Recent advances in the design of SERS substrates and sensing systems for (bio)sensing applications: Systems from single cell to single molecule detection

1.1 Mechanism of Raman scattering

Raman scattering involves the inelastic transfer of momentum from photons of the incident light to molecules in the sample. The interaction of the external electromagnetic field with the electron subsystems of the molecule results in light scattering. These molecular energy transitions during scattering are shown in the Jablonski diagram, which illustrates the electronic states: ground state, singlet state (S_o), and triplet state (T), as shown in Figure 1. A typical Jablonski diagram contains vertically stacked electronic states and horizontally grouped vibrational energy states, according to their spin multiplicity. Each energy state contains sublevels based on the vibrational energy of the molecule (Demtröder, 2015). According to Pauli’s exclusion principle, all electrons are paired in the ground state of the molecule, while a change in the spin of the energy state is observed in the singlet state (S₀) due to half-filled molecular orbitals. The electronic transitions to higher singlet energy states are caused by photon energy absorption from the incident electromagnetic radiation, whereas the emission is a two-photon process, with single-photon emission to the lowest vibrational state (S₁). A triplet state (T₁) is a virtual triple-degenerate excited state for the relaxation of excited photons to the ground state by intersystem crossing (Prochazka, 2016). The standard expression for the energy of the photon transition is illustrated in (Equation 1), where h = Planck’s constant = ~6.626 x10⁻³⁴ J/s and c = speed of light = ~3 × 10⁸ m/s.

Figure 1. Schematic representation of energy levels of a molecule in a Jablonski diagram.

However, in addition to optical absorption and emission, another process termed as scattering is often observed in the emission of a molecular photon owing to the absorption of an incident light photon. This photon scattering may be elastic or inelastic. Rayleigh scattering, an elastic scattering, is associated with energy conservation during photon emission, whereas the inelastic scattering associated with the transfer of energy between photons and molecular subsystems is termed Raman scattering. Furthermore, any Raman active molecule in the lowest vibrational state that absorbs the photon energy causes a decreased frequency and higher wavelength (λ_S > λ_L) of the scattered photon, resulting in a lower energy of the scattered photon (E_S < E_L), termed as the Stokes process, as shown in Figure 2(a). In contrast, the anti-Stokes process occurs because of the emission of excessive energy from the molecule in the excited vibrational state with a lower wavelength (λ_S < λ_L) and a higher energy (E_S > E_L) of the scattered photons (Eric Le Ru, 2009). The Raman shift is only dependent on the material subjected to the Raman effect; a negative shift is associated with the anti-Stokes process, and a positive shift with the Stokes process. Because of the lower energy associated, the anti-Stokes Raman scattering process is weaker than the Stokes process (Blackie et al., 2009).

Figure 2. (a) Stokes and anti-stokes scattering due to energy changes in electronic excitation and emission.

(b) Raman spectrum depicting Stokes scattering and anti-stokes scattering.

Nonetheless, the abundance of photons involved in Stokes and anti-Stokes processes is significantly lower. A comprehensive study by D.A. Long stated that one out of 10⁷ photons incident on a sample may be scattered, resulting in the Stokes or anti-Stokes process. The Raman effect is also significantly affected by the components of the optical system, as it can limit the sensitivity with some major design considerations, such as the wavelength and power of the laser (Glass, 1967), spectral resolution (Meier et al., 1988), collection optics (Greenler & Slager, 1973), range and sensitivity of the detector (Allemand, 1970), and minimizing Rayleigh scattering (Cutler et al., 1980). Typically, near-infrared (NIR) laser-equipped Raman systems are employed to test organic (Košek et al., 2020) and biological specimens (Synytsya et al., 2014) in order to minimize fluorescence and facilitate easier penetration into the sample matrix. Efficient Raman scattering can be achieved only with optimum laser power. A lower power generates weak Raman signals, while high power can generate fluorescence, which in most cases may lead to sample degradation. In addition, spectrometers with a lower spectral resolution cannot distinguish between closely-spaced Raman peaks. The light collection efficiency of a spectrometer is significantly affected by collection optics; hence, high-numerical-aperture (NA) lenses are commonly used in commercial Raman systems. Rayleigh scattered light is a significant impediment to producing better Raman signals, which can be mitigated with notch filters. Other considerations include the choice of optical components for minimal autofluorescence and the stable alignment of the optical components for the highest light throughput.

Notwithstanding these challenges, definitive fingerprinting ability can assist in the development of Raman-based optical sensing systems and their use in sensing diverse inorganic molecules (Mycroft et al., 1990), biomolecules (Larsson & Rand, 1973), and bacterial cells (Layne & Bigio, 1986). Most of the Raman spectrometers before the 1990s used low-energy argon ion lasers with a high laser power of 200–500 mW and silicon photodiodes as detectors, thereby producing fluorescence and a low signal-to-noise ratio (SNR). The use of plasmonic nanoparticle-modified substrates may amplify Raman scattering at low laser power, which enhances the Raman effect. The Raman shift observed in the presence of resonating electronic clouds around noble metal surfaces enhances molecular scattering, and this surface enhancement is called Surface-Enhanced Raman Scattering (SERS) (D.A. Long, 1977).

2.1 Chemical enhancement

Chemical enhancement (CE) is defined as the amplified Raman scatter signal of a target analyte resulting from the interaction between the adsorbed analyte molecule and the plasmonic nanostructure. It is calculated as the sum of differences in Raman polarizability due to the molecule adsorption onto the metal surface and the charge transfer mechanism, given by Equation (4). However, the basis of CE has been the subject of debate for decades, particularly the science that governs the SERS chemical enhancement factor (EF) (Eric C. Le Ru & Pablo G. Etchegoin, 2009). Molecular adsorption can be physisorption or chemisorption with a bond energy of ~40 kJ/mol (Aroca, 2006). Charge transfer in the metal-molecule complex is due to the incident light energy corresponding to the electronic transitions of the molecule with the underlying phenomenon of Resonance Raman Scattering.

One of the widely accepted theories for chemical enhancement is the Charge Transfer (CT) mechanism proposed by the SERS pioneer Andreas Otto, where an adsorbed molecule subsequently changes the molecular polarizability, thereby enhancing the Raman scattering (Otto et al., 1992). For example, an incident photon with a frequency ν_inc, in resonance with the surface-adsorbate complex, causes excitation and return of the metal electron to its ground state because of the CT mechanism. However, if the excited electron resides in the lowest unoccupied molecular orbital (LUMO) for a time period shorter than the absorbed photons, it will be scattered with dissimilar energy levels than the incident photons (ω_s = ω_inc − ω_vib).

Chemical enhancement can be due to intramolecular resonance (Creighton, 1983), ground-state charge transfer, (Lippitsch, 1984) or resonant charge transfer (Lombardi et al., 1986). Intramolecular resonance assists in improving Raman scattering due to coherence of analyte’s molecular vibrations with the frequency of the selected excitation wavelength. However, intramolecular resonance is affected by the area of the Raman cross-section, as larger cross-sections assist in the greater scattering of some weak scattering molecules (Tauber & Mathies, 2002). Other key parameters include the chemical structure of the analyte (Ryder, 2005) and the shape, size, and dielectric properties of the plasmonic nanostructure (Jensen et al., 2000) (Notingher & Elfick, 2005). Another cause of chemical enhancement is ground-state charge transfer, a process of electron transfer between the metal surface and the adsorbed molecules to generate charged species, which typically assist the formation of coordination compounds (Flamigni et al., 2007) and redox reactions (Fukuzumi, 1997). However, the ground state charge transfer is associated with specific limitations, such as the selective enhancement of molecular vibrations of the analyte molecule, (Rurack et al., 2000) thereby limiting the uniform enhancement. Additionally, the charge transfer dynamics are affected by environmental factors (Fleming et al., 1988), the spectral overlap with other vibrational modes of the analyte (Blandamer & Fox, 1970) can limit the selectivity, and ground state saturation with the electrons of the analyte can limit any further enhancement. In the other chemical enhancement mode, resonant charge transfer was observed to overcome the limitations of ground-state charge transfer. This is the electron transfer from the resonating energy levels between the plasmonic metal and analyte molecule by charge transfer between the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO), which alters the local electromagnetic field of the analyte molecule, resulting in greater Raman scattering. The improvement in resonant excitation enhances the plasmonic metal nanostructure and analyte interaction (McNay et al., 2011) and facilitates enhanced Localized Surface Plasmon Resonance (LSPR) of the plasmonic substrates (Kleinman et al., 2013), thereby maximizing the enhancement effect. Despite these modes of signal improvement associated with chemical enhancement, their significance to the overall average enhancement factor was marginal. Hence, the use of noble metal substrates is also being explored to achieve better signal enhancement by altering the surface chemistry of substrates and modifiers.

However, the contribution of the CE to the average SERS enhancement was not significant. Therefore, the choice of materials used for SERS substrate synthesis requires meticulous consideration of the nanostructure morphology, light absorbance range, stability, (Yamada et al., 1987) and hydrophobicity, which may be tailored to improve plasmonic resonance and electromagnetic enhancement.

2.2 Electromagnetic enhancement

Electromagnetic (EM) enhancement refers to the amplification of Raman-scattered photons in the proximity of a resonating electron cloud at the metal-dielectric interface. Surface plasmon resonance (SPR) is generated by the cumulative resonance on the metal surface by the coherence of the electron cloud oscillation frequency at the metal-dielectric interface with the frequency of the excitation laser. In contrast, resonance localized at a position with plasmonic nanoparticles is termed LSPR. The LSPR effect can absorb or scatter the incident laser and potentially enhance the local electromagnetic field. Local field enhancement requires the molecule to be in proximity (within ~ 100 nm) to the metal surface, that is physisorption or chemisorption. The field enhancement factor is also prominently dependent on the laser power and Stokes and anti-Stokes effects (Ding et al., 2016). Similarly, EM enhancement is a coupling effect of the local field and the re-radiated (Raman) field of the SERS substrate, as shown in Figure 3. As previously discussed, the localized electromagnetic field for metals is higher if the excitation wavelength (λ_L) is close to the electromagnetic resonance of the system. Hence, the localized electric field (E_loc) is dependent on light polarization and its excitation wavelength. SERS hotspots are generated if the magnitude of │E_loc│ is greater than │E_inc│. The local field intensity enhancement factor (M_loc (w_L)) would be increased by a factor of:

Figure 3. Schematic illustration of electromagnetic (EM) enhancement with specifics of the two-step enhancement mechanism (This figure has been reproduced with permission from (Ding et al., 2016) Copyright (2016) Nature Reviews Materials).

In addition to local field enhancement, re-radiated field enhancement is also prominent for EM enhancement. In the electromagnetic model, the molecule is considered a dipole that responds to a greater local field near the surface (Kerker, 1984). The interaction of a metal nanostructure with light generates an LSPR effect on the nanostructure, which amplifies the incident EM field and scattered Raman field. Under SERS conditions, the radiation of the Raman dipole of a molecule in proximity to the metal surface modifies the exciting field, termed modified spontaneous emission (MSE). The enhancement of the EM field is a combination of the excitation field and the Raman scattered field, which is proportional to the fourth power of the field enhancement given by equation (7), where M_Loc (ω_L) is the local field enhancement, M_Loc (ω_R) is the re-radiated field enhancement, │E_loc (ω_L)│⁴ is the magnitude of the local electric field amplitude, and │E_loc│⁴ is the magnitude of the incident electric field amplitude.

The maximum EM enhancement for isolated silver or gold nanostructures/nanoparticles was observed in the range of 10⁹-10¹⁰ (Le Ru et al., 2007). This was further improved to approximately 10¹¹ by roughening the substrate surface (Camden et al., 2008). However, some significant limitations associated with electromagnetic enhancement include the near-field effect of isolated plasmonic nanomaterials that require close presence of the analyte on the substrate (Chou et al., 2012), heterogeneous signal enhancement (Schlücker, 2014) on the substrate surface that can alter the reproducibility, and amplification of background signals that obscure the weak Raman signals (Larmour et al., 2012), thereby reducing the signal-to-noise ratio (SNR) of the SERS system. A low SNR was associated with low sensitivity, poor selectivity, fluorescence, higher Rayleigh signal scattering, and low reproducibility and reliability. Thus, some considerations for transcending the low SNR of a SERS system include the synthesis of plasmonic nanoparticles with uniform shape, size, and morphology to ensure sharp absorption peaks for signal enhancement (Berciaud et al., 2005), use of pulsed lasers with short duration that minimize photobleaching of the sample (H. F. Zhang et al., 2007), use of longer-wavelength lasers (Liu et al., 2020) and spectral filters (T. Murphy et al., 2000) to reduce fluorescence of biological samples, and minimizing photobleaching effects by optimizing the laser exposure time and laser power. Ongoing studies suggest that the use of these strategies can effectively improve the SNR of SERS systems and their applicability for on-site sensing in environmental monitoring, characterization, and diagnostics. SERS systems provide significant amplification of the scattered Raman signal, thus allowing trace-level fingerprinting in complex samples with a relatively simpler optical system that allows for point-of-use (POU). Hence, the basic building blocks of the SERS system warrants a discussion.

3.1 Sample illumination setups and associated optics

Lasers are commonly employed in modern Raman spectrometry because of the high coherence necessary to produce efficient Raman scattering with high SNR. However, the design of POU SERS systems mostly uses laser photodiodes because of their small footprints. Table I presents a list of laser diodes along with their corresponding emission wavelengths and samples that are typically tested using the respective SERS systems. Laser diodes are a common choice for light sources in SERS systems because of their high stability, electronic tunability, and wavelength precision. However, the choice of a laser diode in a POU-SERS system requires a comprehensive understanding of the plasmonics of the nanostructure or nanocomposite. Additionally, the physical and chemical properties of the substrate and the optical properties of the substrate material can determine the required laser power, laser diodes with low fluorescence, minimal photodamage to the target analyte, and match the substrate plasmonic resonance with the excitation laser. Hence, SERS systems with laser diodes can generate accurate and reproducible measurements.

The choice of excitation wavelength is a key consideration, as some solvents and colored sample matrices absorb the incident light or the Raman scattered radiation, thus requiring multiple wavelength laser sources. As discussed in Table I, the laser diodes are wavelength-specific, and higher-wavelength lasers can produce a different color with doubled frequency, such as the NIR (1064 nm) laser diode, which can produce green light at 532 nm with doubled frequency. Other considerations for the choice of laser are (i) absorbance and autofluorescence of the sample matrix, (ii) optimized laser power to avoid the generation of fluorescence and sample degradation, and (iii) inexpensive, compact, and easy to integrate with the sensing system.

SERS sample illumination systems are equipped with laser diodes, as in the case of Raman systems; however, the choice of diode is based on the plasmonic absorbance wavelength of the nanostructure used in SERS systems. The sample illumination system also includes other optical components such as a beam expander to regulate the laser beam divergence, collimation optics to ensure a constant diameter of the laser beam, a dichroic mirror that separates laser light from the Raman-scattered light, and a few optical filters that ensure specific transmission of Raman-scattered light to the photodetector. The use of sample illumination systems for POU-SERS sensing applications requires device miniaturization by integration with nanostructure specific laser diodes or fiber optics probes coated with nanostructures, or integration with microfluidic channels for efficient transport of samples to the SERS-active substrate. In addition, the miniaturization of all optical components, such as mirrors, lenses, and spectral filters for specific applications, can also ease device miniaturization. They differ from conventional sample illumination systems in terms of compactness and portability, ease of integration with mobile electronic devices, ease of sample handling and transport to the laser transmission path, and the use of tailored SERS-active substrates for improved sensitivity and specificity.

Raman spectroscopy is widely applied for the analysis of all forms of matter, including the liquid and gaseous phases. Sensing target analytes in solid samples may require manual preparation for accessible target analyte detection, sampling by swabbing or pressing the samples on the substrate, and appropriate instrumentation, such as sample holders or stages integrated with fiber optics or microscopes. Figure 4(i) depicts the optical configuration used for the analysis of solid samples and the staged sample holder attached to the stepper motor that facilitates continuous solid sample rotation (Paiva et al., 2020). Commonly tested solid samples include nanomaterials, pharmaceuticals, semiconductors, metals, alloys, food, biological samples, forensic analysis, and environmental samples. In contrast, liquid samples are typically sealed in capillaries, glass tubes, or ampoules for samples using volatile solvents or detected in standard cuvettes for aqueous samples, as shown in Figure 4(ii) (Kawahara-Nakagawa et al., 2019). Typically, in liquid samples, collection optics are placed perpendicular to the sample position for maximum light collection and transmission to the photodetector. Commonly tested liquid samples include organic solvents, acids, biological fluids, petroleum products, chemical reagents, food, and beverages. Consequently, gaseous samples were tested by filling a capillary or small cavity with a fine ground sample. As shown in Figure 4(iii) The gaseous samples are typically passed through specialized optically transparent chambers, where scattering is observed perpendicular to the sample chamber (Sänze et al., 2013). Some commonly tested gaseous samples include pollutants NO_x and SO_x, molecules that are IR inactive with zero dipole moment, such as H₂, and hydrocarbons, such as methane, ethane, propane, and butane.

Figure 4. Sample illumination systems used in point-of-use (POU)- SERS devices (i) (A) Optical configuration: 1) 785 nm laser, 2) amplified spontaneous emission (ASE) filters, 3) dichroic mirror, 4) −100 mm meniscus lens, 5) 2″ silver-coated parabolic mirror with 4″ of focus and central 3.2 mm hole in the parallel direction, 6) silver mirror, 7) 2″ aspheric sample lens, 8) sample pellet, 9) 600 μm slit, 10) collimator lens, 11) notch filters, 12) focusing lens, 13) 100 μm optical fiber. (B) Sample holder coupled to the stepper motor for rotating the pellet. (This figure has been reproduced with permission from (Paiva et al., 2020) Copyright (2020) Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy), (ii) Schematic representation of the assay system using Raman spectroscopy for sensing in liquid samples (This figure has been reproduced with permission from (Kawahara‐Nakagawa et al., 2019) Copyright (2019) Protein Science), (iii) Raman-FTIR device for continuous measurement of the sensor output (DC conductivity) (This figure has been reproduced with permission from (Sänze et al., 2013) Copyright (2013) Angewandte Chemie).

3.2 Waveguides for SERS

Optic fiber bundles may be used to guide the laser and collect scattered Raman signals with minimal collimating optics, allowing the construction of simpler and more robust optical systems. The integration of fiber bundles allows for flexible optical transmission; hence, devising systems with SERS-enabled cartridges becomes easy (Milenko et al., 2020). Typically, a fiber optic probe integrated with a micro-Raman setup comprises an objective lens focused on a laser beam that brings excitation radiation towards the sample. This excitation fiber can be used to illuminate solid samples or immersed in liquid samples. The other terminal collects the scattered radiation from the aperture of the spectrometer. For example, as illustrated in Figure 5, the fiber-optic SERS probe was etched and modified using Ag nanostructures. The SERS-active probe was connected to a microscope objective and the other end was connected to a chamber with nanostructured Au substrates. Evaporation of a 2-napththalenethiol (2-NP) sample with Au nanostructures results in analyte adsorption onto the Ag film-coated SERS-active probe tip, which determines the analyte concentration. Time-based spectral measurements can determine the variation in sample matrix evaporation and thus facilitate the efficient detection of 2-NP (Agarwal et al., 2016).

Figure 5. (a) Block diagram of the experimental design of the fiber-optic SERS probe and the Au substrate in the measurement chamber, (b) Schematic illustration of the light transmission through the Ag nanoparticle layered fiber-optic SERS probe (This figure has been reproduced with permission from (Agarwal et al., 2016) Copyright (2016) Sensors and Actuators B: Chemical).

3.3 Photodetectors

Photodetectors used in optical systems include CCDs, CMOS, APDs, SiPMTs, and laser diodes, which are comprised of photoresponsive semiconductor materials. They are commonly used because of their spectral range, cost, and rapid capture and analysis. Photodiode arrays are the most commonly used photodetectors for the design of POU systems owing to their scalability and footprint. They offer high sensitivity and rapid responses that are detected in specific visible or near-infrared (NIR) regions. Semiconductor materials with customized bandgaps allow photodetectors to detect predefined wavelengths of light, thereby aiding in specific optical applications. These semiconductors operate based on the difference in bandgap energy, which determines the emitted light absorbance wavelength of the photodetector. Table II. discusses the types of photodiodes, detection ranges, specific characteristics, and commonly detected analytes. However, the tunability of photodiodes for specific applications requires the use of spectral filters for high SNR, optimizing the spectral response with signal processing tools, and miniaturization of photodiodes and allied electronics. Fourier Transform (FT) Raman spectrometers facilitate signal processing of the obtained spectral data by splitting the fundamental constituent frequencies of the signal. It also assists in high-frequency precision, enhancement of spectral resolution, data processing, image reconstruction, and the quantitative analysis of samples. Figure 6 illustrates a visible laser-based FT-Raman system, with optical elements such as parabolic mirrors to focus the excitation light, dielectric mirrors to collect the scattered light, long-pass filters to allow a specific wavelength of scattered light to reach the photodetector, a quartz beam splitter to simultaneously measure the reference and the sample, a pinhole to restrict unwanted light, and a photomultiplier tube that acts as a photodetector. The integration of this Raman system with FT assists in the classification of spectral patterns (Baeten et al., 1998) and multiplex analyte sensing from complex specimens.

Figure 6. Schematic of the visible laser-based FT-Raman spectrometer (P-Pinhole 2mm diameter, PM1, PM2 and PM3- Parabolic mirrors, M1, M2, M3, M4, and M5- Dielectric mirrors, F1 and F2- Long-pass filters, BS-Quartz beam splitter, PMT-Photo-multiplier tube) (This figure has been reproduced with permission from (Dzsaber et al., 2015) Copyright (2015) Journal of Raman Spectroscopy).

A significant challenge in developing a POU-SERS system lies in the intermittent cooling of the photodetector to reduce thermal noise and operate at longer wavelengths. The use of spectral filters, such as notch filters or bandpass filters, can prevent heating, and anti-reflective optics may minimize the laser light absorption and thus the heating of the photodetector. At higher temperatures, thermally generated noise can affect the performance of semiconductor materials used as photodetectors, which require intermittent cooling to ensure stability and high spectral resolution. The cooling of the photodetector in a Raman system reduces the dark noise, which originates from the dark current (Stiff-Roberts, 2011). As these dark currents influenced by the photodetector temperature, which requires intermediate cooling it can aid in reducing background noise and improving the SNR of the sensing system. Most photodetectors employ thermoelectric coolers (TECs) or cryogenic cooling using liquid nitrogen. TECs utilize the Peltier effect, in which a voltage applied across two dissimilar semiconductors creates a temperature gradient. One side serves as a heat sink (at a higher temperature), while the other end cools where the photodetector is attached (Lundgaard & Sigmund, 2019). In the case of cryogenic cooling mechanism, liquid nitrogen at -196° C is used. A double-walled dewar flask with a vacuum between the two walls is filled with liquid nitrogen on the outer surface of the flask, and the photodetector was placed inside the chamber. The outer surface absorbs heat from the inner chamber until equilibrium was achieved. At equilibrium, the photodetector can operate efficiently at lower temperatures, resulting in reduced dark noise and improved SNR (Hubbs, 2000), which are crucial for achieving high-quality measurements in Raman spectroscopy and other applications.

3.4 Allied optical components

The key components and alignment in Raman spectrometers can be optimized and miniaturized on the basis of the type of sample matrix, sensing application, and deployability. Some essential characteristic features of POU-SERS systems include portability, rapid analysis, multiplex detection, user-friendly interfaces, and customization and integration with intelligence-enabled technologies. Other considerations include the choice of stable excitation with high laser power and minimal photobleaching, the selection of optimal spectral filters, lenses and mirrors with higher NA, and a dispersion component, such as a prism. Spectral filters, such as bandpass filters and notch filters, are employed for light transmission of specific wavelengths. In POU-SERS systems, spectral filters are crucial for blocking Rayleigh scattered light, reducing background noise, and reliable detection. In addition, the high NA of the mirrors and lens assists in efficient light collection and precise focusing owing to their parabolic structure that minimizes the loss of scattered light. Furthermore, dispersion components, such as diffraction gratings and prisms, are used to split the incident light into its spectral components to obtain accurate and high-resolution Raman spectra to differentiate the Raman shift. Precise consideration of these challenges and optimization of the systems can assist in reliable, high-performance, and on-site sensing applications.

5.1 Hotspots

SERS hotspots are localized nanozones with intense plasmonic fields that enhance the resonance and thus the Raman scattering. The interaction of incident light with plasmonic nanostructures results in a highly concentrated and localized EM field, and an exceptional enhancement in Raman scattering is termed a hotspot (Freeman et al., 1995). This interaction requires the presence of a target analyte near the plasmonic field of the nanostructure. SERS hotspots assist in the amplification of Raman scattering by multiplicative enhancement of the plasmonic field of nanostructures, resulting in synergistic effects and spatial localization facilitating selective enhancement of signals (Qin et al., 2006) from the hotspot and reduces the background interference (Le Ru & Etchegoin, 2004). Dyanmics of the structure and orientation (Mulvaney & Keating, 2000) of the plasmonic nanostructure and their aggregation affects the dimensionality of the nanosubstrate (Futamata et al., 2003) and thus the signal enhancement. The enhancement factor in the hotspots was calculated as the ratio of the SERS intensity (I_SERS) to the Raman intensity (I_Raman) (Vo-Dinh, 1998). The Raman scattering intensity is mathematically expressed as the square of the electric field energy of the incident light (│E│²) and expressed as (│E│⁴) in the proximity of a plasmonic nanoparticle (Tian et al., 2002). The generation of hotspots is crucial for the design of SERS materials and platforms to improve nanoscale light confinement (Hao & Schatz, 2004). Hence, extensive research is being pursued on the synthesis and incorporation of nanomaterials ranging from 0D to 3D for engineering hotspots. Table IV provides an overview of the common nanostructures used in hotspot generation, and some unique properties that assist in higher signal enhancement.

Hotspot engineering is crucial for the development of SERS-active substrates with nanoscale light confinement in specific localized regions of 0D to 3D materials. In the case of 0D materials, such as quantum dots, a plasmonic field structure with confined dimensionality is not possible (H. P. Lu, 2005). However, quantum confinement of discrete energy levels can generate localized field that can be engineered to generate hotspots (Johnson, 1995), and the near-field effect can tailor the plasmonic field magnitude and structure (Pereyra & Ulloa, 2000). The limited plasmonic field of 0D materials cannot detect larger analytes. Common domains of applications include the detection of small molecules (Weisbuch et al., 2000) and biomolecules (Chan & Nie, 1998). In the case of 1D materials, such as nanorods and nanowires, a greater plasmonic field is observed at the tip curvature, which contributes to longitudinal plasmon absorbance (Clapp et al., 2004). Regularly ordered alignment (Nikoobakht et al., 2002), surface roughness (Fan et al., 2004), and tip engineering (Fan et al., 2004) of 1D materials can be used to tailor plasmonic field structures. These materials are mostly employed for the detection of small molecules (Kneipp et al., 1998), inorganic molecules (Niemeyer, 2001), and heavy metal ions (Muniz-Miranda & Sbrana, 2001) in biological or non-biological samples. For 2D materials such as nanoparticle films, graphene, and graphene oxide, the plasmonic field structure is determined by the nanogaps or crevices of adjacent nanosheets or nanoparticles (Grigorenko et al., 2012). Cumulative surface plasmon polaritons (SPP) can enhance the local field by coupling with 1D and 3D materials to tailor the plasmonic-field structure (Zayats et al., 2005). 2D materials are commonly used for the label-free detection of biomolecules. In the case of 3D materials such as microfabricated nanoholes, nanopillars, and nanopyramids, the plasmonic field structure is generated by bulk plasmon resonance (Wei, 2004). Optimization of the 3D material geometry and customizing the dielectric environment can tailor the plasmonic field structure, and thus assist in hotspot engineering. Three-dimensional (3D) materials are commonly used in the detection of biological and non-biological organic molecules, food sample analysis, and forensic analysis.

5.1.1 Hotspot engineering with 0D & 1D materials

Research in the past decades regarding hotspot engineering of 0D and 1D nanostructures has focused on the synthesis of shape-controlled metallic (Au/Ag/Cu) nanoparticles. 0D materials are nanostructures lacking dimensionality and show quantum confinement effects, whereas 1D materials are relatively larger or elongated nanostructures that generate hotspots by phonon scattering and electron confinement. Common 1D materials include nanorods, nanotubes, and nanowires with high aspect ratios that generate well-defined hotspots. These 1D nanostructures exhibit the lightning rod effect by concentrating on the localized EM field at the tip-surface curvature. This results in a longitudinal enhancement mode of the plasmonic resonance (Cardinal et al., 2017) and absorbance at longer wavelengths (700-1000 nm). Arrays of 1D metal nanoparticles are promising nanoscale optical devices that orient and guide electromagnetic energy (Quinten et al., 1998). An interesting study by Vaidya et al., the use of free-standing fibers of a Au(I)-based coordination polymer (CP) for 4-MBA detection was described. The [Au (SPh)]_n CP flexible fibers are hydrophobic and exhibit high chemical stability under harsh acidic and basic conditions because of the phenyl rings and strong Au(I)–S interactions. Furthermore, calcination can produce a composite, resulting in the formation of AuNPs on CP fibers. Because of the plasmonic resonance of AuNPs, this composite material showed high sensitivity, as demonstrated by SERS (Vaidya et al., 2020). Figure 8(i) shows the 1D gold(I)-thiophenolate [Au (SPh)]_n deposition on a polymer, XRD patterns, and photographs of the modified coordination polymers. Despite the extensive use of electron-beam lithography (EBL)-fabricated 1D metal nanoparticle arrays with defined spacing, the large-scale fabrication of 1D arrays and organized structures is essential for practical applications. Other methods, such as chain assembly synthesis using solution-based protocols, exhibit interesting plasmonic properties. For example, hollow Au nanoparticle-based chains with cobalt nanoparticle (CoNP) chain templates have been assembled using magnetic fields (J. Zeng et al., 2007). Other examples of nanostructures assembly include the ligand exchange method using mercaptoethyl alcohol (MEA) (S. Lin et al., 2005) or cetyltrimethylammonium bromide (CTAB), a cationic surfactant (Dai et al., 2006). For example, as shown in Figure 8(ii), the etching of Ag nanowires using a H₂O₂/NH₃ mixture roughens the nanowire surface resembling beads-on-a-string, which improves the SERS activity across the surface area by 10-fold, while (Goh et al., 2012) another common 1D nanostructure is face-to-face nanodisk arrays fabricated by on-wire lithography comprising cylindrical nanopores, using porous alumina membranes (J. Qin et al., 2005). Therefore, research suggests that the decrease in SERS enhancement is due to excitation in the crevices of nanostructures, for which field enhancement is associated with the nanoscale roughness of the metal surface. Despite significant progress in optimizing protocols for nanoparticle hotspot engineering, these nanoparticles do not possess the intrinsic property of serving as efficient SERS platforms because of the limited SERS active area and insufficient EM hotspot strength required for ultra-trace sensing. Future studies should aim to improve nanoparticle efficiency and enhance SERS signals by optimizing the interparticle distance or using nanostructures with high-order dimensionality.

Figure 8. Hotspot engineering with 0D & 1D materials for SERS signal enhancement (i) (A) TEM images of the composite material obtained after calcination of [Au(SPh)]n fibers at 230^oC for 1 hour, (B) Photographs and (C) Powder X-ray diffraction (PXRD) patterns fibers calcined at various temperatures and time durations, (D) Comparison of the Raman spectra of 4-mercaptobenzoic acid (4-MBA) (5 x 10⁻² M), the fibers calcined at 230^o C–1 h and a solution of 4-MBA with the fibers calcined at 230^o C–1 h, and (E) a zoomed in version of the spectra. (This figure has been reproduced with permission from (Vaidya et al., 2020), Copyright from (2020) Journal of Materials Chemistry) (ii) (a, b) Atomic force microscopy (AFM) images of the synthesized and chemically etched Ag nanowires, (c) Average EF per pixel of etched and synthesized Ag nanowires and Ag film. (This figure has been reproduced with permission from (Goh et al., 2012), Copyright (2012) Langmuir).

5.1.2 Hotspot engineering with 2D materials

The organization of plasmonic nanoparticles in ordered 2D arrays significantly initiates the plasmonic coupling of adjacent nanoparticles, generating uniform EMF enhancement. This will likely enable the design of repeatable SERS systems. 2D hotspot engineering can be achieved using a top-down or bottom-up approach. A recent strategy for engineering hotspots with 2D materials is graphene-enhanced Raman scattering (GERS), which results from the deposition of exfoliated graphene on a SiO₂/Si substrate. The characteristic electronic structure and high electron density of graphene can significantly improve EM interactions (Schultz et al., 2014). Additionally, the first layer effect of the molecules adsorbed on the graphene surface caused by its high surface area and charge transfer can substantially improve EM interactions (Ling & Zhang, 2010), thereby enhancing Raman scattering. In a study by Yu et al., mildly reduced graphene oxide (MR-GO) was drop-casted on a 300 nm SiO2/Si substrate for the detection of Rhodamine B (RhB), which displayed a good EF of 10³ and LoD of 10⁻⁸M. Figure 9(i) Graphical abstract of the 2D MR-GO substrate for Rh B detection and corresponding SERS spectra (X. Yu et al., 2011). Another interesting strategy is the use of graphene-noble metal substrates that can provide substantial SERS enhancement by the coupling of GERS with the plasmonic effect of nanostructures. An interesting study by Xu et al. fabricated a novel SERS substrate by depositing Ag and Au nano-islands on the backside of a graphene monolayer (1LG) for the detection of R6G molecules on the non-coated side of graphene (W. Xu et al., 2012b). Furthermore, the geometry of plasmonic nanostructures deposited on graphene was found to alter the nanoparticle assembly. A recent study by Zhang et al., in 2017, demonstrated the fabrication of gold triangular nanoarrays (Au TNAs) on graphene for the detection of Hg²⁺ in water and sandy soil samples. The use of AuTNAs for substrate fabrication improved the thermal stability and further deposition on the graphene monolayer, which enhanced the SERS signal, facilitating improvement in SERS sensitivity with an LoD of 8.3 nM. A schematic representation of the Au TNA/graphene/Au NP fabrication process and its effect on the SERS spectra is shown in Figure 9(ii) (X. Zhang et al., 2017). Furthermore, other 2D materials, such as hexagonal boron nitride (h-BN), can also be used for hotspot engineering because of its structural analogy with graphite (Pakdel et al., 2014). Signal enhancement using h-BN was different from that of graphene, as variations in the charge transfer process of h-BN do not affect the Raman intensity. Kim et al., in 2016, utilized h-BN to insulate Au SERS substrates. R6G Raman signals were stronger for h-BN/Au/SiO2 than for h-BN/SiO2 and Au/SiO₂ (G. Kim et al., 2016a). Therefore, the use of nanostructured sheets, graphene, and h-Bn as 2D materials assists in hotspot engineering and the enhancement of Raman signals. Although 2D materials possess unique benefits, the use of 3D materials for hotspot engineering is expected to provide further enhancements.

Figure 9. Hotspot engineering with 2D materials for SERS signal enhancement (i) Graphical representation of the 2D mildly reduced-graphene oxide (MR-GO) substrate for Rhodamine B detection and the corresponding SERS spectra depicting increased SERS intensity (This figure has been reproduced with permission from (X. Yu et al., 2011), Copyright (2011) ACS Nano), (ii) (A) Schematic representation of Au TNAs/graphene/Au NP fabrication. (B) SERS spectra of varying concentrations of 4-NTP absorbed on 9nm thick Au TNAs/monolayer graphene. (This figure has been reproduced with permission from (X. Zhang et al., 2017), Copyright (2016) Nano micro small).

5.1.3 Hotspot engineering with 3D materials

Hotspot generation with 3D materials differs from that with 2D materials in terms of dimensionality, spatial distribution, plasmonic field distribution, and accessibility to the analyte. Common 3D materials for hotspot engineering include nanoporous materials, such as nanoholes, nanoarrays, nanopillars, and nanoparticle aggregates. The integration of top-down and bottom-up strategies, as seen in 2D hotspot engineering, was identified to generate open 3D SERS platforms. In the case of a bottom-up approach, nanoparticle self-assembly is widely used to exploit the interface of two immiscible fluids to improve the assembly of 2D nanoparticle meta-crystals. Figure 10(i) shows the schematics of the developed interfacial self-assembly at the oil/water interface, varying configurations of the Ag octahedral nanostructures, Atomic Force Microscope AFM images of functionalized Ag octahedra, and length of octahedra immersed in the oil phase (Y. H. Lee et al., 2015). 3D hotspot engineering also focuses on designing open structures that improve the accessibility of the laser to the analyte to maximize the SERS response. For example, Udayabhaskararao et al. (2017) used non-close-packed gold nanoparticle arrays with Au and Fe₃O₄ building blocks that displayed improved analyte diffusion into the crystal lattice due to the selective etching of Fe₃O₄ nanoparticles and the resulting SERS signal enhancement, as shown in Figure 10(ii) (Udayabhaskararao et al., 2017). Another approach proposed by Lee et al. (2013) used polymeric films for nanoimprint molds to create porous microcylindrical structures and further used metal nanoparticles by electrostatic self-assembly to generate open SERS-active microcylinders with an EF of 6.5 × 10⁴ (S. Y. Lee et al., 2013). The use of 3D porous microcylinders improves the AuNP loading ability, thus improving the SERS signal by 10-fold in comparison with AuNPs on non-porous substrates. Despite the advancements in hotspot engineering, some persistent considerations include: (1) inadequacy of target/analyte detection at the single-molecule level and (2) uniform hotspot density only with specific affinity of analyte molecules to plasmonic surfaces. Hence, most studies still use Raman probes that possess a greater affinity for plasmonic surfaces or larger cross-sectional areas. To overcome the limitations of hotspot engineering, surface fabrication techniques such as in situ growth of 3D nanostructures may be used by electrochemical deposition and bottom-up in situ growth for an enhanced SERS effect.

Figure 10. Hotspot engineering with 3D materials for SERS signal enhancement (i) (a) Schematic illustration of the developed interfacial self-assembly near the oil/water interface, (b) adopting a planar configuration, C3-octahedra- octahedra aligned edge to edge, C16-octahedra-square superlattice with a pyramidal protrusions with square superlattice, (c) AFM characterization of PDMS with various functionalized Ag octahedra and the Ag octahedra length (%) immersed into the oil phase (This figure has been reproduced with permission from (Y. H. Lee et al., 2015), Copyright (2015) Nature Communications), (ii) (A) Schematic representation of the synthesis protocol (B) The measured SERS spectra of vac₁Au₁₁ (blue), vac₁Au₅ (red), vac₁Au₁ (gray) arrays, as displayed in the TEM images (This figure has been reproduced with permission from (Udayabhaskararao et al., 2017), Copyright (2017) Science).

5.2 Substrate hydrophobicity

Hydrophobicity is the intrinsic property of a material to resist water owing to nonpolar interactions, resulting in poor water solubility. Substrate hydrophobicity can alter the contact angle, which defines the ability of the substrate to maintain contact with the liquid sample matrix. Hydrophilic surfaces possess a high affinity for aqueous sample matrices, thus decreasing the contact angle and increasing the surface contact area of the droplet, resulting in rapid sample evaporation (T. Smith, 1980). In contrast, hydrophobic surfaces have a higher contact angle with the substrate surface because of their quasi-spherical shape and require a longer time for solvent evaporation (Ko et al., 1981). Analyte detection with hydrophobic substrates generates a large EM enhancement owing to the high contact angle of the target analyte containing the sample matrix. The confinement of plasmonic nanostructures in a sessile droplet on a hydrophobic substrate can generate enhanced SERS signals. This behavior was not observed with hydrophilic surfaces owing to the spread-out nature of the water droplet on them, offering significantly less confinement of plasmonic nanostructures. Hydrophobic surfaces aid in the concentration of target analyte molecules near plasmonic nanoparticles within a confined region on the substrate (Sakai et al., 2006). The confinement of the analyte molecules with nanostructures generates an intense plasmonic field that assists in SERS signal enhancement. Common hydrophobic surface fabrication methods include silanization (Péron et al., 2009), fluorination (Y. Chen et al., 2009), vapor deposition (Y. Wu et al., 2007), sol-gel coating (Pilotek & Schmidt, 2003), chemical etching (B. Qian & Shen, 2005), and the use of hydrophobic nanostructures (Akagi et al., 2007).

The substrate wettability can be tailored for SERS-based applications by fabricating hydrophobic and hydrophilic surfaces. For example, a microcontact-printing-based hydrophilic surface was fabricated by Shin et al. in 2002 on a hydrophobic polydimethylsiloxane (PDMS) stamp using hydrophilic silver colloids. The hydrophilic nanostructure coated stamp was pressed against a gold-coated Si substrate with a self-assembled monolayer (SAM) of a thiol-containing moiety to develop silver colloidal patterns (Shin et al., 2002). However, the hydrophilicity of this substrate is dependent on electrostatic and van der Waals interactions, which can severely affect silver colloid adsorption on the gold-coated Si substrate. In another study by Wei et al. (2005), a hydrophilic substrate was fabricated on mica using CTAB-based silver colloids, forming a hydrophilic surface because of the -NH₄⁺ cationic group of CTAB (G. Wei et al., 2005). However, CTAB concentrations lower than 10 μM resulted in no SERS signal, and a higher CTAB concentration may result in the formation of surfactant bilayers, resulting in minimal surface adsorption.

As discussed, hydrophobic surfaces can improve the retention time because of the increased contact angle with the sample matrix. An interesting study by Kyle C. Bantz and Christy L. Haynes in 2009, demonstrated the use of SAMs of alkanethiol and perfluoroalkanethiol on the silver film over nanospheres (AgFON) substrates for the detection of polychlorinated biphenyls (PCBs). Cleaned copper discs were deposited with silica nanospheres and vapor-deposited with Ag to form a 200 nm thick Ag film on the nanospheres. Further, it was treated with 1 mM decanethiol (DT) and perfluorodecanethiol (PFDT) to improve the hydrophobicity of the SAM layer. The sensor demonstrated an LoD of 50 pM PCB within 1 min of 532 nm laser exposure, thus facilitating the distinction of PCBs (Bantz & Haynes, 2009). However, manual agitation of the silica nanospheres cannot ensure homogeneous layer formation, which affects substrate repeatability. Rather, spin-coating the silica nanospheres and analysis with microscopic techniques, such as atomic force microscopy (AFM), may provide insight into substrate surface homogeneity. In addition, the SAM layer assists in the partitioning of PCBs from organic solvents, such as tetrahydrofuran (THF), rather than from an aqueous solvent. However, this may limit the use of AgFON substrates for on-site applications because they are abundantly found in aqueous environmental samples. An interesting study by Gentile et al. (2010) developed a micro/nanopatterned superhydrophobic sensor to detect and differentiate biomolecules. Figure 11(i) shows silver grains coated with regularly ordered disk patterns comprising cylindrical micropillars on a Si wafer obtained by optical lithography aiding in the SERS enhancement. A thin Teflon (C4F8) polymer film was coated on the Si wafer to ensure hydrophobicity, which increased the apparent contact angle from 150° to 175°. This hydrophobic SERS sensor exhibited an LoD of 10⁻¹⁸ M with a sample volume of 5 μL of R6G (Gentile et al., 2010). However, the sensitivity may be significantly affected by the reactive-ion etching process, which in turn alters the diameter and height of the Ag mask. Hydrophobic metal surfaces are synthesized by the construction of a micro/nano-metered structure, followed by surface modification with low surface energy molecules (B. Su et al., 2010).

Figure 11. Effect of substrate hydrophobicity on SERS signal enhancement (i) (A-B) MicroPillars of silicon containing ordered hexagonal lattices, coated with electroless grown silver aggregates, (C) SEM image of the residual Rhodamine 6G (R6G) by the end of solvent evaporation, and (D) Characteristic Raman spectrum of R6G. (This figure has been reproduced with permission from (Gentile et al., 2010), Copyright (2010) Microelectronic Engineering), (ii) Graphical illustration of a droplet and its evaporation at specific solute deposition spot due to the hydrophobic substrate surface. (b) SEM images of the drop diameter and the solute suspended on the nanopillars. (c) Recorded contact angles during evaporation at varying time intervals, (e) Measurement of Raman map of rhodamine and (f ) the associated Raman spectrum (This figure has been reproduced with permission from (De Angelis et al., 2011), Copyright (2011) Nature Photonics).

The contact angle of the substrate affects the analyte distribution around a plasmonic hotspot and thus determines the effective SERS enhancement. Another property of the substrate surface, superhydrophobicity, can be obtained with contact angles higher than 150° (Reyssat et al., 2006). Optical lithography-based superhydrophobic surfaces are created to adlayer plasmonic nanostructure arrays by micro/nanofabrication. The structured arrays were coated with a thin Teflon film via reactive ion etching. The deposition of a droplet of aqueous R6G solution on the superhydrophobic surface allows it to concentrate and precipitate to a confined area near the Ag nanostructures owing to substrate hydrophobicity. Figure 11(ii) depicts a droplet of a specific solute deposited on a hydrophobic substrate made of nanopillars. The use of a hydrophobic nanopillar substrate resulted in an increase in the SERS intensity, as shown in Figure 11(ii) and (f ) (De Angelis et al., 2011). The use of a superhydrophobic plasmonic substrate such as silver-decorated polystyrene (PS) nanotubes is highly efficient, with an LoD of 400 ppt for crystal violet (Lovera et al., 2014). Jayaram et al. used a hydrophobic Ag-decorated ZnO nanostructure thin film with a contact angle of 163°. The as-prepared SERS substrate exhibited an LoD of 10⁻¹⁰ mol/L for the detection of R6G (Jayram et al., 2015).

5.3 Substrate periodicities

Periodicity is defined as a highly ordered array or pattern on the substrate surface that ensures hotspot uniformity, with SERS enhancement up to several orders of magnitude higher than that of disordered metal-nanoparticle films. This section focuses on the use of ordered 1D and 2D SERS substrates and the corresponding SERS enhancements. Periodicity using larger metal nanoparticles is challenging because of the increase in long-range van der Waals forces due to the increase in particle size, thus preventing the formation of 2D periodic structures. The tuning of van der Waals attraction was observed using a proper surfactant, ensuring the close packing of larger nanoparticles. For example, the use of calixarene as a surfactant provides greater repulsive forces for the fabrication of highly ordered larger-sized AuNPs, as shown in Figure 12(ii) (A. Wei, 2006).

Figure 12. Effect of substrate periodicity on SERS signal enhancement (i) (A) 1-3: C₁₁ resorcinarene derivatives (Left); C₁₂ thiol capped Au nanocrystals, resorcinarene 1, and cavitand 3 (control), (B) SERS spectra of resorcinarene 6 measured from Au nanoparticle arrays, with a 785 nm laser, (C) (a) G values with excitation wavelengths at 1064, 785, 647 nm; (b) G values with 785 nm using varying solid angles, as determined by the numerical aperture (N.A.) of the collection objective. (This figure has been reproduced with permission from (A. Wei, 2006), Copyright (2007) Chemical Communications), (ii) (a) Schematic illustration of the self-assembled AuNP metafilm (b) SEM image of the fabricated metafilm with 3ml AuNP solution (c) Thiram SERS spectra (1000-0.5 ppm) obtained from a AuNP metafilm by swabbing the orange surface (This figure has been reproduced with permission from (N. Yang et al., 2019) Copyright (2019) Langmuir).

The periodicity of metafilms created by self-assembled nanostructures at the liquid-liquid interface owing to density differences can be utilized for signal enhancement. Metafilms are customizable nanofilms fabricated using precisely structured nanoparticles with unique optical properties that generate highly localized EM hotspots. Yang et al. synthesized a flexible SERS metafilm with self-assembled AuNPs at the water-toluene interface for the detection of thiram, a commonly used moderately toxic fungicide, on orange peel, as shown in Figure 12(i). The metafilm obtained after the evaporation of toluene exhibited high uniformity owing to its ordered nanostructure arrangement. The sensitivity of metafilm was tested using crystal violet as the SERS probe. The sensitivity of the metafilm with 1 mL and 6 mL was low because of the large vacant spaces and due to overspread and close packing, respectively. The metafilm with 3 mL showed the best enhancement, with an LoD of 0.5 ppm thiram (N. Yang et al., 2019).

In addition to chemical synthesis and self-assembled structures, surface fabrication techniques assist in the formation of 2D periodic substrates. Gong et al. employed plasmonic cavity lens lithography for the fabrication of graphene and silver nanohole arrays for the detection of R6G in standard samples. A 100 nm thick silver film was deposited on a quartz substrate, followed by spin coating of the photoresist and another layer of silver film. The substrate was UV-cured with a chromium (Cr) mask, followed by Ag film removal for photoresist development. The pattern was then transferred to the bottom Ag layer by dry etching. The developed SERS-active substrate exhibited an LoD of 10⁻¹¹ mol/L of R6G and an EF of 10⁷ (T. Gong et al., 2019). In another study by Bi et al., an SERS substrate was fabricated using electron-beam lithography (EBL) to detect crystal violet from standard samples. Polyvinylpyrrolidone (PVP) dissolved in ethanol was mixed with chloroauric acid and spin-coated onto the Si substrate. EBL was used to generate nanopatterns of AuNPs on the Si substrate. Polyvinyl alcohol (PVA) gel was then spin-coated onto the nanopatterns. Following the baking and solidification of PVA, the gel was peeled off with the AuNP pattern transferred onto the gel. The fabricated PVA gel with AuNP patterns was used to analyze the sensitivity of the Crystal Violet (CV) probe molecule, which showed an LoD of 10-5 M and an enhancement factor of 9.8 × 10⁵ (Bi et al., 2019). In conclusion, the periodicity of the substrate was observed to improve the surface plasmonic field density and, thus, the EF of the substrate. Furthermore, specific optimizations with physical, chemical, or biological methods may ensure periodicity at the nanoscale, but may not ensure repeatability of the substrate owing to irregular distribution or uneven surface fabrication.

6.1 Paper-based substrates

Paper-based substrates are gaining attention owing to their customizable, biodegradable, and biocompatible properties, and their scalable use in the development of consumer-oriented products. Most paper-based substrates are made of cellulose polymer, which is composed of a linear structure of a few to hundred 1 of 4-linked D-glucose monomers (Shaik et al., 2022). Other common polymers in paper include hemicellulosic paper (Kaushik & Moores, 2016; Xiang et al., 2022), lignin-based paper (Klemm et al., 2005; Mahmoud & Zourob, 2013), and bacterial cellulose-based paper (Basta & El-Saied, 2009; Xiao et al., 2023). However, hemicellulose exhibits high water solubility, (Credou & Berthelot, 2014),whereas lignin-based paper undergoes rapid oxidation in the presence of air, leading to degradation of the substrate (Małachowska et al., 2020). Bacterial cellulose is known to lose its flexibility upon drying, which may serve as a potential limitation for its extensive use in paper-based sensor substrates (Provin et al., 2021). Therefore, cellulose-based paper has been extensively used for the development of paper-based substrates. For instance, in a study conducted by Romo et al. (2021), the intrinsic properties of cellulosic paper, such as porosity, hydrophilicity, and mechanical strength, were exploited to fabricate SERS-based substrates for cell culture applications. The inherent ability to absorb fluid by capillary action and its porous nature contribute to the adhesion and migration of cells (Romo-Herrera et al., 2021). These properties make cellulosic paper an excellent substrate for various applications, including the development of paper-based biosensors. In addition to low-cost, large-scale production and disposability are some major benefits of cellulose-based paper (S. Wang et al., 2012). The physical and chemical properties of cellulose paper have been utilized for the integration of nanoparticles and surface engineering to develop disposable paper-based substrates for SERS-based sensing applications. They are affordable, highly useful in resource-limited settings, and user-friendly alternatives with considerable sensitivity and specificity (W. Zhao et al., 2008). Additionally, the ease of loading liquid samples will further improve sensitivity by restricting the sample flow to a small sensing region, which will aid in lateral flow assay (LFA)-based sensing techniques (W. W. Yu & White, 2010).

In an interesting study by Dong-Jin Lee et al., in 2019, a method was developed for the detection of thiram, using paper as the sensor substrate. Initially, the surface of the filter paper was modified with a diluted PDMS solution to achieve hydrophobicity and confine the porosity, followed by drop-casting gold nanoparticles arranged on graphene oxide (AuNPs@GO) flakes to fabricate a hydrophobic paper (h-paper). The PDMS on the filter paper increased the contact angle (CA) to ~ 128.4° and decreased the surface contact area with an extended retention time of the AuNPs@GO solution. The developed sensor showed an LoD of 1 μM and a linear detection range of 10⁻³-10⁻⁶M, using a Raman spectrometer equipped with a 785 nm excitation wavelength (λ_ex) laser and a laser power of 2.4mW (Dong-Jin Lee & Dae Yu Kim, 2019). Figure 13 depicts the detailed synthesis of gold nanoparticles (AuNPs) and the steps involved in substrate modification. However, this study did not include validation using real samples. Additionally, the presence of interferents or other structurally analogous molecules can result in false positives. This limitation can be overcome by using a bioreceptor specific to the analyte, which can improve the specificity of the sensor, even in the presence of complex samples. However, all these cellulose-based paper substrates are associated with a major limitation of autofluorescence because of the presence of organic materials, lignin, and additives, such as calcium carbonate (CaCO₃), alkyl ketene dimer (AKD), polyacrylamide-based resins, bleaching agents, and antioxidants. Autofluorescence may be mitigated by the use of higher-wavelength lasers and optics that block fluorescence or baseline autofluorescence. Other limitations include moisture sensitivity, flammability, and lower chemical and temperature resistance, which can be surpassed by other disposable substrates such as polymer-based and silica-based substrates. Recent advancements in analyte detection using paper-based substrates are presented in Table V.

Figure 13. (a) Photoreduction process for the synthesis of AuNPs arranged on graphene oxide (GO) flakes (AuNPs@GO). (b) The hydrophobic (h)-paper-based SERS substrate was fabricated by drop-casting 50 μL AuNPs@GO solution onto the h-paper (This figure has been reproduced with permission from (Dong-Jin Lee & Dae Yu Kim, 2019), Copyright (2019) Sensors).

6.2 Fabric-based substrates

Fabrics can serve as excellent substrates for SERS-based detection owing to their uniform interwoven fibers, durability, and resistance to wear and tear (Satani et al., 2023). They possess excellent porosity, microfluidic behavior, and enhanced surface area; thus, they (Ismail, 1991) make ideal materials for efficient contact with samples. In addition, the 300–500 μm coarse fibrous structure in plant fibers (C. H. Lee et al., 2021) can be modified by coating or trapping plasmonic metal nanoparticles for the development of a SERS-active substrate (Moon et al., 2011). However, porosity may be a limitation as it may lead to the requirement of large sample volumes, which can be overcome by optimizing the sample microfluidic flow, which is regulated by the capillary pressure and wicking force of the material. Benltoufa et al. suggested that the capillary kinetics in a knitted cotton fabric depended on the geometry of the fibers (Benltoufa et al., 2008). Furthermore, Bhandari et al. suggested that the wicking rate of yarns is significantly regulated by the number of fiber twists/inch (Bhandari et al., 2011), while Das et al. suggested that the wicking rate significantly decreased with an increase in twists per inch under the effect of gravity (B. Das et al., 2011). Other recent studies have suggested that microfluidic flow can be optimized by surface engineering methods such as patterning or channeling the hydrophilicity of fabrics. Furthermore, hydrophobic cotton fabrics can be used for SERS-based applications by dip coating (Mahltig, 2011), spin coating, (L. Xu et al., 2012a) or printing with hydrophobic materials (Noppakundilograt et al., 2010) to obtain varied levels of hydrophobicity that regulate the sample flow. Wearable sensors that can be integrated into clothing can be used for the continual assessment of patient health. Electronic textiles that can monitor physiological parameters are becoming more common, and smart textiles that can monitor chemical biomarkers are required (Lu et al., 2016).

Robinson et al. conducted an interesting study in 2014 for the detection of 4, 4’-Bipyridine (4, 4’-BiPy) using novel fab-chips made of Zari fabric (metal coated over silk fabric). The Zari fabric-based chip is composed of silver nanoparticles (AgNPs) for roughening the Zari yarns, thereby improving the SERS signal. 4,40’-BiPy was used as a probe to assess substrate uniformity and sensitivity, while the detection of adenine bases in DNA provided solid evidence for SERS detection of biological molecules on treated Zari fabric. However, the data suggest a relatively higher relative standard deviation (RSD) for the substrate and large errors at lower analyte concentrations. In addition, no real sample testing was conducted, which can significantly alter the results owing to pH, the presence of other biomolecules, analogous molecules, and other organic loads (Robinson et al., 2015). In another study, Gong et al. fabricated gel-assembled AgNPs, and their in situ growth on cotton swabs was used for the detection of 2, 4-dinitrotoluene (2,4-DNT) in standard organic solvents, as shown in Figure 14. The cotton Q-tip was transformed into a surface-SERS-active substrate (SERS Q-tip) using a bottom-up strategy. The sensitivity of this direct swab-sensing method was tested with Nile blue A (4-NBA) and further explored for the detection of 2,4-DNT. The swab detection method showed exceptional sensitivity, with an LoD of ∼1.2 ng/cm² and a shelf life of ~30 days (Z. Gong et al., 2014). Table VI discusses some recent advancements in analyte detection with exceptional LoDs and some critical insights for improvement in sensitivity. However, most fabric-based substrates do not optimize the microfluidic flow because of (i) varying inter-fiber sizes between the threads and (ii) limited surface area that does not allow printing. Hence, the use of rigid, inert, non-porous materials, such as polymers or silica, can facilitate the fabrication of surface chemistries.

Figure 14. Schematic illustration of the SERS Q-tip on cotton swab for 2, 4-DNT detection and the associated SERS spectra.

(This figure has been reproduced with permission from (Z. Gong et al., 2014), Copyright (2014) Applied materials & Interfaces).

6.3 Commercial polymer-based substrates

Polymers are large molecules that consist of long chains or networks of covalently bonded monomers. Some commonly used polymers include polyethylene terephthalate (PET) (Y. Wang, Jin, et al., 2018a), poly (methyl methacrylate) (PMMA) (Huebner et al., 2012), (PDMS) (C. Qian et al., 2015), and polyvinylidene fluoride (PVDF) (Szymborski et al., 2014). They possess characteristic features, such as durability, flexibility, and versatility, enabling their broad-spectrum applications, including medical diagnostics and environmental monitoring. Recently, the increased use of polymer-based substrates has been observed due to their chemical inertness, disposability, cost-effectiveness, and optical transparency(Z. Li et al., 2020), which facilitate the fabrication of disposable SERS substrates.

However, some major considerations in the fabrication of polymer-based SERS substrates are the flexibility and transparency of the polymers. The flexibility of a polymer ensures design versatility, adaptability with integrated circuitry (McAlpine et al., 2005), and irregular sample surfaces for proximity between the substrate surface and the analyte molecules (Restaino & White, 2019). Optical transparency facilitates light penetration (L. Li & Chin, 2020) and suppression of background fluorescence (George et al., 2018). Flexible and transparent substrates that can attach conformally to arbitrary solid surfaces are of increasing interest owing to their in situ detection potential. Among these, PDMS (Park et al., 2017) stands out because of its chemical inertness (McDonald & Whitesides, 2002), leak-proof nature (Abidin et al., 2019), gas impermeability (Nakagawa et al., 2002), and thermal stability(X. Liu et al., 2015). In addition, polymer-based nano composites are gaining attention as hybrid SERS substrates, with a typical composition of synthetic or natural polymers as the host matrix and a filler with 1D nanostructures such as metallic nanoparticles dispersed into a large volume of filler followed by curing (Connatser et al., 2004; Giesfeldt et al., 2005). Furthermore, substrates modified with surfactants or fatty acids, such as alkyl dithiols (Kubackova et al., 2015) and oleic acid, (H. Zheng et al., 2015) have been used as promising disposable-SERS substrates, alongside gold (Shahar et al., 2017) and silver nanoparticle-doped or modified polymers (H. Zheng et al., 2015).

In 2017, Singh et al. developed tantalum (Ta)-doped TiO₂ nanofibers (TNFs) in alcoholic solutions via electrospinning with PVP for the detection of methylene blue. The 5% Ta in the TNFs displayed an improved photocatalytic activity of 2.2 times with solar light irradiation because of the newly induced energy levels in TiO₂ (Ti³⁺). These energy levels improve the photoexcited charge division and promote charge transfer, resulting in a higher chemical enhancement of the substrate. However, as the average signal enhancement is primarily influenced by charge transfer, the use of other photocatalytic nanoparticles may improve the sensitivity. In addition, as the homogeneity of Ta in TNFs depends on the properties of PVP and the parameters associated with electrospinning, the repeatability and reproducibility of the substrate may be adversely affected. Rather, PVP electrospinning should be performed prior to Ta doping of the TNFs (Singh et al., 2017).

As discussed above, the flexibility of the polymer is pivotal for its choice as a disposable SERS substrate. Recent studies have shown the increasing attention paid to flexible platforms owing to their durability and adaptability with irregular surfaces and geometries, aiding in practical applications. In a recent study, Zang et al., in 2021, proposed a strategy for the fabrication of a polyethylene terephthalate (PET) film-based flexible SERS substrate using argon (Ar) plasma etching with the physical vapor deposition (PVD) of gold to produce worm-like Au nanostructures. The as-synthesized SERS substrates exhibited a significant signal enhancement with an EF of 1.2 x 10⁸ and the LoD was calculated as 10⁻⁹ M. However, the repeatability and reproducibility of the developed sensor substrate were not analyzed. In addition, no cross-sensitivity testing or real sample testing was performed, which can vary significantly with the point-of-use application (Zang et al., 2021). Another interesting study on the detection of malachite green (MG) presence on fish was reported by Zhao et al. in 2018. They developed a novel method for fabricating a three-dimensional (3D) flexible SERS substrate using graphene oxide/Ag nanoparticle/pyramidal PMMA (GO/AgNP/P-PMMA), as shown in Figure 15. The pyramidal and flexible 3D PMMA film (P-PMMA) was imprinted from pyramidal silicon with a high curvature and triangular geometry; therefore, it acted as an activity site for heterogeneous detection. The larger field enhancement and improved probe capturing are due to the homogenous development of hotspots, as is evident from the higher SERS intensity of R6G in comparison to the flat PMMA surface. The performance of the developed SERS substrate was validated with the AgNP/P-PMMA substrate and GO/AgNP/flat-PMMA substrate using molecular probes, such as R6G and (CV) (X. Zhao et al., 2018). However, the LoD of the developed sensor varied significantly with the sensing MG on the real sample; therefore, the accessibility of specific detection of MG in a complex organic specimen may be a challenge. The use of a specific bioreceptor may improve the sensitivity and specificity of a sensor, as demonstrated by Li et al. (2018). Biofluid analysis was performed for the specific and quantitative assessment of dopamine in the serum (L. Li et al., 2018). Table VII discusses a few recent advancements using polymer-based substrates for analyte detection with exceptional LoDs and some critical insights.

Figure 15. Schematic illustration of the process for the fabrication of the GO/AgNPs/P-PMMA substrate.

(This figure has been reproduced with permission from (X. Zhao et al., 2018), Copyright from (2018) Applied Surface Science).

6.3.1 DVD-based substrate

Digital Versatile Discs (DVDs) can be efficiently repurposed to generate ordered structures with microscale features that amplify the Raman signal. Commercially available DVDs contain a silver-coated spiral arrangement of rectangular grooves (AgDVDs), which is exploited as a regularly ordered substrate for SERS biosensing. A few characteristics of the substrate include a larger surface area, homogeneity of the substrate, easy customizability, multiplexing, and substrate recycling, which enable the versatility of DVDs for SERS-based sensing. A study proposed by Giuseppe Giallongo et al. detailed the fabrication of SERS substrates based on the electrodeposition of silver nanoparticles on the inner silver surface of a commercial DVD, resulting in AgNPs@AgDVD, as depicted in Figure 16(i). The versatility of the DVD facilitated the customization of the AgNPs and substrate uniformity. The AgNPs@AgDVD substrates showed an enhancement factor (EF) up to 7 × 10⁵ with good reproducibility and repeatability. This method provides a practical alternative for inexpensive disposable substrates for SERS and offers further room for improvement (Giallongo et al., 2011). Commercially available blue ray digital versatile discs (BRDVDs) possess a 320 nm structural periodicity and a channel width of 100 nm, thus generating an ideal structure for entrapping nanoparticles. The BRDVD nanochannel’s sidewalls are composed of polycarbonate (PC) material with a refractive index of 1.58 required for guiding the coupled EM field with the trapped nanoparticles in the channel.

Figure 16. Digital versatile disc (DVD)-based disposable substrate for SERS-based sensing (i). (a) Large-scale SEM images from AgNPs@AgDVD sample, (b) AFM image from AgNPs@AgDVD, and (c) Comparison between literature data of several Ag-based SERS substrates and the results of this study (EF values, reproducibility and repeatability) (This figure has been reproduced with permission from (Giallongo et al., 2011), Copyright from (2011) Plasmonics), (ii) (A) FESEM image of BR(Blue ray) DVD substrate. The inset image shows the distribution of AuNPs on the substrate. (B) Characteristic SERS signal intensities scattered from the mixture of albumin, creatinine and urea when mixed in different ratio (This figure has been reproduced with permission from (Chamuah et al., 2019) Copyright (2019) Sensors and Actuators B: Chemical).

In a similar study by Chamuah et al. in 2019, BRDVD was used as a SERS substrate for the detection of albumin, creatinine, and urea in urine samples. The trapped AuNPs in the BRDVD channel produced a guided mode resonance (GMR) field and an increase in the photon lifetime of the coupled EM field, accounting for the overall increase in the local field intensity, as depicted in Figure 16(ii). The LoDs were calculated as 0.1 μg/mL, 0.2 μg/mL and 0.6 μg/mL respectively, which are well below the normal range and thus meeting the requirements for the analysis in various clinical approaches (Chamuah et al., 2019). The sensitivity and LoD obtained by the BRDVD-based sensor were significant because only isotropic nanostructures were employed. However, AuNP absorbance wavelength coherence with the 785 nm laser may not be prominent for efficient light scattering from the synthesized nanostructure. The substrate exhibited an exceptional shelf life of 45 d. However, no cross-sensitivity testing has been conducted to ensure specificity in the presence of structurally analogous molecules or other interferents in a complex sample. In another study, Nguyen et al. (2019) demonstrated the detection of amoxicillin by drop-casting AuNPs onto the surface of a DVD. AuNPs were synthesized by pulsed laser ablation of gold, resulting in colloidal gold, which was then deposited on the DVD surface in a circular pattern. The polycarbonate layer was removed from the DVD, followed by rinsing with ethanol and DI water before the deposition of AuNPs. The average EF of the AuNPs/DVD SERS substrates was calculated as 10⁶, with an LoD of 0.1 ppm and linear detection range of 0.1-1 ppm (Nguyen et al., 2019). However, the sensor showed poor sensitivity despite having a good EF with the developed substrate. Rather, the use of anisotropic nanostructures, such as gold nanorods, nanostars, or nanoflowers, significantly improves the detection limit and sensitivity. Despite the advantages offered by polymer-based substrates for SERS measurements, they are also associated with some major limitations, such as optical absorption or scattering, background interference, lower signal enhancement, and incompatibility with different types of sample matrices. Therefore, another promising disposable material, glass, has been extensively used in SERS systems.

6.4 Silica-based substrates

Crystalline silica, commonly known as quartz, is the most abundant form of silica in nature, whereas other common forms are fused silica or glass, amorphous silica, colloidal silica, and silica nanoparticles. Silica is beneficial because of its high-temperature resistance, chemical inertness (Vaidya et al., 2020), rigidity (Bruno & Svoronos, 2010), transparency, electrical insulation (B. C. Senn et al., 1999), and biocompatibility (Bayliss et al., 1999). Fused silica/glass slides ensure a stable and inert surface necessary for the chemical modification or immobilization of biomolecules. Conversely, Silica nanoparticle-embedded paper and silica nanoparticle-coated substrates possess exceptional physical and chemical stability and regulate sample flow, which aids in the development of disposable microfluidic devices. These properties will assist in the development of SERS-active substrates for on-site environmental monitoring and biomedical diagnostics. In contrast, porous silica (pSi) is extensively used in adsorption and separation (Steinbacher & Landry, 2014), (bio) sensors (Patel et al., 2006), drug delivery (J.-F. Chen et al., 2004), catalysis (K. Yang et al., 2015), and environmental remediation (Nutt et al., 2005), owing to its large surface area, pore size, and sample distribution. However, fused silica/glass-based substrates are extensively used for the fabrication of disposable substrates owing to their characteristic features such as optical transparency, ease of functionalization, homogeneous surface roughness, and biocompatibility. In 2019, Zhou et al. fabricated an SERS-active substrate on ultrathin glass to detect 1,2-bis-(4-pyridyl)-ethene (BPE). AuNPs were annealed on the glass coverslips by metal evaporation at varying temperatures (350°C, 450°C, and 550°C) and at different time intervals (1,3,6, and 9 h), which resulted in varying thicknesses of the gold films, as shown in Figure 17(ii). The variation in the gold film thickness resulted in different colors, such as dark green (8 nm), light green (6 nm), blue (4 nm), and light blue (2 nm). The developed glass substrate showed an EF of 2.71 × 10⁷ with an LoD of 10⁻¹² M and a linear detection range of 10⁻³ to 10⁻¹² M was observed within 120 s of laser incidence on the aqueous solutions. It also showed excellent repeatability with an R² value of 0.9976 and a good shelf-life of 5 weeks, indicating that these gold coverslips can be a good choice for SERS-based sensors (L. Zhou et al., 2019). However, coverslips coated with 4 nm Au at 550°C showed the highest surface coverage and smallest interparticle distance, which assisted as a 2D ordered array structure, aiding in better sensitivity. Rather, the use of anisotropic nanostructures can significantly improve the sensitivity owing to the edge effect. In addition, no cross-sensitivity testing or real-time detection were performed because they significantly affect the sensitivity with interference from other analogous structures, biomolecules, or organic load in the sample matrix.

Figure 17. Silica-based disposable substrate for SERS-based sensing (i). Graphical abstract of mesoporous Ag-TiO₂ nanocage for biosensing application (This figure has been reproduced with permission from (S. Das et al., 2022), Copyright from (2022) Optical Materials), (ii) (A) SEM images of AuNPs on square glass coverslips coated with 4 nm and annealed for different time periods (a) 1 h, (b) 3 h, (c) 6 h, and (d) 9 h at 550°C, (B) SERS spectra of (1,2-bis-(4-pyridyl)-ethene) BPE molecules of different concentrations (10⁻³, 10⁻⁵, 10⁻⁷, 10⁻⁹, and 10⁻¹² M) using 4 nm gold-coated coverslips annealed at 550°C for 3 h on a hot plate. Inset-photo of a coverslip after the deposition of five different BPE concentrations (This figure has been reproduced with permission from (L. Zhou et al., 2019), Copyright from (2019) Biosensors).

A novel study by Das et al. (2022) fabricated a low-cost mesoporous Ag–TiO₂ SERS substrate on glass. The distinctive cage-like structure of the TiO₂ film resulted in the uniform growth of a spherical and porous Ag film with an average interparticle distance of 10 nm, as shown in Figure 17(i), which aids hotspot generation within a small volume. The TiO₂ nanocage (NC) also increased the effective surface area for analyte adsorption and the Ag–TiO₂ NC structure displayed an enhancement of 10⁸ with the R6G probe using a portable handheld Raman spectrometer. The proposed substrate was recyclable, owing to its photocatalytic activity, upon exposure to UV light for 130 min, which resulted in degradation of the dye molecule (S. Das et al., 2022). Furthermore, the substrate exhibited high sensitivity for detecting urea concentrations up to 1 mM, which covers the critical range of blood urea levels. In yet another study, Furu Zhong et al., 2018 prepared a silver nanoparticle (AgNPs)-coated porous silicon photonic crystals (PS PCs) for the detection of Picric acid (PA) in alcoholic solution. The developed sensor exhibited an LoD of 10⁻⁸ mol/L and a linear detection range of 10⁻⁴ to 10⁻⁷ mol/L. The PS PCs substrates displayed a 3.58 times greater signal enhancement than conventional single-layered porous silicon. Incubation of PS PCs with AgNPs for more than 75 s decreased the inter-nanoparticle distance and improved SERS enhancement (Zhong et al., 2018). However, as the PS PCs substrate was incubated with AgNPs, the homogeneity of the SERS substrate was severely affected. Rather, electrospinning or spin-coating of AgNPs on PS PCs can generate uniform hotspots. Furthermore, no cross-sensitivity testing of the AgNP-coated PS PCs was conducted, as it may significantly affect the specificity of SERS-based detection. Therefore, these studies indicate that, despite the challenges associated with the use of silica nanoparticles or glass slides for SERS-based sensing, they can serve as excellent disposable SERS substrates.

In conclusion, the use of all disposable substrates for SERS-based sensing at the small-scale or industrial level was beneficial because of its cost-effectiveness, ease of customizability, portability, and scalability, despite some challenges in substrate performance. Therefore, disposable SERS substrates are promising for sensing target analytes in food safety, biomedical diagnostics, and environmental monitoring, to address real-world challenges.

7.1 Physical conjugation and its effect on Raman enhancement

A few commonly used surface modifiers in physical conjugation include metal–organic frameworks (MOF), covalent organic frameworks (COF), aerogels, and hydrogels. These surface modifiers are porous structures that facilitate the controlled integration of plasmonic nanoparticles or Raman-active molecules for the efficient enhancement of the EM field and Raman scattering. The controlled addition of plasmonic nanostructures with defined shapes and sizes to these surface modifiers can generate uniform hotspots on the substrate surface. In an interesting study by Qiao et al., a zinc-based MOF, ZIF-8, was coated with AuNPs (AuNPs@ZIF-8) for the detection of volatile organic compounds (VOCs) for early lung cancer diagnosis, as shown in Figure 18. Regulated gaseous flow of VOCs resulted in increased adsorption owing to the use of ZIF-8. Schiff’s base reaction between the 4-ATP dye on the AuNPs and the aldehyde group of VOCs resulted in a detection limit of 10 ppb (Qiao et al., 2018). However, the efficiency and sensitivity of the system may be significantly affected by humidity, which may restrict the adsorption of VOCs onto ZIF-8. In addition, the Schiff base interaction may be due to the interaction with another aldehyde-containing interferents. Table VIII presents some recent studies describing the effect of physical conjugation, their effect on Raman enhancement, and some critical insights for improvement.

Figure 18. (A) Schematic diagram of AuNP@ZIF-8 synthesis, (B) Schiff’s base reaction of capture aldehyde vapors and covalent linkage with the GSPs. (C) SEM of gold superparticles (GSPs) from monodispersed AuNPs (D_avg ≈ 5.8 nm), (D) SEM images of GSPs@ZIF-8 (This figure has been reproduced with permission from (Qiao et al., 2018), Copyright from (2017) Advanced Materials).

7.2 Chemical conjugation and its effect on Raman enhancement

Commonly used chemical surface modifiers include long-chain thiol-containing molecules such as 11-mercaptoundecanoic acid (11-MUA), 1, 4-benzenedithiol (1, 4-BZT), and amine-terminated or carboxyl-terminated dendrimers. Typically, chemical conjugation is used for the immobilization of target analytes or bioreceptors to improve sensitivity, mitigate non-specific interactions, and stabilize the sensor. These techniques can significantly alter the Debye length of the developed sensor (L. Xu et al., 2020d) and generate uniform hotspots on a substrate. They also act as binding sites for different functional groups, such as silanes, amides, phosphoric acids, and carboxylic acids, relative to the available substrates. In an interesting study by Gorbachevskii et al., the local electric field density and the hotspot density were observed to vary depending on the kinetics of Raman dye oxidation with hydrogen peroxide using citrate-capped AuNPs (Gorbachevskii et al., 2018), while another study by Kim et al., in 2016, observed irreproducibility in Raman enhancement by selective interaction between the amine and carboxyl groups of dendrimers and rGO of the Au-rGO complex, respectively (K. Kim et al., 2016b). In another interesting study, Xu et al. developed a compact AuNP-templated nanostructure from a mesoporous silica film (MSF) at the air-water interface, as shown in Figure 19. The increased adsorption of AuCl₄ ^- in the MSF channels resulted in the close packing of AuNPs@MSF and facilitated hotspot creation and SERS enhancement. Real sample detection was done in water, milk, and apple samples and obtained an LoD of 0.79 pg/mL for 2,4-D, 1.04 pg/mL for pymetrozine and 1.21 pg/mL for thiamethoxan for a linear detection range of 0.1 to 1000 ng/mL (Y. Xu, Kutsanedzie, et al., 2020a). Table IX presents some recent studies describing the effect of chemical conjugation, their effect on Raman enhancement, and some critical insights for further enhancement.

Figure 19. (A) Synthesis of AuNPs@MSF at air-water interface, (B) UV-Visible absorption spectra of AuCl₄- silica films before and after UV irradiation exposure (C) SERS spectra of AuNPs @MSF mixed with 2,4-D and 2,4-D solid respectively (D) SERS spectra of AuNPs@MSF mixed with pymetrozine and pymetrozine solid respectively (E) SERS spectra of AuNPs@MSF mixed with thiamethoxam and thiomethoxam solid respectively (This figure has been reproduced with permission from (Y. Xu, Kutsanedzie, et al., 2020b), Copyright (2020) Food Chemistry).

7.3 Biological conjugation and its effect on Raman enhancement

Biological conjugation involves the covalent linking of a specific bioreceptor with a substrate surface or modifier to improve sensor specificity. This is primarily dependent on the terminal functional groups of the stabilizing shells. The bio-conjugation of the ligands on the surface of NPs is facilitated by the formation of an amide bond by carbodiimide activation. Because it offers superior stability, bio-conjugation created by thiol group attachment is regarded as a robust and effective conjugate in SERS. Additionally, electrostatic interactions and the attachment of biotin-streptavidin conjugates to nanoparticles (NPs) have also been utilized in the creation of adaptable nanomaterials (Chauhan et al., 2022).

In an interesting approach by Barahona et al., in 2013, an SERS-based aptasensor on micron-sized polymer particles was used for the detection of malathion. Methacrylic acid monomers and ethylene glycol dimethacrylate co-monomers were subjected to precipitation polymerization to synthesize polymer particles in acetonitrile. The use of acetonitrile facilitates a smaller pore size, whereas methacrylic acid supplies carboxyl groups that bind gold. Controlled aggregation of AuNPs resulted from the conjugation of AuNPs with the polymer using 2-aminoethanethiol. Thiolation was used to connect the aptamers to the nanoparticles. The thiol–gold interaction caused the polymer-AuNP-aptamer complex to adhere to the metal surface. The LOD was 33.3 μg/mL, and the concentration ranged from 3.3-33.3 μg/mL (Barahona et al., 2013). Figure 20 depicts the fabrication of polymer-AuNP-aptamer substrates and the specific detection of 2, 4-DNT. Table X presents some recent studies describing the effect of biological conjugation, their effect on Raman enhancement, and some critical insights for improvement.

Figure 20. (A) Illustration demonstrating synthesis of polymer-AuNP-aptamer substrates for the SERS based detection of malathion (B) Comparison of SERS detection of malathion at 495 cm^-1 peak with polymer-AuNPs aptamer substrates 16.5 μg/mL of malathion (A) 3.3 μg/mL of malathion (B) and blank buffer solution (C) Spectra are offset for clarity.

(This figure has been reproduced with permission from (Barahona et al., 2013), Copyright from (2013) Industrial Biotechnology).

10.1 Small molecules

Small molecules are organic compounds with molecular weights lower than 1000 Da that are fundamental for chemical and biological processes. Common small molecules include antibiotics, pesticides, chemicals, and heavy metal ions. The detection of small molecules in food and water sources is crucial for preventing detrimental effects on the environment and human and animal health. Early detection of these small molecules can assist in formulating the necessary regulatory measures to devise strategies for the prevention of adverse health disorders.

10.1.1 Pesticides and chemicals

The indiscriminate use of pesticides poses a serious threat to consumer health and can possibly lead to soil contamination and, eventually, contamination of water bodies. The chronic and fatal effects on human health are proportional to pesticide exposure time, pesticide concentration due to bioaccumulation, and chemical characteristics of the pesticide(L. Zhang et al., 2014). Thus, the detection of pesticides in food, soil or water bodies is essential to prevent their entry into humans. Conventional strategies, including HPLC and GC-MS, are expensive and require longer time for result acquisition. Other strategies include colorimetric detection, RI sensitivity detection, and Raman spectrum-based detection, which may not detect trace levels of pesticides due to their lower sensitivity, although these standard techniques cannot provide a quantitative estimation of the analyte.

In an interesting study by Dies et al. in 2018, the presence of thiram, a fungicide used in agriculture, was checked in apple juice by SERS using dendritic silver nanosubstrate on microelectrodes by maskless photolithography. It recorded an LoD of 115ppb and a liner detection range of 0.01-100 ppb, on the microelectrode chip by using a preconcentrated nanoparticle solution, as shown in Figure 22(i). The key peak at 686–703 cm⁻¹ was used for identification and quantification (Dies et al., 2018). However, the use of microelectrodes as the substrate with a high-energy laser can ablate the electrode surface, resulting in the generation of surface interferents that significantly affect the sensitivity of the sensor. Another interesting approach by Xie et al. (2020) used bipyramidal AuNPs for trace detection of methyl parathion in apple peels. The senor showed an LoD lower than other substrates, i.e., 98.63 ng/cm², 31.56 ng/cm² and 36.58 ng/cm² on peels of cucumber, tomatoes and apples respectively (J. Xie et al., 2020). The “press -peel off” method of detection is shown in Figure 22(ii) is user-friendly, but the recovery rate is not significant and therefore cannot aid in field deployment. Despite the use of a 785 nm laser and spherical AuNPs, a considerable detection limit with a 10-fold EF was achieved. Table XI presents the specific detection of pesticides and chemicals using disposable substrate-based SERS systems.

Figure 22. Disposable SERS system for specific detection of pesticides and chemicals (i) (a) Preparation of the SERS substrate for the detection of food contaminants (thiram and melamine from apple juice and milk respectively), (d, e) SEM images of the dendritic silver nanostructures, (g-f) SERS spectra of thiram (peak at 1384 cm^-1) and melamine (peak at 695 cm^-1) at varying concentrations, in unprocessed apple juice and milk samples (This figure has been reproduced with permission from (Dies et al., 2018), Copyright from (2018) Sensors), (ii) (A) Schematic of the paper-based SERS substrates for the detection of methyl parathion on the fruit peels surface, (B) SERS spectra of methyl parathion (MP) (concentration range-0.018 μg/cm² to 35.368 μg/cm²) with a paper-based substrate, and (C) SERS spectra of MP on the surface of apple using the paper-based substrate (This figure has been reproduced with permission from (J. Xie et al., 2020), Copyright from (2020) Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy).

10.1.2 Antibiotics

Excessive use of antibiotics leads to antibiotic resistance in patients, rendering the drug ineffective. Commonly used antibiotics include tetracyclines, beta-lactams, fluoroquinolones, amphenicols, and carbapenems. Traces of these antibiotics may be found in food and water sources because of their excessive use in animal treatments that remain as residues or by the use of contaminated water for agricultural practices. Currently, SERS detection methods are used extensively to detect trace levels of antibiotics in food and water sources.

An interesting study was conducted by Shi et al. (2017) for the detection of neomycin in milk using lateral flow immunoassay (LFA)-based SERS. This testing strategy displayed good sensitivity with an LOD of 0.216 ppb and a recovery rate between 89.7%- 105.6% (Shi et al., 2017). Although exceptional sensitivity was observed, the sensor lacked coherence of the laser excitation wavelength mismatch and absorbance of the nanoparticles. The use of monoclonal antibodies could not ensure specificity, as better cross-reactivity with other antibiotics such as enoxacin was observed. In another interesting study by Pinheiro et al. in 2018, gold nanostars (GNS) decorated on magnetite nanoparticles (MNP) were used for the trace detection of tetracycline. This method estimated LOD of 44.4ppb and 444ppm in ultrapure water and mixed aqueous solutions respectively (Pinheiro et al., 2018). However, this substrate sensitivity is primarily dependent on the adsorption of tetracycline, which may significantly vary with Au branches and surface irregularities on MNPs that generate hotspots. Instead, the use of a specific bioreceptor may improve the adsorption capacity and specificity of the substrate. pH studies for tetracycline adsorption showed that pH 5-6 was the most favorable for adsorption; however, pH regulation in complex sample matrices may not be possible during rapid analysis. In a recent study by Riswana et al., gold nanostars (AuNSs) were decorated on a PMMA substrate for SERS-based detection of ciprofloxacin and chloramphenicol in chicken wing samples, as shown in Figure 23. The developed system achieved an LoD of 3.41 × 10⁻¹¹ and an EF of 2.03 × 10⁹ (Riswana Barveen et al., 2022). The use of a novel AuNS synthesis method utilizing a UV-C light-based photoreduction process is simple, yet effective. The substrate showed exceptional sensitivity and multiplexing ability; however, no cross-sensitivity testing was conducted. Table XII discusses the disposable substrates for the SERS-based detection of antibiotics.

Figure 23. (A) Photographs of (a) pure PMMA film, (b) Au/PMMA-1 film, (c) A/PMMA-2 film, (d) Au/PMMA=3 film, (e) Au/PMMA-4 film; (B, D) D Photographs of the SERS-based detection of (B) Ciprofloxacin and (C) Chloramphenicol on chicken wings with flexible Au-NSs/PMMA SERS substrate using 532 nm laser and (D, E) their corresponding SERS spectra (This figure has been reproduced with permission from (Riswana Barveen et al., 2022), Copyright (2022) Chemical Engineering Journal).

10.1.3 Heavy metals & other small molecules

Contamination of food and water by heavy metal ions such as mercury (Hg²⁺), lead (Pb²⁺), arsenic (As²⁺), and cadmium (Cd²⁺) can potentially affect the physiological functions of the nervous system, as well as injure vital organs such as the kidneys and liver. Accidental ingestion of food containing these contaminants increases the risk of heart disease, neurological damage, such as tremors, and impairment of cognitive function. The most toxic heavy metal ion in water sources that affects the environment and human health is mercury (Hg²⁺), and its accumulation in humans leads to neurological disorders. A disposable SERS substrate developed by Yang et al., (H. Yang et al., 2017), as shown in Figure 24, by the deposition of thymine for simultaneous detection of Hg²⁺ and Pb²⁺ ions in drinking water employing oligonucleotide functionalized gold coated polystyrene microspheres (PSMPs). The LoD of the fabricated substrate was found to be 0.1ppb for Hg²⁺ ions, which is far lower than the limit prescribed by the World Health Organization (WHO) (Guo et al., 2023). Despite the excellent sensitivity and ability to specifically detect Hg²⁺ from untreated river samples, cross-sensitivity analysis was not conducted. Table XIII discusses the disposable substrates for the SERS-based detection of heavy metals and other small molecules.

Figure 24. (A) Schematic illustration of the SERS based detection of Hg²⁺ ion on the Au NRs@Thymine (This figure has been reproduced with permission from (H. Yang et al., 2017), Copyright from (2017) Nanomaterials). (B) TEM image of the Au NRs@T adsorbed with Hg²⁺ ions, (C) SERS spectra of the Au NRs@T with varying Hg²⁺ ion concentrations.

10.2 Nucleic acids

Nucleic acids, DNA, and RNA are genetic materials of all living beings that may be utilized as biomarkers. Standard detection methods include southern/northern blotting, reverse transcriptase-polymerase chain reaction (RT-PCR), and microarray-based methods. These methods cannot be used for POU applications because of the need for sample preprocessing, longer assay time/incubation time, high cost, and low sensitivity. Optical and electrochemical detection methods can overcome these limitations, thereby increasing the possibility of on-site nucleic acid detection using SERS-based systems. Recently, Mabbott et al. developed a 3D paper-based microfluidic platform, as shown in Figure 25. The fabricated substrate used malachite green XXX isothiocyanate (MGITC)-functionalized AuNPs coated on Whatman 4 chromatography paper for the specific detection of miR-29a in PBS running buffer. This system showed an LoD of 47 ppm using a handheld Raman spectrometer, thus proving its potential as a POU device (Mabbott et al., 2020). Despite the POU applications, the authors used a CBEX handheld Raman spectrometer, which is highly expensive for integration into a POU-SERS system. Table XIV discusses disposable substrates for the SERS-based detection of nucleic acids.

Figure 25. (A) Detailed schematic of the 3D paper-based microfluidic device for the detection method of miR-29a in a 3D paper-based microfluidic device; (B) Sample readout images with both colorimetric and SERS for samples with (+ve) and without miR-29a (−ve). and (C) SERS spectra representing the 18 pg μL⁻¹ (blue) and 360 pg μL⁻¹ (red) concentrations of miR-29a, highlighting three signature methyl green isothiocyanate (MGITC) peaks.

(This figure has been reproduced with permission from (Mabbott et al., 2020), Copyright from (2019) Analyst).

10.3 Proteins

Proteins are biomolecules that are synthesized through an in-vivo process called translation from genetic material, and may be used as biomarkers or indicators in disease diagnostics as efficient target analytes for SERS-based detection. A strategy was developed by Wang et al. by integrating digital microfluidics with a SERS-based immunoassay for the detection of hemagglutinin of H5N1 influenza virus, as shown in Figure 26. The surface protein was detected in both buffer and human serum samples, and the system showed an LoD of 74 ppb but required less assay time than conventional ELISA and a lower sample volume requirement. The characteristic Raman peaks of 4-MBA were observed at 1071 cm⁻¹ and 1580 cm^–1 (Y. Wang, Ruan, et al., 2018c). Another study by Pinzaru et al. used hydroxylamine-reduced AgNPs to detect lipophilic marine biotoxins, namely okadaic acid (OA), Dinophysistoxin-1 (DTX-1), and Dinophysistoxin-2 (DTX-2) in shellfish tissue. A prominent SERS peak at 1017cm⁻¹ was observed for DTX-1 and DTX-2, which was not observed in the OA fingerprint. OA showed LOD of 200ppm, 71.8 ppm for DTX-1 and 32.6 ppm for DTX-1 respectively (Pinzaru et al., 2018). The developed substrate showed good sensitivity; however, the excitation laser wavelength and absorbance range of the nanostructures were not coherent. Rather, the use of AuNPs with a typical spectral absorbance at 530 nm may improve sensitivity. Table XV illustrates some of the disposable substrates for the SERS-based detection of proteins.

Figure 26. (a) Schematic illustration of DMF-SERS-based immunoassay and bottom side of the DMF chip, (b) Characteristic peaks of 4-MBA on the DMF-SERS chip, (c) DMF chip (side-view) with a droplet on the magnetic beads, (d) Immunocomplex functionalized with SERs tags on magnetic beads, (e) SERS spectra of increasing concentrations (0-5000 pg/mL) of H₅N₁ sample, and (f) Cross-sensitivity of the DMF-SERS method (H5N1, 0.085 nM; PSA, CRP, HbsAg, 0.85 nM; cTnT, 0.135 nM) (This figure has been reproduced with permission from (Y. Wang, Ruan, et al., 2018b), Copyright from (2018) Analytical Chemistry).

10.4 Bacteria, viruses and other whole cells

Microbes, specifically viruses, pose a major hazard to the healthcare sector because of their rapid replication cycles and minimal access to instruments capable of ultrasensitive detection. Real-time detection from complex samples using conventional techniques cannot limit bacterial replication and improve the chances of infection after entering a live host. Thus, there is an urgent need for rapid and ultrasensitive detection in real samples for instant treatment of the host (Sreelakshmi et al., 2024). A label-free bacterial SERS detection method was developed by Gao et al., using the selective growth of AgNPs on the bacterial outer membrane conjugated with aptamers. Improved SERS signal by aptamer@AgNPs complex resulted in exceptional LoD of 1.5CFU/mL using micro-Raman system (W. Gao et al., 2017). Additionally, the use of lateral flow assay (LFA)-based strips for the qualitative determination of the analyte of interest is essential for POC diagnostics, which are cost-efficient, reliable, reproducible, and highly sensitive (Ryu et al., 2010). Wang et al. tested a sandwich immunoassay using antibody-conjugated AuNPs with reporter molecules for bacterial detection, as shown in Figure 27(i). Malachite green isothiocyanate (MGITC) in the test line indicated the presence of target bacteria binding to antibodies on the AuNPs. The test concentration was maintained at 10⁷ CFU/mL for all studies. This rapid detection (~15 min) method showed an LoD of 357 CFU/mL for B. anthracis strain with a minimal sample (R. Wang, Kim, et al., 2018b).

Figure 27. Disposable SERS systems for specific detection of whole cells (i) Schematics of (A) conventional LFA strip (B) SERS-based LFA strip. (C) Images of three strips and (D) SERS signal intensity for Y. pestis, F. tularensis, and B. anthracis. (This figure has been reproduced with permission from (R. Wang, Kim, et al., 2018b), Copyright from (2018) Sensors and Actuators B: Chemical), (ii) (A) Overview of SERS-microfluidic droplet platform for single-cell encapsulation and simultaneous detection of three metabolites produced by a single cell using Fe₃O₄@AgNPs nanocomposite, (B) Fe₃O₄@AgNPs nanocomposite for solo metabolite detection (ATP (731 cm^–1), lactate (1030 cm^–1), and pyruvate (1342 cm^–1), respectively), (C) SERS of cellular metabolites content in SGC, HepG2, and MCF-7 cells (one to five cells from bottom to top, respectively) encapsulated in individual droplet. (This figure has been reproduced with permission from (D. Sun et al., 2019), Copyright from (2019) ACS Analytical Chemistry).

Most recent epidemics are caused by viruses that possess multiple surface proteins and genetic material, which are reliable biomarkers for the detection of their presence during disease prognosis and treatment. A mini microfluidic platform named VIRRION (virus capture with Raman Spectroscopy detection and identification) was developed using arrays of carbon nanotubes with different filtration permeabilities for high-throughput virus capture and rapid SERS detection. This method can be employed for multiple-virus capture and detection on a chip by SERS using a Raman system (Yeh et al., 2020). Another method with labelled detection of the virus was developed by Shen et al., based on the immunoassay method using an LFA strip-based reaction for wild-type pseudorabies virus from clinical samples of pig tissue. This method detects glycoprotein E-specific PCR, which differentiates between the wild-type PRV and gE-deleted vaccine. This method requires approximately 15 min and has an LoD of 5ppm and linear detection range of 41-650 ppm(Shen et al., 2019).

Whole cell detection is considered a need of the hour to identify the presence of circulating tumor cells in the bloodstream, which helps in the early detection of cancer metastasis. This early diagnosis involves a liquid biopsy instead of a tissue biopsy. CTC (circulating tumor cells) detection from millions of blood cells is expensive and minute sensitivity needs advancements in the technique for rapid and sensitive detection techniques. In a study by Wu et al., a SERS method with gold nanostructures of different geometries, that is, AuNSs, AuNRs, and AuNPs, was used for CTC detection in liquid biopsy samples. 4-Mercaptobenzoic acid (4-MBA) dye was used with the nanoparticle to induce stability of bovine serum albumin (BSA) and to increase specificity towards conjugation with CTC; folic acid was used. AuNPs displayed the highest specificity, with an LoD of 1-100 cells/mL (X. Wu et al., 2016). In another study by Sun et al., a SERS-based label-free droplet microfluidic sensor was developed for the detection of multiplex metabolites at the single-cell level, as shown in Figure 27. (ii). For the detection of excessive metabolites (lactate, pyruvate, and ATP) secreted from CTC by magnetic isolation of CTCs from complex samples, 400 nm Fe₃O₄ magnetic microspheres were decorated with 30 nm AgNPs ( D. Sun et al., 2019). Table XVI illustrates some of the disposable substrates used for the SERS-based detection of bacteria, viruses, and whole cells.

In conclusion, despite the insignificant size and molecular weight of the analyte, these disposable material-based sensors showed higher sensitivity when anisotropic nanomaterials were used, which improved the nanostructure’s local electric field due to the edge effect and thus contributed to the amplification of the Raman signal.