Scott dissertation 2015
-
Upload
brandon-scott -
Category
Documents
-
view
21 -
download
0
Transcript of Scott dissertation 2015
To the University of Wyoming:
The members of the Committee approve the dissertation of Brandon L. Scott presented on May
13, 2015.
Dr. Keith T. Carron, Chairman
Dr. David T. Anderson
Dr. Jing Zhou
Dr. Franco Basile
Dr. James L. Caldwell
APPROVED:
Dr. Keith Carron, Department Chair, Chemistry
Dr. Paula M. Lutz, Dean, College of Arts and Sciences
Scott, Brandon L., Dynamic Signal Processing for the Characterization of SERS-Active
Nanoparticles, Ph.D., Department of Chemistry, August 2015
Abstract
Since its discovery in the 1970’s, Surface-Enhanced Raman Scattering (SERS) has aided
the development of analytical methods for a wide variety of applications. Raman scattering
enhancements of up to 7 orders of magnitude permit trace detection and identification of
analytes. Furthermore, the ease of use, affordability, and portability of modern Raman
instrumentation makes it a viable candidate for analytical chemistry.
We developed a new direct and indirect SERS assay with buoyant silica microspheres,
termed Lab-on-a-Bubble. Direct assays involve coating silica bubbles with nanoparticles and
indirect assays pair bubbles with Raman reporters in a sandwich assay. These assays have the
unique advantage of buoyancy-driven detection and selection of analytes in solution. To evaluate
these assays we looked at cyanide and 5,5’-dithiobis(2-nitrobenzoic acid) (direct) and cholera
(indirect).
The second part of this dissertation relates to particle aggregation. This work follows a
report from Wustholz et al. that suggested SERS enhancement occurs near gap regions in
nanoparticle aggregates, termed hotspots. Aggregates are difficult to study due to their small
size. They can be probed in vacuum by electron microscopy but they cannot be observed directly
with light microscopy in solution. We developed a statistical method for specific extraction of
SERS signals from colloidal SERS active nanoparticles, termed dynamic SERS (DSERS). Our
first study examined a strongly coordinating monolayer, 4-mercaptopyridine, which exhibits
1
unique SERS spectra in acid and base but invariant DSERS spectra. Our interpretation was that
DSERS results showed only molecules in the gap region between nanoparticles.
Continued work examined a non-coordinating (thiophenol) and a weakly coordinating (4-
mercaptophenol) monolayer and their role in aggregation of NPs. Thiophenol was observed to
not produce unique DSERS spectra as a function of pH. In contrast to 4-mercaptopyridine, we
found that 4-mercaptophenol produced different DSERS spectra as a function of pH. We also
developed additional statistical methods to complement DSERS results: correlograms and
frequency shift histograms.
In addition to these studies we began looking at viologen-functionalized SERS substrates
for the detection of polycyclic aromatic hydrocarbons and chiral molecules. While this work is
very preliminary we observed differences in SERS spectra of (DL)-, (D)- and (L)-cysteine
adsorbed to silver nanoparticles coated with chiral viologen. We also observed adsorption of
polycyclic aromatic hydrocarbons on these substrates.
2
DYNAMIC SIGNAL PROCESSING FOR THE CHARACTERIZATION OF SERS-
ACTIVE NANOPARTICLES
by
Brandon Scott
A dissertation submitted to the Department of Chemistry and the Graduate School of the University of Wyoming in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
in
CHEMISTRY
Laramie, Wyoming
August 2015
Acknowledgements
I would like to thank my graduate advisor, Dr. Keith Carron, for his support and encouragement
throughout my undergraduate and graduate career. Special thanks to research collaborators Dr.
Richard Martoglio, Dr. Virginia Schmit, Dr. Aaron Strickland, Dr. Ed Clennan, Xiaoping Zhang,
and Jacob Williams. Thanks to all of the teachers and professors that inspired me to pursue
scientific research, my friends and my family. I would not be where I am today without all of the
people in my life who believe in me.
ii
Table of Contents
Abstract
Acknowledgements
Table of Contents
1 Introduction............................................................................................................................1
1.1 Raman Spectroscopy.........................................................................................................3
1.2 Surface-Enhanced Raman Scattering (SERS)...................................................................4
1.3 Salt Enhancement of SERS...............................................................................................5
1.4 Lab-on-a-Bubble...............................................................................................................7
1.5 Dynamic SERS...............................................................................................................12
1.6 References.......................................................................................................................14
2 Lab-on-a-Bubble (LoB): Synthesis, Characterization, and Evaluation of Buoyant Gold Nanoparticle-Coated Silica Spheres..........................................................................16
2.1 Introduction.....................................................................................................................16
2.2 Experimental Methods....................................................................................................18
2.2.1 Silanization of Glass Bubbles..................................................................................18
2.2.2 Preparing and Shelling Gold Nanoparticles (AuNPs).............................................18
2.2.3 Modification of Glass Bubbles with AuNPs...........................................................18
2.2.4 Concentration of AuNP-Coated Glass Bubbles.......................................................19
2.2.5 Instrumentation........................................................................................................19
2.2.6 UV-vis Spectroscopy...............................................................................................20
2.2.7 SERS of AuNPs Added to Aqueous Cyanide (CN-) Solutions...............................20
2.2.8 SERS of AuNP-Coated Glass Bubbles Added to Aqueous CN- Solutions.............21
2.2.9 SERS of Varying Amounts of AuNP-Coated Glass Bubbles Added to CN- Solutions of Constant Concentration.......................................................................21
2.3 Results and Discussion....................................................................................................22
2.4 References.......................................................................................................................30
3 Lab-on-a-Bubble Surface Enhanced Raman Indirect Immunoassay for Cholera.........32
3.1 Introduction.....................................................................................................................32
3.2 Materials and Methods....................................................................................................35
iii
3.2.1 LoB Activation and Antibody Attachment..............................................................35
3.2.2 Dynamic Light Scattering (DLS).............................................................................35
3.2.3 Raman Reporter Synthesis.......................................................................................36
3.2.4 Preparing and Shelling AuNPs................................................................................40
3.2.5 LoB Immunoassay...................................................................................................41
3.2.6 Data Acquisition and Analysis................................................................................41
3.3 Results.............................................................................................................................42
3.4 Acknowledgements.........................................................................................................48
3.5 References.......................................................................................................................48
4 Dynamic SERS: Extracting SERS from Normal Raman Scattering...............................51
4.1 Introduction.....................................................................................................................51
4.2 Results and Discussion....................................................................................................52
4.2.1 SERS Signal Extraction...........................................................................................52
4.2.2 Sites Selective Spectroscopy...................................................................................55
4.3 Conclusion......................................................................................................................60
4.4 Acknowledgements.........................................................................................................61
4.5 References.......................................................................................................................61
5 Statistical Analysis of 4-Mercaptophenol and Thiophenol on Gold Nanoparticles.......63
5.1 Introduction.....................................................................................................................63
5.2 Materials..........................................................................................................................67
5.3 Experimental...................................................................................................................68
5.4 Instrumentation...............................................................................................................68
5.5 Data Analysis..................................................................................................................68
5.6 Raman Modes.................................................................................................................70
5.7 Results.............................................................................................................................71
5.7.1 4-Mercaptophenol Analysis.....................................................................................71
5.7.2 Thiophenol Analysis................................................................................................78
5.7.3 4-Mercaptopyridine Analysis..................................................................................82
5.8 Summary.........................................................................................................................84
5.9 Conclusion......................................................................................................................85
5.10 References.......................................................................................................................86
iv
6 Clennan Group Collaboration: Viologen-Functionalized SERS Substrates for the Detection of Polycyclic Aromatic Hydrocarbons and Chiral Molecules.........................88
6.1 Introduction.....................................................................................................................88
6.2 Silver Nanoparticle (AgNP) Synthesis...........................................................................88
6.3 Instrumentation...............................................................................................................89
6.4 Experimental...................................................................................................................89
6.5 Results and Discussion....................................................................................................90
6.6 References.......................................................................................................................99
v
1 Introduction
Since its discovery, analytical assays based on surface-enhanced Raman scattering
(SERS) have developed using wide variety of methods for several applications. Our research
group continued the development of SERS assays by implementing buoyant bubbles with unique
advantages and with dynamic Raman scattering (DRS) to detect very low concentrations of
SERS particles.
The first goal was to optimize SERS enhancement by inducing hotspots via addition of
electrolytes to SERS substrates. Although these solutions showed significant SERS
enhancement, the stability of the SERS substrate was compromised due to rapid aggregation.
This led us to examine methods of improving the stability of colloidal SERS substrates.
Ultimately this gave rise to two novel SERS detection methods. Both methods utilized buoyant
silica microspheres which float to the surface of a solution. The first method involves adsorbing
gold colloids to the microsphere surface, effectively controlling aggregation effects while
maintaining SERS activity. The second method involves coupling SERS-active colloids coated
in silica (Raman reporters) to buoyant silica microspheres via antigen-antibody binding. The
novelty in these two methods came from pairing SERS substrates to the buoyant silica
microspheres to effectively concentrate the SERS-analyte complex to the surface of the aqueous
solution. The term Lab on a Bubble (LoB) was coined to describe this technique and is described
in detail in chapters 2 and 3.
However, both of these techniques affect SERS hotspot phenomena by permanently
fixing colloids to a surface or within a silica shell. Several research groups showed that SERS
hotspots occur in colloidal solutions between coalesced nanoparticles, albeit at very low
1
concentrations relative to nanoparticle monomers. This led to our attempts to develop a new data
analysis technique to distinguish between normal and hotspot-enhanced SERS signaling within a
colloidal solution. We began by monitoring signal fluctuations between multiple spectra of dilute
colloidal SERS substrates collected in rapid succession. Fluctuations in the SERS signal were
found to be inversely proportional to nanoparticle concentration; a result of noise created by
fewer particles passing through the detection beam by Brownian motion. The term Dynamic
SERS (DSERS) was coined to describe this technique. Analytes with pH-dependent SERS
substrate binding sites were examined to induce hotspot formation via chemical bonding and
DSERS results were compared. DSERS provided a tool to investigate shifts in vibrational modes
and anomalous SERS signals due to hotspots that are otherwise lost in conventional SERS
analysis.
We are including a brief collaboration with Dr. Clennan’s group to implement a chiral
viologen they synthesized into achiral SERS assays. Similar research demonstrated the ability of
viologen-functionalized SERS substrates to detect polycyclic aromatic compounds (PAHs) that
are otherwise undetectable by conventional SERS methods and we generated similar results. Our
preliminary results look promising but this topic of research will require further investigation.
Raman spectroscopy and surface-enhanced Raman scattering are the backbone for
modern SERS assays. We finish the chapter by introducing signal enhancement techniques and a
novel SERS detection application. We describe what they are, why they are important to the
technique, and their potential for further advancement. The next three chapters describe
completed work submitted for publication; lab-on-a bubble assays and dynamic SERS,
respectively. The final two chapters describe work to be submitted for publication.
2
1.1 Raman Spectroscopy
Raman spectroscopy is a technique to measure the rotational and vibrational modes of
molecules. Unlike infrared spectroscopy, which involves a change in the dipole moment of a
molecule excited from the ground vibrational state to an excited state, Raman spectroscopy
involves an induced dipole moment that leads to the scattering of light from a vibrational state.
The resulting scattered photon can either have a frequency less than or greater than the incident
light frequency, known as Stokes or anti-Stokes scattering, respectively, as shown in Figure 1.1.
For both cases this inelastically scattered light (Raman scattering) can be separated from the
dominant elastically scattered light (Rayleigh scattering) by dispersion from the spectrum before
a detector. The amount of deformation of the electron cloud of a molecule with respect to the
vibrational coordinate determines the strength of the Raman effect1. Raman spectroscopy and IR
spectroscopy produce similar, but sometimes complementary, results.
3
Excited electronic state
Virtual states
Ground electronic states
v = 0
v = 1
v = 2
Vibrational levels
Infrared Absorption
Stokes
Rayleigh Anti-Stokes
Figure 1.1: Electronic state diagram showing Stokes, anti-Stokes, and Rayleigh scattering events for a molecule interacting with light of suitable frequency.
1.2 Surface-Enhanced Raman Scattering (SERS)
Surface-enhanced Raman scattering (SERS) enhances Raman scattering from molecules
adsorbed to a rough metal surface by up to seven orders of magnitude. This gives it the potential
to be a sensitive tool for analytical chemistry. The phenomenon was first observed by
Fleischmann, et al.2 in 1974 and explained by Jeanmaire and Van Duyne3 three years later. There
are two proposed theories for describing the SERS effect: electromagnetism and the formation of
charge-transfer complexes. The electromagnetic theory attributes the light-induced interaction
between adsorbed molecules and the localized surface plasmon resonance of certain metals, such
as gold and silver, to explain the large enhancement factor of SERS. When metal nanoparticles
are much smaller than the wavelength of incident light, the individual atoms undergo concerted
dipolar electric field oscillations to produce the LSPR phenomenon. This phenomenon can be
explained by the solution for the response of a dielectric sphere in a uniform electric field4
E¿=3 ε (ω )
ε (ω )+2E0
Where Ein is the electric field near the particle, ε(ω) is the dielectric function of the particle
material, and E0 is the electric field of the light incident on the sphere. In the case of free electron
metals, such as copper, silver and gold, the dielectric function has a negative real and small
imaginary component, which correspond to the storage and dissipation of energy within the
medium, respectively. As ε(ω) approaches -2 a resonance occurs and the electric field inside the
particle increases dramatically.
4
These dipolar plasmon oscillations produce an enhancement of the electric field of both the
incident light as well as the scattered Raman light, to produce a combined E4 signal
enhancement. Gold and silver nanoparticles are typically used for SERS since their plasmon
resonance frequencies lie within the visible and near-infrared region.
The electromagnetic theory can be used to explain most of the SERS enhancement of any
species of molecule either chemisorbed or physisorbed to a metal surface. However the charge-
transfer complex theory, or chemical theory, can be used to explain SERS enhancement larger
than the E4 predicted by the electromagnetic theory. Molecules containing lone electron pairs are
capable of forming chemical bonds with the metal surface that may lead to a charge-transfer
(CT) complex. The CT complex may have absorption in the visible region that lead to resonance
Raman.
Recently, SERS signal enhancement due to hotspots has been investigated by several
research groups. By combining experimental and modeling experiments Van Duyne’s research
group determined that hotspots are located near interparticle gap regions where two particles are
in subnanometer proximity or have coalesced to form crevices. SERS signal enhancements of 108
were determined for aggregated nanoparticles containing hotspots.
1.3 Salt Enhancement of SERS
SERS enhancement of analytes adsorbed to gold nanoparticles may be further increased
by the addition of a weak electrolyte solution to the sample matrix (Figure 1.2). This
phenomenon was examined with NaCl, NaF, KBr, NaI and NaBr by adding 250 uL of varying
concentrations of each salt solution to a mixture of 250 uL gold colloids with Nile Blue as our
probe. Salt concentrations 500mM, 250mM, 125mM, 62.5 mM, and 31.3 mM with a constant
5
concentration of Nile Blue-adsorbed nanoparticles in every sample were analyzed. Spectra were
collected for each sample and the Nile Blue peak heights at 589 cm-1 were plotted as a function
of salt concentration. Figure 1.3 indicated a signal increase at low salt concentrations with 62.5
mM NaCl producing the largest enhancement and a reduction of signal enhancement at higher
concentrations. The sample with the highest salt concentration had a weaker signal than the
sample containing no salt indicating degradation of the SERS-active complexes most likely
results from particle aggregation within the sample matrix. Results for NaF, KBr, NaI, and NaBr
experiments showed similar behavior.
Figure 1.2: Raman spectra of colloids in 1.6 µM nile blue (red); colloids in 1.6 µM nile blue and 31.3 mM NaCl (blue).
6
400 450 500 550 600 650 700 400 450 500 550 600 650 700
Inte
nsi
ty
Wavenumber
0 100 200 300 400 500 60012000
13000
14000
15000
16000
17000
18000
NaCl titration of 1 µM NB colloids
[NaCl] (mM)
peak
hei
ght
Figure 1.3: Nile blue peak height at 589 cm-1 vs. NaCl concentration (mM)
While it is clear that the addition of a weak electrolyte solution to a SERS-active
substrate can be used to optimize SERS enhancement controlling this phenomenon is not
reported. Two possible explanations for salt SERS enhancement are an increase in the LSPR due
to Van Der Waals forces from the salt ions or the promotion of hotspot-containing nanoparticle
multimers from the electrolyte-induced reduction in the repelling force of negatively charged
gold nanoparticles. However, it is clear that the stability of the colloidal solution is compromised
by the addition of electrolytes ultimately leading to particle aggregation and precipitation of the
SERS substrate.
1.4 Lab-on-a-Bubble
The use and effectiveness of SERS assays have been demonstrated in a wide variety of
applications5. Common techniques involve either direct detection of analyte-bound nanoparticles
suspended in a colloidal solution or indirect detection of analytes bound to Raman-active
7
nanoparticles via ligand binding interactions. Although somewhat effective, both of these
techniques have major drawbacks. Direct SERS assays often have large limit of detection (LOD)
values due to a small amount of analyte present in a large sampling volume and poor binding
affinity of some analytes to gold nanoparticles. The second problem can be overcome by
implementing indirect techniques if the analyte of interest binds to the modified nanoparticle but
this technique also has its shortfalls, including detection of false positives.
Our work on SERS immunoassays yielded interesting results by sandwiching analytes
between Raman-active nanoparticles and paramagnetic microparticles via antigen-antibody
interactions and concentrating the analyte-bound complex within the sample by introducing a
magnetic field6. Although this method improves the LOD and reduces the detection of false
positives it too has a problem. The attractive magnetic force between a permanent magnet and a
paramagnetic particle decreases exponentially with increasing distance. This requires using
powerful magnets and small sample vials to conduct such paramagnetic pull-down sandwich
immunoassays.
Our research group proposed a solution for both direct and indirect methods by
implementing buoyant silica microspheres into the assays7. The resulting assay is referred to as
Lab-on-a-bubble, or LoB. Figure 1.4 illustrates the concept of a direct LoB assay along with
representative scanning electron micrographs and Raman data acquired from LoB reagents. In a
typical LoB assay, LoB reagents comprised of buoyant SiO2 bubbles and Au or Ag nanoparticles
(NPs) are combined to provide a SERS active particle platform (Figure 1.4A-B) for the
detection of target analytes by localizing them close to the bubble-NP composite (Figure 1.4B-
C). Bubble flotation drives the complex to a specified point in a reaction vessel where the
analyte is selectively detected as a concentrated LoB complex as illustrated in Figure 1.4C. For
8
the current study AuNP-coated LoBs were prepared by activating buoyant silica bubbles (3M
Corporation) with aminopropyltriethoxysilane (APTES) following a standard protocol for glass
coating (Figure 1.4A ,B).8,9 Colloidal gold was incubated with the activated bubbles to adsorb
AuNP aggregates onto the bubble surface (Figure 1.4B, D); aggregates of gold and silver
nanoparticles are known to exhibit strong enhancements in the Raman signal of adsorbed
analytes.10,11 Figure 1.4E shows spectra resulting from AuNP-coated LoBs in the presence
(black spectrum) and absence (red spectrum) of 5 μM 5,5’-dithiobis(2-nitrobenzoic acid)
(DTNB). These spectra were collected by combining SERS active LoBs with DTNB analyte,
allowing the buoyant LoBs to float to the top of a vessel, and collecting the Raman data using an
808 nm Sierra Raman spectrometer (SnRI LLC). Figure 1.4C and the inset in Figure 1.4E
demonstrate a detection scheme for the LoB assay. AuNP coated LoBs were optimized for
SERS activity by starting with a known bubble quantity and saturating the bubble surface using a
progressively larger volumes of colloidal AuNP.
9
Figure 1.4: A-C) The basic components of a Lab on a Bubble (LoB) assay for SERS-based detection of a analyte. (D) Representative scanning electron micrographs of SERS-active AuNP-coated LoBs. (E) Representative Raman spectra of ‘naked’ LoBs, and LoBs in the presence of DTNB. The inset shows picture of SERS-active buoyant LoBs in a microcentrifuge tube.
LoB materials serve as a convenient platform for the detection of analytes in solution and
offer several advantages over traditional colloidal gold and planar SERS substrates. Chapter 3
describes a LoB-based cyanide assay. Cyanide bound to gold-coated LoBs was detected directly
from the corresponding SERS signal. Detection of cyanide in gold colloid is comparable to that
in the presence of LoBs, with a detection limit of ~170 part-per-trillion determined for both
10
cases. Prevention of aggregation common to colloidal nanoparticles is also discussed in relation
to an assay for 5 μM 5,5’-dithiobis(2-nitrobenzoic acid) (DTNB). However, the sensitivity of this
technique still depends on the binding equilibrium between the analyte and LoBs, which limits
the improvement of detection to tightly binding analytes. To overcome this obstacle, our research
group has developed an additional analytical method, known as dynamic SERS, which is
detailed in the following section.
An improved SERS sandwich assay was developed using buoyant silica microspheres,
described above, coated with antibodies against the B subunit of the cholera toxin (CT), and gold
nanoparticles tagged with a Raman reporter, shelled with silica and coated with antibodies
against the B subunit of CT12. Together these components couple to form a sandwich which, after
incubation, floats on the surface of the sample. The buoyant silica microparticle / nanoparticle
reporter combination has been coined a Lab on a Bubble (LoB). LoB materials may provide a
platform for rapid detection of antigen in solution and offers advantages over lateral flow or
magnetic pull-down assays. The Raman reporter provides a unique and intense signal to indicate
a positive analysis. Our limit of detection for the beta subunit of the CT in a buffer based system
is 1100 ng.
11
Figure 1.5: Comparison of LoB sandwich assay (A) and paramagnetic pull-down assay (B)
1.5 Dynamic SERS
Although SERs-based assays have proven to be effective analytical tools, there is still
speculation as to what actually causes enhancement and detection within a sample solution. It is
widely accepted that an analyte binds to gold nanoparticles as predicted by an isotherm model
with a monolayer leading to the greatest signal enhancement. However, inconsistencies of SERS
enhancement between different nanoparticle species within a colloidal solution have been
demonstrated. Such inconsistencies often arise between single nanoparticles and clusters of
nanoparticles. Other researchers demonstrated that clusters of two or more nanoparticles lead to
the largest amount of SERS enhancement due to the presence of so-called “hotspots”13. Hotspots
are regions where two nanoparticles are in close proximity with one another. The Van Duyne
group investigated nanoparticle hotspot regions using a combination of SEM and LSPR
spectroscopy on adjoined nanoparticle pillars14. Other research groups developed high hotspot-
yielding nanoparticle complexes either by novel synthesis or filtration methods.
12
Particle Pairs Float to Surface
Particle Pairs Pulled by Magnet
LDN Assay
Paramagnetic Pull-down
Assay
A B
Our research group developed a much simpler approach to detecting hotspot-containing
nanoparticle complexes involving the time-dependent data analysis of multiple SERS spectra15.
Similarly, time correlation of fluorescence spectroscopy was shown to distinguish instantaneous
light scattering events and delayed fluorescence signals. We demonstrated that the standard
deviation of SERS signal intensity increases as the concentration of nanoparticles in a sample
solution decreases7 due to individual nanoparticles passing through the Raman laser beam as
dictated by Brownian motion within the sample medium. By taking a large number of SERS
spectra in a short amount of time and subtracting the average spectrum from the normalized
standard deviation spectrum we generated unprecedented solvent noise reduction. The result is a
spectrum containing the signal produced specifically by the nanoparticle complexes within the
sample. Furthermore, correlating the data set at specific spectral peaks revealed the presence and
movement of individual nanoparticle-analyte complexes of varying SERS enhancement.
Figure 1.6: (Left) Illustration of an analyte-adsorbed AuNP dimer with a hotspot. (Middle) comparison of SERS spectrum vs. DSERS-corrected spectrum. (Right) Stochastic motion of AuNP complexes within the detection beam.
13
Stochastic Nanoparticle Motion
Analyte monolayer
HotspotS
σWavenumbers
SERS
DSERS
Our first example of shelled nanoparticles at very low concentrations, explained in
further detail in chapter 4, confirmed the benefit of DSERS for removal of an overwhelmingly
strong solvent spectral interference. The second benefit, site selection, was demonstrated with 4-
mercaptopyridine on bare Au nanoparticles to observe a small population of molecules that were
spectroscopically unique from the large population of molecules on the particles. The DSERS
spectrum originated from excess variance between a small population of adsorbates on the
ensemble of nanoparticles. We demonstrated two significant benefits of dynamic SERS
(DSERS) measurements: removal of instrumental and normal Raman interferences in SERS
spectroscopy; and site selective spectroscopy of adsorbate populations on SERS active particles.
1.6 References
1. Raman, C. V.; Krishnan, K. S., A New Type of Secondary Radiation. Nature 1928, 121, 501-502.
2. Fleischmann, M.; Hendra, P. J.; McQuilla.Aj, Raman-Spectra of Pyridine Adsorbed at a Silver Electrode. Chem. Phys. Lett. 1974, 26 (2), 163-166.
3. Jeanmaire, D. L.; Van Duyne, R. P., Surface Raman Spectroelectrochemistry. Part 1. Heterocyclic, Aromatic, and Aliphatic-Amines Adsorbed on the Anodized Silver Electrode. J. Electroanal. Chem. 1977, 84 (1), 1-20.
4. Van de Hulst, H. C., Light Scattering by Small Particles. Dover Publications, Inc.: New York, 1981; p 71.
5. Driscoll, A. J.; Harpster, M. H.; Johnson, P. A., The Development of Surface-Enhanced Raman Scattering as a Detection Modality for Portable In Vitro Diagnostics: Progress and Challenges. Physical chemistry chemical physics : PCCP 2013, 15 (47), 20415-33.
6. Lu, Y.; Yin, Y. D.; Mayers, B. T.; Xia, Y. N., Modifying the Surface Properties of Superparamagnetic Iron Oxide Nanoparticles through a Sol-Gel Approach. Nano Lett. 2002, 2 (3), 183-186.
7. Schmit, V. L.; Martoglio, R.; Scott, B.; Strickland, A. D.; Carron, K. T., Lab-on-a-Bubble: Synthesis, Characterization, and Evaluation of Buoyant Gold Nanoparticle-Coated Silica Spheres. J. Am. Chem. Soc. 2012, 134 (1), 59-62.
14
8. Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter, D. G.; Natan, M. J., Self-Assembled Metal Colloid Monolayers: An Approach to SERS Substrates. Science 1995, 267, 1629-1632.
9. Karrasch, S.; Dolder, M.; Schabert, F.; Ramsden, J.; Engel, A., Covalent Binding of Biological Samples to Solid Supports for Scanning Probe Microscopy in Buffer Solution. Biophys. J. 1993, 65 (6), 2437-2446.
10. Pierre, M. C. S.; Mackie, P. M.; Roca, M.; Haes, A. J., Correlating Molecular Surface Coverage and Solution-Phase Nanoparticle Concentration to Surface-Enhanced Raman Scattering Intensities. J. Phys. Chem. C 2011, 115 (38), 18511-18517.
11. Wang, H.; Levin, C. S.; Halas, N. J., Nanosphere Arrays with Controlled Sub-10-Nm Gaps as Surface-Enhanced Raman Spectroscopy Substrates. J. Am. Chem. Soc. 2005, 127 (43), 14992-14993.
12. Schmit, V. L.; Martoglio, R.; Carron, K. T., Lab-on-a-Bubble Surface Enhanced Raman Indirect Immunoassay for Cholera. Anal. Chem. 2012, 84 (9), 4233-4236.
13. Chen, G.; Wang, Y.; Yang, M. X.; Xu, J.; Goh, S. J.; Pan, M.; Chen, H. Y., Measuring Ensemble-Averaged Surface-Enhanced Raman Scattering in the Hotspots of Colloidal Nanoparticle Dimers and Trimers. J. Am. Chem. Soc. 2010, 132 (11), 3644-+.
14. Wustholz, K. L.; Henry, A. I.; McMahon, J. M.; Freeman, R. G.; Valley, N.; Piotti, M. E.; Natan, M. J.; Schatz, G. C.; Van Duyne, R. P., Structure-Activity Relationships in Gold Nanoparticle Dimers and Trimers for Surface-Enhanced Raman Spectroscopy. J. Am. Chem. Soc. 2010, 132 (31), 10903-10910.
15. Scott, B. L.; Carron, K. T., Dynamic Surface Enhanced Raman Spectroscopy (SERS): Extracting SERS from Normal Raman Scattering. Anal. Chem. 2012, 84 (20), 8448-51.
15
2 Lab-on-a-Bubble (LoB): Synthesis, Characterization, and Evaluation of Buoyant
Gold Nanoparticle-Coated Silica Spheres1
2.1 Introduction
Micro and Nano – Electro – Mechanical systems MEMS and NEMS have made
significant impacts on chemical sensors. For example, the technology behind Lab-on-a-Chip
(LOC) has emerged into a large market defining Point-of-Care (POC) diagnostics2. These novel
systems represent combinations of miniaturized chemical separation methods and a variety of
detection schemes. The drive to miniaturized instrumentation and straightforward single-step
assays has brought about the growth of these research efforts. One example, of a nano-powered
engine is separations that use nanoparticulate paramagnetic materials to couple to analytes. The
paramagnetic engines are powered by external magnets that concentrate the assay results into a
small localized volume for more sensitive analysis. This scheme works well in small sample
volumes and with sufficient time for exponentially decaying magnetic fields to impel the
majority of the particles. In this article we will present a different method of nanopropulsion –
buoyancy from a hollow silica ‘bubble’ to produce a Lab on a Bubble (LoB).
Our initial work with paramagnetic nanoparticles was driven by a fundamental limitation
to Surface Enhanced Raman Scattering (SERS) analysis with colloidal nanoparticles. This
limitation originates with dispersive Raman instruments and the property of étendue. Succinctly,
étendue describes the inverse relationship between spectral resolution and a spectrometer’s
optical throughput. When sampling a nanoparticle solution étendue coupled with a reasonable
spectral resolution requires a focused beam from the excitation laser. Likewise, the colloidal
nature of nanoparticles in a solution requires that they be continually propelled by Brownian
16
motion and thus individual particles are moving into and out of the focused laser beam. It is
often desirable to use a small quantity of nanoparticles to maximize the surface coverage of a
strongly adsorbing analyte; this leads to fluctuations in the SERS signal due to the Brownian
motion induced fluctuation of particles within the focal volume. A chemical analysis for analyte
concentration will be limited by these fluctuations. It is desirable to have the noise in an
experiment be limited by shot noise of the detector, but as we will report, the noise in our
colloidal nanoparticle experiments far exceeds the detector’s shot noise.
SERS active nanoparticles provide valuable information about species in aqueous media.
However their widespread use is limited by their instability. Recently, Pierre et al.3 have shown
the affect of nanoparticle instability on Au nanoparticle (AuNP) assays. They demonstrated the
loss of signal due to changes in AuNP surface as a result of adsorption of a neutral thiol species.
Aggregation is also caused by changes in pH, ionic strength, and mixing parameters. The
limitations of signal noise in excess of the detection system and the instability of nanoparticles
under adsorptive processes is a critical problem for viable SERS diagnostics.
In this study we report results from a different approach to solution phase analysis with
SERS active nanoparticles that combines the separation mechanism directly coupled to the
detection method. This LoB concept is centered on a low density particle that utilizes a buoyant
force to drive assay separation, while Au nanoparticles (AuNP) coupled to the buoyant particles
act as SERS nanosensors. Addition of a selective coating on the AuNP creates the potential for
smart sensors. In the current study we report the detection of a generic thiol containing Raman
active small molecule, and cyanide which is a relevant model analyte in environmental testing.
17
2.2 Experimental Methods
2.2.1 Silanization of Glass Bubbles
0.3g of S60/10000 3M Glass bubbles (average diameter 30 μm, density 0.6 g/mL) were
added to 10N H2SO4 overnight to activate the glass surface. 4 Silanization of the activated glass
bubbles was achieved via exposure to a 10% solution (v/v) of 3-aminopropyltriethoxysilane in
methanol overnight with constant rocking. The glass bubbles were subsequently washed 6 times
with methanol and re-suspended in 3 mL HPLC grade H2O for future use. 5
2.2.2 Preparing and Shelling Gold Nanoparticles (AuNPs)
AuNPs were prepared by the Frens method.6 200mL of HPLC grade H2O was added to a
beaker and warmed on a hot plate. Once the water was warmed to approximately 30˚C, 20 mg of
HAuCl4 was added to the solution and brought to a rolling boil. 1200 μL of 1% (wt/vol)
Na3C6H5O7 was then added all at once. The solution boiled for one hour with a watch glass
placed over the beaker. The solution was then removed from the heat and allowed to cool to
room temperature prior to storage. This method of synthesis produced AuNPs with an average
diameter of approximately 50 nm as determined by SEM. The concentration of the AuNP
solution was 6.0 x 1010 AuNPs/mL by a method similar to that of Haiss et. al.7
2.2.3 Modification of Glass Bubbles with AuNPs
Immediately following sufficient agitation of the silane-treated glass bubble solution, 200
μL was added to a 1.75 mL Eppendorf tube. The glass bubbles were allowed to float to the
surface and the supernatant was removed with a 1 mL syringe and 26-gauge needle. The glass
bubbles were rinsed at least 5 times with 200 μL of 50% (v/v) MeOH solution in water to
18
remove excess APTES: For each wash, 200 μL of the MeOH solution was added and the sample
was agitated at room temperature for ca. 2 minutes. 1 mL of HPLC H2O was then added to the
glass bubble MeOH solution to facilitate floatation of the glass bubbles. The supernatant was
carefully removed and the rinse procedure was performed at least 4 more times with the
supernatant being completely removed on the final rinse. Next, 200 μL of Au nanoparticles
(AuNPs) were added to these rinsed glass bubbles. The mixture was agitated at room
temperature until the solution became almost clear. The glass bubbles were allowed to float to
the top of the solution and the supernatant was removed. AuNPs were added in 200 μL volumes
and agitated until the solution remained purple. The resulting Au coated glass bubbles were re-
suspended in 500 μL of HPLC grade H2O.
2.2.4 Concentration of AuNP-Coated Glass Bubbles
10 μL of the Au-coated glass bubble solution was added to a microscope slide and
allowed to dry. An Olympus BX51 microscope was used to determine the counting area of the
bubble solution and the bubbles in this area were enumerated. Based on the total area of the
solution and the numbers of bubbles counted, we approximated the concentration of Au-coated
glass bubbles to be 1 x 105 bubbles/mL.
2.2.5 Instrumentation
All spectroscopic data was collected using a Snowy Range Instruments IM 52 808 nm laser
Raman system with rastering capability. The rastering addition maintains small laser spot size
while averaging over an elliptical area of ca. 2 mm x 0.5 mm.
19
2.2.6 UV-vis Spectroscopy
UV-vis spectra of aqueous gold nanoparticles can be used to determine the concentration
of the colloidal solution if the approximate nanoparticle diameter is known, as described by
Haiss et. al7. The size of nanoparticles affects how the colloidal solution scatters incident light.
Thus, the wavelength of maximum absorbance changes as a function of nanoparticle diameter.
The amount of relative absorbance at a given wavelength is a function of nanoparticle
concentration, as described by Beer’s law. Although TEM or SEM imaging can be used to
simultaneously determine nanoparticle size and concentration, this technique is much faster and
easier to implement. Gold nanoparticle solutions synthesized using the Frenz citrate method
described in section 2.1 have a concentration of about 0.1 nM.
2.2.7 SERS of AuNPs Added to Aqueous Cyanide (CN-) Solutions
30 μL of AuNPs (1.8 x 109 nanoparticles) were added to an equal volume of sodium
cyanide solution buffered at pH = 9 (4:1 (v/v) 0.1M NaHCO3:0.1M Na2CO3 buffer). Cyanide
solutions of varying concentrations (200 parts per million (ppm) to 2 parts per billion (ppb))
were titrated while maintaining constant volumes from sample to sample. Upon addition of
AuNPs to the CN- solutions, each sample was incubated for 5 minutes with gentle agitation at
room temperature. The entire volume was pipetted onto a steel substrate for interrogation with
the laser. Each spectrum was acquired for 0.5 sec and the intensity was plotted against the
cyanide concentration. Each data point was replicated 5 times for the same integration time and
error bars on graph are +/- 1 standard deviation of all 5 replicates.
20
2.2.8 SERS of AuNP-Coated Glass Bubbles Added to Aqueous CN- Solutions
10 μL of Au-coated glass bubble solution (1.5 x 106 Au-coated glass bubbles) was added
to 40 μL of sodium cyanide solution buffered at pH = 9 (4:1 0.1M NaHCO3:0.1M Na2CO3
buffer). Cyanide solutions of varying concentrations (200 parts per million (ppm) to 2 parts per
billion (ppb)) were titrated while maintaining constant volumes from sample to sample. Samples
were incubated for 5 minutes with gentle agitation at room temperature. The entire volume was
pipetted onto a steel substrate for interrogation with the laser. The Au-coated glass bubbles were
allowed to float to the top of each sample prior to analysis and they formed a small circular
island in the middle of each sample. Once this was observed, each spectrum was acquired for 0.1
sec and the intensity was plotted against the cyanide concentration. Each data point was
replicated 5 times for the same integration time and error bars on graph are +/- 1 standard
deviation of all 5 replicates.
2.2.9 SERS of Varying Amounts of AuNP-Coated Glass Bubbles Added to CN- Solutions
of Constant Concentration
In each trial, the CN- concentration was held at 1 ppm. The amounts of Au-coated glass
bubbles were varied, but the amount of solution containing the Au-coated glass bubbles was held
constant for each sample. Dilutions of the Au-coated glass bubbles were made as follows from
500 μL of the Au-coated glass bubble stock solution: 80 μL stock solution was added to 20 μL
H2O, 60 μL stock was added to 40 μL H2O, 40 μL stock was added to 60 μL H2O, and 20 μL
stock was added to 80 μL H2O. 10 μL of each dilution was added to 30 μL of 1ppm CN-
solution. 10uL of the undiluted stock solution was also added to 30 μL of 1 ppm CN- solution,
and 10 μL water was added to 30 μL of 1 ppm CN as a negative control. Samples were mixed
21
with gentle agitation for 3 minutes at room temperature. The entire volume was pipetted onto a
steel substrate for interrogation with the laser. The Au-coated glass bubbles were allowed to float
to the top of each sample prior to analysis and they formed a small circular island in the middle
of each sample. Each spectrum was acquired for 0.5 sec and the intensity was plotted against the
Au-coated glass bubble concentration. Each data point was replicated 5 times for the same
integration time and error bars on graph are +/- 1 standard deviation of all 5 replicates.
2.3 Results and Discussion
Figure 2.1 illustrates the dynamic properties of AuNP-coated LoBs as compared to
AuNPs in a solution. In Figure 2.1A we illustrate that as the number of nanoparticles in a
focused laser beam decreases the relative error of a measurement sharply increases due to
Brownian motion. Statistically this is expected to follow a Poisson distribution and to increase
according to 1/N1/2 as the number of nanoparticles (N) decrease. The data in Figure 2.1A was
collected with a shot-noise limited detector (Andor) cooled to -80°C (New Dimension Raman
Microscope (SnRI, LLC). SEM analysis of the particles indicated that the average size was
approximately 50 nm and UV-Vis indicated a stock concentration of 6.4 x 1010 AuNP/mL. Our
probe in this study was adsorbed cyanide from a sodium cyanide solution at 1 ppm and pH = 9.
With 16 AuNP in the focal volume of ~ 8 nL the variation in the signal is 24 times that predicted
by a shot-noise limited detection system.
22
Figure 2.1: A) The increase in noise as a function of colloidal AuNP concentration. B) The increase in noise as a function of LoB concentration. The noise is determined by the relative standard deviation from 10 measurements. In both measurements a focus beam was used to collect the data.
A goal in chemical analysis is to reduce the variation in signals such that the limit of
detection (LOD) will decrease. The LOD is defined as: LOD = 3σ/m, where σ is the standard
deviation and m is the slope. Figure 2.1B shows our results with LoB particles. Figure 2.1B
demonstrates the large difference in σ for the static LoBs as compared to colloidal AuNPs; where
σ(LoB) is 0.05 for 1 LoB particle compared to 1.0 for 16 AuNPs in the beam.
We also performed an experimental determination of the isotherm for cyanide adsorption
for on AuNPs and AuNP coated LoBs. The isotherm for cyanide on AuNPs shown in Figure
2.2A exhibits a combination of Frumkin behavior associated with adsorption of charged species
23
at a charged surface, and loss of gold due to dissolution. Figure 2.2B shows the isotherm we
observed for cyanide on our AuNP coated LoBs. Both isotherms have a similar shape with
slightly different dependencies on the cyanide concentration.
Figure 2.2: Cyanide adsorption isotherms for colloidal AuNP (top) and LoB particles (bottom). The k values are calculated from the slope between the first and second data points. The LOD was detected from 3 σ/m.
We found the adsorption coefficient, k, to be quite different from the 0.16 ppb-1 reported
by Tessier, et al.8 Our values calculated from the slope at low concentrations for AuNPs and
LoBs are 0.0059 ppb-1 and 0.0051 ppb-1, respectively. The 30 smaller values for the cyanide
adsorption on our particles may be explained by their surface structure and the pH difference of 9
in our study and 10 in their study. The pKa is 9.5 for HCN and this favors a high pH to keep the
24
solution species as CN-. However, Tessier reported similar k values for both low and high pH
values since the adsorption process is for CN-. Additionally, the Au surface developed by
Tessier is a planar substrate with AuNP coated polystyrene spheres. While Tessier et al. did not
discuss other materials on their AuNPs we observed strongly bound citrate that did not change
intensity through our isotherm titrations.
The zeta potential of our nanoparticles created using the Frens6 protocol is approximately
-35 mV indicating strongly adsorbed citrate. The strong negative charge will repel CN- causing
k to be lower than that from a neutral surface. This may contribute to the smaller k values we
observed. The CN peak we observed is at the same location as reported elsewhere, 2125 cm-
1,8,9,10 and the citrate peaks we observed were also located at the same wavenumbers that other
groups had observed.11,12 Our spectra, shown in Figure 2.3, have citrate peaks at the same
locations noted by Siiman et al.,12 who also reported that the citrate is strongly adsorbed and did
not change in composition or intensity over pH ranges from 2.8 to 9.9. Clearly the saturation of
our surfaces does not represent 100% of the surface coated with cyanide, but rather, the fraction
that is not covered with citrate. Repulsion of CN- by our citrate coated AuNPs may be the best
explanation for the difference in our observed k values relative to the study by Tessier et al.
Tessier et al. reported LOD values of 210 parts-per-trillion (ppt) at high pH. Our values
are similar with 180 ppt for colloidal AuNPs and 173 ppt for LoBs. The sharp drop off of CN-
coverage at the < 100 ppb solution concentration level will dictate the LOD in terms of the slope.
However, Figure 2.1A demonstrates that the σ value increases exponentially for AuNPs. To
alleviate this problem we performed these experiments with a relatively high AuNP
concentration (1.8 x 109 AuNP/mL) and we used a Raman system with a large 1 mm raster area
25
(Sierra ORS, Snowy Range Instruments) to eliminate noise created by dynamic AuNP motion.
The isotherm in Figure 2.2B was collected with identical acquisition parameters and 1000 LoBs.
The cyanide system used in this study demonstrates LoB assays with a fairly weak
reversibly binding species. An examination of the theoretical intensities predicted for colloidal
AuNPs demonstrates a further advantage of the LoB assay. This can be seen from the following
derivation:
I = FΘN
where I (photons/sec) describes the SERS intensity from an analyte from an AuNP colloid with a
fractional analyte coverage of θ and N nanoparticles/mL. F is a factor which converts coverage
into Raman intensity. Assuming a Langmuir isotherm and solving this equation for I as a
function of the number of nanoparticles provides a model to better understand AuNP SERS
assays. Of particular interest are the cases when the analyte concentration c0 is low and the
adsorption coefficient, k, is large. In this case θ is no longer dictated by c0 as the amount of
material adsorbed onto the surface becomes a significant fraction of the total amount of analyte
in the solution. We solved for I as a function of c0 and produced an equation to calculate the
effect of analyte depletion by the AuNPs.
26
Figure 2.3: SERS spectra of cyanide and citrate on LoB and AuNPs. These spectra indicate that citrate is not being displaced by the adsorption of cyanide.
Figure 2.4 illustrates the interplay between k and θ as a function of the number of
particles present. As the concentration of nanoparticles decreases it can be seen that the
coverage increases and as k increases the coverage increases. Intuitively this result is not
surprising; but since θ increases with fewer colloidal AuNPs this result dramatically illustrates
the difficulty of colloidal AuNP assays. For example, the data in Figure 2.1A begins at 3.2 x 106
AuNP/mL and it already is showing significant fluctuations due to dynamic motion into and out
of the laser beam. This simple model predicts that a fundamental limitation occurs as noise
increases while surface coverage increases. Although this may not be observed in a system
examining fairly high concentrations, it will be the fundamental limit of a system examining
trace levels of materials.
27
Figure 2.4: An illustration of the theoretical coverage vs. k. The curves relate to the concentration of nanoparticles in a given sample.
To demonstrate the value of LoBs with a neutral adsorbate and a high k, we chose the
popular tag, 5,5’-dithiobis(2-nitrobenzoic acid) (DTNB). Grubisha, et al.13 reported femtomolar
detection of prostate-specific antigen with the succinimide derivative of DTNB. Specifically,
Grubisha used immobilized particles on a glass slide to avoid aggregation effects from AuNPs in
solution and their ultimate detection limit was detected hypothetically by looking at a ratio of the
22 micron laser beam spot and the 5 mm spot of immobilized AuNPs used in the study. Our
experiment with DTNB consists of a comparison of colloidal AuNPs and LoBs. Figure 2.5
illustrates the signal difference from LoBs and colloidal AuNPs under conditions with an
equivalent amount of AuNP in both analyses. In other words, this demonstrates the
concentration benefit of detecting a single LoB rather than colloidal nanoparticles in a small
beam volume. At 5 μM DTNB we observe a signal that is 28x larger on the LoB than the
colloidal AuNPs. We also do not observe citrate at this concentration as it is displaced from the
AuNP surface by the strongly binding DTNB. This difference can be easily understood from the
28
study by Pierre et al.3 using 2-naphthalene thiol (2-NT). In their study with 2-NT Pierre et al.
found that displacement of the citrate by the strong thiol adsorption led to a time-dependent
signal due to aggregation. The LoB has a stable aggregated surface of AuNPs and through
agitation has the ability to interrogate the solution for DTNB. The colloidal AuNPs are stable
when citrate is strongly adsorbed, but rapidly aggregate and fall out of solution as DTNB is
adsorbed and AuNP surface charge is neutralized.
Figure 2.5: Representative SERS spectra of DTNB at equal concentration on a mass equivalent amount of 50 nm AuNPs. The LoB bound AuNPs do not aggregate and fall out of solution. The colloidal AuNP particles do aggregate and their signal is lost.
The number of LoBs observed in our DTNB experiment is 1. Our 25 μm laser beam is
smaller than a single LoB. We used 200 LoBs on our experiment and made two observations:
we can translate across the surface of our droplet and see signal variations that indicate we are
detecting individual LoBs; and we examined the droplet with a light microscope and found that
our 200 LoBs were uniformly distributed in a monolayer. The localization of our LoB particles
at the top of a droplet is equivalent to the creation of a pellet by a paramagnetic pull-down. The
ability to mix large volumes of samples with a small number of LoBs which localize rapidly
29
through their buoyant force could be advantageous over the paramagnetic counterpart which
requires an external magnetic force that decays rapidly with the distance from the magnetic.
Further, the available chemistries for Au surface modification present many opportunities for the
LoB concept in sensing applications.
2.4 References
1. Schmit, V. L.; Martoglio, R.; Scott, B.; Strickland, A. D.; Carron, K. T., Lab-on-a-Bubble: Synthesis, Characterization, and Evaluation of Buoyant Gold Nanoparticle-Coated Silica Spheres. J. Am. Chem. Soc. 2012, 134 (1), 59-62.
2. Mallouk, T. E.; Sen, A., Powering Nanorobots. Sci.Am. 2009, 300 (5), 72-77.
3. Pierre, M. C. S.; Mackie, P. M.; Roca, M.; Haes, A. J., Correlating Molecular Surface Coverage and Solution-Phase Nanoparticle Concentration to Surface-Enhanced Raman Scattering Intensities. J. Phys. Chem. C 2011, 115 (38), 18511-18517.
4. Aebersold, R. H.; Teplow, D. B.; Hood, L. E.; Kent, S. B. H., Electroblotting onto Activated Glass. The Journal of Biological Chemistry 1986, 261 (9), 4229-4238. S-1.
5. Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter, D. G.; Natan, M. J., Self-Assembled Metal Colloid Monolayers - an Approach to SERS Substrates. Science 1995, 267 (5204), 1629-1632.
6. Frens, G., Controlled Nucleation for Regulation of Particle-Size in Monodisperse Gold Suspensions. Nature-Physical Science 1973, 241 (105), 20-22.
7. Haiss, W.; Thanh, N. T. K.; Aveyard, J.; Fernig, D. G., Determination of Size and Concentration of Gold Nanoparticles from UV-Vis spectra. Anal. Chem. 2007, 79 (11), 4215-4221.
8. Tessier, P. M.; Christesen, S. D.; Ong, K. K.; Clemente, E. M.; Lenhoff, A. M.; Kaler, E. W.; Velev, O. D., On-Line Spectroscopic Characterization of Sodium Cyanide with Nanostructured Gold Surface-Enhanced Raman Spectroscopy Substrates. Appl. Spectrosc. 2002, 56 (12), 1524-1530.
9. Premasiri, W. R.; Clarke, R. H.; Londhe, S.; Womble, M. E., Determination of Cyanide in Waste Water by Low-Resolution Surface Enhanced Raman Spectroscopy on Sol-Gel Substrates. Journal of Raman Spectroscopy 2001, 32 (11), 919-922.
10. Shelton, R. D.; Haas, J. W.; Wachter, E. A., Surface-Enhanced Raman Detection of Aqueous Cyanide. Appl. Spectrosc. 1994, 48 (8), 1007-1010.
30
11. Kerker, M.; Siiman, O.; Bumm, L. A.; Wang, D. S., Surface Enhanced Raman-Scattering (SERS) of Citrate Ion Adsorbed on Colloidal Silver. Appl. Optics 1980, 19 (19), 3253-3255.
12. Siiman, O.; Bumm, L. A.; Callaghan, R.; Blatchford, C. G.; Kerker, M., Surface-Enhanced Raman-Scattering by Citrate on Colloidal Silver. J. Phys. Chem. 1983, 87 (6), 1014-1023.
13. Grubisha, D. S.; Lipert, R. J.; Park, H. Y.; Driskell, J.; Porter, M. D., Femtomolar Detection of Prostate-Specific Antigen: an Immunoassay Based on Surface-Enhanced Raman Scattering and Immunogold Labels. Anal. Chem. 2003, 75 (21), 5936-5943.
31
3 Lab-on-a-Bubble Surface Enhanced Raman Indirect Immunoassay for Cholera1
3.1 Introduction
Surface Enhanced Raman Scattering (SERS) assays are effective analytical methods due
to the robustness of properly prepared nanoparticle materials2; the large dynamic range of single
molecules to high analyte concentrations3; the selectivity of Raman spectroscopy; and
development of small portable Raman devices to read the assays4. We recently demonstrated an
interesting direct SERS assay that employed buoyant silica bubbles derivatized with gold
nanoparticles (AuNP)5. It was demonstrated that the buoyancy could pull the AuNP coated silica
bubbles, coined Lab-on-a-Bubble (LoB), from the sample volume to a compact monolayer of
LoBs on the surface of the sample.
Direct SERS assays have been demonstrated with colloidal AgNP or AuNP, SERS active
substrates, and with AuNP modified paramagnetic particles. Many schemes have been used to
enhance the adsorption of analytes to the fairly unreactive noble metal surfaces. The
significance of the LoB direct assay concept stems largely from the stability of the nanoparticle
coating in contrast to the inherent instability of colloidal particles.
32
Figure 3.1: Conceptualization of an indirect LoB assay for cholera. The components (left) consist of a cholera-antibody derivatized silica bubble (LoB), the cholera-antigen (CT-AG), and an antibody derivatized silica shelled AuNP reporter. For this project, the Raman reporter is 1,2-bis(4-pyridyl)ethylene (BPE). The resulting reaction between antigen and the LoB components is illustrated to the right.
The relative dimensions are exaggerated to show the AuNP reporters. Multiple reporters/bubbles are possible and were observed by SEM imaging.
Figure 3.1 illustrates a LoB indirect assay. This assay, rather than utilizing AuNP coated
LoBs, has LoBs that are coated with an analyte binding reagent. The analyte contains multiple
binding sites such that it can also bind to an AuNP reporter (NPR) coated with analyte binding
reagents. The NPR consists of an AuNP core, single or multiple AuNPs, covered with a
submonolayer coating of a coupled strong Raman scatterer, and a protective shell of SiO2. The
NPRs have the advantage of robustness in comparison to a colloidal AuNP. The relative area of
33
AuNP
CT-AG
LoB
Components Results
the of the silica bubble to the shell nanoparticle is about 4 x 104, making it likely that multiple
analyte bindings can occur at a single LoB.
We chose cholera as the model system to demonstrate a LoB indirect assay. Vibrio
cholerae is the causative agent of cholera, a highly contagious and commonly fatal bacterial
infection of the gastrointestinal tract. Death can occur within hours of infection if not treated
immediately and is usually due to hypovolemic shock or acidosis6. Individuals infected and
actively shedding V. cholerae routinely demonstrate 107 to 108 colony forming units (CFU)/mL
feces. The most common method of identifying cholera in environmental samples is traditional
microbiology: enrich samples for infectious agents by growing them on selective media, and
further selection and identification of a serotype through a series of biochemical tests which take
approximately 8 days for a conclusive determination7. Other tests have been introduced in the
search for a quick and effective V. cholerae identification: Polymerase Chain Reaction (PCR)
following enrichment steps6, direct cell duplexing PCR for immediate identification of infectious
strains8, Digoxigenin labeling (DIG) or radioactive hybridization of colonies for selection of
infectious strains after initial colony growth7, and various immunoassays of V. cholerae colonies
directly imaged by microscopy or Western Blotting9. The US Food and Drug Administration
couples bacterial enrichment steps to PCR identification of pathogenic strains10. A rapid,
accurate diagnostic assay for the presence of CT in either a water sample or a patient sample
would significantly benefit those in outbreak areas.
34
3.2 Materials and Methods
3.2.1 LoB Activation and Antibody Attachment
LoBs (3M S60 glass bubbles) were activated with 10 N sulfuric acid overnight. Bubbles
were silanized with 1:10 3-aminopropyltriethoxysilane APTES in methanol overnight and
washed extensively in methanol (MeOH). Bubbles were resuspended in 3 mL HPLC grade
water. Following APTES silanization, antibodies were activated with the carbodiimide EDC. 1
μg CT Subunit B antibody (anti CT antibody) (Abcam 34992) was added to the reaction with
EDC and activated and silanized LoB solution.
3.2.2 Dynamic Light Scattering (DLS)
Colloidal nanoparticle solutions remain homogeneous for several months due to
Brownian motion of individual particles. The velocity of particles within a solution is a function
of nanoparticle size. Dynamic light scattering determines average particle velocity by measuring
time-correlated fluctuations in the average amount of light scattered by the colloidal solution,
which can be used to calculate nanoparticle diameter. This technique is also useful for
determining the polydispersity of a colloidal sample. DLS measurements were made for bare
colloids and silica-coated nanoparticles. Dynamic light scattering measurements were made
using a ZetaPALs DLS instrument (Brookhaven instruments).
35
Figure 3.2: Example of DLS measurement results, showing the average diameter and polydispersity of a colloidal sample solution.
3.2.3 Raman Reporter Synthesis
Although gold nanoparticle solutions remain stable for several months, addition of
analytes can lead to rapid and irreversible particulate precipitation. One solution to this problem
is to glass-coat the molecule-adsorbed nanoparticles. The result is a stable solution of Raman
reporters which can be further modified (e.g. antibody attachment) for more complex research
applications. The wide array of antibody-antigen combinations permits countless research
possibilities, including pathogen detection, blood glucose monitoring, and detection of
primordial life molecules.
A known method for coating gold nanoparticles with amorphous silica11 was tested using
the synthesized colloids. DLS results showed an increase in particle diameter (~155 nm) and a
36
decrease in polydispersity, indicating successful synthesis of core-shell colloids. The next step
was to coat analyte-adsorbed nanoparticles with silica to make a stable Raman reporter. Thiol
species form a strong bond to metal nanoparticles12 that is unaffected by the silation reaction
process, making it a suitable tag for the core-shell particles. A final diameter of 130-150 nm was
desired to ensure complete silica coverage of the thiophenol-adsorbed nanoparticles while
maintaining SERS properties. Incubation time was adjusted to achieve the desired particle size.
To make these tagged colloids, 4 µL of 1 mM thiophenol (or BPE) was added to 4 mL
gold nanoparticles. This solution was added to 16 mL of 2-propanol while stirring. 500 µL of
ammonia hydroxide, followed by 16 µL of tetraethyl orthosilicate (TEOS) was added to the
reaction mixture to initiate the silation process. After one hour of stirring, the reaction product
was centrifuged for 10 min at 7,200 rpm. The supernatant was poured off and the pellet was re-
suspended in 250 µL H2O. DLS was used to determine the diameter (142 nm) of the thiol-coated,
shelled nanoparticles (Figure 3.3) and a Raman spectrum verified the presence of a strong
analyte peak signature (Figure 3.4) that persisted for several weeks (Figure 3.5).
37
Figure 3.3: Dynamic light scattering results of bare nanoparticles (top) and Raman reporter particles (bottom). Shelled Raman reporters exhibit a larger diameter than bare NPs with little change to the polydispersity.
38
Figure 3.4: Raman spectra of shelled (red) and unshelled (green) colloids in 5 µM BPE.
39
200 400 600 800 1000 1200 1400 1600 1800
Inte
nsi
ty
Wavenumber
Figure 3.5: Raman spectra of thiophenol-adsorbed coated colloids taken on 10/1 (blue), 10/8 (green), and 10/29 (red).
3.2.4 Preparing and Shelling AuNPs
Gold nanoparticles were prepared using the citrate reduction method described by Frens
in 197313. Colloids were sized using SEM and were an average of 50 nm in diameter.
Nanoparticle concentration was determined as described by Haiss et. al14. After UV-vis
spectroscopy and the calculations from that work, we determined the concentration of our
nanoparticles to be 6.02 x 1010 nanoparticles per mL. 4mL fresh colloids were labeled with 50
nM 1,2-bis(2-pyridyl) ethylene (BPE) and added to 20 mL isopropanol (99%) at room
temperature while stirring. Colloids were shelled with silica as detailed in Lu et al.11, 15. The
40
200 400 600 800 1000 1200 1400 1600
Inte
nsity
Wavenumber
SEM image in Figure 3.6D shows that many of the NPR are paired AuNPs. This is significant
as it has been demonstrated that paired AuNPs provide larger enhancements16.
Figure 3.6: SEM images of a positive LoB assay. Images A, B, and C are acquired with refelected electrons to enhance the physical structure of the LoB materials. Image D used backscattered electron detection to visualize the captured AuNP particles. Note that many are AuNP combinations.
3.2.5 LoB Immunoassay
Antibody conjugated LoBs were blocked with nonfat dry milk in PBS and incubated for
10 minutes shaking at room temperature prior to addition to reaction. Shelled, tagged colloids
were incubated with a 1:500 dilution in PBS anti CT antibody (original concentration 1 mg / mL)
and incubated for 20 minutes shaking at room temperature to allow antibodies to adsorb to the
silica surface. Following antibody adsorption, colloids were blocked with nonfat dry milk in
PBS and incubated for 10 minutes shaking at room temperature prior to adding the colloid
component to reaction. Recombinant beta subunit CT (concentration: 1 mg / mL) (Sigma
41
Aldrich C9903) was added at varying concentrations to each reaction. The standard addition
experiment antigen addition description is as follows: (1) Unknown concentration of CT (final
volume in this reaction is 50 μL), (2) Unknown + 2500 ng CT, (3) Unknown + 5000 ng CT.
Antibodies were attached to LoBs in Eppendorf low binding tubes (cat # 0030 108.116) using
EDC. Prior to each assay, antibodies were adsorbed to shelled nanoparticle reporters (NPRs) in
low binding tubes. The LoBs and the NPRs were each added to the reaction tube which was also
a low binding tube. The reactions were incubated shaking for 20 minutes. Following incubation,
the entire reaction volume (85 μL) was transferred to a polished aluminum surface where the
LoBs were allowed to rise to the surface (approximately 5 minutes)5. We did not observe
problems related to evaporation of the droplet in the ~ 5 minute time for LoB floatation and
Raman collection.
3.2.6 Data Acquisition and Analysis
Data were acquired on a Snowy Range Instruments Sierra Raman ORSTM instrument with
an 808 nm rastering laser. By rastering the laser beam over a 2 x 0.5 mm area, the laser spot size
remains small which is a requirement for selectivity in Raman spectroscopy while a larger area is
sampled allowing averaging of possible inhomogeneity. The SEM images in Figure 3.6
illustrates that with the current design, the LoBs appear to have locations where there are many
and few NPRs. This problem is averaged out with the rastering laser. One of the signature peaks
of each Raman tag was chosen for analysis (1600 cm-1 BPE), and 1000 cm-1 glass as an internal
standard was chosen to standardize each data point. The internal standard was a fluorescence
peak generated from the glass of the LoBs. The intensity of the peak from the Raman tag was
divided by the intensity of the internal standard peak to arrive at a standardized intensity for each
42
sample point. This eliminates variations in intensity due to differences in focus in individual
samples. Data from each sample was acquired 5 times to ascertain the standard deviation of the
LoB assay.
3.3 Results
Our Raman measurements were made with an 808 nm Sierra Raman ORS system
(Snowy Range Instruments). This system is capable of maintaining a high etendue with a tightly
focused laser beam, yet it can be adjusted to examine a large sample area. We found that our
LoBs were static and formed a monolayer at the top of the sample droplet, Figure 3.7. Our
focused laser beam’s diameter was approximately 30 µm or about the size of one silica bubble.
We performed a mock assay and obtained a micrograph of the bubbles. We counted the bubbles
in the assay and found a monolayer of ~1000 bubbles. In a monolayer, this equates to a diameter
of 1 mm. We tuned our raster circuitry to produce a spherical pattern of slightly larger than 1
mm to capture the signal from all of the LoBs.
The cholera assay was performed on a droplet placed on an aluminum surface to create a
curved surface to focus the LoBs at the surface, see Figure 3.7A. The underlying concept is that
the indirect LoB assay is to concentrate the positive assays, bubbles conjugated to shelled NPRs,
and to separate the signals from the conjugated NPRs from the unconjugated. Our shelled NPRs
have a density of 2.95 g/cm3, using 200 nm for the SiO2 shell diameter and 50 nm for the AuNP
particle diameter. This causes them to rapidly sink and interfere with the results of a
paramagnetic or centrifugal pull-down assay. Our optical method scans the top of the droplet
and locates the positive LoBs. The focus of the beam and the opacity of the LoBs differentiates
between the silica bubbles on top of the droplet and the material near the bottom. Figure 3.7B
43
illustrates that the focusing of the particles will also produce a spatial differentiation as the
unbound NPRs will disperse to a larger area in the sample.
Figure 3.7: Schematic of the Raman measurement method used in our assay. A) side view illustrating the spatial separation between LoBs and unconjugated AuNPRs. B) Top view illustrating further spatial separation between the focused LoBs and the dispersed AuNPRs.
SEM analysis of the assay materials demonstrates that the assay consists of multiple
AuNPs in each shell and that a single silica bubble binds with multiple NPRs, see Figure 3.6. An
SEM/Raman study by Wustholz et al. demonstrated that the local surface plasmon resonance
(LSPR) responsible for the SERS enhancement shifts with the number of AuNPs and their
orientation16. Their assumption is that the large SERS signals observed from dimers and
multimers stem from single molecules in the AuNP junctions. Our shelled NPRs also show a
large number of dimers and multimers; Figure 2D has 3 monomers, 2 dimers, and one
quadramer.
44
Figure 3.8A is the spectrum obtained from 1 x 104 ng of CT in a LoB assay. The peak
around 1000 cm-1 is due to luminescence from the silica bubbles. We observed this peak in silica
with NIR excitation and it is very strong with 808 nm excitation. We used this as an internal
control to account for the number of LoBs at the droplet’s apex. This accounts for LoBs lost
during the assay development and transfer to the sampling surface. The 1600 cm-1 peak stems
from the reporter molecule, 1,2-Bis(2-Pyridyl) Ethylene (BPE).
Figure 3.8: Assay results for CT. A) Raman spectrum from 10 μg CT pulled out with LoBs and NPRs. B) Standard addition plot with calculated limit of detection.
Cholera detection is commonly required in water supplies or stool samples. Both cases
present a complex sample matrix. Additionally, the CT antigen used for assay development
contains stabilizers and preservatives that affect our assays. We used standard additions to
45
account for interactions between the matrix and the analyte. Figure 3.8B is the standard addition
graph obtained from our experiments. The value of the unknown is found by:
[c] = b/m
where [c] is the unknown concentration, b is the y-intercept, and m is the slope.
The y axis in our plot is the ratio of the silica emission peak around 1000 cm-1 and our
reporter molecule, BPE, peak around 1600 cm-1. Using this method and a linear regression, we
found our predicted unknown to be 3700 ng (actual 5000 ng). The limit of detection (LOD) was
found to be 1100 ng from the linear regressions predicted error in the y-intercept and the slope:
LOD = 3 (σ/m)
where σ in this case is the predicted error in the y-intercept. This may slightly overestimate the
LOD as the calculated predicted error in the y-intercept includes the errors of all the data and
since we see significant heteroscedasticity in the data. However, the calculation provides a
reasonable approximation.
The heteroscedasticity is interesting. It is nearly 20 times larger than the predicted
spectroscopic noise from the signals. We suggest that it is due to the variations in the signals
due to loss of particles during the assay and the transfer of particles to the sampling surface.
This error should be larger when the silica LoBs contain more NPRs. In other words, the loss of
10% of the highly positive LoBs will result in a larger error than 10% of a low positive assay.
All results are discussed as mass rather than concentration since the buoyant LoBs enable
us to detect mass independent of volume. The LoBs will concentrate on top of whatever volume
46
is in the sample. We see this as a significant benefit as the concentration (analyte/volume)
should be very low for samples with large volumes.
Diagnostic assays are not commonly used in developing countries. Reagents are often
refrigerated, trained personnel must operate the instruments, and much laboratory equipment is
required to run diagnostic tests. The LoB platform for the sandwich assay frees the tests from
any volume limitation that the magnet strength would dictate in traditional paramagnetic assays.
It also decreases the likelihood of finding false positives from contamination of the sample to be
interrogated with the NPRs.
There are a number of reports of potentially commercial CT tests in the literature, but we
found only assay, a Lateral Flow Immunoassay (LFI), the SMART Cholera 0117 , which is
actually commercially available. The Cholera 01 SMART II LFI reports an LOD at 2 x 107
colony forming units (CFU) per mL17 and Spira and Fedorka-Cray found that there are
approximately 0.19 fg/CFU Cholera toxin in Vibrio cholerae 0118 . This places their detection
limit at 3.9 ng/mL of CT.
While this appears to be much lower than ours mass detection limit, we do have the
advantage of detecting small levels of CT in large volumes. Additionally, this is proof of
concept study and report that has not been optimized for number of LoBs, antibodies, or
experimental conditions.
Many research groups provide CT detection limits that fluctuate widely. This is not a
comprehensive literature review, but a few CT detection limits are: 1 nM CT on a biosensor19,
from 1 ng/mL to 0.49 ng/mL using ELISAs20,21, sandwich (indirect) assays were reported at 40
ng/mL and 1 μg/mL while direct assays were reported at 200 ng/mL22. Schofield et al. reported a
47
detection limit of 3 μg/mL using glyconanoparticles in a colorimetric assay23 making their
detection limit around 4 μg Cholera toxin.
3.4 Acknowledgements
The authors would like to thank Snowy Range Instruments for the instrumentation and
facility usage. Dr. Martoglio acknowledges the support of DePauw University for his sabbatical
leave.
3.5 References
1. Schmit, V. L.; Martoglio, R.; Carron, K. T., Lab-on-a-Bubble Surface Enhanced Raman Indirect Immunoassay for Cholera. Anal. Chem. 2012, 84 (9), 4233-4236.
2. Penn, S. G.; He, L.; Natan, M. J., Nanoparticles for Bioanalysis. Curr Opin Chem Biol 2003, 7 (5), 609-615.
3. Nie, S.; Emory, S. R., Probing Single Molecules and Single Nanoparticles by Surface-Enhanced Raman Scattering. Science 1997, 275 (5303), 1102-1106.
4. Carron, K.; Cox, R., Qualitative Analysis and the Answer Box: a Perspective on Portable Raman Spectroscopy. Anal Chem 2010, 82 (9), 3419-3425.
5. Schmit, V. L.; Martoglio, R.; Scott, B.; Strickland, A. D.; Carron, K., Lab-on-a-Bubble: Synthesis, Characterization, and Evaluation of Buoyant Gold Nanoparticle-Coated Silica Spheres. JACS 2011, e pub ahead of print (2011 Nov 18).
6. Kaper, J. B.; Morris, J. G.; Levine, M. M., Cholera. Clinical Microbiology Reviews 1995, 8 (1), 48-86.
7. Robert-Pillot, A.; Saron, S.; Lesne, J.; Fournier, J.-M.; Quilici, M.-L., Improved Specific Detection of Vibrio Cholerae in Environmental Water Samples by Culture on Selective Medium and Colony Hybridization Assay with an Oligonucleotide Probe. FEMS Microbiology Ecology 2002, 40, 39-46.
48
8. Goel, A. K.; Tamrakar, A. K.; Nema, V.; D.V., K.; Singh, L., Detection of Viable Toxigenic Vibrio Cholerae from Environmental Water Sources by Direct Cell Duplex PCR Assay. World J Microbiol Biotechnol 2005, 21, 973-976.
9. Wang, D.; Xu, X.; Deng, X.; Chen, C.; Li, B.; Tan, H.; Wang, H.; Tang, S.; Qiu, H.; Chen, J.; Le, B.; Ke, C.; Kan, B., Detection Of Vibrio Cholerae 01 and 0139 in Environmental Water Samples by Immunofluorescent Aggregation Assay. Applied and Environmental Microbiology 2010, 76 (16), 5520-5525.
10. FDA Bacteriological Analytical Manual (BAM). http://www.fda.gov/Food/ScienceResearch/LaboratoryMethods/BacteriologicalAnalyticalManualBAM/default.htm.
11. Lu, Y.; Yin, Y. D.; Li, Z. Y.; Xia, Y. N., Synthesis and Self-Assembly of Au@SiO2 Core-Shell Colloids. Nano Lett. 2002, 2 (7), 785-788.
12. Carron, K.; Peitersen, L.; Lewis, M., Octadecylthiol-Modified Surface-Enhanced Raman-Spectroscopy Substrates - a New Method for the Detection of Aromatic-Compounds. Environ. Sci. Technol. 1992, 26 (10), 1950-1954.
13. Frens, G., Controlled Nucleation for the Regulation of the Particle Size in Monodisperse Gold Suspensions. Nature 1973, 241 (105), 20-22.
14. Haiss, W.; Thanh, N. T. K.; Aveyard, J.; Fernig, D. G., Determination of Size and Concentration of Gold Nanoparticles from UV-Vis Spectra. Anal. Chem. 2007, 79 (11), 4215-4221.
15. Lu, Y.; Yin, Y. D.; Mayers, B. T.; Xia, Y. N., Modifying the Surface Properties of Superparamagnetic Iron Oxide Nanoparticles through a Sol-Gel Approach. Nano Lett. 2002, 2 (3), 183-186.
16. Wustholz, K. L.; Henry, A.-I.; McMahon, J. M.; R.G., F.; Valley, N.; Piotti, M. E.; Natan, M. J.; Schatz, G. C.; Van Duyne, R. P., Structure-Activity Relationships in Gold Nanoparticle Dimers and Trimers for Surface-Enhanced Raman Spectroscopy. Journal of the American Chemical Society 2010, 132.
17. Diagnostics, N. H. SMART Cholera 01 LFI. http://www.nhdiag.com/cholera_bt.shtml.
18. Spira, W. M.; Fedorka-Cray, P. J., Enterotoxin Production by Vibrio Cholerae and Vibrio Mimicus grown in Continuous Culture with Microbial Cell Recycle. Applied and Environmental Microbiology 1983, 46 (3), 704-709.
19. Singh, A. K.; Harrison, S. H.; Schoeniger, J. S., Gangliosides as Receptors for Biological Toxins: Development of Sensitive Fluoroimmunoassays Using Ganglioside-Bearing Liposomes. Analytical Chemistry 2000, 72 (24), 6019-6024.
49
20. Uesaka, Y.; Otsuka, Y.; Kashida, M.; Oku, Y.; Horigome, K.; Nair, G. B.; Yamasaki, S.; Takeda, Y., Detection of Cholera Toxin by a Highly Sensitive Bead Enzyme Linked Immunosorbent Assay. Microbiology and Immunology 1992, 36 (1), 43-53.
21. Edwards, K. R.; March, J. C., GM1 Functionalized Liposomes in a Microtiter Plate Assay for Cholera Toxin in Vibrio Cholerae Culture Samples. Anal Biochem 2007, 368 (1), 39-48.
22. Rowe-Taitt, C.; J., C.; Patterson, C.; Golden, J.; Lingler, F., A Ganglioside Based Assay for Cholera Toxin Using an Array Biosensor. Analytical Chemistry 2000, 281 (1), 123-133.
23. Schofield, C. L.; Field, R. A.; Russell, D. A., Glyconanoparticles for the Colorimetric Detection of Cholera Toxin. Analytical Chemistry 2007, 79 (4), 1356-1361.
50
4 Dynamic SERS: Extracting SERS from Normal Raman Scattering1
4.1 Introduction
Conventional Raman spectrometers improve signal-to-noise by integration of signal in
the wells of CCD chips. With proper cooling and readout circuitry this approach leads to optical
detection that follows Poisson statistics for shot noise-limited spectra. Therefore, within a
spectrum, the variance in the signal is equal to the intensity. When individual spectra are
compared, the dominant source of variation is rms laser noise which follows a normal
distribution and is reduced through spectral averaging. However, this approach of time
indiscriminate signal collection places photons from every possible source into the spectrum.
Conventional Raman spectra contain signal contributions from the desired source in the sample
as well as fluorescence, whether intrinsic or an impurity, stray light from the optical system and
Raman interference from sample containers. Time correlation has been demonstrated as a way
to discriminate between the instantaneous scattering events and delayed fluorescence
signals2{Willis, 1990 #2}.
Colloidal nanoparticles are free floating particles that remain suspended through
Brownian motion. Surface enhanced Raman spectroscopy (SERS) from colloidal nanoparticles
was described very soon after the initial discovery of SERS at electrode surfaces3. The ease of
making colloidal gold and silver particles has made it a popular method for performing SERS
studies and analytical assays4. Additionally, the large velocity imparted on nanoparticles through
Brownian motion leads to an opportunity to discriminate between their spectroscopic signals and
the relatively rapid fluctuations of free molecular species and continuum produced by solid state
interferences.
51
We describe a statistical method for specific extraction of SERS signals from colloidal
SERS active nanoparticles. The difference in these particles’ sizes relative to the molecular
matrix creates an opportunity to statistically differentiate between their signal and the relatively
time indiscriminate fluorescence and matrix Raman signals.
4.2 Results and Discussion
4.2.1 SERS Signal Extraction
Figure 4.1 illustrates the concept of dynamic SERS (DSERS) spectroscopy. The box on
the left illustrates dynamic processes that lead to the theory of DSERS. Raman spectrometers
typically have a tightly focused laser beam to generate the Raman scattering. That small focal
volume is illustrated as the pink cylinder in figure 4.1. This volume of solvent generates a
Raman signal that is shot noise limited and has a standard deviation equivalent to the square root
of the signal. SERS signals are generated by particles moving rapidly into and out of the laser
beam. These fluctuations produce a noise level (σSERS) greater than the square root of the
average signal. The signal (SExcess) due to the excess noise contributed by the dynamic noise from
the SERS active nanoparticles can be found from the difference between the total noise in the
signal (σTotal). and total signal (STotal). The subtraction requires a factor (a) to account for the
difference between the magnitude of the standard deviation and average signals.
The spectra in Figure 4.1 (right) illustrate the results of a DSERS measurement. The top
spectrum (STotal) is from a toluene solution with approximately 8 x 105 particles/cm3 of SiO2
coated nanoparticles with a BPE coating. At this concentration, the presence of the nanoparticles
is undetectable in the average SERS spectrum which is derived from 100 spectra acquired for
100 ms. The middle spectrum (σTotal) represents the standard deviation of the 100 spectra at each
52
data point. This spectrum is still dominated by the variation produced by the laser’s rms power
fluctuations; the variation between the individual spectra is dominated by the laser fluctuations.
This signal independent noise contribution will produce a noise spectrum which has feature
intensities that have values from all sources. The important exception of the instrumental noise
sources is the nanoparticles’ SERS signal. Subtraction of the averaged spectrum (STotal) from the
total noise spectrum (σTotal) divided by the number of averages, 100 in this case, produces the
excess noise spectrum Sexcess. This is shown in the bottom spectrum of Figure 4.1 and closely
represents a SERS spectrum of BPE.
Figure 4.1: DSERS concept. Left) This schematic illustrates a colloidal nanoparticle moving through a focused laser beam. The standard deviation of the continuum, σcontinuum, will scale as the square root of the intensity while the σSERS from the nanoparticle will be larger. Right) An illustration of the signals and standard deviations for a solution of toluene with two nanoparticle events in 10 s.
Examination of the original data set shows that we observed only one major particle
event during the 10 s of data acquisitions. This is observed in Figure 4.2 (top) where an overlay
53
Sexcess = σTotal- aSTotal
σcontinuum = I1/2σSERS > I1/2σTotal = σSERS + σcontinuumWavenumbers
sTotal
σTotal
sExcess
800 1200 1600
10000 ADU
100 ADU
10 ADU
of the 100 spectra in the 1600 cm-1 region indicates that a large event occurred (red) and a
smaller event occurred (violet) during the data collection. Plotting the 1640 cm-1 data point in
time space, Figure 4.2 (bottom), shows the two events in spectrum 67 (major) and 21 (minor)
respectively.
Figure 4.2: Individual Raman spectra from Figure 1 and a plot of intensity at 1640 cm-1 vs the acquisition number.
Most significant about this aspect of the DSERS method is that it removes the interfering
spectral features. Figure 4.1 illustrated this with the observation of a single nanoparticle’s SERS
54
Time10 20 30 40 50 60 70 80 90
50 ADU
500 ADU
Wavenumbers1525 1575 1625 1675
1640 cm-1
spectrum in a neat toluene matrix. In this case, we were able to extract a SERS spectrum with a
one part per thousand relative intensity. The value of this method is its objective (autonomous)
derivation of the pure SERS spectrum in the presence of the overwhelming solvent spectrum.
Even selection of the individual spectra with the nanoparticle present requires subtraction of
toluene of a pure matrix spectrum with an unknown relative intensity to the SERS intensity.
4.2.2 Sites Selective Spectroscopy
Hotspots between nanoparticle aggregates or gaps between nanoparticle features have
been discussed as a possible mechanism for very large enhancements beyond those observed
from single particles5. Examples of experiments to prove this theory have included SEM
combined with LSPR spectroscopy6 and tilted pillar experiments which show larger signals when
pillars are collapsed to produce contact7. The difficulty of proof is the differentiation between
the SERS signal from the majority of the surface’s coverage and that of the small number of
molecules in the gap region. Even with large gap enhancements, the small area associated with
this enhancement will lead to relatively small signals that are difficult to detect in the total SERS
signal.
The challenge of site-selective nanoparticle spectroscopy is that the observed signals are
derived from an ensemble of particles in the laser beam during the integration period. Schmit et
al. recently showed the paradox between signal and fluctuation-induced noise in solution phase
nanoparticle spectroscopy8. As the number of particles decreases, the signal decreases, and the
fluctuation-produced noise, as described by Brownian induced fluctuations, increases. Increased
acquisition times only exacerbate the problem by allowing more particles to traverse the laser
beam and to enter into the observed signal. The DSERS method described herein exploits the
55
negative effect of Brownian motion-induced fluctuations and enhances the individual particle or
site selective signals.
Shelled nanoparticles have particular application as bright reporters to sandwich
paramagnetic9 or lab-on-a-bubble assays8, 10. Direct SERS assays are commonly reported for
chemical analysis and are also affected by the degree of aggregation in the sample. Knowledge
of the site of adsorption and the signal from strongly enhancing sites is valuable for assay
development. For example, if specific sites enhanced more than others and a site specific
spectroscopy existed, then the possibility of more sensitive assays could be realized. The
sensitivity of an assay can be described by the magnitude of the signal produced by an analyte
molecule relative to the noise. If a site selective chemistry can be developed specifically at the
“hotspots” of SERS active nanoparticles, the number of active sites will be dramatically
decreased. In this case, as the number of analyte molecules approaches zero, the signal from
adsorption at hotspots will be higher than it would be at adsorption to poorly enhancing spots,
even at a higher concentration of these poorly enhancing locations.
We performed a second study with unshelled nanoparticles coated with 4MP. Mullen et
al11. demonstrated that the ratio of peaks in the 1000 to 1100 cm-1 region of 4MP SERS spectra is
pH dependent; the ratio of the 1091 cm-1 peak to the ring breathing mode at ~ 1000 cm-1 is
smaller under basic (unprotonated) conditions12. These results are reproduced here and are
illustrated in Figure 4.3(left). We found that the average (SERS) spectra of 4MP-coated Au
nanoparticles exhibiting a ratio of 1091 cm-1/1000 cm-1 is 0.87 at high pH (9) and 2.12 at low pH
(5). This is illustrated in Figure 4.3 A,B. It is important to report that we observed small, but
significant, frequency shifts in the ring breathing mode upon protonation.
56
The DSERS spectra (Figure 4.3 C,D) exhibit very different results. Absent in the
DSERS spectra are the broad interfering contributions from the glass sample vial. This confirms
the ability of DSERS to remove normal Raman interferences discussed above. More
significantly, the DSERS spectra of 4MP are nearly identical in base and acid. In this case, a
drastic deviation from the SERS and DSERS spectra is observed.
Figure 4.3E illustrates the variation between two spectra in the 1000 spectra data set for
pH 9. The spectra come from acquisitions 19 (green circle) and 86 (red circle) illustrated in
Figure 4.3F. The relative intensities of the 1000 and 1091 cm-1 peaks to other features are not
distorted; the anomalous equality of the spectra in Figure 4.3 C and D does not appear to be due
to anomalous particles, rather irregular variations in the intensities of these peaks over an
ensemble of particles. While the sampling of spectra in Figure 4.3E demonstrates large
variations in the 1091 cm-1 /1010 cm-1 ratio, they indicate extremes in the variations and clearly
do not correlate to the spectrum in Figure 4.3C. None of the single acquisitions making up
Figure 4.3A correspond to Figure 4.3C. The equality of the acid and base DSERS spectral
features between 1000 and 1100 cm-1 peak must be due to a small population of identical
molecules present on particles in both the acid and base solutions. Not only are the solvent
spectral features and the glass vial’s features removed, but also the SERS features that are
common to all particles. Note: this experiment was performed with a higher concentration of
nanoparticles than the shelled BPE-coated nanoparticle study. This will lead to a reduction in the
nanoparticle peaks and will enhance the signals from variations between the particles.
57
Figure 4.3: Experimental results for 4MP on Au nanoparticles at basic and acidic pH. A) The average spectrum of 1000-100 ms acquisitions at pH 9; B) The average spectrum of 1000-100 ms acquisitions at pH 5; C) DSERS spectrum from the data set used to produce A; D) DSERS spectrum from the data set used to produce B). E) Two individual acquisitions spectra; F) Intensity vs time subset of the 1000 acquisition at 1091 cm-1.
Figure 4.4 shows an expanded view of the data in Figure 4.3. We observed the ring
breathing peak of 4MP at 1003.9, 1005.7, 1007.7, and 1010.2 at pH 5 (SERS), pH 5 (DSERS),
pH 9 (DSERS), and pH 9 (SERS), respectively. The Raman shifts indicate that the species
observed in the DSERS is inaccessible to protonation and are not located at either the acid or the
base spectral shifts. The most likely conclusion is that we are observing excess noise due to a
small population of adsorbate and, given the inaccessibility of the pyridyl nitrogen to
protonation, these species are not exposed to the solvent. This would be consistent with a model
of SERS involving super enhancements of species in the gap between aggregates or in roughness
features on particle surfaces6. In conventional spectroscopy these molecules would not be
observable due to the large population of species not in the highly enhancing gap relative to the
58
900 1100 1300 1500 1700 Wavenumbers
2.12
0.87
1.81
1.96
50 100 150 200 250 300 350 400 450 50 100 150 200 250 300 350 400 450
50 ADU
Acquisition
0.54
750 850 900 950 1000 1050 1100 800
0.78
Wavenumbers
AB
C
D
E
F
number in the gap. This experiment at relatively high nanoparticle concentration is enhancing
those spectral features which are not present on all particles at the same intensity; it represents
SERS signals that are buried in the spectrum of the ensemble of particles or the ensemble of
molecules on a single particle.
950 1000 1050 1100 1150 950 1000 1050 1100 1150
1003.9
1007.7
1010.2
Wavenumbers
pH 9pH 5
pH 9
SERS
DSERS
pH 5
1005.7
Figure 4.4: Magnified spectra from 4.3 A, B, C, D. The ring breathing mode shifts from 1003.9 cm-1 when protonated to 1010.2 cm-1 when deprotonated.
An alternative explanation might be that we are observing 4MP bound to the Au
nanoparticles through its pyridyl nitrogen. This would account for the invariance to solution pH.
However, it is unlikely that statistically significant variations in the population of 4MP bound
through the thiol or through the pyridyl nitrogen would exist between particles. The DSERS
spectrum is statistically significant and more indicative of a small population of aggregates with
59
a small population of strongly enhanced 4MP molecules in the interparticle gap. Figure 4.2
demonstrated that two particles moving into the beam were sufficient to produce a DSERS
spectrum from the overwhelming signal of neat toluene. The data in Figure 4.4 were acquired
from a large number of SERS-active particles in the beam during any individual acquisition and
the DSERS results from variations within this population. If the DSERS were a small population
of sites on every particle, we would expect it to average and not produce an excess noise signal
(Sexcess).
4.3 Conclusion
We have demonstrated two significant benefits of DSERS: removal of instrumental and
normal Raman interferences in SERS spectroscopy and site-selective spectroscopy of adsorbate
populations on SERS-active particles. Our first example of shelled nanoparticles at very low
concentrations confirmed the benefit of DSERS for removal of an overwhelmingly strong
solvent spectral interference. This benefit would be applicable to colloidal SERS studies in
solvents or mixtures that produce strong interferences that might mask observation of the desired
SERS features.
The second benefit, site selection, provides a powerful method to study small populations
of molecules adsorbed on SERS-active particles. In our example with 4MP, we were able to
observe a small population of molecules that were spectroscopically unique from the large
population of molecules on the particles. This study showed the same feature extraction benefit
as described for the shelled nanoparticles but differed in that the DSERS spectra did not match
any of the individual acquisitions or their average. The DSERS spectrum originated from excess
variance between a small population of adsorbates on the ensemble of nanoparticles.
60
4.4 Acknowledgements
The authors would like to thank Snowy Range Instruments for the instrumentation and
facility usage. Brandon Scott would like to acknowledge the NSF GK-12 grant #0948027 for
their kind support.
4.5 References
1. Scott, B. L.; Carron, K. T., Dynamic Surface Enhanced Raman Spectroscopy (SERS): Extracting SERS from Normal Raman Scattering. Anal. Chem. 2012, 84 (20), 8448-51.
2. Willis, K. J.; Szabo, A. G.; Krajcarski, D. T., The Use of Stokes Raman-Scattering in Time Correlated Single Photon-Counting - Application to the Fluorescence Lifetime of Tyrosinate. Photochem. Photobiol. 1990, 51 (3), 375-377.
3. Jeanmaire, D. L.; Vanduyne, R. P., Surface Raman Spectroelectrochemistry. Part 1. Heterocyclic, Aromatic, and Aliphatic-Amines Adsorbed on Anodized Silver Electrode. J. Electroanal. Chem. 1977, 84 (1), 1-20.
4. Siiman, O.; Bumm, L. A.; Callaghan, R.; Blatchford, C. G.; Kerker, M., Surface-Enhanced Raman-Scattering by Citrate on Colloidal Silver. J. Phys. Chem. 1983, 87 (6), 1014-1023.
5. Chen, G.; Wang, Y.; Yang, M. X.; Xu, J.; Goh, S. J.; Pan, M.; Chen, H. Y., Measuring Ensemble-Averaged Surface-Enhanced Raman Scattering in the Hotspots of Colloidal Nanoparticle Dimers and Trimers. J. Am. Chem. Soc. 2010, 132 (11), 3644-+.
6. Wustholz, K. L.; Henry, A. I.; McMahon, J. M.; Freeman, R. G.; Valley, N.; Piotti, M. E.; Natan, M. J.; Schatz, G. C.; Van Duyne, R. P., Structure-Activity Relationships in Gold Nanoparticle Dimers and Trimers for Surface-Enhanced Raman Spectroscopy. J. Am. Chem. Soc. 2010, 132 (31), 10903-10910.
7. Ou, F. S.; Hu, M.; Naumov, I.; Kim, A.; Wu, W.; Bratkovsky, A. M.; Li, X. M.; Williams, R. S.; Li, Z. Y., Hot-Spot Engineering in Polygonal Nanofinger Assemblies for Surface Enhanced Raman Spectroscopy. Nano Lett. 2011, 11 (6), 2538-2542.
8. Schmit, V. L.; Martoglio, R.; Scott, B.; Strickland, A. D.; Carron, K. T., Lab-on-a-Bubble: Synthesis, Characterization, and Evaluation of Buoyant Gold Nanoparticle-Coated Silica Spheres. J. Am. Chem. Soc. 2012, 134 (1), 59-62.
61
9. Wang, X.; Qian, X. M.; Beitler, J. J.; Chen, Z. G.; Khuri, F. R.; Lewis, M. M.; Shin, H. J. C.; Nie, S. M.; Shin, D. M., Detection of Circulating Tumor Cells in Human Peripheral Blood Using Surface-Enhanced Raman Scattering Nanoparticles. Cancer Res. 2011, 71 (5), 1526-1532.
10. Schmit, V. L.; Martoglio, R.; Carron, K. T., Lab-on-a-Bubble Surface Enhanced Raman Indirect Immunoassay for Cholera. Anal. Chem. 2012, 84 (9), 4233-4236.
11. Mullen, K. I.; Wang, D. X.; Crane, L. G.; Carron, K. T., Determination of pH with Surface-Enhanced Raman Fiber Optic Probes. Anal. Chem. 1992, 64 (8), 930-936.
12. Zhuang, Z. P.; Ruan, W. D.; Ji, N.; Shang, X. H.; Wang, X.; Zhao, B., Surface-enhanced Raman scattering of 4,4 '-Bipyridine on Silver by Density Functional Theory Calculations. Vib. Spectrosc. 2009, 49 (2), 118-123.
62
5 Statistical Analysis of 4-Mercaptophenol and Thiophenol on Gold Nanoparticles
5.1 Introduction
The exact mechanism of Surface Enhanced Raman spectroscopy (SERS) remains elusive
due to the many different types of nanomaterials that demonstrate the effect. Initially Raman
was observed at highly roughened silver electrodes and thought to be observable at monolayer
levels due to an increase in adsorbed molecules through greater surface area than a smooth
surface1. Since this initial observation of Raman scattering at a highly roughened surface
Jeanemaire and Van Duyne demonstrated the effect with much less roughening and even with
polished surfaces2. Kerker proved that SERS could be observed on colloidal particles in
solutions and today SERS is performed on a large variety of solid substrates, membrane
materials, colloidal particles, and even aerosols3.
Despite the large number of SERS active materials the origin(s) of the SERS effect still
remain a topic of discussion. The first widely accepted mechanism termed the electromagnetic
enhancement follows from electrostatics of an ellipsoid. This mechanism, termed the
electromagnetic effect, properly ascribed the wavelength dependence of SERS to the dielectric
functions of the SERS metals: copper, silver, and gold. It is also accepted that the formation of
electronic charge transfer states between the metal surfaces and some adsorbates produces a
resonance Raman enhancement termed the chemical effect.
Early two-dimensional gratings and island films4 were shown to produce collective
particle resonances, but were also found not to produce the large enhancements created by
electrodes, colloids, or stochastic surfaces. Experimentally colloidal nanoparticles of silver and
gold produced an interesting contradiction to the early electromagnetic theory: the electrostatic
63
theory of metal dielectric particle predicts a narrow resonance condition and for silver or gold the
condition is highly dependent on shape and is predicted to be at much shorter wavelengths that is
observed. The contradiction is that SERS is observed over a large range of wavelengths for
silver and gold, even when the particle size distribution is very narrow, and it is observed for
spherical particles at much longer wavelengths than predicted. Part of the discrepancy was
accounted for by corrections for particle size.
Dynamic Light Scattering (DLS) and electron microscopy of colloidal preparations
always indicate a minority of particles that are dimers or higher ordered aggregates. The
presence of aggregates anticipates the potential for size and shape effects to the observed SERS
signal. Eccentricity of a dimer that is freely rotating in solution will break the individual
spherical particle plasmon into two enhancements along the long and short axes of the dimer.
Aggregates also create very large fields at the gaps between the individual particles5. This can be
understood simplistically from the 1/r2 dependence of the electric field between two oppositely
charged particles and by considering the instantaneous induced dipole in the nanoparticles. At
very small distances the oscillating field will be large and will produce significant enhancements
(Figure 5.1). This “gap” effect has been elegantly described through plasmonics.
Figure 5.1: Depiction of large electric field in the gap region.
64
The challenge to understanding the SERS effect from single particles and from
aggregates is the small size of the particles relative to the typical laser beam size. Colloidal
nanoparticles remain suspended through Brownian motion and are subject to:
1¿ x2 (t )=kB T3 πrη
t
where x2(t) is the mean square position of the nanoparticle as a function of time, kB is
Boltzmann’s constant, T is temperature, r is the particles radius, η is the viscosity of the solution.
Van Duyne et al.5 examined similar particles with SEM and Raman spectroscopy and
provided theoretical calculations that indicate that dimers or higher ordered aggregates that are
encapsulated should exhibit a larger enhancement. It is possible that the dispersion of
encapsulated nanoparticles, though highly monodispersed around single particle sizes, contains a
small percentage of aggregates and these are what we observe in our SERS spectra. Equation 1
provides a method to observe different particle sizes and correlate their Raman spectra.
We have reported detection of single silica shelled nanoparticles in an organic solvent by
monitoring Raman spectra rapidly, every 150 ms with 100 ms acquisitions, by monitoring their
monition in and out of a focused 25 µm laser beam6. In a low viscosity solvent such as water or
toluene the particles remain in the beam for less than the acquisition time. We have slowed the
particles down with viscous solvents, glycerol, and observed very slow motion, 500 ms or
longer, through the laser beam (Figure 5.2). The shelled nanoparticles were monodispersed with
reported gold cores of 60 nanometers and overall size of 120 nm. Their signal was large due to
matching between the particle plasmon and the 785 nm excitation of the laser beam.
65
0 50 100 150 200 250
Time / s
Figure 5.2: BPE SERS signal intensity vs. time of shelled Raman reporters in glycerine.
In chapter 4 we examined low concentrations to observe single particles and we also
studied higher concentrations to observe unusual events in an ensemble of gold nanoparticles6.
The average spectra in these cases simply reflect the signal of the ensemble and rare SERS
signals presumably from different particles, such as dimers or higher ordered aggregates, are not
observed. We found that we could visualize the rare events in the ensemble by finding the
standard deviation of the ensemble at each wavenumber and removing a normalized average
signal. The normalization factor for the average spectra was obtained by minimizing the
variation in the subtracted spectra. This method works well to reproduce the spectrum of the rare
spectra from the ensemble average.
In this current study we continue to develop statistical methods to study the ensemble of
particles coated with 4-mercaptopyridine and we will examine two new coatings: 4-
mercaptophenol and thiophenol. Our preliminary work on this concept looked at 4-
mercaptopyridine coating since it has been shown to adsorb through its thiol group with the
potential to bind to additional nanoparticles in a “gap” environment. 4-mercaptophenol and
66
thiophenol also adsorb via thiol groups and the former also has the potential for gap binding. The
pKa values of 4-mercaptophenol and 4-mercaptopyridine are 6.8 and 3.9, respectively. It should
be noted that these are solution values. Yu et al. determined a pKa value of 5.3 ± 3 for 4-
mercaptopyridine at the surface of a self-assembled monolayer7 so it is reasonable to assume that
the pKa value of 4-mercaptophenol adsorbed to a surface is also higher than in solution. Within
the range of pH 5-10 thiophenol and 4-mercaptopyridine do not undergo acid-base reactions but
4-mercaptophenol does.
Novel data analysis methods were implemented to monitor fluctuations both within and
between vibrational modes in acidic and basic conditions for each of the three analytes. In
particular we found that 4-mercaptopyridine was invariant to pH. We studied 4-mercaptophenol
and benzenethiol as slightly acidic and neutral coatings in relation to the slightly basic 4-
mercaptopyridine examined previously to look for better control of the gap molecules. Our
ability to control aggregation and to bind specific materials into the unique environment of
nanoparticle aggregates is fundamental to our capability of manufacturing optimal SERS sensor
materials.
5.2 Materials
All chemicals were purchased from the supplier indicated: HPLC grade water (Fischer),
HAuCl4 (reagent grade, Aldrich), sodium citrate dihydrate (99.0%, EMD), 4-mercaptophenol
(99%, Acros Organics), 4-mercaptopyridine (95%, Aldrich), thiophenol (97%, Aldrich), and
sodium bicarbonate. All glassware was cleaned with aqua regia, followed by thorough rinsing
with Milli-Q water.
67
5.3 Experimental
SERS-active stock solutions were made by adding 10 μL of 1 mM thiophenol, 10 mM 4-
mercaptophenol, and 10 mM 4-mercaptopyridine to vials containing 1 mL bare gold
nanoparticles. The acidic sample solutions (pH 5) were made by adding 0.2 mL of stock solution
to 1.8 mL of 1% (w/v) sodium citrate in water. The basic sample solutions were made by adding
0.2 mL of stock solution to 1.8 mL of 1 M sodium bicarbonate in water. The pH was measured in
the final colloidal solution using pH indicator paper.
5.4 Instrumentation
Raman spectra were acquired with an IM-52 Raman microscope (Snowy Range
Instruments) using its liquid sampling feature. The IM-52 was set to 40 mW of 785 nm laser
excitation at the sample with 8 cm-1 spectral resolution. The IM-52 permits multiple spectra to
be acquired with a delay between acquisitions. An integration time of 250 ms with a delay of
250 ms were used to collect 1000 spectra for the sample solution. Particle size results were
collected by dynamic light scattering (Brookhaven Instruments) using ZetaPALS particle sizing
software.
5.5 Data Analysis
The data were analyzed using Excel to generate statistical results as a correlogram
(Figure 5.3). Prominent peaks were identified from a single spectrum and peak maxima for the
remaining spectra were identified within a ±10 cm-1 window of the reference peaks. Along the
diagonal line of the correlogram lie the marginal distributions of intensity for each peak. Each
marginal distribution is comprised of 1000 data points which correspond to the measured peak
68
intensities for a given peak. Data points of equal intensity are summed as counts for that specific
intensity and the resulting data is plotted as counts vs. intensity. The central limit theorem (CLT)
states that, given certain conditions, the arithmetic mean of a sufficiently large number of iterates
of independent random variables, each with a well-defined expected value and well-defined
variance, will be approximately normally distributed8. The marginal distributions contained
within the generated correlograms are not always normal. Asymmetric and bimodal distributions
are present. A bimodal distribution is a continuous probability distribution with two different
modes. These appear as distinct peaks (local maxima) in the marginal distribution. This is
indicative of the presence of two distinct analyte populations with different detectable signal
intensities. The top right portion of the correlogram lists the coefficients of determination (R2)
between peaks and the bottom left portion contains scatter plots of peak intensities.
815
1004
1091
1576
Peak position (cm-1) and marginal distribution plot (counts vs. intensity)
Coefficient of determination (R2)
Correlation scatter plot
(intensity @ 1091 cm-1
vs. intensity @ 815 cm-1)
Figure 5.3: Example of 4-mercaptopyridine in acid correlogram generated in Excel and descriptions of its main components.
69
5.6 Raman Modes
4-mercaptophenol, 4-mercaptopyridine, and thiophenol belong to the pseudo C2v
symmetry group. There are over 30 normal modes for each of these analytes, some of which are
Raman active. The four C2v symmetry group operators can be split into in-plane (A1 & B2) and
out-of-plane (A2 & B1) ring stretching modes. Four to five Raman modes were selected for each
of the three analytes based on relative signal intensity and variety in vibrational symmetry
(Figure 5.4). Table 5.1 includes a summary of the selected Raman modes, including comparison
to reference values and mode symmetry/descriptions. All of the Raman modes involve ring
stretching modes paired with additional functional group vibrations that are in-plane, unless
otherwise noted.
400 600 800 1000 1200 1400 1600
a bc d
e
ab
c
de
a bc
d
Wavenumber / cm-1
Figure 5.4: Average SERS spectra of 4-mercaptophenol in acid (red), thiophenol in base (green), and 4-mercaptopyridine in acid (blue) with selected peaks (a-e).
70
4-MPOH acid 4-MPOH Raman9 mode702 701 b1
818 816 a1
1006 1007 a1
1168 1172 b2 (βC-O)1589 1585 b2
PhSH base thiophenol (SERS)10 mode419 420 a1, a2
474 470 b1
695 695 a1
1073 1075 a1
1573 1575 a1
4-Mpy acid Ag colloid11 mode815 791 γ(CH)
1004 1004 Ring breathing1091 1095 Ring breathing/C-S1576 1580 ν(CC)
Table 5.1: Experimental and reference vibrational mode frequencies, descriptions and symmetries for 4-mercaptophenol (top), thiophenol (middle), and 4-mercaptopyridine (bottom).
5.7 Results
5.7.1 4-Mercaptophenol Analysis
Figure 5.5 compares correlograms of 4-mercaptophenol at pH 5 and pH 10. None of the
marginal distribution plots along the diagonal appear to be normally distributed for either
correlogram. Instead the distributions of peak intensities appear to be positively skewed. This
means that peak intensities greater than the median value were measured more frequently than
71
peak intensities less than the median value. Additionally, the marginal distributions for most of
the peaks are bimodal at pH 5. This indicates that the average SERS signal of 4-mercaptophenol
in acid is due to two distinct types of SERS enhancement.
The roughly linear correlation plots and their corresponding high determination
coefficients indicate that peak intensities are highly dependent on one another in acid and only
slightly less so in base. In other words there is little change in relative peak intensities between
the 1000 collected spectra at each pH. However, the acidic solution correlation plots contain a
single outlier point with much higher peak intensity than the other 999 collected spectra for the
peak at 1006 cm-1. This is indicative of signaling due to a single hotspot event. Although the
1006 cm-1 peak intensity is much higher in this single spectrum (Figure 5.6), the other peak
intensities are not, indicative of a hotspot spectrum that is unique in relative peak intensities,
possibly due to a unique analyte binding orientation at the hotspot interface.
702
818
1006
1168
1589
4-Mercaptophenol in Acid
702
818
1006
1166
1587
4-Mercaptophenol in Base
Figure 5.5: Correlograms of 4-mercaptophenol in AuNPs SERS spectra in acidic (left) and basic (right) solutions.
72
Figure 5.6: 4-Mercaptophenol in acid outlier spectrum (red) vs. average spectrum (blue).
In Figure 5.7 we show the DLS distributions. Most notable is that we observe
distributions in the pH 5 set of AuNP that matches uncoated nanoparticle distributions. In other
words we did not observe changes due to aggregation at this pH. Conversely at pH 10 we see a
third distribution of particles with 100 times higher concentration due to aggregation. It is also
notable that the number of dimeric aggregates is close to 1000 times less than the monomers
regardless of pH. Multimer aggregates are 100 times more concentrated at pH 10 than at pH 5.
We assume the positive skew observed in the marginal distribution plots (Figure 5.5) is
related to the particle aggregate populations observed in the DLS spectra. Because dimer and
multimer particles contain hotspot regions they likely produce larger SERS enhancements than
monomers. Although the relative population of aggregates is 1000 times less than the monomer
population they still contribute to the overall SERS signal as a result of hotspot enhancement.
73
400 600 800 1000 1200 1400 1600 400 600 800 1000 1200 1400 1600
Figure 5.7: DLS distributions of 4-mercaptophenol-functionalized AuNP aggregates in acidic (top) and basic (bottom) solutions.
The DRS spectra in Figure 5.8 indicate bands which are statistical anomalies. For
example, the 1170 cm-1 peak is unchanged between pH 5 and pH 10. However, the 1006 cm-1
and 1587 cm-1 regions exhibit minor peaks at 993 cm-1 and 1554 cm-1, respectively, in the basic
DRS spectrum. These peaks may represent unique vibrational modes due to particles in the gaps
where binding to between AuNPs occurs. Differential time-dependent signal intensity between
acidic and basic solutions shown in Figure 5.9 indicates that particle aggregation increases with
increasing pH. This observation matches with the DLS results shown in Figure 5.7. This
explains why marginal distributions of peak intensities in base exhibit more asymmetry than the
bimodality observed in the acidic solution which does not have the same degree of aggregation.
74
1000x1000000x
10000x1000x
200 400 600 800 1000 1200 1400 1600 1800
15891006
Wavenumber / cm-1
Wavenumber / cm-1200 400 600 800 1000 1200 1400 1600 1800
993 1006 1554 1587
DRSAverage
Figure 5.8: Comparison of DRS vs. average spectra of 4-mercaptophenol in AuNPs in acidic (top) and basic (bottom) solutions. Raman vibrational modes of interest are expanded.
75
0 100 200 300 400 500 600 700 800 900
Acqusition number
1006
cm-1
Inte
nsity
Figure 5.9: Plots of 1006 cm-1 SERS peak intensity vs. spectrum acquisition number for 4-mercaptophenol in acid (red) and base (blue).
The single spectrum that gave rise to the outlier peak seen in the acidic 4-mercaptophenol
correlogram at 1006 cm-1 was investigated more closely to verify that the signal was not due to
instrument error. Although the spectrum looked similar to the other collected spectra, the peak
maximum was shifted slightly to 1003 cm-1. This observation sparked interest into monitoring
how fluctuations in peak positions compared between vibrational modes and solution pH. To
investigate this behavior, histograms of peak maxima of the five vibrational modes were plotted
in Figure 5.10.
76
-4 -2 0 2 4 6 80
200
400
600
800
4-MpOH pH 5
Wavenumber Shift (cm-1)
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
4-MpOH pH 10
702818100611661587
Wavenumber Shift (cm-1)
Figure 5.10: Histograms of vibrational mode maxima frequency shifts for 4-mercaptophenol in AuNPs in acidic (left) and basic (right) solutions.
The results indicate that the degree of fluctuation in peak position differs between
vibrational modes with the 818 cm-1 peak undergoing the most significant shifts in peak position.
Although slight deviations in peak positions could be attributed to frequency drift the observed
fluctuations are not systematic. Additionally, the amounts of peak shifting increase for three of
the five peaks in basic solution (702 cm-1, 818 cm-1, and 1166 cm-1) while the other two peaks
remain constant. In particular, the out-of-plane ring stretching mode that gives rise to the 702 cm-
1 peak undergoes increased position fluctuations at pH 10. Although DLS data indicates that
aggregation increases in with increasing pH, significant changes in peak fluctuations between
vibrational modes and pHs suggest that pH-dependent chemical binding between 4-
mercaptophenol and AuNPs plays a role in SERS enhancement.
77
5.7.2 Thiophenol Analysis
Correlograms of thiophenol-functionalized AuNPs in acidic and basic solutions are
shown in Figure 5.11. Unlike 4-mercaptohenol the marginal distributions of thiophenol peak
intensities appear normal and invariant to pH. This is because thiophenol adsorbed to AuNPs has
no potential for pH-induced reaction chemistry. The thiophenol in base correlogram shows a
single outlier spectrum indicated by positive skewing of the otherwise normal distributions for
peaks at 419 cm-1, 474 cm-1, and 695 cm-1. This single outlier spectrum manifests as a single data
point in the upper right corner of all but one of the correlation scatter plots. Interestingly, the
peak intensities at 1073 cm-1 and 1574 cm-1 in this outlier spectrum are not significantly higher
than in the other 999 collected spectra. The determination coefficients between thiophenol peaks
vary significantly and with no clear dependence on pH. For example, 1073 cm-1 and 1574 cm-1
peak intensities exhibit a high level of dependence in base (R2 = 0.94) and are nearly independent
in acid (R2 = 0.06).
418
474
694
1073
1574
Thiophenol in Acid
419
474
695
1073
1574
Thiophenol in Base
78
Figure 5.11: Excel correlograms of thiophenol in AuNPs in acidic (left) and basic (right) solution.
The outlier spectrum was compared to the average spectrum in Figure 5.12. Although
there is an apparent shift in overall signal intensity the relative peak intensities of the average and
outlier spectra remain nearly constant. This indicates that analyte binding remains the same
between the outlier and ensemble signals and that SERS enhancement is most likely due to
thiophenol adsorbed to a hotspot region on a coalesced nanoparticle dimer.
Figure 5.12: Thiophenol in base outlier spectrum (red) vs. average spectrum (blue).
In Figure 5.13 we show the DLS distributions. Both distributions show reduced
concentrations of dimeric aggregates compared to 4-mercaptophenol coated nanoparticle
distributions. Multimer aggregates are undetectable at pH 5 and at very low concentration at pH
79
400 600 800 1000 1200 1400 1600 400 600 800 1000 1200 1400 1600
10. This indicates that thiophenol adsorption does not contribute significantly to nanoparticle
aggregation at either pH.
Figure 5.13: DLS distributions of thiophenol-functionalized AuNP aggregates in acidic (top) and basic (bottom) solutions.
Fluctuations in vibrational mode peak positions were plotted in Figure 5.14 to ensure that
the SERS enhancement of thiophenol in base was a result of coalesced AuNPs and not pH-
induced chemical bond formation. Variations in peak maxima are within ±3 cm-1 for all of the
vibrational modes regardless of pH. The largest degree of fluctuation was observed for the 474
cm-1 peak which corresponds to an out-of-plane ring breathing mode.
-4 -3 -2 -1 0 1 2 3 40
100200300400500600700800900
1000
Thiophenol pH 5
Wavenumber Shift (cm-1)
-4 -3 -2 -1 0 1 2 3 4
Thiophenol pH 10
41947469510731574
Wavenumber Shift (cm-1)
80
100000x
1000000000x
10000x
Figure 5.14: Histograms of vibrational mode maxima frequency shifts for thiophenol in AuNPs in acidic (left) and basic (right) solutions.
5.7.3 4-Mercaptopyridine Analysis
Although SERS enhancement phenomena of 4-mercaptopyridine in acid and base were
analyzed using DRS techniques described in chapter 4, correlograms and frequency shift
histograms were generated to ensure the viability of these new data analysis techniques.
Correlograms of 4-mercaptopyridine in acid and base are shown in Figure 5.15. Marginal
distributions of peak intensities exhibit mostly normal behavior with slight positive skewing in
both acid and base. Determination coefficients between peaks decrease significantly in basic
solution. Low R2 values in the first row of both correlograms indicate 815 cm-1 peak intensity is
significantly less dependendent on intensities of the other three peaks.
815
1004
1091
1576
4-Mpy in Acid
4-Mpy in Base
815
1010
1092
1576
Figure 5.15: Excel correlograms of 4-mercaptopyridine in AuNPs in acidic (left) and basic (right) solution.
81
Histograms of peak maxima of the four vibrational modes were plotted in Figure 5.16.
The vibrational modes at 1004/1010 cm-1 and 1576 cm-1 undergo significant shifts in peak
position while the vibrational modes at 815 cm-1 and 1091 cm-1 remain constant. Aside from peak
shifting from 1004 cm-1 to 1010 cm-1 between acidic and basic solutions, the two histograms are
nearly identical. These observations indicate that although the SERS enhancement is sensitive to
chemical bond formation between 4-mercaptopyridine and AuNPs it is independent of pH.
-10
-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 80
100
200
300
400
500
4-Mpy pH 5
Wavenumber Shift (cm-1)
-10
-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
4-Mpy pH 10
815100410911576
Wavenumber Shift (cm-1)
Figure 5.16: Histograms of vibrational mode maxima frequency shifts for 4-mercaptopyridine in AuNPs in acidic (left) and basic (right) solutions.
5.8 Summary
A summary of our results is shown in Figure 5.17 with proposed adsorption mechanisms
for the three analytes in acidic and basic solutions. Our analysis indicates that 4-mercaptophenol-
coated AuNPs undergo pH-induced aggregation resulting from O-AuNP bond formation.
82
Thiophenol-coated AuNPs functioned as a negative control because they are not affected by pH.
4-mercaptopyridine-coated AuNPs undergo aggregation resulting from N-AuNP bond formation
in both acidic and basic solutions.
O
S S
pH 5 pH 10
Figure 5.17: Summary of proposed adsorption mechanisms for 4-mercaptophenol (top), thiophenol (middle), and 4-mercaptopyridine (bottom) on AuNPs in acidic (left) and basic (right) conditions.
83
5.9 Conclusion
Correlograms and frequency shift histograms are a continuation of our previous research
efforts involving specific extraction of unique SERS signals from colloidal SERS active
nanoparticles by DSERS analysis. Outliers in marginal distributions of peak intensities expedited
the identification and extraction of highly enhanced SERS spectra. Marginal distribution profiles
indicated the absence (normal) or presence and type of hotspot formation: chemisorption
(bimodal) or physisorption (positively skewed). Scatter plots of peak intensities and their
corresponding determination coefficients identified unique relationships between specific
vibrational modes.
Frequency shift histograms indicated that certain vibrational modes, particularly out-of-
plane ring stretching modes, fluctuate more than in-plane ring breathing modes. They also
showed that fluctuations in peak intensity increased with the formation of chemisorbed
aggregates. Most importantly, results from these new statistical analysis techniques match with
results from DLS distributions and DSERS analyses.
5.10 References
1. Fleischmann, M.; Hendra, P. J.; McQuilla.Aj, Raman-Spectra of Pyridine Adsorbed at a Silver Electrode. Chem. Phys. Lett. 1974, 26 (2), 163-166.
2. Jeanmaire, D. L.; Vanduyne, R. P., Surface Raman Spectroelectrochemistry. Part 1. Heterocyclic, Aromatic, and Aliphatic-Amines Adsorbed on the Anodized Silver Electrode. J. Electroanal. Chem. 1977, 84 (1), 1-20.
3. Kerker, M.; Siiman, O.; Bumm, L. A.; Wang, D. S., Surface Enhanced Raman-Scattering (SERS) of Citrate Ion Adsorbed on Colloidal Silver. Appl. Optics 1980, 19 (19), 3253-3255.
4. Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter, D. G.; Natan, M. J., Self-Assembled
84
Metal Colloid Monolayers - an Approach to SERS Substrates. Science 1995, 267 (5204), 1629-1632.
5. Wustholz, K. L.; Henry, A. I.; McMahon, J. M.; Freeman, R. G.; Valley, N.; Piotti, M. E.; Natan, M. J.; Schatz, G. C.; Van Duyne, R. P., Structure-Activity Relationships in Gold Nanoparticle Dimers and Trimers for Surface-Enhanced Raman Spectroscopy. J. Am. Chem. Soc. 2010, 132 (31), 10903-10910.
6. Scott, B. L.; Carron, K. T., Dynamic Surface Enhanced Raman Spectroscopy (SERS): Extracting SERS from Normal Raman Scattering. Anal. Chem. 2012, 84 (20), 8448-51.
7. Yu, H.-Z. X., Nan; Liu, Zhong-Fan, SERS Titration of 4-Mercaptopyridine Self-Assembled Monolayers at Aqueous Buffer/Gold Interfaces. Anal. Chem. 1999, 71 (7), 1354-1358.
8. Siegrist, K. The Central Limit Theorem. http://www.math.uah.edu/stat/sample/CLT.html.
9. Li, R.; Ji, W.; Chen, L.; Lv, H.; Cheng, J.; Zhao, B., Vibrational Spectroscopy and Density Functional Theory Study of 4-Mercaptophenol. Spectrochimica acta. Part A, Molecular and Biomolecular Spectroscopy 2014, 122, 698-703.
10. Carron, K. T. H.; L. Gayle, Axial and Azimuthal Angle Determination with Surface-Enhanced Raman Spectroscopy: Thiophenol on Copper, Silver, and Gold Metal Surfaces The Journal of Physical Chemistry 1991, 95 (24), 9979-9984.
11. Y. Wang, H. H., S. Jing, Y. Wang, Z. Sun, B. Zhao, C. Zhao, J. R. Lombardi, Enhanced Raman Scattering as a Probe for 4-Mercaptopyridine Surface-modified Copper Oxide Nanocrystals. Analytical Sciences 2007, 23, 787-791.
85
6 Clennan Group Collaboration: Viologen-Functionalized SERS Substrates for the
Detection of Polycyclic Aromatic Hydrocarbons and Chiral Molecules
6.1 Introduction
A novel helical viologen (N, N'-Dimethyl-5,10-diaza[5]helicene) was synthesized and
characterized by the Clennan research group1. Viologens function as electron-transfer mediators,
DNA photocleaving agents, and as acceptor components of host-guest complexes. A chemical
sensor system for the detection of polycyclic aromatic hydrocarbon (PAH) pollutants using silver
nanoparticles functionalized with a viologen was demonstrated by Lopez-Tocon et al.2 and
sparked interest to develop a similar chemical sensor system using this particular viologen.
Additionally, N, N'-Dimethyl-5,10-diaza[5]helicene is chiral and undergoes racemization
in aqueous solution. Silver nanoparticles functionalized with a chiral viologen may offer
additional sensor capabilities including racemization kinetics and selective adsorption and
detection of chiral analytes. We chose cysteine as our analyte because it is a thiol that binds
readily to AgNPs and occurs as either D or L isomers. Similar research by the Balaz group
showed that CD measurements could differentiate between the chiral optical activity of D and L
cysteine functionalized quantum dots3.
6.2 Silver Nanoparticle (AgNP) Synthesis
Silver nanoparticles (AgNPs) of 30-50 nm diameter were prepared according to a well-
known Frens citrate reduction method. 100 mg of AgNO3 was added to 500 mL of hot HPLC
grade water and brought to a boil with stirring. 10 mL of 1% (w/v) sodium citrate dihydrate was
added at once and the reaction mixture was covered and left to boil with stirring for 1 hour. After
86
1 hour, the heat was turned off and the Ag colloid solution was cooled to room temperature and
transferred to a foil-covered plastic container. λmax = 425 nm of the colloidal solution was
determined by UV-vis analysis.
6.3 Instrumentation
Raman spectra were acquired with an IM-52 Raman microscope (Snowy Range
Instruments) using its solid and liquid sampling features. The IM-52 was set to 40 mW of 785
nm laser excitation at the sample with 8 cm-1 spectral resolution. Spectral data was collected for
0.5 s and 10 sequential collections were used to generate average spectra, unless otherwise noted.
Model structures and Raman spectra of phenanthrene and N, N'-Dimethyl-5,10-diaza[5]helicene
(viologen) were calculated using Gaussian 09 (B3LYP/6-311+G(2d,p)).
CD spectra were acquired using an Aviv model 430 CD spectrometer using a xenon lamp
light source. The spectrum range was set to 550-180 nm with a bandwidth of 4 nm. Spectral data
was collected for 100 ms every 0.5 nm and 5 sequential collections were used to generate an
average spectrum
6.4 Experimental
200 µL of ~20 mM viologen in acetonitrile was added to 2 mL of AgNPs in a glass
sample vial and an average spectrum was collected. 200 µL of 31.4 mM phenanthrene in
acetonitrile was added to the sample vial and an average spectrum was collected. The sample
solution was vortexed and additional spectra were collected hourly for 10 hours. An average
Raman spectrum of solid phenanthrene (i.t. 3 s) was collected.
87
6.7 mg of D-cysteine was dissolved in 2mL AgNPs in a glass sample vial and an average
spectrum was collected (i.t. 5 s, avg/5). 1 mL of 1% (w/v) sodium citrate was added to the
sample vial and indicator paper was used to determine the pH of the acidic solution (pH ~5). A
SERS spectrum was collected (i.t. 5 s, avg/5). The procedure was repeated using 10.4 mg of L-
cysteine.
Saturated aqueous solutions of pyrene, phenanthrene, napthalene, chrysene, benzene, and
anthracene were prepared by adding 10-15 mg of each PAH to 2mL water. Due to the low
solubility of these PAHs in water, most of the solid material remained undissolved, even after the
solutions were vortexed intermittently for several days. 200 µL of ~20mM viologen in
acetonitrile was added to 2mL AgNPs. 200 µL of saturated aqueous PAH solutions were added
to separate aliquots of the vNP solution and SERS spectra of the PAH-vNP solutions were
collected.
20 µL of ~20mM viologen in acetonitrile was added to 2mL AgNPs. 6-8 mg of D/L/DL
cysteine was added to separate aliquots of the vNP solution, followed by 1 mL of 1% (w/v)
sodium citrate. 0.5 mL of the cysteine-vNP solution was added to of 1.5 mL water in a quartz
cuvette and CD spectra were collected. SERS spectra of the remaining 1.5 mL of cysteine-vNP
solutions were collected at 0 and 1 hours.
6.5 Results and Discussion
Figures 6.1 and 6.2 compare experimental and model spectra of phenanthrene and
viologen. Both figures show some agreement between experimental and model spectra with
deviations in relative peak intensity and wavenumber. Although phenanthrene shows better
agreement than viologen, it should be noted that a normal Raman spectrum and SERS spectrum
88
were collected for phenanthrene and viologen, respectively, and that model spectra were
calculated using a monomer in gas phase not associated with a silver nanoparticle.
400 600 800 1000 1200 1400 1600 1800 Wavenumber/ cm-1
Figure 6.1: Model Raman spectrum of phenanthrene (red), Raman spectrum of solid phenanthrene (green). Model spectrum peak at 3180 cm-1 not shown.
89
400 600 800 1000 1200 1400 1600 1800 Wavenumber / cm-1
Figure 6.2: Model Raman spectrum of viologen (red), baseline corrected SERS spectrum of viologen on AgNPs (blue). Model spectrum peaks at 3070 and 3207 cm-1 not shown.
Figure 6.3 compares SERS spectra of phenanthrene in viologen-functionalized AgNPs
taken at different times over ten hours. The low solubility of PAHs in aqueous solutions and the
highly ordered structure of the self-assembled monolayer of viologen on the NP surface make
adsorption and detection of phenanthrene a slow process. Changes in the SERS spectrum are
most apparent between the 0 and 3 hours, and taper off with increasing time, most likely due to
NP aggregation. Boxes indicate regions where significant changes in the SERS spectrum occur.
90
400 600 800 1000 1200 1400 1600 Wavenumber/ cm-1
Figure 6.3: Baseline-corrected SERS specta of viologen on AgNPs (blue), addition of phenanthrene taken at 0h (violet), 3h (green), and 10h (red).
Figure 6.4 contains SERS spectra of AgNPs adsorbed with D and L cysteine at neutral
and acidic pH. These results demonstrate the need to acidify the solution for adsorption and
detection of cysteine on AgNPs. At neutral pH, thiol deprotonation and dimerization via
disulfide bonding inhibits adsorption to the AgNP surface. A 1% sodium citrate buffer was used
to adjust the pH because the AgNPs are synthesized using a similar sodium citrate solution. Both
D and L cysteine SERS spectra are identical, as expected. CD spectra of these samples were
collected to further investigate the role of chirality on NP adsorption.
91
400 600 800 1000 1200 1400 1600 Wavenumber / cm-1
Figure 6.4: SERS spectra of L-cysteine at pH 5 (blue) and pH 7 (violet); SERS spectra of D-cysteine at pH 5 (green) and pH 7 (red).
Figure 6.5 compares the SERS spectra of various PAHs in viologen-functionalized NP
solutions. The experimental method was varied slightly from that used in figure 3. Saturated
aqueous PAH solutions were made by adding PAHs to water in excess and mixing the solutions
for several days. This procedure was implemented to reduce the amount of time required for
PAH adsorption to the AgNPs. Differences between PAH-adsorbed spectra are subtle but are
distinct from the SERS spectrum without adsorbed PAH. It should be noted that the PAHs used
have different solubilitites in water but equal volumes of each solution were added to sample
vials. Adjusting for this will assure equal surface coverage of PAHs in solution.
92
400 600 800 1000 1200 1400 1600
Wavenumber / cm-1
Figure 6.5: Baseline corrected SERS spectra of viologen on Ag NPs from top to bottom: No PAH, pyrene, phenanthrene, napthalene, chrysene, benzene, anthracene.
Figure 6.6 compares average CD spectra of D/L/DL cysteine-coated AgNPs at pH 5.
Although CD spectra of D and L cysteine are not complementary, as is expected for chiral
molecule-coated colloids, all 3 spectra show substantial and similar optical rotation. This could
be a result of chiral citrate molecules on the nanoparticle surface. However, the mechanism and
selectivity of a chiral citrate self-assembled monolayer on the nanoparticle surface has yet to be
understood.
93
-7
-6
-5
-4
-3
-2
-1
180 200 220 240 260 280 300 320 340 180 200 220 240 260 280 300 320 340
Wavelength / nm
CD /
mde
g
Figure 6.6: Average CD spectra of D-cysteine (red), L-cysteine (blue) and DL cysteine (violet) on AgNPs.
Figure 6.7 compares average CD spectra of D-cysteine (red) and L-cysteine (blue) on
viologen-coated AgNPs. The two spectra are complementary to one another, as expected for
chiral molecules.
94
-4
-3
-2
-1
180 200 220 240 260 280 300 320 340 360
Wavelength / nm
CD /
mde
g
Figure 6.7: Average CD spectra of D-cysteine (red) and L-cysteine (blue) on viologen-coated AgNPs.
Figure 6.8 shows the time dependence of the CD spectrum of DL cysteine on vNPs. The
five CD spectra collected for this sample showed a gradual transition from the observed CD
spectrum of viologen-functionalized AgNPs to the observed CD spectrum of DL cysteine-
functionalized AgNPs. Only the first and fifth collected CD spectra of the sample solution are
included in the figure but the other three CD spectra exhibited a similar trend.
95
-8
-6
-4
-2
0
180 200 220 240 260 280 300 320 340 Wavelength / nm
CD /
mde
g
Figure 6.8: CD spectra of viologen-coated AgNPs (violet), DL cysteine on vNPs at t=0h (red), DL cysteine on vNPs at t=1h (green), DL cysteine-coated AgNPs.
Figure 6.9 compares the average SERS spectra of D/L/DL cysteine on vNPs. Although
the spectra are nearly identical, there is a distinct splitting of vibrational modes in the 660-680
cm-1 region. Thiocarboxylic acids generally exhibit a strong Raman band in the region of 500-
750 cm-1 due to the C-S stretching mode. Multiple C-S bands may be observed due to rotational
isomerism4. The DL cysteine on vNPs SERS spectrum exhibits a single broad peak in this
region, while SERS spectra of samples containing D and L cysteine isomers exhibit peak
maxima that are red and blue shifted, respectively. Although this phenomenon is subtle, it
suggests that viologen-functionalized AgNPs may serve as a chiral-sensitive SERS substrate.
96
400 500 600 700 800 900 Wavenumber / cm-1
Figure 6.9: Average SERS spectra of D cysteine (red), L cysteine (blue) and DL cysteine (green) on viologen-coated Ag NPs. Peak splitting at 660-680 cm-1 appears to be indicative of D and L cysteine isomers.
6.6 References
1. Zhang, X.; Clennan, E. L.; Arulsamy, N., Photophysical and Electrochemical Characterization of a Helical Viologen, N,N'-Dimethyl-5,10-Diaza[5]Helicene. Organic letters 2014, 16 (17), 4610-3.
2. Lopez-Tocon, I.; Otero, J. C.; Arenas, J. F.; Garcia-Ramos, J. V.; Sanchez-Cortes, S., Multicomponent Direct Detection of Polycyclic Aromatic Hydrocarbons by Surface-Enhanced Raman Spectroscopy Using Silver Nanoparticles Functionalized with the Viologen Host Lucigenin. Anal Chem 2011, 83 (7), 2518-25.
3. Tohgha, U.; Varga, K.; Balaz, M., Achiral CdSe Quantum Dots Exhibit Optical Activity in the Visible Region Upon Post-Synthetic Ligand Exchange with D- Or L-Cysteine. Chem Commun (Camb) 2013, 49 (18), 1844-6.
97
4. Lin-Vien, D.; Colthup, N. B.; Fateley, W. G.; Grasselli, J. G., The Handbook of Infrared and Raman Characteristic Frequencies of Organic Molecules. Academic Press: San Diego, 1991, 234.
98