment.

Equation (2.13) also shows the dependence of vibrationally induced scatter on a

change in polarizability. If a molecule lacks polarizable modes, B=0 and only Rayleigh scatter


is possible. Although selection rules are used to determine which modes lead to inelastic

scattering, the rules indicate only whether a band may appear; no indication of the intensity

is given. A brief qualitative description of the selection rules is included in the next section.

Detailed discussions were prepared by Herzberg (1950) and Steinfeld (1985).

Practical Considerations

Of the absorption-emission techniques, Raman spectrometry is most closely associated

with infrared (IR) absorption spectroscopy as the two methods are complementary in the type

of information elicited from a sample. Both are vibrational techniques primarily used for the

identification of unknown compounds and the elucidation of structure and conformation. IR

and Raman spectroscopies are also frequently applied to quantitative analyses, the study of

chemical reactions, and the analysis of surfaces. A thorough comparison of the techniques

is provided in a review by Domingo and Escribano (1992) with characteristic frequencies of

the resulting spectra contrasted by Lin-Vien et al. (1991).

Although spectra from the techniques often contain bands at identical frequencies, the

mode intensities usually differ due to excitation mechanisms based on different selection rules

(Barahska, 1987). For instance, Raman activity is detected if there is a change in

polarizability during a molecular vibration (da/dx), whereas a mode is considered active in

the IR only if there is a change in the dipole moment (dp/dx). Raman modes mainly originate

from symmetric vibrations and movements within carbon skeletal bonds, while infrared modes

come mostly from asymmetric vibrations and vibrations related to polar groups in the

molecule. For species possessing a center of symmetry, the rule of mutual exclusion is

observed. For those without such a center, vibrational modes may be shared, often appearing


in the spectrum of each technique. Selection rules based on molecular symmetry models have

been described for both processes (Cotton, 1971; Drago, 1977). One additional point to

note, selection rules which are usually devised for isolated species in solution or gas phases

(sample types often studied by Raman methods) may be greatly altered by intermolecular or

intrachain effects in denser samples (Campion, 1983). In extreme cases, the rules are simply


Being complementary, Raman techniques have certain advantages over infrared

analysis methods. (1) Although any wavelength may be used, Raman measurements are

usually made in the UV or visible range which simplifies instrumental requirements, ie., optics,

sample cells, and detectors. (2) Unlike traditional IR spectroscopy which may require several

instruments, a Raman spectrum covering the entire range of frequencies may be obtained from

a single instrument. (3) Although water absorbs infrared radiation strongly, it is such a poor

scatterer that Raman samples are routinely prepared in aqueous solution. (4) Fewer method

constraints are placed on Raman samples than on samples prepared for an IR study: Materials

in trace or bulk amounts, and in solid, liquid, or gaseous phases are amenable to examination

by Raman without requiring alteration to the instrument other than modification of the sample

holder. (5) The unique properties of the laser as an excitation source for Raman scatter

(Porto and Wood, 1962) allow the study of a wide range of sample types and variations on

the method of study, eg., Raman microprobe. Infrared absorption spectroscopy is somewhat

more limited in both respects. (6) Because the intensity of a Raman mode is linearly related

to the concentration of the scattering species, quantitation is simplified when compared to IR

methods which are often plagued with nonlinearities and overlapping bands. (7) Qualitative


analysis is also simpler as Raman spectra are characteristically less complex than infrared

spectra which routinely contain overtone and combination modes.

Compared to IR, Raman spectra exhibit a lack of sensitivity to minor structural details

and show a strong dependence on instrumental effects (such as, laser power, cell geometry)

which makes comparisons of spectral intensities between systems difficult (Fryling, et al.,

1993). The Raman system is also expensive, and technical skill is required for both

instrument operation and spectral interpretation. An additional hindrance to compound

identification is the limited library of reference spectra; the collection of IR spectra is far more

complete. Despite these problems, the greatest limitation facing conventional Raman

scattering is the weak cross section (-10-'3 cm2 for a moderate scatterer). Because of this,

the fluorescence of minor constituents, contaminants, or even the analyte itself may obscure

the Raman signal reducing the overall sensitivity of the technique.

For scatter to be useful, the detected intensity of the vibrational modes (IR) must be

enhanced sufficiently to overcome the fluorescent background/interferents. A number of

approaches have been used, each of which is suggested by the definition of the Raman

intensity. The following expression is specific to sample molecules adsorbed onto a solid


IR = p P (do/dQ) Q (2.14)

where p is the density adsorbatee species cm') of the scattering species, P (photon sec"1) is

the laser power, (do/dQ) is the differential Raman scattering cross section (cm2 sr1), and a

(sr) is the solid angle of collection. From this equation, four possible means of improving IR

are suggested: (1) Increasing the solid angle of collection, possible when the efficiency of the


detector and collection system is improved. With modem instruments, optics, and detectors,

the collection angle has already been optimized. (2) Increasing the density of adsorbed

species within the probe area by roughening the surface to enlarge the available surface area

which has permitted signal improvements up to 10'. (3) Increasing the amount of radiant

energy reaching the surface which is primarily dependent on the choice of excitation source.

In the past (pre-laser), the source of choice had been the low-pressure mercury arc lamp with

its respective filters. Later, the laser renewed interest in Raman methods as coherent,

monochromatic light became available, and, at high power densities. Increasing signal

intensity by increasing the intensity of light at the surface is limited, however, by the potential

for sample degradation and/or desorption (due to the low thermal dissipation rates of most

samples). To minimize damage, laser powers are rarely permitted to exceed 1 W at the

sample surface. (4) Increasing the scattering cross section which requires that the interaction

between the radiant energy and the molecule be strengthened. When the excitation frequency

is selected to overlap the molecular electronic absorption band, signal enhancements of two

to five orders of magnitude are commonly observed with improvements up to eight orders of

magnitude being possible (Asher, 1993a). This phenomenon forms the basis for the first major

enhanced-Raman technique, discussed in the next section.

Enhanced Raman Techniques

Resonance Raman Spectroscopy

The Resonance Raman (RR) effect, first described by Placzek (Bernstein, 1979) in

1939, occurs when the laser frequency lies very close to an electronic transition in the


molecule. (The theoretical treatments of Spino and Stein (1977) and Bernstein (1979) are

recommended for additional reading.) Molecules having resonances or groups with

resonances in the visible are preferred as samples, and the source is selected for having an

excitation line in the same region as the absorption band. To extend usage to inactive or

weakly active molecules, the sample species may be labelled with a chromophore possessing

the desired resonance properties. Depending on the labelling process, sample preparation can

become quite complicated.

RR spectroscopy is well-suited for trace analysis. The method is sensitive (Limits of

detection as low as 108 M have been reported.) and highly selective, having spectra that are

less complex and more easily interpreted than those obtained by conventional Raman (CR)

techniques. Since spectral enhancements are restricted to those vibrations coupled to the

electronic transition, bands appearing in the spectrum are related to the chromophore. Band

positions and relative intensities contain information describing molecular structure,

orientation, and the chromophoric environment. Both CR and RR spectra can be gathered

from the same instrumentation, except when source requirements demand otherwise.

Because the laser frequency is in close proximity to or overlaps with the molecular

transition, spectra frequently show intense broad band fluorescence backgrounds that obscure

even enhanced modes. Although contaminants are sometimes at fault, such interference

generally come from the sample species themselves. If available, sources emitting excitation

lines that avoid fluorescing modes are chosen to alleviate the problem. Otherwise, the analyte

is usually studied under flowing-stream or low-temperature conditions (Mathies, 1979).

Where fluorescence is a persistent interference, time-resolved methods (Terner and El-


Slayed, 1985; Watanabe et al., 1985; VanHoek and Visser, 1985) have been applied to

separate the two signals.

In addition to the ever-present fluorescence problem, certain analytes are predisposed

to self-absorption or photolytic decomposition. Self-absorption is usually reduced by diluting

the sample (as low as 10--10-7 M) in a thermally conductive matrix (Nimmo et al., 1985).

When decomposition occurs, dilution and /or rotation of the cell or holder (Eng et al., 1985)

is employed to decrease sample heating and the potential for molecular breakdown.

The recent use of excitation in the UV (Review: Asher, 1993a and b) and IR (Chase,

1987 and 1991) has relieved the above restrictions to a large extent, making RR methods

applicable to greater numbers of molecules. Few molecular species fluoresce appreciably

below 250 nm (Hudson et al, 1986). To minimize sample decomposition, UV lines are kept

below 50 mW and signal collection from flowing stream systems is standard practice (Asher,

1993a). Enhancements obtained with IR excitation are often weak. With further solvent

restrictions and instrumental considerations IR seems to have little benefit over visible

sources. When, however, RR obtained with a near-IR line is equivalent to or more intense

than enhancements from visible excitation, the use of IR offers the advantages of fewer

photochemical effects and no fluorescence (Chase, 1991). Use of either UV or IR affects the

selection of instrumentation; optics and detectors must be suited to the wavelength range in


RR spectroscopy is mainly applied to biochemical studies (Review: Parker, 1983),

especially for the chemical analyses of aqueous biological and environmental samples.

Polycyclic aromatic hydrocarbons, for instance, have been characterized (Asher, 1984; Jones


and Asher, 1988), detected in coal-derived liquid distillates (Rummelfanger et al., 1988), and

the effect on and presence ofpyrene in the DNA of calf thymus (Cho and Asher, 1993) have

been studied. Industrially, RR spectra are used in petroleum (Williams and Klenerman, 1992)

and polymer chemistries (Bowley et al., 1985; Kuzmany, 1990) to monitor reactions and

degradation processes, as well as, to identify and characterize molecules. Experiments

describing chemical dynamics in the solution phase and the existence/behavior of transient

intermediates (Oyama et al., 1993; Temer and El-Slayed, 1985) have also attracted serious


Because the technique is sensitive, low concentrations of the analytic species are

acceptable and the actual volume of sample required is minimal (nanoliters). Molecules of

interest are frequently quite large, but the spectra remain simple as the only bands exhibited

are related to the chromophore. Thus, the study of specific chromophores in complex

mixtures or environments is permitted without interference from even closely related

structures (ie., the study of heme in a protein matrix; Asher, 1981). Because the

chromophore in a biological species is generally the centre at which most chemical/biological

function takes place, information regarding structural change during performance may be

obtained and utilized to clarify questions regarding reaction mechanisms (Heald et al., 1988).

In spite of its successful application to specialized areas of study, RR spectroscopy

is limited by fluorescent interference and compound suitability. The second enhanced

Raman technique, discovered accidentally, has attained a broader range of application by

quenching fluorescence.

Surface-Enhanced Raman Spectroscopy

The surface-enhanced Raman (SER) effect was first observed by Fleischman et al.

(1974), during a Raman study of pyridine adsorbed onto a silver electrode. The unusually

large signal was initially attributed to anodization of the electrode surface, the roughness

thought to permit a greater density of molecules in the probe area. Shortly thereafter, two

teams of scientists, Jeanmarie and van Duyne (1977) and Albrecht and Creighton (1977),

observed even larger enhancements from silver electrodes roughened to a lesser extent. The

increased intensity was then determined to be the result of an enhancement of the apparent

Raman cross section due either to the presence or roughness of the metal surface, rather than

to a simple increase in the number of molecules sampled. At this point, interest in the field

grew quickly with some groups attempting to explain and predict the phenomenon, and others

searching for practical applications and experimental limitations.

Currently, the SER phenomenon is described by two theories. Electromagnetic

theories account for long range effects, and explain the necessity of appropriately sized metal

surface features. Chemical theories account for short range effects and describe charge-

transfer and active site mechanisms, processes which necessitate a direct interaction between

molecule and surface. The electromagnetic mechanism is based on an enhancement of both

incident and scattered light. When the incident wavelength is in resonance with conduction

electrons in the metal, surface plasmons are excited and act to amplify the electromagnetic

field along the surface. Molecules in close proximity to the surface then experience not only

the electric field directly from the source, but also the field propagated by the plasmons. If

the light scattered from the sample is also within the resonance of the plasmons, it, too, is


enhanced similarly to the incident radiation. During this process, the actual Raman cross

section is unaffected, although the apparent cross section increases by -10'. Though

independent of molecule-substrate interactions, field amplification is dependent on the

exciting frequency, morphology (size and shape) of the surface features, and the dielectric

constants of both the metal and surrounding medium.

Chemical models, on the other hand, involve a true modification of the scattering

cross section via the alteration of sample molecular orbitals sorbed/complexed with the metal

surface. Compared to RR, the chemical SERS effect is less dependent on analyte species, but

certain analytes, better able to interact with the surface than others, respond very strongly.

Typically ionic, very dipolar, or easily polarized, these compounds contain functional groups

(nitro, amino, or carboxylate) and/or species (sulfur, phosphorus, or nitrogen) that promote

electron exchange. Aromaticity is also common. Detected surface-enhancements result from

a combination of both electromagnetic and chemical mechanisms, each contributing to the

overall effect. Theories and mechanisms will be described in greater detail in a later section.

SERS has been observed from a number of different metals: noble metals (Ag, Au,

and Cu), free-electronlike metals (ie., Na, K, Al; Moskovits and DiLella, 1982), and transition

metals (ie., Ni, Pd, Pt; Benner et al., 1983; Krasser and Renouprez, 1981 and 1982). To

achieve large enhancements, the excitation frequency must overlap the metallic resonance, a

function of the material's dielectric constant. Among the three noble metals this broad band

is located in two distinct regions, the near-IR for Cu and Au (Chase and Parkinson, 1991;

Creighton, 1982), the visible for Ag (Chase and Parkinson, 1991). To date, the largest EFs

have been obtained from Au and Ag, especially from silver, which, because of its high activity


and convenient resonance near the strong line of an easily accessible source (514.5 nm line

of the argon ion laser) has been used in more surface-enhancement studies than the other

metals combined. When the activities of silver and equivalent gold substrates were compared

(Laserna, 1993), EFs from silver substrates were found to be larger by a factor of 102. For

silver, the imaginary component of the dielectric function has a flat dependence on the

incident wavelength (Figure 2.2), thus the enhancement depends to a lesser extent on the

excitation than it does for other metals (Laserna et al., 1988).

Although SER spectra are vibrational in nature and provide details relating

conformational and environmental information about the molecule of interest, they are less

complex than Raman spectra, since there is a selective enhancement of scattering modes;

those closest to the substrate experience the strongest fields and are preferentially intensified,

ordinarily by enhancement factors (EFs) of 103-106 over vibrations displayed in CR spectra.

In addition to an increase in sensitivity, the fluorescence of adsorbed species is effectively

quenched, nonradiative decay into the metal being preferred to de-excitation by fluorescence.

Although species not in contact with the surface may still fluoresce, the large SER signals

are rarely obscured. Spectral interference from solvents in liquid systems are of little

concern, since water is still an exceptionally weak scatterer, and few organic solvents adsorb

appreciatively (Yamada et al., 1986).

Typically, the SER response is linear over 2-3 decades of analyte concentration with

detection limits in the low ng- to mid pg-range, lower sometimes, depending on whether

applied or probed molecules are used in the calculations. Precision, defined by the relative









N cn I I I

1 1 1 1 I I




x/u-uu / 7 \ s D


standard deviation, is extremely sensitive to the uniform distributions of both substrate

features and sample molecules, and ranges between a reported 3% for p-aminobenzoic acid

in a flowing stream of colloids (Laserna et al., 1987) to 30% for some of the poorer solid

substrates (Sutherland and Winefordner, 199 la).

SER spectra are obtained from three main types of substrates: electrodes, colloids,

and metal island films. Each of these will be described separately with major advantages,

weaknesses, and applications briefly mentioned. Unless otherwise specified, silver will be the

only metal specifically referred to in these discussions. Methods, observations and theoretical

trends, are, however, generally applicable to any enhancing metal.

Silver electrode surfaces are activated by oxidation-reduction cycles (ORC). In

addition to creating the rough features necessary for enhancement, the ORC also cleanses the

surface permitting a stronger interaction between molecule and metal (promoting chemical

mechanisms). Parameters affecting the enhancing capability of a given electrode include the

number of ORCs run, the amount of current passed during each cycle, electrode potential,

exciting frequency, and the natures and concentrations of the electrolyte and solute species.

Because the activity of an electrode varies during the course of its lifetime (with the total

number of ORCs), a spectrum may not be directly comparable to those obtained earlier or

later from the same electrode or from equivalent electrodes, used in the same system.

Reproducibility commonly varies between 10 and 30%.

Electrodes in electrochemical cells are frequently chosen as substrates for theoretical

experimentation because control of the conditions (potentials, roughness, etc) allows

separation of the electromagnetic and chemical effects. For instance, detection of weak SERS


from a polished electrode surface gave credence to proposed chemical models; subsequent

spectra, obtained after roughening by ORC, showed a further enhancement attributed to EM

mechanisms (Allen et al., 1980). Interfacial processes such as corrosion (Melendres et al.,

1992), adhesion, and catalysis (Plieth, 1992) have also been examined from in situ electrode

surfaces. Recent advances made by SERS toward the understanding of electrode-electrolyte

interfaces are reviewed by Pettinger (1992).

Colloids, generally silver, gold or copper dispersed in water, are simple to prepare

and easy to characterize. Preparation of silver hydrosols (Neddersen et al., 1993) almost

exclusively involves the reduction of aqueous silver nitrate by sodium citrate or sodium

borohydride, although Neddersen et al. (1993) have recently produced colloids by ablating

small particles of metal directly into the solvent of choice. The loss of simplicity by laser

ablation is compensated by long term stability of the colloids, a more uniform size distribution

(<10% between batches), and the lack of potential interferents (such as, borate; Herne and

Garrell, 1991) co-produced during traditional reduction procedures. Characterization is

performed using absorptiometric techniques. From an extinction spectrum, the average

particle size and optimum exciting frequency can be determined from the position of the band

maximum; the polydispersity is estimated from the band width.

Unlike other substrates, the signal enhancement observed from colloids is directly

dependent on particle shape and size (Creighton, 1982), and so adheres closely to theoretical

predictions. This remains true, even when the colloids aggregate in response to the sample

molecules. Initially, anions released during reduction adsorb onto newly formed colloids and

generate a dielectric bilayer that serves to isolate each particle via repulsive common charges.


On sample introduction, anions are replaced by adsorbing molecules and the bilayer is

compressed. Diffusional collisions then result in aggregation, and the SER signal reflects

changes in the colloidal structure. Enhancement increases as the optimum size and shape is

approached, then decreases as aggregation proceeds toward precipitation. Although

qualitatively useful, spectra collected during unstable periods are distorted with regard to the

relative intensities between bands and are, therefore, virtually useless for quantitative

measurement (Calvo et al., 1993).

For SER scattering to be observed a certain amount of aggregation is desirable.

Freshly prepared Ag hydrosols are spherical in shape and customarily have resonances -400

nm, somewhat out of the range of the most common excitation (514.5 nm). With aggregation

the hydrosols do not merge to form a larger particle, rather the sols adhere into randomly

branched chains in which each participating particle retains its original dimensions. In

response to chain length, the resonance shifts to longer wavelengths and a second maximum

appears. Ideally, partially aggregated sols would be stabilized in the resonance of the

excitation. To this end, static cells have been largely abandoned for flowing streams which

decrease signal variance by arranging for colloids to pass the point of detection only at the

optimum size and shape. The rate of aggregation, however, is dependent on the analyte

species and its concentration, so that change in either of these factors requires rearrangement

of the system to accommodate the new rates. Adding measured quantities ofhalides (CI, Br-,

or I; Liang et al., 1993; Fu and Zheng, 1992) into static and flowing systems has further

reduced the temporal problem by first promoting coagulation, then slowing the process to a

near-stop. This solution is somewhat awkward for frequent use, because the extent of


aggregation reached before quasi-stabilization is dependent on pH, temperature, and the

concentrations of all species present. Although slowed down considerably, the halide-

modified colloidal system is not static; slow kinetics govern a continual change that limits

reproducibility to -5%, when exceptional care is used with flowing streams, to 15-20% with

static cells (Cook et al., 1993).

Occasionally, metallic hydrosols have been converted to dry-state substrates by

nebulization or filtration onto filter papers (Tran, 1984), TLC plates (Koglin, 1988), and

membranes (Sutherland and Winefordner, 1992). Although precision was shown to improve

with physical stability (-5-10%, for filtered membranes), little research has been done with

this type of colloidal substrate. Physically, colloids have been used to study the effects of size

and shape on SERS enhancement factors (Creighton, 1982) and to confirm theoretical

predictions. Despite temporal problems, most biological and biomedical applications seem

to prefer colloidal substrates (Nabiev et al., 1994; Thornton and Force, 1991; Ruperez et al.,

1991) to other available surfaces.

Substrates, such as roughened electrodes or colloids in solution, do not require a

secondary support for their metal features. Metal island substrates, however, consist of a

base material coated with a discontinuous (island-like) layer of silver. The base acts as a

stable support for the active surface and often provides textural features to assist in the

formation of islands within a specific size or shape range. A wide variety of materials have

been used as supports--filter papers, TLC plates, microscope slides, semiconductors, capillary

tubes, polymer microspheres spin-coated onto glass slides, plasma-etched quartz post arrays,

and polymer grating replicas, among others.


Silver is typically applied by vapour deposition. The specific size and shape of the

generated islands is determined by film thickness, deposition rate, substrate temperature,

target orientation, pressure during coating, pretreatment of the support, and, if present, the

size and shape of underlying surface features. Most of these parameters can be controlled

fairly easily. The major drawbacks are access to the necessary equipment and limitations to

the number or sizes of substrates that can be prepared at one time.

As a less expensive alternative, island films are sometimes generated by chemical

reduction which generally involves the reduction of silver nitrate solution by some organic

component (Boo et al., 1985; Ni and Cotton, 1986). Rather than expensive equipment, a

degree of technical skill is required to achieve reproducible films. Colloidal preparation

methods have also been modified to coat solid substrates (Sutherland et al., 1991b).

Parameters affecting the final surface include pretreatment of the support, concentration of

the reactants, temperature, and reduction time. Supports coated by chemical reduction

generally respond with only moderate signal enhancements. Reduction methods do not

permit the fine control possible with thermal evaporation.

Metal island substrates divide easily into two main groups, those that are composed

of commercially available materials having simple preparations and giving low to moderate

enhancements reproducible within 30%, and those that involve complex preparation

procedures, but give large enhancements reproducible within 10%. This latter group consists

of quartz post array and crossed grating substrates. The former contains everything else with

the exception of spin-coated microspheres which fall between categories. Monolayers of

closest-packed microspheres spun onto glass slides create uniformly distributed textural

features, that, in turn, make reproducible substrates (15%; Vo-Dinh et al., 1989), but with

only moderate enhancing capabilities. The lack of large EFs compared to those of plasma-

etched post arrays is due primarily to the behavioral differences of the continuous thin Ag

layer coating the spheres and the discontinuous Ag particles capping each post of the quartz

post array or crossed grating substrates. Regardless of the grouping, metal island substrates

require special equipment (vapour deposition chambers, spin-coating devices, and/or etching

apparatus) for their preparation.

The more precise substrates have been used exclusively for physical studies, probing

the effects of particle size, shape, and isolation on enhancement. Most of the other substrates

have been developed to test the potential of SERS as a routine method of analysis. Separated

spots have been identified in situ from TLC plates (Koglin, 1990), and a fiber-optic based

Raman spectrometer has used silvered surfaces for remote analysis (Alarie et al., 1992).

Silver island films have also increased our knowledge of the order and structure of Langmuir-

Blodgett films (Aroca and Guhathakurta-Ghosh, 1989; Battisti and Aroca, 1992), materials

such as C60 (Akers et al., 1992), and the contaminants and interfacial reactions that can be

expected from semiconductor materials (Quagliano et al., 1993).

Surface-Enhanced Resonance Raman Spectroscopy

The enhancements separately obtained by SERS and RR may be combined to give a

larger Surface-Enhanced Resonance Raman (SERR) effect. Because the processes are based

on different mechanisms, enhancements are multiplicative and signals up to twelve orders of

magnitude have been predicted. In practice, however, enhancements are commonly only 1-3

orders of magnitude above observed RR or SER signals. Intensities are maximized when the

molecular absorption transition overlaps the excitation profile of the metal substrate and the

excitation line has been suitably selected to be in resonance with both sample and surface.

These requirements are a combination of the parent conditions: appropriately roughened

surfaces, species with an electronic transition near the excitation wavelength.

A problem inherent to the combined method is the tendency for the metallic resonance

maximum to shift due to the effect of the adsorbate on the dielectric constant of the medium

immediately adjacent to the particles. A submonolayer may be sufficient to shift the plasmon

resonance slightly or even completely out of range of the exciting line (Kim et al., 1989;

Zeman et al., 1987). Expected shifts in resonance frequencies are not at this point well-

defined theoretically; thus, the appropriate excitation line must be determined experimentally.

To its advantage, though, the signal intensities of many molecules do increase; a general

reduction in spectral fluorescence is also seen, since a nonradiative decay path is provided by

the metallic substrate.

The strengths of SERRS are contributed from both parent techniques. The surface-

enhanced response contains information regarding the distance of the chromophore from the

surface and/or interaction of the species with the metallic surface. The resonance

enhancement, on the other hand, indicates changes in the environment surrounding the

chromophore (usually bulk molecular) and changes induced in the chromophore by activities,

such as photoisomerization.

Biologically important molecules such as heme-proteins and porphyrins have been

studied by SERRS in an effort to clarify the red-ox properties of the heme group (Cotton et

al., 1989; Hildebrandt and Stockburger, 1986). Intact bacterial membranes and spinach


vesicles (from chloroplasts) have been adsorbed onto silver electrodes in topological studies

to determine the position of certain carotenoids in the photosynthetic membranes (Picorel et

al., 1988 and 1992). Because SERR spectra reflect photo-induced changes in isomeric

structures, the method has become a major tool in the study of phycocyanins, "light-

harvesting" antenna pigments that conduct light energy into chlorophyll reaction centers as

a part of the photosynthetic process (Farrens et al., 1989; Holt et al., 1989; Debreczeny et al.,

1992). SERR spectra for a variety of dyes, crystal violet (Chou et al., 1986), Rhodamine 6G

(Hildebrandt and Stockburger, 1984), and others chosen for their strong fluorescence

emmisions (Chambers and Buck, 1984), have been collected for fundamental studies,

involving the differentiation between electromagnetic and chemical contributions to the

enhancements, the effect of colloid activation (chloride ions) on available adsorption sites, the

mechanism of fluorescence quenching, and the extent to which fluorescence can be

suppressed. SERRS has also been used to evaluate the selectivity of shape-specific molecular

sensors (Kim, 1988) and to indicate the extent of pollution in bodies of water by detecting

subnanomolar amounts of nitrite from fresh seawater samples (Xi, et al., 1992). All SER

substrates are suitable for SERR applications.

Enhancement Mechanisms

The SER effect is described by two main theoretical models, chemical and

electromagnetic. Although neither mechanism alone suitably describes the observed

enhancements, most phenomena are adequately explained by a combination. Still under

discussion is the degree to which each mechanism contributes to the one signal, although it

is generally accepted that the contributions depend strongly on the system under observation.

The individual mechanisms have been thoroughly reviewed along with their contributions for

a number of systems (Chang and Furtak, 1982; Chang, 1987; and Campion, 1984). For a

unified view of the major models, Chase and Parkinson (1988) and Otto et al. (1992) are


Electromagnetic Theories

Spherical and spheroidal theories. Electromagnetic theories involve non-local or long range

effects and account for signal enhancements up to 10'. These models describe a surface-

induced increase in the electric field experienced by the analyte rather than a change in the

molecular cross section. The field amplification is the result of three effects (Nitzan, 1981):

(1) localized surface plasmons, the most efficient means of field amplification and the

mechanism on which most EM models dwell, (2) the lightening rod effect, a concentration

of local fields found near sharp metal edges or points of high curvature, and (3) the image

effect, a secondary effect related to a localized dipole induced in the metal surface nearest the

molecule by the molecule. Only the theories surrounding contributions from the surface

plasmons will be considered in this discussion. To simplify the interdependencies of the

parameters involved in the plasmon theories, the theories will be dealt with in a more or less

chronological fashion.

Metal particles are described by an aspect ratio, a/b, where a is the semi-major axis

and b, the semi-minor axis at right angles to a:


Figure 2.3 Spheroidal particle with semi-major axis a and semi-minor axis b, where a
> b.

According to convention, a is aligned with the axis of symmetry (z). In the first complete

EM theory, Kerker et al. (1980a and b) dealt exclusively with spherical particles, a/b=l.

Spheroidal shapes, a/b>1, however, better describe actual particles, and Gersten and Nitzan

(1980 and 1982) expanded the spherical theories to include the more realistic shapes. The

main difference between these two theories is the consideration of a bound vs. unbound

particle. Kerker was primarily interested in colloids thus his model described a free sphere

which was extended to a free spheroid. Gersten and Nitzan, on the other hand, were

influenced by the vapour deposition of metal islands onto crossed grating substrates which

were approximated as hemispheroids emerging from a flat plane.

In both spherical and spheroidal theories, metal conduction electrons are assumed to

act as a free electron gas. The electric field of the incident radiation polarizes these electrons

exciting a dipolar surface plasmon. The particle then becomes a source of the electric field

with the plasmon effectively amplifying the field along the surface of the particle. A nearby

molecule is, thus, exposed to a field intensity larger than that of the incident radiation alone,

and an enhanced signal is detected. Both theories also carefully consider the effects of

incident energy on the total enhancement and recognize the enhancement to occur at the


frequency for which metallic electrons oscillate in phase with the oscillating field. The

strongest couplings occur at different resonances with the plasmon frequencies strongly

dependent on particle shape. For spherical particles, the dipolar resonance is triply

degenerate, and coupling of incident radiation to this resonance promotes SERS. As the

shape becomes elongated, however, the dipolar modes split into longitudinal and transverse

components. The longitudinal component red shifts with the increasing aspect ratio of the

particle; SERS occurs through coupling of the incident wavelength with this component. The

doubly degenerate transverse component remains at a frequency near that of the original

sphere and does not contribute to the SERS effect (Blatchford, 1982). Thus, the plasmon

frequencies of spheroids are at longer wavelengths than spheres. Although originally

described for the unbound aggregates of colloidal particles, this is also an apt description of

the enhancement source for bound hemispheroids.

The Gersten and Nitzan theory (1980) which models the features of a moderately

rough surface forms the basis for the later EM theories. Its constraints include: (1) the

description of a rough surface as a series of regular protrusions rising from a plane (Figure

2.4), (2) the plane is perfectly flat and perfectly conducting; (3) the protrusions, prolate

hemispheroids, can be described by the aspect ratio a/b; (4) the complex dielectric constant

of the particle is modified by the aspect ratio; (5) the molecule is a point-dipole located on the

symmetry axis at a distance d from the particle tip; (6) the molecular dipole is oriented parallel

to the symmetry axis of the particle; (7) the dimensions of the system are small, with a and

d less than two percent of the exciting wavelength; being within the Rayleigh limit, the

particle therefore experiences the electric field as homogenous (van de Hulst, 1981);


} d Bo

Figure 2.4 Isolated prolate hemispheroid protruding from flat plane as described in Gersten
and Nitzan spheroidal electromagnetic theory (1980).

only the surface plasmons are excited, and Rayleigh scattering predominates; and, (8) the

incident radiation is directed parallel to the symmetry axis. With these conditions fulfilled, the

molecule experiences the maximum field. If the molecule is not in contact with the surface,

the experienced field decreases with increasing distance between molecule and surface, as a

function of particle shape (Gersten and Nitzan, 1980).

Despite its limitations, the spheroidal EM theory (Gersten and Nitzan, 1980) is useful

for determining general trends as shown by Aroca and Martin (1985), who used it to ascertain


EFs for a number of metals having various aspect ratios. Of significance is the suggestion that

a system can be tuned for optimum performance by careful selection of the roughness features

and/or incident wavelength. (Initially, the matching of exciting frequency to plasmon

frequency was not predicted to be too stringent a requirement as bulk metal plasmon

resonances are quite broad--FWHM = 0.5 eV; Moskovits, 1985). As expected, maximum

EFs were predicted for silver, gold, and copper, possessors of the most favorable dielectric

properties. Additionally, the effect of particle shape on the exciting frequency and

enhancement factor was demonstrated to agree with experimental observations from silver

and gold particles. The field enhancements calculated using this theory were, however, often

larger than the observed values by four or five orders of magnitude. For a molecule near a

silver particle where a/b = 50/10, an enhancement of 10" was calculated, yet only 106 was


Electrodynamic theories. To improve the agreement between predicted and observed

enhancements, the spheroidal model (Gersten and Nitzan, 1982) was modified to include two

of the main factors limiting the observed responses, surrounding medium and particle size.

These had been previously addressed by Gersten and Nitzan (1982) as significant, but had

been excluded from the model for simplicity.

Starting with the surrounding medium, a brief explanation of each effect, its

significance, and pertinent references will be given. For any system not maintained under

vacuum, the surrounding medium through which the electric field is projected may red shift

the resonant plasmon frequency. The extent of the shift is a function of the medium dielectric


constant and may have a deleterious effect on the enhancement, especially if the incident

wavelength cannot be adjusted (Figure 2.5; Barber, et al., 1983a).

Theoretical upper and lower size limits were computed by Wokaun, et al. (1982) and

agreed well with those determined experimentally. For large particles, the upper boundary

is reached as the Rayleigh limit is surpassed (a > 0.02 Xo; Barber et al., 1983a, 1983b; Kraus

and Schatz, 1983a). At first, a "dynamic" depolarization briefly increases the enhancement

and red shifts the resonance. This is quickly followed by expansion of the dipolar plasmons

into multipolars; subsequent broadening of the resonance frequency results in a weakened

field enhancement. Radiational self-interference has been observed for Ag spheres larger than

30 nm (Furtak, 1983).

The lower limit is approached as the particle becomes smaller than the mean free path

of the electrons in the metal (Kraus and Schatz, 1983a, 1983b). Surface scattering of the

conduction electrons then occurs, surface plasmons are not sustained, and the field

enhancement decreases. Silver has one of the longest mean free paths of any metal; scatter

effects have been observed from particles larger than 10 nm (Creighton and Eadon, 1991).

Wokaun, et al. (1982) showed that when boundary conditions are considered, a maximum

field enhancement can be determined for a particle of given size and shape. One of the major

implications of the size effects is that an optimum size exists for any given shape.

Consequently, particle morphology can, theoretically, be optimized to obtain a desired


Incorporation of size and medium considerations into the EM model comprise the

electrodynamic (ED) theory (Cline et al.,1986). Cline showed that inclusion of these

30 -

25 -


15 -




400 500 600 700


Incident Wavelength (nm)

Figure 2.5 Effect of surrounding medium on the EM field at the surface of an
isolated spheroid (a/b = 2; a = 100 nm; Barber, et al., 1983a).







corrections decreased the discrepancies between calculated and observed EFs to a difference

of three orders of magnitude for silver particles of a 2:1 aspect ratio.

Corrected electrodynamic theory. In the last major theory renovation, Zeman and Schatz

(1984, 1987) further amended the ED model by including corrections for molecule position,

particle-particle interactions, and adsorbate effects. These will be considered separately. (1)

Not every molecule is located at the tip of a particle; therefore, not every molecule

experiences the maximum field. The random placement of molecules on a surface features

is better depicted by averaging calculated EFs for a molecule in a variety of positions (Barber,

et al., 1983a, 1983b). (2) Silver features are not perfectly isolated. Fields of neighboring

particles may influence the field of another particle to red shift (Laor and Schatz, 1981, 1982).

(3) A mono- or submonolayer is sufficient to modify the dielectric constant of the

surrounding medium. This, in turn affects the shift of the plasmon resonance induced by the

medium from the behavior predicted in the ED theory. In response to some species, the

modified shift becomes excessive and coupling between the plasmon and incident wavelength

is disrupted.

Using the corrected electrodynamic (CED) model, the size and shape dependencies

of the incident field amplification and the SERS enhancement for a number of metals,

including silver, were calculated (Figure 2.6). Enhancements were noted to be larger for

spheroids than for spheres of similar dimension. Theoretically, the CED model permits

surface features to be tuned to optimize the EF for a given set of conditions in a manner

similar to tuning a laser for SERRS. The predicted trends are in good agreement with those

observed experimentally, although the theoretical enhancements tend to be lower than those




o a


.il )
a '

"a oa io

o o=
DO *

o 0 0 0 0 0 0
o tr> 0 >rl 0 tf
e'm -


o 0 0 0 0^
o 0 0 0

(ioO x) luou~i9U1qug 8JS




1 \ -- C/
Ill 0
\ -: 1 \ \

^t jn^ I0
1 ^ ^^ ---

0 0 0 0 0 0
0 0 0 0 0 0
M ('N -< \o i Tfm

uoWv3oildmuV Plpij umwnido

(IuU) qlpu3I3ABA luop!3uI

uoiltogidfV pl!aij oupoaIg


observed by about two orders of magnitude. It is at this point that chemical theories are

suggested to answer the discrepancy between predicted and calculated values.

Electromagnetic models are the most complete and best understood of the theories,

yet there are observations that they fail to explain which point to the presence of an

interactive mechanism: (1) The SER effect is not universal. The magnitude of observed

signal enhancement varies with molecular species. (2) Within a single Raman active molecule,

different active bands are enhanced to different degrees. (3) Relative enhancements often

follow known/probable adsorption geometries. (4) The observed enhancement is often

modified by conditions that should not affect a strictly electromagnetic interaction.

Chemical Theories

Chemical theories, accounting for local or short range effects, are usually credited

with increasing the signal by one or two orders of magnitude. The primary models involve

active sites on the metallic substrate or some type of charge-transfer process. Each requires

direct interaction (physisorption or chemisorption) between the molecule and the metal

surface and results in a true modification of the molecular cross section albeit, of a molecule-

metal complex. Otto (1991) published a comprehensive review of the chemical mechanisms

contributing to the SER effect.

In the active site model (Pettenkofer, et al., 1985), molecules preferentially adsorb to

specific sites (defects or atomic scale roughnesses) on the surface of the metal and exhibit a

greater activity than usual. These sites have been proven to be real physical sites (Sobocinski

and Pemberton, 1988) and have been described by Pettinger (1987). Small enhancements,

up to 65-fold have been achieved for pyridine adsorbed onto defects located on a smooth


metal surface (Jiang and Campion, 1987). Further enhancement was expected with the

addition of microscopic surface features.

The charge-transfer model (Adrian, 1982; Lippitsch, 1984) describes the movement

of an electron between a molecular orbital and the metallic conduction band. Although an

electron may transfer in either direction, the usual path is from the metal to an excited

molecular electronic state by tunnelling (physisorption) or hybridization chemisorptionn). The

molecule is left in a higher vibrational state when the electron returns to the conduction band

to recombine with an available hole. Non-resonant recombination (where the energy

difference between the excited molecular level and the Fermi level is not of the order of the

incident energy) results in photon emission contributing to the background continuum

(Lasema, 1993) seen in SERS spectra. Not surprisingly, many molecules exhibiting a strong

SERS effect have either lone pair electrons and/or n transition orbitals which can form

complexes with noble and transition metals. In each of these models, the molecule itself is

changed in a way that modifies the effective Raman cross section of the analyte species.

SERS is a product of both electromagnetic and chemical mechanisms as indicated by

its dependence on local field effects and on the molecular species involved. These theories

have been successfully combined to describe the phenomena associated with SERS with EM

theories describing the mechanism by which a major portion of the enhancement is produced

and chemical mechanisms answering specific questions regarding the nonuniversality of the

effect, the existence and orientation of molecule-surface complexation, and the effect of

specific molecular characteristics on the enhancement of certain modes over others.

Selection Rules / Practical Experimental Considerations

Despite completeness of the SER theories, there is still difficulty reconciling

variabilities within the prescribed behavioral trends. Specifically, CR processes follow well-

defined selection rules. SER processes do not. Common spectral differences between CR

and SER can be used to clarify differences in the selection rules. Although good agreement

is generally found between SER and CR spectra, SER peaks are broader and frequently

shifted compared to CR features (Laserna, 1993). The relative intensities among detected

peaks are often altered. And, SER spectra frequently contain modes forbidden by CR

selection rules.

Three main factors (Hallmark and Campion, 1986) influence the presence and intensity

of a surface-enhancing spectral mode: image dipole, symmetry, and quadrupole polarizability.

The image dipole is a composite of the dipole induced in the conduction electrons of the

metal surface by a nearby molecular dipole and that molecular dipole itself If these dipoles

are normal to the surface, the effect will be reinforced; if parallel, the effect of the molecule

will be screened by the response of the conduction electrons (Figure 2.7). Because metals are

less reflective in the visible than in the IR, the dipole will be only partially reinforced or

partially screened under common SER excitation conditions. This factor strongly affects the

relative enhancement of modes in a molecular spectrumand may be used to indicate which

modes will appear. Of the three factors listed, it is usually dominant.

Symmetry reduction occurs when the presence of a plane near a molecule removes a

plane of symmetry from the point group of the molecule. Formerly forbidden modes may then

Perpendicular: z Parallel:

Reinforced: Shielded:
Ideally, the total Ideally, the net
effective strength transition dipole
increases by a factor moment is zero.
of two.

Figure 2.7 Effect of the molecular dipole orientation on the induced dipole in the metallic
surface. For simplicity, z, the axis of symmetry, is aligned with the incident radiation, normal
to the metallic surface.

appear and the frequency of allowed modes may shift. This effect is further modified with

the absorption or complexation of the molecular species.

Sometimes a negligible effect, the quadrupole polarizability is surface dependent and

becomes dominant only when very large gradients are present in the electric field immediately

adjacent to the surface. The largest gradients occur for surfaces having a high radius of

curvature, such as the so-called "needles" that are sometimes vapor deposited onto crossed

grating type substrates. This effect also allows forbidden modes.

Although the relative importance of these factors is highly dependent on the system

under consideration, each is represented within a single rule: Scatter depends on the


orientation of the mode with respect to the surface and, to a lesser extent, the incident

radiation; therefore, the strongest modes are oriented normal to the surface and the weakest,

if detected, parallel. This "propensity" rule denotes the principles to which most SER

behaviors adhere. In addition to predicting and explaining the presence of CR forbidden

modes, probable molecular orientations based on the relative intensities of spectral modes are

also suggested. The propensity rule is sufficiently vague to allow variation in the

experimental system.

While SERS will never be an all-purpose analysis and detection method due to the

extreme dependence of the response on specific system characteristics, the development of

reproducible substrates purchased ready to use or easily prepared with little time, labour, or

equipment will strengthen the position of SERS in the fields it has already been found useful.

It is for this reason and with these considerations in mind that we looked to commercially

available substrates.



SERS from TLC Materials

Thin-layer chromatography (TLC) is a simple and inexpensive technique for

separating nonvolatile organic compounds. Despite being an older method, TLC is still

considered ideal for the separation of natural products in fields such as pharmaceuticals,

cosmetics, and petrochemicals, and is adept at handling large numbers of these samples for

routine analysis. A long term weakness of TLC has been detection of the compounds after

separation. Traditional methods, based on comparative procedures, such as retardation

factors and colour reactions, are inherently nonspecific. Optical techniques (UV-vis

absorbance, fluorescence, and IR methods) give more information but are limited by the

spectral properties of the molecules and, for IR, by severe interference from both moisture

and certain adsorbent materials common to TLC plates.

Although Raman spectroscopy was a recognized identification technique known to

be insensitive to typical adsorbent materials, CR was not initially considered for in situ TLC

detection due to the high analyte amounts (gg) and laser powers (0.2 W) required to

overcome the phenomenon's weak cross section (Gomez-Taylor et al., 1976). In the first

reported use of an enhanced Raman technique to identify separated spots from a solid

chromatographic material (Tran, 1984b), three dyes (crystal violet, malachite green, and basic

fuchsin), chosen for their similarities in structure, absorptive properties, and low fluorescence

yields, were separated by paper chromatography. After drying, the substrate was saturated

with silver hydrosols sprayed from an atomizer, and SERR was observed from the wet paper.

The LOD was determined to be 2 ng/cm2 in the spot or 6.4 finol in the laser. SER and SERR

spectroscopies have since detected dyes (Soper et al., 1990; Rau, 1993), purine derivatives

(Sequaris and Koglin, 1987), nitro-polynuclear aromatics (Koglin, 1988), and surfactants

(Koglin, 1993) from TLC plates. For a majority of samples, the plates were never developed,

being treated, instead, as any other SER substrate. In one unusual application, the effluent

of an HPLC system was collected onto a TLC plate for later analysis by SERS (Soper et al.,


Silica Gel 60 high performance plates have been used more often than any other type

of TLC plate for SER applications. The adsorbent layer is somewhat thinner (200 pm) than

on a regular plate (250 pm) and the irregularly shaped silica particles are more narrowly

distributed around the 5 im average diameter. The pore size of the silica gel is 60 A. The

plate support is typically glass.

The only metal, thus far, applied to TLC materials for the purpose of surface-

enhancement has been silver in the form of colloids, predominately from the Creighton

(Creighton et al.,1979) and citrate (Freeman et al., 1988) procedures. Hydrosols were either

sprayed onto the plates or the plates were dipped into a silver nitrate solution and

subsequently sprayed with a reducing agent for in situ formation of the particles. Use of

either method has been found to leave unstable colloids susceptible to further coagulation.

Analyte-induced aggregation of plate-applied colloids has been observed at separated

spots as a change in colour (Sequaris and Koglin, 1987; Soper et al., 1990), colour being

dependent on the metallic geometry. Because the extent of aggregation in a particular spot

was a function of the nature and concentration of the analyte in that spot, spots differing by

concentration or species typically possessed colloids of different average sizes and shapes.

When this occurred, comparisons between relative SER responses could not be made and

quantitation was not possible; the enhanced field at each spot differed from the others on the

plate. For all molecules to experience the same enhancement-inducing field, metal particles

need to be uniform in morphology and evenly distributed across the surface. Due in part to

the instability problems of silver colloids on TLC materials, SERS has not yet been reported

from spots separated in the presence of silver particles. Neither has SERS been reported for

analytes applied to a TLC plate previously prepared with silver. Colloids have always been

applied with or after the analyte.

Scatter from TLC materials has been primarily used for purposes of identification

rather than quantitation. While the detection limits have stressed the small amounts of sample

needed for identification, values relating to the linearity or precision of spectra obtained from

these substrates have seldom been reported. Quantitation and its corresponding figures of

merit have been limited by the irreproducible nature of the colloidal particles on the surface

(Cook et al., 1993), the nonuniform distribution of the molecules within the spots

(Sutherland and Winefordner, 1991), and by the small percentage of analyte molecules

appearing in the probe beam compared to the total number applied on the surface (diameter

of a sample spot, mm, compared to probe beam, pm). Qualitatively, however, the spectra

have been in good agreement with SER spectra from other substrates. Advantages of using

SERS have included the application of minute amounts of sample, the identification of

analytes, and the lack of interference by solvent, adsorbent, and fluorescence bands.

Description of TLC Plates

For this project, a material close in nature to the silica gel plates from previous studies

has been chosen. Bakerflex Silica Gel IB2 (BAK; J.T. Baker Chemical Co., Philipsburg, NJ)

is composed of a 200-pm layer of irregularly shaped Silica Gel 60 particles (2-20 um average

diameter) bound to a flexible polyester support by inert gypsum. Retention of compounds

on the silica gel occurs by the activity of silanol (Si-OH), siloxy anion (Si-O-), and silyl ether

(Si-O-Si) functional groups. The main difference between this plate and those used in the

past is that the BAK plate has a flexible polymeric backing which simplifies handling. Silver

was applied to the plate by room temperature chemical reduction of a silver salt.

For comparison, argentation plates (ARG) were purchased from Analtech (Newark,

Delaware). These consisted of a 250-pm layer of irregularly shaped silica gel particles

(Adsorbasil Plus; 6-13 pm diameter) held onto a glass support by a silicon dioxide/aluminum

oxide binder. Separation on the silica gel was modified by silver ions, added by soaking the

plate in a methanolic solution of silver nitrate, to enhance the retention of olefinic derivatives.

During plate development, the solvated ions form rapidly reversible complexes with accessible

olefinic double bonds, causing those substances to be retained more strongly in the stationary

phase (Snyder, 1975). Ultimately, retention is based on the cis-trans configuration about the

double bond, as well as, on the degree of molecular saturation. Neither plate had been

treated with fluorescent indicator.

TLC Application Study

Purpose. In this project, SERS was used to detect small amounts of three structurally similar

pharmaceuticals after development on a TLC plate. Both qualitative and quantitative aspects

of SERS as an in situ detection method for TLC were considered. The stability of silver

particles embedded in the stationary phase was also surveyed as was the effect of the presence

of silver particles on the separation of the three drugs.

Sulfonamides--background. The sulfonamide family consists of more than 5500 individual

compounds of which thirty or so exhibit sufficient therapeutic effect to be recognized for

practical application in human and veterinary medicine, including the prevention of bacterial

infection among livestock and poultry (Bevill, 1984; Eli Lilly, 1951). Sulfadiazine (SDZ),

sulfamerazine (SMR), and sulfamethazine (SMT) are members of the sulfapyrimidine

subfamily (Figure 3.1). Compared to other sulfa drugs, sulfapyrimidines demonstrate a higher

antibacterial activity, lower host toxicity, and an improved absorption rate after oral and

parenteral administration (Eli Lilly, 1951). Although their current use in human medicine has

declined due to an increase in bacterial resistance and the production of more effective agents,

their application in animal husbandry has continued since they are highly effective, relatively

inexpensive, and easily administered via food or water. SDZ, SMR, and SMT are three of

the thirteen most frequently used sulfonamides throughout the world (Mooser and Koch,


Because sulfa drugs have a limited solubility in aqueous systems, life threatening side

effects may occur if the drugs crystallize within the kidneys or bladder (crystalluria). For this

reason, the dosage level of the individual drug is limited. Greatest medical benefit, however,




0 c

1 I

o z

0 Cu

( C '
O .c

or '


is derived when a high internal concentration is maintained. Because each compound acts

independently within the body regarding penetration, transport, and excretion, the

administration of two or more agents in a mixed therapy (approved by the American Medical

Association in 1949) allows a higher therapeutic concentration to be maintained while side

effects are minimized.

Outside of direct drug therapy, the incidence of crystalluria is unusual; however, the

hyper-sensitive (allergic) response of certain persons (Sophian et al., 1952; Steele and Beran,

1984; Bevill, 1984), and the suspected carcinogenic properties of SMT (Sundlof et al., 1988),

combined with the widespread use of these agents among livestock prompted concern over

residues in consumable pork, milk and poultry products. In 1973, the United States Food and

Drug Administration (FDA) set a tolerance of 100 ng sulfonamide per gram of uncooked

edible tissue (Code of Federal Regulations, 1973). Random assays for drug and agricultural

residues have since been made monthly in all commercial U.S. packing plants that handle

meats destined for human consumption (Bevill, 1984). From a summary of studies filed in

1986 and 1987, the FSIS (Food Safety and Inspection Service) reported that over 10% of

slaughtered swine contained residue levels exceeding the tolerance (Van Dresser and Wilcke,

1989). By 1991, this level had dropped to -0.3% of samples taken in the spot checks (FSIS,

1991). Despite the significant improvement in violative rates, a more comprehensive

monitoring regimen is desired to ensure that a truly representative sampling of animals is


Sulfonamides--analysis. The FDA and FSIS relied until recently on the Bratton-Marshall

colorimetric test (Bratton and Marshall, 1939) for the analysis of sulfonamides in meat and

dairy products. The test was nonselective and marginally sensitive. In the procedure, a

derivatized product formed a purple compound absorbing at 545 nm. Because all

sulfonamides form compounds absorbing near 545 nm, the method could not indicate which

drugs were or were not present, but gave, instead, a single response for the total

concentration, even though the tolerance had been specified for the individual compound.

The limit of reliable measurement was also the same order of magnitude as the tolerance the

method was expected to enforce (Bevill, 1984). Insensitivity, however, was a common

problem of the assay methods used into the 1980's (Sato. et al., 1980; Horwitz, 1981a and b).

Currently the AOAC (Association of Official Analytical Chemists) endorses several

methods useful down to the 0.050 ppm level (Unruh et al., 1991), but these are typically

expensive, time consuming, and materially wasteful. Methods of current interest include the

photochemically induced fluorescence of derivatized SMT after FIA (Mahedro and Aaron,

1992) which cannot discern SMT from the structurally similar SMR and SDZ, liquid

chromatography (LC) with post-column derivatization (Bui, 1993) which has recovered SMR

and SDZ from animal tissues, GC/MS (Mooser and Koch, 1993) which has studied all three

agents with good results (LOD = 20-50 ppb), and others, mostly variations of these HPLC

or GC methods. Considering the sheer volume of samples that are scheduled for testing and

the extensive matrix clean-up required for LC and especially for GC/MS, neither of these

methods are practical for the type of monitoring the AOAC wishes to do.

A fast, sensitive (LOD= 0.25 ppb) method was described by Unruh, et al (1991),

utilizing TLC with detection of a fluorescing derivative. TLC methods meet most of the

requirements for routine use: They are simple and inexpensive with less sample preparation


than most of the other methods, and TLC is suitable for routine international application. The

main issue of concern is the limited nature of its detection methods.

Earlier work performed in this laboratory (Sutherland et al., 1990) identified and

quantitated SDZ, SMR, and SMT in aqueous colloidal solutions by SERS with detection

limits between 1 and 10 ng/mL. Although these were laboratory standards that did not

require matrix clean-up or separation steps, SERS was shown to be a potentially sensitive

detection technique. From this point, references to sulfa drugs or sulfonamides will refer

exclusively to SDZ, SMR, and SMT.

General Application Study

The potential of silver-treated TLC materials as routine SER substrates was evaluated.

Figures of merit, such as the limit of detection and linear dynamic range, were determined

using crystal violet as the model molecule. The reproducibility of responses from surfaces

prepared on different days was assessed. The effect of changing the dielectric constant by

surface dampening was also considered. Spectra were collected from several analytes having

a variety of functional groups.



Spectra were collected with a 0.85 m double grating spectrometer (Spex Industries,

model 1403). The spectral resolution was 10 cm1 Excitation was provided by the 514.5 nm

line of an argon ion laser (Spectra-Physics, Series 2000). At the sample surface, the laser

power ranged between 2.5 and 30 mW depending on the sensitivity of the analyte to high


power densities. Specific values are included in the text where appropriate. Scatter collection

generally utilized right angle geometry from the illumination chamber and backscattering

geometry from the Raman microprobe (SPEX micramate), a Zeiss 20 research grade

microscope (with 10x and 40x objectives) modified to pass excitation and collect scatter. All

SER and liquid CR spectra were collected from the illumination chamber. CR spectra from

solid powdered analytes were obtained using the microprobe. Signals were detected by a

cooled gallium arsenide photomultiplier tube (RCA, model C31034) coupled to standard

photon counting electronics. All spectra were obtained from single scans without smoothing.

For visual characterization of the submicrometer surface features, photographic images were

obtained by a JEOL JSM-35C scanning electron microscope.

Chemicals and Materials

All solvents were spectroscopic or HPLC grade. For the TLC study, Dye Mixture IV

(Analtech), containing Sudan Orange G, Sudan II, Solvent Green 3, Solvent Blue 35, and Fat

Red 7B in toluene, was used as obtained. Sulfadiazine (4-amino-N-2-pyrimidinyl-

benzenesulfonamide, CAS 68-35-9), sulfamerazine (4-amino-N-[4-methyl-2-pyrimidinyl]-

benzenesulfonamide, CAS 127-79-7), and sulfamethazine (4-amino-N-[4,6-dimethyl-2-

pyrimidinyl]-benzenesulfonamide, CAS 57-68-1) were purchased from Sigma Chemical Co.

(St. Louis, MI) and used without further purification. For the general application study,

crystal violet (95%; Aldrich Chemical Co., Milwaukee, WI) was selected as the model

compound. Spectra from 9-aminoacridine hydrochloride (Aldrich, 98%), benzoic acid

(Sigma), p-aminobenzoic acid (Eastman-Kodak; Rochester, NY), p-nitrobenzoic acid

(Eastman-Kodak), pyrene (Aldrich, 99+%), 1-aminopyrene (Aldrich, 97%), and 1-nitropyrene


(Aldrich, 97%) were included to provide a variety of functional groups covering a range of

interests. Analyte solutions were prepared in ethanol unless otherwise noted.


Brashear reduction. The BAK material was coated with a discontinuous layer of silver by the

Brashear process, one of three room-temperature chemical reduction methods considered for

this project. The Brashear method (Boo et al., 1985) was chosen over the Rochelle salt and

Tollen's procedures (Ni and Cotton, 1986) for its simplicity, minimal time requirements, and

for the excellent front reflecting properties of the silver layers it produces (Setapen, 1940).

The Brashear procedure described by Boo et al. (1985) for glass microscope slides was

generally followed.

Two grams of silver nitrate were dissolved into 300 mL of water. Approximately 2

mL of this solution were removed and set aside. About 45 mL of 0.7 M ammonium

hydroxide were added in small portions to the bulk silver solution. At the first addition, a

reddish-brown precipitate formed. With further additions the precipitate darkened, then

began to redissolve. From this point, base was added dropwise until the precipitate just

redissolved and the solution was once again clear. The endpoint was checked by replacing

the 2 mL of silver nitrate solution removed earlier, whereupon, formation of a small black

precipitate indicated that no reactive excess of ammonia remained in the solution; the

precipitate itself had no influence on the activity of the Brashear reagent. The volume was

brought up to 500 mL in water. Surfaces prepared from silver solutions between the ages of

one and fourteen days showed no significant variation in appearance or performance. As the

Brashear reagent aged past this point, the reduction was observed to slow and the reagent


took on a grayish cast. To slow aging effects, the ammoniacal silver solution was stored at


The reducing agent was prepared by dissolving 10.8 g of dextrose into 500 mL of

water. Before its first use, dissolved gases were driven out of the solution by allowing it to

sit at room temperature for 48 hr (Setapen, 1940). After activation, it was quite stable and

could be stored indefinitely under normal laboratory conditions.

The silver nitrate and reducing solutions were mixed 2:1 by volume in the presence

of the substrate to be coated, the substrate having been saturated in the metal salt solution

prior to addition of the reducer. During deposition, the solutions were slowly stirred by

magnetic stirrer to maintain homogeneity. Factors affecting the particle size and distribution

included the concentrations of the reducing media and the temperature (23.0 0.05 C) which

were held constant, and the reaction time which was varied between 1 and 6 min. On meeting

the specified exposure time, the substrate was removed from solution, rinsed in a large

volume of water, and air-dried. A BAK plate with a 3 min exposure to the mixed Brashear

reagents would then be designated BAK(Br-3).

TLC separation of standard dye mixture. BAK and ARG plates were activated at 110 *C for

30 and 60 min, respectively. Analtech Dye Mixture IV was applied to the plates in volumes

of 1 utL by a calibrated capillary. The chamber was saturated with toluene for 30 min before

separation began. The plate was developed for 10 min. After drying, the coloured spots were

detected visually. All separations were reported in terms of the retardation factor (R), a ratio

of the distance traveled by the spot to the distance traveled by the solvent front. Both

distances were measured from the origin.

TLC separation of sulfonamide mixture. Before sample loading, BAK and ARG plates were

activated at 110 C for 30 and 60 min, respectively. SDZ, SMR, and SMT (10-2 M in

acetone) were each applied in volumes of 1 pL by calibrated capillary. Individual standards

were developed on the same plate with the mixtures. The chamber was saturated for 30 min

before separation in the mobile phase, ethyl acetate/methanol/ammonium hydroxide

(85:15:0.6 v/v; Wooley, Jr. et al., 1980). Development times varied slightly with the

substrate, 45 min for BAK and 40 min for ARG. Because the sulfapyrimidines are colourless

and do not fluoresce naturally, the dried plates were dipped into fluorescein (30 mg/150 mL

acetone) and examined under UV. The effect was to enhance the fluorescence of the

background so that dark purplish spots appeared on a lighter green plate. For greater

contrast, ARG plates were heated for 30 min at 110 C after treatment with fluorescein.

Detection of separated sulfonamides. After separation, portions of the plates containing the

spots of interest were sectioned into squares (-1.5 x 1.5 mm) and inserted into the Raman

sample chamber. Spectra were collected from the TLC plates in both dry and liquid state

modes. In the dry state mode, the plate section was admitted into the sample holder and a

scan was taken. In the liquid state mode, the segment was placed into the holder and the

surface was saturated with a small volume (5 pL) of water. A cover glass, held by a tension

spring, was positioned over the sample to prevent drying during the scan. Spectra were

obtained through the cover glass. Fluorescein was not applied to plates inspected by SER

General applications study. When determining figures of merit, a 0.5 uL aliquot of each of

a series of ethanolic crystal violet (CV) standards (ranging in concentration from 0.01 to 1000

ng/pL) was delivered by microsyringe to the TLC plate segments. The sections were dried


under a stream of dry nitrogen before analysis. FOM were determined from dry and liquid

state substrates.

As has been noted by Banes (1969), only the best results are published, often after

painstaking manipulations impractical to reproduce on a routine basis in a service-oriented

facility. With this in mind, the FOM study was conducted in two parts. Each CV standard

was applied and analyzed from six BAK and six ARG substrates. At random, three substrates

of each six were selected to be prepared and analyzed as carefully as possible. The remaining

three were examined following good laboratory practice, but in a routine manner. In the

results section, the LOD and LDR information was derived wholly from the "carefully"

prepared surfaces. The repeatability (%RSD) was, however, determined for precisions both

"carefully" and "routinely" obtained. These values were listed side-by-side for comparison.

When obtaining general spectra from a variety of samples, analyte was applied by

immersion of the substrate into a 103 M sample solution for 20 s with subsequent drying

under a nitrogen stream. Weitz et al. (1983) had earlier demonstrated that these dipping

conditions would provide a sufficient number of molecules to saturate the active sites of the

substrate surface, approximating one monolayer. Each spectrum was obtained from a single

scan without smoothing. No more than three spectra were collected for any analyte-substrate

combination, so that the spectra included in this chapter would be representative of spectra

obtained during routine use.

Results and Discussion

TLC Study

Effect of silver on plate development. Before separating the sulfonamides, the effect of the

silver on a standard dye mixture was examined to gain a better understanding of the activity

of Ag particles in the plates. The mixture, composed of six strongly coloured dyes (structures,

Figure 3.2), was designed for reproducible separation on silica gel with a non-polar mobile

phase. Although changes in adsorbent have been known to alter the elution sequence of

compounds (Bieganowska et al, 1993), this phenomenon was not observed when comparing

the bare and coated BAK plates. Indeed, the presence of metallic silver had less effect on

development than had been anticipated (Table 3.1). Reported Rf values were averaged from

three independent experiments with the %RSD < 10%, as predicted for careful manual

analysis (Ripphahn and Halpaap, 1975). Retentions on all plates, coated and bare, were

exceptionally close, typically within the uncertainty of the BAK(Bare) values.

Table 3.1
The effect of the presence of silver on the separation of a standard dye mixture.

Retardation Factors (B._)
Sudan Solvent Sudan II Solvent Sudan Fat Red
Orange G Blue 35 Green 3 Orange G 7B
(orange) (blue) (red) (green) (yellow) (purple)

Bare 0.05 0.24 0.35 0.48 0.63 0.74
Br-3a 0.04 0.23 0.33 0.46 -- 0.72

20% 0.06 0.22 0.39 0.49 -- 0.66
aExposed to Brashear reagents for 3 min.






M=0 a
^ /
\ j-





z *



0 -




From each plate containing silver, the Rf (Table 3.1) of the second Sudan Orange G

dye was undetermined. The most probable explanation is that the yellow colour of the spot

could not be discerned against the dark background. Despite examination under UV light and

treatment with fluorescein, the response remained ambiguous. It is also possible that the two

Sudan Orange G spots resulted from reversible isomers and that, in the presence of silver, one

was strongly preferred over the other. It is known that naturally and synthetically derived

dyes seldom have purities exceeding 94-96% since dyes are rarely single compounds, but are

more often a series of stable structures or unstable isomers.

It was surprising that the two coated plates behaved so similarly, since the silver

particles impregnating the BAK plate were initially deposited in metallic form rather than ionic

and were expected to be considerably smaller than the particles formed from the ions

dispersed throughout the ARG plate. Differences in silver quantity were expected because

the solutions used to prepare the plates varied widely in silver content. The ARG plate was

prepared from a 20% AgNO, solution, the BAK from a 0.4% solution (before addition of the

reducing agent). At purchase, ARG plates were white in colour, but darkened to black on

exposure to light. Freshly coated BAK materials were also white in colour, but darkened to

a rust brown, the lighter colour indicating a lesser quantity of silver. Both coatings took

about forty-eight hours to fully develop.

Separation of sulfonamides. The sulfonamides separated quickly and efficiently (Table 3.2)

with the larger, more polar compounds, containing hydrophobic methyl groups on the

pyrimidine ring, less well retained. For three independent experiments, the precision was <

10%. The diameter of each separated spot was -4 mm. The presence of silver apparently


increased the activity of the stationary phase which resulted in lower Rf values, when

compared to the BAK(Bare). This trend was seen to a slight extent in Table 3.1. With the

sulfonamide separation, however, the retardation factors from the silvered BAK plate were

considerably different from the bare BAK values. This, and the increased retention seen on

the ARG material, was in contrast to the behavior previously observed with the test dyes.

Table 3.2
The effect of silver on the separation of sulfonamides.

An explanation, sought in the structural differences of the analytes, was found in the

sulfur content of the drugs. Silver is known to bond quickly and strongly with the sulfur

atoms of adjacent compounds to form a variety of products (Zickrick, 1940). The

BAK(Bare) plates lacked silver; thus, the sulfa drugs interacted with only the silica-based

functional groups and were less strongly retained than were the drugs on the silver-coated

plates. The marked difference in behavior of the BAK(Br-3) and ARG plates, characterized

by an increase in the retention, would then be due to the larger quantity of silver present in

the ARG plates. The decreased mobility of SDZ compared to SMR and to SMT on each

Retardation Factors .)

Bare 0.35 0.44 0.54
Br-3 0.25 0.35 0.46

20% 0.12 0.20 0.29


coated plate could be explained by the molecule's less polar nature and stronger interaction

of silver with the less hindered sulfur atom; SDZ has no methyl groups on the pyrimidine ring.

Unlike the components of the standard dye mixture which were brightly coloured, the

sulfonamides were colourless. Thus, the spots, in addition to the spot channels, were

examined for variations in silver particle size, as indicated by changes in plate colour. If the

silver particles were unstable and joined in further aggregation, one would expect to see

darker streaks along the path passed by the analyte or in the spots on the brown BAK plate.

On the ARG plate, which is already black, flecks of silvering in channel path or spot would

be expected. The plates, however, appeared uniform: Once dried, neither the positions of

the solvent front nor the sample spots could be distinguished from the rest of the plate,

indicating that (1) the silver particles were immobile and did not move with either the solvent

front or the analyte in spite of the silver-sulfur attractions, and that (2) no large scale

aggregations of silver occurred in response to the analyte presence.

Changing the reduction time and, therefore, the quantity of silver on the BAK surface

affected the separation (Table 3.3). As silver content increased, all analytes showed a greater

affinity for the stationary phase, a phenomenon discussed previously. The presence of silver

had the largest effect on the drug which was least attracted to bare silica gel. Increasing the

amount of silver in the plate increased the retention of SMT to a greater extent than the

retention of SDZ which was already strongly retained, possibly as steric hindrances due to the

bulkier ring structure were overcome by the greater availability of silver. As predicted by

these observations, SMR gave an intermediate response.

Table 3.3
The effect of increasing amounts of silver on plate activity.

Retardation Factors (-3)

Bare 0.33 0.44 0.54
Br-1 0.32 0.38 0.46
Br-2 0.27 0.35 0.44
Br-3 0.25 0.35 0.46
Br-4 0.27 0.34 0.42
Br-5 0.23 0.32 0.41
Br-6 0.25 0.32 0.39

In-situ detection of sulfonamides. SER spectra were collected from ARG and BAK(Br-3)

plates, where selection of the 3 min coating was based on separation performance. In the

initial trials, no SER response was detected from separated drugs on either plate. Elimination

of the separation step did not improve the response. In both cases, small amounts of graphitic

carbon were detected which indicated a photothermal degradation of the analytes. The laser

power had been 27 mW at the sample surface. Although SER spectra of these same

compounds had been collected in an earlier study (Sutherland et al, 1990) using powers of

100 mW, the enhancing surface had consisted of colloids suspended in aqueous solution,

where the thermally conductive solvent minimized damage of the type detected from the dry

TLC plates. Even so, a high background between 1200 and 1700 cm' had been noted by

Sutherland for spectra of small amounts of drug in solution.

CR spectra of the pure crystalline solids (Figure 3.3) were obtained with a power of

27 mW at the sample surface. The fact that no damage was detected in the spectra could be

due to less intense fields interacting with the molecules in the absence of the enhancing silver











I 0





) 0
" 0r1

2,4.~uoJ uI o-- [O



presence, or because the isolated silver particles were not present to act as concentration

points for heat. These spectra compared favorably with those previously published

(Sutherland et al., 1990).

To improve spectral quality and decrease the background noise of dry state

substrates, Li and Wang (1992) suggested that a solution method be used where the solid

substrate is positioned in a cuvette of the analyte solution and scatter is collected from the

immersed substrate. Spectral comparisons of dry and liquid state samples showed a strong

improvement for the liquid state surfaces. Unfortunately, the solvent modes, ethanol in this

case, enhanced along with the sample, suggesting that a less active solvent would be more


The principles of the liquid state procedure were adapted to the TLC plates without

loss of the separating attributes, by saturating the plate sections with small volumes of water.

Water was selected for its qualities as a good conductor and poor Raman scatterer. Local

drying effects from the laser were prevented by use of a cover glass through which the spectra

were collected. With this modified solution method, responses of the sulfa drugs improved

somewhat, although not to the extent found with the analytes examined by Li and Wang


Of the three drugs, SDZ, gave the strongest response, presumably due to a stronger

sulfur-silver interaction. For comparison, the CR spectrum of solid SDZ is included in Fig

3.4 along with SERS spectra of the silver-coated substrate blanks and of separated SDZ on

dampened ARG and BAK(Br-3) plates. Dry state spectra were not included because they did

not significantly differ from the coated blanks at this scale. To observe the SER effect, silver,

& x ,


cao W

on a

U a


*g a

a- *

og/ a

I -
eg o '0
1- g -


04 S

"o 0

ha iua
00 I g
E S ^E-

-no coooo
2 e ? 2 -

*/i) 0 O 0 0 0 0

l l &e & lel- 0 a
- 'C 'fi 00 In -i e 'w 0 N W 0 0 0 0
0- It Ct 00 ro 'o oo w w w 0 C\
*^ vO' C'C 0>i0>00o o o o o o 0 2^


*^1- 00 vo
*" .r 'C aU
*" SE

NS fi en a, Cte enfl^
0 ;- It( m-r- 0 W00 T
O4 e m 0 It0 W 06

3? U
S 4






C) .~ ~

a) .~











i a
g ^

C) S


?isuOuIiq OATIVIO-d


analyte, and water each had to be present on the surface. Assuming a uniform distribution

across the 4-mm spot diameter, and that half of all molecules remained on the surface rather

than penetrating into the depth, 20 ng (80 pmol) of SDZ were within the probe beam. No

surface-enhancing activity was detected from the binder or adsorbent of either plate. The

power at each surface was 27 mW.

Major modes of the enhanced and unenhanced SDZ spectra are compared in Table

3.4. Band assignments for only the strongest visible modes have been suggested, the very

sharp peak at 1151 cm-' corresponding to the SO2 symmetric stretching mode, and the band

at 1583 cm-' to a ring stretching vibration (Lin-Vien et al., 1991). Both of these bands are

also found in the SER spectrum from BAK(Br-3). The raised background on which the SER

features are superimposed is attributed to graphitic carbon.

The SER spectrum from ARG shows no modes assignable to SDZ; it appears that

most or all of the applied drug degraded under the probe. The cathedral peaks centered near

1355 and 1580 cm1 are characteristic of the disordered and graphitic microcrystalline

vibrations of amorphous carbon, respectively (Okada et al., 1992; Cooney et al., 1983). The

great extent of breakdown on a wet substrate was somewhat unusual; solution phase Raman

generally minimizes the local heating effects responsible for graphite formation. The

behavioral difference between substrates may have been prompted by the greater amplification

of EM fields on the ARG substrates, due to the greater quantity of silver particles.

While the separation behaviors were promising, coating by chemical reduction

produced surfaces of marginal activity. Neither material was found suitable for this

application. Reasons for the poor response include the uneven surface and high reflectivity


of the adsorbent layer, the weak scattering character of the drugs, spot spreading with plate

development which increased the spot-to-probe diameter ratio, and the probable loss of

analyte into the depths of the adsorbent which could render large quantities of the drugs

inaccessible to excitation. Loss by flaking of the sorbent phase during sectioning of the TLC

plate was also of concern, but was unavoidable as the sample chamber could not

accommodate an intact plate. Additionally, the silver particles had not been optimized for the

incident wavelength, primarily because chemical reduction limited control to the quantity of

silver rather than to the size and shape of individual particles.

General Applications Study

SEMS. Although the presence of silver noticeably affected the visible appearance and

separating abilities of the TLC plates, differences were not so readily apparent on a smaller

scale. Figures 3.5 and 3.6 consist of the scanning electron micrographs obtained for the BAK

and ARG materials. A variety of scan ranges were used to comprehensively describe the

surface textures of the plates. Irregularities in the sizes, shapes and orientations of the silica

gel particles are obvious. The surface is uneven with large height distributions (on the Pm-

scale) among the uppermost levels. A magnification factor of 1000 reveals the roughness of

the individual particle surfaces and allows bits of dust and debris to be seen. Despite the

uneven surfaces, these materials, like the earlier TLC plates, hold some potential as SERS

substrates because of averaging effects and the relative size of the laser beam compared to

surface irregularities. Averaging effects permit the probe to be focused -100 Pm above the

surface of the material allowing the consequent signal to be less affected by the general

unevenness of the substrate. This effect is further enhanced by the relative size of the beam











4) 0






















IV 79




,.. .'
3 .4) r 'l

compared to the textural features. The focused probe (dia = 0.5 mm) covered an area equal

to 3.5 times that seen in Figure 3.5a, a large portion of which would be"uppermost" surfaces.

The major pictorial difference between the ARG (Fig. 3.6b) and uncoated BAK (Fig.

3.6c) sections is the diffuse quality of the features on the ARG plate, a result of the

discontinuous nature of the silver layer. Distortions in this image occurred during collection

onto the film. Minor fluctuations in the scanning path of the electron beam occurred as the

negative probe encountered mild charge build-ups on isolated metal particles. Continued

scanning of an area where negative charges were not readily disseminated eventually led to

distortional charging and surface damage.

Image distortions were not an issue with the uncoated BAK section. The bare

nonconductive plate was not amenable to study by SEM without first increasing the surface

conductivity. To do so, the surface was coated with a thin layer of C-Pt (20-80%) or Au.

Well-defined images were then obtained. ARG sections were scanned without this step both

to prevent possible obscuring of the silver particles and because the silver particles

sufficiently lowered the surface resistance, permitting a reasonable field of study.

At a higher magnification (Figure 3.6), silver particles appeared as a large number of

small specks on the ARG surface. The grains, in loose clusters of three or four, averaged

between 30 and 40 nm in diameter. Lighter streaks crossing the image were due to

distortional charging. The large clumps visible on both materials consisted of dust from the

environment and from fractured silica gel particles. Images of BAK(Br-3) suffered from

severe charging and were, therefore, not included. The particles on this material were about

the same size as those on the ARG plate, but the distribution was sparse in comparison. Silver

particles formed by the Brashear procedure on smooth glass slides during 3 min exposures

were reported to be spherical with diameters in the 100-150 nm range (Boo et al., 1985). On

rough surfaces more numerous and smaller particles could reasonably be expected.

Figure of merit study. The low activity of the ARG and BAK substrates in the sulfa drug

study led to the use of Crystal Violet (tris[p-(dimethylamino)phenyl]carbonium chloride) as

the molecular model. With absorption maxima at 550 and 590 nm (Chou et al., 1986), the

compound emits a RR response from solution or SER/SERR from an enhancing surface,

when a 514.5 nm source is used. Whether SER or SERR occurs is dependent upon the

particle morphology; SERR is observed only if the molecular electronic level couples to the

surface plasmon resonance and the incident radiation. If this does not occur, CV remains a

highly active molecule, giving a strong SER signal in response to the amplified EM field of

the surface. This, combined with a low fluorescence quantum yield (10-4; Stork et al., 1973)

makes CV useful for studying weakly active surfaces. An additional benefit of using CV as

a probe is that the symmetric nature of the structure (Figure 3.7) limits the number of


(CH3)2N N(CH3)2

Figure 3.7 Structure of Crystal Violet

possible adsorbing orientations. Thus, SER spectra of CV from substrates as widely varied

as sandblasted silver foil (Ling et aL, 1991), silver-coated alumina ( Li and Wang, 1992), and

isolated colloids on chromatographic papers (Tran, 1984b) are extremely similar.

Figures of merit for the SER signals (Table 3-5) were obtained from CV applied to

BAK(Br-3) and ARG surfaces in both dry and liquid states. No CR or RR response was

observed in the absence of silver, in accordance with past observations (Tran, 1984b). The

intensity of the most prominent peak, 1174 cmru, was plotted for the calibration curve with

each point in the series averaged from three independent samples.

Table 3.5
Figures of Merit for TLC Materials
(Power at surface, 4 mW; spot dia, 3 mm)

Substrate LODa LDR %RSD
Range (ng) Correl, R2 Careful Routine
(n=3) (n=6)
Bakerflex, Br-3
Dry State 0.46 0.5-50 0.994 10 28
Damp State 0.01 0.05-25 0.9988 11 38
Argentation, 20%
Dry State 3.6 5-250 0.9999 14 21
Damp State 0.06 2.5-500 0.994 10 22
SAmount of analyte applied to surface for S/N = 3.

Although the linear region shifted somewhat with the substrate, little variation was

observed in the correlation coefficients (R2) of the curves. Comparing dry plates, BAK(Br-3)

had a larger LDR than ARG, and the range extended to lower applied amounts of analyte.

The behaviors of the dampened substrates were similar, each being linear over -2.5 decades

of concentration, again, with BAK(Br-3) including lesser analyte amounts. In each case,

altering the dielectric constant through dampening expanded the LDR and improved lower

end sensitivity.

There are a number of potential sources of nonlinearity in a SER/SERR response, any

combination of which may occur, including, nonuniformity of the substrate features, uneven

deposition of molecules on the surface, or concentration-dependent analyte orientation

(Bulkin, 1991). At the high end of the curve, nonlinearity is due primarily to a limited number

of active sites. Once all available sites are saturated and a single monolayer is surpassed, the

scatter vs concentration curve flattens as upper molecules are less influenced by and

eventually exceed the reach of the EM field. Additional considerations must be made for

species, such as, CV (Tran, 1984a), that self-absorb at higher concentrations, or, in cases

where the LDR is limited by analyte solubility. Compounds of high solubility and moderate

absorptivity typically have the largest linear ranges.

Having a greater density of silver particles, ARG would thus reasonably be expected

to accommodate larger amounts of dye within its linear range than would BAK(Br-3).

Although the particles on BAK(Br-3) are of the same approximate size and shape, the

distribution is less dense, allowing fewer active sites. Flattening of the response at lower

levels was expected and found as analyte overfilled the substrate.

The %RSD for most of the repeated measurements is within the range generally

accepted for a SER method (. 30%). While the ARG plates had been prepared commercially

in a single episode, the BAK plates were coated at various times over a two week period.

Separate coating incidents and solutions were not observed to affect the CV response.


Reproducibility between dry and damp surfaces also varied little, showing precision to be

more a function of the actual surface than of the dampness. Sample handling, however, was

demonstrated to influence the repeatability of the responses. Reproducibility among carefully

prepared samples was improved by 50% over those prepared and analyzed in a routine

manner. These results indicate that the method followed for the coating procedure was

suitable and that a great deal of care must be taken during the analysis step for a consistent


The limits of detection for this study have been conservatively defined as the total

amount of sample applied to a surface yielding a signal three times the noise. This differs

from the definition usually applied to SER samples, where the amount of analyte within the

probe area is estimated and reported, rather than the total applied quantity. The dampened

surfaces provided the lowest LODs by at least one order of magnitude over the dry materials.

In both dry and damp cases, BAK(Br-3) was more sensitive than ARG.

An estimation of the Raman EF was made from a comparison of the intensity of the

1174 cm7' band of CV adsorbed onto the silvered substrate with the intensity of this same

band on the uncoated surface for an equal number of molecules in the scattering area. For

a monolayer of CV (-pg-fg in the probe beam for most surfaces) on BAK(Bare), no CR

response was detected. From dry state BAK(Br-3), the bands are quite strong (Fig 3.8), and

Raman enhancement factors from this spectrum were estimated to be 103. Dampened, the EF

increased to 10'. ARG sections were of similar activity, with EFs slightly greater than three

and four orders of magnitude for dry and liquid states, respectively. It is important to note

that an EF is specific to the analyte species applied to a particular substrate, to the

Dry State


BAK(Br-3) + CV

8 -

6 -



600 800 1000 1200
Raman Shift (cm-1)


Liquid State

1400 1600 1800



......A/A./ "-

- \ BAK(Br-3) +CV


400 600 800 1000 1200 1400 1600 1800
Raman Shift (cm-1)


Figure 3.8 SER spectra of CV on ARG and BAK(Br-3) and of the substrates
alone in (a) dry and (b) liquid state modes.














concentration of the analyte, and to the mode selected from the spectrum. Changing one of

these factors usually alters the EF, sometimes by several orders of magnitude.

SER spectra. In this portion of the study, SER spectra of a popular molecular model and

two compound series, containing a variety of functional groups (Figures 3.9) are considered.

Each spectrum, unless otherwise noted, originated from a substrate segment dipped for 20

s into an adsorbate solution of 10-' M. This process is known to deposit approximately one

monolayer of molecules onto the surface in an even distribution (Weitz et al., 1983). An

application method that provides complete surface coverage has the advantage of requiring

less strenuous laser-sample alignment for data collection, an advantage when scanning

multiple surfaces. With application by syringe, as in the FOM study, alignment to a small

spot is understandably more rigorous and careful attention is essential to achieve consistent

results. Before considering the spectra, the benefits and limitations of immersion vs. syringe

applications will be briefly discussed.

Analyte coverage has been determined to be dependent on the concentration of the

doping solution (Garoffet al., 1982). As less concentrated solutions are used, less than one

monolayer is adsorbed without loss of uniformity, assuming the time of immersion remains

constant and short. For dipping times under 30 s, the amount of analyte introduced onto a

substrate is governed by kinetics; the period of exposure being too short for diffusion to have

an effect. At longer times, diffusion plays a prominent role and monolayers are often achieved

with dose solutions less concentrated than 10'3 M. For maximal coverage an immersion of

at least 30 min was recommended. The main disadvantage to sample doping is uncertainty

regarding the exact amount of sample applied.

Molecular Structures



NH2, heterocyclic N



NO2 --/x- COOH

Benzoic Acid

p-Aminobenzoic Acid

p-Nitrobenzoic Acid



Aromatic Ring Structures Only





Molecular Structures and Functional Groups





Functional Groups

Figure 3.9

In contrast, the amount of material introduced by syringe is known with certainty, but

adsorbate coverage is rarely uniform. As demonstrated by Sutherland et al. (1991), the

distribution is highly dependent on the solvent carrying the molecules and on the substrate

support, as described by the basic rules of chromatography. If the adsorbate, for instance, has

a greater affinity for the solvent than the substrate, a majority of molecules will deposit in a

ring along the edge of the spot. If the attraction is reversed, the molecular distribution across

the spot profile approaches a Gaussian (Hamilton and Hamilton, 1987). In either case, the

spectral response is highly dependent on the location of the probe within the spot boundary,

and on the relative sizes of probe and spot.

The highly fluorescent dye, 9-aminoacridine, has often used as a molecular model for

surface characterization (Laserna et al., 1992; Sutherland and Winefordner 1991). Its

modes, however, were not sufficiently strong for quantitation from the TLC surfaces, where

a detected response was obtained only from the solution state (Figure 3.10). The CR

spectrum was gathered from 10"2 M 9AA in methanoL Subtraction of the solvent background

enhanced the clarity of the features, so that the major bands are easily discerned. They are

in good agreement with SER from BAK(Br-3). The SER spectra of 9AA from ARG are

dominated by the cathedral peaks associated with a-C.

Surveyed compounds included the benzoic acid series, which consisted of benzoic

acid (BA), p-aminobenzoic acid (PABA), and p-nitrobenzoic acid (PNBA) and the pyrene

series, consisting of pyrene (PYR), 1-aminopyrene (APYR), and 1-nitropyrene (NPYR).

Comparisons are made among spectra from the dry and liquid state surfaces of both ARG and

BAK(Br-3). Single monolayers, applied by a 20 s immersion, approximated pg-fg within the