Surface-Enhanced Raman Scattering: Introduction and Applications

1.1 Classical theory of the Raman effect

Raman scattering can be explained using the molecular polarizability [5]. If a molecule is placed in an electric field, electrons and nuclei get displaced. Due to the separation of charged species, an electric dipole moment is induced in the molecule, and it is said to be polarized. If E is the strength of the electric field and μ is the magnitude of the induced dipole moment, then

where α is the polarizability of the molecule. If a sample is subjected to an electromagnetic wave of frequency ν, the electric field experienced by each molecule of the sample varies as

where E₀ is the amplitude of the electromagnetic wave. From Eq. (1)

Thus, Eq. (3) implies that interaction of electromagnetic radiation of frequency ν induces a molecular dipole moment that oscillates and emits radiation of the same frequency, and this is the classical explanation of Rayleigh scattering. However, the ability to perturb the local electron cloud of a molecular structure depends on the relative location of the individual atoms; hence, the polarizability is a function of the instantaneous position of the constituent atoms. So, the polarizability changes with small displacement from equilibrium position (i.e., molecular vibration) and is given by

where α0 is equilibrium polarizability and qeq and q are bond lengths at equilibrium position and any instant, respectively. If a molecule executes simple harmonic motion, the displacement can be represented as

where νvib is the vibrational frequency of a molecule and qmax is the maximum separation distance between atoms relative to their equilibrium position. Substituting Eq. (5) into Eq. (4) gives

Substituting Eq. (6) into Eq. (3) gives

or

or

The first term in Eq. (9) represents Rayleigh scattering and occurs at the excitation frequency ν. The second and third terms correspond to Stokes (ν − νvib) and anti-Stokes (ν + νvib) scattering. In both inelastic scatterings, the excitation frequency is modulated by the vibrational frequency of the bond.

Besides, from Eq. (9), the molecules that have Raman-active vibration modes must experience a change in polarizability during a vibration ∂α∂q, i.e., the electron density in the molecule must distort from its typical shape (inducing a dipole). Molecules with symmetrical bends and stretches, therefore, are generally better Raman scatterers. So, for a molecule to be Raman active, its molecular rotation or vibration must cause a change in a component of molecular polarizability.

1.2 Quantum theory of the Raman effect

Raman scattering can be easily understood in terms of the quantum theory of radiation. In the quantum model, the molecules exist in quantized energy levels corresponding to possible stationary states of the molecule. When radiation having an energy hν incident on a sample, it is considered that the photons undergo collisions with the molecules. When the collision is elastic, the photons will be deflected unchanged, but it is also possible that during the collision, energy is exchanged between the photon and molecule, and as a result, the molecule can gain or lose energy ΔE, where ΔE = hνvib represents a difference in the vibrational or rotational energy levels of that molecule [6]. In quantum mechanical terms, the scattering can be considered as an excitation to a virtual state lower in energy than a real electronic state. When the molecule gains energy ΔE, the photon will be scattered with energy hν − hνvib, and the scattering is known as Stokes’ scattering. Conversely, if the molecule loses energy ΔE, the scattered photon will have energy hν + hνvib, and this type of scattering is known as anti-Stokes’ scattering. Generally, Stokes’ radiation is stronger than the anti-Stokes’ radiation. Figure 1 illustrates the energy level diagram for scattering [7].

2.1 Chemical enhancement mechanism

CE requires the probe to be chemically bound to the SERS substrate. The CE can be grouped into three contributions to the chemical mechanism: (i) a resonance Raman (RR) effect due to the incident light matching an electronic transition in the molecule (10³–10⁶ contribution), (ii) a charge-transfer (CT) effect where the incident light is in resonance with a metal-molecule or molecule-metal transition. (10–10⁴ contribution), (iii) a nonresonant chemical (CHEM) effect due to ground-state orbital overlap between the molecule and the metal (≤10–100 contribution) [21].

RR is a molecular resonance mechanism that arises from the incident light being resonant with a molecule, and without a metal surface, this leads to resonance Raman scattering. RR involves the formation of a surface complex involving the metal and the analyte, leading to a change in the properties of the molecule (such as the possibility of resonance Raman scattering). The RR effect is typically thought of as a molecular property, and it has been included as a SERS mechanism since the presence of the metal surface can alter where this resonance lies. The CT effect only appears when the molecule and metal are close enough to allow for a sufficient overlap of their wave functions. In this mechanism, tunneling of electrons between the metal and adsorbate molecules takes place. Due to the transfer of an electron from metal to molecule or from molecule to metal, a negative ion is formed. Enhancement occurs when the energy of the negative ion is resonant with the incident photon. This mechanism is explained by considering the molecule and metal system as a whole. It is considered that the Fermi level of the metal layer lies between the molecular ground level and one or more excited states of the molecule. The charge transfer mechanism is short-ranged (0.1–0.5 nm) and strongly dependent on the geometry, bonding, and the molecule energy level [22]. The CHEM effect is the least studied and most difficult to quantify experimentally due to its small contribution to the overall enhancement. The formation of metal-molecular complexes mainly causes the CHEM effect due to chemical bonding [23]. This modifies the ability of the dipole to radiate energy, i.e., it can effectively oppose or amplify the dipole amplitude (Figure 2).

2.2 Electromagnetic mechanism

In theory, EM enhancement is analyte independent, while CE is probe-dependent and requires some chemical interaction to the metal surface [24]. Most of SERS enhancements are due to the EM enhancement mechanism and are a direct consequence of the roughness present on the noble metal surfaces [25]. The nanostructure can be formed on the substrate itself or by depositing noble metal nanoparticles. These metal nanoparticles can interact with the excitation light because of unique properties caused by their low dimensions (10–100 nm). The small size of the metal nanoparticles makes a special kind of light-induced electric polarization possible for their surface electrons. Collective oscillations of these electrons, driven by the alternating electric field of the light wave, are called surface plasma oscillations. At a particular frequency, plasmon oscillations are resonant with light; then electric field intensity and Raman scattering from the molecules attached to the nanostructures are enhanced [8, 26]. A locally strong light-induced electric field of plasmons in metal nanoparticle causes the increase of Aeωeand Asωsfactors. It is because nanoparticles work as a kind of optical antenna, redistributing and concentrating light energy near a nanoparticle. As a result, the cross-section of the light scattering processes, including Raman scattering, can be much larger than the geometrical cross-section of the metal nanoparticle.

The surface electron oscillations in metal nanoparticles can be derived from the classical Drude model, describing metal as a lattice of ions immersed into the “gas” consisting of the free electrons [27, 28]. In a static electric field, the internal field of the metals, generated by the displacement of free electrons, shield the external electric field. As a result, the external electrostatic field cannot create the electric field inside the metal. As a result, if electrostatic fields are applied to the metals, their dielectric permittivity is ambiguous. The dielectric permittivity is the measure of how much the electric field inside a material differs from that of a vacuum. However, when a high-frequency electric field is applied, the free electrons inside the metal cannot completely follow in time with the high-frequency oscillations of the electric field. It creates a situation where at very high frequencies, metal can pass the electric field from the incident light, i.e., behave as a dielectric. The high transparency of these metals in the ultraviolet region can be explained by the fact that they have a lot of free electrons. Electrons of such metals as Al, Cu, Au, and Ag are not completely free but partially bounded.

The surface plasmon frequency ωspin small spherical metal nanoparticle includes the frequency of the volume plasma ωpand permittivity of the surrounding dielectric [29]:

Hence at the resonant frequency ωsp=ω, from Eqs. (11) and (12),

From Eq. (12), it follows that the permittivity of metal should have a negative value. Few metals such as Cu, Ag, and Au exhibit strong visible light plasmon resonance, whereas other transition metals show only broadband in the ultraviolet region. Ag, in particular, is suitable for SERS applications in the visible and near IR because it has a tiny imaginary component in this region and thus is less “lossy” than other metals [30].

When monochromatic radiation of frequency ν0 and electric filed E interacts with a molecule, it induces a Raman dipole oscillating at a frequency μ=αE. The oscillating Raman dipole radiates a power proportional to |μ|² at frequency ν and is the frequency detected as Raman signal in far-field. The same phenomenological description can be applied to SERS. However, the presence of nanostructured metal surface alters the effects in the following ways [31, 32, 33, 34, 35]:

1. The electromagnetic field at the metallic surface can be dramatically increased and may result in a possible local field enhancement.

2. The radiation properties of the Raman dipole, μ, are affected by the metallic environment and may result in a possible radiation enhancement.

3.1 Raman spectroscopy for transition metal dichalcogenides (TMDs)

Transition metal dichalcogenides, as the names suggest, are a class of material that is made up of the transition metals (M = Mo, W, Ta, Pt) and chalcogenides (X = S, Se, Te). The unit cell of bulk MX₂ consists of X–M–X units, where one M plane is sandwiched between two X planes. Depending upon how these units are stacked, different kinds of polytypes are formed, for example:

1. 1T (one X–M–X unit in the unit cell, octahedral coordination, tetragonal symmetry, T stands for trigonal)

2. 2H (two X–M–X layers per repeat unit, trigonal prismatic coordination, hexagonal symmetry, H stands for hexagonal)

3. 3R (three X–M–X layers per repeat unit, trigonal prismatic coordination, rhombohedral symmetry, R stands for rhombohedral)

The schematic diagram of these polytypes is given in Figure 3.

In recent years, Raman spectroscopy is adopted to address the challenges in the characterization of these TMDs due to: (1) Many possible structures (generally the stacking of layers has a rotational (or translational) freedom with reference to an axis perpendicular to (or also along) a 2D plane, giving them non-uniqueness); (2) weak interlayer interaction, which is much smaller than the intralayer interactions; and (3) observation of inner layers (the physical properties of each layer within the stacked TMDs are not selectively accessible except for the outermost layer). Other than resolving these challenges, Raman spectroscopy can identify functional groups, structural damage, unwanted by-products, and chemical modifications introduced during synthesis, processing, or its placement on the various substrates during device fabrication. It is spectroscopy, which is nondestructive, quick, and noninvasive for characterizing the TMD materials with high selectivity [64, 65, 66].

In general, the Raman spectroscopy has been widely used to determine the layers of the TMDs. For all the layered materials, which also includes TMDs, there are typically two categories of Raman vibrations. One is intralayer vibrations, which occur within a layer and normally appear in the high-frequency region of the spectra. The second category of Raman vibrations is observed due to the relative motion of the layers. These vibrations give the interlayer Raman modes, which are normally observed in the low-frequency region of the spectra (<100 cm⁻¹).

In the layered TMDs, to determine the number of layers N for few-layer TMDs, high-frequency intralayer Raman modes can be used [67, 68]. For example, the layer number (N) dependence of the peak position and width for E2g1and A_1g of NL-MoS₂ is shown in Figure 4(a). The A1g and E2g1modes undergo blue and redshifts with increasing numbers of layers for MoS₂.

The position of (E2g1) and (A_1g) with decreasing thickness from the bulk to 1 L shows opposite trends, as shown in Figure 4(b).

However, these two modes would decrease in frequency from 2L to 1L based on the linear chain model (only van der Waals interactions are included). This unexpected behavior of E2g1suggests that interactions other than van der Waals forces also exist [68, 69, 70], which is also disclosed by the anomalous Davydov splitting between E2g1and E_1u [71].

Molina-Sanchez et al. [72] carefully examined and reported the relationship between the number of monolayers and the Raman active modes (A_1g and E2g1). They demonstrated that the weak interlayer interaction is the leading cause of the frequency increase (i.e., for A_1g) with the number of layers. Moreover, the decrease in the E2g1phonon frequency is associated with a stronger dielectric screening of the long-range Coulomb interaction (which is induced by the effective charges resulting from the relative displacement between Mo and S atoms) in a few layers and in bulk. Thus, it is expected to exhibit an anomalous frequency trend in which the A_1g mode increases in frequency with an increasing number of layers, while the E2g1mode decreases.

In the case of the WS₂, the proximity of the 2LA(M) and E2g1peaks makes it challenging to measure the accurate difference between the A_1g and E2g1peaks and increases the chance of error in determining the frequency shift of both the modes. To resolve this problem, Berkdemir et al. [74] reported a new method based on the ratio of the intensities of I_2LA and I_A1g peaks. In this method, the authors reported that the absolute intensity of the 2LA(M) mode increases with decreasing the number of layers, while the intensity of the A_1g displays the opposite behavior [74]. The behavior of the A_1g mode with a decreasing number of layers presumably results from weaker interlayer contributions to the phonon restoring forces.

3.2 Application of Raman spectroscopy in photovoltaics

Conducting polymers are widely used in organic light-emitting diodes, heterojunction diodes, organic thin-film transistors, solar cells, actuators, sensors, etc. [75, 76, 77, 78, 79]. Poly(3,4-ethylenedioxythiophene)-poly(styrene sulfonate) (PEDOT:PSS) is extensively used conducting polymer because of its high conductivity, excellent thermal stability, transparency, structural stability. PEDOT:PSS polymer is a promising candidate as a transparent electrode for optoelectronic devices. The solvent treatment of PEDOT:PSS films may affect the conformation of the polymer. The structure of the PEDOT chain changes from benzoid to quinoid structure after solvent treatment [75, 80]. The effect of the conformation of the PEDOT chains in the PEDOT:PSS film before and after the dimethyl sulfoxide (DMSO) treatment was studied by Raman spectroscopy. Figure 5 shows the Raman spectra of PEDOT:PSS with different concentrations of DMSO. The most obvious change was observed for the strongest band between 1400 and 1500 cm⁻¹. The highly conductive PEDOT:PSS film exhibits a narrower band. This change is similar to that of PEDOT:PSS film treated with ethylene glycols reported by Xia et al. [81]. These vibrational modes correspond to the stretching vibrations of C_α = C_β on the five-member ring of PEDOT.

The band at 1440 cm⁻¹ is associated with the C_α〓C_β symmetric vibration. The band near about 1368 cm⁻¹ is associated with the C_β▬C_β stretching. Raman peaks located at 1508 and 1568 cm⁻¹ are associated with the C_α〓C_β asymmetric stretching vibrations. The band at 1540 cm⁻¹ has been related to the splitting of these asymmetrical stretching vibrations [79, 81, 82, 83, 84]. Two kinds of resonant structures have been proposed for PEDOT, namely, benzoid and quinoid structure. For coil conformation, benzoid structure is the favorite structure, and quinoid structure is the favorite structure for linear and expanded-coil structure. Both benzoid and quinoid resonant structures exist simultaneously in pristine PEDOT:PSS film. The benzoid structure may be transformed into the quinoid structure after DMSO treatment so that quinoid structure becomes dominant in the highly conductive PEDOT:PSS film. The conducting PEDOT:PSS films are vastly used in optoelectronic devices.

4.1 Random morphology SERS substrates

Random morphology SERS substrates are inhomogeneous and are not highly reproducible [86]. Roughened electrodes are the most primitive SERS substrate and were discovered by Fleischmann et al. [9]. These substrates are typically created by running the redox cycle in an electrochemical cell containing a metallic salt solution. Such substrates have gained popularity due to an ability to adjust electrode potential to understand the charge transfer phenomenon between adsorbate and metallic surface [87]. Regardless, the importance of this substrate is decreasing substantially due to relatively low enhancement factors.

Among the random morphology SERS substrates, silver or gold colloids are the most common substrates used in both early and more recent studies. Since colloids are easy to produce in a laboratory and tend to generate large enhancement factors, most researchers are still involved in colloid-based SERS rather than more sophisticated substrates [88, 89]. Metallic colloids are also of historic significance related to SERS development, as the first single-molecule SERS detection was reported using colloid substrates [90]. In colloid-based methods, nanoparticle size and geometry can be controlled by altering experimental conditions. One of the most popular methods for controlling nanoparticle morphology stems from the polyol synthesis of silver nanocubes by Sun and Xia [78]. In addition to the nanocubes, various groups have produced octahedra and cuboctahedra [91] and octapods [92]. El-Sayed et al. have contributed to the control of particle morphology, yielding a variety of interesting and useful structures [93, 94, 95]. Since the size, shape, and material of the particles govern the resulting plasmonic resonance characteristics, significant effort has been exerted in the control of plasmon resonance via core-shell and alloyed particles, to which the Halas group has a large contribution [96, 97, 98].

SERS substrates have also been fabricated by depositing nanoparticles onto different surfaces. These simple deposition approaches include micro pipetting [99], soaking [100], screen-printing [101], filtration [102], and inkjet printing [103]. However, a major concern with these simple deposition processes is that the hotspots are generated and distributed randomly over the substrate. The differences in metallic particle sizes and their shapes due to the differences in preparation recipes can lead to several orders of magnitude difference in the SERS enhancement factor.

In order to obtain more consistent hotspots from nanoparticles, researchers have explored both self-assembly [99] and directed assembly techniques, such as the Langmuir-Blodgett techniques [91], to create regular arrays of nanoparticles. However, these techniques also introduce more complexity to the fabrication process.

4.2 Periodic or uniform SERS substrate

Although metallic colloidal particles are known for their high SERS EF and possibility to accomplish SERS spectra of a single molecule, it is often challenging to reproduce or routinely deliver such a high performing SERS feature. To overcome this issue, a few alternatives have been introduced on engineering periodic arrays of metallic nanostructured SERS substrates. Nanosphere lithography (NSL), developed by Van Duyne et al., is one of the most extensively used nano-fabrication procedures used in understanding SERS phenomena and performing plasmonic-based sensing [104, 105, 106]. It involves the assembly of polystyrene nanospheres into a regular array, using this as a mask to create periodic nanostructures, sometimes called “metal island films,” by the evaporation of Ag or Au through the gaps created by the packing of the nanospheres, followed by removal of the nanospheres. A variation on this method, called “metal film over nanospheres,” is to evaporate a metal film directly onto the nanosphere template, using closely-packed nanospheres to pattern the substrate surface itself [107]. Electron beam lithography (EBL) is another most widely used conventional nanofabrication technique in designing uniform and controlled morphology SERS substrates [108, 109]. In the literature, there are many SERS-active surface designs prepared with EBL [110, 111]. These designs are mainly periodic arrays of simple nano-structures, and generally, the relation between the LSP resonance wavelength and SERS signal enhancement is studied. For example, Le Ru et al. have taken SERS measurements, from periodic gold dot, square, and triangle arrays, from “Rhodamine 6G” [112]. They demonstrated that the localized plasmon resonances, which are at the origin of visible-NIR extinction spectra and the SERS effect, can be tuned to any desired wavelength by varying the particle shape/size and spacing, thus tuning the Raman amplification. In a very similar study, Gunnarsson et al. studied similar Ag structures on a silicon wafer for the same molecule and reported that better results are obtained than nano-roughened Ag film [110]. They investigated the size and geometry dependence of the SERS effect on supported particles, by manufacturing artificial structures by modern nanofabrication techniques. Arrays of 100–200 nm silver particles of different shapes were prepared on a Si wafer by electron beam lithography. Kahl et al. have shown that the SERS measurements of “Rhodamine 6G” on gold periodic nano-dot arrays and grating structures resulted in order of magnitude better SERS enhancement when compared with metal-island film substrates [113]. They demonstrated two different methods of substrate fabrication by e-beam lithography. In the first method, regular fields of nanoparticles are produced by the lift-off technique. A silver layer is evaporated on the structured resist, and the resist is removed afterward. In the second method, gratings or crossed gratings are transferred into a silicon wafer with a thermal oxide surface layer by reactive ion etching (RIE). Then, the e-beam resist is removed, and finally, a silver layer is evaporated. Hatab et al. have demonstrated significant SERS enhancement factors exceeding 10¹¹, resulting from a new configuration of elevated gold bowtie nano-antenna arrays with optimized array periodicity [114]. A process combining nanofabrication steps of pattern definition by EBL, metal deposition, lift-off, and RIE arranged in a particular sequence was used to fabricate the elevated gold bowtie arrays on Si wafers. The elevated bowties allow the manifestation of intrinsic plasmonic coupling effects in suspended nanocavities, or the tip-to-tip nanogaps, from structures that are not in physical contact with a substrate. This configuration results in up to two orders of additional magnitude enhancement in SERS response compared to that of nonelevated bowtie arrays. The diversity of designs is endless when fabrication with EBL is considered. However, these techniques, while excellent at making SERS substrates with defined characteristics, are hampered by the slow, serial nature and high cost of the processes used in their fabrication.

4.3 SERS from silver columnar film

The SERS enhancement strongly depends on the substrate. As already discussed earlier, various techniques have been proposed and identified for the fabrication of the SERS substrate. However, only a few methods are available to develop uniform, reproducible, robust, stable, and cost-effective SERS substrates. Recently, silver columnar thin films fabricated by glancing angle deposition (GLAD) have been identified as high sensitivity SERS active substrates [115, 116, 117, 118, 119, 120, 121, 122]. A remarkable SERS enhancement factor with applications in sensing the biomolecules at very low concentrations has been observed on the silver nanorod arrays [123]. To understand the SERS mechanism and attain a maximum possible enhancement, large numbers of studies have been performed on the Ag nanorod (AgNR) arrays. In an interesting study, Chaney et al. have investigated the SERS response as a function of the nanorod length using trans-1,2-bis(4-pyridyl)ethane (BPE) as a probe molecule at an excitation wavelength of 785 nm [124]. They found that the SERS intensity increases dramatically with nanorod length. Zhou et al. fabricated aligned, single-crystalline AgNRs on planar Si substrates by GLAD technique, with sample substrate cooled by liquid nitrogen in the e-beam deposition system [125]. They were successful in detecting aqueous solution of 10⁻¹² molL⁻¹ Rhodamine 6G by the porous Ag film with nanorods. They also deposited AgNRs on Ag, Al, Si, and Ti thin films with a thickness of 100 and 400 nm, respectively, to achieve layers with different reflectivities. The SERS intensity of the AgNRs grown on Ag thin film was found to be higher than others, and the SERS intensity of the AgNRs on Al film was larger than that on Ti film, and the AgNRs on Si film showed the minimum SERS intensity. They concluded that the larger the under-layer reflectivity, the larger the SERS performance of substrate. So, the pre-deposition of Ag layer under AgNRs can be an effective way to promote the SERS performance of AgNRs. Zhang et al. have made AgNRs in film grow into periodic patterns at a micro-nano scale, and they showed that the AgNR film with periodic patterns exhibits better SERS performance than Ag film with nanorods arranged randomly as before [126]. He et al. also reported a new scalable strategy based on dynamic shadowing growth (DSG) to fabricate large-scale chiral Swiss roll nanostructures. They developed a chiral conical Swiss roll nanostructure by helically stacking Ag films on a SiO₂ frustum with SiO₂ films as insulating layers [127]. They also showed that the chiral dichroism (CD) spectral feature can be tuned by changing the bead diameter. They achieved a broadband CD response in visible to near-IR region by making the bead diameter a few hundred nanometers. Mark et al. combined the low-temperature shadow deposition with nanoscale patterning to fabricate nanocolloids with anisotropic three-dimensional shapes, feature sizes down to 20 nm [128]. They first deposited a uniform hexagonal array of Au nanodots deposited onto a Si wafer by micellar nanolithography. Then they deposited material onto the substrate by physical vapor deposition at grazing incidence. To reduce the adatoms’ mobility and reduce the diffuse during growth, they cooled the substrate. So, by combining the uniform nano-seeding and low-temperature growth, they fabricated various complex hybrid nanostructures of many materials like from Al₂O₃, Ti, and Cu. GLAD has also emerged as a powerful tool for the fabrication of 3D chiral plasmonic nanostructures. Titus et al. investigated the optical properties of Ti-doped Ag helices in the visible and near-infrared ranges using transmission ellipsometry and spectroscopy fabricated by GLAD [129]. Nair et al. reported the fabrication of wafer-scale 3D chiral nanoplasmonic substrates with different dielectric templates, namely, silica, magnesium fluoride, and titanium dioxide using GLAD [130]. They have also investigated the effect of interparticle separation on the chiroptical response of chiral nanohelices [131].

Hence, we can see that the development of fabrication and application of substrates for SERS is driven by nanotechnology and the development of high-end fabrication processes. Increasingly SERS substrates with high sensitivity and reproducibility are invented by electrochemical deposition, physical vapor deposition of the metal film, metal nanoparticle colloids, and so forth and applied into various fields, such as detection of pollutants at trace level, surface analysis, biomolecule, and bacteria detection. With the development of SERS substrate, advancement in Raman spectrometers, and tip-enhanced Raman scattering (as the combination of SERS and atomic force microscopy), SERS is becoming increasingly popular as a detection and diagnostic tool.