Surface Enhanced Raman Spectroscopy: Applications in Agriculture and Food Safety

Abstract

Recent global warming has resulted in shifting of weather patterns and led to intensification of natural disasters and upsurges in pests and diseases. As a result, global food systems are under pressure and need adjustments to meet the change—often by pesticides. Unfortunately, such agrochemicals are harmful for humans and the environment, and consequently need to be monitored. Traditional detection methods currently used are time consuming in terms of sample preparation, are high cost, and devices are typically not portable. Recently, Surface Enhanced Raman Scattering (SERS) has emerged as an attractive candidate for rapid, high sensitivity and high selectivity detection of contaminants relevant to the food industry and environmental monitoring. In this review, the principles of SERS as well as recent SERS substrate fabrication methods are first discussed. Following this, their development and applications for agrifood safety is reviewed, with focus on detection of dye molecules, melamine in food products, and the detection of different classes of pesticides such as organophosphate and neonicotinoids.

1. Introduction

Climate change is manifesting itself with increased temperatures more favorable to spread of pests and diseases. This has resulted in more challenging food production as well as higher numbers of foodborne disease outbreaks. To try to protect yield and ensure food safety, farmers have little choice other than treating their crop with a range of pesticides [1]. Unfortunately, there is more and more evidence showing that these phytosanitary products would do harm to humans, and also lead to the loss of biodiversity, leading to accumulation in soil and deleterious effects on indigenous microbiome [2]. As a result, it is necessary to monitor biological and chemical contamination in food systems throughout the whole food chain.

Traditional detection methods in food and agriculture systems are based on chromatography techniques coupled with mass spectrometry. These methods are time consuming, high cost, and laboratory based. Therefore, implementation of new portable technologies is needed to provide pesticide usage monitoring and regulation.

In this regard, Surface Enhanced Raman Spectroscopy (SERS) has recently attracted a lot of attention as it addresses these requirements. The advantages of SERS include ultrasensitive detection, fast turnover, in-situ sampling, on-site monitoring, low cost, portability of sensors, and the suitability for large-scale screening. Sensors based on SERS have been shown to provide real-time data on soil nutrients [3,4], monitor water run-off and contaminants in water supplies [5], and also detect pesticide residues in food [6,7]. The technique has also found applications in a wide range of sectors including the following: industrial, material, forensic, biological, food safety and electrochemical fields [8,9,10,11,12,13,14,15,16,17,18], with the most common chemical contaminants within environmental and food sectors being pesticides, adulterants, antibiotics and illegal drugs and illegal food dyes.

In this paper, we review the main SERS substrate fabrication methods and bring a special focus on applications for the detection of hazardous chemicals in both food and agriculture.

2. Raman Spectroscopy and Surface Enhanced Raman Scattering (SERS)

Raman spectroscopy was first discovered in 1928 by Sir Chandrasekhara Venkata Raman. It was observed that when light excited a molecule, the majority of light was elastically scattered—this is Rayleigh scattering where E_incident = E_scattered, but also a small fraction of photons was inelastically scattered—this corresponds to Raman scattering where E_incident ≠ E_scattered. When E_scattered < E_incident, it is called a Stokes shift and when E_scattered > E_incident, it is an anti-Stoke shift. Figure 1 shows an energy diagram for the two types of scattering [19,20]. Each peak on the Raman spectrum corresponds to a vibrational mode of the bonds of the molecule under Investigation. Raman scattering can occur in the near ultraviolet, visible, or near-infrared ranges [21,22]. However, Raman scattering efficiency is very low, with typically only 1 in 10⁸ incident photons being Raman scattered. This results in a low signal/noise ratio and can make it impractical and vulnerable to background interferences [23]. Fortunately, the efficiency of this scattering can be increased by using nanostructured plasmonic metals in a technique known as Surface Enhanced Raman Spectroscopy (SERS).

SERS effects were first observed by Fleischman et al. in 1974, when acquiring Raman spectra of pyridine on electrochemically roughened silver [25]. In SERS, lasers are used to excite plasmons in nanostructured metallic surfaces (Au, Ag, Cu, etc.). The SERS enhancement is thought to be two-fold. The first enhancement is based on an electromagnetic enhancement [26,27]. due to localised surface plasmons, while the second originates from chemical resonant energy charge transfer [25,28,29,30]. The enhancement factor (EF) typically reaches 10⁶, which drastically improves the sensitivity of the plasmonic based device.

2.1. Localized Surface Plasmon Resonance (LSPR)—Electromagnetic Enhancement

When laser light excites a metallic nanostructure, the free electrons on its surface oscillate. This collective oscillation is known as Localised Surface Plasmon Resonance (LSPR) [31]. The excited LSPR makes a target molecule highly polarisable and forms a large electric field on the surface. This electric field induces dipole moments in a molecule on the surface of nanostructures, and sequentially produces Raman enhancement. These large, localised electromagnetic fields present around the nanostructures or in nano-gaps between closely-spaced nanostructures are known as “hot spots” [32], see Figure 2. The intensities of the Raman scattered photons are susceptible to enhancement, if their wavelengths are in resonance with the plasmon mode of the nanostructure. LSPR enhancement depends on the size, shape, composition, orientation and local dielectric of the nanostructure. This will be discussed in the next section.

2.2. Chemical Enhancement

The origins of the chemical mechanism for SERS enhancement are still under discussion [33]. One hypothesis is that it is based on the change in polarisability of molecule adsorbed to a metal surface. Upon absorption of the incident laser light, charge transfer occurs between the molecule and the metal [34,35]. It is generally estimated that the chemical effect contributes to a factor of 10² of the total SERS enhancement [36].

2.3. Enhancement Factor (EF)

The enhancement factor (EF) depends on molecular adsorption on plasmonic surface, as well as the morphology, roughness and homogeneity of this surface and the laser wavelength, etc. The calculation of the enhancement factor is $\mathrm{EF}=\frac{{\mathrm{I}}_{\mathrm{SERS}}/\left({\mathsf{\mu }}_{\mathrm{M}} {\mathsf{\mu }}_{\mathrm{S}} {\mathrm{A}}_{\mathrm{M}}\right)}{{\mathrm{I}}_{\mathrm{RS}}/\left({\mathrm{C}}_{\mathrm{RS}} {\mathrm{H}}_{\mathrm{eff}}\right)}$ [37], where I_SERS is the intensity of Surface Enhanced Raman signal, I_RS is the intensity of the normal Raman signal, µ_M (m⁻²) is the surface density of NPs contributing to enhancement, µ_S (m⁻²) is the surface density of molecules adsorbed to NP, A_M (m²) is the surface area of metallic NPs, C_RS (M) is the concentration of the solution used for non-SERS measurements and H_eff (m) is the effective height of the scattering volume.

Experimental values of EFs are typically in the range of 10⁴ to 10⁶. However, electromagnetic “hotspots” are claimed to have provided massive enhancements of between 10¹¹ to 10¹⁴ orders of magnitude of the SERS signal [38]. This theory is still being largely researched, as it is key to single molecule detection, which may be achieved from selective excitation of single molecules [32,39,40].

3. Fabrication of SERS-Active Substrates

SERS substrates need nanostructured metallic surfaces with well-defined distances in the region of 10–100 nm between nano clusters [41]. By decreasing the distance between the nanostructures, the electric field becomes more localized and concentrated, and the corresponding SERS intensity signal increases accordingly. An example of this was discussed by Lee et al. where the distance between metallic clusters was decreased from 30 nm to 10 nm, and an intensity increase of over 200-fold was observed [42]. There are two fundamentally different approaches to the development of SERS-active nanostructures with “hotspots”: bottom-up assembly and top-down fabrication have been used [43].

3.1. Bottom-Up Assembly

Bottom-up approaches refer to the fabrication of nanostructures by chemical synthesis [44,45], colloid aggregation [41,46], electrochemical deposition [47,48,49,50,51], and self-assembly [52,53,54]. These methods have been used to fabricate a variety of nanostructures ranging from a few nanometers to a few hundred nanometers in size. Metal nanoparticles can be synthesised chemically at low cost with tailored geometries such as nanoparticles [55,56], nanowires [57,58], nanospheres [55,59,60], nanorods [61,62,63,64], nanotubes [65,66], nanotriangles [67], nano flower [68], nano-urchins [69], and nanoshells [70,71], see Figure 3. Besides pure metallic nanoparticles, composite materials such as bimetallic or hybrid nanostructures [72], Graphene Oxide/Au nanostars, [73], SiO₂@TiO₂@Ag [74] etc. and other composite material based on molecular imprint (MIP) have also attracted research attention [68,74,75,76,77] and have been reviewed recently [78].

Metal NPs such as gold or silver possess great potential for numerous applications in SERS [86,87]. The most common fabrication of SERS substrates are gold (Au) and silver (Ag) colloids in diameters between 10 and 100 nm, as they yield the greatest enhancements at their “hot spots”. These nanoparticles constitute the fundamental SERS “building blocks” and can be assembled in different ways.

For example, they can be presented in a suspension or sol-gel in the presence of the analyte of interest [88,89,90]. The nanostructures, together with the analyte suspension, can then be drop-casted onto a substrate to create hot spots for Raman enhancement. The disadvantage of this method is that the nanoparticle suspensions must be mixed with the analyte solution for SERS applications [91,92]. However, this drawback was addressed by Yang et al., who grew Ag nanoshells on thiol-modified silica NPs, and deposited them directly on apple skin for analysis [93]. Although there is a greater enhancement observed with these substrates, it is hard to obtain a homogeneous surface to get uniform enhancement. Additionally, they are not suitable for field analysis due to their complex sample preparation steps. In contrast, solid based devices may be more suited for portable and remote sensing, i.e., NPs that are immobilised on a solid substrate [94,95,96,97]. For example, Fan et al. fabricated self-assembled Ag NPs onto glass slides by using 3-mercaptopropyltrimethoxysilane (3-MPTMS) [97] This transformation stabilizes the Ag NPs, avoids the usual aggregation process and produces self-sustaining and portable SERS active substrates. Yu et al. fabricated silver colloidal nanoparticles for SERS analysis but alternatively injected them through a filter membrane, thus entrapping them in the filter. The filter therefore was used as the solid portable substrate. It demonstrated 1–2 orders of magnitude better SERS enhancement than the typical approach [98] In addition, Shiohara et al. fabricated gold nanostars and deposited them onto a polydimethylsiloxane (PDMS) platform for SERS evaluation. They used back side illumination for the detection of selected pesticide on fruit skin [99].

Other supports also add additional functionalities to SERS devices. Optical fiber-based SERS sensors have in this regard generated steady interest as a versatile means of extending SERS for portable field applications [63]. There is also a growing interest in the fabrication of flexible SERS substrates. These substrates are ultra-low cost, disposable, easy to use and highly suitable for on-site sensing applications. Polavarapu et al. fabricated SERS substrates by directly writing on paper using a pen filled with plasmonic nanoparticle inks to detect thiabendazole, which is a fungicide and parasiticide [100]. Lee et al. also fabricated SERS paper substrates impregnated with gold nanorods by dip coating [101] Chen et al. combined adhesive tape and SERS activity of Au nanoparticles to fabricate a “SERS tape” substrate. The Au particles were deposited onto the sticky side of the tape that was used to extract pesticides from different kinds of fruit and vegetable peels [102]. These flexible sensors fabricated by different methods dramatically improve the portability and feasibility of SERS detection for pollutants as a promising technique for both laboratory and field-based detection [102].

Overall, bottom up assemblies have shown very high enhancement but they often give inconsistent performance. This is mainly because of a lack of structural uniformity over the entire area of the substrate which could result in poor reproducibility and inhomogeneity, as different size, shape, composition, orientation and local dielectric of the plasmonic structure have different enhancement factors as mentioned before. The arrangements of the aggregates on the nanostructured surface are also hard to control.

3.2. Top-Down Synthesis

Top-down approaches for nanofabrication are scalable and highly reproducible. Top-down approaches include lithography techniques (electron-beam (E-beam) [103,104,105] and nanoimprint lithography [106,107]), laser etching together with film deposition (sputtering, metal evaporation, atomic layer deposition) [108,109], templating (using anodic aluminium oxide [65], porous polymer (Polyester(PS) [110], masks or molds [57]), inkjet printing [111].

E-beam and nanoimprint lithography are fabrication methods used to create patterns with dimension down to 10 nm. In E-beam lithography, the photon resist is crosslinked after being exposed to the electron beam. The exposed resist can then be washed away, leaving only the unexposed resist on the substrate. Metal layer is then evaporated onto the whole substrate and the unexposed polymer is then removed together with its metal over-layer, leaving only the metal pattern on the substrate. SERS substrates with various geometries such as nanoparticle dimers have been fabricated using E-beam lithography, see Figure 4a–e.

Besides, Hu et al. fabricated polymer nanofinger structures on Si wafers using nanoimprint lithography, and coated the nanofingers with 70 nm of gold by e-beam evaporation, followed by exposure to solvent, inducing a leaning or self-closing of the nanowires, creating hot spots [119]. The fabricated arrays of electromagnetically coupled Ag nanoparticles on Si, could increase Raman efficiency by controlling the interparticle separation between Ag nanoparticles. These substrates showed high SERS enhancement with good control and reproducibility. However, lithography based methods, although extremely tuneable and scalable, suffer from high cost, slow throughput and are time consuming.

Laser-induced fabrication of SERS substrates has attracted research attention as it is scalabe and cost-effective. The fabrication of laser-induced SERS substrates always involves two steps. Firstly, fabrication of a nano/micro patterned substrate using ultra-fast laser pulses followed by physical vapor deposition to deposite the metal layer on the nano/micro patterned substrates in order to get plasmonic structures. Yang et al. used a nanosecond pulsed laser (1064 nm, pulse duration (τ) (full width at half maximum, FWHM) 5 ns, pulse repetition rate (PRR) 100 kHz, spot size~20 μm and laser ablation speed (ν) 100 mm/s) ablation system to create micropatterns and generate different size nanoparticles [108]. The ablated Si surfaces were then deposited with Ag by electron beam evaporation. The enhancement factor of the fabricated substrates was estimated to be ~5.5 × 10⁶. Diebold et al. fabricated Ag SERS substrate using a femtosecond laser (100-fs pulses at a repetition rate of 1 kHz, 800-nm center wavelength) structuring process [120]. This pulse train was frequency-doubled to a center wavelength of 400 nm through a thin BiBO₃ crystal. They used an n-type silicon wafer as the substrate. These laser pulses had an average fluence of 10 kJ/m² at the surface of a silicon wafer. A thermal deposition at a rate of 0.15 nm/s onto the structured silicon with different thicknesses of 10, 30, 60, 80, 100, and 200 nm was undertaken to get the optimized Ag nanostructure sizes for enhanced SERS performance. Similarly, Indrė Aleknavičienė et al. fabricated fast and scalable SERS substrates at low cost using ultrashort-pulse laser-induced (280 fs, 100 kHz, 350–380 nJ) plasma-assisted ablation (LIPAA) of soda-lime glass. The fabrication speed was as fast as 150 mm/s [121]. After the amorphous nanostructure formation on the glass surface, deposition of a 170 nm silver layer by vacuum deposition (around 100 nm in diameter, forming 1–3 μm size dendrimers) was applied to the glass substrate. This SERS substrate achieved an average enhancement factor (EF) of 3.0 × 10⁵ evaluated using thiophenol.

Inkjet printing combined with electrochemical deposition is more convenient than e-beam lithography in terms of time and cost, and allows larger area fabrication at the same time. Inkjet printing method can be used both on solid substrate or on flexible substrate [122,123,124,125,126,127,128,129,130,131,132,133,134]. However, the choices of ink for inkjet printing are currently limited.

SERS substrates fabricated by templating methods is another widely used approach and has been demonstrated by a number of groups, see Figure 4f–i. Shanshan Shen et al. have studied substrates based on CdTe quantum dots modified polystyrene (PS) spheres with Ag nanoparticle caps. These substrates showed high enhancement factor (0.71 × 10⁶) by using 4-ATP as the model molecule. [135]. Another very popular method is to use Anodised Aluminium Oxide (AAO) as a template to produce nanotubes [136]. Aluminium foil is anodised in acid to create nanopores, which are then used as a template to fabricate SERS substrates. Metals can be pulsed electrodeposited inside the template channels [137], or deposited via electron beam evaporation [138]. Alternatively, polymers can be employed to template the AAO. Lovera et al. fabricated super hydrophobic PS nanotubes by wetting commercial AAO filters and depositing silver onto the resulting PS nanotube structures [65]. Similarly, Zhang et al. patterned Polyethylene terephthalate (PET) on a nanostructured AAO template and deposited Au onto the polymer to create Au ananosturcture arrays. [115]. Other templates used for the fabrication of nanowires and nanotubes for SERS substrates include Polycarbonate membranes (PCM) [139,140,141], Polystyrene microspheres (PSM) [142,143,144] and nano-channel glasses. Charconnet et al. fabricated superlattices by templated self-assembly of gold nanoparticles on a flexible support, with tunable lattice-plasmon resonances through macroscopic strain. They found that the highest SERS performance was achieved by matching the lattice plasmon mode to the excitation wavelength, by post-assembly fine-tuning of long-range structural parameters [145]. The above substrates fabricated with “top-down” methods are manufactured reproducibly with high throughput, but often produced weaker signals than ‘bottom-up’ method due to larger distance between the nano structures and smaller surface density for nano particles that contribute to SERS signal. It involves also the use of expensive equipment and/or complex procedures. Combination of bottom-up and top-down fabricate method could increase the Raman intensity by creating a rougher and more homogeneous plasmonic surface [146].

4. Chemical Functionalization of SERS Substrates

SERS substrates without functionalization have limitations when dealing with real samples, e.g., strong background noise from the environment, or dealing with macro molecular such as DNA/proteins, which would block the SERS signal. For this reason, functionalization of SERS substrates is often undertaken to improve the selectivity or sensitivity in identifying the specific target analyte, which could lower detection time and limit of detection (LOD) [147]. The most widely used functionalization method is the attachment of self-assembled monolayer (SAM) of thiols to the silver or gold metallic nano structures. Functional groups along the thiol are used to phyisorb or chemically bind the analyte of interest to the substrate, thereby pre-concentrating the analyte at the surface leading to a subsequent increased in sensitivity. The thiol attaches to the metal nano structure surface owing to the strong affinity of sulphur with metals. However, the mechanism for the thiol group binding to the metal surface is not totally understood yet. One theory is that the thiol moieties chemisorb to planar gold surfaces, with the loss of hydrogen during the formation of the bond [148]. In 2003, Meirav Cohen-Atiya et al. studied adsorption of thiols on different metal surfaces by potentiometric measurements [149]. This adsorption process involves several complex steps including negative charge transfer and discharge through a reduction process. These steps related to the metal, the surface state of the metal, and the end functional groups in thiols. In 2014, Xue et al. used AFM to study the force between Au NPs and thiols under different experimental condition including oxidized Au surface and reduced Au surface with different pH effects; they stated that the bond between Au and thiol is a covalent bond [150]. In 2019, Inkpen and colleagues suggested that in gold–thiol SAMs prepared from solution deposition of dithiols, the gold–sulfur coupling had a physisorbed character, by both experiment and density functional theory (DFT) caculation [151]. Henrik Grönbeck et al. suggested the role of the thiol chain length must be considered when accessing the stability of monolayer systems on surfaces and clusters [152]. The formation of the Metal-S bond was not always very strong and the attachment of the thiol to the metal surface often happens in a few seconds. In order to gain homogeneous thiol self-assembly layer, longer incubation times, typically 24 h, are required. Thiol concentration affects binding result [153], and spacing between thiols can be adjusted by adding different thiols. Some bio-sensors based on thiolated single stand-DNA [154] or thiolated anti-body [155] have been used to detect DNA/RNA/antigen in environment, while thiol-aptamers are also used for DNA detection, see Figure 5.

The interaction between antibody-antigen is considered strong due to the presence of electrostatic forces such as hydrophobic interactions, hydrogen bonds, van der Waals forces or ionic bonds [156]. Functionalization of SERS substrates with antibodies has gained much interest in recent years because of the significant level of sensitivity and selectivity that can be achieved with them [157]. Aptamers are single-stranded DNA sequences that can be designed to capture specific chemicals. Thiolated aptamers conjugated onto metal nano structures can capture and combine with the targeted chemicals to get surface enhanced Raman spectra. Sensors based on aptamers and SERS have been used for detecting pesticides, DNAs, RNAs, uranyl, biotoxin, pathogen, hazard foodborne etc. [158,159,160,161,162,163]. Gold/silver nanoparticles modified with polyethylene glycol (thiol-PEG) provide a capping system that stabilizes the antibody and avoid the reticular endothelial system [129,164,165].

5. Application of SERS in Agri-Food

5.1. Detection of Pesticide Residues

According to The United Nations Population Division, it is estimated that in 2050, the global population will reach ~9.7 billion, 30% more people than in 2017. A key challenge therefore, is that food production must keep pace with population growth. To this end, a variety of interventions have been put in place over the years to reduce losses due to disease and pests. Pesticide usage is essential in modern agricultural practices to protect crops and increase yield. There are currently more than 1000 pesticides used commercially around the world to ensure food is not damaged or destroyed by pests. Each pesticide has different properties and toxicological effects. The main drawbacks of pesticides is their potential toxicity to humans and other non-targeted organisms, which can result in significantly reduced biodiversity, through environmental contaminations in soil, water, and other vegetation [166,167]. The widespread use of pesticides therefore needs to be monitored and controlled [168].

Pesticides are classified by (i) the mode of entry, (ii) their function and the pest organism they kill, and (iii) their chemical composition. Based on chemical composition, pesticides are classified into four main groups, namely, organochlorines, organophosphorus, carbamates and pyrethrin/pyrethroids; see

Table 1. Four categories of pesticides classified by chemical composition.

Pesticides Category	Examples	Degradation in the Environment
Organo-chlorines	DDT, Chlorinated cyclodienes (aldrin, dieldrin, endrin, heptachlor, chlordane and endosulfan), dicofol, mirex, kepone, and pentachlorophenol	Long term residual effect in the environment
Organo-phosphates	Parathion, malathion, methyl parathion chlorpyrifos, diazinon, dichlorvos, phosmet, fenitrothion, tetrachlorvinphos, azamethiphos, azinphos-methyl, terbufos	Biodegradable
Carbamates	Aldicarb, carbofuran, carbaryl, ethienocarb, fenobucarb, oxamyl, and methomyl.	Easily degraded under natural environment with minimum environmental pollution
Pyrethroids	delatmethrin, cyfluthrin, befenthrin, lambda-cyhalothrin, permethrin.	Non-persistent, and break down easily on exposure to light.

. Organochlorines pesticides (also known as chlorinated hydrocarbons) are organic compounds attached with five or more chlorine atoms. Organophosphates are derivatives of phosphoric acid, while carbamates derived from carbamic acid. Synthetic pyrethroid pesticides are group of organic pesticide that can be synthesized by duplicating the structure of natural pyrethrins.

Current detection methods of pesticides include high pressure liquid chromatography (HPLC), gas chromatography (GC), liquid chromatography (LG), mass chromatography (MC), spectrofluorimetric techniques as well as electrochemical methods. Han et al., for example, reported 302 targeted contaminants in catfish muscle by fast low-pressure GC–MS/MS and UHPLC-MS/MS methods [169]; Velkoska-Markovska, L. et al. detected malathion using liquid chromatography [170]; Wang et al. detected organic phosphates (OPs) using fluorescent probe [118]. Geto et al. used screen-printed carbon electrodes electrochemical sensors to detect bentazone in water source [171]. Santana et al. detected Carbendazim using electrochemical detection [172].

There are growing demands for the development of novel analytical techniques for a variety of pollutants affecting crops, for example, pesticides. In this respect, methods based on SERS have attracted attention. The first SERS study of pesticides was the detection of organophosphorus pesticides in 1987 by Alak et al. [173]. Since then, the potential toxicity to humans, animals, and the environment has been reported, and tolerance levels were introduced for a large number of harmful pesticides [174,175]. SERS detection methods have developed considerably, resulting in a large number of the more recent reports employing in-situ SERS detection methods on the surface of different foods [93,176]. A search on Web of Science with key words combining the topics of “surface enhanced Raman” and “pesticides” revealed 432 publication results up to 12 November 2021. The majority of SERS and pesticide-related researches demonstrated detection of organophosphate (OP) insecticides, for example, phosmet [176,177], parathion-methyl [178], malathion [170,179,180], chlorpyrifos [181,182,183]; see Table 2. Other SERS pesticide studies included fungicides (thiram [184], thiabendazol [185,186]), herbicides [179,180,187,188], and neonicotinoids insecticides (imidacloprid [189], thiacloprid [190], acetamiprid [191,192]).

Amongst the pesticides, organophosphates (OP) represent the largest class, making up to 50% of the neurotoxic agents in chemical pesticides [193]. Most OP usage is agricultural, since the Environmental Protection Agency banned their residential use in 2001 [194]. However, their human and animal toxicity still make them a societal health and environmental concern [195,196,197]. Moreover, pesticides at low concentration have been detected in food and drinking water [182,198,199,200,201]. Liu et al. reported on the use of silver-coated gold bimetallic nanoparticles; their Raman Enhancement depends on the silver shell thickness, for in situ detection of a range of pesticides on fruit peels without further sample preparation, with a limit of detection (LOD) below the required maximum residue levels (MRL) [202]. However, as described above, colloidal based solutions are not ideal for portable applications while solid SERS substrates that can be prepared in advance are more suitable. For example, Chen et al. used “SERS tape” to extract OP pesticides (thiram, chlorpyrifos, methyl parathion) from different kinds of fruit and vegetable peels [102]. The tape is placed on to the surface of the produce and peeled off for SERS analysis. This is non-invasive and requires no sample or substrate preparation. Additionally, Li et al. have created a ‘smart dust’ that easily spreads over a probed surface for in-situ SERS measurements [203]. This method requires no preparation or particle aggregation/concentration on the substrate. Their shell-isolated nanoparticles are used to analyse residues of OP pesticide parathion, on a fresh orange. They present comparable results between a normal Raman and a portable Raman, demonstrating the substrates potential use in-field.

Table 2. Organophosphates (OP) detection by SERS in food industry and environment monitoring.

Organophosphates (Matrix)		SERS Substrate	LOD (Reported)	LOD (Normalised)	Excitation Wavelength
Phosmet	Ooling tea [204]	Ag NPs	0.1 mg/kg	0.1 ppm	633 nm
	fruit [205]	multi-walled carbon nanotubes	0.5 mg/kg	0.5 ppm	785 nm
	paddy water [206]	Au nanorods	0.25 mg/L	0.25 ppm	Portable Raman spectrometer, 785 nm
	fruit skin [207]	polyurethane-Ag NPs	0.6 µg/mL	0.6 ppb	785 nm
	Plant Surfaces [208]	polyurethane micelle/Ag NP	0.08 g/mL	80 ppm	/
Parathion-methyl	fruit or vegetable peels [209]	Snowflake-like Au NPs	0.026 ng/cm	/	638 nm
	solvent [210]	Ag NP decorated ZnO-nanorods	10−8 M	2.63 ppb	532 nm
	solvent [211]	nanoporous structure	12 ppb	12 ppb	785 nm
Malathion	solvent [212]	nanostructured Ag	10 nM	3.30 ppb	632.8 nm
Chlorpyrifos	tomato surface [183]	Ag colloid	10−9 mol/L	0.35 ppb	638 nm
	soil [213]	Au NP	10 ppm	10 ppm	785 nm
	fruits [214]	Ag NP	10 ng/mL	10 ppb	633 nm

Table 2. Organophosphates (OP) detection by SERS in food industry and environment monitoring.

Organophosphates (Matrix)		SERS Substrate	LOD (Reported)	LOD (Normalised)	Excitation Wavelength
Phosmet	Ooling tea [204]	Ag NPs	0.1 mg/kg	0.1 ppm	633 nm
	fruit [205]	multi-walled carbon nanotubes	0.5 mg/kg	0.5 ppm	785 nm
	paddy water [206]	Au nanorods	0.25 mg/L	0.25 ppm	Portable Raman spectrometer, 785 nm
	fruit skin [207]	polyurethane-Ag NPs	0.6 µg/mL	0.6 ppb	785 nm
	Plant Surfaces [208]	polyurethane micelle/Ag NP	0.08 g/mL	80 ppm	/
Parathion-methyl	fruit or vegetable peels [209]	Snowflake-like Au NPs	0.026 ng/cm	/	638 nm
	solvent [210]	Ag NP decorated ZnO-nanorods	10−8 M	2.63 ppb	532 nm
	solvent [211]	nanoporous structure	12 ppb	12 ppb	785 nm
Malathion	solvent [212]	nanostructured Ag	10 nM	3.30 ppb	632.8 nm
Chlorpyrifos	tomato surface [183]	Ag colloid	10−9 mol/L	0.35 ppb	638 nm
	soil [213]	Au NP	10 ppm	10 ppm	785 nm
	fruits [214]	Ag NP	10 ng/mL	10 ppb	633 nm

Neonicotinoids are a more recent and relatively powerful class of insecticide, and since the introduction of imidacloprid in 1991, they have been the fastest-growing class of insecticides in modern crop protection [215], representing almost 17% of the global market [216]. Typical detection methods of neonicotinoids are based on enzyme linked immuno-sorbent assays (ELISA) [217,218], HPLC- or GC- mass spectrometry [219,220], surface plasmon resonance [221], and fluorescence spectroscopy [222], none of which are suitable for field analysis. Neonicotinoids are extremely effective against herbivorous insects [223], while having perceived low toxicity to mammals, birds and fish [224]. This has led to their widespread uptake for use on a variety of crops. However, concerns have been raised recently about environmental impact in affecting the homing capacity of honey bees, resulting in global colony collapse of the pollinator population [2,225]. Consequently, the European Union enforced a temporary ban (Dec 2013) [226] reducing the MRL of neonicotinoids to between 0.01 to 3 mg/kg for many fruits and vegetables [227,228].

Studies of detection of neonicotinoids are shown in

Table 3. Neonicotinoids detection by SERS in food industry and environment monitoring.

Neonicotinoids		SERS Substrate	LOD (Reported)	LOD (Normalised)	Excitation Wavelength
Acetamiprid	Solvent [229]	AuNP/MOF (metal–organic framework) composite	0.009 μM	2 ppb	780 nm
	apple juice [232]	Gold nanoparticles (AuNPs) bonded with polyadenine (polyA)-mediated aptamer and Raman tag (MMBN-AuNPs-aptamer)	6.8 nM	1.514 ppb	532 nm
	solvent [233]	co-doped N/Ag carbon dot	0.006 µg/L	6 ppt	633 nm
	green tea [234]	Au NPs	1.76 × 10−8 M	3.91 ppb	micro-Raman spectroscopy, 785 nm
Imidacloprid	solvent [235]	3-D Ag dendrites on Paper substrate	0.02811 ng/mL	0.02811 ppb	633 nm
	extract solution from apple [214]	Ag NP coated glass	50 ng/mL	50 ppb	633 nm
	green tea [236]	flower shaped Ag nanostructure	10−4 µg/mL	10−4 ppm	785 nm
	waste water treatment [237]	r-GO supporting Ag meso-flowers and phenyl-modified graphitic carbon nitride	10 mg/mL	104 ppm	632.8 nm
	Solvent [238]	Ag nanostructures on PVDF	1 ng/mL	1 ppb	514 nm
	fresh tea leaves; apple peels [239]	Au NPs	0.5 mg/kg; 0.02 mg/kg	0.5 ppm; 0.02 ppm	780 nm
Thiamethoxam	solvent [240]	Au NPs	0.1 ng/mL	0.1 ppb	785 nm
	solvent [241]	Ag nano structure	g/mL order	106 ppm order	532 nm
Thiacloprid	fruit [242]	Au@Ag NPs	0.1 mg/kg	0.1 ppm	633 n

. For example, Cao et al. synthesized three types of AuNP/MOF (metal–organic framework) composite to investigate the interaction between acetamiprid and the bridging molecules of the MOFs. Acetamiprid in this case was used to evaluate the characteristics of the SERS substrates. LODs of 0.02 μM, 0.009 μM, and 0.02 μM were achieved for the three composites, which could satisfy the requirement of detection according to the MRLs of acetamiprid. [229] Yang et al. used SERS to evaluate the penetration behaviors of four pesticides (acetamiprid, thiabendazole, ferbam and phosmet) in a variety of fresh produce matrices. They used a pesticide/AgNP complex deposited onto the external surfaces of different fresh produce and measured the penetration depth of the complex using SERS [230]. Although the results are promising, this method requires complex sample preparation with the pesticide and AgNP, including washing steps, and is not ideal for farm-side analysis. On the other hand, Wijaya et al. employed silver dendrites for SERS-based detection of acetamiprid in apple juice and from swabs of the apple surface [231]. The TQ Analyst Software (Thermo Fisher Scientific) was used for SERS spectral data analysis, with second-derivative transformation employed to remove baseline and separate overlapped peaks. Outlier peaks were also removed to gain a more accurate quantification results. Acetamiprid detection was determined using principal component analysis (PCA) and retains the principal components (PCs) that capture the variation between sample treatments. This method does not need pre-treatment for the apple juice samples and the use of the swab is non-invasive to the fruit. This method therefore has the potential to be used for on-site pesticide detection.

5.2. Detection of Chemical Additives

5.2.1. Dye Molecules

Dye molecules are used in industries to colour different materials such as silk, wool, cotton and paper. Unfortunately, the wasted water from these industries cause pollution in aquaculture and also cause serious toxic, carcinogenic and mutagenic effects in mammalian cells [243]. Besides, Malachite green and Crystal violet have been used for the treatment of fungal, parasitic and protozoan diseases in fish, and it is found to absorb and metabolise in tissues of fish [244]. The detection of dye molecules such as rhodamine 6G (R6G), malachite green, crystal violet (CV) and 4-aminobenzenthiol are the most reported chemical contaminants due to their ease of detection. These dye molecules are highly Raman active and used indiscriminately as antimicrobials in aquaculture. Moreover, the most significant and influential papers in this field have employed these dye molecules to study single molecular SERS detection [245], enhancement factors [37], and the mechanisms of SERS [246].

5.2.2. Melamine

In 2008, a sanitary scandal involved the intentional contamination of milk powder with melamine to give a false appearance of high protein levels [247]. Monitoring the level of residual melamine has since become important for the dairy industry. Regarding SERS, melamine is probably the most widely documented food adulterant, with Web of Science literature search revealing 309 articles up to 12 November 2021. The majority of melamine detection researches that employ SERS substrates are based on Au and Ag nanoparticle fabrication. Gold substrates include: Au colloids [248], Au NP agglomerates [249,250,251], 4-mercaptopyridine-modified Au NPs [252], and magnetic Au NPs [253], to name a few. Similarly the silver SERS substrates include: Ag Colloids [254], Ag NP agglomerates [255], Ag NP coated Ag/C nanospheres [256], Ag NP coated polystyrene nanospheres [144,257], cyclodextrin-coated Ag NPs [258], functional graphene/Ag nanocomposite [259].

Peng et al. used self-assembled vertical arrays of nanorods to detect melamine in methanol [260] and similarly, Hu et al. coated Ag nanoparticles on the surfaces of Fe₃O₄@SiO₂ composite microspheres to detect a melamine methanol solution [261]. Neither of these articles demonstrated melamine detection in real samples. Zhang et al. demonstrated melamine detection in milk using silver colloid solution. They reported on an easy pre-treatment for the milk, which, however, still required large instrumentation and, additionally, the colloid NPs required mixing with the diluted and filtered milk [246]. Alternatively, Guo et al. developed self-assembled hollow gold nanospheres to detect melamine in milk on a solid chip platform, which is ideal for remote sensing. However, they employed centrifugation as their only method of sample pre-treatment, which is complex and not suitable for transporting [262].

Another novel method by Betz et al. used a copper tape and a penny coin to fabricate Ag micro- and nanostructures. It was used to analyse infant formula adulterated with melamine [255]. The fact that these substrates form in five minutes on-site without the need for complex equipment, sample pre-treatment, or harsh chemicals enabled the possibility of remote point-of-sampling. However, their LOD (5 ppm) is not sufficient for remote melamine detection.

Finally, Chen et al. reported on the detection of melamine in egg white using fabricated ZnO/Au composite nanoneedle arrays, see Figure 6. The results showed some background interferences from the egg proteins but the characteristic peak for melamine at ~682 cm⁻¹ remained detectable and was well resolved [263]. The only sample preparation is a filtering of the egg solution through four layers of gauge, which can easily be employed on-site as it requires no complex instrumentation. Similarly, Kim et al. applied their previously reported gold nanofingers to melamine detection in milk [264]. Although they also require sample pre-treatment, the authors avoid using centrifugation, as it is neither portable nor low cost. Instead, they employ a mini dialysis kit and detect characteristic melamine peaks from the dialysis filtered solutions at 1 ppm. They also demonstrate melamine detection at 100 ppb in infant formula using a solution gel filtration chromatography treatment. These are a few articles that report SERS substrates and methods of sample pre-treatment, that both are fully compatible for field applications in a limited-resource environment.

6. Conclusions and Perspective

In this review paper, SERS as a useful technology for Agri-Food and environmental sensing has been investigated. Compared to other detecting technologies such as chromatography or electrochemical based methods, SERS has the potential to deliver rapid, ultra-sensitive and highly specific detections of a wide range of chemicals and biomolecules. SERS substrates fabrication methods include bottom-up and top-down. Bottom-up approaches refer to the fabrication of nanostructures by chemical synthesis, colloid aggregation, electrochemical deposition and self-assembly; top-down methods include lithography techniques (electron-beam (E-beam) and nanoimprint lithography, laser etching together with metal film deposition (sputtering, metal evaporation, atomic layer deposition), templating (e.g., anodic aluminium oxide), inkjet printing, etc. The combination of top-down and bottom-up methods could improve SERS intensity. The common plasmonic materials are Ag, Cu, Au or their composite material. A wide choice of the support substrate is variable Si, Cu, tissue, paper, leaves and aluminium cans, etc. SERS related devices could widely be used in agri-food and environment monitoring, following the SERS device fabrication method section; this review paper presents SERS applications in detection of pesticides such as organophosphate (OP) insecticides and neonicotinoids insecticides as well as illegal food additive such as dyes and melamine.

However, despite the advancement and strategies presented above, some challenges still need to be overcome for SERS to be widely used in agri-food and environment analytical applications. These include (i) repeatability of SERS substrates and SERS signals for each Raman measurements; (ii) weak interaction—or even repulsion because of the surface energy—between some analytes and SERS surface; (iii) stability of SERS substrate and functional layers that can in some instance react with the targeted analytes or degrade over time (e.g., oxidation) or under continuous laser excitation; (iv) general non suitability for direct detection of heavy metal or macromolecular such as protein; (v) current lack of standardised optical setup or methodologies to compare results obtained by different research groups and (vi) in real sample verification, SERS devices performance could be affected by contamination or interferences from the environment.

Regarding sample reproducibility, SERS substrates could benefit from constant advances in nanofabrication processes and instrumentation as well as in surface chemistry. Regarding quantification, techniques such as isotope labelling [265] or standard addition method have proven promising. Also, new emerging techniques such as electrochemical SERS (EC-SERS), shifted excitation Raman difference spectroscopy (SERDS) or surface enhanced spatially offset resonance Raman spectroscopy (SESORRS) such could help overcome the limitations mentioned above.

All in all, SERS devices benefit from a wide range of choice of plasmonic structures such as Ag/Au NPs, nano rods, nano flowers, nano cubes, and their composite materials etc; supporting substrate, fabrication method and functionalization method. Besides, they also have a numerous of advantages such as rapid response time, high sensitivity and selectivity and possibility to use handheld devices for onsite measurements. However, several challenges still exist in terms of reliability and durability for SERS sensing platform. Ongoing advances in nanofabrication and chemistry have the potential to overcome the current limitations of SERS sensing. As a result, we believe that SERS will soon be a widespread analytical technique for sensitive detection of contaminants in agri-food and environmental applications.