nanomaterials

Review
Polarization- and Angular-Resolved Optical Response
of Molecules on Anisotropic Plasmonic Nanostructures
Martin Šubr * and Marek Procházka * ID

Faculty of Mathematics and Physics, Institute of Physics, Charles University, 121 16 Prague 2, Czech Republic
* Correspondence: subr.Martin@seznam.cz (M.S.); prochaz@karlov.mff.cuni.cz (M.P.);
Tel.: +420-221-911-463 (M.S.); +420-221-911-474 (M.P.)




Received: 10 May 2018; Accepted: 7 June 2018; Published: 9 June 2018


Abstract: A sometimes overlooked degree of freedom in the design of many spectroscopic (mainly
Raman) experiments involve the choice of experimental geometry and polarization arrangement
used. Although these aspects usually play a rather minor role, their neglect may result in a
misinterpretation of the experimental results. It is well known that polarization- and/or angular-
resolved spectroscopic experiments allow one to classify the symmetry of the vibrations involved or
the molecular orientation with respect to a smooth surface. However, very low detection limits in
surface-enhancing spectroscopic techniques are often accompanied by a complete or partial loss of
this detailed information. In this review, we will try to elucidate the extent to which this approach
can be generalized for molecules adsorbed on plasmonic nanostructures. We will provide a detailed
summary of the state-of-the-art experimental findings for a range of plasmonic platforms used in the
last ~15 years. Possible implications on the design of plasmon-based molecular sensors for maximum
signal enhancement will also be discussed.

Keywords: SERS; Raman; plasmonics; metallic nanostructures; polarization and angular dependences;
molecular orientation; ellipsometry

1. Introduction
Methods of optical spectroscopy, such as absorption spectroscopy, Raman scattering spectroscopy
or fluorescence spectroscopy, can be performed in a wide range of experimental setups relying
on a proper choice of instrumentation, light sources, waveguides, monochromators, detectors etc.
Interplay between all the factors mentioned above determines the final spectral pattern observed in
spectroscopic experiments. An obvious degree of freedom in the experimental design is the polarization
and (especially in the case of Raman experiments) angular arrangement used. Although for many
types of experiments this additional information is not required, variation in light polarization and/or
the scattering angle give rise to a family of phenomena that both theoreticians and experimentalists
should be aware of when interpreting their results. For example, confusion of laser polarization may
lead to very strong Raman bands to appear as rather weak in the Raman spectra, and vice versa. Thus,
understanding these effects helps to obtain deeper insight into the properties of the system studied.
In this review, the main emphasis is placed on polarization and directional/angular characteristics
occurring in Raman spectroscopy and surface-enhanced Raman spectroscopy (SERS). However,
it briefly touches on the polarization and angular dependences of other optical phenomena such as
absorbance, reflectance or fluorescence. It is well known that polarization- and/or angular-resolved
Raman experiments allow one to classify the symmetry of the vibrations involved [1–3] or molecular
orientation with respect to a smooth surface [4–6]. We will try to extend some of these results to
molecules adsorbed on plasmonic nanostructures. Generalizing these results to the case of SERS is not

Nanomaterials 2018, 8, 418; doi:10.3390/nano8060418 www.mdpi.com/journal/nanomaterials

Nanomaterials 2018, 8, 418 2 of 38

straightforward due to the existence of one more element—the nanostructure, which imposes its own
anisotropic pattern in the polarization and directional properties of the SERS signal.
Although some reviews focusing on anisotropic metallic nanostructures for SERS and their
fabrication already exist in literature [7–10], our review covers the topic from a rather untraditional
point of view, focusing primarily on their polarization and directional properties. The review is
organized as follows. Firstly, basic overview of polarization and angular dependences in classical
Raman spectroscopy will be introduced. Different experimental geometries for Raman scattering
measurements will be presented and consequences for the Raman signal coming from randomly
oriented molecules as well as crystals will be derived. We will summarize basic results of the group
theory, allowing classification of normal vibrational modes depending on the symmetry of the vibration
involved. Theoretical predictions will be briefly compared to the experimental results measured in
our laboratory. It is important to note that we will focus only on the phenomena within the dipole
approximation, leaving aside the vast body of experimental techniques such as dichroic methods or
Raman optical activity [11]. Secondly, we will concentrate on the polarization effects for molecules
adsorbed on smooth surfaces. We will show that interference between the incident and the reflected
radiation must be taken into account in order to correctly describe the observed spectral pattern.
The crucial part will involve the derivation of the surface-selection rules, allowing determination of
molecular orientation on smooth surfaces. Although this approach may seem to go far beyond the
original scope of this review, we will show that a very similar principle also applies on roughened
metallic surfaces and contributes to the SERS enhancement. Finally, we will try to generalize previous
considerations to the case of surface-enhanced Raman scattering. We will summarize the up-to-date
findings regarding polarization- and angular-resolved SERS experiments and survey the contributions
to this field from the last 15 years. The influence of both different coupling efficiency of different laser
polarizations to plasmons as well as symmetry of the Raman tensor of the analyte on polarized SERS
will be critically discussed. Described phenomena will be supported by our own experimental results
as well as by published ones.

2. The Origin of Polarization Characteristics in Spectroscopy

Before going deeply into the definitions of the depolarization ratios or formulations of the selection
rules, we devote a short paragraph to the theoretical description of the polarization of the light waves.
A monochromatic plane wave travelling along the z axis in a given Cartesian coordinate system can be
characterized by its electric field vector in any instant restricted to the xy plane. Thus, the variables in
an equation describing a plane wave can be decoupled in the form [12]:
→ → →
E ( x, y, z, t) = E0x cos (ωt − kz) e x + E0y cos (ωt − kz + δ) e y , (1)

where E0x and E0y denote amplitudes of the electric field intensities in the x and y direction, respectively
(traditionally, the polarization direction is assumed to be determined by the electric field vector instead
of the magnetic field vector), k denotes the wavevector and ω the angular frequency. The way that
the electric field vector oscillates in space and time defines the light polarization. In the simplest
case, these oscillations are confined to a single spatial direction, which is termed linear polarization.
In this case, there does not exist any phase shift between the x and y constituents of the electric field
vector. If there does exist a phase shift δ between these constituents (δ 6= kπ, k ∈ Z ), the electric field
vector experiences a more complex motion in the xy plane, which gives rise to circular polarization
(E0x = E0y , δ = π2 + kπ, k ∈ Z) or elliptical polarization (otherwise). Vice versa, any polarization state
can be built up from two mutually perpendicular linearly polarized components.
Further, it is instructive to delve somewhat into a brief theoretical description of the interaction
between (polarized) light and matter. Absorption is an optical process whereby energy of an
electromagnetic wave (a photon) is absorbed in a medium, accompanied by a transition of the
atom/molecule in a higher energetic state. Depending on the photon energy and quantum states
Nanomaterials 2018, 8, 418 3 of 38

participating in the process, one can distinguish the infrared (IR) absorption, which is connected with
a molecular transfer to a higher vibrational level within a given electronic state, and the UV/Vis
absorption, which involves an electron excitation to a higher electronic state. In the context of this
review, absorption will be mainly understood as the UV/Vis absorption. In order to restore the
thermodynamic equilibrium, a molecule in an elevated electronic state can lose its energy and return
to the ground state, which can be basically performed via one of the two processes: (i) The energy
excess may dissipate in the form of heat (non-radiative crossing); or (ii) by a spontaneous emission
of radiation, which is termed luminescence (fluorescence). Raman scattering belongs to the group
of light scattering processes including annihilation of a primary photon and creation of a secondary
photon with the frequency shifted from the primary photon by frequency of a molecular vibration.
Thus, Raman scattering is a two-photon, inelastic process, which does not require the participating
photons match the resonance frequency. For this reason, Raman scattering is a very weak process, but
can be greatly amplified for molecules adsorbed in the vicinity of nanostructured metallic surfaces.
This technique is called surface-enhanced Raman scattering (SERS) and it provides enhancement of the
Raman scattering by a factor of 104 or higher due to resonance excitation of plasmons localized in the
metallic nanostructures [13]. Due to its high sensitivity, SERS has found use in numerous bio-related
applications [14,15], e.g., for the detection of pharmaceuticals and drugs [16–18], food additives [19,20],
explosives [21,22], disease biomarkers [23–25] and many others.
Although “classical” models of the interaction between light and matter appeared already in the
1900s [12] (followed by a semi-classical theory of Raman scattering in 1930s by Placzek [26]), interaction
of light with atoms or molecules cannot be fully described unless quantum theory comes into play.
In the lowest degree of approximation, transition probability (to which intensity of a given spectral
line is proportional) between two quantum states of an atom/molecule is governed by the Fermi’s
golden rule. This rule says that in order for the transition to be allowed, the matrix element

h2|Ŵ |1i (2)

must be different from zero. In Equation (2), |1i represents the wavefunction of the initial state, h2| the
wavefunction of the final state and Ŵ is a perturbation operator, which usually corresponds to the
presence of the electromagnetic wave.
In the case of absorption spectroscopy, the perturbation operator can be rewritten using the
→
electric dipole moment of the molecule ( d ) and the probability of the transition then derives from the
transition dipole moment matrix element:
h2|dˆ|1i. (3)

In the case of Raman spectroscopy, the situation is more complex because in order to obtain
the non-zero contribution, second-order perturbation theory has to be employed. The transition
probability then derives from the second-rank tensor [27,28]:
" #
h2|dˆp |sihs|dˆq |1i h2|dˆq |sihs|dˆp |1i
α12
pq = ∑ E1 − Es + }ωi
+
E2 − Es − }ωi
, (4)
s6=1,2

where E1 and E2 denote the energies of the initial (|1i) and final (|2i) quantum states of the molecule,
→ˆ
d is the electric dipole moment operator and the s-indices refer to all remaining quantum states
of the molecule. } denotes the reduced Planck’s constant and ωi is the angular frequency of the
incident radiation.
Equations (3) and (4) demonstrate basic differences between absorption and Raman measurements.
While the molecular dipole moment is a vector quantity, the transition probability is expected to scale
with the projection of the light polarization in the direction of the transition dipole moment of the
molecule. However, Raman scattering is a two-photon process, which means that both the polarization
of the incident beam as well as that of the scattered beam dictate the total Raman intensity.
Nanomaterials 2018, 8, 418 4 of 38

In the case of absorption spectroscopy, polarization effects in an ensemble of many randomly

oriented molecules, such as liquid samples, are averaged out. That is because the total absorption
intensity is given by spatial averaging over positions of all absorbing molecules. This can be computed
explicitly in azimuthal coordinates:

x = sin ϑ cos ϕ, y = sin ϑ sin ϕ, z = cos ϑ (5)

as:
cos2 ϑdΩ
R
→ → 2
2 2 2
h( E · d ) i = ( Ed) hcos ϑi = ( Ed) R (6)
dΩ
where dΩ = sin ϑdϑdϕ, ϑ ∈ (0, π ) and ϕ ∈ (0, 2π ) are standard azimuthal coordinates.
After standard integration, we obtain:
1
hcos2 ϑi = , (7)
3
which implies that for many randomly oriented molecules, the absorbance does not depend on the
direction of polarization. Since any polarization state can be built up from two mutually perpendicular
linearly polarized components, absorbance of randomly oriented molecules never depends on the state
of polarization including unpolarized (natural) radiation. This is not the case for molecules exhibiting
some preferential orientation (linear dichroism) or phenomena beyond the dipole approximation in
the case of chiral molecules (circular dichroism). Both of these effects are, however, beyond the scope
of this paper.
In the case of the Raman scattering, intensity of a given spectral line is given by a combination of
→i →sc
the molecular Raman tensor and unit vectors of both the incident field e and the scattered field e :
2
I ∼ (esc 12 i
p α pq eq ) . (8)

→i →sc
Directions e and e are determined by the illumination-observation geometry, including
the scattering angle and polarization of the incident beam (and possibly also polarization of the
collected beam). Unlike in absorption spectroscopy, Raman measurements allow much wider choice
of the experimental geometry including the polarization arrangement used as well as the scattering
angle. Equation (8) naturally explains why the molecular quantity describing the disposition to
Raman transition must intrinsically be of tensorial nature and also predicts that detected Raman
intensities depend on the polarization arrangement used even in the case of liquids, i.e., randomly
oriented molecules.
In both absorption and Raman measurements, molecular symmetry imposes certain restrictions
on the number of vibrational modes being active or non-active in respective spectroscopic techniques.
According to the number of symmetry elements characterizing a given isolated molecule, molecules
can be classified into one out of 32 molecular point groups [29]. Having assigned a molecule to
a point group, the methods of group theory allow one to decompose the vibrational signature of
the molecule into so-called irreducible representations. The number of normal modes belonging
to each irreducible representation (out of the 3N − 6 normal modes in total) is calculated using a
well-established algorithm, making use of character tables of molecular point groups [29,30]. For the
purpose of spectroscopy, it is important to note that each irreducible representation may transform
as a particular linear or quadratic function of coordinates. Since the constituents of the electric
dipole moment have the same transformation properties as x, y and z and elements of a Raman
tensor as x2 , y2 , z2 , xy, xz and yz, it is relatively easy to predict which vibrations will be active in
Raman/IR absorption spectra: In order for a given vibrational mode to be allowed in IR absorption
spectra, its irreducible representation must span at least one of x, y and z species (in the case of
randomly-oriented molecules). For the transition to be allowed in Raman spectra, its irreducible
Nanomaterials 2018, 8, 418 5 of 38

representation must span at least one of the quadratic (x2 , y2 , z2 , xy, xz and yz) species (again in the
case of randomly-oriented molecules).

3. Polarization and 8,Angular

Nanomaterials 2018, Dependences
x FOR PEER REVIEW in Raman Spectroscopy 5 of 37

3. Polarization and Angular Dependences in Raman Spectroscopy

3.1. Derivation of Raman Intensities as a Function of the Polarization and Angular Arrangement
3.1. us
Let Derivation
consider of a
Raman
general Intensities as a Function
experimental of the Polarization
geometry for Ramanand Angular measurements,
scattering Arrangement depicted in
Figure 1. LetLet us consider a general experimental geometry for Raman scattering measurements, depictedin its
us suppose that a Cartesian coordinate system is chosen so that the sample is placed
origin,
inthe excitation
Figure 1. Let usbeam
suppose travels
that along the xcoordinate
a Cartesian direction withsystem polarization
is chosen soalong z direction
thesample
that the and the
is placed
detector
in itsisorigin,
oriented to collectbeam
the excitation lighttravels
scatteredalong the𝑥 xy
inthe plane with
direction at anpolarization
angle ϑ with along the 𝑧 direction
respect to the x axis.
and the
The angle detector
ϑ is termed is the
oriented to collect
scattering lightand
angle scattered
it canin take 𝑥𝑦 plane
the any valueat from 0◦ to
an angle 180◦respect
𝜗 with . The xy to plane
the is
𝑥 axis. The angle 𝜗 is termed the scattering angle and it can take any value
termed the scattering plane. In the following, we will refer to the z components of the electric field asfrom 0° to 180°. The 𝑥𝑦
to theplane
normal is termed the scattering
or vertical plane. In
(with respect to the
thefollowing,
scatteringwe will refer
plane, to theas
denoted 𝑧 components
v or ⊥) andofthe thecomponents
electric
field as to the normal or vertical (with respect to the scattering plane, denoted as 𝑣 or ⏊) and the
lying in the scattering plane as to the horizontal (denoted as h or ||). In general, the scattered light
components lying in the scattering plane as to the horizontal (denoted as ℎ or ||). In general, the
contains both vertical and horizontal components—that is why an analyser is often inserted between
scattered light contains both vertical and horizontal components—that is why an analyser is often
the monochromator
inserted between the andmonochromator
the sample toand distinguish
the samplethese two polarizations.
to distinguish In order to
these two polarizations. In overcome
order
different grating different
to overcome responses for different
grating responses light polarizations,
for different a scramblera must
light polarizations, be inserted
scrambler must be between
inserted the
analyser and the
between monochromator
the analyser (not shown in(not
and the monochromator Figure
shown 1 for simplicity).
in Figure 1 for simplicity).

Figure 1. Scheme of a general Raman scattering geometry. The excitation beam travels along the 𝑥
Figure 1. Scheme of a general Raman scattering geometry. The excitation beam travels along the x
direction with polarization in the 𝑧 direction and the detector collects light scattered in the 𝑥𝑦 plane
direction
at anwith
anglepolarization
𝜗 with respectin to
the z 𝑥direction
the axis. An and the detector
analyser collectsbetween
may be inserted light scattered andxy
in the
the sample theplane
at an monochromator
angle ϑ with respect to the x axis. An analyser may be inserted between
to allow only one specific Raman polarization enter the detector. the sample and the
monochromator to allow only one specific Raman polarization enter the detector.
With the fixed angle 𝜗, 4 different polarization combinations can be thought of to retrieve Raman
spectra:
With theThere
fixed are twoϑ, basic
angle possibilities
4 different of setting
polarization the polarization
combinations can be of thought
the incident
of tobeam (either
retrieve Raman
perpendicular to the scattering plane as depicted in Figure 1, i.e., in the 𝑧 direction,
spectra: There are two basic possibilities of setting the polarization of the incident beam (either or lying in the
scattering plane, i.e., in the 𝑦 direction) as well as two basic possibilities of rotating the analyser to
perpendicular to the scattering plane as depicted in Figure 1, i.e., in the z direction, or lying in the
pick out either horizontal or vertical polarization. Thus, we adopt the symbols:
scattering plane, i.e., in the y direction) as well as two basic possibilities of rotating the analyser to pick
𝐼(𝜗, ⊥ 𝑠𝑐 ,or
out either horizontal ⊥𝑖 )vertical 𝐼(𝜗, 𝑠𝑐 , ⊥𝑖 ) =Thus,
= 𝐼𝑣𝑣 , polarization. 𝐼𝑣ℎ , we𝐼(𝜗,
adopt
𝑖
⊥ 𝑠𝑐 , the 𝐼(𝜗, 𝑠𝑐 , 𝑖 ) = 𝐼ℎℎ
) =symbols:
𝐼ℎ𝑣 , (9)
to account for each
i of these polarization
I (ϑ, ⊥sc , ⊥ ) = Ivv , I (ϑ, ksc , ⊥i )arrangements.
= Ivh , I (ϑ,The ⊥scfirst
, ki )symbol
= Ihv , in each
I (ϑ, bracket
ksc , ki ) denotes
= Ihh the (9)
scattering angle, the second symbol refers to polarization of the scattered beam (with respect to the
scattering plane) and the third one refers to polarization of the incident beam. Sometimes the
to account for each of these polarization arrangements. The first symbol in each bracket denotes the
scattering angle is unimportant and as a shorthand we introduce the notation 𝐼𝑣𝑣 , 𝐼𝑣ℎ , 𝐼ℎ𝑣 and 𝐼ℎℎ with
scattering angle, the second symbol refers to polarization of the scattered beam (with respect to the
the first subscript standing for polarization of the incident beam and the latter standing for
scattering plane) and
polarization of thethe third one
scattered refers
beam. Ourtotask
polarization
is to deriveofexpressions
the incident forbeam. Sometimes
intensities of a given theRaman
scattering
angleline,
is unimportant and as a shorthand we introduce the notation
measured under respective arrangements. It is important to note that for Ivv , I , I
vh 𝜗 hv I
= 0° or 𝜗hh= 180° the
and with the first
subscript
scattering plane is not well-defined. Nevertheless, we will show that the general considerations above of
standing for polarization of the incident beam and the latter standing for polarization
the scattered beam.
will become Our
trivial in task
theseisspecial
to derive
cases.expressions for intensities of a given Raman line, measured
Nanomaterials 2018, 8, 418 6 of 38

under respective arrangements. It is important to note that for ϑ = 0◦ or ϑ = 180◦ the scattering plane
is not well-defined. Nevertheless, we will show that the general considerations above will become
trivial in these special cases.
Equation (8) says that each of the 4 possible polarization arrangements mentioned above
picks out different elements of the Raman tensor of the molecules. In other words, specific
illumination-observation geometry determines which elements of the Raman tensor will appear
in the final expressions dictating the intensity of a given Raman line. The Cartesian components of the
electric field vector of the scattered beam are:
Exsc = E(ϑ, ksc ) sin ϑ, Eysc = E(ϑ, ksc ) cos ϑ, Ezsc = E(ϑ, ⊥sc ), (10)

where E(ϑ, ksc ) and E(ϑ, ⊥sc ) denote components of the amplitude of the scattered beam lying in the
scattering plane and in the direction perpendicular to the scattering plane, respectively. By combination
of Equations (8)–(10), the following explicit formulas can be obtained (for simplicity, we further omit
the upper indices 12 in denomination of the molecular Raman tensor α):
I (ϑ, ⊥sc , ⊥i ) ∼ α2zz , I (ϑ, ksc , ⊥i ) ∼ (α xz sin ϑ + αyz cos ϑ )2 , (11)

I (ϑ, ⊥sc , ki ) ∼ α2zy , I (ϑ, ksc , ki ) ∼ (α xy sin ϑ + αyy cos ϑ )2 . (12)

In the case of an ensemble of randomly oriented molecules, respective squares or Raman

tensor elements in Equations (11) and (12) are to be replaced with their spatially-averaged values as
contributions from each individual molecule sum up incoherently. This can be performed in a manner
analogous to Equation (6), however, molecular polarizability responsible for the Raman scattering is
a tensor rather than a vector quantity, which makes the determination of spatially-averaged tensor
elements a more tedious procedure. Any spatial rotation of the molecule with respect to a given
reference frame can be expressed unambiguously by a rotational matrix associated with 3 Euler angles,
whose elements must be averaged in a manner analogous to Equation (6) [31]. An alternative and
less time-consuming approach involves computation of isotropic averages of respective products
of direction cosines between two Cartesian axis systems (again only 3 out of 9 direction cosines are
independent) [27,32]. For an ensemble of many randomly oriented molecules, intensity of a given
Raman line in a given experimental geometry may be expressed as a linear combination of Raman
tensor invariants. These are usually referred to as:
The square of the mean polarizability a:

(α xx + αyy + αzz )2
a2 = , (13)
9
the anisotropy γ:

(α xx − αyy )2 + (αyy − αzz )2 + (αzz − α xx )2 (α xy + αyx )2 + (αyz + αzy )2 + (αzx + α xz )2

γ2 = +3 , (14)
2 4
and the antisymmetric anisotropy δ:

(α xy − αyx )2 + (αyz − αzy )2 + (αzx − α xz )2

δ2 = 3 . (15)
4
For non-resonance Raman scattering only the isotropic invariant a and the anisotropic invariant
γ are relevant since in this case the Raman tensor is symmetric and the antisymmetric invariant δ is
identically zero. The result of the averaging procedure is [27,32]:

45a2 + 4γ2 3γ2 + 5δ2

hα2xx i = hα2yy i = hα2zz i = , hα2xy i = hα2xz i = hα2yz i = , (16)
45 45
hα xz αyz i = hα xy αyy i = 0. (17)
Nanomaterials 2018, 8, 418 7 of 38

This result shows the intuitive notion that the average values of all diagonal elements of the
Raman tensor are equal and also all off-diagonal elements are equal, irrespective of a specific choice of
the system of coordinates. Also all Raman tensor products which involve one common subscript are
zero. Putting these values in Equations (11) and (12) yields:

45a2 + 4γ2 3γ2 + 5δ2

I (ϑ, ⊥sc , ⊥i ) ∼ , I (ϑ, ksc , ⊥i ) ∼ , (18)
45 45

3γ2 + 5δ2 45a2 + 4γ2 3γ2 + 5δ2

I (ϑ, ⊥sc , ki ) ∼ , (ϑ, ksc , ki ) ∼ cos2 ϑ + sin2 ϑ. (19)
45 45 45
Equations (18) and (19) reveal that actually the only polarization arrangement in which the
intensity of a given line will depend on the scattering angle ϑ is the arrangement in which both the
incident beam as well as the scattered beam are polarized in the scattering plane.
In order to avoid misunderstanding, it is important to realize that any combination of Raman
tensor invariants (13)–(15) is again an invariant. In literature, many alternative definitions of Raman
tensor invariants may be found [27]. For example, Raman tensor invariants used by Placzek [26] are:
2 2 2 2
G0 = 3a2 , Ga = δ , Gs = γ . (20)
3 3
A quantity which can be very easily measured experimentally is the Raman depolarization
ratio, which is defined as the ratio of Raman intensities measured under two selected configurations.
Most commonly, the depolarization ratio ρ is assumed as the ratio between the intensity of the Raman
field polarized orthogonal to the incident field and the intensity of the Raman field polarized parallel
to the laser field [27,28]. The most common way of retrieving the depolarization ratio is fixing the
incident polarization and rotating the analyser. In our notation, this yields to:

I (ϑ, ksc , ⊥i ) I
ρ(ϑ, ⊥i ) = sc i
= vh . (21)
I (ϑ, ⊥ , ⊥ ) Ivv

Although many textbooks resort only to the definition of the depolarization ratio given by
Equation (21) [28,30], this formula can be easily generalized and many alternative definitions of the
depolarization ratio can be found in the literature [27], such as:

I (ϑ, ⊥sc , ki ) I I (ϑ, ⊥sc , ki ) I

ρ(ϑ, ki ) = sc i
= hv , ρ(ϑ, ⊥sc ) = sc i
= hv , (22)
I (ϑ, k , k ) Ihh I (ϑ, ⊥ , ⊥ ) Ivv

I (ϑ, ⊥sc , ⊥i ) Ivv

ρ(ϑ, ⊥sc , ⊥i , ksc , ki ) = = , (23)
I (ϑ, ksc , ki ) Ihh
etc. The depolarization ratio provides unique information on the Raman polarizability tensor of a
given molecular vibration and is therefore a very useful tool. Indeed, in the case of randomly oriented
molecules, we have:
I 3γ2 + 5δ2
ρ(ϑ, ⊥i ) = vh = . (24)
Ivv 45a2 + 4γ2
Similar formulas can be derived for the depolarization ratios given by Equations (22) and (23).
It is obvious that the introduction of the depolarization ratios is very convenient because it allows to
exclude the insignificant multiplicative constant relating Raman intensities to specific combinations
of squares of Raman tensor elements. The depolarization ratio defined by Equation (24) is usually
termed the molecular depolarization ratio since it is an intrinsic property of the Raman probe (contrary
to the depolarization ratio measured in presence of smooth/nanostructured metallic surfaces where
the mutual interaction between light/analyte/metal substantially affect the measured depolarization
ratio). From Equation (24) it immediately follows that for non-resonance Raman scattering (δ = 0)
the molecular depolarization ratio is always bound between 0 and 3/4. A specific value of ρ will
Nanomaterials 2018, 8, 418 8 of 38

depend on the symmetry of the vibration involved. For totally symmetric vibrations γ2 = 0 and the
depolarization ratio tends to zero. Conventionally, such a Raman line is said to be completely polarized.
This aspect can be explained straightforwardly from a classical point of view: For totally symmetric
vibrations, the polarizability tensor will become a scalar and the direction of the oscillating dipole
will match the direction of the incident field polarization. Since total Raman intensity is proportional
to the scalar product of the oscillating dipole moment and the detected field, one would expect that
when setting the analyser parallel to the direction of the oscillating dipole (the incident polarization),
the intensity will reach its maximum, while after rotating the analyser about 90◦ the intensity will drop
to its minimum.
Nanomaterials 2018, On 8,the other
x FOR PEER hand,
REVIEW for non-totally symmetric vibrations a2 = 0 and the depolarization 8 of 37
ratio is 3/4 [27,28]. Conventionally, such a Raman line is said to be depolarized. When 0 < ρ < 3/4,
90° the intensity will drop to its minimum. On the other hand, for non-totally symmetric vibrations
the Raman line is said to be (partially) polarized [27].
𝑎2 = 0 and the depolarization ratio is 3/4 [27,28]. Conventionally, such a Raman line is said to be
Previous considerations are valid for any angle ϑ apart from ϑ = 0◦ (forward scattering) and
depolarized. When 0 < 𝜌 < 3/4, the Raman line is said to be (partially) polarized [27].
ϑ = 180◦ (backscattering).
Previous considerations In this areinstance
valid for any theangle
scattering
𝜗 apartplane
from 𝜗is=not well-defined
0° (forward and the
scattering) and symbols
𝜗=
sc i
⊥ , ⊥180° sc i
, k (backscattering).
and k as well asInIvvthis , Ivhinstance
, Ihv and the Ihh lose their sense.
scattering plane Nevertheless,
is not well-defined the situation
and theis symbols
trivial in this
⊥ 𝑠𝑐 , us
case. Let 𝑠𝑐 and 𝑖the
⊥𝑖 , resketch experimental
as well as 𝐼𝑣𝑣 , 𝐼𝑣ℎ , 𝐼geometry in Figure
ℎ𝑣 and 𝐼ℎℎ lose 2, adopting
their sense. the direction
Nevertheless, of theisztrivial
the situation axis parallel
in
to thethis
wavevector
case. Let us of resketch
the incident (scattered) light.
the experimental Afterinilluminating
geometry the sample
Figure 2, adopting with light
the direction 𝑧 axis
with
of the a given
parallel to
polarization, the wavevector
confined to the xyofplane,the incident (scattered)
the analyser maylight. After
be set illuminating
to collect Raman thefield
sample with light
polarized parallel
with a given
or orthogonal to thepolarization,
laser field. confined to the 𝑥𝑦Raman
Alternatively, plane, the analyser may may
measurements be setbetoperformed
collect Raman withfield
a fixed
polarized parallel or orthogonal to the laser field. Alternatively,
position of the analyser and varying polarization of the incident beam. Respective arrangements lead Raman measurements may be
performed with a fixed position of the analyser and varying polarization of the incident beam.
to probing of the averages of the squares of the diagonal/off-diagonal Raman tensor elements. In this
Respective arrangements lead to probing of the averages of the squares of the diagonal/off-diagonal
case, the z components of the Raman tensor do not figure in the equations. Formally, the 4 respective
Raman tensor elements. In this case, the 𝑧 components of the Raman tensor do not figure in the
polarization
equations. combinations
Formally, the discussed
4 respective above split intocombinations
polarization since I (0◦above
two groups discussed i ) = I (0◦ , ksc , ki )
, ⊥sc , ⊥split into two
◦ sc
(0 , k since i
, ⊥ ) 𝐼(0°, ◦
= I (⊥0𝑠𝑐,,⊥ sc i ◦
and Igroups ⊥𝑖 ) ,=k 𝐼(0°,
) in this
𝑠𝑐 , 𝑖case,
) and the
𝐼(0°,same
𝑠𝑐 , ⊥𝑖holding 𝑠𝑐 𝑖for ϑ = 180 . The depolarization
also
) = 𝐼(0°, ⊥ ,  ) in this case, the same holding
ratio for =𝜗0= ◦ or ϑ = 180◦ can be expressed simply as the ratio of intensities obtained with crossed
alsoϑfor 180°. The depolarization ratio for 𝜗 = 0° or 𝜗 = 180° can be expressed simply as the ratio
polarizations
of intensities to intensities obtained
obtained with crossed with parallel polarizations.
polarizations Formally:with parallel polarizations.
to intensities obtained
Formally:
I (0◦ or 180◦ , ksc , ki )
ρ = 𝐼(0°◦or 180°,◦𝑠𝑐 ,sc⊥𝑖 ) i , (25)
𝜌 = I (0 or 180 , k , k, ) (25)
𝐼(0° or 180°, 𝑠𝑐 , 𝑖 )
which takes the form given by Equation (24) in the case of randomly oriented molecules.
which takes the form given by Equation (24) in the case of randomly oriented molecules.

FigureFigure 2. Scheme
2. Scheme of a of a forward-scattering
forward-scattering geometry.
geometry. Linearlypolarized
Linearly polarizedlight
light is
is falling
falling on
onthe
theplane
planeofof the
the sample (𝑥𝑦) from above. Analyser can be set to retrieve either a perpendicular component of the
sample (xy) from above. Analyser can be set to retrieve either a perpendicular component of the laser
laser field (𝐼⊥ ) or a parallel one (𝐼 ). Analogous situation occurs in the case of a backscattering
field (I⊥ ) or a parallel one (Ik ). Analogous situation occurs in the case of a backscattering geometry.
geometry.

Further, let us inspect another widely-used geometry in which 𝜗 = 90° (right-angle geometry).
After inserting 𝜗 = 90° in Equation (19), we obtain:

𝐼(90°, 𝑠𝑐 , ⊥𝑖 ) = 𝐼(90°, ⊥ 𝑠𝑐 , 𝑖 ) = 𝐼(90°, 𝑠𝑐 , 𝑖 ) ≠ 𝐼(90°, ⊥𝑠𝑐 , ⊥𝑖 ), (26)

𝐼𝑣ℎ = 𝐼ℎ𝑣 = 𝐼ℎℎ ≠ 𝐼𝑣𝑣 . (27)

Nanomaterials 2018, 8, 418 9 of 38

Further, let us inspect another widely-used geometry in which ϑ = 90◦ (right-angle geometry).
After inserting ϑ = 90◦ in Equation (19), we obtain:

I (90◦ , ksc , ⊥i ) = I (90◦ , ⊥sc , ki ) = I (90◦ , ksc , ki ) 6= I (90◦ , ⊥sc , ⊥i ), (26)

Ivh = Ihv = Ihh 6= Ivv . (27)

It means that for retrieving information on the Raman depolarization ratio (and consequently on
the symmetry of the vibration involved), Ivv and any of Ivh , Ihv and Ihh are required.
Nanomaterials 2018, 8, x FOR PEER REVIEW 9 of 37

3.2. Simple Illustration of the Depolarization Ratio Measurement

3.2. Simple Illustration of the Depolarization Ratio Measurement
In this paragraph, we demonstrate previous considerations on a simple example of liquid CCl4
In this paragraph, we demonstrate previous considerations on a simple example of liquid CCl 4
measured in our laboratory in the right-angle geometry (excitation wavelength 532 nm). The default
measured in our laboratory in the right-angle geometry (excitation wavelength 532 nm). The default
polarization of the incident beam in this experimental setup is perpendicular to the scattering plane,
polarization of the incident beam in this experimental setup is perpendicular to the scattering plane,
but may be rotated by 90◦ using a half-wave plate or a Fresnel rhomb retarder. The CCl4 is a highly
but may be rotated by 90° using a half-wave plate or a Fresnel rhomb retarder. The CCl 4 is a highly
symmetric molecule belonging to the Td point group. Polarized Raman spectra obtained in all four
symmetric molecule belonging to the 𝑇𝑑 point group. Polarized Raman spectra obtained in all four
polarization arrangements are displayed in Figure 3.
polarization arrangements are displayed in Figure 3.

Figure 3.
3. Polarized
Polarized Raman
Ramanspectra
spectraofofCCl
CCl4. The excitation wavelength was 532 nm, laser power 100
Figure 4 . The excitation wavelength was 532 nm, laser power
mW and the accumulation time 1 min. No further
100 mW and the accumulation time 1 min. No further corrections have
corrections been
have made.
been made.

Analysis of respective depolarization ratios suggests that the band centered around 465 cm −1 can
−1 can
Analysis of respective depolarization ratios suggests that the band centered around 465 cm
be classified as totally symmetric (the “disrupted” structure of the 465-cm−1 band is owed to different
be classified as totally symmetric (the “disrupted” structure of the 465-cm−1 band is owed to different Cl
Cl isotopes which lift the degeneracy of this peak) while the bands around 222 and 319 − 1
cm−1 as non-
isotopes which lift the degeneracy of this peak) while the bands around 222 and 319 cm as non-totally
totally symmetric modes. Indeed, group theory predicts the 3N−6 representation of CCl4 to be A1 ⊕
symmetric modes. Indeed, group theory predicts the Γ3N−6 representation of CCl4 to be A1 ⊕ E ⊕ 2F2
E ⊕ 2F2 (E being doubly degenerate and F being triply degenerate, making 3 ∙ 5 − 6 = 9 vibrational
(E being doubly degenerate and F being triply degenerate, making 3·5 − 6 = 9 vibrational modes
modes in total).
in total).
Basically, those values obtained from the experiment confirm that 𝐼𝑣ℎ , 𝐼ℎ𝑣 and 𝐼ℎℎ configurations
Basically, those values obtained from the experiment confirm that Ivh , Ihv and Ihh configurations
are equivalent. However, closer analysis (inset of Figure 3) reveals slight differences in respective
are equivalent. However, closer analysis (inset of Figure 3) reveals slight differences in respective
polarization arrangements, which is obviously attributable to transfer optics imperfections. The
polarization arrangements, which is obviously attributable to transfer optics imperfections.
depolarization ratios obtained from measurements with varying incident polarization are larger than
The depolarization ratios obtained from measurements with varying incident polarization are larger
from the measurements where the incident polarization was kept constant, which shows some subtle
than from the measurements where the incident polarization was kept constant, which shows some
imperfection of the half-wave plate, transmitting a small fraction of orthogonal polarization as well
(making the relative error particularly large for highly polarized lines where 𝜌 ≈ 0 ). Thus,
polarization-dependent Raman spectra of CCl4 serve well for checking the correct function of the
scrambler and other optical components used for more complex polarization-dependent
measurements.
Last but not least, the literature is not consistent about whether peak heights (above spectral
Nanomaterials 2018, 8, 418 10 of 38

subtle imperfection of the half-wave plate, transmitting a small fraction of orthogonal polarization
as well (making the relative error particularly large for highly polarized lines where ρ ≈ 0). Thus,
polarization-dependent Raman spectra of CCl4 serve well for checking the correct function of the
scrambler and other optical components used for more complex polarization-dependent measurements.
Last but not least, the literature is not consistent about whether peak heights (above spectral
background) or rather integrated areas should actually enter Equations (21)–(23). In many instances
(such as in the case of CCl4 ) these two ways produce almost identical results as the line width is
not expected to change with polarization. On the other hand, in the case of several overlapping
bands, determination of the spectral background and delimiting the peak area may be difficult and the
two methods may give somewhat different results. In these cases, fitting the spectrum with the sum of
the Lorentz curves would be advisable.

3.3. Other Possible Polarization Arrangements Used

Previous survey of the most widely-used geometries for the depolarization ratio measurements
is still not exhaustive. All other possible geometries can be resolved exploiting the fact that any
polarization state can be decomposed into a superposition of two independent, linearly polarized
waves with their electric field vectors perpendicular to each other. Other possible experimental
geometries (maybe even the most frequent) involve the situations when no analyser is used, i.e.,
the excitation field is either vertical or horizontal, but polarization of the detected field is not
characterized (or a far less common case when the excitation field is unpolarized and respective
scattered field components are detected). For the sake of completeness, we adopt the symbols:

I (ϑ, nsc , ⊥i ) = I (ϑ, ⊥sc , ⊥i ) + I (ϑ, ksc , ⊥i ), (28)

I (ϑ, nsc , ki ) = I (ϑ, ⊥sc , ki ) + I (ϑ, ksc , ki ), (29)

and the less commonly used symbols:

I (ϑ, ⊥sc , ni ) = I (ϑ, ⊥sc , ⊥i ) + I (ϑ, ⊥sc , ki ), (30)

I (ϑ, ksc , ni ) = I (ϑ, ksc , ⊥i ) + I (ϑ, ksc , ki ). (31)

Equations (28) and (29) show that investigating a total Raman intensity without an analyser is
equivalent to performing two experiments in a row: setting the analyser to collect the horizontal
polarization, then rotating the analyser by 90◦ to collect the vertical polarization and summing up both
results. An analogous situation is true for Equations (30) and (31). Moreover, respective left-hand sides
of Equations (28)–(31) can be combined and yield further alternative depolarization ratio definitions.
Surprisingly, information on the orientation-averaged squares of the Raman tensor elements (and
consequently information on the symmetry of molecular vibrations involved) can be extracted even
without an analyser—all that one needs to do is to record the Raman spectra with the incident field
polarization perpendicular to the scattering plane and parallel to the scattering plane, respectively.
Let us, for instance, determine the expected depolarization ratio in the right-angle geometry with
varying incident polarization and no analyser used. By combining Equations (18), (19), (28) and (29),
we obtain:
I (90◦ , nsc , ki ) 6γ2 + 10δ2
ρ(90◦ , nsc ) = ◦ = . (32)
sc
I (90 , n , ⊥ ) i 45a2 + 7γ2 + 5δ2
The depolarization ratio for non-resonance Raman scattering (δ2 = 0), defined in the manner of
Equation (32), is always bound between 0 and 6/7.
Nanomaterials 2018, 8, 418 11 of 38

3.4. Directional Properties in Raman Spectroscopy

At the end of this section, implications of previous considerations on the angular dependences
in Raman spectroscopy will be discussed. We have already seen that out of the 4 basic polarization
arrangements, the only one where the Raman intensity actually depends on the angle ϑ is the I (ϑ, ksc , ki )
arrangement (Equation (19)). Thus, let us resketch the Raman measurement scheme as in Figure 4.
Supposing the scattering process takes place in the plane of the sheet, different Raman intensities
are expected to arise when moving the detector about the axis perpendicular to the scattering plane.
This intensity modulation occurs only if both the excitation and the detected field are polarized parallel
to the scattering plane, or when the detected field is unpolarized (or a far less common case when
the excitation field is unpolarized and parallel field component is detected). If at least one of the
excitation/detected field components is perpendicular to the scattering plane, Raman intensity will no
longer be the function of ϑ as follows directly from Equation (18). By contrast, if both the excitation
and detected fields were perpendicular to the plane of the sheet, Raman intensity would be modulated
when moving
Nanomaterials the
2018, 8, xdetector
FOR PEERinREVIEW
the plane perpendicular to the sheet. 11 of 37

Figure 4. Angular dependence of the Raman signal. When both the incident and scattered fields are
Figure 4. Angular dependence of the Raman signal. When both the incident and scattered fields
polarized parallel to the scattering plane or when one field is polarized parallel and the latter is
are polarized parallel to the scattering plane or when one field is polarized parallel and the latter
unpolarized, the detected Raman intensity will be angular-dependent. Angular dependences for
is unpolarized, the detected Raman intensity 2will2 be angular-dependent. Angular dependences for
molecules with a dominant contribution of 𝑎 , 𝛾 and 𝛿 2 respectively when moving the detector by
molecules with a dominant contribution of a2 , γ2 and δ2 respectively when moving the detector by an
an angle 𝜗 from 0° to 360° and no analyser is used are indicated in the inset.
angle ϑ from 0◦ to 360◦ and no analyser is used are indicated in the inset.

Let us explore in more detail the case when the excitation line is polarized parallel to the
Let us explore in more detail the case when the excitation line is polarized parallel to the scattering
scattering plane and polarization of the scattered beam is not characterized, i.e., the analyser is
plane and polarization of the scattered beam is not characterized, i.e., the analyser is missing. Again,
missing. Again, we will suppose that the Raman scatterers are randomly oriented molecules. In our
we will suppose that the Raman scatterers are randomly oriented molecules. In our notation:
notation:

𝑖 𝑖 𝑠𝑐 𝑖 3𝛾22 +
3γ 5𝛿22 45𝑎
+ 5δ 45a22++4𝛾
4γ22
3𝛾 23γ
+ 25𝛿+2 5δ22
I (ϑ, nsc , 𝐼(𝜗, 𝑛 𝑠𝑐I, (ϑ,
ki ) = )= ⊥sc , ki⊥)𝑠𝑐+, I)(ϑ,
𝐼(𝜗, ksc , k i ),  ∼
+ 𝐼(𝜗, )~ +
+ cos 2 𝜗2+
cos ϑ+ sin2 ϑ. (33)
sin 𝜗. (33)
45
45 45
45 45 45
Rearranging Equation
Rearranging Equation (33),
(33), we
we obtain:
obtain:
𝛾2 𝛿2
𝐼(𝜗, 𝑛sc𝑠𝑐 , i𝑖 ) ~ 𝑎2 cos 2
𝜗 + γ 2 + cos 2 𝜗) +
(6 2 + sin2 𝜗).
(1 (34)
I (ϑ, n , k ) ∼ a2 cos2 ϑ + 45 (6 + cos2 ϑ ) + 9 δ (1 + sin2 ϑ ). (34)
45 9
Angular dependence of the three terms on the right-hand side of Equation (34) is displayed in
Angular
the inset dependence
of Figure 4. From of here
the three terms that
it follows on thetheright-hand side of Equation
angular dependence (34) is symmetric
of a totally displayed
in the inset(𝑎of≠Figure
vibration 4.𝛿 From
0, 𝛾 = 0, = 0) is here it follows
dictated by the the2 𝜗angular
thatcos dependence
function, of a as
i.e., the same totally symmetric
encountered in
vibration (a 6 = 0, = 0, = 0) is dictated by the cos 2 ϑ function, i.e., the same as encountered in
investigating the angular dependence of the dipole radiation in classical physics [12]. For a vibration
γ δ
investigating
whose dominant the angular
invariant is 𝛾 2 , the detected
dependence of the dipole radiation
intensity in classical
will only slightlyphysics
depend [12].
on For a vibration
the angle 𝜗 via
2
the term 6 + cos 𝜗. Finally, in the case of resonance Raman scattering, the anisotropic contribution
will scale as 1 + sin2 𝜗. It is the only term which experiences its maximum for 𝜗 = 90° (anomalous
angular dependence).

4. Polarization Effects for Crystals and Molecules Adsorbed on Smooth Surfaces

Nanomaterials 2018, 8, 418 12 of 38

whose dominant invariant is γ2 , the detected intensity will only slightly depend on the angle ϑ via
the term 6 + cos2 ϑ. Finally, in the case of resonance Raman scattering, the anisotropic contribution
will scale as 1 + sin2 ϑ. It is the only term which experiences its maximum for ϑ = 90◦ (anomalous
angular dependence).

4. Polarization Effects for Crystals and Molecules Adsorbed on Smooth Surfaces

We have demonstrated that Raman scattering measurements of systems consisting of freely
rotating molecules allow retrieving only spatially-averaged values of respective squares of Raman
tensor elements. However, more detailed information can be extracted from anisotropic systems such as
single crystals [33] or molecules adsorbed on planar surfaces possessing some sort of preferential
orientation. In single crystals, all molecular axes are lined up within the unit cell in the same direction
→
for each cell. Under these circumstances, polarization of the incident light set to e q and polarization of
→
the scattered light set to e p will pick out only the α pq component of the Raman tensor (see Equation (8)).
Thus, it is possible to analyse each of the Raman tensor elements separately by employing different
polarization arrangements. According to Porto’s notation, each configuration for Raman scattering
measurements of single crystals is unambiguously described by four symbols, such as i ( jk )l (where
i, j, k, l = x, y, z). The symbols inside the parentheses are, left to right, the polarization of the incident
and the scattered light, the leftmost symbol denotes the wavevector of the incident light and the
rightmost symbol the wavevector of the scattered light [33]. However, in the case of anisotropic systems
such as single crystals, the selection rules are a bit stricter in some sense. In the geometry determined
by the Porto’s notation i ( j, k)l, only those vibrations covering the same irreducible representation as
jk (j, k = x, y, z) will be allowed. As a consequence, symmetry species of single crystals of known
orientation may be distinguished with the use of polarized radiation. In this paper, we will not deal
further with crystal samples, but we will demonstrate how exploiting the formalism of the group
theory together with experimental observation can determine the orientation of molecules on (whether
planar or rough) surfaces.
In 1982, Moskovits coined the term “surface selection rules” to account for modification of the
absorption, emission and Raman scattering intensities of molecules in the vicinity of planar metallic
surfaces [34,35], although some considerations concerning variation in near-field intensities with
the angle of incidence/polarization had been made before [4,5]. A year later, Creighton derived
expressions enabling him to determine the orientation of molecules adsorbed at the surface of a small
metallic sphere [36]. Both results show that depending on the molecular orientation with respect
to the surface, differences in the relative enhancement of modes belonging to different irreducible
representations (symmetry classes) may be expected. Let us first consider the situation on a flat
surface and proceed further to molecules adsorbed on plasmonic nanostructures. The latter is a more
complicated case due to the fact that in this instance both the symmetry of the Raman tensor and
coupling efficiency of the laser light to nanostructures influence the final spectral pattern.
What lies behind the surface selection rules for molecules adsorbed on planar (metallic) surfaces
is the fact that the field felt by the analyte may be viewed as a superposition of the incident wave
and the reflected wave. In principle, this interference can be both constructive and destructive and
the total intensity of a given optical process will depend, among other things, on reflectivity of the
substrate (and thus wavelength, of which reflectivity is a function), angle of incidence and polarization.
For molecules adsorbed on planar surfaces, the surface selection rules may be interpreted simply in
the framework of the Fresnel coefficients. Briefly, when light in vacuum (air) is incident on a planar
interface between two different media, the latter possessing a (complex) refraction index n e = n + ik,
it is common procedure to resolve the electric field vectors of both the incident and the reflected
wave into two orthogonal components. Assuming that the interface lies in the xy plane, the plane of
incidence is yz and light of frequency ω falls on the interface at an angle ϑ with respect to the surface
normal, the wavevectors of both the incident and the reflected light may be expressed as:
Nanomaterials 2018, 8, 418 13 of 38

→ ω → ω
ki = (0, sin ϑ, − cos ϑ), kr = (0, sin ϑ, cos ϑ), (35)
c c
and their electric field vectors as:
→ p p
E i = ( Eis , Ei cos ϑ, Ei sin ϑ ), (36)
→ p p p p
E r = ( Ers , − Er cos ϑ, Er sin ϑ ) = (rs Eis , −r p Ei cos ϑ, r p Ei sin ϑ ), (37)

where the amplitude of the electric field vector of the incident beam projected on the plane of incidence
p p
is denoted as Ei and on the direction perpendicular to the plane of incidence as Eis , similarly Er and
Ers are amplitudes of the electric field vector of the reflected beam projected on the plane of incidence
and on the direction perpendicular to the plane of incidence, respectively. rs and r p are the Fresnel
reflection coefficients for s- and p- polarized light. By applying boundary conditions to the solutions of
Maxwell’s equations, it can be shown that [12,13]:
p p
cos ϑ − e2 − sin2 ϑ
n e2 cos ϑ −
n e2 − sin2 ϑ
n
rs = p , rp = p . (38)
cos ϑ + e2 − sin2 ϑ
n e2 cos ϑ +
n e2 − sin2 ϑ
n

In general, both Fresnel reflection coefficients become complex, thus rs ≡ e rs and r p ≡ e rp.
Interference between the incident and the reflected beam at the interface suggests that the electric field
intensity felt by the adsorbed molecules has the components:
Nanomaterials 2018, 8, x FOR PEER REVIEW 13 of 37
p p
Ex = Eis (1 + rs ), Ey = Ei cos ϑ(1 − r p ), Ez = Ei sin ϑ (1 + r p ). (39)
𝑝 𝑝
𝐸𝑥 = 𝐸𝑖𝑠 (1 + 𝑟𝑠 ), 𝐸𝑦 = 𝐸𝑖 cos 𝜗 (1 − 𝑟𝑝 ), 𝐸𝑧 = 𝐸𝑖 sin 𝜗 (1 + 𝑟𝑝 ). (39)
Since the real part of rs is negative and the real part of r p is usually positive, it is clear that only
Sincecomponent
the z (normal) the real partof 𝑟𝑠 iselectric
of the negativefield
and at
thethe
real 𝑟𝑝 is benefit
part ofwill
surface usually from
positive, it is clear that
constructive only
interference,
the 𝑧 (normal) component of the electric field at the surface will benefit from constructive
whereas the other two (tangential) components exhibit rather destructive interference. For most metals,
interference, whereas the other two (tangential) components exhibit rather destructive interference.
rs approaches −1 and r p approaches 1 in the IR and the visible spectral region at angles around ∼60◦ .
For most metals, 𝑟𝑠 approaches −1 and 𝑟𝑝 approaches 1 in the IR and the visible spectral region at
For example, for silver, z direction 2 ) can be as much
angles around ~ 60°.light intensityfor
For example, associated
silver, lightwith the field
intensity in thewith
associated the field (E
in zthe 𝑧 direction
as ∼6(𝐸×𝑧2stronger than theasintensity in other 2
(Exin+other2
Ey ) around
) can be as much ~ 6× stronger thandirections
the intensity the(𝐸wavelength
directions 2 2 of 500 nm
𝑥 + 𝐸𝑦 ) around the
(Figure 5). The difference tends to be even more pronounced in the IR [37].
wavelength of 500 nm (Figure 5). The difference tends to be even more pronounced in the IR [37].

Figure 5. Calculated near-field intensities in the vicinity of a smooth Ag surface for an arbitrarily
Figure 5. Calculated near-field intensities in the vicinity of a smooth Ag surface for an arbitrarily
polarized light wave, 𝜆 = 500 nm. Calculations were performed for different angles of incidence and
polarized light wave, λ = 500 nm. Calculations were performed for different angles of incidence and
the refractive index 𝑛̃𝐴𝑔 = 0.05 + 3.09i [38].
the refractive index n e Ag = 0.05 + 3.09i [38].
Since IR absorption intensity derives from the scalar product of the transition dipole moment
and direction of light polarization, one can deduce that the IR absorption by molecular vibrations
with a nonzero component of the transition dipole moment perpendicular to the surface will be
strengthened, whereas the absorption by vibrations with the only nonzero component parallel to the
surface will be weakened with respect to a “free” molecule [37]. This observation together with the
aid of the group theory becomes useful when aiming to elucidate molecular orientation on the
surface. Let us illustrate this aspect with a simple example. For each individual molecule, a local
Nanomaterials 2018, 8, 418 14 of 38

Since IR absorption intensity derives from the scalar product of the transition dipole moment and
direction of light polarization, one can deduce that the IR absorption by molecular vibrations with a
nonzero component of the transition dipole moment perpendicular to the surface will be strengthened,
whereas the absorption by vibrations with the only nonzero component parallel to the surface will
be weakened with respect to a “free” molecule [37]. This observation together with the aid of the
group theory becomes useful when aiming to elucidate molecular orientation on the surface. Let us
illustrate this aspect with a simple example. For each individual molecule, a local Cartesian system of
coordinates x 0 , y0 z0 may be introduced to describe its orientation in space (with z0 usually being the
principal axis of symmetry). Providing that the molecule binds to the surface so that x = x 0 , y = y0
and z = z0 , the strongest modes observed in the IR spectrum will be those belonging to the irreducible
representation with the same transformation properties as z. By contrast, in case the molecule is
positioned on the surface so that its z0 axis lies parallel to the surface, the modes spanning z coordinate
are expected to be very weak.
In the case of Raman scattering of molecules adsorbed on flat surfaces, a very similar principle
applies in the case of the scattered radiation. Providing that the scattered radiation is collected at an
arbitrary angle ϑ0 as shown in Figure 6, the wavevector of the scattered light will be:
→ ω
k sc = (0, sin ϑ0 , cos ϑ0 ), (40)
c
the directly scattered laser field is:
→ p p
s
E sc = ( Esc , − Esc cos ϑ0 , Esc sin ϑ0 ), (41)

and the contribution to the total detected field experiencing a single reflection from the surface will be:
p p
(rs0 Esc
s
, r 0p Esc cos ϑ0 , r 0p Esc sin ϑ0 ). (42)

Thus, the total detected field will have the components:

p p
s
Ex = Esc (1 + rs0 ), Ey = Esc cos ϑ0 (r 0p − 1), Ez = Esc sin ϑ0 (1 + r 0p ). (43)

p s denote the amplitudes of the electric field vector of the scattered radiation
As usual, Esc and Esc
projected on the plane of incidence and on the direction perpendicular to the plane of incidence,
respectively. The prime in rs0 and r 0p refers to the fact that the scattered radiation is slightly shifted in
frequency and thus also the Fresnel reflection coefficients may be slightly shifted. Combination of
Equations (8), (39) and (43) gives:

2
Ivv ∼ |α xx (1 + rs )(1 + rs0 )| , (44)
2
Ihv ∼ |α xy (1 + rs0 )(1 − r p ) cos ϑ + α xz (1 + rs0 )(1 + r p ) sin ϑ | , (45)
2
Ivh ∼ |αyx (1 + rs )(r 0p − 1) cos ϑ0 + αzx (1 + rs )(1 + r 0p ) sin ϑ0 | , (46)

Ihh ∼ |αyy (1 − r p )(r 0p − 1) cos ϑ cos ϑ0

+αyz (1 + r p )(r 0p − 1) sin ϑ cos ϑ0 + αzy (1 − r p )(1 + r 0p ) sin ϑ0 cos ϑ (47)
+αzz (1 + r p )(1 + r 0p ) sin ϑ sin ϑ0 |2 .
Equations (44)–(47) are known as the surface selection rules [34,35].
 𝐼𝑣ℎ ~ |𝛼𝑦𝑥 (1 + 𝑟𝑠 )(𝑟𝑝′ − 1) cos 𝜗′ + 𝛼𝑧𝑥 (1 + 𝑟𝑠 )(1 + 𝑟𝑝′ ) sin 𝜗′| , (46)

𝐼ℎℎ ~ |𝛼𝑦𝑦 (1 − 𝑟𝑝 )(𝑟𝑝′ − 1) cos 𝜗 cos 𝜗′

+ 𝛼𝑦𝑧 (1 + 𝑟𝑝 )(𝑟𝑝′ − 1) sin 𝜗 cos 𝜗′ +𝛼𝑧𝑦 (1 − 𝑟𝑝 )(1 + 𝑟𝑝′ ) sin 𝜗′ cos 𝜗 (47)
′ 2
+ 𝛼𝑧𝑧 (1 + 𝑟𝑝 )(1 + 𝑟𝑝′ ) sin 𝜗 sin 𝜗 | .
Nanomaterials 2018, 8, 418 15 of 38
Equations (44)(47) are known as the surface selection rules [34,35].

Figure6.6.Scheme
Figure Schemeofofthethegeometrical
geometricallayout
layoutand
anddefinition
definitionof
ofthe
thecoordinates.
coordinates.All
Allsymbols
symbolshave
haveusual
usual
meanings as described in the text. The components of the scattered field experiencing a reflection
meanings as described in the text. The components of the scattered field experiencing a reflection from
⃗ 𝑟𝑠′→and 𝑝′
→s0 (𝐸
from the surface p0 𝐸⃗𝑟 ) are not included in the scheme for simplicity.
the surface ( E r and E r ) are not included in the scheme for simplicity.

5. Polarization Effects for Molecules on Plasmonic Nanostructures

Polarization-dependent effects for molecules located in presence of plasmonic nanostructures
are much more complex in comparison to “free” molecules. The principal reason is that the
enhancement provided by plasmonic nanostructures is polarization-sensitive and is therefore
commonly accompanied by dramatic alteration of polarizations of both the incident beam as well as
of the scattered beam. Local field polarization is dictated mainly by coupling of the given light
polarization in the metallic nanostructures and the similar process applies in the case of the scattered
field. As a result, polarization felt by the molecule can be drastically different from polarization
of the incident beam. It causes the depolarization ratio to no longer depend only on the intrinsic
properties of the Raman tensor of the analyte, but also (and often mainly) on the geometry of the given
nanostructure [39–41]. This difference occurs even in the simplest nanostructure geometries, such as
a small sphere in the dipole approximation, it can be commonly observed in arrays of elongated
nanoparticles (NPs) and is most pronounced for molecules in the location of hot-spots which exhibit
anisotropic behaviour [40,42]. Since different polarizations couple to metallic nanostructures with
different efficiency, the enhancement factor (EF) can no longer be regarded as a scalar but it takes the
form of a tensor. In this case, the polarization and angular dependence of the SERS signal is therefore
mostly dictated by the coupling of the laser to the plasmons while symmetry of the Raman tensor of
the analyte often plays a minor role [39,40,43].

5.1. Polarization Properties of Isolated Spherical Particles

The case of an isolated metallic sphere in the dipole approximation was theoretically studied
in [36,44]. Theoretical analysis reveals that the enhanced local field in the vicinity of a small metallic
sphere after illumination with a light wave may be expressed in the electrostatic approximation in
spherical coordinates as:
→ → →
E = E0 (1 + 2g(ω )) cos ϑ e r − E0 (1 − g(ω )) sin ϑ e ϑ , (48)

i.e., the normal field components with respect to the sphere undergo the enhancement by the factor of
Ar = 1 + 2g(ω ) while the tangential component is enhanced by the factor of Aϑ = 1 − g(ω ), where
ε(ω )−ε
g(ω ) = ε(ω )+2εr is the geometrical factor for a sphere. The relative permittivity of the surrounding
r
Nanomaterials 2018, 8, 418 16 of 38

medium is denoted as ε r . Therefore, amplification of the surface field for any point on the surface of
the sphere is determined by the matrix:
 
↔
1 + 2g(ω ) 0 0
Ar,ϑ,ϕ = 0 1 − g(ω ) 0 . (49)
 
0 0 1 − g(ω )

As a consequence, molecules with certain elements of symmetry and preferential orientation

with respect to the surface will exhibit differences in the relative enhancements of modes belonging
to different symmetry classes. Averaging over all solid angles (all positions on the metallic sphere)
reveals that the ratio of tangential to normal intensities is:

2
cos2 ϑdΩ 1 1 + 2g(ω ) 2
R
Er2 1 + 2g(ω )
= = . (50)
Eϑ2 1 − g(ω ) sin2 ϑdΩ 2 1 − g(ω )
R

In the early 1980s, Moskovits pointed out the existence of three classes of vibrational modes with
distinct spectral behaviour (unique excitation profile) [35,45]: (1) those excited only by the normal
component of the field and resulting in an induced dipole with a strong component only in the
direction perpendicular to the surface; (2) those excited only by the tangential component of the field
and resulting in an induced dipole with a strong component tangential to the surface; and (3) mixed
cases. EF for respective molecular vibrational modes will thus depend on the orientation of the
molecule on the surface. Let us assume that a molecule possessing certain symmetry is adsorbed on
the surface so that its principal axis (z axis) is perpendicular to the sphere. Then, the modes of the first
type are modes spanning the same irreducible representation as αzz . SERS intensity of these modes is
expected to obey the relationship:

Inn ∼ |1 + 2g(ωi )|2 |1 + 2g(ωs )|2 . (51)

By contrast, modes spanning the same irreducible representation as α xx , αyy or α xy benefit only
from the existence of the electric field tangential to the surface. Thus,

Itt ∼ 4|1 − g(ωi )|2 |1 − g(ωs )|2 . (52)

Finally, modes excited by the normal component of the field and resulting in an induced dipole
with a strong component only in a direction parallel to the surface, or vice versa, will be characterized
by the excitation profile:

1
Int,tn ∼ (|1 + 2g(ωi )|2 |1 − g(ωs )|2 + |1 + 2g(ωs )|2 |1 − g(ωi )|2 ). (53)
2
As usual, ωi denotes frequency of the excitation radiation and ωs frequency of the scattered
radiation. It is clear that the SERS excitation profiles associated with the three types of modes obey the
wavelength dependence of the right-hand sides of Equations (51)–(53) (Figure 7). By contrast, when
the molecule adsorbs with the z axis tangential to the surface, the role of z and x (or y) in the character
table is to be swapped. This fact provides a unique tool for determination of molecular orientation on
nanostructured metallic surfaces [45–48].
Due to disrupted spherical symmetry in the case of non-spherical particles, the polarization
insensitive SERS EF is expected to split into two (or more) geometry-dependent components.
For example, in the case of prolate spheroids, only the incident field polarized parallel to the long axis
results in the greatest surface fields [49]. Such electrodynamic calculations were performed already in
the 1980s and their understanding is a key for elucidation of the polarization-resolved optical response
of more complex systems such as elongated particles or nanorods (NRs), which will be discussed in
more detail in the second part of this section.
 As usual, 𝜔𝑖 denotes frequency of the excitation radiation and 𝜔𝑠 frequency of the scattered
radiation. It is clear that the SERS excitation profiles associated with the three types of modes obey
the wavelength dependence of the right-hand sides of Equations (51)(53) (Figure 7). By contrast,
when the molecule adsorbs with the 𝑧 axis tangential to the surface, the role of 𝑧 and 𝑥 (or 𝑦) in the
character 2018,
Nanomaterials table8, is
418to be swapped. This fact provides a unique tool for determination of molecular
17 of 38
orientation on nanostructured metallic surfaces [45–48].

Figure 7. Calculated wavelength-dependent relative enhancement factor for molecules adsorbed in

Figure 7. Calculated wavelength-dependent relative enhancement factor for molecules adsorbed in the
the vicinity of a small silver sphere for the three classes of Raman modes. 𝜀 was assumed to be 1.33.
vicinity of a small silver sphere for the three classes of Raman modes. ε r was 𝑟assumed to be 1.33.

Due to disrupted spherical symmetry in the case of non-spherical particles, the polarization
5.2. Local Field Distribution in Presence of Hot-Spots and Its Relation to Molecular Orientation
insensitive SERS EF is expected to split into two (or more) geometry-dependent components. For
Unfortunately,
example, in the casethe
of optical response of
prolate spheroids, very
only thefew plasmonic
incident systems parallel
field polarized can be to computed as
the long axis
straightforwardly as shown
results in the greatest in Section
surface 5.1.Such
fields [49]. Bothelectrodynamic
the experimental results aswere
calculations well performed
as the theoretical
already
calculations
in the 1980sshow that electromagnetic
and their understanding isfield canforbeelucidation
a key confined inofnanometer-sized metallic clefts.
the polarization-resolved optical
Analytical
response treatment of a twosystems
of more complex sphere system,
such as performed already in
elongated particles orthe 1980s [50,51],
nanorods (NRs),suggests
which willthatbe
discussedresiding
molecules in moreindetail in the second
a nanogap between part
theofNPs
thisare
section. 5
subjected to ∼10 × stronger electromagnetic
field in comparison to the field enhanced by one single NP. Due to mutual interaction, electromagnetic
5.2. in
field Local
suchField Distribution
cavities is muchinlarger
Presence
thanofthe
Hot-Spots
sum ofandthe Its Relation
fields causedto Molecular Orientation spheres,
by two non-interacting
which Unfortunately,
has been explained the by strongresponse
optical electromagnetic
of verycoupling between the
few plasmonic NPs, giving
systems can berisecomputed
to coupledas
plasmon resonances [13,35,52–54]. Coupled plasmon resonances are characterized by
straightforwardly as shown in Section 5.1. Both the experimental results as well as the theoretical much dramatic
difference
calculationsin responses
show thattoelectromagnetic
different polarizations.
field canSuch an extreme
be confined in enhancement occurs
nanometer-sized only ifclefts.
metallic the
incident polarization is along the interparticle axis. For light polarized across the
Analytical treatment of a two sphere system, performed already in the 1980s [50,51], suggests thatinterparticle axis,
the enhancement is almost negligibly different from its value at a single, isolated particle [40,42].
Closer analysis shows that with NP radii of 45 nm (average size of the silver NPs used) and the
separation distance of d = 5.5 nm (diameter of a hemoglobin molecule, [52]), the enhancement effect
adds another ∼2–3 orders of magnitude in comparison to isolated NPs and further ∼2–3 orders of
magnitude after decreasing the NP gap down to ∼1 nm, making the total EF of ∼1011 . Nowadays, SERS
EFs of ∼1011 –1012 are considered to be theoretical limits (for silver) on account of the electromagnetic
contribution, sufficient even for observation of single-molecular SERS [55–57]. Not surprisingly, there
often exists an inverse relationship between the EF and reproducibility of the substrate used (“SERS
uncertainty principle”).
The extremely spatially-localized sites providing extreme enhancement were dubbed hot-spots in
the literature [13]. However, geometry of hot-spots is not restricted only to two mutually interacting
spheres. Generally, for systems of aggregated particles with nanoscale crevices or junctions where
interparticle separation can be made very small, extremely localized regions of ultrahigh enhancement
were predicted [53] as well as experimentally demonstrated [56,58–60]. It means that the enhancement
in the gap region where majority of the electromagnetic energy is packed completely overwhelms
the surface average and produces the dominant contribution to single-molecular sensitivity in SERS.
Besides, theory also predicts very strong enhancement of the electromagnetic field at sharp metallic
 interacting spheres. Generally, for systems of aggregated particles with nanoscale crevices or
junctions where interparticle separation can be made very small, extremely localized regions of
ultrahigh enhancement were predicted [53] as well as experimentally demonstrated [56,58–60]. It
means that the enhancement in the gap region where majority of the electromagnetic energy is
packed completely overwhelms the surface average and produces the dominant contribution to
Nanomaterials 2018, 8, 418 18 of 38
single-molecular sensitivity in SERS. Besides, theory also predicts very strong enhancement of the
electromagnetic field at sharp metallic tips and large curvature regions due to so-called lightning-rod
effect.
tips andIn hot-spot
large curvaturedominated
regions due systems, information
to so-called on the effect.
lightning-rod symmetry of thedominated
In hot-spot Raman tensor
systems,(or
information on the symmetry
molecular depolarization of Equation
ratio, the Raman tensor
(24)) (or molecular
is almost depolarization
completely overriddenratio, Equation pattern
by anisotropic (24)) is
of the nanostructures.
almost completely overridden by anisotropic pattern of the nanostructures.
Letus
Let ustake
takeaa closer
closer look
look at
at the
the simplest
simplest case
case of
of aa hot-spot—a
hot-spot—a dimer
dimer formed
formed by by two metallic
metallic NPs.
Letus
Let usadopt
adoptsuch
suchaaCartesian
Cartesiansystem
system of ofcoordinates
coordinates thatthat the
the dimer
dimer axis
axis isisparallel
parallel with the x𝑥axis
withthe axisasas
depictedin
depicted inFigure
Figure8.8. In
Inthis
thiscase,
case, the
the optical
optical response
response of ofaamolecule
moleculesituated
situated ininthe
thegap
gapbetween
between thethe
twoNPs
two NPsisiscompletely
completelydominated
dominatedby bypolarization
polarizationin thex𝑥direction
inthe direction(Figure
(Figure9). 9).

Figure 8. Geometry of a dimer. The Cartesian coordinate system is chosen so that the dimer axis lies
Figure 8. Geometry of a dimer. The Cartesian coordinate system is chosen so that the dimer axis lies in
in the 𝑥 direction, the excitation beam travells along the 𝑧 direction with the incident polarization
the x direction, the excitation beam travells along the z direction with the incident polarization making
making an angle 𝜗 with the 𝑥 axis. Scattered radiation is assumed to be collected in a backscattering
an angle ϑ with the x axis. Scattered radiation is assumed to be collected in a backscattering geometry,
geometry, allowing only the polarization making an angle 𝜑 with respect to the incident polarization
allowing only the polarization making an angle ϕ with respect to the incident polarization (i.e., making
(i.e., making an angle 𝜗 + 𝜑 with respect to the 𝑥 axis) to be detected.
an angle ϑ + ϕ with respect to the x axis) to be detected.

As already mentioned, the ratio of obtained intensities for polarization parallel/perpendicular to

the dimer axis may amount to 105 . Therefore, the SERS response for polarization perpendicular to the
dimer axis can be neglected (the enhancement effect of the dimer is comparable to that of one isolated
particle in this case). Thus, assuming that the incident wavevector lies in the z direction, local field
strength for polarization making an angle ϑ with respect the x-axis will be:

Elocal = γE0 cos ϑ. (54)

In other words, amplification of the surface field is of tensorial nature and can be expressed as:
 
↔
γ 0 0
A= 0 1 0 . (55)
 
0 0 1

The factor γ represents amplification of the laser field polarized along the x-axis. Similarly to
the incident radiation, also the scattered radiation exhibits different coupling efficiency depending
on the angle made between the dimer axis. Thus, the local field should scale as ∼ cos ϑ and the local
intensity as ∼ cos2 ϑ. Assuming the backscattering geometry and for simplicity the same EF for both
the incident and the scattered radiation, respective intensities obey the following trends:

I (180◦ , ⊥sc , ⊥i ) = I (180◦ , ksc , ki ) ∼ cos4 ϑ, (parallel polarizations), (56)

I (180◦ , ksc , ⊥i ) = I (180◦ , ⊥sc , ki ) ∼ cos2 ϑ sin2 ϑ, (crossed polarizations), (57)

Nanomaterials 2018, 8, 418 19 of 38

and the depolarization ratio is:

ρ = tg2 ϑ, (58)

which diverges for ϑ = 90◦ . In this crudest approximation, the depolarization ratio is determined solely
by the plasmonic nanostructures and is not by any way linked to the molecular depolarization ratio.
Nanomaterials 2018, 8, x FOR PEER REVIEW 18 of 37

𝐸
Figure 9. Local intensity enhancement 𝑀1 =( 𝑙𝑜𝑐 )22 in a logarithmic scale in a plane through the
𝐸
Figure 9. Local intensity enhancement M1 = EEloc00 in a logarithmic scale in a plane through the
centers of the Ag spheres and perpendicular to the incident → wavevector 𝑘⃗ versus incident
centers of the Ag
polarization: (a)spheres and(c)
0°; (b) 30°; perpendicular to the
60° and (d) 90°. The incident
incident wavevector
wavelength k was
versus incident
514.5 nm inpolarization:
all cases. The
0◦ ; (b)represent
(a)arrows 30◦ ; (c) 60 ◦
theand
◦ . The incident wavelength was 514.5 nm in all cases. The
(d) 90polarizations.
different In (e), the SERS enhancement factor 𝑀 = 𝑀12 is shown arrowsas
represent the different polarizations. In (e), the SERS enhancement factor M = M 2 is shown as a
a function of the incident polarization 𝛼 for a point in the nanogap located at the dimer 1 axis 𝛿 = 0.5
function
nm away of the
from incident polarization
one spherical surface,α for
anda the
point fit in the nanogap
(solid line) to a located at the dimerIn
cos4 𝛼 dependence. (f),δ𝑀=averaged
axis 0.5 nm
away from one spherical surface, and the fit (solid line) to a cos 4 α dependence. In4 (f), M averaged
over all points 𝛿 = 0.5 nm outside the Ag sphere surface versus 𝛼 and fit to a cos 𝛼 dependency are
over all points
shown. δ = 0.5𝑅 nm
The radius = 45outside the Ag sphere
nm corresponds to thesurface versus
average and
sizeαof thefitAg cos4 α dependency
to ananoparticles used inarethe
shown. The radius R = 45 nm corresponds to the average size of the Ag
experiment while the separation distance 𝑑 = 5.5 nm corresponds to the diameter of a hemoglobinnanoparticles used in the
experiment while the separation
molecule. Reprinted distance
with permission d=
from 5.5 nm corresponds to the diameter of a hemoglobin
[54].
molecule. Reprinted with permission from [54].
As already mentioned, the ratio of obtained intensities for polarization parallel/perpendicular to
the The
dimer axis may predictions
theoretical amount to 10 5. Therefore, the SERS response for polarization perpendicular to the
above may be easily verified experimentally in the case of colloids.
dimer axis
Colloids maycanbe be neglected
regarded as a(the enhancement
collection effectoriented
of randomly of the dimer is comparable
hot-spots where thetooverall
that ofSERS
one isolated
signal
particleoriginates
usually in this case).
fromThus,
a fewassuming
moleculesthat the incident
residing wavevector
in the hot-spot sites lies
and in
thethe 𝑧 direction,
rest occupyinglocal field
the less
strength for
enhancing polarization
sites making
[40]. Although anan angle 𝜗 with
extremely smallrespect the 𝑥-axis
percentage will be:
(difficult to determine precisely) of
molecules actually cover the hottest sites, their presence is decisive for the total enhancement [61,62].
𝐸𝑙𝑜𝑐𝑎𝑙 = 𝛾𝐸0 𝑐𝑜𝑠 𝜗. (54)
In other words, amplification of the surface field is of tensorial nature and can be expressed as:
𝛾 0 0
𝐴⃡ = (0 1 0). (55)
0 0 1
The factor 𝛾 represents amplification of the laser field polarized along the 𝑥-axis. Similarly to the
incident radiation, also the scattered radiation exhibits different coupling efficiency depending on
the angle made between the dimer axis. Thus, the local field should scale as ~ cos 𝜗 and the local
Nanomaterials 2018, 8, 418 20 of 38

Integration over random orientations of a collection of dimers in standard azimuthal coordinates

(Equation (5)) yields: s
cos2 ϑ sin2 ϑ cos2 ϕdΩ 1
ρ= s = , (59)
cos4 ϑ dΩ 3
which is the expected value of the depolarization ratio observed experimentally [41,43]. Extremely low
detection limits for molecules adsorbed on hot-spot dominated SERS active systems opposed to almost
complete loss of more detailed information on the symmetry of the vibrations involved may be viewed
as another consequence of the SERS uncertainty principle.
Let us demonstrate some of the previous theoretical aspects on the Raman spectra of methylene
blue (MB) measured in our laboratory in both (90◦ , ⊥s , ⊥i ) (Ivv ) and (90◦ , ks , ⊥i ) ( Ivh ) experimental
configurations (Figure 10). MB was in the form of a water solution (concentration 10−4 M) and
in hydroxylamine-reduced Ag NPs (final MB concentration of 10−6 M), prepared by the standard
procedure [63]. In both cases the excitation wavelength was 532 nm, therefore both spectra should
be considered as preresonance ones. This brings certain difficulty with fluorescence and the need
for baseline subtraction in the case of non-SERS measurements, which makes determination of
respective peak intensities more difficult, however, all basic principles highlighting the differences in
Raman/SERS depolarization
Nanomaterials 2018, ratios are retained.
8, x FOR PEER REVIEW 20 of 37

Figure 10. Polarized (preresonance) Raman spectra of methylene blue (MB) in a water solution
Figure 10. Polarized (preresonance) Raman spectra of methylene blue (MB) in a water solution
(excitation wavelength 532 nm; after fluorescence baseline subtraction) and corresponding
(excitation wavelength 532 nm; after fluorescence baseline subtraction) and corresponding
depolarization ratios (black dots). Symmetry species of respective vibrational modes are indicated in
depolarization ratios (black dots). Symmetry species of respective vibrational modes are indicated in
the upper part. Comparison with the depolarization ratios obtained from hydroxylamine-reduced Ag
the upper part. Comparison with the depolarization ratios obtained from hydroxylamine-reduced Ag
NPs is indicated by red dots (SERS spectra are not shown for better clarity). MB concentration was
NPs is indicated by red dots (SERS spectra are not shown for better clarity). MB concentration was
10−−44 M,
10 M, laser
laser power
power 100
100 mW,
mW, accumulation
accumulation time
time 55 min.
min.

MB belongs to the C2 point group possessing 108 normal vibrational modes distributed among A
and B symmetry species as Γ3N−6 = 54A ⊕ 54B (A denoting the totally-symmetric vibrations and B
the non-totally symmetric vibrations). Respective bands in Figure 10 are labelled as A or B on the basis
of theoretical calculations adapted from [64].
Nanomaterials 2018, 8, 418 21 of 38

From Figure 10, several conclusions can be drawn. First, it is obvious that the most intense bands in
preresonance Raman (non-SERS) spectrum are almost entirely the A species, which is fully consistent
with the theory of resonance Raman scattering. Second, evaluation of the Raman depolarization
ratios reveals that for the A species ρ lies between ∼0.19 and 0.23 while for the B species ρ is up
to ∼0.45, approximately twice as much as for A. This is again in full compliance with the theory
according to which the depolarization ratio of symmetric vibrations (A) is always lower than for
non-totally symmetric vibrations (B). However, this difference in the depolarization ratios between
A and B species is wiped out completely when the measurement is performed on a silver colloid
where the depolarization ratios are (within the experimental error) all around 1/3, irrespective of the
symmetry
Figure of10.the vibration
Polarized involved. Since
(preresonance) Raman at spectra
lower analyte concentrations,
of methylene blue (MB) molecules are expected
in a water solution
to adopt sub-monomolecular
(excitation wavelength 532surface coverage
nm; after and tendbaseline
fluorescence stack upsubtraction)
in multiple and
layers at higher analyte
corresponding
concentrations,
depolarization ratios (black dots). Symmetry species of respective vibrational modes are indicated inwhile
series of concentration-dependent polarization-resolved SERS measurements
monitoring
the upper the depolarization
part. ratios
Comparison with thecould represent
depolarization a unique
ratios way
obtained to distinguish
from these two regimes.
hydroxylamine-reduced Ag
Figure
NPs 11 showsby
is indicated our
redresults from concentration-dependent
dots (SERS MB SERS
spectra are not shown for better measurements.
clarity). Singular
MB concentration was value
decomposition
10−4 M, laseralgorithm
power 100 [65]
mW,based on factor
accumulation analysis
time 5 min. was employed to treat the spectral set.

Figure
Figure 11.
11. Results
Results of of factor
factor analysis
analysis of
of concentration-dependent
concentration-dependent SERS SERS measurements
measurementsof of methylene
methylene
blue (MB). Profile of the V i1 coefficients reflects the variation in MB SERS intensity with concentration.
blue (MB). Profile of the Vi1 coefficients reflects the variation in MB SERS intensity with concentration.
Profile
Profileof
ofthe
theVVi2i2coefficients
coefficientsandandthe
theshape
shapeofofthe
thesecond
secondsubspectrum
subspectrumreveal
revealaasystematic
systematicdecrease
decreasein
in
the ratio of EFs for B/A modes with increasing concentration (cf. Figure
the ratio of EFs for B/A modes with increasing concentration (cf. Figure 10). 10).

Profile of the Vi1 coefficients reflects the variation in total MB intensity with concentration.
As expected, Raman intensity rises when MB concentration is increased from 10−8 M to ~10−6 M
concentration but starts to decrease when further increasing MB concentration, probably due to MB
aggregation and less efficiency in occupying the highest-enhancing sites. After normalizing the original
spectral set using the Vi1 coefficients, the Vi2 coefficients were found almost perfectly linear (in a
semi-logarithmic scale) with concentration with the second subspectrum highlighting the differences
between A and B symmetry modes. In other words, when increasing MB concentration, the ratio of EFs
for B/A modes is systematically decreasing. This finding clearly reveals a change in MB adsorption
Nanomaterials 2018, 8, 418 22 of 38

geometry with concentration, adopting a face-on adsorptive stance at lower concentrations and a
preferable edge-on adsorptive stance at higher concentrations [64,66].
Efforts to make use of the enhancement capability of dimers are still very lively in literature,
which can be documented by numerous studies from the last years [67–74], benefiting from extremely
high enhancement between two adjacent metallic NPs or between a metallic NP and a substrate.
A very promising analytical tool for biosensing applications include DNA origami, [75–77] where
NPs with defined distance can be constructed. However, all these illustrations raise a question about
whether or not there exists a general rule to reveal some information regarding the Raman tensor
elements of the analyte (intrinsic properties of the SERS probes) by measurements on nanostructured
metallic surfaces. Correlation between the morphological anisotropy and symmetry of the vibrational
species was found in [78]. Using a well-established pyridine molecule, the authors demonstrated that
the in-plane (ring stretching) vibrational modes are sensitive to the morphological anisotropy while
the out-of-plane (ring deformation) modes are not. A more in-depth approach was developed by P.
Gucciardi et al. [31,39,43], who in his analysis went beyond the commonly used E4 approximation.
For a hot-spot represented by a dimer with the axis lying in the x direction, the authors used the field
enhancement tensor in the form:  
↔
γ 0 0
A ( ωi ) =  0 1 0 , (60)
 
0 0 1
and the re-radiation enhancement tensor as:
 
↔
γ (1 + ε ) 0 0
A (ωsc ) =  0 1 0 , (61)
 
0 0 1

where ε is a small perturbation accounting for the difference between the enhancement for the incident
and the scattered radiation. The local field felt by the analyte and the amplified scattered field are thus:
→ ↔ → → ↔ →
E local = A (ωi ) E i , E SERS = A (ωsc ) E sc . (62)

These two fields are actually those that enter Equation (8). The theoretical SERS intensity for
many randomly oriented molecules as a function of the angle ϑ and ϕ (see Figure 8) was found to be:

I ∼ γ4 (1 + ε)2 hα2ii i cos2 ϑ cos2 (ϑ + ϕ) + γ2 (1 + ε)hα2ii i[1/2 sin 2ϑ sin 2(ϑ + ϕ)]
(63)
+γ2 hα2ij i[ε sin ϑ cos (ϑ + ϕ) + sin ϕ]2 + hα2ii i sin2 ϑ sin2 (ϑ + ϕ).

Equation (63) indicates that the total intensity contains a term proportional to γ4
(E4 approximation) but also other terms beyond the E4 approximation. More importantly, this equation
also suggests how the molecular depolarization ratio (the quantity corresponding to hα2ij i/hα2ii i that
would be measured for a “free” molecule without the presence of metallic nanostructure) is influenced
by the EF and the experimental geometry. The diagonal terms of the Raman tensor of the analyte
are enhanced as ∼ γ4 but the non-diagonal terms only as ∼ γ2 , which suggests that with higher γ,
the effect of the geometrical shape of the nanostructure will be dominant. Since the role of the EF γ in
Equation (63) is crucial, virtually causing linearization of the polarization of both the incoming and the
outgoing light, it remains very difficult to verify this formula experimentally [31,79]. More detailed
discussion can be found in [31].

5.3. Polarization Properties of Trimers and Aggregated NP Domains

Polarization properties of trimers have been studied in [80]. The third particle breaks the dipolar
symmetry of the two-particle junction, generating a wavelength-dependent polarization pattern,
Nanomaterials 2018, 8, x FOR PEER REVIEW 22 of 37

5.3. Polarization Properties of Trimers and Aggregated NP Domains

Nanomaterials 2018, 8, 418 23 of 38
Polarization properties of trimers have been studied in [80]. The third particle breaks the dipolar
symmetry of the two-particle junction, generating a wavelength-dependent polarization pattern,
which
which waswas not
not observed
observed in in the
the case
case of
of aa dimer.
dimer. InIn general,
general, polarization-dependent
polarization-dependent profile
profile of
of more
more
complex
complex NP NP clusters is expected
clusters is expected to toexhibit
exhibitmultiple
multiplemaxima
maximaand andminima,
minima,often
oftenreflecting
reflecting symmetry
symmetry of
of the nanocluster array [40] (Figure 12). If the enhancement at one specific point were
the nanocluster array [40] (Figure 12). If the enhancement at one specific point were monitored (such as monitored
(such
a gap as a gap 2between
between adjacent 2NPs),
adjacent NPs), full would
full anisotropy anisotropy would be
be observed. observed.
Although the Although the SERS
SERS characteristics
characteristics of any single probe-adsorption
of any single probe-adsorption were found to weredepend found to depend onfor
on polarization polarization for trimers
trimers in [81] as well in
as
[81] as well as supported by finite-difference time-domain (FDTD) simulations,
supported by finite-difference time-domain (FDTD) simulations, the periodic array of regular Ag the periodic array of
regular
NP trimersAg NPwastrimers
found towasformfound to form a polarization-independent
a polarization-independent SERSfact
SERS signal. This signal. This fact was
was attributed to a
attributed to a global structural D symmetry of the nanopit area. The polarization
global structural D3h symmetry of the nanopit area. The polarization independence was even better as
3h independence
was even better
compared as compared
to large-area to large-area
Ag nanowires and Ag nanowires
randomly and randomly
disordered disordered
Ag NPs of uneven Agsize,
NPs producing
of uneven
size,
clearlyproducing
discrepantclearly discrepant
SERS signals underSERS signals
different under
incident different incident
polarizations. polarizations.
Thus, supposing Thus,
the surface
supposing the surface is uniformly covered with an analyte, such clusters will show
is uniformly covered with an analyte, such clusters will show a considerable degree of isotropy as a a considerable
degree of isotropy as a whole [58,82,83].
whole [58,82,83].

Figure 12. A simple example of a cluster with quasi-isotropy. In this case an example of a 2D cluster
Figure 12. A simple example of a cluster with quasi-isotropy. In this case an example of a 2D
formed by threebycylinders
cluster formed (radii =(radii
three cylinders 25 nm) = 25with
nm)a with
smalla separation gap (𝑑gap
small separation = 6.7 (d nm)
= 6.7isnm)
shown. The
is shown.
problem
The problem is solved in the electrostatic approximation, wavelength = 357 nm. The polarization is
is solved in the electrostatic approximation, wavelength = 357 nm. The polarization is
changed the plane for different angles 𝛽
changed in the plane for different angles β as shown in (a) where the electromagnetic intensity (on aa
in as shown in (a) where the electromagnetic intensity (on
logarithmic
logarithmicgrey-scale)
grey-scale)isisexplicitly
explicitlyshown
shown forfor
thethe
case 𝛽=
case β 90°.
= 90A◦ .hot-spot in between
A hot-spot in between cylinders 2 and2
cylinders
3and
can3becan
easily seen. If we look at the maximum enhancement at any point on
be easily seen. If we look at the maximum enhancement at any point on the surface of the surface of the cluster,
there will always
the cluster, there be
willa always
place thatbe profits
a placethe thatmost from
profits thethe particular
most from the orientation
particularoforientation
the field and of we
the
obtain thewe
field and curve labelled
obtain as ‘‘total’’
the curve in (b).
labelled The three
as “total” maxima
in (b). in themaxima
The three latter areinthe
thethree
latterpossible two-
are the three
cylinder hot-spots in this
possible two-cylinder clusterinthat
hot-spots thisachieve theirachieve
cluster that highesttheir
value whenvalue
highest the field
when is the
aligned
field along the
is aligned
axis joining any two of them. If the surfaces were uniformly covered with analyte,
along the axis joining any two of them. If the surfaces were uniformly covered with analyte, an almost an almost isotropic
response
isotropic is obtained.
response is The degreeThe
obtained. of isotropy
degree increases
of isotropy with the complexity
increases with theofcomplexity
the cluster.ofHowever,
the cluster.if
the enhancement
However, at one specific
if the enhancement point
at one is monitored
specific (in this case
point is monitored the enhancement
(in this case the enhancement in the gap in
in the
between 2 and 3)
gap in between we obtain
2 and the curve
3) we obtain labelled
the curve as ‘‘single’’
labelled in (b)inwhich
as “single” shows
(b) which the the
shows full full
anisotropy
anisotropy of
any single
of any hot-spot.
single Adapted
hot-spot. Adapted from [40][40]
from withwithpermission
permission from TheThe
from Royal Society
Royal Society of Chemistry.
of Chemistry.

Similar conclusion was supported in [84] using arrays of NRs prepared by inkjet printing
Similar conclusion was supported in [84] using arrays of NRs prepared by inkjet printing
technique, or by Shegai in [85]. Shegai attributed different depolarization ratios of different
technique, or by Shegai in [85]. Shegai attributed different depolarization ratios of different rhodamine
rhodamine 6G bands adsorbed on silver nanocrystal aggregates to charge transfer resonances since
6G bands adsorbed on silver nanocrystal aggregates to charge transfer resonances since the
the electromagnetic field around the nanocrystals was found near-isotropic. A more theoretical
electromagnetic field around the nanocrystals was found near-isotropic. A more theoretical approach
approach was adopted in [86]. Making use of the group theory, the authors showed that
was adopted in [86]. Making use of the group theory, the authors showed that a nanostructure that
a nanostructure that belongs to Cn symmetry point group (𝑛 ≥ 3) has an optical response that is
belongs to Cn symmetry point group (n ≥ 3) has an optical response that is insensitive to the light
insensitive to the light polarization when the wavevector is parallel to the Cn axis. Attempts have
polarization when the wavevector is parallel to the Cn axis. Attempts have been made to verify this
been made to verify this claim experimentally using arrays of nanotriangles, nanostars and other
claim experimentally using arrays of nanotriangles, nanostars and other nanoobjects [87]. However,
nanoobjects [87]. However, due to non-negligible relative standard deviation of the SERS
due to non-negligible relative standard deviation of the SERS measurements and the fact that the
theoretical criteria mentioned above are not perfectly met, the decision whether or not there exists a
polarization-dependent pattern is often difficult.
Nanomaterials 2018, 8, x FOR PEER REVIEW 23 of 37

measurements and the fact that the theoretical criteria mentioned above are not perfectly met, the
decision whether or not there exists a polarization-dependent pattern is often difficult.
Nanomaterials 2018, 8, 418 24 of 38

5.4. Polarization Properties of Regular Solid Plasmonic Platforms

5.4. Polarization
ConsiderationsProperties
aboveof Regular
can beSolid Plasmonic extended
analogously Platforms to many other plasmonic platforms
possessing anisotropic
Considerations abovemorphology
can be analogouslyfabricated using
extended various
to many otherprocedures [88]. For possessing
plasmonic platforms example,
polarization properties fabricated
anisotropic morphology of metallic usingnanocubes [89–91], [88].
various procedures arraysForof silver polarization
example, NP rows [92], gold
properties
nanoassemblies [93], metallic gratings structures [94], half-shells [95] etc. were
of metallic nanocubes [89–91], arrays of silver NP rows [92], gold nanoassemblies [93], metallic gratingsstudied in literature.
A frequently-studied
structures SERS-active
[94], half-shells [95] etc. system from the
were studied point of view
in literature. of polarization and
A frequently-studied angular
SERS-active
characteristics
system from the arepoint
arraysofofview
elongated NPs appearing
of polarization under different
and angular names in
characteristics areliterature
arrays ofsuch as NRs
elongated
[96–99], nanowires [43,60,100–103], nanoantennas [104,105], nanorattles [106,107],
NPs appearing under different names in literature such as NRs [96–99], nanowires [43,60,100–103], nanobones [108] or
NP-nanowire
nanoantennas systems
[104,105],[109–111]
nanorattles where the ~nanobones
[106,107], cos 𝜗 trends[108]were or often observedsystems
NP-nanowire (Figures 13–15).
[109–111]
Published
where the results
∼ cos ϑindicate that the
trends were optical
often response
observed for light
(Figures polarized
13–15). parallel/perpendicular
Published results indicate that to the
long axis
optical of the nanoobjects
response (relatedparallel/perpendicular
for light polarized to excitation of longitudinal plasmon
to the long axis modes (LM) or transverse
of the nanoobjects (related
plasmon modes
to excitation (TM), respectively)
of longitudinal is indeed
plasmon modes (LM)different. However,
or transverse the SERS
plasmon modesintensity profile with
(TM), respectively) is
varying angle/polarization is a function of a wide range of parameters, such as
indeed different. However, the SERS intensity profile with varying angle/polarization is a function of athe dimensions of the
metallic
wide rangeobjects, their aspect
of parameters, suchratios
as theand spatial arrangement,
dimensions of the metallic material (Ag aspect
objects, their or Au), the and
ratios excitation
spatial
wavelength
arrangement,ormaterial
orientation
(Ag or ofAu),
the the
probe molecules
excitation on the surface
wavelength [85,112,113],
or orientation whichmolecules
of the probe resulted onin
seemingly contradictory accounts appearing in literature. Last but not least,
the surface [85,112,113], which resulted in seemingly contradictory accounts appearing in literature. what hampers more
precise
Last butanalyses
not least,ofwhat
the angular
hampersand morepolarization effectsofisthe
precise analyses sample bleaching
angular and photodecomposition
and polarization effects is sample
induced
bleachingbyandthephotodecomposition
incident laser, whichinduced causes poorer reproducibility
by the incident and brings
laser, which causes further uncertainty to
poorer reproducibility
the
and set of experimentally
brings further uncertainty measured points.
to the set This aspect was
of experimentally usuallypoints.
measured resolved Thisinaspect
the literature
was usually by
applying
resolved ina the
correction
literature assuming
by applying exponential
a correctiondecay
assuming of the SERS signal
exponential decay withof the time
SERS [81,97,109].
signal with
Polarization-dependent properties of SERS
time [81,97,109]. Polarization-dependent microarrays
properties also found also
of SERS microarrays usefound
in elimination of the
use in elimination
polarization insensitive
of the polarization spurious
insensitive bands bands
spurious originating from the
originating frombulk
thematerial [94]. [94].
bulk material

Figure
Figure 13. 13. Unpolarized
Unpolarized SERS SERS intensity
intensity of of methylene
methylene blueblue adsorbed
adsorbed on on gold
gold nanowires
nanowires (NWs)
(NWs) versus
versus
excitation polarizationθ as
excitation polarization 𝜃 as measured
measured in a backscattering
in a backscattering geometry.
geometry. SERS intensity
SERS intensity was foundwas found
maximum
→
for θ = 0◦ for
maximum 𝜃 = black
(inset, 0° (inset, black
line), i.e.,line), i.e., for incident
for incident polarization
polarization vector (denoted
vector (denoted 𝑒𝑒𝑥𝑐 ) parallel
as e exc )asparallel to the
→ →
to the nanocavity axis 𝑛⃗ , and minimum for 𝜃 ◦= 90° (inset, red line),
nanocavity axis n x , and𝑥 minimum for θ = 90 (inset, red line), i.e., e exc parallel i.e., 𝑒𝑒𝑥𝑐 parallel
to the NWs longlong
to the NWs axis
→ 2 𝜗 law (blue line), plus low order
axis 𝑛
⃗ . The SERS intensity profile was well fitted with the ~
2 cos
𝑦 SERS intensity profile was well fitted with the ∼ cos ϑ law (blue line), plus low order terms
n y . The
terms
(Equation(Equation (63)). Excitation
(63)). Excitation wavelength
wavelength was 633was nm, 633
thusnm,
the thus
SERSthe SERS spectrum
spectrum is considered
is considered as a
as a resonance
resonance
one (SERRS). oneAdapted
(SERRS). Adapted
with permission with from
permission from [43].
[43]. Copyright Copyright
2011 American2011 American
Chemical Chemical
Society.
Society.
Nanomaterials 2018, 8, 418 25 of 38
Nanomaterials 2018, 8, x FOR PEER REVIEW 24 of 37

Figure
Figure 14. Polarized
14. (a): (a): Polarized
SERSSERS spectra
spectra of trans-1,2-bis(4-pyridyl)-ethene
of trans-1,2-bis(4-pyridyl)-ethene (BPE)
(BPE) on on Ag nanorod
Ag nanorod substrates
substrates as a function of the incoming polarization angle measured in a backscattering geometry.
as a function of the incoming polarization angle measured in a backscattering geometry. (b): Polarized
(b): Polarized UV/Vis/NIR absorption spectra. Cross-section of the NR array is schematically depicted
UV/Vis/NIR absorption spectra. Cross-section of the NR array is schematically depicted in the inset
in the inset (inclination angle of the nanorods was ~ 71° with respect to the surface normal). Both
(inclination angle of the nanorods was ~71◦ with respect to the surface normal). Both diagrams show
diagrams show the p-polarization direction with the incident polarization almost parallel to the major
the p-polarization direction with the incident polarization almost parallel to the major long axis of the
long axis of the nanorods. The maximum SERS intensity was observed in the polarization direction
nanorods. The maximum SERS intensity was observed in the polarization direction perpendicular
perpendicular to the long axis of the Ag nanorods, while the UV/Vis absorbance was strongly
to the long axis of the Ag nanorods, while the UV/Vis absorbance was strongly polarized along
polarized along the direction of the long axis of the NR array. Adapted with permission from [97].
the direction of the long axis of the NR array. Adapted with permission from [97]. Copyright 2006
Copyright 2006 American Chemical Society.
American Chemical Society.
Although theory predicts the difference in the EFs for different light polarizations up to ~ 105
(for an ideal theory
Although structure [53,54]),the
predicts in difference
reality, thisin
factor is much
the EFs more modest.
for different light For example, the
polarizations to ∼of
upratio 105
(forSERS responses
an ideal structure in [53,54]),
the directions parallel/perpendicular
in reality, this factor is much moreto themodest.
NRs wasFor found to be the
example, around
ratio0.8 for
of SERS
silver NRs as reported in [97]. This finding was attributed to the lateral arrangement
responses in the directions parallel/perpendicular to the NRs was found to be around 0.8 for silver NRs of the NR lattice
and strong
as reported inelectromagnetic
[97]. This finding coupling between adjacent
was attributed metallic
to the lateral NRs instead
arrangement ofof preferential
the NR latticemolecular
and strong
orientation of coupling
electromagnetic the probebetween
moleculeadjacent
on the surface.
metallicANRs
similar trendofwas
instead observed molecular
preferential for coupledorientation
metallic
nanowires [60], oriented silver nanowire films [101], silver nanowire rafts [100]
of the probe molecule on the surface. A similar trend was observed for coupled metallic nanowires or self-organized gold
[60],
nanowires [43]. In all these cases the SERS intensity ratios for light polarized perpendicular/parallel
oriented silver nanowire films [101], silver nanowire rafts [100] or self-organized gold nanowires [43].
with respect to the NR axes was found around 5 (see Figure 13) and explained by excitation of a new
In all these cases the SERS intensity ratios for light polarized perpendicular/parallel with respect
plasmon mode trapped in the interstices between adjacent, parallel wires, similarly to the case of a
to the NR axes was found around 5 (see Figure 13) and explained by excitation of a new plasmon
dimer. However, this only happens if the interwire distance is sufficiently small (< ~ 10 nm) [60]. By
mode trapped in the interstices between adjacent, parallel wires, similarly to the case of a dimer.
contrast, antenna arrays similar to bent needles [104], nanorattles [106], single gold NRs [99] or
However, this only happens if the interwire distance is sufficiently small (<~10 nm) [60]. By contrast,
aligned Ag NRs prepared by a dynamic oblique deposition technique [98] were all found to exhibit
antenna arrays similar to bent needles [104], nanorattles [106], single gold NRs [99] or aligned Ag NRs
the enhancement higher for excitation light polarized parallel to the needle/NR direction than for the
prepared by a dynamic
perpendicular oblique
case. Such deposition
results were, on technique
the other [98] wereusually
hand, all found to exhibitby
rationalized theintense
enhancement
local
higher for excitation light polarized parallel to the needle/NR direction than
electromagnetic fields emanating from points of high curvature, such as NR tips (“lightning-rod for the perpendicular
case. Such and
effect”) results
thewere, on the
dominant other
role hand,
of the usually rationalized
longitudinal plasmon modes. by intense
To sum local
up,electromagnetic fields
the experimentally
emanating from points of high curvature, such as NR tips (“lightning-rod
measured polarization dependences often may be understood as a result of “competition” between effect”) and the dominant
role of the longitudinal plasmon modes. To sum up, the experimentally measured polarization
Nanomaterials 2018, 8, 418 26 of 38

dependences often may be understood as a result of “competition” between the enhancement provided
by longitudinal plasmon modes and by transverse plasmon modes. For example, for NRs growing
onNanomaterials
a substrate in8,one
2018, specific
x FOR direction as indicated in Figure 14, the obtained SERS intensities
PEER REVIEW 25 of 37are
expected to be:
the enhancement provided by longitudinal plasmon modes and by transverse plasmon modes. For
example, for NRs growing→i → on 2 a substrate →iin→ one
2 specific direction
→ 2 as indicated
→sc →in 2Figure 14, the
  
0 →sc 0
I ∼
obtained SERS a|| ( e · are
intensities + a⊥ (1 −to( be:
l ) expected e · l ) ) · a|| ( e · l ) + a⊥ (1 − ( e · l ) ) , (64)
2 2 2 2
→i 𝐼 ~ [𝑎|| (𝑒 𝑖 ∙ 𝑙 ) + 𝑎⊥ (1 − (𝑒 𝑖 ∙ →
𝑙 )sc)] ∙ [𝑎|| ′ (𝑒 𝑠𝑐 ∙ 𝑙 ) + 𝑎⊥ ′ (1 − (𝑒 𝑠𝑐 ∙ 𝑙 ) )],→ (64)
where e is the unit incident field vector, e is the unit scattered field vector, l is the unit vector
where 𝑒 𝑖 isby
determined the
theunitlongincident
axis of field
the NRs and𝑒a𝑠𝑐|| and
vector, is thea⊥unit
are scattered 𝑙 isplasmonic
field vector,the
factors determining the unit vector
response
todetermined
polarization byparallel/perpendicular
the long axis of the NRs and to the𝑎|| and
long𝑎axis
⊥ areof
factors
the NRsdetermining the plasmonic
for the excitation lightresponse
(unprimed
to polarization
values) parallel/perpendicular
and the scattered light (primed tovalues).
the long Theaxis first
of thebracket
NRs forinthe excitation
Equation (64)light (unprimed
accounts for the
enhancement of the incident radiation and the latter accounts for the enhancement of the for
values) and the scattered light (primed values). The first bracket in Equation (64) accounts the
scattered
enhancement of the incident radiation and the latter accounts for the enhancement
radiation. What matters is which of the two factors a⊥ and a|| is larger. In general, the answer to this of the scattered
radiation.
question What
is not matterssince
uniform is which
it is of the two factors
extremely 𝑎⊥ and
sensitive 𝑎|| specific
to the is larger.preparation
In general, the answer toofthis
procedure given
plasmonic nanostructures. As a rule of thumb, for arrays of regular NRs stacked in closeofproximity
question is not uniform since it is extremely sensitive to the specific preparation procedure given
toplasmonic
each othernanostructures.
(gap < ~10 nm), As a rule of thumb, for arrays of regular NRs stacked in close proximity
the role of factor a⊥ usually prevails. Similar conclusion holds true
to each other (gap < ~ 10 nm), the role of factor 𝑎⊥ usually prevails. Similar conclusion holds true for
for NP-nanowire systems. On the other hand, for rather isolated or well-separated NRs, the most
NP-nanowire systems. On the other hand, for rather isolated or well-separated NRs, the most
contributing term to the biggest SERS sensitivity will usually be the factor a|| [114,115].
contributing term to the biggest SERS sensitivity will usually be the factor 𝑎|| [114,115].
Sensitivity of the plasmonic anisotropy to the excitation wavelength was investigated in more detail
Sensitivity of the plasmonic anisotropy to the excitation wavelength was investigated in more
in [116] where two different excitation wavelengths were used to study the polarization-dependent
detail in [116] where two different excitation wavelengths were used to study the polarization-
SERS on anisotropic Ag NP array. The study demonstrates that SERS nanostructures can possess
dependent SERS on anisotropic Ag NP array. The study demonstrates that SERS nanostructures can
completely different polarization
possess completely characteristics
different polarization when using
characteristics whentwo different
using wavelengths.
two different The cos2The
wavelengths. ϑ and
2
sincosϑ2 𝜗trends
and sin 𝜗 trends were found using the 514 nm and 647 nm excitation wavelength, respectively 15).
were
2 found using the 514 nm and 647 nm excitation wavelength, respectively (Figure
Completely
(Figure 15).different
Completely polarization characteristics
different polarization were in good
characteristics werecorrespondence
in good correspondencewith the dielectric
with the
function
dielectric parallel
functionand perpendicular
parallel to the long
and perpendicular axislong
to the of the
axisnanostructures as determined
of the nanostructures by spectral
as determined by
ellipsometry.
spectral ellipsometry. This approach is expected to open novel possibilities in biosensing applications to
This approach is expected to open novel possibilities in biosensing applications due
increase in the specificity
due to increase of targetofdetection
in the specificity by using
target detection bymultiple excitation
using multiple wavelengths.
excitation wavelengths.

Figure15.15.Left:
Left:Setup
SetupforforRaman
Raman measurements
measurements used. −1 −1
Figure used. Middle:
Middle:Normalized
Normalizedintensity
intensityofof
the 1079
the cmcm
1079
band
band ofof 4-mercaptobenzonitrile
4-mercaptobenzonitrile (MBN)
(MBN) adsorbed
adsorbed onon anisotropic
anisotropic AgAg nanoparticle
nanoparticle arrays
arrays created
created by by
metal
metal evaporation on rippled silicon substrates as a function of 𝜃 obtained with the 514
evaporation on rippled silicon substrates as a function of θ obtained with the 514 nm (green triangles) nm (green
andtriangles)
the 647 and the 647
nm (red nm (red
squares) squares)The
excitation. excitation. Thesolid
respective respective solid lines to
lines correspond correspond to2fits
fits of a cos of
θ (green)
a cos 2 𝜃 (green) and sin2 𝜃 (red) function to the data. Right: Imaginary part of the effective dielectric
2
and sin θ (red) function to the data. Right: Imaginary part of the effective dielectric function of the Ag
function
film parallelof and
the Ag film paralleltoand
perpendicular the perpendicular
long axis of theto ripples.
the longAdapted
axis of the
withripples. Adapted
permission fromwith
[116].
permission from [116]. Copyright 2016 American Chemical Society.
Copyright 2016 American Chemical Society.

5.5. Angular Dependences in SERS

5.5. Angular Dependences in SERS
As far as angular dependences are concerned, the situation in SERS is again more complicated
As far as angular
in comparison dependences
to classical are concerned,
Raman measurements the situation
(Figure 4) due tointhe
SERS is again
presence of more complicated
a third element— in
comparison to classical Raman measurements (Figure 4) due to the presence of a third element—the
the nanostructure, which imposes its own anisotropic pattern in the directional properties of the SERS
nanostructure, which
signal. Moreover, theimposes its own
term angular anisotropicitself
dependences pattern in theambiguous
is rather directionalsince
properties
“angle”ofmay
the SERS
be
understood either as the angle between laser wavevector/polarization vector and a specific
nanostructure, or as the angle determining the specific illumination-observation geometry in a way
analogous to that described in Section 3. From the point of view of the latter meaning, the vast
Nanomaterials 2018, 8, 418 27 of 38

signal. Moreover, the term angular dependences itself is rather ambiguous since “angle” may
be understood either as the angle between laser wavevector/polarization vector and a specific
nanostructure, or as the angle determining the specific illumination-observation geometry in a way
Nanomaterials 2018, 8, x FOR PEER REVIEW 26 of 37
analogous to that described in Section 3. From the point of view of the latter meaning, the vast
majority of experiments
experiments have have been
been performed
performedonly onlyininbackscattering
backscatteringgeometry,geometry,which whichmeans
meansthat thata
adetailed
detailedinspection
inspectionon onthetheangular
angulardependence
dependenceofofthe theSERSSERS signal
signal is is somewhat
somewhat exceptional in the the
literature.
literature. However,
However, possible
possibleoptimization of plasmon-based
optimization of plasmon-based sensors for maximum
sensors for signal enhancement
maximum signal
relies, among other
enhancement relies,things,
amongon the right
other things,choice
on theof the
right excitation
choice of angle
the and/or
excitation theangle
scattering
and/or angle.
the
Geometric
scattering angle.reasons, such as change
Geometric reasons, in such
the size of the illuminated
as change in the size of areatheand possibly different
illuminated area andefficiency
possibly
in collection
different of the in
efficiency scattered
collection radiation must be also
of the scattered taken must
radiation into account when into
be also taken aiming to retrieve
account when
angular
aiming to dependences
retrieve angular[112,113,117]. This was
dependences often neglected.
[112,113,117]. This Thewasissue oftenof neglected.
optimum excitation
The issueand of
collection efficiency and
optimum excitation is ofcollection
great importance
efficiencyalsois ofingreat
the construction
importance also of optical waveguides of
in the construction with the
optical
aim to minimize
waveguides withlosses
the aim where coupling of
to minimize the SERS
losses where field to the waveguide
coupling of the SERSmodes field tomust
the be carefully
waveguide
optimized.
modes mustThis approach
be carefully is forecasted
optimized. This to open the
approach way towards
is forecasted lab-on-a-chip
to open sensing lab-on-a-
the way towards systems,
allowing
chip sensing multiplexed detectionmultiplexed
systems, allowing of extremely weak Raman
detection signals
of extremely on aRaman
weak highlysignals
dense on integrated
a highly
platform [118,119].
dense integrated platform [118,119].
A basic
basic idea
ideaofofthe thesimplest
simplestmodelsmodels trying
trying to to
elucidate
elucidate thetheSERSSERS angular
angularcharacteristics again
characteristics lies
again
in
liesthe
infact
the that
fact the
thatprimary
the primaryelectric field felt
electric by felt
field the bymolecule is the sum
the molecule of the
is the sum incident and the reflected
of the incident and the
field,
reflectedwhich induces
field, whichan oscillating
induces dipole in the
an oscillating adsorbed
dipole in the molecule. A modified
adsorbed molecule. A Greenler
modifiedmodelGreenler [5]
based
modelon [5]classical
based on electrodynamic dipole radiation
classical electrodynamic wasradiation
dipole used to explain
was used the anisotropic
to explain the nature of tilted
anisotropic
Ag
natureNRs,ofproducing
tilted Ag maximum
NRs, producing SERS intensities
maximumatSERS approximately
intensities45 at◦ approximately
relative to the surface normal
45° relative to[96],
the
but measured
surface normal still in the
[96], but backscattering
measured stillconfiguration (Figure 16).
in the backscattering Although the
configuration length
(Figure of Although
16). the NRs used the
(~868
lengthnm) of thewas NRscomparable
used (~ 868 to nm)
the excitation
was comparablewavelengthto the(785 nm) in wavelength
excitation this case, the(785 modelnm)was treating
in this case,
the surface
model of wasthetreating
NR as planar, neglecting
the surface of thethe NR
diffraction
as planar, effectneglecting
and calculating near-field effect
the diffraction intensities
and
using the Fresnel
calculating equations.
near-field Later,
intensities the authors
using the Fresnel upgraded their Later,
equations. model,the assuming
authorsthat the primary
upgraded their
field
model, at the NR surface
assuming that theis the sum offield
primary the incident
at the NR and reflected
surface fields
is the sumfrom theincident
of the Ag NR and from the
reflected
supporting
fields from Ag the film.
Ag NR Theandauthors
fromshowed that the Ag
the supporting presence
film. The of anauthors
underlying showed Ag film
that plays a crucialofrole
the presence an
underlying
for the SERSAg film plays due
enhancement a crucial
to its role
high for the SERS
reflectivity enhancement
[120]. Although due thesetosimple
its high reflectivity [120].
considerations were
proven
Although to these
capture the essential
simple angular
considerations characteristics
were proven to capture in SERS, theNRs of subwavelength
essential dimensions
angular characteristics in
SERS,
can be NRs of subwavelength
misleading to treat as planar dimensions can be misleading
as demonstrated to treata strong
in [121]. There, as planar as demonstrated
difference between the in
optical constants
[121]. There, of the difference
a strong NR films and those of
between thetheoptical
constituent materials
constants of the wereNRfound
films using generalized
and those of the
constituent materials
ellipsometry were found using generalized ellipsometry (Jones formalism).
(Jones formalism).

Figure 16. Left: Schematic illustration of the modified Greenler’s model of an induced dipole on a Ag
Figure 16. Left: Schematic illustration of the modified Greenler’s model of an induced dipole on a Ag
nanorod: Case I, where the dipole is perpendicular to the incident plane; Case II, where the dipole is
nanorod: Case I, where the dipole is perpendicular to the incident plane; Case II, where the dipole is on
on the
the incident
incident plane.
plane. All the the induced
All induced dipoles
dipoles are perpendicular
are perpendicular to thetoNR.
theRight:
NR. Right: Representative
Representative SERS
spectra of trans-1,2-bis(4-pyridyl)-ethene adsorbed on the Ag NR substrate at different incidentincident
SERS spectra of trans-1,2-bis(4-pyridyl)-ethene adsorbed on the Ag NR substrate at different angles
angles
θ. θ. The
The peak peak intensity
intensity was strongest
was strongest at the
at the angle angle θ45around
θ around ◦ , which 45°,
waswhich was in agreement
in agreement with
with the model.
the model. Adapted with permission
Adapted with permission from [96]. from [96].

Unlike optical constants of homogeneous materials, optical constants of nanostructured layers

depend (due to the presence of subwavelength structures) also on the incident angle. For example,
island-like Ag films in [122] exhibited uniaxial optical properties and NRs in [121] exhibited biaxial
properties with the complex refractive index different for different orientations of the incident angle
with respect to the NRs, similarly to the case of crystals [123]. A very similar conclusion was drawn
Nanomaterials 2018, 8, 418 28 of 38

Unlike optical constants of homogeneous materials, optical constants of nanostructured layers

depend (due to the presence of subwavelength structures) also on the incident angle. For example,
island-like Ag films in [122] exhibited uniaxial optical properties and NRs in [121] exhibited biaxial
properties with the complex refractive index different for different orientations of the incident angle
Nanomaterials 2018, 8, x FOR PEER REVIEW 27 of 37
with respect to the NRs, similarly to the case of crystals [123]. A very similar conclusion was drawn
from ourincident
both the work [112] where
angle we as
as well showed that
rotation of the
the optical
substrate constants
by 90°. On of silver NR hand,
the other arraysthey
are sensitive
were fairlyto
both the incident angle as well as rotation of the substrate by 90 ◦ . On the other hand, they were fairly
insensitive to flipping the substrate by 180°. Moreover, surface roughness or the presence of
insensitive
plasmonic to flipping may
resonance the substrate
invoke aby 180◦ . depolarization,
partial Moreover, surface roughness
i.e., a loss of or the presence
coherence of phase
of the plasmonic
and
resonance may invoke a partial depolarization, i.e., a loss of coherence of the
amplitude of the electric field [124]. These phenomena lie beyond the validity of the Jones model and phase and amplitude
of thefull
their electric field [124].
description These
requires phenomena
using lie beyondStokes
more complicated the validity
formalismof the[125].
Jones model and their full
description requires contribution
An important using more complicated
to angular-Stokes formalism [125].
and polarization-resolved SERS was made in [113],
An important contribution to angular- and polarization-resolved
probing adsorbate-covered gold NPs immobilized on a metallic substrate SERS was made
(Figure in Mathematical
17). [113], probing
adsorbate-covered gold NPs immobilized on a metallic substrate (Figure
background combining local field enhancement tensor, intrinsic molecular Raman tensors 17). Mathematical background
and
combining
rotation tensors has been developed in order to elucidate the surface coverage depending on has
local field enhancement tensor, intrinsic molecular Raman tensors and rotation tensors the
been developed
analyte in order
concentration. The to authors
elucidate thepointed
also surface out
coverage depending
the role on thetension
of the surface analyteofconcentration.
the solution
The
used,authors also pointed
preventing accessout of the
therole of the surface
analyte molecules tension of the solution
in between used, preventing
the nanospheres access
in the highof
the analyte molecules
concentration regime. in between the nanospheres in the high concentration regime.

Figure 17. Left: Scheme of the angular-resolved polarized SERS of aminobenzenothiol immobilized
Figure 17. Left: Scheme of the angular-resolved polarized SERS of aminobenzenothiol immobilized on
on gold nanospheres. Right: Observed depolarization ratio ( 𝜂) profile. The solid lines represent
gold nanospheres. Right: Observed depolarization ratio (η ) profile. The solid lines represent different
different 𝜂 profiles simulated for different 𝜃𝑐𝑎𝑝 angles. Adapted with permission from [113].
η profiles simulated for different θ angles. Adapted with permission from [113]. Copyright 2012
Copyright 2012 American ChemicalcapSociety.
American Chemical Society.
Another approach used for elucidation of polarization and angular dependences in SERS was
Another
employed approach
in our used for
recent work usingelucidation
MB adsorbedof polarization
on silver NR and angular
arrays [112]dependences in SERS was
(Figure 18). Although the
employed in ourwere
nanostructures recent work using MBanisotropic,
morphologically adsorbed onthe silver NR arrays
plasmonic [112] (Figure
properties 18). the
around Although the
excitation
nanostructures were morphologically anisotropic, the plasmonic properties around
wavelength used (532 nm) were found rather isotropic. Therefore, it could not be the main reason for the excitation
wavelength
the anisotropicused (532 nm)observed
behaviour were found rather
in the isotropic.
SERS Therefore,
experiments it couldinnot
(anisotropy thebedepolarization
the main reason for
ratios
the anisotropic
measured underbehaviour
selectedobserved in the SERS
configurations). experiments
Instead, difference(anisotropy in the depolarization
in the depolarization ratios ratios
after
measured under selected configurations). Instead, difference in the depolarization
rotating the sample by 90° was attributed to different refractive indices along different directions ratios after rotating
the sampleusing ◦ was attributed to different refractive indices along different directions (obtained
by 90spectral
(obtained ellipsometry) and explained within the framework of the surface selection
using spectral ellipsometry)
rules. This fact was explained andbyexplained within theperiodicity
slightly different frameworkalong of thedifferent
surface selection
directionsrules. This of
instead facta
was explained by slightly different periodicity along different directions instead
specific angle made between the NR axes and incident/collected radiation. These results enabled of a specific angle
made between
quantitative the NRof
analysis axesMB andRaman
incident/collected
tensor elements, radiation. These that
indicating results
theenabled
moleculesquantitative
adsorb
analysis of MB Raman tensor elements, indicating that the molecules adsorb
predominantly with the symmetry axis perpendicular to the surface. Interestingly, the depolarization predominantly with
the symmetry axis perpendicular to the surface. Interestingly, the
ratio computed using Equation (24) produced the value of 0.29, which was in-between the depolarization ratio computed
using Equationratio
depolarization (24) of
produced the value
MB molecules of 0.29, in
measured which was solution
a water in-between theand
(0.22) depolarization
the value ofratio 1/3,
of MB molecules measured in a water solution (0.22) and the value of 1/3,
predicted for a set of randomly-oriented hot-spots. It suggests that although the NRs were found predicted for a set of
randomly-oriented hot-spots. It suggests that although the NRs were found plasmonically
plasmonically isotropic, the role of hot-spots is still manifested in the value of the depolarization isotropic,
the role
ratio. Theofimportance
hot-spots isofstill
themanifested in the value of enhancement
Fresnel mechanism-based the depolarizationat the ratio. The importance
metal-dielectric of
interface
has also been highlighted in [126] where the origin of the additional SERS enhancement was
explained on account of optical interference effects, giving rise to an enhancement ~ 3 times higher
than in the case of direct illumination. The role of the surface-selection rules in SERS has also been
experimentally demonstrated in [117] using a Nile blue molecule. By comparing the EF of various
modes and relating these to their bare Raman tensors, the molecular orientation on a flat gold surface
Nanomaterials 2018, 8, 418 29 of 38

the Fresnel mechanism-based enhancement at the metal-dielectric interface has also been highlighted
in [126] where the origin of the additional SERS enhancement was explained on account of optical
interference effects, giving rise to an enhancement ~3 times higher than in the case of direct illumination.
The role of the surface-selection rules in SERS has also been experimentally demonstrated in [117]
using a Nile blue molecule. By comparing the EF of various modes and relating these to their bare
Raman tensors, the molecular orientation on a flat gold surface was inferred. Rich information
Nanomaterials 2018, 8, x FOR PEER REVIEW 28 of 37
about the Raman tensor components was obtained by SERS measurements as a function of the
incident angle and
authors reported theboth
EF upincident
to ~ 3, and scattered
but still polarizations.
retained the term SERS Theforauthors reported
simplicity. In thisthe EF the
case, up EF
to
~3,
wasbut still retained
caused solely by the the term SERSfactors
geometric for simplicity. In this in
and interpreted case,
termstheofEFthewas caused
Fresnel solely byThis
coefficients. the
geometric
concept was factors
extendedand interpreted in terms ofthe
in [127], investigating the Fresnel coefficients.
wavelength Thisindex
and refractive concept was extended
dependence of the
in [127], investigating
geometrically enhanced the wavelength
SERS signal. andInrefractive index dependence
this approach, of theoptical
the effective geometrically
constantsenhanced
of the
SERS signal. In metallic
nanostructured this approach, the effective
film were determined optical constants
by spectral of the nanostructured
ellipsometry measurements,metallic film
applying
were determined by spectral ellipsometry measurements, applying a homogeneous
a homogeneous uniaxial model. This approach highlights the need for application of the surface- uniaxial model.
This approach
selection ruleshighlights
not onlytheinneed the for application
case of smoothof the surface-selection
surfaces, but also rules
in thenotcase
only ofin the case of
plasmonic
smooth surfaces, but also in the case of plasmonic nanostructures.
nanostructures.

Figure 18.
Figure 18. Upper
Upper left: Scheme
left: of the
Scheme of experimental
the experimentalgeometry used for
geometry polarization-
used and angular-
for polarization- and
resolved SERS on silver nanorod arrays using methylene blue (MB) as the probe
angular-resolved SERS on silver nanorod arrays using methylene blue (MB) as the probe molecule. molecule. Lower
part: Depolarization ratio for
ratio profiles the 1628-cm
for the MB band−1forMB
different angular arrangements
−1
Lower part: Depolarization profiles 1628-cm band for different angular
(colour points) and their fit by the surface-selection rules with pseudo-refractive indices
arrangements (colour points) and their fit by the surface-selection rules with pseudo-refractive indices obtained
from ellipsometry
obtained measurements
from ellipsometry (lines). Excitation
measurements wavelength
(lines). Excitation was 532 nm.
wavelength wasUpper
532 nm. right: Real
Upper and
right:
imaginary part of pseudo-refractive index 𝑛
̃ =
Real and imaginary part of pseudo-refractive index n 𝑛 + 𝑖𝑘 of silver NR arrays and their variation with
e = n + ik of silver NR arrays and their variation both
angles
with as angles
both measured by standard
as measured by ellipsometry [112].
standard ellipsometry [112].

5.6. Plasmonic Anisotropy in Polarized Absorption and Emission and Their Influence on SERS
Anisotropic optical response of morphologically anisotropic plasmonic nanostructures is not
only restricted to SERS. Polarization dependence of plasmon-enhanced fluorescence on individual
Au NRs was reported in [99,128], producing maximum intensities with the excitation polarization
oriented along the NR axis. Intensity profile obtained when the angle between the excitation
polarization direction and the NR axis was varied was well fitted by a cosine squared function
Nanomaterials 2018, 8, 418 30 of 38

5.6. Plasmonic Anisotropy in Polarized Absorption and Emission and Their Influence on SERS
Anisotropic optical response of morphologically anisotropic plasmonic nanostructures is not only
restricted to SERS. Polarization dependence of plasmon-enhanced fluorescence on individual Au NRs
was reported in [99,128], producing maximum intensities with the excitation polarization oriented
along the NR axis. Intensity profile obtained when the angle between the excitation polarization
direction and the NR axis was varied was well fitted by a cosine squared function together with an
exponential decay. Very similar trends were observed in [129] using Au NRs of different aspect ratios,
or from a single Ag nanowire in [130]. Contradictory results appeared in [101].
Other instances where plasmonic anisotropy of metallic substrates is manifested include
absorbance or reflectance measurements. Although the SERS properties of plasmonic nanostructures
are very often derived from absorption or reflection measurements, this relationship works well only
in the average SERS regime where the role of hot-spots is rather negligible, such as regular arrays
of isotropic silver nanoislands [131]. In the case of more complicated (typically anisotropic and/or
hot-spot dominated) plasmonic nanostructures, little correlation between these two phenomena (“near
field” and “far field” properties of SERS substrates) was found [132,133]. Anisotropic SERS response
surely is, in a more or less straightforward way, manifested in polarization-resolved absorbance
or reflectance spectra with the difference in absorbance for different light polarization often even
more pronounced than in the case of SERS. However, the exact relationship between the polarization
providing the highest SERS enhancement and the polarization exhibiting the highest absorbance
(or transmittance/reflectance) remains highly non-trivial.
In the case of NR arrays, absorbance typically features a sharp transverse peak (for Ag generally
below 400 nm) and a red-shifted broad longitudinal peak which spill one into another when rotating
the polarization direction [97,112,134]. For aligned silver NRs in [98], the magnitude of the SERS
enhancement and of absorbance of the same polarization were almost a perfect match (Figure 19)
with the biggest response provided by polarization along the NRs. Counterintuitively, in the case
of metallic grating structures [94], light polarization exhibiting the biggest transmission was also
found to dominate in the SERS spectrum over perpendicular polarization for which the transmitted
intensity was lower. Such trends were also found for silver NR arrays in [97], although a bigger
SERS response was found for polarization perpendicular to the adjacent NRs. In a more elaborate
study [135], the authors investigated changes in absorbance and its polarization characteristics as a
function of the NR height, concluding that smaller NRs are rather isotropic. With the increase of the
NR size (>1000 nm), absorbance splits into at least two plasmon resonant modes.
The works mentioned above suggest that in the case of metallic NRs, the basic features
or polarization-resolved absorbance is influenced by the specific preparation procedure of a
given nanostructure less than polarization-resolved SERS. However, it has to be emphasized that
polarization-resolved absorbance has been studied less intensively in literature than polarization-resolved
SERS and many aspects regarding their mutual correspondence are yet to be fully resolved.
In a study devoted to angle resolved SERS on metallic nanostructured plasmonic crystals [136],
a correlation between plasmon-related absorption in the reflectivity and the Raman enhancement with
the varied angle of incidence was found. The need for absorbance or reflectance spectra may also be
supplemented by ellipsometry measurements [112,116,127]. Although the plasmonic properties of
silver NRs in [112] were found rather isotropic using the 532 nm wavelength, a clear distinction between
the extinction profiles (computed using the ellipsometry characteristics) after rotating the sample by
90◦ was observed. This distinction was attributed to different periodicity in different directions as
the NRs used were almost perfectly aligned in one direction, but still exhibiting slight deviations
from the perfectly-ordered state in the perpendicular direction. This slight inhomogeneity was
reflected in the inhomogeneously broadened longitudinal plasmon peak and was found responsible
for orientation-dependent depolarization ratios of the adsorbed molecule, although the extinction
profiles suggested rather plasmonic homogeneity. In [116], the SERS response for two different
 enhancement and of absorbance of the same polarization were almost a perfect match (Figure 19)
with the biggest response provided by polarization along the NRs. Counterintuitively, in the case of
metallic grating structures [94], light polarization exhibiting the biggest transmission was also found
to dominate in the SERS spectrum over perpendicular polarization for which the transmitted
intensity was
Nanomaterials 2018,lower.
8, 418 Such trends were also found for silver NR arrays in [97], although a bigger SERS
31 of 38
response was found for polarization perpendicular to the adjacent NRs. In a more elaborate study
[135], the authors investigated changes in absorbance and its polarization characteristics as a function
wavelengths was in
of the NR height, agreement
concluding with
that the imaginary
smaller partisotropic.
NRs are rather of the effective
With thedielectric
increase offunction of the
the NR size
Ag(>1000
film. nm), absorbance splits into at least two plasmon resonant modes.

Figure
Figure 19.19.Left:
Left:The
Theabsorbance
absorbance spectra
spectra of
of the
the Ag
Ag NR
NRarrays
arraystuned
tunedtotothe
theRaman
Ramanspectroscopy at at
spectroscopy
excitation wavelength of 785 nm (1.58 eV). The incident light was either p or s polarized, where an
excitation wavelength of 785 nm (1.58 eV). The incident light was either p or s polarized, where
electric field vibrates parallel or perpendicular to the deposition plane, respectively, as shown
an electric field vibrates parallel or perpendicular to the deposition plane, respectively, as shown
schematically in the upper right box. Right: Polarization dependence of the Raman spectra (a) and
schematically in the upper right box. Right: Polarization dependence of the Raman spectra (a) and the
the peak intensity at 1014 cm−1 (b). The polarization dependence of the absorbance at the wavelength
peak intensity at 1014 cm−1 (b). The polarization dependence of the absorbance at the wavelength of
of 785 nm is also shown (b). Adapted with permission from [98].
785 nm is also shown (b). Adapted with permission from [98].
The works mentioned above suggest that in the case of metallic NRs, the basic features or
6. polarization-resolved
Conclusions absorbance is influenced by the specific preparation procedure of a given
nanostructure lesswethan
In this review, havepolarization-resolved SERS. However,
focused mainly on polarization it has to be emphasized
and directional/angular that
characteristics
polarization-resolved absorbance has been studied less intensively in literature than
occurring in Raman spectroscopy and surface-enhanced Raman spectroscopy (SERS) on anisotropic polarization-
resolved
metallic SERS and manyWe
nanostructures. aspects
have regarding
summarizedtheirbasic
mutual correspondence
formulas for Raman areintensities
yet to be fully
as aresolved.
function of
light polarization and angular arrangement used for both liquid samples as well as crystals. We have
demonstrated that polarization- and angular-resolved measurements allow one to retrieve information
on the (orientation-averaged) Raman tensor components, which are useful for symmetry determination
and/or obtaining information regarding the molecular orientation. The complexity of this problem
rises when molecules are adsorbed in the vicinity of nanostructured metallic surfaces, where not only
coupling between light and molecules, but also (and often mainly) between light and the anisotropic
NP arrays determine the total intensity profile. The observed spectral pattern results from interplay
between the material used, size and shape of the nanoobjects, their spatial arrangement as well as the
excitation wavelength used. Periodicity of the NP array is also deciding since it is manifested in the
NP optical properties, which in turn influences the SERS signal. Since details of the NP fabrication are
often very subtle, this process is very difficult to reproduce and therefore the exact results often vary in
literature. It is one of the reasons why all aspects of polarization- and angular resolved measurements
are not yet fully understood and remain a challenge for future research. It is expected that study of
the polarization and angular properties of molecules on plasmonic nanostructures will contribute to a
deeper theoretical understanding of the enhancement mechanism. It will also provide a theoretical
background for the design of plasmon-based molecular sensors for maximum signal enhancement.

Author Contributions: M.S. surveyed the literature, prepared the figures and wrote the manuscript.
M.P. participated in editing the article and helping with final version of the manuscript.
Funding: This research was funded by Charles University research project GA UK No. 572216, Charles University
Research Centre program UNCE/SCI/010, and the Czech Science Foundation No. 18-10897S.
Nanomaterials 2018, 8, 418 32 of 38

Acknowledgments: The authors thank Ondřej Kylián from Department of Macromolecular Physics, Charles
University, for fabrication of SERS-active substrates.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Allemand, C.D. Depolarization Ratio Measurements in Raman Spectroscopy. Appl. Spectrosc. 1970, 24,
348–353. [CrossRef]
2. Dawson, P. Polarization Measurements in Raman Spectroscopy. Spectrochim. Acta Part A-Mol. Spectrosc. 1972,
28, 715–723. [CrossRef]
3. Porto, S.P.S. Angular Dependence and Depolarization Ratio of the Raman Effect. J. Opt. Soc. Am. 1966, 56,
1585–1589. [CrossRef]
4. Greenler, R.G. Infrared Study of Adsorbed Molecules on Metal Surfaces by Reflection Techniques.
J. Chem. Phys. 1966, 44, 310–315. [CrossRef]
5. Greenler, R.G.; Slager, T.L. Method for Obtaining Raman-Spectrum of a Thin-Film on a Metal Surface.
Spectrochim. Acta Part A-Mol. Spectrosc. 1973, A 29, 193–201. [CrossRef]
6. Pearce, H.A.; Sheppard, N. Possible Importance of a Metal-Surface Selection Rule in Interpretation of
Infrared-Spectra of Molecules Adsorbed on Particulate Metals—Infrared-Spectra from Ethylene Chemisorbed
on Silica-Supported Metal-Catalysts. Surf. Sci. 1976, 59, 205–209. [CrossRef]
7. Murphy, C.J.; Sau, T.K.; Gole, A.M.; Orendorff, C.J.; Gao, J.; Gou, L.; Hunyadi, S.E.; Li, T. Anisotropic Metal
Nanoparticles: Synthesis, Assembly, and Optical Applications. J. Phys. Chem. B 2005, 109, 13857–13870.
[CrossRef] [PubMed]
8. Li, N.; Zhao, P.X.; Astruc, D. Anisotropic Gold Nanoparticles: Synthesis, Properties, Applications,
and Toxicity. Angew. Chem. Int. Ed. 2014, 53, 1756–1789. [CrossRef] [PubMed]
9. Thorkelsson, K.; Bai, P.; Xu, T. Self-assembly and applications of anisotropic nanomaterials: A review.
Nano Today 2015, 10, 48–66. [CrossRef]
10. Paramasivam, G.; Kayambu, N.; Rabel, A.M.; Sundramoorthy, A.K.; Sundaramurthy, A. Anisotropic noble
metal nanoparticles: Synthesis, surface functionalization and applications in biosensing, bioimaging, drug
delivery and theranostics. Acta Biomater. 2017, 49, 45–65. [CrossRef] [PubMed]
11. Barron, L.D. The development of biomolecular Raman optical activity spectroscopy. Biomed. Spectrosc. Imaging
2015, 4, 223–253. [CrossRef]
12. Born, M.; Wolf, E. Principles of Optics, 7th ed.; Cambridge University Press: Cambridge, UK, 1999;
ISBN 0-521-642221.
13. Le Ru, E.C.; Etchegoin, P.G. Principles of Surface-Enhanced Raman Spectroscopy and Related Plasmonic Effects;
Elsevier: Amsterdam, The Netherlands, 2009; ISBN 978-0-444-52779-0.
14. Procházka, M. Surface-Enhanced Raman Spectroscopy: Bioanalytical, Biomolecular and Medical Applications;
Springer: Cham, Switzerland, 2016; ISBN 978-3-319-23990-3.
15. Moore, T.J.; Moody, A.S.; Payne, T.D.; Sarabia, G.M.; Daniel, A.R.; Sharma, B. In Vitro and In Vivo SERS
Biosensing for Disease Diagnosis. Biosensors 2018, 8, 46. [CrossRef] [PubMed]
16. Štolcová, L.; Peksa, V.; Proška, J.; Procházka, M. Gold film over very small (107 nm) spheres as efficient
substrate for sensitive and reproducible surface-enhanced Raman scattering (SERS) detection of biologically
important molecules. J. Raman Spectrosc. 2017, 49, 499–505. [CrossRef]
17. Zhu, Q.; Yu, X.; Wu, Z.; Lu, F.; Yuan, Y. Antipsychotic drug poisoning monitoring of clozapine in urine by
using coffee ring effect based surface-enhanced Raman spectroscopy. Anal. Chim. Acta 2018, 1014, 64–70.
[CrossRef] [PubMed]
18. Mukanova, Z.; Gudun, K.; Elemessova, Z.; Khamkhash, L.; Ralchenko, E.; Bukasov, R. Detection of
Paracetamol in Water and Urea in Artifical Urine with Gold Nanoparticle@Al Foil Cost-efficient SERS
Substrate. Anal. Sci. 2018, 34, 183–187. [CrossRef] [PubMed]
19. Peksa, V.; Jahn, M.; Štolcová, L.; Schulz, V.; Proška, J.; Procházka, M.; Weber, K.; Cialla-May, D.; Popp, J.
Quantitative SERS Analysis of Azorubine (E 122) in Sweet Drinks. Anal. Chem. 2015, 87, 2840–2844.
[CrossRef] [PubMed]
Nanomaterials 2018, 8, 418 33 of 38

20. Ai, Y.-J.; Liang, P.; Wu, Y.-X.; Dong, Q.-M.; Li, J.-B.; Bai, Y.; Xu, B.-J.; Yu, Z.; Ni, D. Rapid qualitative and
quantitative determination of food colorants by both Raman spectra and Surface-enhanced Raman Scattering
(SERS). Food Chem. 2018, 241, 427–433. [CrossRef] [PubMed]
21. Hakonen, A.; Wang, F.C.; Andersson, P.O.; Wingfors, H.; Rindzevicius, T.; Schmidt, M.S.; Soma, V.R.; Xu, S.;
Li, Y.Q.; Boisen, A.; et al. Hand-Held Femtogram Detection of Hazardous Picric Acid with Hydrophobic Ag
Nanopillar SERS Substrates and Mechanism of Elasto-Capillarity. ACS Sens. 2017, 2, 198–202. [CrossRef]
[PubMed]
22. Liyanage, T.; Rael, A.; Schaffer, S.; Zaidi, S.; Goodpaster, J.V.; Sardar, R. Fabrication of a self-assembled and
flexible SERS nanosesor for explosive detection at parts-per-quadrillion levels from fingerprints. Analyst
2018, 143, 2012–2022. [CrossRef] [PubMed]
23. Cheng, H.-W.; Huan, S.-Y.; Yu, R.-Q. Nanoparticle-based substrates for surface-enhanced Raman scattering
detection of bacterial spores. Analyst 2012, 137, 3601–3608. [CrossRef] [PubMed]
24. Morelli, L.; Zor, K.; Jendersen, C.B.; Rindzevicius, T.; Schmidt, M.S.; Nielsen, A.T.; Boisent, A. Surface
Enhanced Raman Scattering for Quantification of p-Coumaric Acid Produced by Escherichia coli. Anal. Chem.
2017, 89, 3981–3987. [CrossRef] [PubMed]
25. Sanchez-Purra, M.; Carre-Camps, M.; de Puig, H.; Bosch, I.; Gehrke, L.; Hamad-Schifferli, K.
Surface-Enhanced Raman Spectroscopy-Based Sandwich Immunoassays for Multiplexed Detection of Zika
and Dengue Viral Biomarkers. ACS Infect. Dis. 2017, 3, 767–776. [CrossRef] [PubMed]
26. Placzek, G. Rayleigh-Streuung und Raman-Effekt. In Handbuch der Radiologie; Marx, E., Ed.; Academische
Verlag: Leipzig, Germany, 1934; Volume 6, pp. 205–374.
27. Long, D.A. The Raman Effect: A Unified Treatment of the Theory of Raman Scattering by Molecules; John Wiley &
Sons Ltd.: Chichester, UK, 2002; ISBN 0-471-49028-8.
28. Smith, E.; Dent, G. Modern Raman Spectroscopy—A Practical Approach; John Wiley & Sons Ltd.: Chichester,
UK, 2005; ISBN 0-471-49668-5.
29. Cotton, F.A. Chemical Applications of Group Theory, 3rd ed.; John Wiley & Sons Ltd.: Chichester, UK, 1990;
ISBN 0-471-51094-7.
30. Ferraro, J.R.; Nakamoto, K.; Brown, C.W. Introductory Raman Spectroscopy, 2nd ed.; Elsevier: New York, NY,
USA, 2003; ISBN 978-0-12-254105-6.
31. Gucciardi, P.G.; de la Chapelle, M.L.; Lidgi-Guigui, N. Handbook of Enhanced Spectroscopy; Pan Stanford
Publishing Pte. Ltd.: Singapore, 2016; ISBN 978-981-4613-32-3.
32. Strommen, D.P. Specific Values of the Depolarization Ratio in Raman-Spectroscopy—Their Origins and
Significance. J. Chem. Educ. 1992, 69, 803–807. [CrossRef]
33. Damen, T.C.; Porto, S.P.S.; Tell, B. Raman Effect in Zinc Oxide. Phys. Rev. 1966, 142, 570–574. [CrossRef]
34. Moskovits, M. Surface Selection-Rules. J. Chem. Phys. 1982, 77, 4408–4416. [CrossRef]
35. Moskovits, M. Surface-Enhanced Spectroscopy. Rev. Mod. Phys. 1985, 57, 783–826. [CrossRef]
36. Creighton, J.A. Surface Raman Electromagnetic Enhancement Factors for Molecules at the Surface of Small
Isolated Metal Spheres—The Determination of Adsorbate Orientation from Sers Relative Intensities. Surf. Sci.
1983, 124, 209–219. [CrossRef]
37. Aroca, R. Surface-Enhanced Vibrational Spectroscopy; John Wiley & Sons Ltd.: Chichester, UK, 2006;
ISBN 0-471-60731-2.
38. Johnson, P.B.; Christy, R.W. Optical Constants of Noble Metals. Phys. Rev. B 1972, 6, 4370–4379. [CrossRef]
39. Foti, A.; D’Andrea, C.; Messina, E.; Irrera, A.; Marago, O.M.; Fazio, B.; Gucciardi, P.G. On the SERS
depolarization ratio. Nanospectroscopy 2015, 1, 26–62. [CrossRef]
40. Etchegoin, P.G.; Galloway, C.; Le Ru, E.C. Polarization-dependent effects in surface-enhanced Raman
scattering (SERS). Phys. Chem. Chem. Phys. 2006, 8, 2624–2628. [CrossRef] [PubMed]
41. Le Ru, E.C.; Meyer, M.; Blackie, E.; Etchegoin, P.G. Advanced aspects of electromagnetic SERS enhancement
factors at a hot spot. J. Raman Spectrosc. 2008, 39, 1127–1134. [CrossRef]
42. Moskovits, M. Surface-enhanced Raman spectroscopy: A brief retrospective. J. Raman Spectrosc. 2005, 36,
485–496. [CrossRef]
43. Fazio, B.; D’Andrea, C.; Bonaccorso, F.; Irrera, A.; Calogero, G.; Vasi, C.; Gucciardi, P.G.; Allegrini, M.;
Toma, A.; Chiappe, D.; et al. Re-radiation Enhancement in Polarized Surface-Enhanced Resonant Raman
Scattering of Randomly Oriented Molecules on Self-Organized Gold Nanowires. ACS Nano 2011, 5,
5945–5956. [CrossRef] [PubMed]
Nanomaterials 2018, 8, 418 34 of 38

44. Wang, D.S.; Chew, H.; Kerker, M. Enhanced Raman-Scattering at the Surface (SERS) of a Spherical-Particle.
Appl. Opt. 1980, 19, 2256–2257. [CrossRef] [PubMed]
45. Moskovits, M.; Suh, J.S. Surface Selection-Rules for Surface-Enhanced Raman Spectroscopy-Calculation and
Application to the Surface-Enhanced Raman Spectrum of Phthalazine on Silver. J. Phys. Chem. 1984, 88,
5526–5530. [CrossRef]
46. Macdonald, I.D.G.; Smith, W.E. Orientation of Cytochrome c Adsorbed on a Citrate-Reduced Silver Colloid
Surface. Langmuir 1996, 12, 706–713. [CrossRef]
47. Gao, X.; Davies, J.P.; Weaver, M.J.A. Test of Surface Selection-Rules for Surface-Enhanced Raman-Scattering:
The Orientation of Adsorbed Benzene and Monosubstituted Benzenes on Gold. J. Phys. Chem. 1990, 94,
6858–6864. [CrossRef]
48. Carron, K.T.; Hurley, L.G. Axial and Azimuthal Angle Determination with Surface-Enhanced
Raman-Spectroscopy: Thiophenol on Copper, Silver, and Gold Metal-Surfaces. J. Phys. Chem. 1991, 95,
9979–9984. [CrossRef]
49. Barber, P.W.; Chang, R.K.; Massoudi, H. Surface-Enhanced Electric Intensities on Large Silver Spheroids.
Phys. Rev. Lett. 1983, 50, 997–1000. [CrossRef]
50. Aravind, P.K.; Nitzan, A.; Metiu, H. The Interaction between Electromagnetic Resonances and Its Role in
Spectroscopic Studies of Molecules Adsorbed on Colloidal Particles or Metal Spheres. Surf. Sci. 1981, 110,
189–204. [CrossRef]
51. Aravind, P.K.; Metiu, H. The Effects of the Interaction between Resonances in the Electromagnetic Response
of a Sphere-Plane Structure; Applications to Surface Enhanced Spectroscopy. Surf. Sci. 1983, 124, 506–528.
[CrossRef]
52. Xu, H.; Bjerneld, E.J.; Käll, M.; Borjesson, L. Spectroscopy of single hemoglobin molecules by surface
enhanced Raman scattering. Phys. Rev. Lett. 1999, 83, 4357–4360. [CrossRef]
53. Xu, H.; Aizpurua, J.; Käll, M.; Apell, P. Electromagnetic contributions to single-molecule sensitivity in
surface-enhanced Raman scattering. Phys. Rev. E 2000, 62, 4318–4324. [CrossRef]
54. Xu, H.; Käll, M. Polarization-Dependent Surface-Enhanced Raman Spectroscopy of Isolated Silver
Nanoaggregates. ChemPhysChem 2003, 4, 1001–1005. [CrossRef] [PubMed]
55. Le Ru, E.C.; Blackie, E.; Meyer, M.; Etchegoin, P.G. Surface Enhanced Raman Scattering Enhancement Factors:
A Comprehensive Study. J. Phys. Chem. C 2007, 111, 13794–13803. [CrossRef]
56. Camden, J.P.; Dieringer, J.A.; Wang, Y.; Masiello, D.J.; Marks, L.D.; Schatz, G.C.; Van Duyne, R.P. Probing the
Structure of Single-Molecule Surface-Enhanced Raman Scattering Hot Spots. J. Am. Chem. Soc. 2008, 130,
12616–12617. [CrossRef] [PubMed]
57. Hakonen, A.; Svedendahl, M.; Ogier, R.; Yang, Z.-J.; Lodewijks, K.; Verre, R.; Shegai, T.; Andersson, P.O.;
Käll, M. Dimer-on-mirror SERS substrates with attogram sensitivity fabricated by colloidal lithography.
Nanoscale 2015, 7, 9405–9410. [CrossRef] [PubMed]
58. 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, 10903–10910. [CrossRef] [PubMed]
59. Zang, Y.; Shi, H.; Huang, Y.; Zeng, X.; Pan, L.; Wang, S.; Wen, W. SERS polarization dependence of Ag
nanorice dimer on metal and dielectric film. Chem. Phys. Lett. 2017, 684, 373–377. [CrossRef]
60. Tao, A.R.; Yang, P.D. Polarized surface-enhanced Raman spectroscopy on coupled metallic nanowires. J. Phys.
Chem. B 2005, 109, 15687–15690. [CrossRef] [PubMed]
61. Le Ru, E.C.; Etchegoin, P.G.; Meyer, M. Enhancement factor distribution around a single surface-enhanced
Raman scattering hot spot and its relation to single molecule detection. J. Chem. Phys. 2006, 125, 204701.
[CrossRef] [PubMed]
62. Moskovits, M. Persistent misconceptions regarding SERS. Phys. Chem. Chem. Phys. 2013, 15, 5301–5311.
[CrossRef] [PubMed]
63. Leopold, N.; Lendl, B. A New Method for Fast Preparation of Highly Surface-Enhanced Raman Scattering
(SERS) Active Silver Colloids at Room Temperature by Reduction of Silver Nitrate with Hydroxylamine
Hydrochloride. J. Phys. Chem. B 2003, 107, 5723–5727. [CrossRef]
64. Roy, S.D.; Ghosh, M.; Chowdhury, J. Adsorptive parameters and influence of hot geometries on the SER(R)S
spectra of methylene blue molecules adsorbed on gold nanocolloidal particles. J. Raman Spectrosc. 2015, 46,
451–461. [CrossRef]
Nanomaterials 2018, 8, 418 35 of 38

65. Malinowski, E.R. FactorAnalysis in Chemistry, 3rd ed.; John Wiley & Sons: New York, NY, USA, 2002;
ISBN 0-471-13479-1.
66. Zhong, L.; Hu, Y.J.; Xing, D.; Gu, H.M. Adsorption of Methylene Blue on Colloidal Silver-A Surface-Enhanced
Raman Spectroscopy Study Combined with Density Functional Theory Calculations. Spectrosc. Spectr. Anal.
2010, 30, 90–94. [CrossRef]
67. Timmermans, F.J.; Lenferink, A.T.M.; van Wolferen, H.A.G.M.; Otto, C. Correlative SEM SERS for quantitative
analysis of dimer nanoparticles. Analyst 2016, 141, 6455–6462. [CrossRef] [PubMed]
68. Sergiienko, S.; Moor, K.; Gudun, K.; Yelemessova, Z.; Bukasov, R. Nanoparticle-nanoparticle vs.
nanoparticle-substrate hot spot contributes to the SERS signal: Studying Raman labelled monomers, dimers
and trimers. Phys. Chem. Chem. Phys. 2017, 19, 4478–4487. [CrossRef] [PubMed]
69. Xiang, Q.; Li, Z.; Zheng, M.; Liu, Q.; Chen, Y.; Yang, L.; Jiang, T.; Duan, H. Sensitive SERS detection at the
single-particle level based on nanometer-separated mushroom-shaped plasmonic dimers. Nanotechnology
2018, 29, 105301. [CrossRef] [PubMed]
70. Zhang, Q.; Li, G.-C.; Lo, T.W.; Lei, D.Y. Polarization-resolved optical response of plasmonic particle-on-film
nanocavities. J. Opt. 2018, 20, 024010. [CrossRef]
71. Cao, Z.; He, Y.; Cheng, Y.; Zhao, J.; Li, G.; Gong, Q.; Lu, G. Nano-gap between a gold tip and nanorod for
polarization dependent surface enhanced Raman scattering. Appl. Phys. Lett. 2016, 109, 233103. [CrossRef]
72. Haidar, I.; Levi, G.; Mouton, L.; Aubard, J.; Grand, J.; Lau-Truong, S.; Neuville, D.R.; Felidj, N.;
Boubekeur-Lecaque, L. Highly stable silica-coated gold nanorods dimers for solution-based SERS. Phys. Chem.
Chem. Phys. 2016, 18, 32272–32280. [CrossRef] [PubMed]
73. Tang, L.; Li, S.; Han, F.; Liu, L.; Xu, L.; Ma, W.; Kuang, H.; Li, A.; Wang, L.; Xu, C. SERS-active Au@Ag
nanorod dimers for ultrasensitive dopamine detection. Biosens. Bioelectron. 2015, 71, 7–12. [CrossRef]
[PubMed]
74. Aswathy, V.G.; Patra, P.P.; Kumar, G.V.P. Geometry-dependent anti-Stokes SERS radiation patterns from gold
nanorod dimers. J. Opt. 2015, 17, 114011. [CrossRef]
75. Maune, H.T.; Han, S.-P.; Barish, R.D.; Bockrath, M.; Goddard, W.A., III; Rothemund, P.W.K.; Winfree, E.
Self-assembly of carbon nanotubes into two-dimensional geometries using DNA origami templates.
Nat. Nanotechnol. 2010, 5, 61–66. [CrossRef] [PubMed]
76. Thacker, V.V.; Herrmann, L.O.; Sigle, D.O.; Zhang, T.; Liedl, T.; Baumberg, J.J.; Keyser, U.F. DNA origami
based assembly of gold nanoparticle dimers for surface-enhanced Raman scattering. Nat. Commun. 2014, 5,
3448. [CrossRef] [PubMed]
77. Prinz, J.; Heck, C.; Ellerik, L.; Merk, V.; Bald, I. DNA origami based Au-Ag-core-shell nanoparticle dimers
with single-molecule SERS sensitivity. Nanoscale 2016, 8, 5612–5620. [CrossRef] [PubMed]
78. Pal, A.; Gohil, S.; Ghosh, S.; Ayyub, P. Polarization Selection Rules for Surface-Enhanced Raman Scattering
from Anisotropic Microstructured Surfaces. J. Phys. Chem. C 2012, 116, 18504–18507. [CrossRef]
79. Le Ru, E.C.; Grand, J.; Félidj, N.; Aubard, J.; Lévi, G.; Hohenau, A.; Krenn, J.R.; Blackie, E.; Etchegoin, P.G.
Experimental Verification of the SERS Electromagnetic Model beyond the |E|4 Approximation: Polarization
Effects. J. Phys. Chem. C 2012, 116, 18504–18507. [CrossRef]
80. Shegai, T.; Li, Z.; Dadosh, T.; Zhang, Z.; Xu, H.; Haran, G. Managing light polarization via plasmon-molecule
interactions within an asymmetric metal nanoparticle trimer. Proc. Natl. Acad. Sci. USA 2008, 105, 16448–16453.
[CrossRef] [PubMed]
81. Feng, C.; Zhao, Y.; Jiang, Y.J. Periodic array of regular Ag nanoparticle trimers: A reliable
polarization-independent surface-enhanced Raman spectroscopy substrate. RSC Adv. 2016, 6, 83273–83279.
[CrossRef]
82. Shegai, T.; Brian, B.; Miljkovic, V.D.; Käll, M. Angular Distribution of Surface-Enhanced Raman Scattering
from Individual Au Nanoparticle Aggregates. ACS Nano 2011, 5, 2036–2041. [CrossRef] [PubMed]
83. Quinten, M. Local fields close to the surface of nanoparticles and aggregates of nanoparticles. Appl. Phys.
B-Lasers Opt. 2001, 73, 245–255. [CrossRef]
84. Dai, Z.; Xiao, X.; Wu, W.; Liao, L.; Mei, F.; Yu, X.; Guo, S.; Ying, J.; Ren, F.; Jiang, C. Side-to-side alignment
of gold nanorods with polarization-free characteristic for highly reproducible surface enhanced Raman
scattering. Appl. Phys. Lett. 2014, 105, 211902. [CrossRef]
85. Shegai, T.O.; Haran, G. Probing the Raman Scattering Tensors of Individual Molecules. J. Phys. Chem. B 2006,
110, 2459–2461. [CrossRef] [PubMed]
Nanomaterials 2018, 8, 418 36 of 38

86. Shen, H.; Rouxel, J.; Guillot, N.; de la Chapelle, M.L.; Toury, T. Light polarization properties of three fold
symmetry gold nanoparticles: Model and experiments. C. R. Phys. 2012, 13, 830–836. [CrossRef]
87. De la Chapelle, M.L.; Guillot, N.; Frémaux, B.; Shen, H.; Toury, T. Novel Apolar Plasmonic Nanostructures
with Extended Optical Tunability for Sensing Applications. Plasmonics 2013, 8, 475–480. [CrossRef]
88. Reguera, J.; Langer, J.; de Aberasturi, D.J.; Liz-Marzán, L.M. Anisotropic metal nanoparticles for surface
enhanced Raman scattering. Chem. Soc. Rev. 2017, 46, 3866–3885. [CrossRef] [PubMed]
89. McLellan, J.M.; Li, Z.Y.; Siekkinen, A.R.; Xia, Y. The SERS Activity of a Supported Ag Nanocube Strongly
Depends on Its Orientation Relative to Laser Polarization. Nano Lett. 2007, 7, 1013–1017. [CrossRef]
[PubMed]
90. Zhou, Y.; Zhou, X.; Park, D.J.; Torabi, K.; Brown, K.A.; Jones, M.R.; Zhang, C.; Schatz, G.C.; Mirkin, C.A.
Shape-Selective Deposition and Assembly of Anisotropic Nanoparticles. Nano Lett. 2014, 14, 2157–2161.
[CrossRef] [PubMed]
91. Prezgot, D.; Ianoul, A. Probing the Anisotropy of SERS Enhancement with Spatially Separated Plasmonic
Modes in Strongly Coupled Silver Nanocubes on a Dielectric Substrate. J. Phys. Chem. C 2015, 119, 3293–3301.
[CrossRef]
92. Luo, W.; van der Veer, W.; Chu, P.; Mills, D.L.; Penner, R.M.; Hemminger, J.C. Polarization-Dependent Surface
Enhanced Raman Scattering from Silver 1D Nanoparticle Arrays. J. Phys. Chem. C 2008, 112, 11609–11613.
[CrossRef]
93. Hossain, M.K.; Huang, G.G.; Tanaka, Y.; Kaneko, T.; Ozaki, Y. Anisotropic gold nanoassembly: A study
on polarization-dependent and polarization-selective surface-enhanced Raman scattering. Phys. Chem.
Chem. Phys. 2015, 17, 4268–4276. [CrossRef] [PubMed]
94. Hong, K.Y.; Brolo, A.G. Polarization-dependent surface-enhanced Raman scattering (SERS) from microarrays.
Anal. Chim. Acta 2017, 972, 73–80. [CrossRef] [PubMed]
95. Farcau, C.; Vallée, R.A.L.; Boca, S.; Astilean, S. Polarized SERS on linear arrays of silver half-shells: SERS
re-radiation modulated by local density of optical states. J. Opt. 2015, 17, 114017. [CrossRef]
96. Liu, Y.; Fan, J.; Zhao, Y.P.; Shanmukh, S.; Dluhy, R.A. Angle dependent surface enhanced Raman scattering
obtained from a Ag nanorod array substrate. Appl. Phys. Lett. 2006, 89, 173134. [CrossRef]
97. Zhao, Y.P.; Chaney, S.B.; Shanmukh, S.; Dluhy, R.A. Polarized Surface Enhanced Raman and Absorbance
Spectra of Aligned Silver Nanorod Arrays. J. Phys. Chem. B 2006, 110, 3153–3157. [CrossRef] [PubMed]
98. Suzuki, M.; Maekita, W.; Wada, Y.; Nakajima, K.; Kimura, K.; Fukuota, T.; Mori, Y. In-line aligned and
bottom-up Ag nanorods for surface-enhanced Raman spectroscopy. Appl. Phys. Lett. 2006, 88, 203121.
[CrossRef]
99. Ming, T.; Zhao, L.; Yang, Z.; Chen, H.; Sun, L.; Wang, J.; Yan, C. Polarization Dependence of
Plasmon-Enhanced Fluorescence on Single Gold Nanorods. Nano Lett. 2009, 9, 3896–3903. [CrossRef]
[PubMed]
100. Jeong, D.H.; Zhang, Y.X.; Moskovits, M. Polarized Surface Enhanced Raman Scattering from Aligned Silver
Nanowire Rafts. J. Phys. Chem. B 2004, 108, 12724–12728. [CrossRef]
101. Zhou, W.; Hu, A.; Bai, S.; Ma, Y.; Bridges, D. Anisotropic optical properties of large-scale aligned silver
nanowire films via controlled coffee ring effects. RSC Adv. 2015, 5, 39103–39109. [CrossRef]
102. Chen, W.; Shi, H.; Wan, F.; Wang, P.; Gu, Z.; Li, W.; Ke, L.; Huang, Y. Substrate influence on the polarization
dependence of SERS in crossed metal nanowires. J. Mater. Chem. C 2017, 5, 7028–7034. [CrossRef]
103. Yoon, I.; Kang, T.; Choi, W.; Kim, J.; Yoo, Y.; Joo, S.-W.; Park, Q.-H.; Ihee, H.; Kim, B. Single Nanowire on a
Film as an Efficient SERS-Active Platform. J. Am. Chem. Soc. 2009, 131, 758–762. [CrossRef] [PubMed]
104. Chen, L.; Zhai, T.; Zhang, X.; Unger, C.; Koch, J.; Chichkov, B.N.; Klar, P.J. Polarization-dependent SERS
effects of laser-generated sub-100 nm antenna structures. Nanotechnology 2014, 25, 265302. [CrossRef]
[PubMed]
105. Kessentini, S.; Barchiesi, D.; D’Andrea, C.; Toma, A.; Guillot, N.; Di Fabrizio, E.; Fazio, B.; Marago, O.M.;
Gucciardi, P.G.; de la Chapelle, M.L. Gold Dimer Nanoantenna with Slanted Gap for Tunable LSPR and
Improved SERS. J. Phys. Chem. C 2014, 118, 3209–3219. [CrossRef]
106. Liu, K.K.; Tadepalli, S.; Kumari, G.; Banerjee, P.; Tian, L.; Jain, P.K.; Singamaneni, S. Polarization-Dependent
Surface-Enhanced Raman Scattering Activity of Anisotropic Plasmonic Nanorattles. J. Phys. Chem. C 2016,
120, 16899–16906. [CrossRef]
Nanomaterials 2018, 8, 418 37 of 38

107. Zhang, X.; Wang, T.; Li, Y.; Fu, Y.; Guo, L. Optimally designed gold nanorattles with strong built-in hotspots
and weak polarization dependence. Nanotechnology 2017, 28, 495201. [CrossRef] [PubMed]
108. Zhu, J.; Zhang, Q.; Zhang, C.-H.; Weng, G.-J.; Zhao, J.; Li, J.-J.; Zhao, J.-W. Synthesis of colloidal gold
nanobones with tunable negative curvatures at end surface and their application in SERS. J. Nanopart. Res.
2017, 19, 364. [CrossRef]
109. Lee, S.J.; Baik, J.M.; Moskovits, M. Polarization-Dependent Surface-Enhanced Raman Scattering from a
Silver-Nanoparticle-Decorated Single Silver Nanowire. Nano Lett. 2008, 8, 3244–3247. [CrossRef] [PubMed]
110. Dai, Z.G.; Xiao, X.H.; Liao, L.; Ying, J.J.; Mei, F.; Wu, W.; Ren, F.; Li, W.Q.; Jiang, C.Z. Enhanced and
polarization dependence of surface-enhanced Raman scattering in silver nanoparticle array-nanowire
systems. Appl. Phys. Lett. 2015, 107, 129901. [CrossRef]
111. Li, J.; Zhang, W.; Lei, H.; Li, B. Ag nanowire/nanoparticle-decorated MoS2 monolayers for surface-enhanced
Raman scattering applications. Nano Res. 2018, 11, 2181–2189. [CrossRef]
112. Šubr, M.; Petr, M.; Kylián, O.; Štěpánek, J.; Veis, M.; Procházka, M. Anisotropic Optical Response of Silver
Nanorod Arrays: Surface Enhanced Raman Scattering Polarization and Angular Dependences Confronted
with Ellipsometric Parameters. Sci. Rep. 2017, 7, 4293. [CrossRef] [PubMed]
113. Nagai, Y.; Yamaguchi, T.; Kajikawa, K. Angular-Resolved Polarized Surface Enhanced Raman Spectroscopy.
J. Phys. Chem. C 2012, 116, 9716–9723. [CrossRef]
114. Ranjan, M.; Facsko, S. Anisotropic surface enhanced Raman scattering in nanoparticle and nanowire arrays.
Nanotechnology 2012, 23. [CrossRef] [PubMed]
115. Khlebtsov, B.N.; Khlebtsov, N.G. Multipole Plasmons in Metal Nanorods: Scaling Properties and Dependence
on Particle Size, Shape, Orientation, and Dielectric Environment. J. Phys. Chem. C 2007, 111, 11516–11527.
[CrossRef]
116. Gkogkou, D.; Schreiber, B.; Shaykhutdinov, T.; Ly, H.K.; Kuhlmann, U.; Gernert, U.; Facsko, S.;
Hildebrandt, P.; Esser, N.; Hinrichs, K.; et al. Polarization- and Wavelength-Dependent Surface-Enhanced
Raman Spectroscopy Using Optically Anisotropic Rippled Substrates for Sensing. ACS Sens. 2016, 1, 318–323.
[CrossRef]
117. Le Ru, E.C.; Meyer, S.A.; Artur, C.; Etchegoin, P.G.; Grand, J.; Lang, P.; Maurel, F. Experimental demonstration
of surface selection rules for SERS on flat metallic surfaces. Chem. Commun. 2011, 47, 3903–3905. [CrossRef]
[PubMed]
118. Peyskens, F.; Dhakal, A.; Van Dorpe, P.; Le Thomas, N.; Baets, N. Surface Enhanced Raman Spectroscopy
Using a Single Mode Nanophotonic-Plasmonic Platform. ACS Photonics 2016, 3, 102–108. [CrossRef]
119. Dhakal, A.; Peyskens, F.; Clemmen, S.; Raza, A.; Wuytens, P.; Zhao, H.; Le Thomas, N.; Baets, R. Single mode
waveguide platform for spontaneous and surface-enhanced on-chip Raman spectroscopy. Interface Focus
2016, 6, 20160015. [CrossRef] [PubMed]
120. Liu, Y.J.; Zhao, Y.P. Simple model for surface-enhanced Raman scattering from tilted silver nanorod array
substrates. Phys. Rev. B 2008, 78, 075436. [CrossRef]
121. Benson, M.; Shah, P.; Marciniak, M.; Sarangan, A.; Urbas, A. Optical Characterization of Silver Nanorod Thin
Films Grown Using Oblique Angle Deposition. J. Nanomater. 2014, 694982. [CrossRef]
122. Perera, M.N.M.N.; Schmidt, D.; Gibbs, W.E.K.; Juodkazis, S.; Stoddart, P.R. Effective optical constants of
anisotropic silver nanoparticle films with plasmonic properties. Opt. Lett. 2016, 41, 5495–5498. [CrossRef]
[PubMed]
123. Grundmann, M.; Sturm, C.; Kranert, C.; Richter, S.; Schmidt-Grund, R.; Deparis, C.; Zuniga-Perez, J. Optically
anisotropic media: New approaches to the dielectric function, singular axes, microcavity modes and Raman
scattering intensities. Phys. Status Solidi-Rapid Res. Lett. 2017, 11, 1600295. [CrossRef]
124. Oates, T.W.H.; Ranjan, M.; Facsko, S.; Arwin, H. Highly anisotropic effective dielectric functions of silver
nanoparticle arrays. Opt. Express 2011, 19, 2014–2028. [CrossRef] [PubMed]
125. Jellison, G.E. Spectroscopic ellipsometry data analysis: Measured versus calculated quantities.
Thin Solid Films 1998, 313, 33–39. [CrossRef]
126. Jayawardhana, S.; Rosa, L.; Juodkazis, S.; Stoddart, P.R. Additional Enhancement of Electric Field in
Surface-Enhanced Raman Scattering due to Fresnel Mechanism. Sci. Rep. 2013, 3, 2335. [CrossRef] [PubMed]
127. Perera, M.N.M.N.; Gibbs, W.E.K.; Juodkazis, S.; Stoddart, P.R. Wavelength and refractive index dependence of
the geometrical enhancement in surface-enhanced Raman scattering. J. Raman Spectrosc. 2017, 48, 1182–1189.
[CrossRef]
Nanomaterials 2018, 8, 418 38 of 38

128. Ming, T.; Chen, H.; Jiang, R.; Li, Q.; Wang, J. Plasmon-Controlled Fluorescence: Beyond the Intensity
Enhancement. J. Phys. Chem. Lett. 2012, 3, 191–202. [CrossRef]
129. Liaw, J.-W.; Tsai, H.-Y.; Huang, C.-H. Size-Dependent Surface Enhanced Fluorescence of Gold Nanorod:
Enhancement or Quenching. Plasmonics 2012, 7, 543–553. [CrossRef]
130. Song, M.; Chen, G.X.; Liu, Y.; Wu, E.; Wu, B.T.; Zeng, H.P. Polarization properties of surface plasmon
enhanced photoluminescence from a single Ag nanowire. Opt. Express 2012, 20, 22290–22297. [CrossRef]
[PubMed]
131. Šubr, M.; Petr, M.; Kylián, O.; Kratochvíl, J.; Procházka, M. Large-scale Ag nanoislands stabilized by
a magnetron-sputtered polytetrafluoroethylene film as substrates for highly sensitive and reproducible
surface-enhanced Raman scattering (SERS). J. Mater. Chem. C 2015, 3, 11478–11485. [CrossRef]
132. Le Ru, E.C.; Galloway, C.; Etchegoin, P.G. On the connection between optical absorption/extinction and
SERS enhancements. Phys. Chem. Chem. Phys. 2006, 8, 3083–3087. [CrossRef] [PubMed]
133. Kleinman, S.L.; Sharma, B.; Blaber, M.G.; Henry, A.I.; Valley, N.; Freeman, R.G.; Natan, M.J.;
Schatz, G.C.; Van Duyne, R.P. Structure Enhancement Factor Relationships in Single Gold Nanoantennas by
Surface-Enhanced Raman Excitation Spectroscopy. J. Am. Chem. Soc. 2013, 135, 301–308. [CrossRef] [PubMed]
134. Aizpurua, J.; Bryant, G.W.; Richter, L.J.; de Abajo, F.J.G.; Kelley, B.K.; Mallouk, T. Optical properties of
coupled metallic nanorods for field-enhanced spectroscopy. Phys. Rev. B 2005, 71, 235420. [CrossRef]
135. Zhao, Y.P.; Chaney, S.B.; Zhang, Z.Y. Absorbance spectra of aligned Ag nanorod arrays prepared by oblique
angle deposition. J. Appl. Phys. 2006, 100, 063527. [CrossRef]
136. Baumberg, J.J.; Kelf, T.A.; Sugawara, Y.; Cintra, S.; Abdelsalam, M.E.; Bartlett, P.N.; Russell, A.E.
Angle-Resolved Surface-Enhanced Raman Scattering on Metallic Nanostructured Plasmonic Crystals.
Nano Lett. 2005, 5, 2262–2267. [CrossRef] [PubMed]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).