-
Nanobiotechnol (2007) 3:172–196DOI 10.1007/s12030-008-9015-z
Tip-Enhanced Raman Imaging and Nanospectroscopy:Sensitivity,
Symmetry, and Selection Rules
Catalin C. Neacsu · Samuel Berweger ·Markus B. Raschke
Published online: 25 February 2009© Humana Press Inc. 2009
Abstract The fundamental mechanisms of tip-enhanced Raman
spectroscopy (TERS) have beeninvestigated, including the role of
the plasmonic ex-citation of the metallic tips, the nature of the
opticaltip–sample coupling, and the resulting
local-fieldenhancement and confinement responsible for ultra-high
resolution imaging down to just several nano-meters. Criteria for
the distinction of near-fieldsignature from far-field imaging
artifacts are addressed.TERS results of molecules are presented.
With en-hancement factors as high as 109,
single-moleculespectroscopy is demonstrated. Spatially resolved
vi-brational mapping of crystalline nanostructures anddetermination
of crystallographic orientation and do-mains is discussed making
use of the symmetry prop-erties of the tip scattering response and
the intrinsicRaman selection rules.
Keywords tip-enhanced Raman imaging ·nanospectroscopy ·
tip-enhanced Raman spectroscopy
Introduction
Optical spectroscopy provides nondestructive tech-niques for
obtaining both structural and real-timedynamic information of
molecules and solids. Vibra-tional spectroscopy in particular, by
directly couplingto the nuclear motion, offers insight into
chemical
C. C. Neacsu (B) · S. Berweger · M. B. RaschkeDepartment of
Chemistry, University of Washington,Seattle, WA 98195-1700,
USAe-mail: [email protected]
composition, molecular bonds, intermolecular cou-pling, and
zone-center phonons in crystalline solids[1, 2]. IR and Raman
spectroscopy provide complemen-tary information. Raman spectroscopy
features fewerconstraints in terms of selection rules, readily
providesaccess to low frequency vibrations [3–6], and is car-ried
out in the visible to near IR spectral range in acomparably simple
experimental design. However, withscattering cross-sections of ∼
10−27 − 10−30 cm2, theRaman response is weak, generally requiring
probinga large molecular ensemble or bulk solids [7, 8].
It is therefore desirable to combine the intrinsicchemical
specificity of Raman spectroscopy with opticalmicroscopy for the
investigation of the spatial hetero-geneity and composition of the
analyte. In that regard,the optical far-field Raman microscope has
become anestablished tool for material characterization on the
mi-crometer scale and in a confocal implementation withspatial
resolution down to several hundred nanometers[9]. However, for most
applications, the desired spatialresolution often exceeds the
resolution imposed by far-field diffraction [10, 11].
Scanning near-field optical microscopy (SNOM) pro-vides access
to subwavelength scale spatial resolution[12–19]. Aperture-based
SNOM using tapered glassfiber tips has been employed for nano-Raman
spec-troscopy [20–23]. However, the low optical throughputof the
aperture probes (10−3–10−5) severely limits thespatial resolution
and the sensitivity that can be ob-tained, resulting in a long
imaging time and parasiticRaman signal from the glass tip that
could be an im-pediment [22].
High sensitivity in Raman scattering, in general,can be achieved
by surface-enhanced Raman spectros-copy (SERS), providing a
strongly enhanced Raman
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 173
response from molecular adsorbates on rough metallicsurfaces or
colloidal aggregates [24–26]. SERS is due tothe near-field
enhancement of the electromagnetic fieldat single or coupled metal
nanostructures often reso-nantly excited at their surface plasmon
polariton (SPP)eigenmodes [27–34]. Together with a corresponding
butweaker (∼ 101–102) chemical contribution [35] origi-nating from
surface bonding or charge transfer, theelectromagnetic enhancement
leads to a total increasein Raman signal by up to 14 orders of
magnitude,allowing for detection down to the single moleculelevel
[36–41]. Despite its potential for chemically spe-cific detection
of minute amounts of analytes, it hasremained challenging to
develop SERS into a routineanalytic spectroscopic tool mostly due
to difficultiesassociated with the reproducible fabrication of
SERS-active substrates [42–45].
Better control over the SERS active sites and theirfield
enhancement can be achieved by what may beviewed as resorting to an
inverse geometry with respectto SERS: suspension of the metal
nanostructure provid-ing the field enhancement at a small distance
above theanalyte [46]. This is the basis of tip-enhanced
Ramanscattering (TERS) making use of a single plasmon-resonant
metallic nanostructure provided in the form ofa scanning probe tip
of suitable material and geometry.
Fundamentally, TERS is a variant of scattering-type scanning
near-field optical microscopy (s-SNOM)[47–51]. All-optical
resolution down to just severalnanometers is provided by s-SNOM, in
the visible[52–54] and IR spectral regions [55–58]. TERS is
theextension of this technique to inelastic light scatteringwith
the metallic tip used as an active probe, whichprovides both the
local-field enhancement and servesas an efficient scatterer for the
Raman emission.
s-SNOM and special aspects of TERS have been ad-dressed in
recent reviews [18, 19, 59–62] (and referencestherein), but no
comprehensive discussion of the under-lying physical mechanisms has
yet been provided. Here,we review our recent contributions to the
understand-ing of near-field Raman enhancement and
sensitivity,tip–sample coupling, spatial resolution, the
importanceof the plasmonic character of the tip, and tip
fabrica-tion. In addition, we show that the symmetry proper-ties of
the tip-scattering geometry in combination withthe Raman selection
rules allows for the determina-tion of crystallographic information
on the nanoscale.This, together with the results of other groups,
showsthe potential of TERS as a nanoanalytical tool withdiverse
applications in material and surface scienceand analytical
chemistry for the study of biomolecularinterfaces, molecular
adsorbates, nanostructures, andnanocomposites.
Tip-Enhanced Raman Spectroscopy
TERS combines the advantages of SERS with thoseoffered by
s-SNOM: the single nanoscopic tip apex pro-vides the local field
enhancement at a desired samplelocation without requiring any
special sample prepara-tion [63, 64]. With the spatial resolution
mainly limitedby the tip apex size, chemical analysis on the
nanometerscale is made possible. By raster scanning the
sample,spatially resolved spectral Raman maps with nanome-ter
resolution can be obtained simultaneously withthe topography in
atomic force microscopy (AFM)or surface electronic properties in
scanning tunnelingmicroscopy (STM).
The origin of the field enhancement at the tip apexis attributed
to the singular behavior of the electro-magnetic field (akin to the
lightning-rod effect). Inaddition, the spatial confinement allows
for the possibleexcitation of localized SPPs (tip-plasmons) for
certaintip materials [65]. With the first effect being
geometricalin origin, its magnitude is mainly dependent on
thecurvature of the apex. Taking advantage of the exci-tation of
tip-plasmons can further increase the overallenhancement by several
orders of magnitude, as will bediscussed further on.
Ultra-high sensitivity and nanometer spatial resolu-tion imaging
using TERS were obtained on variousmaterials and molecular systems
adsorbed on both flatand corrugated surfaces [66–85]. Having large
Ramancross-sections, several dye molecules (e.g., malachitegreen
(MG), rhodamine 6G, brilliant cresyl blue) wereused and near-field
Raman enhancement factors upto ∼ 109 were achieved [68, 71, 77, 81,
82]. UsingAg-coated AFM tips, spatial resolution below 50 nmwas
obtained on surface layers of Rhodamine 6G dyemolecules [68, 71].
Lateral resolution as high as 14 nmand a maximum Raman enhancement
factor estimatedat ∼104 were obtained in spatially resolved
probingvibrational modes along individual carbon nanotubes[72, 73,
78].
In studies of adenine, as well as C60 molecules,the tip-induced
mechanical force was shown to leadto mechanical strain-induced
frequency shifts of thenormal Raman modes [74, 80]. Furthermore, it
wasobserved that, when interacting with individual metalatoms of
the tip apex, adenine molecules form differentisomers,
demonstrating the potential for TERS to havefor atomic site
selective sensitivity [83].
Extension of TERS implementation for coher-ent spectroscopy was
shown for coherent anti-StokesRaman scattering (CARS) of adenine
molecules in-cluded in a DNA network [75]. Owing to the third-order
nonlinearity of the CARS process, the induced
-
174 Neacsu et al.
polarization at the tip apex is further confined, andhigher
lateral resolution is, in principle, possible [76].Concomitant,
theoretical studies on TERS report fieldenhancements up to three
orders of magnitude in par-ticular wavelength regions [86–91].
However, the ex-pected resulting TERS enhancement of 12 orders
ofmagnitude has not yet been observed experimentally.
In recent work from our group, we have refinedthe metallic tip
fabrication and experimentally identi-fied and theoretically
discussed the importance of theplasmonic properties of the scanning
tip to achievehigh Raman sensitivity [92–95]. This has enabled
near-field Raman enhancement factors of up to 109 fromMG molecules
adsorbed on smooth Au surfaces to beobtained, allowing for the
detection of TERS responsewith single molecule sensitivity [81,
96].
The review is organized as follows: The experimen-tal
arrangement is presented in the “Experimental”section. This
includes laser excitation, Raman detec-tion, metallic tip
fabrication by electrochemicaletching, and molecular systems used.
“Optical TipCharacterization” discusses the experimental
charac-terization of the optical properties of the tips in-cluding
their plasmonic behavior and the local-fieldenhancement factor as
determined by second harmonicgeneration (SHG). “Calculation of the
Near-FieldDistribution at the Tip-Apex” describes the
theoreticalanalysis of the near-field distribution at the tip
apextogether with its spectral characteristics. The tip–sample
optical coupling is discussed in the “OpticalTip–Sample Coupling”
section, where its effect onsensitivity, spatial resolution, and
spectral shift ofthe plasmon resonance are derived. The
near-fieldcharacter vs far-field imaging artifacts in TERS and
itspolarization dependence are addressed in the “Near-Field
Character and Far-Field Artifacts in TERS”section. The procedure
for estimating the near-fieldenhancement factor is detailed and
representativevalues discussed. Tip-enhanced near-field spectra
ofmonolayer (ML) and a submonolayer of molecularadsorbates on a
smooth Au surface are given in the“TERS of Molecular Adsorbates”
section. The highsensitivity obtained and the dependence of the
spectralfeatures on the enhancement level is discussed. In
the“Molecular Bleaching” section, we also address theimportant
question of molecular bleaching and possiblechemical contamination
paths and show a number ofcontrol experiments. Near-field
tip-enhanced Ramanresults with single molecule sensitivity are
shown inthe “TERS with Single Molecule Sensitivity” section.This is
concluded from the ultra-low molecular cover-age and the observed
intensity and spectral temporalfluctuations. We identify and
propose in the “Raman
Imaging of Nanocrystals: Near-Field CrystallographicSymmetry”
section a new and promising extensionof TERS for determination of
both chemical andstructural properties of nanocrystals. The
“Outlook”section gives an outlook on TERS, and novel waysto
circumvent current instrumental difficulties arediscussed.
Experimental
Various experimental schemes have been employed forTERS
experiments. A tip axial illumination and detec-tion geometry has
been used, allowing for high numer-ical aperture (NA), but
requiring transparent samplesor substrates [73, 97]. A high-NA
parabolic mirror canbe used to probe nontransparent samples [98,
99]. Inboth schemes, the tip is illuminated along the
axialdirection with the tip apex positioned in the laser focus.For
these geometries, polarization conditions requireeither a
Hermite–Gaussian beam [100] or radial inci-dent polarization [98,
101]. While allowing for efficientexcitation and detection with the
tip, independent po-larization and k-vector control is limited but
desirablefor symmetry selective Raman probing.
In contrast, side-on illumination and detection allowsfor
greater flexibility in the selection of polarization andk-vector,
as well as the use of nontransparent samples.The scanning and
tip–sample distance are controlledusing either an atomic force
microscope (AFM) or ascanning tunneling microscope (STM), with STM
re-stricted to the use of conducting samples.
Figure 1 shows the experimental layout of our side-illuminated
TERS experiment. The incident radiation(νi) is focused onto the
tip–sample gap and the tip-scattered Raman light (νs) is detected.
For the ex-periments described here, a shear-force AFM is
used.Shear-force AFM maintains a constant height of sev-eral
nanometers above the sample. Due to the short-range tip–sample
distance dependence of the opticalfield enhancement, dynamic
noncontact AFM is lesssuitable. The time-averaged signal is greatly
reduceddue to the oscillating tip. Contact AFM maintains aconstant
and small tip–sample distance but the com-parably large forces make
it unfavorable for probingmolecular or soft matter samples. In
contrast, with thespatial range of shear-forces confined to within
25 nm[102], the shear-force AFM tip is controlled in closeproximity
to the sample without actual physical contact.
The control mechanism in shear-force AFM is basedon the near
surface vibrational damping of a probetip oscillating parallel to
the surface. The nature ofthe shear-force damping mechanism is not
yet fully
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 175
Fig. 1 Schematic of the experimental arrangement for TERS.The
incident light is focused onto the tip–sample gap. The tip-enhanced
and -scattered Raman response is spectrally filteredusing a notch
filter and is spectrally resolved by an imagingspectrometer with a
liquid nitrogen-cooled charge-coupled de-vice array, or integrally
detected by means of an avalanchephotodiode (APD). Polarization
directions of both incident andscattered light beams can be
controlled independently. The bluelayer indicates a thin water film
as may be present on the sampleunder ambient conditions.
understood [19], with a variety of mechanisms beingdiscussed
[103–106]. It has been suggested [107, 108]that the tip experiences
viscous damping from a thinwater layer adsorbed on the surface of
the sampleunder ambient conditions [108–110]. This water
layer,present on most hydrophilic samples, may play animportant
role in the surface diffusion of the analytemolecules and possibly
the transition of molecules toadsorb onto the tip.
For an incident light source, a continuous waveHelium–Neon
laser, with λi = 632.8 nm (1.92 eV) iscommonly used [2]. In our
experiments, after passingthrough a laser-line filter, the light is
focused onto thetip–sample gap by means of a long working distance
mi-croscope objective (NA = 0.35). The tip-backscatteredlight is
collected with the same objective and spectrallyfiltered using a
notch filter. The signal is detected usingeither an avalanche
photodiode or spectrally resolvedusing a fiber-coupled imaging
spectrograph with a liq-uid nitrogen-cooled charge-coupled device
detector.Even for large enhancements, the signal intensities
areweak, and detector noise is one limiting factor. Wetherefore
limit the spectral resolution to 25 cm−1 forthe tip-enhanced
experiments. Far-field spectroscopicstudies of molecular MLs
serving as reference toquantify the enhancement are conducted using
a micro-Raman confocal setup, based on an inverted micro-scope
(Zeiss Axiovert 135).
For our experiments, we chose MG, an organic triph-enylmethane
laser dye with an absorption peak aroundλ ≈ 635 nm. The absorption
peak of MG is very close
to the laser energy used, leading to a resonant Ramanexcitation
via the S0–S1 electronic transition of theconjugated π -electron
system, as discussed below [111].To limit the rate of the molecular
decomposition, themaximum fluence in the focus of the microscope
objec-tive was 5 × 103 − 3 × 104 W/cm2. However, molecularbleaching
prevails under ambient conditions under res-onant Raman excitation
in TERS [81, 112].
Tip Fabrication
The metallic scanning probe tips hold the central func-tion in
TERS studies, providing the enhanced elec-tromagnetic field at
their apex. Ideally, as discussedbelow, they present strong plasmon
resonances in thespectral region of interest, leading to enhanced
pump(νi) and scattered Raman fields (νS) at the apex. Sincethe
first experiments, the fabrication of suitable tipshas been a major
experimental challenge. A vari-ety of methods for their fabrication
have been used:including angle-cutting the metal wire [113], DC
orAC voltage electrochemical and milling procedures[114–117],
focused ion beam milling [54], metal-coatingcommercially available
cantilever AFM tips [118, 119],and attaching spherical or other
plasmonic nanopar-ticles to a glass tip [120]. Electrochemical
etching[98, 121] is most commonly used due to perceivedadvantages
[122] of tips fabricated in this manner.
We employ a DC voltage electrochemical etchingmethod for both Au
and W tips. It involves the anodicoxidation of the metal wire
[123]. In the case of Au, itinvolves the formation of soluble
AuCl−4 , which subse-quently diffuses away from the electrode
[121].
The schematic of the electrochemical cell used isdepicted in
Fig. 2a. After careful cleaning with acetone,the Au wire (φ = 0.125
mm, purity 99.99%, temperas drawn, Advent Research Materials) is
partially im-mersed (∼2–3 mm) into the electrolyte solution. As
anelectrolyte, a 1:1 mixture of hydrochloric acid (HCl, aq.37%) and
ethanol is used. As cathode, a platinum wire(φ = 0.3 mm) circular
ring electrode with a diameterof ∼1 cm is utilized. It is held at
the surface of theelectrolyte with the Au wire positioned at the
centerof the Pt ring. For the etching, a potential of +2.2 V
isapplied to the Au anode with respect to the Pt cathode.This
voltage was determined by us, as well as others[124], to produce
the best tips. This value is well abovethe Au oxidation potential
due in part to the activationenergy along the reaction pathway
[114].
When placed in the electrolyte solution, the surfacetension
causes a concave meniscus to form around thewire, as shown
schematically in Fig. 2b. The overallshape and aspect ratio of the
tip after etching are
-
176 Neacsu et al.
Fig. 2 Schematics of theelectrochemical etching cell.a The Au
wire (anode) ispartially immersed into theelectrolyte solution. It
issurrounded by the Pt-ringcathode. The evolution of theetching
process is closelyfollowed using a videomicroscope. b The
etchingtakes place at the meniscusformed around the Au-wire.The
flow of AuCl−4 is shown.For W-tips, a similarprocedure is used (see
text).c Time evolution of theetching current for a potentialof 2.2
V. The currentoscillations are periodic forabout 100 s, after which
theybecome less frequent,although they maintain theamplitude. For
periodiccurrent oscillations (inset),tips with smooth surface
andconsistent taper are obtained.
primarily determined by the shape of the meniscus[121]. During
etching, a downward flow of AuCl−4 alongthe wire can be observed.
The resulting ion concen-tration gradient partially inhibits
etching of the lowerportion of the wire, resulting in a necking of
the wirenear the meniscus [125]. This proceeds until the
lowersection of the wire falls off. The remaining upper partof the
wire is then used as a TERS/AFM tip.
Since the tip remains in the solution under themeniscus after
the detachment of the lower part, thecircuit has to switch off as
rapidly as possible, as furtheretching would result in blunt tips.
A comparator breaksthe etching voltage when the current value
becomessmaller than an adjustable reference value determinedfrom
the current change associated with the drop of thelower tip.
Monitoring the etching current reveals periodic os-cillations of
the current, as is shown in Fig 2c. Forthe potential of 2.2 V, the
etching will generally reachcompletion in approximately 150 s with
an averagebaseline current of ∼2 mA. After a short time of ini-tial
fluctuations, the period equilibrates and remainsconstant for the
first ∼100 s of the etching duration.Towards the end of the etching
process, the oscillationperiod continuously decreases until
completion.
These current oscillations have been attributed tothe depletion
of Cl− near the surface of the elec-trode [121]. Initially, the Au
will react rapidly toform AuCl−4 , depleting the Cl
− near the electrode–
electrolyte interface and resulting in a period of highcurrent.
With decreasing local Cl− concentration, theAu will more readily
form an electrode–passivatinglayer of Au(OH)3 [126], leading to
extended periodsof decreased current. Upon restoration of the
localCl− concentration, the passivating layer dissolves andanother
current spike occurs [127]. The details of thisoscillating
electrochemical process are sensitively de-pendent on the applied
potential. An empirically es-tablished etching voltage leads to the
most periodicoscillations and results in the highest quality
tips.
It is desirable to have a criterion to select suitableTERS tips
other than scanning electron microscopy(SEM) which is known to
deposit Raman-visible car-bon contamination onto the tips due to
electron-beaminduced decomposition of trace organics in the
residualgas [128, 129]. Tips etched under a constant
oscillationperiod exhibit a smooth surface and consistent taper.In
contrast, tips etched with an irregular periodicityof the etching
current frequently present deformitiesand large surface
irregularities. We could verify in ourexperiments a link between
homogeneous and smoothtaper with TERS performance by comparison of
TERSactivity with SEM of tip shape as well as the study ofSPP of
the tip apex. Our observations here are in goodagreement with
previous work by Wang et al. [124].
A similar etching procedure is used for the tungstentips. A W
wire (φ = 200 μm) is partially immersed (∼2–3 mm) in aqueous 2 M
KOH. A DC voltage of 3 V
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 177
is applied between the wire and a stainless steel ringcathode.
Using these procedures, tips with apex radiias small as 10 nm are
obtained. After etching, the tipsare cleaned in distilled water and
stored in isopropanolprior to usage to avoid possible contamination
in anotherwise uncontrolled atmosphere.
Optical Tip Characterization
Efficient local-field enhancement and TERS activity is,in
general, associated with the excitation of local modesof SPPs at
the metallic tip [63, 73, 130–132]. However,determining the details
of this correlation has remainedan open problem. In the following
section, we presentthe investigation of the spectral
characteristics of theelastic light scattering from individual
sharp metal tipsand discuss the results in the context of the local
plas-monic resonant behavior.
SPPs of Au Tips
Dark-field scattering spectroscopy with white lightillumination
would lead to a largely unspecific responsewith the scattering
dominated by the tip shaft [133], andthe SPP characteristics of the
apex itself would becomedifficult to distinguish.
Therefore, for the plasmonic light tip-scattering ex-periments,
we spatially limit the optical excitation tothe near-apex region by
use of evanescent wave excita-tion. For that purpose, the tip
frustrates the evanescentfield formed by total internal reflection
on a prismbase [134, 135], and the tip-scattered light is
detectedand spectrally analyzed. This confines the excitationto
just several 100 nm from the tip apex. The com-plete description of
the setup and the results are givenelsewhere [93].
Figure 3 shows representative scattering spectra fordifferent Au
(a, b) and W (c, d) tips. Both the excitationand detected light
fields are unpolarized. All spectraare acquired for the tips within
a few nanometersabove the prism surface, as controlled by
shear-forceAFM. The intensity scale is the same for all four
cases,and the spectra are offset for clarity. Electron micro-graphs
for the tip structures investigated are shown asinsets.
The pronounced wavelength dependence of the scat-tering of Au
tips is characteristic of a plasmon resonantbehavior. Both
scattering intensity and spectral posi-tion of the resonance are
found to be sensitive to thestructural details of the tips. In
general, for regular tipshapes, the resonance is characterized by
one (Fig. 3a)distinct spectral feature. Inhomogeneities in the
geo-
Fig. 3 Scattering spectra for different Au and W tips.
Aplasmon-resonant behavior is observed for Au tips (a, b).
Weakerand, in general, spectrally flat signals are observed in the
case ofW tips (c, d). The spectral data are juxtaposed with the
electronmicroscope image from the corresponding tip. The scale
barcorresponds to 100 nm [93].
metric shape are reflected in spectral broadening
and/oroccurrence of multiple spectral features (Fig. 3b).
Inaddition, the spectral position and shape of the plasmonresonance
depend sensitively on the aspect ratio ofthe tip.
For comparison, spectral light scattering by tungstentips of
similar dimensions is shown in Fig. 3c and d.Overall weaker
emission intensities are observed com-pared to Au tips. For W, a
metal with strong polar-ization damping due to absorptive loss in
the visibleand near-IR region, no SPP resonance is expected.Here, a
spectrally flat optical response is observed withweak overall
scattering intensities (Fig. 3c). The spec-tral behavior is found
to show little variation with tipradius and tip cone angle, except
for the case of veryslender tips, where a modulation is observed as
shownin Fig. 3d.
From polarization-dependent studies of the tip scat-tering
process, we have found that more intense scat-tering is observed
for emission polarized parallel withrespect to the tip axis
(p-polarized emission), corre-sponding to the excitation of
longitudinal plasmonicmodes. The intensity ratio of p to s
(emission ortho-gonal to the tip axis) is typically found to range
be-tween 2 and 5, with a maximum value of ∼ 10. Inthe context of
the TERS experiments, it must benoted that the s-polarized field is
not expected tobe enhanced near a surface. In this case, the
optical
-
178 Neacsu et al.
polarization of the tip and the corresponding imagepolarization
induced in the sample are oriented antipar-allel, as will be
discussed in the “Optical Tip–SampleCoupling” section.
Local-Field Enhancement from Bare Tips
The highly resonant characteristics observed for Autips suggest
strong local-field enhancements in thevicinity of the apex. The
quantification of the near-field enhancement factor is a difficult
task in general.Without an absolute reference, the enhancement
factorcannot readily be quantified from linear optical
exper-iments. We therefore make use of the symmetry selec-tivity of
the second-order nonlinear optical response inthe form of SHG from
the apex region of the tips.
SHG is forbidden in the dipole approximation formedia with
inversion symmetry [136]. In scatteringgeometry for sagittal
illumination and tip-parallel po-larization, the SHG response from
the tip is dominatedby the apex region where the macroscopic
transla-tional invariance is broken in the axial direction. Withthe
symmetry being radially conserved, little signalis expected from
the conical near-apex shaft area.1
For the experiments, linearly polarized incident lightfrom a
mode-locked Ti:sapphire oscillator (pulse dura-tion < 15 fs, λ =
805 nm) is directed onto the sharp endof the free standing tip and
the scattered SHG signal isspectrally selected and detected.
The contribution of the local field-enhancement ofSHG from the
metal tips is derived by comparing thesignal strength obtained to
that of a planar surface ofthe same material. With the
SH-enhancement expectedto be dominated by the tip apex, a
SH-enhancementof ∼ 5 × 103 − 4 × 104 was observed for Au tips withr
� 20 nm. For a SH-power ∝ E4 [137], this corre-sponds to an
amplification of 8–25 for the averageelectric field near the apex,
in agreement with estimatesbased on other SHG experiments [138].
For W tips,significantly lower values for the SH enhancement
arefound corresponding to local field factors between 3and 6. These
results are also in good agreement withtheoretical models despite
microscopic variations in thedetails of the tip geometry.
The excitation of the localized SPP in the axial direc-tion is
responsible for the field enhancement observed
1In addition to the unique symmetry properties, the local
surfaceand nonlocal longitudinal bulk polarizations contributions
to thenonlinear polarization and their directional and polarization
se-lection rules are directly distinguishable here due to the
geometryof the tip as a partial asymmetric (∞mm) nanostructure with
themirror symmetry broken along the axis.
experimentally for Au tips. A systematic investigationof the
influence of these geometric parameters in termsof cone angle and
tip radius would be highly desirable;however, the limitations due
to the electrochemicalpreparation procedure render this
difficult.
Calculation of the Near-Field Distributionat the Tip-Apex
The near-field distribution and enhancement have beenderived
theoretically for a variety of tip model geome-tries and tip and
sample material combinations usingdifferent theoretical methods
[81, 92, 95, 130, 132, 139–142] (and references therein). The
accurate theoreticaltreatment of the problem involves the solutions
ofMaxwell’s equations. This can be performed numeri-cally for a
chosen model tip-geometry [94, 131, 143,144]. Although this may
closely reproduce the exper-imental observations, the approach is
computationallyvery demanding. In addition, it has remained
difficult toextract the underlying relevant microscopic
parametersresponsible for the optical response observed given
thatthe effects of tip geometry, tip material, tip–sampledistance,
and optical field are coupled.
Taking advantage of the small dimensions of the tipapex compared
to the optical wavelength (kr � 1, withk the wave vector and r the
tip-apex radius), the prob-lem can be treated in the quasistatic
approximation,which allows solving the Laplace equation
analyticallyfor certain geometries [95, 130, 132, 140–142] to
derivethe local field distribution [87, 145]. For a size of theapex
region of r ∼ 10 nm, this implies that the electricfield has the
same amplitude and phase across thestructure at any time; thus,
retardation effects can beneglected [133, 146]. Despite constraints
in terms ofthe tip geometries that can be treated in this
approach(paraboloidal, spheroidal, hyperboloidal), this
methodprovides direct insight into how the solutions scale
withseveral experimentally relevant structural and
materialparameters.
Here, we approximate the tip geometry as a hyper-boloid. The
influence of different dielectric and struc-tural parameters on the
near-field enhancement anddistribution is systematically derived
for both bare tipsand tip–sample systems [95]. The optical
wavelengthdependence is explicitly taken into account
consideringthe frequency dependence of the dielectric functions
oftip and sample media [147].
The field distributions and enhancement and theirspectral
dependence calculated within the quasistaticapproximation for a
hyperbolical tip are found to agreewith other detailed theoretical
observations. Using a
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 179
fully 3D finite-difference time-domain method, we havecalculated
the field distribution [94] with results simi-lar to those obtained
in [140, 144, 148]. The compar-ison with both the exact theoretical
treatments andexperimental results validates the approach of
treatingthe probe tip in the quasistatic approach to a
goodapproximation. Despite the simplicity of the model,the
essential optical properties and the physical trendscharacteristic
for the optical response of the tip–samplesystem are accurately
predicted.
Figure 4 shows characteristic local field distributionsand the
corresponding enhancement near the apexregion of free standing tips
of gold (a) and tungsten(b). The equipotential surfaces are
indicated by solidlines. In both cases, the radius of the apex and
thecone semiangle are fixed to r = 10 nm and to θ =
20◦,respectively. This closely resembles the conditions inthe TERS
experiments by setting the incident light ata wavelength of λ = 630
nm and the polarization alongthe tip axis.
At optical frequencies, even for metallic tip material,the tip
surface does not represent an equipotentialin contrast to the pure
electrostatic case. The finiteresponse time of the charge carriers
with respect to theoptical frequency results in the decay of the
field insidethe tip on the length scale given by the skin depth.For
gold as a representative material with high con-ductivity, this
results in the strongest field enhancement
Fig. 4 Dependence of field enhancement (E/E0) on tip materialfor
free-standing Au and W tips with apex radius r = 10 nm,
andwavelength of λ = 630 nm. The solid lines represent contours
ofconstant potential [95].
of E/E0 ≈ 50 at the apex.2 In contrast, tungsten is apoor
conductor in the optical frequency range leadingto a comparably
moderate enhancement of ∼12. Thedegree of field enhancement E/E0 at
the tip apexdepends sensitively on apex radius and cone
semiangledue to their influence on the plasmon resonance,
asdiscussed briefly below and in [95]. Typical values rangebetween
10 and 100 for gold tips with 10–20 nm radiusand realistic
semiangles.
Despite the necessary approximations inherent tothe quasistatic
approach, the theoretical results pre-sented here prove to be
sufficiently accurate for mostpractical purposes. This is drawn
from comparison ofthe field enhancement factors with the
experimentalresults of the “Optical Tip Characterization”
section.From the tip-scattered SHG experiments, the local
fieldenhancement of 8–25 for Au and 3–6 for W was esti-mated for r
= 20-nm apex radii [92]. Considering thatthe experimental
enhancement factors are obtained asa spatial average over the apex
region, these valuesfall well within the range of the theoretically
predictedenhancements given in Fig. 4.
Optical Tip–Sample Coupling
The local field enhancement, as well as the lateralconfinement,
can change significantly for the tip in closeproximity to a surface
plane. This behavior is of crucialimportance for the optical
contrast in scattering near-field microscopy. The optical
tip–sample coupling isthe result of the forcing of the boundary
conditionsat the surface plane on the field emerging from theapex.
With the incident electric field inducing an opticaldipole
excitation in the tip, the presence of the samplecan be accounted
for by considering a virtual imagedipole located inside the sample,
with the resultingfield distribution being a superposition of the
fields ofthe two dipoles [149]. This gives rise to a mutual
andconstructive tip–sample optical polarization when theelectric
field is oriented parallel with respect to the tipaxis
(p-polarized). For an s-polarized incident field, thetip-dipole is
induced parallel to the sample surface, andthe correspondent
image-dipole aligned antiparallel.This leads to a partial
cancellation and reduced fieldintensity and scattering [58].
Figure 5 (top panel) displays the evolution of thefield in the
tip–sample gap calculated along the axial
2Note that for the calculations that refer to the field at the
tipapex, the field is calculated at 0.125 nm below the apex to
avoidnumerical artifacts due to finite grid size.
-
180 Neacsu et al.
Fig. 5 Top: Variation of field enhancement E/E0 along the
axialdirection across the tip–sample gap region for different
distancesd for an Au tip (r = 10 nm) and Au sample at an
excitationwavelength of λ = 630 nm. The tip is at variable z = d
posi-tions and extends to the right. The sample surface is located
atz = 0 nm and its bulk occupies the range of negative
z-values.Bottom: Spectral dependence of field-enhancement with
tip–sample distance d for an Au tip (r = 10 nm) approaching anAu
surface. The pronounced red shift of the plasmon responseis
associated with the strong near-field tip–sample coupling ford ≤ r.
The lines represent contours of equal optical field enhance-ment
[95].
direction for different distances d for an Au tip ap-proaching a
flat Au sample. The tip–sample approach isaccompanied by a
significant increase in field enhance-ment in the tip–sample gap.
The tip–sample interactionis correlated with apex radius and
becomes significantat distances below about twice the tip radius
(here,d = 20 nm). Here, the near-field interaction becomeseffective
with a particularly fast rise of the field at the
sample surface. The local field enhancement reachesvalues of up
to several hundred for d = 2 nm, showingan increase of more than
one order of magnitude whencompared with the free-standing tip. In
addition, thefield enhancement with decreasing tip–sample
distanceis accompanied by a strong lateral confinement of thefield
underneath the apex [95]. The equipotential sur-face is forced to
align nearly parallel with respect tothe metallic substrate plane,
which gives rise to anenhanced lateral concentration of the field.
This is im-portant, as it leads to an increased spatial
resolutionin scanning probe near-field microscopy for small
tip–sample distances.
The details of the near-field distribution in the tip–sample gap
depend on the optical properties of thetip and the sample materials
that affect both field en-hancement and lateral confinement. The
lateral spatialresolution that can be obtained is determined by
thenear-field spatial extent and is given to the first orderby the
tip apex radius; however, deviations from thissimple scaling
behavior can be expected as the apexdimensions become smaller than
the typical metal skindepth of ∼20–25 nm at visible frequencies. In
addi-tion, the dielectric properties of both tip and
sampledetermine the details of the field distribution in the
gapregion. In particular, the lateral confinement decreaseswith a
decrease of the material optical polarizability, asdiscussed in
[95]. Probing metallic samples with metallictips at small
tip–sample gaps results in the highestspatial resolution. In
contrast, the lateral resolution willbe lower when imaging
dielectric surfaces in otherwiseidentical experimental
conditions.
One of the virtues of the quasistatic model is thedirect access
to the spectral variation of the field en-hancement and its
distribution for different tip–samplegeometries. The spectral
tip-scattered response can be-come a complex superposition of the
tip and the sampleoptical properties, the understanding of which is
im-portant in nanospectroscopy. Figure 5 (bottom panel)shows the
calculated spectral dependence of the fieldenhancement near the
surface for an Au tip (r = 10 nm)approaching an Au surface. As
expected, a structuralplasmon resonant behavior is observed.
Associatedwith the increase in field enhancement for shorter
dis-tances a spectral shift in the plasmon response to
longerwavelengths is observed. This red shift is especially
pro-nounced for distances d ≤ r, correlated with the onsetof the
sharp rise in field enhancement in the regime ofstrong coupling. It
is the result of the superposition ofthe dielectric functions of
the tip and the sample mate-rial mediated by the tip–sample optical
coupling. Thisis a general phenomenon and it is found in
calculationsof spheres and other plasmonic nanostructures in
close
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 181
proximity to metal surfaces [132, 144, 150, 151], and ithas been
observed experimentally in TERS and lightemission in inelastic
tunneling [152–154].
Predicted spectral peak widths of the order of 0.2 to0.3 eV
correspond to what is expected from the elec-tronic dephasing times
for SPP in Au of ∼ 4–7 fs [155].Using tungsten as tip material, no
plasmon behavior isobtained except for the case when it is combined
witha metallic sample that can itself sustain an SPP at
thecorresponding wavelength [95].
The determination of the field enhancement for atip–sample
coupled system has been experimentallyachieved by TERS from surface
MLs of molecular ad-sorbates [73, 77, 81, 85, 94]. From tip–sample
distance-dependent Raman measurements in comparison
withcorresponding far-field experiments using MG dyemolecules or
single-walled carbon nanotubes, near-field enhancements of 60–150
at the sample surfacewere measured, as detailed below. These
experimentalvalues are in good agreement with the theoretical
pre-dictions, which range from ∼50 to ∼300, as shown inFig. 5 for
small tip–sample distances.
Near-Field Character and Far-Field Artifacts in TERS
TERS manifests itself in an enhancement of the Ramanresponse,
with the increase confined to the region un-derneath the tip-apex.
However, with the illuminationextended on a larger surface region
determined by thefar-field focus, the discrimination of the
variation of far-field response due to the presence of the tip
inside thefocus is difficult (Hartschuh et al. [73]).
Without any lateral scanning or systematic verticaltip–sample
distance variations, this does not allow forthe unambiguous
assignment of the observed opticaleffect to a near-field process.
The apparent Ramansignal rise may be due to far-field effects
associated withthe tip being scanned inside the tight laser focus.
Thiscan influence both signal generation and detection. Asthe tip
penetrates into the focus region, it would scatteradditional
otherwise forward-scattered (nonenhanced)far-field Raman light back
into the detector. Further-more, the interference of the
tip-scattered and surface-reflected contributions with direct
incident pump lightresults in locally enhanced pump intensities.
With bothprocesses affecting a surface region not confined by
theapex area, near-field effects may be obscured by anincrease in
far-field signal.
In Fig. 6, tip-scattered Raman results are shownfor single-wall
carbon nanotubes and MLs of MG
Fig. 6 Near-field signature vs far-field artifact: Raman
spectraof single-walled carbon nanotubes and MG molecules with
tipretracted (SWNT, MG) and tip engaged (SWNT tip, MGtip).
Inset:tip–sample distance dependence of the Raman signal
obtainedunder similar conditions but displaying very different
behaviors:the ∼20-nm-length scale increase is characteristic for
the near-field signal origin (circles); the few hundred nanometer
decaylength (squares) shows a far-field artifact—leading to
similarsignal increase as the near-field response. Dashed lines
added asguide for the eye.
molecules with the tip in force feedback at d = 0 nm3vs tip
retracted by several 100 nm. When the tip iswithin several
nanometers above the sample surface, astrong increase in Raman
intensity is observed for bothadsorbates (spectra denoted SWNTtip
and MGtip). Notethat the difference in noise level from the SWNT to
theMG spectra is due to different spectral resolution set-tings of
the spectrometer. Although frequently used toassign the observed
TERS signal, a simple comparisonof surface vs tip-scattered Raman
intensities rendersnear- and far-field processes a priori
indistinguishable.The inset of Fig. 6 shows the Raman peak
intensity as afunction of the tip–sample distance obtained in two
sim-ilar experiments for MLs of MG molecules adsorbedon a flat Au
surface. The overall increase in signalis comparable in both cases,
and an estimate of theRaman enhancement factor gives G > 106
(vide infra).However, with the distance variation occurring on
alength scale correlated with pump wavelength or focusdimensions,
in one case, the enhancement can solelybe attributed to far-field
effect (squares). A true near-field effect manifests itself in a
correlation of the spatialsignal variation with the tip radius (∼20
nm). Here,with high-quality tip (sharp apex, smooth tip shaft),
3Here, d = 0 nm is defined ascorresponding to a 20–30% de-crease
in the shear-force amplitude
-
182 Neacsu et al.
the near-field contribution can dominate the overallsignal
(circles). Therefore, for the tip-scattered Ramansignal, only the
demonstration of a clear correlation ofthe lateral or vertical
tip-molecule distance dependencewith tip radius allows for an
unambiguous near-fieldassignment of the optical response [73]. This
is true forall near-field microscopies including s-SNOM and
thespecial case of TERS [16, 19, 59, 65].
Experimental Quantification of the Near-FieldRaman
Enhancement
In contrast to SERS, where the quantification of theenhancement
is generally a difficult task, for the tip-scattering experiment,
the Raman enhancement factorcan be derived from comparison of
tip-enhanced vsfar-field response of the same surface ML.
Figure 7 shows the spectrally resolved tip-scatteredRaman signal
during approach of ∼ 1 ML of MG ongold (2 nm/step, 1 s/spectrum
acquisition time). Thepump light is polarized along the tip axis
(pin) andthe Raman signal is detected unpolarized. Althougha faint
Raman signature of the molecules is observedwith the tip at d >
100 nm, a clear molecular fingerprintis obtained only when the tip
is within ∼20 nm fromthe sample. The prominent bands around 1,615
and1,365 cm−1 are assigned to combinations of the C=Cstretching
vibrations of the phenyl ring and the modeat 1,170 cm−1 is due to a
methyl group rocking mode oran in-plane C-H bending mode of the
phenyl ring [156].
The enhancement is confined to a tip–sample spac-ing of just
several nanometers and correlated with theapex radius of the tip,
as expected for the near-fieldsignature. The increase in Raman
response is accom-panied by a weak rise in a spectrally broad
fluorescencebackground that has been subtracted. With the
molec-ular fluorescence being quenched due to the electronic
Fig. 7 Tip–sample distance dependence of spectrally
resolvedRaman signal during approach of ∼1 ML of MG on gold.
Eachspectrum is acquired for 1 s and the approach is realized with
2-nm increments. Near-field tip-enhanced signal is observed withthe
tip within ∼20 nm above the sample, displaying typicalRaman modes
for MG molecules [81].
coupling to the metal substrate, this emission couldlargely be
attributed to the enhancement of the intrinsictip luminescence in
conjunction with excitation of plas-monic modes in the tip–sample
cavity [152, 157], i.e., itsorigin is independent of the molecular
adsorbates.
With the near-field character of the Raman responseverified, the
experimental field-enhancement factor canbe derived from comparison
of the tip-enhanced vs far-field response from the same surface ML.
For the ex-periments presented here, the integrated Raman
signalover the 1,150–1,650 cm−1 spectral region is used
afterbackground subtraction.
As shown in the “Calculation of the Near-FieldDistribution at
the Tip-Apex” section, the electromag-netic near-field enhancement
originates from a sam-ple surface area approximated by the area of
the tipapex. For the evaluation of the enhancement factor,the
different areas probed in the near-field (TERS) andfar-field (FF)
cases are then taken into account. Forthe TERS setup, the
illumination focus has a diameterd � (λ/NA) × 1.5 = 2.7 μm (the
empirical factor of 1.5accounts for the deviation of the laser beam
from aperfect gaussian profile). Considering the ∼70◦ angleof
incidence of the pump light with respect to thesurface-normal in
our setup, the actual surface regionilluminated is elongated
elliptically and larger, with atotal area of ∼ 17μm2. In the case
of the confocal–Raman setup (FF) used to record the far-field
spectrumof MG, the same laser illuminates an area of 870 nm2.The
molecule is estimated to occupy a surface areaof ∼ 0.87 nm2 (actual
3D space filling: 1.18 × 1.39 ×0.98 nm). One ML surface coverage
then correspondsto approximately 106 molecules in the focus of
thefar-field setup, and only < 200 are responsible for
thetip-enhanced signal for a tip with 10-nm apex radius.
In addition, the detection efficiencies for the two dif-ferent
experimental arrangements are considered. Thefar-field response is
emitted only in half space (solidangle � = 2π) assuming an
isotropic dipolar intensitypattern, and with NA = 0.9, a total of ∼
27% of theRaman signal is detected, in contrast to the
TERSexperiments with ∼2% for NA = 0.35 and assuming theemission
pattern following a cos2(θ + π2 , ) dependence(θ ∈ [0, π ]).
Taking all these factors into consideration, theRaman
enhancement factor can be estimated, and it isfound to range from
106 to 109, with variation mostlydepending on the tip used. This
enhancement is ex-pected to be electromagnetic in origin. The
moleculesexperience the tip-enhanced local-field Eloc(νi) =
L(νi)E(νi), with L(νi) as the enhancement factor of theincident
field E(νi). The concomitant enhancement ofthe polarization P(νs)
at the Raman-shifted frequency
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 183
Fig. 8 Field enhancement (E/E0) for two different Au
tipsapproaching ∼ 1 ML of MG on a gold surface. The Ramanresponse
is integrated over the dominant peaks in the 1,150–1,650 cm−1
spectral region after background subtraction. Theenhancement factor
is calculated according to ITERS ∝ E4.
νs is L′(νs). Therefore, the total field enhancementis given by
L(νi) · L′(νs) [93, 137]. With the Ramanintensity I ∝ |L(νi) ·
L′(νs) · E(νi)|2, the total Ramanenhancement factor G is given by G
= |L(νi) · L′(νs)|2[158]. Although different in general, the field
enhance-ment factors L(νi) and L′(νs) for pump and
Ramanpolarizations, respectively, can be assumed to be similarin
this case [159, 160]. This is motivated by both thespectrally broad
plasmonic resonance of the tip and itssmall red-shift upon
approaching the sample surface[95, 132, 161].
Figure 8 shows the effective field enhancement fac-tors
(L(νi)·L′(νs) = E/E0) for the integrated Raman
signal from ∼ 1 ML of MG on gold as a function oftip–sample
distance for two different Au tips (bothwith r < 15 nm). Maximum
enhancements of ≥90 and≥60 (black and blue curves, respectively)
are obtainedconsidering a variation of the tip-scattered Raman
re-sponse with the fourth power of the electrical field,
asindicated above. The molecules adsorbed on the planarAu surface
already experience a field enhancementgiven by the Fresnel factor
due to reflection of theincident light at the surface [162]. At 633
nm and θ =70◦, the Fresnel factor is equal to 0.95, resulting in
anelectric field enhancement of 1.95. We will thereforealso derive
the total enhancement with respect to thefree molecule
response.
The study of the polarization dependence of theRaman response
offers additional insight into the elec-tromagnetic enhancement of
TERS. Figure 9 showsnear-field Raman spectra from ∼ 1 ML of MG
ongold for the different polarization combinations of bothpump and
Raman light. Incident laser power and ac-quisition times are
identical for all spectra. The po-larization directions are defined
as parallel (p) andperpendicular (s) with respect to the plane of
incidenceformed by the incoming wave vector k(ωi) and thetip axis.
No background has been subtracted and thedata are normalized with
respect to the intensity ofthe 1,615 cm−1 mode measured in the
pin/pout polar-ization configuration (upper left panel).
With the incident field polarized perpendicular tothe tip axis
(sin), almost no Raman signal is observed,irrespective of the
polarization of the scattered light.In contrast, with the pump
polarized along the tipaxis (pin), clear Raman fingerprints of MG
molecules
Fig. 9 Polarizationdependence of the near-fieldtip-enhanced
Ramanresponse originating from∼ 1 ML of MG on gold. Allspectra are
acquired for 30 sand normalized to themaximum intensity value ofthe
1,615-cm−1 mode inpin/(unpol.)out geometry. Avery weak Raman
response isobserved for polarizationcombinations other
thanpin/pout, where strongnear-field coupling gives riseto a strong
Ramanenhancement. The spectrallybroad background is due tothe
intrinsic luminescencefrom the Au tip itself.
-
184 Neacsu et al.
are observed, with the Raman response being pre-dominantly
polarized parallel to the tip (pin/pout) asexpected for near-field
TERS from isotropically dis-tributed molecules with diagonal Raman
tensor com-ponents as is the case of MG. For both the sin/pout
and the pin/sout configuration, only a weak specificvibrational
Raman signal is observed due to the ab-sence of the tip–sample
optical coupling. For sin/sout,a larger background is observed,
albeit with no Ramanenhancement, as expected.
The highly polarized TERS response observed inour experiments
indicates the absence of significantnear-field depolarization for
the homogeneous andslender tip geometries used. This is required
forsymmetry-selective probing in TERS and other non-linear
tip-enhanced processes [163–165] (Neacsu et al.,unpublished
manuscript) that rely on polarization se-lective and conserving
light scattering. In contrast, fora Ag-particle-topped quartz AFM
probe as used forprobing the 520-cm−1 Raman band of Si [166],
theobserved Raman depolarization has been attributed tothe wide
cone angle of the tip [119].
TERS of Molecular Adsorbates
In Fig. 10, representative tip-enhanced Raman spectraare shown
for MG on smooth Au surfaces. They areacquired for the same surface
coverage of ∼1 ML, butusing different tips exhibiting enhancements
of 3 × 108(a), 7 × 107 (b), 1 × 107 (c), and 1 × 106 (d),
respec-tively. The experimental uncertainty is estimated at afactor
of 3–5 for each value. The tip-enhanced Ramanspectra are
reproducible for a given tip. However, asseen from the data, the
spectral details vary from tipto tip with increasing enhancement.
With the lateralconfinement of the tip-enhancement within a
∼10-nm-diameter surface region and a molecular density of∼ 1/nm2,
we estimate that the signal observed inFig. 10a–d originates from
∼100 molecules.
In the lower panel of the figure, the far-fieldRaman spectrum
from the same sample is shown forcomparison (black line). The
far-field spectrum closelyresembles that in aqueous solution [156],
indicatingthat the molecules are physisorbed in isotropic
orien-tation at the surface. The blue bars represent normalRaman
modes of the MG anion calculated using densityfunctional theory as
implemented in Gaussian03 as dis-cussed in [81]. The assignment and
spectral position ofthe calculated modes agree well with literature
values[156, 167].
For moderate near-field enhancements ≤107, bothspectral
positions and relative intensities of the modes
800 1000 1200 1400 1600 1800
Far-field
Ram
an In
tens
ity (
arb.
u.)
Wavenumber (cm-1)
d)
c)
a)Tip-enhanced
b)
Fig. 10 Tip-enhanced Raman spectra for ∼ 1 ML of MG fordifferent
degrees of enhancement (a, b, c, d) in comparisonwith the
corresponding far-field Raman spectrum (bottom graph)and DFT
calculation for the mode assignment (blue bars). TheRaman
enhancement factors derived are 3 × 108 (a), 7 × 107 (b),1 × 107
(c), and 1 × 106 (d), respectively. Data are acquired for1 s (a),
100 s (b, c), and 30 s (d), respectively. Spectral resolu-tion is
25 cm−1 for the near-field and 1 cm−1 for the far-fieldspectra
[96].
resemble the far-field signature, as seen in Fig. 10c andd and
in accordance with other TERS results for similarenhancements [79].
However, with increasing enhance-ment of 7 × 107 (b), and most
pronounced for 3 ×108 (a), the vibrational modes start to look
markedlydifferent. While some of the modes are present inboth
far-field and near-field spectra (e.g., those at: 920,1,170, and
1,305 cm−1), some others are only presentin the highly enhanced
near-field results (e.g., 1,365,
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 185
1,544 cm−1). The vertical dashed lines in Fig. 10 areadded as a
guide to the eye for easier comparison.
The DFT calculation allows for the identificationof the spectral
features in the far-field data of Fig. 10.The majority of the
intense Raman modes can be at-tributed to modes either localized at
the phenyl ring ordelocalized over the two dimethylamino phenyl
groups.In the spectral region of 910 to 980 cm−1,
severalvibrational modes, typically characterized by
in-planeskeletal bending and/or out-of-plane C-H motions, arefound.
The 1,170-cm−1 mode may be assigned to amethyl group rocking mode
or an in-plane C-H bendingmode of the phenyl ring. Around 1,300
cm−1, in-planeC-H deformation modes and C-C stretching modes ofthe
methane group are located.
Furthermore, the calculations show that the newspectral features
seen in the highly enhanced near-fieldspectra in Fig. 10 mostly
correspond to vibrational nor-mal modes of MG. Among the
characteristic near-fieldenhanced modes, e.g., the peak at 1,365
cm−1, whichis very strong in the far-field spectrum but
decreaseswith increasing enhancement, can be assigned to
com-bination of the C=C stretching motions of the aromaticring. In
contrast, the prominent peak at 1,544 cm−1,which dominates for the
highest enhancement is veryweak for small enhancements or in the
far-field spectra.Here, the calculation shows a mode characterized
bystretching motions combined with in-plane C-H bend-ing motions of
the conjugated di-methyl-amino-phenylrings. The two modes at 1,585
and 1,615 cm−1, whichcan be assigned to C=C stretching vibrations
of thephenyl ring, decrease with enhancement.
This change in both intensity and spectral signa-ture with
increasing near-field enhancement togetherwith the vibrational
analysis shows that the Ramanpeaks observed correspond to
vibrational modes ofMG, whereby different selection rules seem to
applyfor the Raman spectra obtained under condition of
highenhancements [158].
In the following, we discuss different physical mech-anisms
possibly responsible for this mode selectivity.Due to the high
localization of the optical near-field,the molecules in the
tip–sample gap experience a largefield gradient and different Raman
symmetry selectionrules can come into play [168–170]. The relevant
termsof the dipole moment μa of a molecule situated in thepresence
of a high field-gradient are:
μa ={ (
dαabdQ
)0
E +(
dμpadQ
)0
E + αab(
dEdQ
)0
+ 13
∂ E∂c
(dAabc
dQ
)0
}Q,
with E as the amplitude of the applied electric field,μ
pa the permanent molecular dipole moment, αab the
polarizability tensor at vibrational frequency, and Aabcthe
quadrupole polarizability [171]. Q represents thecoordinate of
vibration and {a, b , c} are a permutationof the coordinates {x, y,
z}.
The first and second terms of the equation describethe normal
Raman emission and the IR absorption,respectively. For an IR mode
to be active, a variationin the dipole moment along the vibration
coordinateis necessary [172]. The presence of the field
gradientfulfills this requirement and thus allows for IR modes
tobecome Raman active in that case. However, for MG,it was found
that the few IR modes show no significantresemblance with the
highly enhanced TER spectra(details in [81]).
The third term denotes the gradient field Raman(GFR) effect
[171]. The mechanism by which strongfield gradients can influence
the molecular Ramanspectra by altering the selection rules require
that thepolarizability tensor (αab ) and (dEb/dQ)0 must
simul-taneously be nonzero. The resulting selection rules re-semble
the surface selection rules [173] and give rise toGFR lines
observed in addition to the normal Ramanmodes.
The optical field gradient can also couple to vibra-tions via
the derivative of the quadrupole polarizabilityAijk of a mode, as
described by the last term of theabove equation [170]. With αab and
Aijk transformingdifferently in terms of symmetry and their ratio
beinghighly mode-dependent, this could also account for themode
selectivity observed in tip-enhanced Raman scat-tering [174]. This
might help to resolve the striking ob-servation that the strength
of the calculated four modesat 846, 988, 1,029, and 1,544 cm−1 are
overestimated byDFT calculation as compared to the far-field
spectra,but coincidentally represent modes that are
relativelystrongly enhanced in the TER spectra.
While it is difficult to quantify the contribution ofGFR in
general [171], the occurrence of normally for-bidden modes in the
TERS results suggest that theprocess contributes in this case.
Similar differences be-tween far- and near-field Raman spectra were
reportedpreviously in fiber-based SNOM experiments on Rb-doped
KTiOPO4 [20, 21].
In addition, the resonant Raman excitation leads toa coupling of
electronic with vibrational transitions,resulting in different
selection rules. Such vibronicallyallowed transitions may further
be influenced by thepresence of the strong field gradient and will
affect therelative peak amplitude of the observed Raman modes[172].
This effect would resemble the “chemical effect”observed in SERS
[35].
-
186 Neacsu et al.
The pronounced spectral differences between thetip-enhanced and
far-field Raman response have sim-ilarities to observations made in
SERS, where vibra-tional modes that are normally not Raman
allowedhad been found [175–177]. Aside from orientational ef-fects,
these spectral variations are typically interpretedto arise from
conformational changes and/or transientcovalent binding of the
molecule at “active sites” [173].With that being unlikely in the
tip-enhanced Ramangeometry discussed here, this indicates that the
kindof spectral selectivity observed in our TERS experi-ments
should be attributed primarily to electromag-netic mechanisms.
Molecular Bleaching
An important question regarding the appearance of dif-ferent
spectral features in the near-field tip-enhancedRaman spectra is
the influence of the molecular bleach-ing or other decomposition
products [178, 179].
With the molecules under investigation exposedto the strongly
localized and enhanced near-field,this leads to a sometimes rapid
photo-decompositionprocess, especially in a resonant Raman
excitation.
To probe for the possible appearance of photoreac-tion products
and their signature in the Raman spectra,the evolution of the Raman
emission is monitored intime-series experiments. Figure 11 (left
panel) showsconsecutive near-field Raman spectra acquired for 1
s
each for an enhancement of 1.3 × 107 with an inci-dent laser
fluence of 3 × 104 W/cm2. The moleculesbleach on a time scale of ∼
100 s, depending on theenhancement level, and the decay and
subsequent dis-appearance of the spectral response is uniform,
i.e.,the relative peak amplitudes are maintained. Duringthe
bleaching, no new spectral features appear frompossible
photoreaction products and the signal decayswith the relative peak
amplitudes remaining constant.After complete bleaching, no
discernible Raman re-sponse can be observed. The fluctuation
observed in thetime series is expected given the small number of ∼
100molecules probed in the near-field enhanced regionunder the tip
apex. The sum over all spectra (top graph)or the sum of any large
enough subset even at latertimes, i.e., after substantial bleaching
has already oc-curred, closely resembles the far-field response of
MGand, thus, allows us to attribute the Raman response toMG
molecules.
The same behavior of a gradual and homogeneousdisappearance of
the Raman response without a rel-ative change in peak intensity is
also observed forlarger enhancements, i.e., the case where a
differentmode structure is observed. However, the larger localfield
experienced by the molecules leads, in general, todecreasing decay
time constants.
Figure 11 (right panel) shows the decay kineticsof the
spectrally integrated Raman intensity for thetime series data shown
on the left. This is shown incomparison to the integral intensity
of the region from
Fig. 11 Left: Time series of100 consecutive near-fieldRaman
spectra (acquired for1 s each) for ∼1 ML MG onAu (Raman
enhancement1.3 × 107). The signal decaysto zero due to bleaching
andno new spectral featuresappear from thephotoreaction products.
Thesum Raman spectrum(top panel) clearly resemblesthe far-field
spectrum. Right:Bleaching kinetics derived forthe spectrally
integratedRaman time series and theregion from 1,550 to1,650 cm−1
(inset) [96].
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 187
1,550 to 1,650 cm−1 encompassing only the two promi-nent modes
(inset). Assuming an exponential decaybehavior of I/I0 = exp(−t/τ)
for the Raman intensity,a decay time τ = 40 ± 5 s is derived for
both cases(solid lines). From the applied laser fluence of 3 ×104
W/cm2 and the enhancement of the pump intensityof ∼ √1.3 × 107, the
bleaching is induced by a localpump fluence of 4.7 × 107 W/cm2.
An extreme case of molecular bleaching is shown inFig. 12. A
time series of 100 consecutive Raman spectrafrom a sample with
submonolayer molecular coverage,with each spectrum acquired for 1
s, is recorded. Theestimated Raman enhancement factor is 9 × 107,
andthe spectral features deviate from the far-field Ramanspectrum
presented in Fig. 10b, akin to the spectrashown in panels a and b
of the same figure for highenhancement level (e.g., the mode at
1,544 cm−1). Afteran illumination time of about 50 s, the overall
Ramanintensity drops suddenly over the whole spectral region,and
most visible for the peak at 1,306 cm−1, but no newspectral
features emerge. Thus, even for extreme cases
Fig. 12 Time series of 100 near-field Raman spectra acquiredfor
1 s each. After ∼50 s, the Raman intensity reduces due tobleaching,
but no signature of secondary products is visible.
of bleaching, the molecular decomposition products donot
contribute to the observed Raman signal.
It was suggested that the molecular bleaching ratecould be used
for the derivation of the enhancementfactor [112]. It should be
noted that the bleaching rateis not a characteristic physical
quantity universal fora given molecule. For example, MG
isothiocyanate—asister dye of MG—was found to bleach with a rate
con-stant more than two orders of magnitude higher thanthe MG
studied here [112], if renormalized to the sameexperimental
conditions. Bleaching mechanisms canbe quite diverse [180]. They
may include irreversiblephotoinduced or even multiphoton induced
reactionssuch as rearrangements, dissociation and fragmenta-tion,
elimination or hydrogen abstraction, or perhapsphotooxidation with
ambient oxygen via triplet states.It can depend on, e.g., humidity
or cleanliness of the tipor sample, and is, hence, not a useful
measure to com-pare experiments performed under different
conditionsin different laboratories.
With the experiments carried out under ambientconditions,
special care must be taken to use clean sam-ples and tips. It is
well known [181] that contaminatingorganic compounds and/or carbon
residues can adhereto either tip or sample and reveal their Raman
signa-ture in the tip-enhanced spectra, e.g., Raman bands
ofappreciable intensity at frequencies above 1,750 cm−1for carbon
[182].
A carbon cluster Raman response (see Fig. 6 in [96])manifests
itself in characteristic spectral features muchdifferent from the
spectral response discussed abovefor ML or submonolayer MG. In
accordance withprevious observations [181–183], the carbon
Ramanresponse is comparatively large and fluctuates rapidlyin an
uncorrelated way. In contrast to the data on MG,a distinct spectral
feature emerges around 2,000 cm−1,which has been assigned to, e.g.,
modes within thesegments of carbon chains [182], and which is
absentin the TERS spectra of MG. This absence of the carbonRaman
response can be understood since the bleach-ing of ML and
submonolayer MG coverage leads tosmaller molecular fragments and,
subsequently, to adilute surface carbon distribution and, hence,
does notlead to extended carbon chains and aggregates, whichcan
readily form by multilayer MG decomposition atambient
temperatures.
Besides the degradation of the analyte, anotherpotential source
of carbon contamination is the near-field probe itself. At room
temperature and in anoncontrolled atmosphere, contaminants from the
en-vironment could adsorb onto the tip surface and re-veal highly
enhanced Raman signals. TERS controlexperiments with the bare tip
and clean Au samples
-
188 Neacsu et al.
prior to MG deposition were carried out to confirm theabsence of
vibrational Raman signature. Only intrinsictip luminescence has
been observed in that case [81].
TERS with Single Molecule Sensitivity
With the measured tip-enhanced Raman spectra pre-senting a
signal-to-noise ratio of more than 40:1 whenprobing ∼100 molecules
for 1 s accumulation time, thisdemonstrates the potential for even
single-moleculesensitivity. Figure 13 shows near-field
tip-enhancedRaman spectra measured in a time series with 1
sacquisition time for each spectrum for a sample pre-pared with
submonolayer surface coverage, adjusted toexpect on average < 1
molecule under the tip-confinedarea of ∼ 100 nm2.
Here, the tip has been held at a constant distanceof d = 0 nm
above the sample and the total Ramanenhancement was estimated at 5
× 109.4 The observedRaman signal exhibits temporal variations of
relativepeak amplitudes and fluctuations in spectral position.These
are characteristic signatures of probing a singleemitter in terms
of an individual molecule. Similar ob-servations have been made
before in SERS [37–39, 41]with the fluctuations in the
spectroscopic signature ofa single emitter typically attributed to
changes in itslocal environment, its structure, molecular
diffusion[36, 184], and changes in molecular orientation [179].With
MG only physisorbed, it has to be considered inparticular that the
molecules can diffuse in and out ofthe apex-confined probe region
while experiencing dif-ferent degrees of enhancement. The diffusion
dynamicscan be facilitated by the thin water layer present on
thesample surface under ambient conditions.
In the time series in Fig. 13, the apparent bleachingrate seems
reduced compared to what is expected fromthe analysis of the
ensemble bleaching discussed above.This is a result of the dilute
surface coverage where newmolecules directly neighboring the
tip–sample gap dif-fuse into the near-field enhanced region.
However, thesignal vanishes rapidly after 100 s due to the
depletionof molecules in the immediate near-apex region.
A different mode structure of single-moleculeMG Raman is
observed compared to the far-field response—especially pronounced
for the 1,500–1,600-cm−1 spectral region, akin to the results for
theensembles under strong enhancements shown above.Figure 13
(middle panel) displays the sum spectrum
4Here, d = 0 nm is defined as corresponding to a 20–30%
de-crease in the shear-force amplitude.
(olive) for the single-molecule response time series(left) in
comparison with ensemble spectra of ∼100molecules probed for 1 s
(red) showing different de-grees of enhancement: 3 × 108 in a and 7
× 107 in b.Panel c presents the sum spectrum (blue) of the
datashown in Fig. 11a that, as for a moderate enhance-ment, closely
resembles the far-field spectrum. Theresemblance of the Raman
spectrum of a molecularensemble with the sum spectrum over the
whole timeseries offers strong evidence of probing single intactMG
molecules [40]. Note that the sum over the timesseries shall
therefore resemble an ensemble spectrumfor large enhancement rather
than the far-field Ramanspectrum. In assessing the resemblance of
the spectralcharacteristics, it has to be considered that, at
lowcoverage, the molecules have more degrees of freedomto
dynamically change orientation and they can diffuse.Given the
rather weak response, the signal detected canonly be expected to
emerge from the region of largestenhancement, and with the
diffusing molecules probingthe spatial variation of the enhancement
under the tip,this corresponds to an extreme case of
inhomogeneousbroadening. Therefore, while individual spectral
fea-tures at positions in accordance with the strongly tip-enhanced
near-field response are observed in the timeseries, the sum
spectrum no longer exhibits clearly re-solved lines. This
interpretation is further corroboratedconsidering, e.g., the
improved resemblance of the peakin the 1,550–1,600-cm−1 region of
the single moleculesum spectrum with the sum of the two near-field
spectra(in a and b) of different enhancement.
Further insight is obtained by studying the statisti-cal
behavior of the single-molecule Raman response[36]. Figure 13
(right panel top graph) displays theintegrated 1,430 to 1,650 cm−1
spectral intensity forthe data in the left panel. The signal
intensities clusterwith intervals of 170–230 counts·s−1, as already
evidentfrom visually inspecting the time-series of
integratedintensities (dashed lines), and this manifests itself in
thecorresponding histogram in an asymmetric distributionwith
discrete peaks (inset). This behavior is qualita-tively
reproducible for experiments with the same sur-face coverage, and
it can be interpreted as the Ramanemission from n = 0 (noise peak),
1, 2, and 3 moleculesbeing probed under the tip, as suggested for
similarfindings in SERS [34, 36]. This assignment is corrob-orated
from experiments with different surface cover-ages: for lower
coverages, only the n = 0 and 1 peaksremain, and with increasing
coverage, the distributionconverges to a narrow random Gaussian
distribution,which is observed from a large molecular ensemble,
asseen in the lower panel, where 100 consecutive far-fieldspectra
were acquired for 100 s each and the signal is
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 189
Fig. 13 Left: Time series of tip-scattered Raman spectra fora
submonolayer MG surface coverage with Raman enhance-ment of 5 ×
109. The spectral diffusion observed is character-istic for
observing single MG molecules. Middle: Comparisonbetween sum
spectrum of data shown in left panel (olive) and tip-enhanced Raman
spectra for different degrees of enhancement(red, a and b) and sum
spectrum of data shown in Fig. 11 (blue).Right: Temporal variation
of the Raman intensity of the inte-
grated 1,480–1,630 cm−1 of time series shown in left panel
(top).From the corresponding histogram (inset) a discretization
ofRaman intensities can be seen with 170–230 counts·molecule−1 ·s−1
(dashed line increment). For comparison, temporal variationof the
Raman intensity of the integrated 1,480–1,630 cm−1 of bigmolecular
ensemble (far-field) is shown (bottom) together withthe
corresponding histogram (inset) [81, 96].
integrated over the same spectral range as in the caseof the
single-molecule experiment. The details of thehistogram, however,
depend on the binning procedure,especially for a small data set, as
has been shown tobe insufficient as the sole argument for
single-moleculeobservation [185].
The optical trapping and alignment of MG under thetip must also
be considered as a possible source of theobserved surface diffusion
and intensity fluctuations[186]. However, for MG together with our
particularexperimental conditions, this cannot explain the
dis-cretization of Raman peak intensities in the single-molecule
response, as detailed elsewhere [81].
Our single-molecule TER results are similar to otherrecent
experimental findings [85] where brilliant cresylblue (BCB)
molecules adsorbed on planar Au wereprobed. For enhancements of
∼107 similar to our inter-mediate values, temporal fluctuations in
both intensityand mode frequency were observed, albeit with
nosignificant spectral differences between the far-field
andnear-field response, as seen here for high enhance-ments.
Recently, spatially resolved TERS imaging andidentification of
individual BCB molecules directly con-
firmed our singe-molecule assignment in the Ramanresponse in the
results discussed above [187].
Raman Imaging of Nanocrystals: Near-FieldCrystallographic
Symmetry
With the TERS capability of detecting molecules downto the
single emitter level, it is highly desirable toextend the technique
to also probe other classes ofmaterials such as crystalline
nanostructures. While far-field Raman microscopy has been
successfully used forsuch investigations [188, 189], the
applicability of TERShas not yet been explored.
In the following, we propose and discuss the po-tential of TERS
for the spatially resolved determina-tion of the crystallographic
orientation of crystallinematerials by taking advantage of the
fundamental in-teraction of the Raman process with lattice
phononstogether with the symmetry properties and selectionrules of
the tip-enhanced scattering. While the applica-tion of the symmetry
properties of the Raman selectionrules in a tip-enhanced geometry
has been emphasized
-
190 Neacsu et al.
previously [119, 190–192], no discussion of the ap-plicability
to crystalline nanostructures has yet beenprovided.
With Raman scattering being less invasive than elec-tron or
X-ray techniques and applicable in situ, thisapproach will fill a
much needed gap in the characteri-zation of nanostructured
materials with increasing com-plexity [193, 194]. Transmission
electron microscopy,capable of providing atomic resolution [195],
requiressamples thin enough to be transparent for
electrons,extensive sample preparation, and vacuum condi-tions,
making in situ experiments difficult [196]. Like-wise, X-ray
microscopy is capable of characterizingnanostructures with atomic
resolution [197], but it re-quires a monochromatic brilliant
synchrotron radiationsource and radiation beam damage remains a
con-cern [198]. Here, the comparable simplicity of TERSfrom an
instrumentation perspective makes it highlyattractive, providing
complementary information andeven avoiding some of the
disadvantages of the existingtechniques.
In Raman spectroscopy, the specific phonon modesprobed depend on
the chosen experimental geometry,in terms of the incident and
detected polarization, aswell as the propagation direction of light
[5, 199]. Thesephonon modes allow determination of the
crystallo-graphic orientation of a sample. This has been shownin
far-field Raman in, e.g., the study of 90◦ domainswitching in bulk
BaTiO3 [200] or the observationof ferroelastic domains in LaNiO4
[201]. However, inRaman microscopy, in the commonly used
confocalepi-illumination and detection geometry, this reducesthe
available degrees of freedom, thus resulting in theloss of the
general capability to probe the symmetry-specific Raman tensor
elements. In extending the use ofthe Raman selection rules to a
side-illuminated TERSgeometry, these degrees of freedom can be
regainedand even further refined by taking into account the
tipgeometry.
The intensity of the Raman scattered light from amedium is given
by: Is ∝ | �es · R̄ · �ei|2, where �ei and �esare the polarization
of the incident and scattered light,respectively, and the Raman
tensor R̄ is the derivativeof the susceptibility tensor [202]. As
an example, for aRaman-active phonon mode of tetragonal BaTiO3, R̄
isgiven by:
A1(ξ̃ ) =⎛⎝ a 0 00 a 0
0 0 b
⎞⎠ ,
where �ξ denotes the polarization direction of the mode(for
polar modes). The symmetry of a given mode, in
this case A1, is determined from group theory and maycontain
multiple component phonon modes of differentfrequency [203]. Thus,
when the polarization condi-tions, determined from the
susceptibility derivative, aresatisfied for a given symmetry mode,
the Raman shiftdue to the phonons belonging to that mode can
beobserved.
In addition, one can selectively isolate specificphonon modes
within a symmetry mode. For polarmodes, the phonons will separate
into transverse op-tical (TO) and longitudinal optical (LO)
components[204], which, being distinct in frequency, can be
spec-trally resolved. The Raman tensor methods describedabove do
not account for the distinct frequencies ofTO and LO modes nor
describe how to selectively ex-cite these, requiring further
refinement of the selectionrules
For a given geometry, the wavevector �q of the prop-agating
phonon can be determined by conservationof momentum from the
wavevectors of the incidentand scattered light. Based on the
relative orientationbetween �q and �ξ , one can selectively excite
the LOmode for �q ‖ �ξ , or the TO mode for �k ⊥ �ξ . Thus,
theobservation of either a TO or an LO mode providesthe additional
information about the orientation of thecrystallographic axes.
Furthermore, drawing on the nanoscopic apex of aplasmonic tip
for preferential enhancement of incidentlight polarized along the
tip axis allows us to exploitthe unique symmetry selection rules
associated with thetip [92], selecting modes with polarizations
parallel tothe tip axis. Modes for which either the incident
orscattered polarization coincide with the enhancementaxis may also
be observed, albeit with a lower intensity.
By appropriate selection of the polarization andpropagation
directions of the incident and scatteredlight, specific
Raman-active phonon modes may beisolated and detected. From the
Raman tensor, onecan obtain the angular dependence of the Raman
scat-tering intensity for each specific combination of inci-dent
and scattered polarizations. A comparison of thecorresponding
experimental and theoretical emissionintensities will allow one to
deduce the orientation ofthe crystallographic axes. Thus, in
combination with ana priori knowledge of the space group of the
crystaland the Raman modes of the material, one can deter-mine the
crystallographic orientation of a nanostruc-ture (Berweger et al.,
unpublished manuscript).
Although the study of nanocrystalline samples opensup a wide
range of potential applications for TERS,some fundamental aspects
are not yet fully understood.Recent far-field studies of wurtzite
CdS nanorods in-dicate a possible depolarization effect in
dielectric
-
Tip-Enhanced Raman Imaging and Nanospectroscopy 191
nanostructures, leading to a breaking of the Ramantensor
selection rules [189]. Furthermore, it has beenshown that the
presence of a sharp edge within thenear-field of a photoemitter can
affect the polarizationof the emitted light [205], although the
resulting effecton Raman scattering is yet unclear. In addition,
thelarge field gradient near the tip can fundamentally al-ter the
selection rules, making previously silent modesvisible [171].
Although this may render IR and othermodes Raman-active [206],
making mode assignmentmore difficult, it would still shed further
insight into thefundamental material properties.
Outlook
TERS may emerge as an important analytical tool forchemical and
structural identification on the nanoscale.It offers chemical
specificity, nanometer spatial reso-lution, single-molecule
sensitivity, and symmetry se-lectivity. However, with both
sensitivity and spatialresolution critically dependent on the well
definedgeometry and related optical properties of the tip,
re-producibility has remained an issue in TERS.
Reproducibility can be enhanced performing TERSunder controlled
experimental environmental condi-tions. Performing experiments
under, e.g., ultrahighvacuum (UHV) conditions offers variable
sample tem-perature and combination with other UHV techniquesfor
surface analysis [82, 153, 187]. Furthermore, fordirect far-field
illumination conditions, it is difficult toa priori distinguish the
near-field response from theunspecific far-field imaging artifacts
(vide supra).
Future developments in tip design and fabricationmay prove
critical. We have recently demonstrated anovel way to generate a
nanoconfined light emitter ona nanoscopic probe tip obtained by
grating-coupling ofSPPs on the tip-shaft [207]. The adiabatic field
con-centration of the propagating SPP, determined by theboundary
conditions imposed by the tapered shape ofthe tip, offers an
intrinsic nanofocusing effect and, thus,gives rise to confined
light emission only from the apexregion, as theoretically predicted
[208].
In this experiment, linear gratings are written ontothe shaft of
Au tips by focused ion beam milling,∼ 10 μm away from the apex, as
schematically shownin Fig. 14a. Upon grating illumination with a
broadbandlight source (150 nm spectral bandwidth of
Ti–sapphireoscillator), SPPs are excited and launched towards
theapex [155, 209], where they are reradiated, as shown inFig. 14b
(details discussed in [207]).
a)
b)
Fig. 14 a Principle of the nonlocal excitation of the tip
apex.Far-field radiation excites SPP on the grating, which
propagatealong the shaft towards the tip apex, where they are
reradiatedinto the far-field. b Microscope image recorded for
illuminationof the tip-grid demonstrating the efficient nonlocal
excitation ofthe tip apex via illumination of the grating
[207].
Spatially separating the excitation from the apexitself, this
approach is particularly promising as itavoids the otherwise
omnipresent far-field backgroundpresent for direct apex
illumination. In addition, itwould provide the spatial resolution
needed for near-field optical techniques including s-SNOM and
TERS.
Summary
A systematic understanding both experimentally andtheoretically
of the fundamental processes responsi-ble for field enhancement,
spectral tip plasmonic re-sponse, and tip–sample coupling has
allowed for reach-ing Raman enhancement factors as high as 109
leadingto single-molecule sensitivity. With lateral
resolutiondetermined by the interplay of the tip apex radius,the
tip–sample distance and the dielectric propertiesof tip and sample
materials, nanometer spatial reso-lution can be obtained. Criteria
have been discussedfor experimental TERS implementation and
distinc-tion of the near-field signature from far-field
imagingartifacts. The combination of the inherent sensitivityand
spatial resolution of TERS with the Raman selec-tion rules and the
unique symmetry of the scanningtip makes possible the spatially
resolved vibrationalmapping on the nanoscale. Its implementation
for
-
192 Neacsu et al.
determining the orientation and domains in crystal-lographic
nanostructures has been proposed. Futuredevelopments in tip design
and more efficient illumi-nation and detection geometries and
scanning probeimplementation will allow for TERS to become a
pow-erful nano-spectroscopic analysis tool.
Acknowledgements The authors would like to thank NicolasBehr,
Jens Dreyer, Thomas Elsaesser, Christoph Lienau, andClaus Ropers
for valuable discussions and support. Funding bythe Deutsche
Forschungsgemeinschaft through SFB 658 (“Ele-mentary Processes in
Molecular Switches at Surface”) the Na-tional Science Foundation
(NSF CAREER grant CHE 0748226and IGERT Fellowship) is greatly
acknowledged.
References
1. Chalmers J, Griffiths P, editors. Handbook of
vibrationalspectroscopy. Chichester: Wiley; 2002.
2. Kuzmany H. Solid-state spectroscopy. New York:
Springer;1998.
3. Wilson EB, Decius JC, Cross PC. Molecular vibrations:the
theory of infrared and Raman vibrational spectra.New York: Dover;
1980.
4. Hendra PJ, Stratton PM. Laser-Raman spectroscopy. ChemRev.
1968;69:325.
5. Hendra PJ, Vear CJ. Laser Raman spectroscopy: a review.The
Analyst. 1970;95:321.
6. Yu PP, Cardona M. Fundamentals of semiconductors. 3rded. New
York: Springer; 2005.
7. Aroca R. Surface-enhanced vibrational spectroscopy.New York:
Wiley; 2006.
8. Demtröder W. L