Nanostructure-Enhanced Optical Sensing Platforms for Pesticide Analysis in Food and Water Samples: A Review

Abstract

Pesticides are applied to promote performances in the agricultural field, sustaining crop productivity by counteracting the damages induced by pests and weeds. Under conditions of uncontrolled application, their negative influences exerted on soil, water and biodiversity mean contamination of food and impact on human health. The reactive oxygen species generation induced by pesticides impair the antioxidant protective ability. For humans, pesticides can have cytotoxic, carcinogenic, and mutagenic potential. They can be classified relying on the chemical structure or on the targeted organism. Optical sensors are based on UV-Vis absorption, fluorescence, chemiluminescence, surface plasmon resonance or Raman scattering. Based on their coloring features, nanomaterials are used in optical sensing platforms. They impart high specific surface area, small sizes, facility of surface modification by biorecognition elements (enzyme, antibody, aptamer, molecularly-imprinted polymer) and promote sensitivity and selectivity in biosensing platforms. The present paper highlights the performances of the optical sensing platforms in pesticide assay. Relevant novel applications are discussed critically, following the attempts to improve analytical features of chemical and biochemical sensors. Critical comparison of the techniques is performed in the last section. Advances in nanofabrication like the inclusion of novel nanomaterials and optimizing data interpretation by integration of algorithms can further enhance performances.

1. General Aspects: Classification, Role and Mechanism of Action

Pesticide treatment is useful to increase performances in the agricultural field, enhancing crop productivity by diminishing the damages induced by pests and weeds [1,2]. They are required for food security and public health, protecting the amount and quality of harvested crops. Pesticides help in hampering the action of mosquitoes, ticks, or fleas, impeding disease spreading (malaria, dengue, or Lyme disease) [2,3,4].

The unrestricted use of pesticides exerts a negative impact on both the environment and human health. Impressive quantities of pesticide residues build-up in the soil, a vital resource sustaining human existence. Under conditions of uncontrolled and unreasonable application, their negative influences exerted on soil and water, biodiversity, food cycle, or human health, grow more and more significant. It has been reported that contamination with pesticides reaches 90% of farmland water sources, impacting aquatic and terrestrial ecosystems and consequently food chains. It has been established that even at amounts regarded as environmentally secure, pesticide use in Europe lowers the population of species by 42% [5,6].

Cancer, endocrine, neurological and reproductive imbalances are health issues provoked by pesticide residues [6,7]. Hence, the impact of pesticides should be thoroughly investigated, limiting their application and securing human health.

Governments have established policies for managing pesticide application and have stipulated maximal residue amounts for foodstuffs and agricultural products [8]. Though most pesticides were found within stipulated limits, the impact of bioaccumulation and ongoing exposure can imply human health hazards [9].

Unhindered pesticide application leads to environmental, aquatic and agricultural pollution meaning contamination of food, including vegetables and fruits, processed foods, air, aquatic media and soil [10,11]. Dietary pesticide residues become severe public health issues, with acute and chronic health negative influences, reported mainly in developing countries [11].

The reactive oxygen species generation induced by pesticides impair the antioxidant profile and the protective ability against oxidative decay. The oxidative stress, meaning imbalance between oxidative and antioxidant species, affects biomolecules like lipids, proteins, and nucleic acids, as well as the signaling pathways in the cell, with associated negative health impact [4,12]. The often indiscriminate and flawed application of pesticides, may result in effects from acute intoxication to chronic illnesses like cancer (brain, colon, breast, bladder, prostate) [13,14], neurodegenerative diseases like Alzheimer’s disease [15], Parkinson’s disease [16], neurotoxicity [17], leukemia [18], diabetes [19], or infertility [20].

Pesticides are organic pollutants, degradable via microbial, chemical or photochemical processes. Microbial decay implies mineralization that decomposes the molecule into carbon dioxide, followed by co-metabolization, involving microbial processes that convert pesticides into other chemical forms. During photolysis, the pesticides are broken down by ultraviolet light. Chemical degradation of pesticides involves redox reactions and hydrolysis under the action of water, air or other compounds present in soil [21].

Several criteria are considered when classifying pesticides. Depending on the chemical structure, more categories are distinguished: organochlorines, organophosphorus, carbamate, pyrethrin, and pyrethroid pesticides.

In conformity to the targeted organisms, pesticides are classified as herbicides, fungicides, insecticides, molluscicides, rodenticides, nematicides, and compounds regulating plant growth [22].

Herbicides can be structurally triazines, amides, urea derivatives, uracil derivatives, bipyridyls, carbamates, or dinitroanilines. Triketones inhibit pigment synthesis in broad-leaved weeds resulting in bleaching and eventually death. Insecticides are mainly chlorinated derivatives, but also organophosphates, carbamates, or pyrethroids. Fungicides and bactericides are benzimidazoles, morpholines, diazines, diazoles and triazoles [4].

Organochlorines, like DDT or chlordane, are largely applied inducing convulsions, paralysis and eventually death of the targeted organism. Organophosphate pesticides like malathion, chlorpyrifos, and diazinon, derived from phosphoric acid, are employed as acaricides and insecticides. Their functioning relies on inhibiting the activity of cholinesterase, an enzyme required for insects’ nerve function [23]. They contaminate soil, groundwater and the atmosphere [24] and are linked to poison by contact, or during fumigation [25]. Though these compounds possess a moderate pest resistance, they are biodegradable [25].

Carbamates possess a carbamate group, with a nitrogen atom linked to both a carbonyl group and an oxygen atom. They disrupt nerve signal transmission in the targeted insect. Carbamates inhibit the activity of acetylcholinesterase, not by forming a covalent bond with the enzyme, like organophosphates, but by establishing a reversible bond with AChE [26]. Carbamates like carbaryl, methomyl, aldicarb or propoxur follow a mechanism analogous to that of organophosphates, but persist less in the environment. Their degradation is pivotal given their toxic potential to living systems. Carbamate microbial degradation depends on the availability of microorganisms with proper biodegradative enzymes. The progresses in genetic engineering and biotechnology can be used to exploit the degradative features of microorganisms, promoting bioremediation protocols [27].

Pyrethroids like permethrin, cypermethrin, and deltamethrin are synthetic compounds derived from pyrethrins that are isolated from chrysanthemum flowers. They have insecticidal attributes analogous to those of pyrethrins, but are characterized by higher stability and longer persistence [28,29].

Neonicotinoids present a conjugated bond system, with an aromatic heterocyclic group, a flexible linkage, and a hydroheterocyclic guanidine/amidine group, along with an electron-withdrawing group [30]. Neonicotinoid pesticides like imidacloprid, clothianidin, and thiamethoxam are categorized as systemic insecticides, being absorbed by plants and distributed in plant tissues. They are also obtained by synthetic routes and act on nicotinic acetylcholine receptors present in the nervous system of the target insect [31].

Triazines (atrazine, simazine, and metribuzin, etc.) are a class of herbicides possessing a triazine ring. They block the electron transport chain, inhibiting photosynthesis in the chloroplasts [29,32].

Some of the representatives of the most common pesticide classes are presented in Figure 1 [29].

By binding to the Q_B protein in the photosystem II, these herbicides obstruct the electron flow in the electron transport chain [33]. Some fungicides like triazoles and strobilurins are transferred to various organs of the plant after ingestion, impairing key metabolic functions in the target [4,34]. Fungicides influence plant physiology, encompassing pivotal metabolic processes like photosynthesis. The negative impact on photosystems leads to progressive chlorosis, impaired enzyme antioxidant defense, pigment biosynthesis and root development [35].

It has been reported that around 700 pesticides, encompassing herbicides, insecticides, and fungicides, act on approximately 95 biochemical targets in pest weeds, insects, and harmful fungi. Common insecticides act mainly on four nervous targets, like the voltage-gated chloride channel, acetylcholinesterase, the acetylcholine receptor, and the γ-aminobutyric acid receptor, found in animals but not plants. Herbicides impact plant-specific pathways, obstruct photosynthesis, carotenoid synthesis, or branched-chain and aromatic amino acid synthesis, vital for plants. Most fungicides disrupt ergosterol, tubulin biosynthesis, or cytochrome c reductase, while others interfere with essential cell functions. An important limiting factor in the ongoing application of pesticides is the strain selection. The challenges are the resistance to the selected compounds, but also the cross-resistance to other pesticides that attack the same target [36].

Given the toxicity associated to pesticide application, continuous monitoring of their level is pivotal for minimizing their health impact. Elaborating fast, simple, and sensitive technologies for simultaneously detecting multiple pesticide residues is necessary, to comply with the requirements for food security, environmental monitoring, and human health. The bioanalysis methods, such as enzyme inhibition assays, enzyme-linked immunosorbent assay, surface plasmon resonance, and fluorescence methods [37,38,39,40] provide a series of advantages and can be coupled with, or employed as an alternative to LC–MS/MS or GC–MS, for the detection of organophosphates in agricultural samples. Though sensitive and able to provide results in real time, they require further exploration with respect to stability and reproducibility. Biosensors and immunoassays rely on the use of biomolecules, such as enzymes, antibodies, and nucleic acids, to specifically analyze chemical species [41,42]. These methods are largely applied to pesticide residue assay, for their simplicity, rapidity, sensitivity, and specificity, as shown in food control, clinical analysis, and environmental monitoring [43].

A chemical sensor as analytical device is comprised of two major components: the sensitive element or receptor, which interacts with the targeted analyte, and the transducing element, which transforms the recognition event into a measurable signal [44]. The latter is processed and correlated to the analyte concentration [45].

Various detection mechanisms can be exploited: electrochemical, optical, thermal or piezoelectrical. Optical sensors’ functioning rely on different optical phenomena and may be subdivided according to the type of optical properties quantified, such as absorbance, fluorescence, luminescence, reflectance, refractive index, and the light scattering [46].

Optical detection employs optical waveguides, surface plasmon resonators, interferometers, and photonic crystals, quantifying absorption, fluorescence, phosphorescence, refraction, or Raman scattering [47,48,49]. An optical biosensor encompasses a recognition element (enzyme, antibody, aptamer, molecularly-imprinted polymer, and host-guest recognizer) interacting with the targeted pesticide, and a transducer that signals the binding event. Increasing attention is focused towards improving the sensors’ analytical features [50].

2. Colorimetry

Colorimetric assay relies on the quantification of the color behavior associated to a chemical process. Nanomaterials have gained increasing attention in material science, chemical catalysis, medicine, or computer science. In the development of sensing devices, many nanomaterials are employed for qualitative analysis purposes, but also for signal output or amplification, in selective and sensitive chemical and biochemical quantitative assays. Other fields are food packaging, environmental, medicinal, agricultural, pharmaceutical applications and photoelectronics [51,52].

Synthetic routes for obtaining nanostructures include hydrothermal and solvothermal techniques, the sol–gel method, chemical precipitation, pyrolysis of the metal homolog, thermal decomposition, wet chemical synthesis, and microwave synthesis. Green synthesis makes use of plant extracts, algae, biopolymers, or active biosurfactants, with elevated specificity, biodegradability, and biocompatibility, counteracting the use of toxic compounds [52,53].

Hydrothermal, solvothermal, photochemical, electrochemical, microwave, microemulsion, pyrolysis, redox, co-precipitation, sol-gel, and chemical vapor deposition are classed as bottom-up approaches, and the other methods like ball milling, arc discharge, laser ablation, sonication, and nanolithography, are viewed as top-down methods [54,55]. Physical synthesis relies generally on top-down approaches, involving high energy expenditure and costs, but has as benefits elevated purity, and also control on shape, size, and crystallinity. Chemical methods encompass bottom-up approaches, allowing for large-scale production, but use toxic reagents and instrumentation, as well as significant energetic expenses. Green synthesis is characterized by the facility of the experimental procedure, does not harm the environment, but requires control for providing an aseptic cultivation environment [55].

The Figure below (Figure 2) depicts the main techniques applied for carbon quantum dots synthesis [56].

A comprehensive synthesis of the main techniques applied for nanoparticle preparation is presented below (Figure 3) [57].

Based on their variable coloring features, nanomaterials are used in optical sensing platforms. As light reaches a noble metal nanoparticle, the oscillating electric field induces a collective electron oscillation at the same frequency, a phenomenon recognized as localized surface plasmon resonance. At the resonant wavelength, localized surface plasmon resonance results in a powerful electromagnetic field near the nanoparticle surface, causing extinction. The features of localized surface plasmon resonance are influenced by the size, morphology, aggregation state, and the neighboring medium of the nanoparticles. Moreover, the electromagnetic field is maximal at the surface of the particle, but swiftly diminishes, when increasing the distance from the surface. The localized surface plasmon resonance effect of noble metal nanoparticles induces a specific characteristic color correlated with the extinction peak, employed to assess the analyte concentration. The peculiar nanoparticle color is exploited to develop test strips to quantify pesticides [6].

Metal nanoparticles such as gold or silver nanoparticles possess remarkable localized surface plasmon resonance features, elevated molar extinction coefficients, and facility of surface functionalization [58,59] enabling their extensive application in colorimetric sensors. Aggregation-based colorimetric sensing platforms are the most broadly used. Metal nanoparticles aggregation is directly or indirectly promoted by the presence of analyte molecules via hydrogen bonds, electrostatic forces, van der Waals interactions, and covalent linkages. This can lead to major localized surface plasmon resonance and color modifications, the resulted colorimetric signals being linearly correlated to the analyte amount [60,61].

Localized surface plasmon resonance properties of metal nanoparticles result in powerful colors in the visible range. Aggregation results in distinct linear and non-linear optical features. A well-established chemical interaction between the nanoparticles’ surface and the analyte molecules can lead to increasing optical modifications (red to blue in the case of gold nanoparticles, and bright yellow to red-brown in the case of silver nanoparticles), enabling naked-eye pesticide analysis. These techniques proved cost-effective, rapid, and allow for on-site detection. Functionalization and modification of nanoparticles highly promote the sensitivity and selectivity of such techniques [62]. Gold nanoparticles proved to be reliable signal transducers in colorimetric sensors for insecticide detection. The van der Waals interactions established between particles and the charge of nanoparticles are tightly linked to their colloidal stability, which can be disintegrated by coordination bonds, adsorption, or salt ions, leading to color changes induced by nanoparticle aggregation [46].

A paper-based sensor was developed for the swift colorimetric assay of six main organophosphate and carbamate pesticides. The sensor was obtained by dropping gold and silver nanoparticles on the hydrophilic area of the paper support. L-arginine, quercetin, and polyglutamic acid were used as nanoparticle modifiers. The sensing mechanism relied on the pesticide–nanoparticle interaction, that promoted color changes recorded by a digital camera. The technique resulted in a single analytical signal for each tested pesticide, carbaryl, parathion, malathion, paraoxon, diazinon, and chlorpyrifos. The limits of detection for these pesticides were 29.0, 32.0, 17.0, 22.0, 45.0, and 36.0 ng mL⁻¹, respectively. The method was applied for the simultaneous quantification of these pesticides in a mixture, using the partial least square method. The reported root mean square errors of prediction were 11.0, 9.2, 10.0, 8.7, 12.0, and 11.0 for carbaryl, parathion, malathion, paraoxon, diazinon, and chlorpyrifos, respectively. The discrimination of the two types of studied pesticides (organophosphate and carbamate), as well as the differentiation between the two types of functional groups (oxon and thion), were performed by the paper-based device. The potential interferent species such as metal ions, anions, sugars, amino acids, and vitamins, did not impact the analytical response of pesticides, showing selectivity. The method showed analytical viability at the analysis of apple juice, tap water, and rice samples [63].

Colorimetric sensing platforms with direct visualization can be integrated in point-of-care assays. The conversion of the recognition event into a visual color change is possible by integrating nanomaterials, like gold nanoparticles. Acetamiprid detection relied on the robust interaction between the cyano group of acetamiprid and the gold nanoparticles. This interaction led to nanoparticle aggregation associated to a color change. By visual monitoring of the color modification or by employing a spectrometer, it was possible to qualitatively and quantitatively determine acetamiprid. The sensing platform functioned relying on the state change in gold nanoparticles from dispersion to aggregation. The analyte concentration could be qualitatively evaluated from the red to blue change. This color modification caused by nanoparticle aggregation was significantly impacted by their distance and concentration value. UV-Vis spectroscopy and transmittance electron microscopy have been applied to follow the process. The operational conditions parameters were subject to optimization with respect to the size of the nanoparticles, pH value, or incubation time. At pH values smaller than 4.0, gold nanoparticles can readily aggregate, given the absence of the negative charge on citrate (required to synthesize nanoparticles), that can determine repulsion [64]. At pH values higher than 8.0, no relevant differences were noticed concerning the absorption proper to gold nanoparticles in the presence of the analyte. This was ascribed to the hydrolysis of cyano groups in alkaline medium, resulting in a lack of ability of acetamiprid to destabilize gold nanoparticles. Hence, a working pH value of 6.0 was chosen. A linear relationship between the logarithm of acetamiprid concentration and the absorbance was reported in the range from 0.66 to 6.6 μM for nanoparticles with diameters of 22.0 ± 1.0 nm, and in the range of 6.6–66 μM for gold nanoparticles having diameters of 15.0 ± 1.0 nm. No significant interferences were noticed except for cyanide. The analysis of greens, plant eggs and cucumber was successfully performed [65].

Terbuthylazine and dimethoate residues were rapidly detected by a novel colorimetric sensor based on citrate-stabilized gold nanoparticles, with high selectivity and excellent sensitivity. The detection mechanisms have been checked via Fourier Transform-Infra Red, UV-Vis spectra, Zeta Potential assay, Transmission Electron Microscopy and Dynamic Light Scattering. Other tested environmental pollutants exerted no interference on the terbuthylazine and dimethoate detection. The limits of detection of terbuthylazine and dimethoate assessed by naked eye visualization were 0.3 μM (68.9 μg L⁻¹) and 20 nM (4.585 μg L⁻¹), respectively, and those based on calculating using the 3σ/S formula, were 0.02 μM (4.59 μg L⁻¹) and 6.2 nM (1.42 μg L⁻¹), respectively. The minimal quantified levels of terbuthylazine or dimethoate were much smaller than the maximum residue limit stipulated by the governments of EU and China [66].

In China, the GB 2763 series of National Standards established for terbuthylazine and dimethoate a maximum residue limit of 50 μg kg⁻¹ and 20 μg kg⁻¹, respectively. The EU Regulation established a default maximum residue level of terbuthylazine to 10 μg kg⁻¹, relying on LOQ. The Environmental Protection Agency establishes a maximum dimethoate residue limit of 20 μg kg⁻¹.

The linear relationships between the UV-Vis spectrometric signals and concentration corresponded to 0.1–0.9 μM terbuthylazine and 1–40 nM dimethoate. The analytical viability of the gold nanoparticle-based sensor was proved at green tea, apple juice and tap water analysis [66].

The dispersion–aggregation behavior of gold nanoparticles was the underlying mechanism of a colorimetric sensor array, to identify five organophosphate pesticides. The sensor array consisted of citrate-capped gold nanoparticles, of 13 nm dimension. Organophosphate addition was followed by aggregation-induced spectral changes in gold nanoparticles, that were analyzed with pattern recognition techniques, encompassing hierarchical cluster analysis and linear discriminant analysis. The developed sensor array could identify individual organophosphates or mixtures in real samples [25].

Functionalization of gold nanoparticles has been applied to improve selectivity in colorimetric pesticide analysis. The p-amino benzenesulfonic acid-functionalized gold nanoparticles are endowed with outstanding optical features and zeta potential-induced instability. The surface zeta potential determinations were regarded as a novel tool to explore the aggregation mechanism. At carbaryl addition, both the color and zeta-potential of the sensor changed, enabling pesticide assay with a limit of detection of 0.25 μM (0.05025 mg L⁻¹), in agreement with the maximum level of 0.05 mg L⁻¹ for carbaryl, as established by the Chinese Environmental Protection Agency. The sensing platform presented anti-interference abilities and allowed for carbaryl determination in water samples from Yangtze River and East Lake with limits of detection as low as 0.25 μM and 0.10 μM, respectively. Moreover, the reported detection platform can constitute a viable tool for effective carbaryl removal [67].

Also integrated in optical transducers, C-dots can be employed for replacing conventional chromogenic groups, or for amplifying the efficacy of a colorimetric probe. The photosensitive characteristics of C-dots are also exploited, when developing probes with dual responses, to boost the detection mechanism. C-dots were reported as activators of classical chromogenic agents [68]. They promote the anti-interference abilities in colorimetric assays. C-dots can enhance the visible range absorbance or the color modifications of gold or silver nanoparticles, improving the detection ability [69]. Plasmonic colorimetric detectors based on “non-aggregation” or “aggregation” have received increasing attention. The etching or the growth of metal nanoparticles results in plasmon changes, underlying mainly the development of “non-aggregation” plasmonic colorimetric sensors, with elevated sensitivity, broad range of color changes, and prone to application in test strips. Of peculiar interest, the test strips developed by immobilization of nanoparticles on the paper substrate, hamper the impact of nanoparticles’ auto-aggregation and promote the facility during both storage and functioning [70]. Zheng et al. [71] found that C-dots originating from Lycii Fructus can be associated with silver nanoparticles. The C-dots present numerous nitrogen- (-NH₂) and oxygen-containing (-OH and -COOH) functional groups, able to establish interactions with the analytes. The amino groups can get absorbed on the silver nanoparticles’ surface. The free functional groups (amino, hydroxyl, carboxyl) can strengthen the C-dots–silver nanoparticle interaction. Phoxim promotes the C-dots–silver nanoparticle cross-linking, leading to the aggregation of C-dots–silver nanoparticles and subsequent color change. The aggregation of the dispersed C-dots–silver nanoparticles in the presence of the analyte, was associated to a red shift of the absorption peak from 400 nm to 525 nm, accompanied by a color change from yellow to red. The absorbance ratio A525 nm/A400 nm was linearly correlated to the phoxim concentrations in the range of 0.1–100 μM. The detection limit was 0.04 μM, smaller than the maximum residue limits of phoxim stipulated in China. The maximum residue limit specified for phoxim, by the National Standard of China GB 2763-2021 is 0.01 mg kg⁻¹ (as default or, detection limit for different pesticides) for many food products including grains/vegetables and potentially 0.02 mg kg⁻¹ for some fruits.

The colorimetric sensor was employed for phoxim quantitation in environmental and food (fruit) samples with good recoveries, ranging from 87% to 110.0% [71].

Gold nanoparticles encoded with 4-aminothiophenol were synthesized, and it was proved that their aggregation can be enhanced by silver ions, but can also be hampered in a specific manner by the competitive reaction between thiram and silver ions. By observing the color change in the probe, thiram concentration could be determined within 15 min with a detection limit of 0.04 μM and a linearity corresponding to 0.05–2.0 µM. The probe’s functioning relied on an increase in the 542 nm UV-Vis absorbance in the presence of thiram and silver ions, with a negligible influence from interfering species, even at concentrations 5 times greater than that of the analyte. This visualization technique proved sensitive and specific, characterized by simplicity of operation and instrumentation, rapidity, and potential for on-site analysis. The analytical viability has been confirmed at apple and soil analysis [72].

A colorimetric sensor array consisting of five cost-effective and commercial thiocholine and H₂O₂ sensitive indicators was constructed for the qualitative and quantitative analysis of organophosphates and carbamates. The sensor array relied on the irreversible inhibitive ability towards acetylcholinesterase activity, hampering thiocholine and H₂O₂ generation from S-acetylthiocholine and acetylcholine, leading to a color diminution or absence of color, given by the corresponding indicators. Pattern recognition and standardized statistical data treatment (hierarchical clustering analysis and principal component analysis), proved that the sensor array had discrimination ability, distinguishing organophosphates and carbamates from other pesticide classes, but also, could differentiate between the tested pesticides. Furthermore, semiquantitative assays encompassing combination of recognition patterns (hierarchical clustering analysis and principal component analysis), were also performed, resulting in the correspondent fitting curves. The limits of detection for chlorpyrifos, triazophos, methamidophos, phoxim, dimethoate, methomyl, metolcarb, carbaryl, fenobucarb, and isoprocarb were 4.6 × 10⁻⁸, 3.0 × 10⁻⁸, 3.5 × 10⁻⁸, 3.8 × 10⁻⁸, 3.3 × 10⁻⁸, 2.5 × 10⁻⁸, 2.1 × 10⁻⁸, 2.3 × 10⁻⁸, 2.3 × 10⁻⁸ and 2.4 × 10⁻⁸ g/L, respectively, with these values meeting the requirements of the Chinese food security standards regarding maximum residue limits.

The reported characteristics of the method were the elevated sensitivity and selectivity, with efficient anti-interference ability, at the simultaneous detection of organophosphates and carbamates. The application of the sensor consisted of the analysis of apple juice and green tea drinks [73].

The features of colorimetric sensing are portability, fastness, naked eye detection, equipment simplicity, nevertheless, it may require selectivity and sensitivity improvement. Nanozymes, mimicking enzyme behavior, benefit from cheapness, facility in preparation and modification when compared with most natural biocatalysts, and can be included in colorimetric sensing platforms, nevertheless, specificity and catalytical efficacy are prone to improvement [74].

The colorimetric reactions function mainly via redox mechanism. The typical nanozymes are mostly oxidoreductases such as peroxidases, thus, the selectivity is still a challenge due to various redox species with similar or close redox capacity. Both the selectivity and sensitivity can be promoted by the development of the appropriate colorimetric reagents, by reshaping the elemental composition and the colorimetric assay principle [75].

Studies focused on promoting the catalytic activity of nanozymes, have proved that nanozymes with smaller size have higher catalytic activity because more active sites are exposed. By molecular imprinting, specific recognition sites can be developed on the surface of nanozymes, mimicking enzyme–substrate interactions [76].

To overcome the limitations of natural enzymes, artificial enzymes can be used in sensor development. A novel L-aspartic acid-copper nanozyme possessing multiple enzyme-like activity was designed as sensing unit, and included in a colorimetric sensor array, relying on the distinct impact exerted by different pesticides, on nanozyme catalytic activities. It was found that laccase-like nanozymes were affected to a different extent by organophosphates and carbamates. Four nanozymes, guanosine 5′-monophosphate-copper, 4,4′-bipyridine-copper, 2-methylimidazole-copper and L-aspartic acid-copper, catalyzed the colorimetric reaction of 2,4-dichlorophenol and 4-aminoantipyrine. At organophosphate addition, the laccase-like activities of nanozymes were subject to different changes, this impact relying on the divalent copper–phosphate interaction, giving noticeable color variations, and absorbance value diminutions at 510 nm. Organophosphate identification and individual distinction were successfully performed with lack of interference from carbamates or other potential interferent species (other pesticides, ionic species, antibiotics). Relying on the nanozyme sensor array, a portable smartphone-based technique was developed, allowing for the selective identification and discrimination of organophosphates in complex samples, apples, pears, nectarines, tomatoes, celeries, and cabbages, demonstrating the feasibility of on-site detection. It was reported that the technique had more powerful anti-interfering abilities than acetylcholine esterase-based methods, as the latter were more prone to interference from carbamate [77].

Glyphosate detection was performed by a label-free, simple colorimetric method based on the hindrance of peroxidase-like activity of copper ion, inhibiting the oxidation of 3,3′,5,5′-tetramethylbenzidine in the presence of hydrogen peroxide, and resulting in absorbance change. The mechanism underlying the hindrance of peroxidase-like activity was the generation of glyphosate–Cu²⁺ complexes. Optimized experimental conditions involved 652 nm for absorbance readings, and 40 °C incubation temperature. At increasing glyphosate concentrations, the absorbance diminished, and then it was maintained constant at analyte concentrations higher than 200 μM. The linear range of analytical response corresponded to 2–200 μM and the detection limit was 1 μM. Moreover, the analyte could be distinguished by naked eyes at concentrations as low as 10 μM [78].

Zhan et al. used cube-shape Ag₂O particles with enhanced oxidase-mimicking catalytic activity and designed a colorimetric platform for dimethoate detection, based on nanozyme activity amplification. The analytical strategy in this case relied on the promotion of the electron transfer from silver oxide to dissolved molecular oxygen, by dimethoate. This leads to the abundant release of superoxide anion radical and singlet oxygen, accelerating the oxidation of the chromogenic substrate. The detection platform relied on the adsorption of dimethoate on the surface of the cube-shaped Ag₂O through the Ag–S bond. The lone electron pair present in the amide group of dimethoate could function as intermediate to facilitate the electron transfer from silver oxide to dissolved oxygen. The colorimetric method had a broad linear range of 20–160 μg/L, and a limit of detection of 14 μg/L. Samples of pepper, green beans and cabbage were analyzed by the colorimetric sensing platform. Different analyte amounts were uniformly sprayed on the surface of these samples, which were dried in a fume hood for 12 h. The samples were repeatedly eluted with 5 mL sodium acetate buffer, pH 5.0, and then the eluate was filtered through a 0.22 µm needle filter before determination. The absorbances increased with the catalytic reaction time and the signal was stabilized after 10 min [79].

It was found that CeO₂ embedded in N-doped carbon possess phosphatase-like activity, exploited in paraoxon colorimetric determination. Ce(III) and Ce(IV) present in cerium dioxide nanoparticles behaved as active sites for polarization and hydrolysis of paraoxon, yielding the colored p-nitrophenol. The catalytic hydrolysis capacity versus the phosphomonoester bonds, proper to CeO₂ embedded in N-doped carbon, gave the remarkable detection specificity. Moreover, the catalytic features of CeO₂ embedded in N-doped carbon were enhanced, as a consequence of the efficient paraoxon adsorption by N-doped carbon via π–π stacking mode. Following the absorbance maximum at 400 nm, resulted in paraoxon quantitation with a broad linearity, corresponding to 3.0–100.0 μM. Spiked garlic chives samples were analyzed with recoveries of 93–104.6% [80].

Another cerium-based nanozyme with phosphatase-like activity (grain-boundary-rich ceria metallene nanozyme) could play the role of line or interfacial defects, leading to an increase in the number of Ce⁴⁺/Ce³⁺ site pairs to 72.28%, associated to a 49.28-fold activity enhancement. Moreover, rich grain boundaries achieved band structure optimization, promoting the photoelectron transfer under irradiation, which further increased the number of metal site pairs to 88.96%. A 114.39-fold optimized activity over that of CeO₂ in the absence irradiation, was achieved. Relying on different inhibition effects of pesticides exerted on catalyst activity in the presence and in the absence of irradiation, the grain-boundary-rich ceria metallene nanozyme, proved its viability in the successful recognition of toxic pesticides in mixtures [81].

A MnO₂ nanoflowers-labeled mouse monoclonal chlorpyrifos antibody was integrated in an immunochromatographic dual read-out colorimetric/chemiluminescent chlorpyrifos assay. The MnO₂ nanoflowers-labeled antibody functioned as the signal tracer for the immunochromatographic test strip. After 10 min of competitive immunoreaction, the antibody tracer interacted with the immobilized immunogen in the test strip, capturing MnO₂ nanoflowers on the test line, leading to the occurrence of a brown color, easily noticeable by the naked eye, giving the qualitative signal. MnO₂ nanoflowers significantly impacted the luminol–H₂O₂ system, and the chemiluminescent signal led to quantitative detection of trace chlorpyrifos concentrations. 1,3-Diphenylisobenzofuran quenching and tetramethylbenzidine–H₂O₂ colorimetric assays were driven for investigating the boosting mechanism of MnO₂ nanoflowers, relying on H₂O₂ decomposition, yielding reactive oxygen species. Under optimized conditions, the linearity corresponded to 0.1–50 ng mL⁻¹, with a detection limit as low as 33 ng L⁻¹, at a signal-to-noise ratio of three [82]. The accurate monitoring of pesticides at such low level is vital. EPA has proposed to revoke all tolerances for chlorpyrifos, except for the tolerances to cover its application in 11 food and feed crops. Within the European Union, chlorpyrifos is severely restricted due to neurotoxicity, with a near-total ban and a maximum residue limit, at a “zero tolerance”.

In the other proposed dual analytical platform, chlorpyrifos and carbaryl antibodies were labeled to graphitic carbon nitride/bismuth ferrite nanocomposites, for developing a multianalyte immunochromatographic assay. Based on the luminol–H₂O₂ catalytical system, the nanocomposites were employed in a colorimetric/chemiluminescent immunochromatographic assay of chlorpyrifos and carbaryl residues.

After completing the competitive immunoreactions on the test strip, the tracer antibodies interacted with the immobilized antigens on the test lines. The accumulation of graphitic carbon nitride/bismuth ferrite nanocomposites yielded a brown color, monitored as a colorimetric semi-quantitative analytical signal. The graphitic carbon nitride/bismuth ferrite nanocomposites-induced generation of the colorimetric signal enabled sensitive quantitation, after initiation of the luminol–H₂O₂ reaction. The limits of detection of chlorpyrifos and carbaryl were both equal to 0.033 ng mL⁻¹ [83].

In another study, a biotin-labeled immunoglobulin G-modified gold nanoparticle probe, constituting an immunosensor performing in a dual mode, was designed for detection of chloroacetamides. Dephosphorylation of ascorbic acid 2-phosphate by alkaline phosphatase was followed by sequential ascorbic acid-driven deposition of silver on gold nanostars and the fluorogenic reaction of dehydroascorbic acid with o-phenylenediamine. Thus, the probe functioned in a dual mode, monitoring both the red-green-blue color modification, and the fluorescent crystal growth-induced in situ signal. The obtained limits of detection were as low as 1.20 ng mL⁻¹ for acetochlor, 0.89 ng mL⁻¹ for metolachlor, 1.22 ng mL⁻¹ for propisochlor, and 0.99 ng mL⁻¹ for the herbicides in mixture by using a smartphone, and 0.44 ng mL⁻¹ for acetochlor, 1.59 ng mL⁻¹ for metolachlor, 2.80 ng mL⁻¹ for propisochlor, and 0.72 ng mL⁻¹ in their mixture, by a spectrofluorometer. The analysis of corn samples was performed with recoveries of 91.4–105.1% operating in the colorimetric mode, and 92.4–106.2%, functioning in the fluorescent mode [84].

Aptamers represent short-chain (25 to 90 nucleobases), single-stranded oligonucleotides (DNA, single stranded DNA, or RNA oligonucleotides) developed by a procedure named SELEX that enables identification of molecules having increased affinity and binding specificity. They can selectively attach to various targets, allowing for their incorporation in sensors, alongside many other applications [85].

Acetamiprid detection was performed sensitively and selectively by aptamer-based colorimetric sensing. Associating the target analyte and the aptamer, results in salt-induced aggregation of gold nanoparticles, and specific color modifications imposed by the interparticle plasmon coupling. Even pesticides possessing analogous structure to chlorpyrifos or imidacloprid, did not exert interferent effect. Mechanistic studies inferred that the specificity could be assigned to the characteristic supramolecular interaction between acetamiprid binding aptamer and acetamiprid, and to the induced conformational changes in the biorecognition element from random coil to hairpin structure. The practical applicability was proved at the detection of acetamiprid in soil samples and at the evaluation of its natural degradation [86].

Nevertheless, it was opinionated that the salt-induced procedures present several drawbacks such as a high number of experimental steps and an impact on aptamer–pesticide affinity interaction. Thus, an aptamer-based colorimetric assay was reported, based on gold nanoparticles, without salt incorporation. Positively charged gold nanoparticles can directly react to conformational changes in the aptamer induced by the analyte, resulting in the aggregation and color changes in the nanoparticles. Besides not requiring salt addition, the method avoided the problems related to the use of negatively charged gold nanoparticles in colorimetric analysis [6,87].

Carbendazim colorimetric determination was achieved by associating specific unlabeled carbendazim aptamers, gold nanoparticles and poly-diallyldimethylammonium chloride as cationic polymer. In the analyte’s absence, the carbendazim specific aptamer electrostatically interacts with the cationic polymer forming a complex structure. The gold nanoparticles remain therefore dispersed due to the lack of availability of the polymer. When carbendazim is present into the sensing platform, the carbendazim-specific aptamer can in a selective manner capture the analyte to form a stable complex. Due to aptamer molecules blockage, poly-diallyldimethylammonium chloride cannot form a complex with the aptamer and promotes gold nanoparticles aggregation, engendering a color change from red to blue. Colorimetric determination of carbendazim relying on specific recognition imparted by the aptamer and polymer-induced aggregation of nanoparticles, results in a detection limit of 2.2 nM, a linear range of analytical response from 2.2 to 500 nM with an excellent correlation coefficient of 0.9960. The method was characterized by sensitivity, specificity, and applied to the analysis of water samples with recoveries of 94.9–104.8%. This colorimetric method proved to be simple and rapid, with low instrumentation requirements [88].

A dual modal (colorimetric and fluorescent) sensor used adenosine triphosphate- and rhodamine B-modified gold nanoparticles, aiming at the selective and sensitive organophosphate pesticide analysis, by visualization of colorimetric and fluorescence imaging changes (Figure 4) [89].

Molecularly imprinted polymers are selective biorecognition elements developed to specifically bind to target molecules by generating molecular cavities within their polymeric structure [90]. To obtain these cavities, a polymerization process is applied using the analyte molecule as a template. After template removal, the remaining cavities promote recognition and binding of the analyte found at exceedingly small levels. It was reported that this capacity to “mimic” specific molecular interactions offers benefits over other biomolecules used for recognition, such as antibodies or aptamers [45].

A simple and cost-effective colorimetric technique was designed for 3-phenoxybenzaldehyde detection. A layer of molecularly imprinted polymer was deposited using the sol-gel process onto silica nanoparticles. The 3-aminopropyltriethoxysilane and phenyltrimethoxysilane were chosen due to their ability to interact with 3-phenoxybenzaldehyde via hydrogen bonds and π–π stacking interactions. This double functionality imparts better affinity and absorption ability towards the target, than that proper to a single functional monomer. After 3-phenoxybenzaldehyde elution from the molecularly imprinted polymer, the analyte determined distinct color fading in the potassium permanganate solution. Under optimized conditions, the method’s linearity range corresponded to 0.1–1 μg mL⁻¹, with a detection limit of 0.052 μg mL⁻¹ for the pyrethroid metabolite. The recovery rates at fruit juice, beverages and river water, ranged from 90.0% to 98.9%, proving its potential for detecting pyrethroid pesticide residues [91].

3. Fluorimetry

Fluorescence’s principle of measurement relies on light emission by molecules excited by radiation of a specific wavelength, that can be modified by pesticide intervention. Sensing platforms using fluorescent detection have proved their potential in pesticide residue assay, with sensitivity and rapidity. Rational design systems applied for pesticides consider signal reliability of fluorescent materials, hazard risk of pesticides, the requirement for operation advantages, multiple detection and real sample assessment [92].

It is highly important to choose and develop an appropriate recognition system that can be associated to the fluorescent probe, and able to respond to the fluorescent “turn off”, “turn on”, or “ratiometric” analytical signal [50].

Generally, the alterations in the fluorescent signal in the presence of pesticides are “turn-off” signals, the consequence of fluorescence quenching by pesticides. The quenching can proceed via energy transfer from the sensor to the pesticide, which can be the result of Förster resonance energy transfer (FRET), photoinduced electron transfer (PET), electron exchange (EE), or inner filter effect (IFE). Förster resonance energy transfer (FRET) stipulates the presence of two participants, an energy donor and an acceptor, considered as two dipoles that interact at significant distances [93]. FRET-based assays involve spectral overlap between the emission spectrum of the energy donor and the absorption spectrum of the acceptor. Generally, FRET can be applied as an energy transfer mechanism, in most cases the pesticide molecule acting as the acceptor and determining a fluorescence quenching upon an efficacious energy transfer [94].

In photoinduced electron transfer, a direct electron transfer takes place from the energy donor to the energy acceptor, by a difference from FRET, in which no electron transfer occurs [95]. In this case, the electron transfer is promoted by photochemical excitation, requiring vicinity between the donor and the acceptor molecules [96]. PET mechanism relies on the establishment of bonds [97] or interaction at distance [98]. Fluorescence-based methods largely employ PET mechanisms [99,100]. During pesticide detection via PET, the pesticide molecules hamper the photoinduced electron transfer already taking place, resulting in signal amplification, quenching, or alteration, considering the nature of both the donor and the acceptor that are part of the mechanism [101].

Electron exchange (EE) is also named Dexter energy transfer, the electron transfer takes place from the LUMO of an excited state donor to the LUMO of an acceptor, and simultaneously an electron undergoes transfer from the HOMO of the acceptor to the HOMO of the donor [102]. EE necessitates orbital overlap and close contact between the donor and the acceptor [103]. Electron exchange is mainly effective when the donor and the acceptor can be captured in conformations that promote the required proximity [104], such as when supramolecular associations binding inside a macrocycle strengthens interaction at short distance [105]. Compared to Förster resonance energy transfer, electron exchange depends less on the spectral overlap integral, and therefore can result in turn-on fluorescence detection with a thoroughly dark background, encountered for instance when residual donor emission lacks [106]. Pesticides generally have electron deficiency; therefore, they can behave as electron acceptors. Mass spectrometry-based techniques are applied to detect the radical species that result from the single electron acceptance by pesticides [107].

Inner filter effect mechanism functions relying on an important spectral overlap between the donor and the acceptor [108,109]. While this effect has initially been considered an unwanted process as it hinders emission from the target analyte [110], control exerted on the inner filter effect relying on rational system design allows for effective fluorescence-based sensing [111]. Inner filter effect proved successful in facilitating pesticide assay, given their significant absorbance in the UV range that offers an inner filter of the light that attains the target luminophore [112,113].

It is important to mention that the inner filter effect, relies not on the deactivation of an excited state, but on the role of the absorbers present in the sample (such as the high concentration fluorophores) that can reabsorb the emitted light, a mechanism that is distinguished from a real quenching (loss of energy of an excited state), that happens for instance during FRET. Inner filter effect functions based on a non-radiative energy conversion pattern in spectrofluorometric assays, where absorption of the excitation or emitted light by the absorber takes place in the detection environment [114].

Gold nanoclusters can be successfully integrated in fluorescent probes, owing to their significant Stokes shift and lifetime, as well as good photostability. Given these reported features, they promote the analytical signal in biosensing or molecular imaging [115].

Gold nanoclusters possess improved stability when compared to their larger nanoparticle homologues, as they are characterized by stronger bonds between metal atoms and denser architecture. This increased stability is believed to bring benefits mainly in biological systems [116]. Furthermore, gold nanoclusters have different surface features that markedly lower the aggregation tendency, commonly occurring when larger gold nanoparticles are used [115,117].

Carbon dots-derived photoluminescent label-free sensors have gained increased attention. The photoluminescent assays involving carbon dots can rely on previously detailed mechanisms, but also on aggregation-induced emission (AIE), static quenching effect (SQE), or dynamic quenching effect (DQE) [69]. Static quenching involves formation of a non-fluorescent complex in the ground state between the quencher and the fluorophore that alters the absorptive features of the fluorophore, diminishing fluorescence intensity. Dynamic or collisional fluorescence quenching relies on intermolecular collisions that can involve a small analyte molecule and a biorecognition element (macromolecule), resulting in the diminution of both fluorescence quantum yield and fluorescence intensity [118].

Furthermore, the surface modification or development optimization, integrating detection materials such as quenchers and enzymes, highly improved the selectivity and sensitivity of C-dots-based photoluminescent sensors, aiming at accurate and precise pesticide analysis by monitoring the influence on amplified, quenched, or analytical response ratio of recorded signals [69].

Carbon dots–glyphosate functioned as an energy donor–energy acceptor pair system for glyphosate detection. The analyte followed fluorescence resonance energy transfer to quench the fluorescence intensity of the carbon dots synthesized using citric acid and employing the hydrothermal technique, and this process has been exploited to develop an “AND” logic gate to sensitively quantify the organophosphate pesticide. The linear range of analytical response corresponded to 0.02–2 μmol/L and the detection limit to 0.6 μmol/L [119].

The reported quenching effect of diazinon on carbon-dots solution underlies the construction of a sensitive nanosensor for the organophosphate pesticide. The analyte could interact with the hydroxyl groups of carbon dots via an efficient electron transfer mechanism, which causes the probe’s fluorescence to be quenched. The results pointed that the chemical species originating from the alcoholic extraction applied to yellowish rose, are characterized by better stability and higher quantum yield than others, underlining the requirement for proper extraction procedures applied to natural complex matrices, to achieve the synthesis of high-performance carbon dots. The fluorescence decrease in carbon dots in the presence of the analyte could be linearly correlated with concentration in the range of 0.02–1.0 µM, the relative standard deviation at 0.01 μM being 3.5%. A neutral pH was correlated to inert conditions and most enhanced fluorescent properties. The analysis of samples presenting a constant level of diazinon (0.5 µM) and different additional concentrations of other interferent pesticides, showed a less than ±5% error in the results, proving the specificity in diazinon detection [120].

Complex elements are present in the analyzed samples, which do not interact with carbon dots (distanced at more than 10 nm) and do not influence carbon-dots fluorescence duration or absorption peaks. Nevertheless, when carbon dots interact with the target pesticides, their emission or excitation spectra can overlap with the absorbance spectra of some interferents, markedly diminishing the reported fluorescent responses. When signal shifts are minor or specificity is low, an absorbent material with the potential of enhancing selectivity and sensitivity can be integrated in the sensing device [69].

Aldicarb assay has been performed by a simple, sensitive fluorescent method based on the inner filter effect. The probe integrated gold nanoparticles and carbon quantum dots that determine aggregation of the former via electrostatic interaction. Amine groups of carbon quantum dots interact with gold nanoparticles via ligand replacement of citrate ions, which led to the aggregation of gold nanoparticles and subsequent quenching of the carbon quantum dots fluorescence. At aldicarb addition upon gold nanoparticles, an intercalated layer resulted between them via Au-N and Au-S linkages, which lowered the inner filter effect of gold nanoparticles. Hence, increasing pesticide concentration in the range of 3.8–76 µg L⁻¹, reverted the impact of gold nanoparticles on carbon quantum dots, resulting in fluorescence intensity recovery of the latter, with the detection limit being as low as 3.02 µg L⁻¹. The analytical platform proved its viability at the assay of fruits, soft drinks and vegetables, with efficacy and repeatability. The sensor was adaptable and resulted in analytical viability at on-site environmental and food monitoring [121].

It has been shown that FRET mechanism underlies carbon quantum dots fluorescence quenching by gold nanoclusters [122]. The energy transfer takes place when the absorption band of the quencher overlaps with the emission band of the fluorescent material [123]. The photoluminescence intensity of nitrogen-doped carbon quantum dots, lost due to FRET, was restored by carbendazim addition [117,124]. Photoluminescence and second-order Rayleigh scattering signals of a nitrogen-dopped carbon quantum dots/gold nanoclusters system were recorded for the ratiometric detection of carbendazim. Carbendazim addition resulted in a lowering of nitrogen-dopped carbon quantum dots–gold nanocluster interaction, leading to the gold nanoparticles aggregation and the recovery of nitrogen-dopped carbon quantum dots fluorescence. UV-Vis absorption, zeta potential and fluorescence lifetime assays pointed that the fluorescence activation was due to the aggregation of gold nanoparticles, that exerts inhibitive effect on the FRET process [124]. At low pH value, the carboxyl groups present on gold nanoclusters are undissociated, lowering the FRET effect between nitrogen-doped carbon quantum dots and gold nanoclusters [125].

Furthermore, carbendazim is not stable in alkaline media, attenuating the fluorescence. An optimal pH value of 6, with maximum fluorescence enhancement, was determined after the addition of carbendazim, under conditions of overlapping between the absorption band of gold nanoclusters and emission wavelengths of nitrogen-doped carbon quantum dots. The nitrogen-doped carbon quantum dots–gold nanocluster probe allowed for sensitive and selective carbendazim assay, with two linear ranges of analytical response (1−100 μM, 150−1000 μM), and low detection limits of 0.83 μM and 37.25 μM. The sensing platform proved selective, with positively charged ionic species and other pesticides exerting a minimal impact on the analytical signal, and only carbendazim being able to boost the fluorescence intensity of nitrogen-doped carbon quantum dots [124]. In conformity to China’s food safety regulations (GB 2763–2021), the maximum allowed residue limit for carbendazim in citrus fruit is 5 mg kg⁻¹ [126]. EFSA identified acute health risks, proposing to reduce MRLs to LOQ for citrus fruits, mangoes and papayas. Only residues below the limit of detection are allowed, setting near-zero tolerances.

Moreover, it was reported that ratiometric fluorescent sensors hamper interferences in environmental samples, by exploiting the features of carbon dots–gold nanoclusters incorporated in dual-fluorescence nanomaterials [127].

Gold nanoclusters were integrated in enzyme optical probes, as they exhibit variations in fluorescence intensity after binding to the target molecules. Thiocholine, resulted from acetylcholine esterase-catalyzed cleavage of acetylthiocholine is positively charged; therefore, it can establish electrostatical interactions with gold nanoclusters, accompanied by their aggregation and fluorescence diminution. Paraoxon, by inhibition of the enzyme activity and subsequent decrease in thiocholine level in the reaction medium, lowers gold nanoclusters’ ability to aggregate. The fluorescence determinations based on gold nanoclusters and yeast-acetylcholine esterase E69Y/F330L exhibited an outstanding sensitivity, with a detection limit of 3.3 × 10⁻¹⁴ M, reported as 2–6 orders of magnitude lower than in other wild type acetylcholine esterase-based method. The fluorescence intensities of gold nanoclusters presented an enhanced rise with the increase in paraoxon concentration. In conformity to the obtained results, F330L mutant enzyme exerted a role in enhancing paraoxon sensitivity. The E69Y and F330L strain combination gave the best sensitivity in paraoxon detection. To confirm the efficacy of this methodology, tap water, seawater, sewage water and cucumber juice were analyzed. For tap water, seawater, and sewage, the recoveries were comprised in the range of 96.4–106.8%, the fluorescent signal being characterized by precision, and being not impacted by interfering substances [128].

Another analytical platform relying on the hydrolysis of acetylcholine by acetylcholine esterase into thiocholine and acetic acid, enabled quantitation of organophosphates based on their inhibition exerted on the enzyme activity. The efficacy of disulfide-functionalized gold nanoclusters was applied to the analysis of pakchoi. Gold nanoclusters functionalized by disulfide bonds interact with the hydrolysate, without employing additional materials. A simple redox process was applied for their synthesis. Advantages were powerful red-light emission due to disulfide bonds, very small size, and convenient water dispersion abilities. At L-cysteine addition, these bonds were converted into −SH groups, with an important fluorescence diminution. Nevertheless, disulfide-functionalized gold nanoclusters proved a reversible change in fluorescence in the presence of L-cysteine and hydrogen peroxide, at three cycles exposure. The detection limit was 0.015 ng mL⁻¹, meeting the regulatory standards for methidathion residue in food. This facilitated the detection by not requiring quenchers. The efficacy of disulfide-functionalized gold nanoclusters in the detection of various organophosphates was evaluated by replacing methidathion with malathion, chlorpyrifos, and paraoxon, efficient inhibitors of acetylcholine esterase in the assay. These results were consistent with those of HPLC, with higher residual levels of methidathion reported in roots compared to leaves; therefore, the process of degradation was slower in roots. The response to methidathion was specific, with minimal impact of agricultural interferents. The recoveries approached 100%, with relative standard deviations smaller than 3.5% [129].

A label-free acetylcholine esterase-based technique allowed for the sensitive detection of organophosphates, relying on the controlled release of thiocholine and the associated quenching of carbon dots fluorescence. Interaction between the carbon dots obtained by the hydrothermal technique and the enzyme reaction product, leads to fluorescence quenching. Organophosphate pesticides hinder enzyme activity, leading to both fluorescence restoring and absorbance decrease. The dual-signal enzyme sensor allowed for the monitoring of both absorbance decrease and fluorescence recovery with good sensitivity and specificity [130].

The accurate detection of carbamate pesticides was performed by using a gold nanocluster-manganese dioxide composite. Bovine serum albumin functioned as a co-template. FRET mechanism was responsible for the MnO₂-induced fluorescence quenching of gold nanoclusters. The bi-enzyme technique relied on acetylcholinesterase and choline oxidase, leading to the catalytic breakdown of acetylcholine and choline, generating hydrogen peroxide. The latter enhances MnO₂ decomposition, accompanied by color modification and fluorescence recovery. The inhibition of acetylcholinesterase activity in this dual-output assay, hinders hydrogen peroxide generation, and associated MnO₂ degradation, leading to fluorescence shift, with emphasis on carbaryl identification [131].

Methyl parathion could be analyzed by a simple and sensitive enzyme fluorescent sensor based on tyrosinase and carbon dots functionalized with L-tyrosine methyl ester, synthesized by a hydrothermal technique employing citric acid as carbon source and L-tyrosine methyl ester as modifier. Carbon dots characterization was performed by transmission electron microscopy, high resolution transmission electron microscopy, X-ray diffraction spectrometry, Fourier transform infrared spectroscopy, and X-ray photoelectron spectroscopy. The carbon dots exhibited powerful and steady photoluminescence with a quantum yield of 3.8%. On the surface of carbon dots, the enzyme can catalyze tyrosine methyl ester oxidation to the correspondent quinone products, quenching the carbon dots fluorescence. A diminution of the fluorescence quenching rate was noticed at organophosphorus pesticides addition, as a result of the decrease in enzyme activity. Methyl parathion, typical organophosphate, could be detected relying on the tyrosinase inhibition rate, proportional to the logarithm of the analyte concentration in the range of 1.0 × 10⁻¹⁰–1.0 × 10⁻⁴ M, with an outstandingly low detection limit of 4.8 × 10⁻¹¹ M, at a signal-to-noise ratio of 3. Thus, the method shows a low detection limit, wide linear range, good selectivity and high reproducibility. This sensing system has been successfully used for the analysis of cabbage, milk and fruit juice samples [132].

The inhibition on alkaline phosphatase, dephosphorylation enzyme, exerted by organophosphates, underlied the construction of a fluorescent sensing platform relying on L-ascorbic acid 2-phosphate sesquimagnesium salt hydrate, as an ascorbic acid precursor. The latter interacts with o-phenylenediamine yielding 3-(1,2-dihydroxyethyl)furo[3,4-b]quinoxalin-1(3H)-one, with an enhanced emission signal at 425 nm. The inhibitive ability of organophosphates resulted in a decrease in the fluorescence of the quinoxaline skeleton present in the fluorophore molecule. Under optimized conditions, the fluorescent signal was linearly correlated to the logarithm of chlorpyrifos concentration in a broad range of 20 pg mL⁻¹–1000 ng mL⁻¹ with a detection limit of 15.03 pg mL⁻¹, at a signal-to-noise ratio of 3. Leeks and celery samples were analyzed with precision given by an inter-assay relative standard deviation below 11.51%, and accuracy given by recovery values of 94.5–106.7%. The color changes in ultraviolet light could be applied for semiquantitative determination [133].

A sensitive wide-spectrum monoclonal antibody with high binding affinity against carbofuran and 3-hydroxycarbofuran was employed to design an optical waveguide-based fluorescent immunosensor. The competitive assay relied on the inhibition of the antibody affinity to carbofuran by 3-hydroxycarbofuran. The hapten-protein conjugate immobilized on a waveguide was represented by 3-succinyl-carbofuran-OVA conjugate. During the first step, the analyzed sample was mixed with the Cy5.5-labeled antibody and left to react for a certain time. Then, the mixed sample crossed a chip surface. The unbound antibody reacted with the hapten immobilized on the chip surface. For the quantitative analysis, the fluorescence intensity of the Cy5.5-labeled antibody attached on the chip surface was eventually measured. The linear ranges of detection for carbofuran and 3-hydroxycarbofuran were 0.29–2.69 and 0.12–4.59 μg L⁻¹, respectively. The limits of detection were 0.13 and 0.06 μg L⁻¹ for carbofuran and 3-hydroxycarbofuran, respectively. The fluorescent immunosensor was enabled for the simultaneous, cost-effective and quick detection of both analytes. Samples of river water necessitated mere filtration on a 0.45 µm filter membrane and tap water was analyzed as such. In the case of long bean and apple sample, the weighed samples (5 g) were spiked with various analyte concentrations, followed by homogenization with 10 mL acetonitrile. The mixture was ultrasonicated for 20 min and centrifuged for 10 min at 5000 rpm. After extraction of the supernatant, saturated NaCl solution was added, followed by the organic phase addition to the solid phase extraction dispersion tube of QuEChERS (Quick, Easy, Cheap, Effective, Rugged, and Safe), and centrifugation for 5 min at 12,000 rpm. Additionally, 800 s was selected as the reaction time, and 300 s as the pre-reaction time, reaching a compromise of the sensitivity, detection time and cost. The schematic representation and the results of the optimization are presented in Figure 5 [134].

The sensor chip was characterized by reusability, allowing for repeatable application for more than 100 times, with full automation, the detection process lasting for around 30 min for every cycle. The technique proved good recoveries at the analysis of fruits, vegetables or water samples [134].

A fluorescent competitive immunosensor enabled the analysis of agricultural products, relying on dual signal amplification followed by a catalytic hairpin self-assembly reaction. The surface coating of three pesticide antigens in 96-well microtiter plates constituted an immunocompetitive system that competed with analyte molecules to form bonds with antibodies modified by three gold nanoparticle probes. The unbound material was discarded, the biobarcode modified on the gold nanoparticles was released, and a hairpin structure was added. Given the free energy value, the released biobarcode behaved as chain initiator, forming a double-stranded structure after complementary pairing with the corresponding hairpin structure, by that transferring the target chain, which can keep on circulating in a catalytic hybridization process, to fulfill double-signal enhancement. The double-stranded structure was fluorescently labeled, enabling quantitative multiresidue detection of organophosphate. Experimental groups were constituted, containing eight combinations of triazophos, parathion, and chlorpyrifos in the immunocompetitive system—single/mixed OVA-haptens, single/mixed gold nanoparticles, and single/mixed hairpin structures, respectively. Each group was subject to five parallel studies. The dynamic range of the method corresponded to 0.01–50 ng mL⁻¹ triazophos, parathion, and chlorpyrifos, with limits of detection of 0.012, 0.0057, and 0.0074 ng mL⁻¹, respectively. The recoveries obtained at spiked samples’ analysis were comprised between 82.8% and 110.6%, with variation coefficients ranging between 5.5% and 18.5%. The viability and accuracy were proved at the analysis of cabbage, cucumber, rice and apple. The competitive fluorescent immunosensor results were in good agreement with those provided by the LC–MS/MS technique, proving that the catalytic hairpin self-assembly-based biosensor can function as faster, viable, less costly alternative to the LC–MS/MS technique in organophosphate detection, at trace amounts [43].

DNA-templated silver nanoclusters and a Cu²⁺-dependent strategy, were used to develop a label-free fluorescent detection technique for glyphosate. Cupric ions had the ability to quench the fluorescence of the DNA-templated silver nanoclusters, that was in turn restored by glyphosate. Specific DNA-templated silver nanoclusters were selected for constructing the glyphosate sensor, after the analysis of the storage stability of the developed nanoclusters, employing different DNA templates. After optimization of the buffer pH and of the copper ion concentration, a pH value of 7.5 and a concentration of copper ions of 60 nM were selected, yielding the most significant recovery rates. The linear analytical response for glyphosate corresponded to 1.0–50 ng mL⁻¹ and the limit of detection to 0.2 ng mL⁻¹. The method was applied to glyphosate analysis in tap and spring water samples [135].

The sensitive detection of dimethoate was performed by a novel fluorescent aptasensor encompassing carbon quantum dots labeled with double-stranded DNA and Ti₃C₂T_x (2D transition metal carbide) flakes. The fluorescence of DNA-labeled carbon quantum dots was quenched by MXene flakes following Förster resonance energy transfer, with subsequent recovery at dimethoate addition. The linearity corresponded to 1 × 10⁻⁹–5 × 10⁻⁵ M dimethoate with a correlation coefficient of 0.996, and an outstanding sensitivity given by the low detection limit of 2.18 × 10⁻¹⁰ M. The good precision as reproducibility was given by a RSD value of 3.06%. High selectivity for dimethoate was obtained in the presence of potential interferents, the biorecognition element giving specificity. Apple juice and tap water were analyzed with minimum pretreatment, and good recovery values ranging between 96.2 and 104.4%. The most representative results, along with the interference studies are presented in Figure 6 [136].

Given the confirmed analytical performances, the reported aptasensor has proved its ability to detect food contaminants, with analytical potential for other hazardous compounds. Nevertheless, some reported limitations were the relatively long incubation time (1 h) and lack of suitability for real-time or on-site detection [136].

A fluorescent sensor prepared by a surface-imprinting strategy, and integrating a molecularly imprinted polymer synthesized by a catalyst-free imprinting polymerization, was designed for glyphosate analysis. The molecularly imprinted polymer-coated paper sensor had selectivity imparted by the biorecognition element, with a linear detection range from 0.5 to 10 µmol and a limit of detection of 0.29 µmol. The assay proved rapid, with a detection time of about 5 min. The accuracy of the specific paper-based sensor was proved by recovery rates between 92 and 117% in spiked food samples. Minimum matrix interferences, short sample pretreatment time, good stability, cost-efficacy, rapidity, operation facility, and potential for on-site detection of glyphosate were the advantages of the paper sensor [137].

4. Chemiluminescence

Chemiluminescence relies on the release of electromagnetic radiation (ultraviolet, visible or infrared), noticed when a reaction generates an electronically excited chemical species: intermediate or product. This can either directly release radiation (direct chemiluminescence) or transfer its energy to another molecule that eventually emits radiation (indirect or sensitized chemiluminescence). The radiation can be released via energy transfer, the process being currently named chemo-excitation; when the reaction involves biocatalysts or takes place in living organisms, the phenomenon is called bioluminescence. The optical instrumentation required is simple with no need of external light source. Powerful background light is avoided, thus lowering the background signals. The impact of stray light and the effects that can result from light source instability are lowered, meaning improved sensitivities and low limits of detection. The method has been characterized as “dark-field technique”, the acquisition of the chemiluminescent signal necessitating a photomultiplier tube.

The most often reported direct and indirect chemiluminescent systems rely on luminol/peroxyoxalate reactions, tris (2,2′-bipyridine) ruthenium (III) system or direct oxidative processes. The technique can be integrated in flow-injection systems or can be associated with high-performance liquid chromatography. It is considered that chemiluminescent phenomena can address some shortcomings related to the analysis of pesticide residues in the liquid media, and it can be regarded as a viable alternative to other detection techniques like UV/Vis spectrometry, fluorescence or mass spectrometry, mainly with respect to sensitivity and instrumentation simplicity [138].

Recently, nanomaterials such as metallic organic frameworks, black phosphorus quantum dots, carbon dots, molybdenum disulfide, layered double hydroxides, or noble metal nanoparticles, have proved their ability to enhance the chemiluminescent signal, given their extensive surface area, peculiar photoelectric features, and outstanding reactivity [139]. Owing to performances imparted by large surface area, elevated surface energy, porous configurations, excellent reactivity and peculiar photoelectric features, nanomaterials are broadly employed to further develop the chemiluminescent chemical sensors and biosensors. In chemiluminescent assays, nanoparticles act as energy acceptors and donors, luminophores, response substrates, and catalysts [140].

Liquid chromatography with post-column photoinduced chemiluminescence detection was developed for carbamate pesticide determination, like aldicarb, butocarboxim, ethiofencarb, methiocarb, methomyl, thiodicarb, thiofanox and thiophanate-methyl. After performing chromatographic separation that took less than 13 min, quinine as sensitizer was integrated in the system, the flow crossed a UV lamp (with an irradiation time of 67 s) and the resulted photoproducts reacted with acidic Ce(IV), leading to chemiluminescent emission. Mechanistically, the carbamate molecule is excited by UV radiation, then interacts with Ce(IV), reducing it and yielding Ce(III) in excited state. The latter transfers energy to quinine, that eventually emits radiation. The photoinduced chemiluminescent technique was highly selective for sulphur-containing carbamate pesticides. The applied solid-phase extraction process promoted sensitivity (with a limit of detection ranging from 0.06 to 0.27 ng mL⁻¹) and enabled carbamates determination in ground and surface water, with recoveries comprised in the range of 87–110% (except for thiophanate-methyl, that gave recoveries between 60 and 75%). The intra- and inter-day precision was illustrated by RSD values ranging from 1.1 to 7.5% and from 2.6 to 12.3%, respectively. The UV irradiation in the presence of quinine has turned out to be a convenient methodology to enhance the number of chemical species with adequate chemiluminescent features [141].

A peroxyoxalate-based chemiluminescent nanosensing system composed of bis(2,4,6-trichlorophenyl)oxalate, hydrogen peroxide, and specific-sized gold nanoparticles allowed for extremely sensitive, quick, cost-effective and selective detection of thiram, via a chemiluminescent “on–off” signal. The nanosensing technique yielded a powerful chemiluminescent signal in a chemical manner, using specific-sized excited gold nanoparticles and functioned via energy transfer from peroxyoxalate; the chemiluminescent system responded very sensitively to thiram and the response diminished progressively, as the energy level of gold nanoparticles changed after interaction with the analyte to generate an aggregated complex sustained by strong gold–sulfur bonds. The determination of thiram with the developed chemiluminescent nanosensor was performed sensitively, with a low detection limit of 0.35 nM, and a detection time of 40 s. The nanosensing technique was selective for sulfhydryl-containing hazardous chemical species present at residue level in vegetables, fruits, milk, and water, ensuring food and environmental safety [142].

A chemiluminometric sensor array exploited the simultaneous utilization of the triple-channel features of the luminol-functionalized silver nanoparticle/H₂O₂ system. The triple-channel properties encompassed chemiluminescent intensity, the time of chemiluminescent emissions and the time to attain the signal’s peak value, measurable via a single experiment. The triple-channel properties simultaneously varied after interaction with the pesticides, yielding different chemiluminescent response patterns associated to each specific pesticide. Principal component analysis was applied to generate a clustering map. Five organophosphate and carbamate pesticides, dimethoate, carbaryl, dipterex, carbofuran, chlorpyrifos have been differentiated at a 24 μg mL⁻¹ concentration [143].

A novel luminescent chemosensor using europium(III)-vitamin B1 complex, with a 1:2 stoichiometric ratio, was employed for chlorfenvinphos and malathion detection in water samples. The detection method relied on luminescence quenching of a Eu(III)–vitamin B1 probe in solution at the increase in pesticide concentration. Methanol was confirmed as the optimum solvent. The detection limits were 0.31 and 0.12 µM for chlorfenvinphos and malathion, respectively, with a better sensitivity versus malathion. The ratiometric method showed that malathion had a 13-fold higher binding affinity for the Eu(III)–(vitamin B1)₂ complex than chlorfenvinphos [144].

Iron-based metal-organic gels nanosheet/gold nanoparticles hybrid immobilization was performed at ambient temperature, by applying a simple in situ grown methodology. The hybrids had enhanced mimicking peroxidase-like activity, and proved remarkable chemiluminescent features in the presence of H₂O₂. The modification of gold nanoparticles on iron-based metal-organic gels nanosheets, could synergistically promote the chemiluminescent process by accelerating the generation of oxygenated species like hydroxyl radical OH^•, singlet oxygen ¹O₂, or superoxide anion radical O₂^•−. The synthesized gold nanoparticles/metal-organic gel hybrids allowed for the sensitive chemiluminometric detection of organophosphates, in the analytical domain from 5 to 800 nM, with a detection limit as low as 1 nM. It was inferred that the applicability of the novel enzyme-mimicking catalyst can be extended to other similar hybrid-based chemiluminometric systems for different analytes [145].

Organophosphate pesticides and D-aminoacids could be detected by a bienzyme chemiluminescent imaging sensor using N-(4-aminobutyl)-N-thylisoluminol/Co²⁺/chitosan hydrogels, and Pt-based metal organic frameworks to highly enhance the sensitivity. H₂O₂ was generated in the reaction medium by using as substrate acetylcholine chloride, and acetylcholinesterase and choline oxidase as biocatalysts, specific for the sensing platform. The decrease in H₂O₂ generation_, due to acetylcholinesterase inhibition by pesticides reduced the chemiluminescent signal. The reported linear range and the limit of detection for chlorpyrifos were 0.5–1.0 μg mL⁻¹ and 0.21 ng mL⁻¹, respectively [146].

An indirect competitive chemiluminescent enzyme immunoassay was designed for the quantitation of acetamiprid residues in vegetables. The technique was optimized in experimental conditions: the concentrations of the coating antigen (acetamiprid–bovine serum albumin) and of the anti-acetamiprid monoclonal antibody were 0.4 and 0.6 µg mL⁻¹, respectively; the pre-incubation time of the analyte (sample) with its corresponding monoclonal antibody was 30 min; the dilution factor applied to the anti-mouse-HRP antibody was 1:2500; the chemiluminescence reaction time was 20 min. The concentration yielding 50% of maximum inhibition (IC50), the detection range (IC10–IC90), and the detection limit (IC10) of the indirect competitive chemiluminescent enzyme immunoassay were 10.24, 0.70–96.31, and 0.70 ng mL⁻¹, respectively. The ratio between the luminescence value in the analyte’s absence and the IC50, was used to assess the impact of specific factors on the method’s performances. At the increase in the antibody concentration, this ratio increased, then diminished. At an antibody concentration of 0.6 µg mL⁻¹, it attained a maximum value, given the small IC50. Therefore, this resulted in a chosen optimal value of 0.6 µg mL⁻¹. The impact of four neonicotinoid analogues (thiacloprid, nitenpyram, thiamethoxam, and clothianidin) was not significant (resulting in an error smaller than 10%), proving specificity. The method showed its analytical performances at the assay of Chinese cabbage and cucumber, with recoveries from 82.7 to 112.2%, and a variation coefficient smaller than 9.19%. The matrix effects at the determination of acetamiprid in cabbage and cucumber samples are presented in Figure 7 [147].

Hence, the developed indirect competitive chemiluminescent enzyme immunoassay required minimum pretreatment and simple detection process, being characterized by good sensitivity, accuracy, and compliance with the requirements for fast acetamiprid residues screening in vegetables. The reported results showed very good correlation with those furnished by HPLC [147].

Using a novel bifunctional antibody as a biorecognition element synthesized by the hybrid hybridomas technique, a multifunctional time-resolved chemiluminescent immunochromatographic test strip was developed for the simultaneous quantitation of methyl parathion and imidacloprid residues. Horseradish peroxidase and alkaline phosphatase were employed as labels for the haptens of methyl parathion and imidacloprid, respectively. The labeled haptens competed with methyl parathion and imidacloprid to bind to the bifunctional antibody immobilized on the test strip. The two chemiluminescent enzyme-catalyzed reactions were simultaneously boosted after injection of coreactants. Horseradish peroxidase and alkaline phosphatase were characterized by different kinetics; therefore, methyl parathion and imidacloprid signals were acquired at 2.5 s and 300 s, respectively. The linearities for both analytes were 0.1–250 ng mL⁻¹, with a detection limit of 0.058 ng mL⁻¹ at a signal-to-noise value of 3. The complete analytical procedure could be finalized within 22 min. The applicability of the technique was proved at the analysis of spiked traditional Chinese herbs. The reported analytical tool proved rapid, unexpensive, and straightforward, for multiple pesticide screening relying on a single antibody [148].

The rapid screening of glyphosate pesticide residue in soybeans was performed by a novel aptamer-based chemiluminescent sensor. After specific binding of glyphosate, the resulting unbound aptamers underwent adsorbtion onto gold nanoparticles. The resulted signal consisted of luminol–H₂O₂ emission, promoted by gold nanoparticles’ aggregation, the chemiluminescent process emerging from the glyphosate–glyphosate binding aptamer complex. A strong linear relationship was established between the chemiluminescent intensity of glyphosate–glyphosate binding aptamer complex and the analyte concentration. With respect to specificity, the chemiluminescent sensor responded only to glyphosate and profenofos. Moreover, the aptamer chemiluminescent sensor could achieve glyphosate residue quantitation in organic soybeans immersed in glyphosate. The linear detection range post extraction corresponded to 0.001–10 mg L⁻¹, with compliance with the regulations in the field. The sensor proved applicable in monitoring food safety [149].

Compliance is proved with the standards stipulated for crops and foods. The maximum residue limit for glyphosate is 0.05 mg kg⁻¹ in the European Union, 0.70 mg L⁻¹ in the United States (Environmental Protection Agency), 0.10 mg kg⁻¹ in Japan and 1.0 mg kg⁻¹ in China [150].

Another enhanced, improved chemiluminescent sensing platform was designed for highly sensitive and selective acetamiprid assay. The key advantages of the technique relied on the aptamer’s enhanced binding affinity for the analyte, on the significance of gold nanoparticles’ morphology and its catalytic impact, beneficially influencing the chemiluminescent signal in the presence of H₂O₂ and luminol. Furthermore, aptamers’ conformation in the course of analyte binding, leads to nanoparticle minor morphological change, significantly altering catalytic properties. The designed analytical platform for pesticide residue analysis enabled highly sensitive quantitation of acetamiprid with a detection limit as low as 62 pM, reported as 100-fold lower than that of other previous acetamiprid aptamer-based sensors. This was due to the innate specificity of aptamer’s recognition. The technology provided a label-free, successful and unexpensive tool for sensitive and selective pesticide residue assay in contaminated soil, wastewater, and cucumbers [151].

Undiscriminated use of organophosphorus residues can be considered a threat to milk quality. Coumaphos was used as template molecule, and the bioelement, a molecularly imprinted polymer, could selectively recognize seven types of organophosphates. Then, the molecularly imprinted polymer served as an identification element for designing a chemiluminescent sensor relying on a 96-well microplate for the screening of milk samples for organophosphates. The 4-(imidazol-1-yl)phenol-enhanced luminol–H₂O₂ chemiluminescent system used imparted high sensitivity; the detection limits of the seven organophosphates subject to analysis were comprised between 1 and 3 pg mL⁻¹, and the half maximal inhibitory concentrations ranged between 1 and 20 ng mL⁻¹. The intraday recoveries were comprised between 86.1 and 86.5%, and the interday recoveries were found in the range of 83.6–94.2%. The sensor had an up to 5 times reusability. Therefore, the molecularly imprinted polymer-based chemiluminescent sensor proved to be a useful analytical tool applied for milk organophosphate residues [152].

Electrochemiluminescence implies the generation of electroactive chemical species at the electrode’s surface. The process can be followed by electron transfer reactions yielding excited states that can emit light [153].

A highly sensitive, specific luminol–H₂O₂ electrochemiluminescent system integrating gold nanoparticles was designed for atrazine detection. Gold nanoparticles promote H₂O₂ decomposition, releasing various intermediate reactive oxygen species, amplifying luminol electrochemiluminescence. The cyclic voltammetric and electrochemiluminescent determinations proved the viability of the atrazine aptasensor. The developed aptasensor was stable, sensitive, specific, rapid and reproducible. The relative standard deviation was 3.42%, for four electrodes prepared under the same experimental conditions, at 100 pg mL⁻¹ atrazine. The linear range of analytical response corresponded to 1 × 10⁻³–1 × 10³ ng mL⁻¹, and the limit of detection was 3.3 × 10⁻⁴ ng mL⁻¹. The analytical viability was proved at the analysis of cabbage, tap water and soil samples with good results. The recovery rates ranged from 89.13% to 123.03%. The signal intensities and calibration curve are presented in Figure 8 [154].

5. Vibrational Spectroscopy

5.1. Near-Infrared Spectroscopy

The method relies on the use of electromagnetic waves in the region of 700–2500 nm, that interact with materials for analytical purposes, being individualized as the non-destructive detection technique that provides the samples’ molecular composition and structure [155].

The Near-Infrared Spectroscopy increasing application in recent years encompasses the fields of agriculture and food safety, including pesticide residue detection. As vibrational technique, NIR detects vibrations determined by changes in the electric dipole moment, whereas SERS detects vibrations that resulted from molecular polarization [156]. The accuracy and sensitivity can be enhanced by combining with novel sensors, or with modern data processing and machine learning algorithms [157].

Portable Near-Infrared Spectroscopy exploiting accurate models with specific wavelengths was associated with data mining to analyze fresh fruits for their pesticide content at residue levels. Reference analyses were carried out, by applying liquid chromatography. Mathematical pre-processing methods, alongside variable selection were applied to the spectral data collected [158]. For cherry tomatoes, the best calibration and cross-validation performances obtained by Orthogonal Projection for Latent Structures models with selection of Recursive Feature Elimination or Sequential Feature Selection variables (coefficient of determination in calibration R²c from 0.86 to 0.94; coefficient of determination in cross-validation R²cv from 0.83 to 0.93), were close to those reported in other studies employing dry extract systems associated with NIR spectroscopy and partial least squares models with second derivative, that presented R²c of 0.96 and R²cv of 0.95, to assess various diclofluanide concentrations (1800–2500 nm) [159], as well as metiram/pyraclostrobin concentrations (1850–2048 nm) [160] in tomato samples, respectively. Nevertheless, in this investigation, the dry extract system associated with NIR spectroscopy aimed to eliminate water to lower its impact on the spectra. Tomato samples were treated with chlorpyrifos, difenoconazole, lambdacyhalothrin, and tetraconazole. Strawberry samples were treated with azoxystrobin, chlorothalonil, chlorpyrifos and difenoconazole. Subsequently, NIR spectra could be obtained in a non-destructive manner, applied to intact fresh tomatoes and strawberries, rendering the technique faster and enhancing applicability for pesticide residues under commercial conditions. To promote real commercial conditions and cover the rich variability in the pesticide residue amounts, 32 samples subject to each pesticide treatment were analyzed at 2 h after spraying, and 28 samples after the pre-harvest interval, set for each pesticide. Reflectance spectra were recorded with a resolution of 3–10 nm, a reading time of 100 ms, and an accuracy of ±1 nm [158].

Visible/near-infrared (376–1044 nm) and near-infrared (915–1699 nm) with hyperspectral imaging systems were applied to detect pesticide residues in grapes. Logistic regression, support vector machine, random forest, convolutional neural network, and residual neural network models enabled the development of classification models for pesticide residue amounts. The saliency maps of convolutional neural network and residual neural network allowed for the visualization of the wavelengths’ contribution. The results of near-infrared spectra resulted in better performances than those furnished by visible/near-infrared spectra. For visible/near-infrared spectra, the best model, resulting in an accuracy of over 93% was residual neural network. Concerning near-infrared spectra, logistic regression yielded best results, with 97% accuracy, but support vector machine also resulted in good results comparable with those of convolutional neural network and residual neural network models. The saliency map of convolutional neural network and residual neural network presented close ranges of key wavelengths. Globally, it was inferred that deep learning provided better results than conventional machine learning. The study highlighted the efficacy of hyperspectral imaging technology associated to machine learning in pesticide residue detection in grapes [161].

The Near-Infrared Spectroscopy-excitation strategy was employed for the detection of organophosphate and carbamate pesticides, with anti-interference ability in colored samples. An acetylcholine esterase-activated fluorescent probe was biomimetically developed by associating an acetyl-ammonium complex (enzyme identifying unit), to an extended xanthene scaffold, the latter functioning as a fluorophore supporting NIR excitation and emission. Bench and computational experiments proved the high affinity of the probe towards acetylcholine esterase molecules, resulting in rapid and sensitive response. The probe responded sensitively to acetylcholine esterase activity and both organophosphate and carbamate concentrations. Also important, by hampering the inner filter effect in both excitation and emission steps, the novel probe proved reliable in the analysis of colored samples. Curcumin, lycopene, β-carotene, betanin and cyanidin cation, when present in the probe-acetylcholine esterase mixture, exerted no obvious impact on the fluorescence intensity. Given the powerful absorbance of chlorophyll molecules at around 650 nm, the anti-interference of the probe functioned at chlorophyll concentrations below 10 μg mL⁻¹. The matrix-matched calibration curve of dichlorvos using standards in buffer and spiked samples, also confirmed anti-interference ability. Sample pre-treatment for chlorophyll or the use of a probe with higher excitation wavelength were proposed as solutions. The limits of detection for dichlorvos, carbofuran, chlorpyrifos and methamidophos were 0.0186 μg L⁻¹, 2.20 μg L⁻¹, 12.3 μg L⁻¹ and 13.6 μg L⁻¹, respectively. Pesticide assay in spiked vegetable samples, beet, carrot and lettuce was performed with higher sensitivity than that provided by the UPLC–MS/MS method [162].

5.2. Surface-Enhanced Raman Spectroscopy (SERS)

This is viewed as a robust vibrational spectroscopic technique that enables detection and sensitive quantitation of analytes at low concentration by enhancement of the electromagnetic fields occurring after excitation of localized surface plasmons [163].

SERS is considered an association of two techniques, namely Raman spectroscopy and nanotechnology. Raman and Krishnan, noticed the inelastic light scattering which involves about one in a million photons of incident light impacting a surface. The rest of the radiation is elastically reflected, and is known as Rayleigh scattering. It has been proven that the frequency changes occurring after the inelastic scattering of light, are consistent with the discrepancies in the vibrational energy levels. Therefore, each molecule type yields a distinct Raman spectral imprint, dictated by the various characteristic vibrational levels, proper to different functional groups. The ability to collect the molecular fingerprint peculiarity for every distinct molecule/analyte is considered the major advantage of Raman spectroscopy. However, given the intrinsically weak Raman signals, the technique could be improved in sensitivity. Furthermore, it has been found that Raman signals were enhanced if the analyte was placed in the vicinity of a rugged noble metal surface. Though the underlying mechanism was not precisely delineated, two theories have been elaborated. The electromagnetic theory states that the amplification of Raman signals is the consequence of the excitation of the localized surface plasmon resonance of nanoparticles when the incident radiation impacts the surface of the targeted analyte, found in proximity of the nanomaterial. The excitation frequency of the nanomaterials has to be in resonance with that of the incident light (for noble metals, this is placed in the UV-Vis range), yielding intensity peaks with at least 10⁴–10⁶ enhancement, at particular Raman shifts. In the chemical theory, the chemisorption of the analyte on the substrate, shifts the electronic state of the complex to a new absorption maximum, enabling resonation with the laser excitation frequency, and enhancing the Raman signals [164].

The main signal amplification mechanism is considered the electromagnetic one: the analyte’s Raman scattering can be promoted by the enhanced electromagnetic field in the proximity of the plasmonic nanostructures, denoted as the localized surface plasmon resonance [165,166,167].

Usually, noble metals (silver or gold) are employed to generate plasmonic nanostructures [168,169]. Different morphologies and compositions (metal nanorods, nanostars or nanoarrays, metal-doped semiconductors, metal/metal-organic frameworks) have been developed to get sensitive and robust SERS substrates [170,171,172,173,174,175] largely applicable, as they meet the practical requirements in food safety detection, environmental monitoring, and biomedical analysis [176,177,178].

Gold nanoparticles are considered the most broadly applied, given their efficacy, preparation facility, capacity to promote sensitivity, and biocompatibility [11].

There is an increasing interest in pursuing cost-effective, sensitive and reproducible materials applicable in surface-enhanced Raman spectroscopy. A SERS method for the rapid pesticide detection has been developed, and applied to the analysis of fruit juices and milk. The SERS substrate consisted of assembling gold nanorods arrays, placed on a gold-coated silicon slide. The standing nanorod arrays were orderly placed and could promote a powerful electromagnetic field for SERS determination. The SERS substrate was applied to carbaryl detection in acetonitrile/water media, milk, and fruit (orange and grapefruit) juices. The obtained concentrations of carbaryl in spiked milk or in fruit juice samples were linearly correlated with the concentrations predicted by chemometrics treatment. The partial least-squares model application resulted in r values of 0.95, 0.91 and 0.88, for milk, orange juice, and grapefruit juice, respectively. The technique was able to detect carbaryl originating from the mentioned samples, at levels as low as 50 ppb. The analyte’s detection limits were 391, 509 and 617 ppb in milk, orange juice and grapefruit juice, respectively. Acceptable recoveries from 82 to 97.5% were obtained. These reported levels were below the maximum residue limits established by EPA. The SERS technique integrating as substrate standing gold nanorod arrays, proved reliable, sensitive, rapid and reproducible for pesticide detection in food and beverages [179].

Nanocellulose-based substrates have also proved their applicability in surface-enhanced Raman spectroscopy. Cellulose nanofibers were functionalized with ammonium ions and established electrostatic interactions with citrate-stabilized gold nanoparticles yielding homogenous, reliable nanocomposites. The cellulose nanofiber-based nanostructures were loaded on gold nanoparticles strongly adhering to the cellulose nanofiber surfaces, resulting in a three-dimensional plasmonic SERS platform. Four-aminothiophenol acted as a Raman-active probe, to assess the reproducibility and sensitivity of the cellulose nanofiber-based SERS substrate. The intensity of SERS spectra that resulted from the cellulose nanofiber-gold nanoparticle composite was 20 times more enhanced than that obtained from the filter paper-gold nanoparticle substrate. The good uniformity of the nanocomposite substrate was proved by the SERS intensity mapping. The nanocomposites were applied to rapid, sensitive thiram detection in apple juice, by attaining a limit of detection of 52 ppb for the dithiocarbamate derivative [178,180].

An extraction-integrated plasmonic sensing platform was designed to perform pesticide residue detection sensitively, via surface-enhanced Raman scattering. A floating SERS substrate was designed by uniformly decorating gold nanoparticles on hollow silica microspheres. The hollow silica microspheres modified with gold nanoparticles can locate at the water/oil interface and then dispose on a filter paper, thus providing approachable and adaptable hot spots for the detection of analytes slightly enriched in the extraction phase. The floating substrate can place itself at both the water/air interface and water/oil interface. In the course of organic extraction, the analyte molecules present in the aqueous phase can be transferred to the organic phase, and can be concentrated in a very low organic solvent amount. Upon combining with the floating SERS substrate, the concentrated analyte molecules are reachable by hot spots, giving more sensitive SERS signals than can be obtained via the sol-based assay. Furthermore, the floating substrate can be relocated onto a filter paper by a capillary tube, hampering the formation of a coffee-ring. By using ethyl acetate as an extraction solvent, the analytical platform was sensitive and applicable in multiphasic systems, providing low detection limits for thiabendazole in tea drink (10² ppb) and thiram in apple juice (10 ppb), which were reported as lower than that in previous investigations [178].

The surface-enhanced Raman effect proved governed by the powerful electromagnetic field amplification at surfaces, which was linked to the localized surface plasmon resonance of several noble metal nanomaterials. Thus, noble metal nanostructures have gained increasing interest in application, mainly given the high intensification of commonly weak Raman signals, compared to other materials. Silver nanostructures are broadly used as SERS-active substrates, given their wide plasmon resonance in the visible region, significant signal amplification and preparation facility, compared with other metallic materials.

Ag-nanoplates decorated graphene-sheets constituted a highly sensitive SERS substrate for outstandingly sensitive organic pesticide detection. The “hot spots” responsible for signal amplification are due to graphene capacity to keep the nanoplatlets together. Furthermore, graphene sheets proved strong absorptive capacity for pesticide molecules, and the ability to establish π–π interaction with them. Therefore, the silver-nanoplates@graphene hybrids employed, proved performance in highly sensitive detection of pesticides at trace amounts, applicable to thiram and methyl parathion, individually and in mixture. The detection limits for thiram and methyl parathion were as low as 40 nM and 600 nM, respectively. A good linear response was noticed for thiram concentrations ranging between 10⁶ nM and 10 nM with an excellent correlation coefficient of 0.995. For methyl parathion, a good linear dependence on the concentration corresponded to 5 × 10⁵ nM–1 × 10³ nM, and a correlation coefficient of 0.993 was obtained. The signal reproducibility given by the silver-nanoplates@graphene sheets hybrid substrate was illustrated by a relative signal deviation smaller than 5.6%, showing the technique’s potential to detect organic pesticides at trace amounts [181].

A novel microneedle patch-based surface-enhanced Raman spectroscopy sensor was designed, for the simultaneous detection of pesticide residues in agricultural products. The silver nanoparticles and sodium hyaluronate/poly(vinyl alcohol) hydrogel employed in this microneedle patch-based sensor, led to the amplification of the Raman signals. Furthermore, it was noticed that the hyaluronate/poly(vinyl alcohol) hydrogel can efficiently and rapidly concentrate the residues, facilitating the detection by the sensor. An increase in the sensor’s surface area was due to the stepped structure of the microneedles. The detection procedure was minimally invasive, could perform at the surface and inside agricultural products, was sensitive, reaching detection limits of 10⁻⁷ and 10⁻⁸ M for thiram and thiabendazole, respectively. It was concluded that the microneedle patch-based SERS sensor can be applied to the analysis of plants and animals, aiming at safety and health monitoring [182].

Antibody-conjugated gold nanoparticles were used as SERS substrate, while fluorescent tags labeled on the gold nanoparticles functioned as Raman reporters, resulting in a SERS-based immunochromatographic sensor for the cypermethrin and esfenvalerate detection. When the binding sites on the immunoprobes were blocked by the analyte molecules, the excess immunoprobes interacted with the coating antigen. The biorecognition event took place, then the number of immunoprobes captured by the coating antigen present on the test line was negatively correlated with the number of analyte molecules. By monitoring two test lines, a concomitant dual detection was carried out. The immunochromatographic-SERS system promoted an excellent sensitivity for both analytes with limits of detection of 2.3 × 10⁻⁴ ng mL⁻¹ and 2.6 × 10⁻⁵ ng mL⁻¹ for cypermethrin and esfenvalerate, respectively, reported as three to four times better than in the case of ELISA and fluorescent-based immunochromatographic methods. The accuracy for the analyzed samples of river water, tap water, and milk samples was shown by recoveries comprised between 94.9% and 112%. The precision was given by relative standard deviation values of 1.3–13.5%. It was considered that this approach paves the way for using the SERS immunosensors in the point-of-care trace level pesticide assay [183].

A SERS-based technique employed 4-aminothiophenol as signal molecule and aminated mesoporous silica nanoparticles coated with aptamers, integrated in a one-pot technique. The chlorpyrifos-specific aptamer was loaded by using electrostatic interaction. The specific binding of the aptamer and chlorpyrifos resulted in 4-aminothiophenol release, whose amount was positively correlated with the analyte concentration. Antigen-coated mesoporous silica was also synthesized, resulting in efficiently amplified SERS detection. A good linear correlation between the Raman intensity and chlorpyrifos concentration was obtained for the 25–250 ng mL⁻¹ analytical domain, with a detection limit of 19.87 ng mL⁻¹. Apple and tomato samples were analyzed with recoveries ranging from 90.08 to 102.2%, and a precision given by a relative standard deviation smaller than 3.32%. This method relying on the use of the aptamer to modulate the release of the signal molecule was specific, highly sensitive, reproducible and stable, and could be used for the quantitation of chlorpyrifos in agricultural and environmental samples [184].

Feng et al. applied molecularly imprinted polymer recognition with surface-enhanced Raman spectroscopy, to detect thiabendazole fungicide in orange juice. The molecularly imprinted polymers were synthesized by precipitation polymerization. Kinetic studies and static adsorption tests confirmed the selective and efficient adsorption of thiabendazole. Synthesized molecularly imprinted polymers underwent packing into solid phase extraction cartridge functioning as sorbents, for thiabendazole separation from orange juice. The surface-enhanced Raman spectroscopy-active substrate used to quantify the eluted analyte was constituted by silver colloids prepared by reduction in silver nitrate by trisodium citrate. The whole process, preparation and detection, necessitated 23 min and the limit of detection achieved by the chemosensor was 4 ppm [185].

A NIR-SERS mixed analytical technique developed for pesticides, resulted in highly sensitive recognition via SERS, but also helped in evaluating the global chemical composition using NIR, thereby enhancing the detection accuracy of pesticides at low concentration. It was suggested that while the NIR spectral data model achieves some predictive abilities in pesticide residue detection, its rather low Relative Percent Deviation value would indicate that the model has smaller robustness and generalization capacity, related to a difficulty in preserving high precision in complex analyzed media. Conversely, the SERS spectral data model resulted in better performances than the NIR model. The optimal number of latent variables for this model was 8, with a Root Mean Square Error calibration of 0.214, a calibration R² of 0.976, a prediction root mean square error of 0.238, a prediction R² of 0.972, and a Relative Percent Deviation of 5.650. Compared to the NIR model, it was reported that the SERS spectral data model led to significantly lower errors in both the calibration and prediction sets, higher R² values, and a significant increase in Relative Percent Deviation, revealing that this method proved more accurate, more robust and able to be extended to a broader range of complex sample detection.

The Hilbert-Schmidt Independence Criterion-based Variable Space Iterative Optimization algorithm was applied for characteristic variables selection from the NIR and SERS spectra. After combination of feature variables, a multivariate calibration model was developed for rapid and accurate quantitation of food pesticides. NIR and SERS integration leads to the development of a sensitive, efficient, and cost-effective detection technique. This method significantly enhances the speed and accuracy in food safety assay. It was asserted that the technique can enable regulatory agencies and enterprises to identify food samples with high pesticide residue level, allowing timely measures to be undertaken, hampering the access of harmful food to the market, and sustaining public health [156].

A novel technique for trace assay of neonicotinoid residues in extra virgin olive oil, was based on a Covalent Organic Frameworks—Au@Ag nanoparticles composite substrate, and applied deep learning algorithms. The highly sensitive SERS substrate responded due to the excellent specific surface area and adsorption features of Covalent Organic Frameworks, associated with the powerful electromagnetic field triggered by Au@Ag nanoparticles. The outstanding detection performances for the tested pesticides, were limits of detection as low as 6.31 pM (clothianidin), 8.91 pM (imidacloprid), and 8.32 pM (acetamiprid), significantly smaller than the maximum residue limits established by China and the U.S. Environmental Protection Agency. China’s National Food Safety Standard GB 2763-2021 stipulated maximum residue limits for pesticides in olive oil, 0.05 mg kg⁻¹ for acetamiprid and clothianidin, and 0.10 mg kg⁻¹ for imidacloprid. The U.S. Environmental Protection Agency set maximum residue limits of 0.01 mg kg⁻¹, 0.02 mg kg⁻¹, and 0.05 mg kg⁻¹, for the three tested pesticides, respectively. This reported SERS sensor benefited from a dual enhancement in the sensing mechanism, and included deep learning technology for trace pesticide assay in lipid samples. The developed sensitive platform proved its reliability for contaminant assay in food, being compatible with ecological and public safety requirements [186].

6. Surface Plasmon Resonance

From the classical electrodynamics’ standpoint, a plasmon represents the common, cohesive oscillation of the conductive electrons in a material. Plasmons can be sustained by noble metal nanoclusters of a few nm dimensions [187]. These quanta of collective oscillations of the charge density of electromagnetic waves can be generated and amplified along a metal/dielectric interface. The electromagnetic field intensity exponentially decays in both media, under irradiation with a transverse magnetic-polarized light. Performing determinations with a surface plasmon resonance fiber-optic sensor, involves matching of the two momentum vectors: that of the evanescent wave created at the fiber core/metal interface due to reduced total internal reflection, and that of the surface plasmon polariton at the metal/dielectric interface [188,189]. In surface plasmon resonance sensors, a surface plasmon undergoes excitation at the metal/dielectric interface, with subsequent measurement of the changes in the refractive index [189]. At resonance, some energy of the incident light is transferred to the surface plasmons for their excitation, which is visible by a shift at an established value of the wavelength (which is the resonance wavelength). In fiber-optic surface plasmon resonance biosensors, the sensing layer covers a metal layer, whose interaction with the analyte in the sample determines the occurrence of a variation in its refractive index. The latter is transduced by a shift in the resonance wavelength values in surface plasmon resonance spectra, correlated to different analyte concentrations, which is the underlying detection principle of such a sensor [190].

A surface plasmon resonance-based optical fiber sensor was developed for fenitrothion identification and quantitation. For plasmon generation, a thin layer of silver covered the unclad core of the silica optical fiber. Subsequently, a sensing surface consisting of a layer of tantalum (V) oxide nanoparticles trapped in a reduced graphene oxide nano-scaled matrix was deposited. Fenitrothion interaction with the silver film, led to a modification in the refractive index. Relying on the monitoring of the wavelength, the fiber-optic sensor had a red shift of 56 nm, related to the fenitrothion analytical domain in the range of 0.25–4 μM, that included the blank solution. The sensitivity was 24 nm μM⁻¹, and the detection limit, 38 nM. The technique was very fast, repeatable, selective, with a response time of 23 s, enabling determinations at ambient temperature [190].

The optical fiber technology was further applied to a novel sensor development relying on localized surface plasmon resonance, for thiram detection. An etched silica optical multimode fiber covered with gold nanoparticle layers, tuned the reflected signal. Gold nanoparticles were synthesized by applying the single-phase aqueous reduction in tetrachloroauric acid by sodium citrate. The interaction between the analyte and gold nanoparticles altered the electrical properties at the surface, imparting a surface plasmon resonance response. The sensor was individualized by its capacity to trigger localized surface plasmon resonance modes for a single nanoparticle layer, as well as different aggregation states of nanoparticles. This ability has been proved after both simulations and experiments. The real-time response as resonance wavelength of the absorption peak, when the fiber probe was successively introduced in analyte solutions of different concentrations, was recorded. The developed sensor was reported as simpler and less expensive compared to SERS-based techniques, providing a lower limit of detection, when compared to UV-VIS absorption-based methods. Thiram was detected within the linear concentration range of 0.1–100 µM, and the reported detection limit was below 5 nM [191].

A surface plasmon resonance-based fiber optic sensor was developed, characterized and applied to chlorpyrifos detection. The probe’s fabrication implied acetylcholinesterase immobilization on a plastic-cladded silica fiber core, coated with silver. On the metal surface, the enzyme immobilization has been performed, and the detection relied on the ability of the analyte to act as an inhibitor via competitive binding. The spectral interrogation technique indicated that the surface plasmon resonance wavelength diminished at increasing pesticide concentration, for a fixed substrate (acetylthiocholine iodide) concentration in the liquid medium around the probe. The response time of the sensor was 6–8 min. Furthermore, the effect of the substrate concentration on the resonance wavelength, the detection accuracy, sensitivity, reproducibility and stability of the sensor have been also assessed. The sensitivity was defined as the change induced in the surface plasmon resonance wavelength per unit change in concentration of pesticide (δλsp/δc). At the increase in acetylthiocholine iodide level, the resonance wavelength increased (diminishing the impact of the competitive inhibitor) and saturation occurred after attaining a specific substrate concentration.

Regarding the influence of the analyte concentration, it has been reported that the sensitivity decreased with increasing pesticide concentration, the sensitivity variation with the analyte concentration being non-linear. A reverse trend was noticed for the detection accuracy; the latter being correlated to the width of the analytical curve. As the concentration of the pesticide increased, the reciprocal of the width increased, improving detection accuracy. It was asserted that the polyacrylamide gel present on the silver coating protects the metal from the oxidative processes and ensures the stability of the sensor [192].

A flow surface plasmon resonance device was designed using Kretschmann geometry enabling plasmon excitation on a gold surface. The major component of the sensing device was a retroreflection prism allowing for a maximum operational angle of 17°, placed on a rotating table, automatically controlled by a computer. The flow cell was constructed on the gold sensor placed on the prism surface. The flow modulated by a peristaltic pump allowed for the samples to be analyzed in the flow system. A polarized light was focalized on the surface of the sensor, with subsequent recording of the angular dependence of the reflected light. The modification of the gold sensing surface encompassed a self-assembled monolayer developed by immersion of the gold disk in an ethanol solution containing 1,1-mercaptoundecanoic acid (2 mM). The calibration graph for atrazine, plotted by using the results of the surface plasmon resonance device, exhibited a linear range from 1.0 × 10⁻⁷ to 1.5 × 10⁻⁶ M and the sensitivity was given by a limit of detection of about 5.0 × 10⁻⁸ M. The analytical signals at milk assay are presented in Figure 9 [193].

In this comparative study, the linear range of the developed screen-printed and conventional amperometric immunodevices was of about three and five concentration decades, respectively, reaching limits of detection of about 10⁻⁸ M and 5.0 × 10⁻¹¹ M, respectively. Nevertheless, the SPR device had a measurement time of about half that necessary for the two competitive devices [193].

The simultaneous detection of six different pesticides azoxystrobin, boscalid, chlorfenapyr, imazalil, isoxathion, and nitenpyram was performed by a surface plasmon resonance-based immunosensor. The dynamic ranges were 3.5–19 ng mL⁻¹ for azoxystrobin, 4.5–50 ng mL⁻¹ for boscalid, 2.5–25 ng mL⁻¹ for chlorfenapyr, 5.5–50 ng mL⁻¹ for imazalil, 3.5–50 ng mL⁻¹ for isoxathion, and 8.5–110 ng mL⁻¹ for nitenpyram. The recovery values for the analyzed tomato samples showed suitable accuracy: 104–116% for azoxystrobin, 94–101% for boscalid, 90–112% for chlorfenapyr, 96–106% for imazalil, 107–119% for isoxathion, and 104–109% for nitenpyram. The analytical viability was proved at the analysis of vegetables, and the results correlated well with those of HPLC or LC–MS/MS without significant bias, with the exception of nitenpyram for which the reported immunosensor’s sensitivity was too low [194].

A surface plasmon resonance-based nanosensor integrating a molecular imprinting technique was applied to the detection of coumaphos, an organophosphate used as insecticide and veterinary drug. To achieve sensitive and selective determination, UV polymerization was applied to develop polymeric nanofilms using as functional monomers N-methacryloyl-l-cysteine methyl ester, ethylene glycol dimethacrylate, and 2-hydroxyethyl methacrylate. These functioned also as cross-linkers, and imparted enhanced hydrophilicity. Scanning electron microscopy, atomic force microscopy and contact angle measurements were applied for the nanofilms’ characterization. The kinetic behavior in coumaphos assay was studied using coumaphos-imprinted surface plasmon resonance and non-imprinted surface plasmon resonance nanosensor chips. The created coumaphos-imprinted surface plasmon resonance nanosensor was highly selective for the analyte molecule in the presence of competitive, interferent chemical species, including pyridaphenthion, pirimiphos-methyl, phosalone, N-2,4(dimethylphenyl) formamide, 2,4-dimethylaniline, dimethoate, diazinon and phosmet. A strong linear relationship between the analytical signal and the concentration was noticed in the range of 0.1–250 ppb, with a low detection limit of 0.001 ppb and a limit of quantification of 0.003 ppb, respectively, as well as a high coumaphos imprinting factor. The Langmuir adsorption model was considered the most appropriate thermodynamic approach applied to the nanosensor. The coumaphos-imprinted surface plasmon resonance nanosensor’s reusability was assessed by intraday trials carried out three times with five repetitions for statistical evaluation.

Interday analyses performed for two weeks pointed out the three-dimensional stability of the nanosensor. The outstanding precision, in both reusability and reproducibility, was illustrated by a relative standard deviation smaller than 1.5. It was inferred that coumaphos detection in aqueous media, by using the coumaphos-imprinted surface plasmon resonance sensor, has been performed rapidly, selectively, with sensitivity and simplicity of the analysis procedure [195].

The quantification of pesticides at trace level was possible at the terahertz frequency range, using carbon nanotubes-based plasmonic metasurface sensors. Using an incident wave in the terahertz range, the surface plasmon polariton resonance was generated at the carbon nanotube film/silicon interface. The surface plasmon polariton resonance could couple an incident terahertz wave of one particular polarization, to carbon nanotubes. The resonance features changed when the analyte covered the plasmonic surface of the metasurface. The developed plasmonic metasurface responded to different concentrations of 2,4-dichlorophenoxyacetic and chlorpyrifos solutions deposited on the sensor surface. The sensor was characterized by stability, reliability and good linear dependence between transmission amplitude and analyte concentration. The lowest analyte mass detected was 10 ng and the sensitivities were 1.38 × 10⁻²/ppm and 2.0 × 10⁻³/ppm, respectively [196].

A dual sensing platform integrated surface plasmon resonance and electrochemical catalysis within a compact fiber, for in situ monitoring of dimethoate decay. Electrochemistry promoted degradation and monitored the catalytic behavior, whereas surface plasmon resonance allowed for the continuous following of the dynamic formation and disappearance of intermediate species. The 10 wt% WO₃/TiO₂ composite sample resulted in a degradation efficacy of 90.3%, due to WO₃-induced amplification of the TiO₂ interfacial electronic structure. This compact fiber sensor, with both optical and electrochemical features, provides an insight on the mechanism and kinetics of catalytic pesticide degradation [197].

Sensitivity improvement was achieved by integrating photonic crystal fibers in SPR sensors. Light propagation in the photonic crystal fibers relies on two main mechanisms. The first mechanism is index guiding, the core having higher refractive index than the cladding. The second one refers to the photonic bandgap effect, the light being confined despite a lower refractive index of the core, when compared to the cladding. One-dimensional photonic crystals consist of alternating layers of high- and low-dielectric materials disposed in a regular manner. Their performance can be improved by incorporating 2D materials like graphene, promoting light–matter interactions [198].

The micro structured construction of photonic crystal fibers sustains efficacious coupling of light to surface plasmons, improving signal intensity and detection limits. The sensing platform is sensitive to small changes in refractive index. The photonic crystal fibers-based SPR sensors have compact structure, being included in handheld devices. They allow for on-site detection in different media, being employed in point-of-care diagnostics and field applications, as they are characterized by miniaturization and portability, adequate for pollutants’ monitoring [199].

Interferometric waveguide sensors use the principles of interferometry to measure refractive index variations. The essential component of such a sensor is the waveguide structure, which guides and hinders the light within a well-established zone, for sensitive assay. Integrated interferometers imply configurations like Mach-Zehnder, Young, or bimodal waveguide interferometers. These represent viable, highly sensitive, label-free transducers, due to low limits of detection, corresponding to subtle variations in the refractive index measured in the environment above the transducer, in the domain from 10⁻⁶ to 10⁻⁸ refractive index units [200].

Chemical and biochemical sensors applied for the assessment of water pesticide contamination rely on interferometry, measuring the phase shifts of the reflected light, correlated to a refractive index change. The SPR-interferometer in association with thin recognition films has been tested to detect pesticide contamination of water. The SPR-interferometer related to multi-channel cells and thin binding films for recognition, can lead to “optoelectronic tongues” for simultaneous multi-analyte assay in liquid samples [201].

7. Analytical Parameters Obtained in Relevant Applications

The next table (

Table 1. Analytical parameters obtained at pesticide analysis in representative samples.

Method	Analyte	Matrix	RSD %	Linear Range	LOD	LOQ	Ref.
colorimetric biosensor	paraoxon	vegetables, irrigation water		9.9–79.8 μg L−1	4.678 μg L−1		[202]
colorimetry	dimethoate	pepper, green beans, cabbage	1.9–5.4	20–160 μg⋅L−1	14 μg⋅L−1	46 μg⋅L−1	[79]
colorimetry	glyphosate	soil, water and soybean samples		0.169–16.9 μg⋅L−1	0.00319 μg⋅L−1		[203]
colorimetry	deltamethrin	pears, apples	0.13–6.12	0.1–1.2 mg·L−1	54.57 μg L−1		[204]
colorimetry, fluorescence	organophosphorus pesticides	extracts from pears, tomatoes, cucumbers		0.003–100 mg L−1	0.37 μg⋅L−1 colorimetry 0.41 μg⋅L−1 fluorescence		[205]
colorimetry, fluorescence	acetamiprid, thiamethoxam	cabbage, potato	1.97–2.13 (acetamiprid) 3.41–4.34 (thiamethoxam)	4.454–44.54 μg⋅L−1 (acetamiprid) 2.917–43.75 μg⋅L−1 (thiamethoxam)	0.251 μg⋅L−1 (acetamiprid) 0.507 μg⋅L−1 (thiamethoxam)		[206]
colorimetry	carbosulfan	rice, soybeans, wheat		50–20,000 μg L−1	25 μg L−1		[207]
fluorescence	parathion-methyl	lake water, apple, and cucumber	1.01–2.67 (lake water) 2.59–5.78 (apple) 1.20–7.72 (cucumber)	0.33–6.67 μg L−1	0.14 μg L−1		[208]
fluorescence	glyphosate	tap water		0.04–0.4 μg L−1	0.035 μg L−1		[209]
fluorescent chemosensor	acetamiprid, imidacloprid, dimethoate, edifenphos, λ-cyhalothrin, quinalphos	cabbage		0–33,405 μg L−1 0–38,349 μg L−1 0–34,389 μg L−1 0–46,555 μg L−1 0–67,477 μg L−1 0–44,745 μg L−1	445.4 μg L−1 511.32 μg L−1 458.52 μg L−1 620.74 μg L−1 899.7 μg L−1 596.6 μg L−1		[210]
fluorescence	glyphosate	cabbage, potatoes	2.51–4.85	200–1800 μg L−1	15 μg L−1		[150]
fluorescence	malathion	lettuce, tap water, soil samples	<3.1	330.36–3303.6 μg L−1	16.52 μg L−1		[211]
fluorescence	chlorpyrifos	water		350–6980 μg mL−1	15.5 μg mL−1		[212]
fluorescence	phoxim	red grapes, green grapes, tomatoes, pitayas	1.4–3.80	7457.5–74,575 μg L−1	1.338 μg L−1		[213]
fluorescence, chemiluminescence	glyphosate	drip bag coffee, instant coffee, cold brew freeze-dried coffee, barley grass powder, green tea	1.528–3.711 (fluorescence) 0.879–2.971 (chemiluminescence)		81 μg L−1 (fluorescence) 38 μg L−1 (chemiluminescence)		[214]
chemiluminescence	diquat	river, tap, mineral and ground waters	3.1	10–600 μg L−1	2 μg L−1		[215]
chemiluminescence	acetamiprid	waste water, cucumber, apple, soil	2.9–4.6	0.00467–2.0 μg L−1	0.00198 μg L−1		[216]
chemiluminescence	thiram	fruits, vegetables, milk, environmental water			0.084 μg L−1		[142]
chemiluminescence	acetamiprid	Chinese cabbage, cucumber		0.70-96.31 μg L−1	1.26 μg Kg−1		[147]
chemiluminescence	glyphosate	soybean samples	1.44–2.60	0.001–10 mg L−1	0.0009 μg L−1		[149]
chemiluminescence	malathion	apple, citrus, tomato	minimal 0.82	100–1000 μg L−1	16 μg L−1	50 μg L−1	[217]
electrochemiluminescence	profenofos, isocarbophos, phorate, omethoate	spinach, rape, baby cabbage	0.93–5.48	0.001–1000 μg L−1 0.001–1000 μg L−1 0.01–1000 μg L−1 0.1–1000 μg L−1	0.0003 μg L−1 0.0003 μg L−1 0.0033 μg L−1 0.034 μg L−1, respectively		[218]
electrochemiluminescence	atrazine	tap water, soil, cabbage samples	5.12	1 × 10−3–1 × 103 μg L−1	3.3 × 10−4 μg L−1		[154]
electrochemiluminescence	carbaryl	milk powder, fruit wine	1.9–4.2	0.02-100,600 μg L−1	0.0094 μg L−1		[219]
electrochemiluminescence	malathion	cucumber, cabbage, spinach	1.4–4.2	1.321 × 10−5 μg L−1 −1.321 μg L−1	0.429 × 10−5 μg L−1		[220]
electrochemiluminescence	malathion	oranges, cabbages, eggplants	3–6	3.3 × 10−5 −3.3 μg L−1	72.348 × 10−9 μg L−1		[221]
electrochemiluminescence	isoprocarb		2–5	1.159–28.986 μg L−1	0.985 μg L−1	3.014 μg L−1	[222]
NIR	dichlorvos, carbofuran, chlorpyrifos methamidophos	beet, carrot, lettuce	<14.5		0.0186 μg L−1 2.20 μg L−1 12.3 μg L−1 13.6 μg L−1, respectively		[162]
NIR	dichlorvos	lettuce, tomato	1.46–4.65				[223]
NIR	paraoxon, chlorpyrifos, oxon, diazoxon	Chinese cabbage, oilseed rape, pakchoi (paraoxon)	0.7–6.3 (paraoxon)	1.0–10 μg L−1 (paraoxon)			[224]
NIR	glyphosate	lentil types	5.12–6.66	0.008–19.80 μg g−1 (red lentils) 0.007–19.08 μg g−1 (large green lentils) 0–19.87 μg g−1 (black beluga) 0.019–19.34 μg g−1 (French green lentils)			[225]
NIR	azoxystrobin, chlorothalonil, chlorpyrifos, difenoconazole, lambda-cyhalothrin, tetraconazole	cherry tomatoes, strawberries	1.61–3.80 as residual predicted deviation in cross-validation (RPDCV)	0.0–0.8 mg g−1 (azoxystrobin) 0.01–0.7 mg g−1 (chlorothalonil) 0.0–025 mg g−1 (chlorpyrifos) 0.02–0.20 mg g−1 (difenoconazole) 0.0–0.30 mg g-1 (tetraconazole)			[158]
NIR	glyphosate	tap water, Songhua river water, licorice extract, soil extract and canned cola	0.28–2.69	845.35–4226.75 μg L−1	6.171 μg L−1		[226]
NIR	acetamiprid	cabbage	0.6 (intra-assay) 2.33 (inter-assay)	0.0445–0.178 μg L−1			[227]
SERS	phosmet	apple	6.6–14	0.5–5 μg g−1	0.1 μg g−1		[228]
SERS	2,4-D, imidacloprid	milk		0.001–100 μg L−1	2.73 × 10−4 μg L−1 (2,4-D) 4.25 × 10−4 (imidacloprid) μg L−1	0.001 μg L−1	[229]
SERS	ferbam	peach layers (peach surface, inner skin, peach flesh)			0.012 μg g−1		[230]
SERS	pymetrozine, carbendazim	apple	0.82–12.32 (pymetrozine) 3.92–7.20 (carbendazim)				[231]
SERS	carbendazim	tea leaves	<10	191.19–0.0191 μg L−1	0.0169 μg L−1		[232]
SERS	thiabendazole	tea samples	5.92		0.1 μg L−1		[233]
SERS	thiram	pear juice	4.62–5.06	25–100 μg L−1	1.01 μg L−1		[234]
SERS	ofloxacin	egg white	8.1720 (residual predictive deviation)	0.5–45.0 mg L−1		0.5 mg L−1	[235]
SPR	triazophos	Chinese cabbage, cucumber, apple		0.98–8.29 μg L−1	0.096 μg L−1		[37]
SPR	dimethoate, carbofuran	water	0.04–0.09	0.009–1 μg L−1 (dimethoate) 0.011–0.995 μg L−1 (carbofuran)	8.37 ng L−1 (dimethoate) 7.11 ng L−1 (carbofuran)		[236]
SPR	azoxystrobin boscalid, chlorfenapyr imazalil, isoxathion nitenpyram	potato		3.5–19 μg L−1 (azoxystrobin) 4.5–50 μg L−1 (boscalid) 2.5–25 μg L−1 (chlorfenapyr) 5.5–50 μg L−1 (imazalil) 3.5–50 μg L−1 (isoxathion) 8.5–110 μg L−1 (nitenpyram)			[194]
SPR	coumaphos	honey		0.1–250 μg L−1	0.001 μg L−1	0.004 μg L−1	[195]
SPR	topramezone	cucumber, corn	2.4–7.3	1–200 μg L−1	0.61 μg L−1		[237]

) presents the most important analytical parameters obtained at pesticide analysis by optical sensors, applied on a wide range of food and water samples. Linear ranges, relative standard deviations illustrating precision, low limits of detection and determination, characterize the optical techniques applied to a variety of analyzed samples like fruits, vegetables, cereals, juices, milk, coffee, tea, eggs, honey and different types of water samples. The Limit of Detection (LOD) is the smallest analyte amount that can be discriminated from the background noise, whereas the Limit of Quantification (LOQ) is the lowest concentration that can be reliably quantified with satisfactory reproducibility, accuracy and precision.

The table below (

Table 2. Performances of several representative optical sensors in food, water and soil assay.

Sensor	Sensitivity	Selectivity (Interferences)	Stability (Life Time)	Pre-Analysis Time (Not Including Sample Pre-Treatment)	Response Time (Without Sample Pre-Treatment)	Sample Pre-Treatment Steps	Ref.
Colorimetric sensor for dimethoate based on Ag2O particles	0.011 as the slope of the calibration, against μg L−1 as concentration units	The massive anions can markedly interfere with dimethoate detection due to inhibition on Ag2O catalytical activity; outstanding selectivity against other competitive pesticides	Reported remarkable stability	Ag2O mimicking oxidase-like activity, and dimethoate were added to a 1.5 mL centrifuge tube, mixed and incubated for 10 min at room temperature	Tetramethylbenzidine molecules in the catalytic solution could be oxidized to form blue products within 10 min, leading to the analytical absorption peak at 652 nm	The analyte was evenly sprayed on the surface of the vegetable samples, that were dried into the fume hood for 12 h. The samples were eluted with 5 mL acetate buffer, and then the eluate was filtered through a needle filter having a 0.22 µm pore size	[79]
Colorimetric biosensor for paraoxon based on iodine-starch	4.7 ppb, reported lower than the maximum residue limits in the EU pesticide database (10 ppb)			Paraoxon was incubated with acetylcholine esterase at room temperature for 30 min; acetylcholine and choline oxidase were added, followed by further incubation for 30 min. After potassium iodide, horseradish peroxidase and starch addition, the absorbance was measured at 572 nm	High sensitivity for the assay of paraoxon residues with a reaction time of about 60 min	Vegetable irrigation water samples were twice filtered through a 0.22 µm membrane, and the filtrate was collected. The analyte was spiked, and then the colorimetric biosensor was applied to the detection of the spiked paraoxon concentration in the samples.	[202]
Colorimetric and fluorescent dual-mode biosensor for organophosphorus pesticide based on polysaccharide stabilized core-shell nanoflowers	0.3/log c of dimethoate (mg L−1) in colorimetry; −374.5/log c dimethoate (mg L−1) in fluorimetry	Fenvalerate, tebuconazole, thiamethoxam, glucose, fructose, and metal ions including Na+, K+, Zn2+, Mg2+ and Ca2+ did not interfere, only dimethoate being significantly responsible for the biosensor’s signal	Stable for 7 days in the liquid state	50 μL of glyphosate solution was mixed with 50 μL acetylcholinesterase at 150 U/L followed by 15 min incubation. 150 μL acetylcholine 1 mM was added, and the mixture was incubated for 30 min. The organophosphate-acetylcholine esterase-acetylcholine solution was combined with 200 μL bimetallic nanoflowers and 50 μL of an acetic acid/sodium acetate buffer solution. 1 mL of tetramethyl benzidine 0.6 mM pH = 4.0 was added to the mixture. The reaction was performed in a thermomixer at 40 °C for 10 min.	Bimetallic nanoflowers with oxidase-like activity reacted between 3 and 5 min The absorbance was read at 652 nm	Lettuce, cucumber and melon were crushed first, the supernatant was centrifuged and diluted 10 times with deionized water. Then, different amounts of glyphosate were added. The spiked samples were assayed.	[205]
Colorimetric and fluorescent dual-response aptamer sensor for thiamethoxam and acetamiprid based on gold nanoparticles	0.006629/nM for thiamethoxam and 7.782/nM for acetamiprid as slopes of the calibration	The dual target sensor functioned without interference for the two target analytes	Within 55 min, for thiamethoxam, the corresponding aptamer was degraded by more than 95%; only less than 50% of the thiamethoxam aptamer/gold nanoparticles was degraded	15 min as the best incubation time between aptamer and gold nanoparticles	Fluorescent signal stabilized after 3000 s	Cabbage, tomato and potato samples were homogenized with an appropriate amount of Tris-HCl buffer. The formed free-flowing puree was centrifuged at 5000 rpm for 30 min. Then, the precipitate was removed and the supernatant was filtered using a 0.22 μM membrane. The samples were then mixed with certain analyte amounts and analyzed	[206]
Colorimetric and fluorescent detection for carbosulfan based on Fe-N/C single-atom nanozyme integrated smart hydrogel	8.28 as slope of the calibration, developed against concentration (μg mL−1)	Common cations and anions (Ca2+, Na+, Mg2+, Al3+, CO32−, SO42−, NO3−), biomolecules (L-Histidine, BSA, glucose, glycine, citric acid), and other common pesticides (dimethoate, glyphosate, carbaryl, thiamethoxam, metiram, acetamiprid, deltamethrin) had a minimal impact at 10 times higher concentration; glutathione, cysteine, and L-ascorbic acid, at concentrations higher than 1 mg mL−1 affected colorimetric and fluorescent signals	After 14 days of storage at 4 °C, the catalytic activity of the sodium alginate hydrogel probe preserved its stability, and the fluorescence intensity retained more than 89% of its initial value	The hydrogel beads adopted a stable light blue coloration after 40 min of incubation	The hydrogel probe has short detection time (15 min)	Known concentrations of carbosulfan were sprayed on brown rice, wheat, and soybean (10 g samples). The treated samples were left to stay 1 h at room temperature, followed by refrigeration overnight. Each sample was subject to extraction with ethyl acetate/acetone solution, via shaking for 10 min. Centrifugation at 8000 rpm for 5 min was followed by supernatant filtration through a 0.22 μm membrane. The filtrate was evaporated in a fume hood, and the residue was redispersed in 1 mL methanol	[207]
Fluorescent sensor for organophosphorus pesticides relying on gold nanoclusters	0.38 as slope of the calibration, against concentration μmol L−1(μg L−1)	Na+, Mg2+, Hg2+, K+, Cr3+, Ca2+, Cd2+, SO42−, PO43−, CO32−, Cl− did not interfere at 1 μmol L−1		Parathion methyl was added to the bovine serum protein-protected gold nanoclusters/acetylcholine esterase system. After incubation for 20 min, acetylthiocholine iodide was added to the mixture solution	The fluorescence signal of the solution was measured after 15 min	The water samples were spiked with different amounts of parathion methyl (1–5 μg L−1), then filtered, and centrifuged. Phosphoric acid and ferrous sulfate (0.1 mol L−1 solution) were added to remove any free chloride ions and oxidants. 1.0 mL copper sulfate solution, served to remove microorganisms. Finally, the water samples were distilled. The apples and cucumbers sprayed with analyte were chopped then crushed. The resulted homogenates (20 g) were dissolved in 20 mL methanol, the obtained dispersion was filtered with a membrane, and the juice was subject to further experiments.	[208]
Fluorescence sensor for glyphosate based on papain-stabilized gold nanoclusters		Na+, Mg2+, D-glucose, D-fructose, soluble starch, glycine, L-tryptophan, L-glutamic acid, BSA (30 μg·L−1) did not give major interferences	Characterized by good photostability, preserving approximately 96% of the initial fluorescence intensity after continuous irradiation for one hour	Highest fluorescence intensity after 6 h incubation time for papain-gold nanoclusters system	Tap water samples with different concentrations of glyphosate and tyrosinase/dopamine were dropwise added onto the test strip and incubated for 15 min; the glyphosate concentration was assessed relying on the fluorescence color.	40 μL of different glyphosate concentrations (0–10 μg·L−1) were mixed with 40 μL tyrosinase (250 U·mL−1), and the solution was shaken for 30 min at 4 C. After addition of 80 μL dopamine 10 mM, the solution was incubated for 1 h, and 60 μL of papain-gold nanoclusters were added. The solution was diluted to 400 μL with Tris-HCl buffer and shaken for 10 min. Tap water samples were initially filtered through a 0.22 μm membrane and then diluted 10 times using Tris-HCl buffer. Analyte solutions (0.1, 0.2, and 0.4 μg·L−1) were added.	[209]
Fluorescence sensor for organophosphorus pesticides based on enzyme inhibition	189.519 as slope of the calibration plotted against concentration mg L −1	The system was not affected by interfering substances (Na+, K+, Cl−, Mg2+, Zn2+, Ca2+, Hg+, Fe3+, carbaryl, quintozene) when detecting organophosphates	After 40 days, it was noticed that the quantum dots’ fluorescence intensity still maintained 98% of the initial value, with outstanding stability	Under the optimal detection conditions, glyphosate was added to acetylcholinesterase (1 U/mL, 50 μL) and incubated for 5 min	The fluorescence intensity of the quantum dots was quenched with increasing acetylcholinesterase-acetylthiocholine iodide reaction time. At 60 min, the fluorescence intensity attained about 80% quenching level, considered for detection	The real samples of cabbage and potatoes were pretreated: 1 g was immersed into 0.01 M phosphate buffer solution and sonicated for 5 min to extract the analyte. After standing for 1 min, the supernatant was collected as the real sample for fluorescent assay. If the food was acidic or alkaline, the dosage could be tuned to hamper steep pH changes. Also, phosphate buffer solution can be employed to ensure the pH value at 7.50	[150]
Fluorescence imprinted sensor for malathion based on N-doped carbon dots and metal organic frameworks	0.1882 as the slope of the calibration plotted against concentration μM	K+, Ca2+, Na+, Cl−, Cu2+, SO42−, Pb2+, Mg2+, Zn2+, Mn2+, NO3−, Ni2+, and Co2+, malathion, thiamethoxam, imidacloprid, diazinon, phoxophos, glypho sate, dimethoate, and chlorpyrifos did not give major interferences	The stability of the sensor’s fluorescent response did not significantly alter during the first 20 days	The optimal imprinting time was 12 h	The fluorescence enhancement efficiency of the ratiometric sensor reached stability within only 1.0 min	Lettuce samples were washed with water and then spiked with malathion standard solutions (1 μM, 5 μM, and 10 μM). After drying at room temperature for 6 h, the spiked samples were cut into small pieces. 25 mL of methanol–water mixture (6:4, v/v) was added to the final solution, followed by sonication for 30 min and centrifugation at 10,000 rpm for 10 min. The obtained supernatant was filtered on a 0.22 μm membrane. The soil samples were filtered through qualitative 0.22 µm filter membranes followed by centrifugation for 15 min at 8000 rpm. The samples of lettuce, soil, and tap water, were spiked with malathion (1 μM, 5 μM, and 10 μM), then added to the ratiometric imprinted fluorescence sensor	[211]
Dual fluorescence-chemiluminescence sensor for glyphosate based on green biological metal organic framework	111.31 per μg mL−1 as slope of the calibration in fluorescence and 5891 per μg mL−1 in chemiluminescence	Methionine, lysine, aspartic acid, glutamic acid, ions Na+, K+, Mg2+, Ca2+, Fe2+, Fe3+, glucose, caffeine, methyl cellulose, chlorpyrifos, dichlorvos, parathion at 10-fold higher concentration did not interfere	The green biological metal organic framework preserved over 90% of its initial activity after 30 days of storage showing its long-term stability and suitability for the detection platform	The incubation time of glyphosate with biological metal-organic frameworks was 20 min	10 min was selected as the reaction time for detection	Drip bag coffee, instant coffee, cold brew freeze-dried coffee, green juice solid drink and green tea samples were analyzed. 2.00 g of the sample were transferred into a conical flask. Next, 50 mL of deionized water was added followed by ultrasonic extraction for 20 min. The resulting solution was filtered through a 0.22 μm aqueous membrane filter. A standard addition method was subsequently applied	[214]
Chemiluminescence sensor for thiram based on gold nanoparticles and peroxyoxalate chemiluminescence system		Anti-interference potential of the chemiluminescence-based sensing system against thyocyclam, methamidophos, atrazine, dimethoate, malathion, ammonium glyphosate, 2,4-D	Photostability of the chemiluminescent sensing system—120 h		Detection time of 40 s		[142]
Chemiluminescence sensor for acetamiprid based on indirect competitive immunoassay	IC50 10.24 μg L−1	The cross-reactivity rates of four neonicotinoid analogues (nitenpyram, thiacloprid, thiamethoxam, and clothianidin) were all less than 10%		The pre-incubation time of anti-acetamiprid monoclonal antibody with the analyte solution was 30 min	Optimal reaction time of chemiluminescence 20 min	Homogenization, acetone addition to the juice, filtration, centrifugation at 6000 rpm for 5 min, and filtration through a 0.22 µm filter membrane	[147]
Chemiluminescence sensor for glyphosate based on aggregation of gold nanoparticles and the chemiluminescent signal of the glyphosate-glyphosate binding aptamer	1817.9 as the slope of the calibration built against analyte concentration (μg L−1)	In the specificity test of the glyphosate-binding aptamer, only glyphosate and profenofos were distinguished among the fifteen tested pesticides		Chemiluminescence intensity of glyphosate binding aptamer-glyphosate indicated a complete reaction within 15 min, which remained stable thereafter	Chemiluminescence signals of glyphosate were recorded versus concentrations at 15 min, under the optimal conditions	The organic and free-spraying soybeans (2 g) were introduced in a 15 mL polypropylene centrifuge tube with analyte solution (10 mg L−1) at 4 °C for 1 h, and dried in a hood for 2 h. The QuEChERS extraction was applied. Grinding, placing 2 g in a centrifuge tube, addition of 10 mL deionized water for 10 min, addition of 10 mL methanol solution containing 1% formic acid, homogenizing in a high-speed homogenizer, followed by 5 min of agitation, centrifugation for 10 min at 15 °C, and eventually filtration through 0.22 μm polyvinylidene difluoride membrane.	[149]
Chemiluminescence FIA sensor for diquat based on oxidation with ferricyanide	85.54 as the slope of the calibration built against concentration μg mL−1	Ca2+, Mg2+, K+, NH4 +, Fe3+, Pb2+, Cu2+, Hg2+, Mn2+, Cl−, SO4 2−, NO3−, HPO4 2−, C2O4 2−, NO2−, urea, paraquat gave an error smaller than 5%.			Fastness given by a high sample throughput—144 samples per hour	Ground and river waters were filtered with polyamide membrane filters of 0.45 μm (for tap and mineral waters filtration was not necessary) and stored at 4 °C in the refrigerator. They were used within 1 week. In some cases, it was required to remove anionic interferences by prior passage of the spiked sample, through an anionic-exchange resin	[215]
Chemiluminescence sensor for acetamiprid based on graphene oxide/gold nanoparticles	467.00 per acetamiprid concentration (10−10 mol L−1) 654.87 per concentration in nmol L−1)	Specificity in the presence of 2,4-D, chlorpyrifos, imidacloprid, bisphenol A, omethoate, dipterex, parathion methyl, isoprocarb	The storage stability of the gold based nanoparticles sensor was about half a month; graphene oxide/gold nanoparticles had good stability when stored for six months	The aptamer solution was incubated with the analyte for 15 min at room temperature yielding aptamer folded conformation; then, graphene oxide/gold nanoparticles (150 μL at 10 μg mL−1) were added to the acetamiprid-aptamer solution and interacted with the remaining unfolded single-stranded DNA aptamers for 5 min	The chemiluminescent signal was recorded immediately after injection of luminol–H2O2 solution; aptamer/acetamiprid/graphene oxide/gold nanoparticles nanocomposite yielded a maximum response at 150 s	The wastewater samples were filtered through a 0.22 μm membrane. The soil samples were dried to constant weight and ground, and then spiked with acetamiprid standards, followed by ultrasonication, extraction with dichloromethane, filtration, centrifugation, evaporation of the extracts in a rotary evaporator. The residue was dissolved in 10% ethanol–water. The cucumbers and apples were homogenized in a mortar, centrifuged and filtered through a 0.22 μm membrane. The filtrates were employed as solutions of agricultural samples. Standards were added to the sample solutions.	[216]
Portable microfluidic point-of-source-testing chemiluminometry for malathion	80.139 as the slope of the calibration developed as signal intensity against concentration (ppm)	The technique was specific for malathion, with negligible interferences from other compounds, including inorganic and organic species		Microfluidic paper-based analytical devices were heated for one minute of incubation before all the experiments	Chemiluminometric reactions were instantaneous	0.5 kg of apples, tomato or citrus fruits were cleaned with deionized water; 50 mL of rinsed water were collected in a centrifuge tube. 12 μL of the retrieved solutions were coated to the microfluidic paper-based analytical devices pre-treated with chemiluminometric reagents, followed by drying in the oven at 30 °C overnight. To assess the sample concentration, known concentrations were spiked over the microfluidic paper-based analytical devices.	[217]
Electrochemiluminescence aptasensor for profenofos, isocarbophos, phorate, and omethoate based on copper-gold bimetallic nanoparticles	−732.48, −804.61 −813.55 −938.21 for profenofos isocarbophos phorate omethoate, as the slopes of the calibration graphs, developed versus log c (μg L−1)	The aptasensor had good specificity for four organophosphates	The electrochemiluminescence intensity of the electrodes after seven days was 97.95% of the electrodes’ signal before the seven days of assay (RSD = 3.33%); fourteen days later, the electrochemiluminescence intensity of the four electrodes decreased by only 7.32% (RSD = 6.95%)	The electrochemiluminescence intensity progressively diminished as the incubation time became longer, but the change began to become insignificant after 50 min; it took about 50 min for the organophosphorus pesticides to fully combine with the aptamer	16 s detection time of the electrochemiluminescent aptasensor based on copper-gold bimetallic nanoparticles	The vegetables were cut into small pieces (1–2 mm). 2 g were weighed into a centrifuge tube, then certain analyte amounts were sprayed on the vegetables. The samples were left in a greenhouse for 24 h. 1 mL of acetone and 9 mL of 0.01 M phosphate buffer solution pH 8.0 were added, followed by sonication for 20 min, centrifugation at 12,000 rpm for 15 min, and collection of the supernatant.	[218]
Electrochemiluminescence sensor for atrazine using silver nanoparticles and hydrogen peroxide decomposition	−549 per lg C (μg L−1)	Six single interfering pesticides (imazine, bromoxynil, 2,4-D, propanil, paraquat, and malathion) could not effectively obstruct the luminescence intensity given by the analyte, under the same concentrations	Six electrodes were stored in a dark experiment room for detection, and three of them were tested every 5 days. After 5 days, the electrochemiluminescent intensity of the three electrodes was reduced by 2.70% (RSD = 4.99%). After 10 days, the electrochemiluminescent intensity of the remaining three electrodes was reduced by 3.71% (RSD = 3.90%).	Signal stabilized at 30 min incubation time	Maximum electrochemiluminescent signal obtained after 2–3 s for luminol, silver nanoparticles, aptamer, bovine serum albumin and atrazine, respectively	The cabbage was washed, dried, and ground, then 10 g of the cabbage homogenate were weighed. 28 mL of methanol and 12 mL of phosphate buffer solution (0.01 M, pH 9.0) were added and fully mixed under ultrasonication for 20 min. After passing on filter paper, the filtrate was centrifuged at 10,000 rpm for 5 min. Eventually, the extracted supernatant was employed as the sample solution. For the soil samples, the same treatment was applied, except for the washing, drying, and grinding steps.	[154]
Near-infrared-excitable acetylcholinesterase-activated fluorescent probe for dichlorvos, carbofuran, chlorpyrifos, methamidophos	IC50 values of the tested pesticides, showed by the slopes were 0.344, 11.5, 82.9 and 89.0 μg L−1, respectively showing the reported highly sensitive response in acetylcholine esterase inhibition	The anti-interference potential of the developed probe was certified only at chlorophyll concentration below 0.01 μg μL−1		10 μL of acetylcholine esterase was incubated with 250 μL sample at 37 °C for 10 min; 10 μL of acetylcholine esterase-activated NIR fluorescent probe was then added and the signal was measured after another 5 min of incubation		Lettuce, beet and carrot were chosen as matrix models, and dichlorvos as pesticide model. Matrix-matched calibration curves of dichlorvos in various matrices were assayed. Each unspiked sample was extracted with buffer solution, and the obtained extract was either diluted 20 times in volume or used directly to obtain two series of dichlorvos solutions. All these solutions were employed to develop the individual matrix-matched calibration.	[162]
SERS sensor based on gold nanoparticles for chlorpyrifos	373.22 as the slope of the calibration, plotted against concentration mg mL−1	The applied QuEChERS technique during pre-treatment served to remove carbohydrates, proteins, fats and other potential interfering compounds	The gold nanoparticles were reported for their outstanding chemical stability, reproducibility, ultrasensitivity, and the limit of detection was as low as 10 μg L−1		Using a portable Raman spectrometer system combined with a 785-nm excitation wavelength diode-stabilized stimulator, the acquisition time was 10 s with three accumulations	5 mL of ultra-pure water was added to a 10 g soil sample and vortexed for 30 s. After addition of 10 mL acetonitrile 1%, the mixture was vortexed at 400 rpm for 3 min, followed by 2 min ultrasonic oscillation; the sample was left for 15 min, and then 4 g sodium acetate and 3 g NaCl were added. The resulted solution was vortexed for 1 min at 400 rpm and centrifuged at 5000 rpm for 5 min. 1.5 mL supernatant, 50 mg N-propyl ethylenediamine, 10 mg graphite carbon black, 150 mg magnesium sulfate and 50 mg C18 were added, followed by centrifugation of the supernatant for 1 min to remove carbohydrates, proteins, fats and other compounds. Eventually, the solution was centrifuged for 5 min at 5000 rpm and then the supernatant was passed through a 0.22 μm organic film.	[238]
Direct surface plasmon resonance biosensor for triazophos based on sensor chip with immobilized antibody	Sensitivity to the analyte given by an IC50 around 1 μg L−1	Other eight tested pesticides gave negligible responses	The sensor chip could be regenerated for 160 cycles at least	Activation time of the carboxylic acid groups on the chip surface within 20 min, and the monoclonal antibodies were immobilized to reach the saturation plateau above 30,000 resonance units	Assay time of 7 min per cycle	Homogenized cabbage, cucumber or apple samples (10 g) were spiked with standard triazophos (10–100 ng g−1). After 2 h of incubation at room temperature, extraction and purification by QuEChERS was applied. Sample pretreatment gave efficiency in cleaning-up complex samples; dilution (10-fold for Chinese cabbage and cucumber, 20-fold for apple) was needed for the purified samples	[37]
SPR nanosensor for coumaphos based on molecular imprinting	0.0314 as slope of the calibration graph between 50 and 300 ppb; 0.2839 as slope of the calibration graph between 0.1 and 25 ppb	High ratio of selectivity values versus diazinon, pirimiphos-methyl pyridaphenthion, phosalone, N-(2,4-dimethylphenyl)formamide, 2,4-dimethylaniline, dimethoate, phosmet, amitraz and parathion-ethyl	After two weeks, the activity of the coumaphos-imprinted sensor was 87% of the initial one	500 s	300–2000 s	Honey samples (1:5 ratio in water) were spiked with 100 ppb analyte and then passed through the sensor system	[195]
SPR sensor based on Fe3O4@Au@polydopamine core-shell magnetic nanoparticles for tebuconazole	0.03 as the slope of the calibration graph, plotting the angular shifts against concentration (μg L−1)		RSD between parallel films of the sensor was 3%, indicating very good stability and controllability. When not carrying out experiments, the sensor was stored at 4 °C to lower molecular activity and retard decay	The bare gold film was immersed in 1 mg mL−1 dopamine Tris buffer, for self-polymerization for 30 min. It was kept in a dark environment, then removed, rinsed with deionized water, and dried with nitrogen. Then it was introduced in a HAuCl4 solution 0.01 for 30 min to reduce gold nanoparticles. After rinsing and drying, it was re-introduced to the flow tank. Monoclonal antibody 100 μg mL−1 was injected in the flow cell and kept for 3 h to achieve immobilization; incubation with BSA 10 mg mL−1 solution followed for 30 min, to hamper nonspecific binding.	Sensor response at various analyte concentrations stabilized after 6 min	Samples were homogenized into pulp, 10 g of this homogenate were mixed with 10 mL methanol, the mixture was shaken for 20 min, and then filtered. Subsequently, tebuconazole standard solution with concentrations ranging from 30 μg L−1 to 100 μg L−1 was added to cucumber and corn extracts	[237]

) also presents essential analytical performances of the developed optical sensors. Various detection modes and development strategies are developed. The chemical and biochemical sensors are sensitive, selective, and fast. Nevertheless, as shown, factors like the sample-pretreatment steps and the incubation time of the biorecognition element may influence total duration.

8. Critical Conclusions and Future Prospects

Sensitive and rapid trace detection technologies are necessary to preserve food safety. Monitoring key food analytes is essential; therefore, analytical technologies are required for ensuring health and a good life quality [239]. With the ongoing necessity of preserving food quantity and quality [240], pesticides have been applied in agriculture to enhance productivity and hamper disease spreading [241]. On the other hand, it is important to stress out the relevance of contaminant assessment [242].

Optical sensors have been employed for the detection of significant markers in fields such as food, environment, and health. Optical sensors’ functioning relies on colorimetry, fluorescence, chemiluminescence, near-infrared spectroscopy, surface-enhanced Raman spectroscopy and surface plasmon resonance, included in a plethora of applications. The interaction between analyte molecules and radiation (ultraviolet, visible, or infrared) is followed and transduced into a measurable signal [243].

Optical fiber sensors use the principle of total light reflection, have been proved as sensitive, cost-effective, with facility of operation procedure, resistance to interferent radiation, and ability to be integrated in online or at distance monitoring [244].

The integration of nanomaterials, metallic nanoparticles, carbon-based materials, nanopolymers, nanozymes, in optical sensors, leads to improved analytical performances, given their particular physico-chemical features. Nanomaterials, with sizes between 1 and 100 nm, given their catalytic properties, elevated surface areas and high signal-to-noise ratios, can avoid some shortcomings, promoting increased sensitivity and real-time use [245].

The application of gold/silver nanoparticles or quantum dots involved for many years the use of traditional techniques, with associated toxicity of the required chemicals. Recent, novel trends are based on green technologies, renewable resources and low energy expenditure, leading to sustainability of designed sensors in the health and the environment sector. Employing safer chemicals and precursors, and different plant extracts as a source of reducing agents, proved as viable alternatives for silver or gold nanoparticle synthesis [242,246,247,248].

Novel multiplexed sensors, detecting two or more analytes or biomarkers, have been designed for complex assay, in the environmental or medical field. The multi-analyte sensing platforms can be developed by the modification with various sensing probes or by selection of the distinct characteristic wavelength for each analyte [242].

Colorimetric analyzers provide the possibility of real-time determination, given the low energy and reagent requirements of microfluidic systems [239]. Nevertheless, the colorimetric signal is susceptible to temperature variations and requires constant calibration by reference standards to diminish the impact of these alterations [249]. Moreover, the analysis accuracy can be impacted by the sample’s turbidity [250]. The colorimetric reagents and standards required are often hazardous and toxic, rendering the design of the remote sensing platforms more difficult. Interferent effects can contribute to the sensor’s response, at a certain wavelength [251,252,253].

Colorimetry is cost-effective, portable and characterized by simplicity of analytical procedure. Nevertheless, it is not applicable for measuring high concentrations. Errors can occur in the presence of unstable light sources, that give changes in light intensity [254].

Although fluorescence does not provide minutious information regarding structure, fluorescence spectroscopy is successfully applied to quantitative assay of complex samples, associated with multivariate statistical treatment of acquired data.

Fluorescence occurs in various systems, irrespective of the aggregation state. Only a few molecules show fluorescent features exploitable for analytical purposes, making the use of fluorescence less extensive than UV-Vis absorption, though fluorescent measurements are considered more selective. Nevertheless, in some cases, fluorescence spectrometry is preferred to molecular absorption spectrometry given the low detection limits and high selectivity. More chemical species can contribute to the fluorescent properties of a sample, but selectivity can be improved by choosing excitation and emission wavelengths [255].

Fluorescent responses can be easily impacted by pH changes and oxygen amounts. Autofluorescence can influence the analytical signal. The presence of foreign materials in biological samples can induce toxicity. The fluorophores can have a short lifespan, and other shortcomings are linked to photostability and impairment of recognition ability [256].

A novel dual-mode sensor was developed, relying on Cu²⁺/Cu⁺ redox-cycling, that underlied the optical signal transduction for organophosphorus pesticide residues, relying on acetylcholinesterase inhibition. In this innovative study, both UV-Vis absorption and fluorescence intensities exhibited gradual attenuation with the increase in triazophos concentration. UV-Vis assay had a LOD of 8.98 ng mL⁻¹. Fluorescence achieved an improved LOD of 1.72 ng mL⁻¹ [257].

Luminescence benefits from good sensitivity and quick detection, broad dynamic range, simplicity of the equipment and reduced costs, not requiring external optical light sources [258].

Fluorescence and chemiluminescence techniques are recognized for their high sensitivity, accuracy and specificity, able to lower the impact of interferents (complex substrates or colorants) in samples such as coffee or medicinal plants. Self-calibration of the analytical signal using dual-mode sensing techniques further improves the afore-mentioned analytical parameters in complex media, extending the application range. A Cerium-based biological metal organic framework, denoted as Suc-Ce-OH, was prepared using a one-step hydrothermal technique employing Ce⁴⁺ metal centers, to which succinic acid molecules were associated as ligands. After alkaline treatment, the resulted Suc-Ce-OH as hydroxylated metal organic framework rich in Ce–OH active sites, adopted significant phosphatase-like activity and substrate affinity, promoting both the hydrolysis of 4-methylumbellione phosphate disodium salt and the dephosphorylation of 3-(2′-spiroadamantyl)-4 methoxy-4-(3′-phosphoryloxyphenyl)-1,2-dioxetane, leading to blue fluorescent and strong chemiluminescent signals, respectively. The hydroxylated biological metal organic framework enabled dual-mode glyphosate detection, relying on the pesticide’s inhibition on phosphatase-like activity of Suc-Ce-OH. The developed sensing platform had operational facility, was green, fast, sensitive, and proved specificity in complex media like coffee or food samples. The detection limit in fluorimetry was 0.081 μg mL⁻¹ and about twice smaller in chemiluminescence, 0.038 μg mL⁻¹ [214].

The main advantages of chemiluminescence lie in the broad dynamic range, excellent signal intensity and high specificity, low background signal, rapid data acquisition, low reagent consumption, stability of reagents and their derivatives, the possibility of integration in immunoassays with low incubation time [259,260].

Nevertheless, different external physicochemical factors may negatively impact the sensitivity and selectivity of chemiluminescent measurements [261].

Moreover, the maintenance of appropriate operational features may be conditioned by skilled operation of expensive, complex equipment [262].

Vibrational spectroscopy is applicable for the study of a broad range of sample categories and can be performed for detection, but also for detailed qualitative and quantitative analysis [263]. NIR spectra are characterized by complexity and absorption bands may have broad overlapping, leading to difficulty in interpretation [155]. Chemometrics and machine learning algorithms can be integrated to improve accuracy.

The main mentioned shortcoming of different types of IR and Raman spectroscopy is unsatisfactory specificity at the assay of multi-analyte samples; nevertheless, it is considered that this disadvantage can be thoroughly counteracted by the rapidity and the non-destructive nature of the technique [264].

The localized surface plasmon resonance of metal nanoparticles can be associated with color modifications and subsequent absorbance measurement (as shown at the section colorimetry), or can be linked to a variation in the resonance wavelength/refractive index, correlated to the analyte concentration, as shown at the section surface plasmon resonance. Surface plasmon resonance is characterized by accuracy, high sensitivity, real-time monitoring without requiring labeling and possibility of miniaturization. High-throughput SPR biosensors have good specificity [265]. Also, the technique is non-destructive and non-invasive [266]. Sample detection can require bulky and costly instrumentation. However, recently, significant efforts have been directed towards the development of portable instruments [265].

The pros and cons of the biorecognition elements integration have been discussed critically. Aptamers are obtained via chemical synthesis, not necessitating animal sources, avoiding batch-to batch variations, lowering the costs and speeding the production process. Antibodies have been used as capture probes in biosensors, given their intrinsic affinity for their target and specificity. Nevertheless, the developed immunosensors can experience some limitations, associated to the biomolecule size, its conjugation abilities, stability, and expenses. Immobilization of antibodies is conditioned by protein functional groups, resulting in an unknown immobilization configuration, lack of uniformity, where the control exerted on orientation may prove difficult. The conjugation of the aptamer to the transducer’s surface or to a functional material relies on non-covalent interactions, like coordination interactions between DNA bases and gold surfaces, hydrogen bonds or π–π stacking established with aromatic-like structures like graphene oxides. Appropriate aptamer density on a certain surface is pivotal for obtaining enough binding sites for the target, counteracting steric hindrance and electrostatic repulsion between close aptamer molecules. Special attention should be paid at aptamer immobilization on nanostructures that have the propensity for steric hindrance [267]. Antibodies can be irreversibly denatured at ambient or higher temperature, but in the case of aptamers, the original conformation can be restored when optimal temperature is reached. Aptamers have lower molecular weight than antibodies, and can attain previously blocked or intracellular targets, facilitating detection [268]. Also, aptamers are prone to labeling by using linkers, reporter molecules, and other functional groups [267,269] and can detect a broad range of analytes [268,270]. They also have high binding efficacy and less complex structure [271]. In aptasensors, carbon-based nanomaterials like carbon nanotubes and graphene, given the high surface, electrical conductivity, and functionalization facility, enable the immobilization of high aptamer amounts, improving the electroactive area of the sensors. Carbon nanomaterials function as nanocarriers for signaling probes or as electroactive probes, providing an enhanced output signal [272]. Nevertheless, for aptamers, as nucleotides, there is an intrinsic limitation in their functional variability, when compared to proteins and the SELEX process needs ongoing improvement, to optimize aptamers availability. It has been highlighted that the comparison between the two biorecognition elements is not simple, as they should be regarded not necessarily as opposing, but rather as complementary [267].

Nanomaterials play a protective role, extending the enzyme lifetime, and promoting the stability and recyclability of co-immobilized enzymes [52]. Nanozymes overcome the limitations of natural enzymes, having improved stability and resistance to pH and temperature variations. Nevertheless, higher catalytic activity and selectivity, as well as better biocompatibility characterize natural biocatalysts. On the other hand, surface modification may be required to enhance performance of nanomaterials, and nanoparticles may induce cytotoxicity [273].

Pesticides can be detected by exploiting the recognition properties of calix[n]arenes. The role of the macrocycle proved pivotal in the recognition process [274]. Host–guest complexes rely on the coordination of a cationic, anionic, or neutral guest within an individual host molecule or supramolecular host assembly. Macrocyclic ligands are typical hosts, the complex cohesion being ensured by hydrogen bonding, van der Waals forces, ionic forces, or hydrophobic interactions [275].

Cucurbiturils, pillararenes, calixarenes and cyclodextrins are employed, and the resulted host–guest complexes show in many cases less toxicity than the pesticides, but preserve the strong herbicidal activity. These host–guest systems show high selectivity and specificity, tolerating a broad range of impurities without giving interferences. Many host–guest systems are characterized by wide linear ranges, with detection limits of the order of ng mL⁻¹, and their repeatable use does not impact efficiency. The formation of host–systems also helps in improving solubility and can delay pesticide decomposition [276].

Hybrid methods using electrochemiluminescent detection and aptamer biorecognition, enabled highly sensitive malathion quantitation, relying on the steric hindrance, able to restrict electron transfer and analyte diffusion on the electrode surface. The novel aggregation-induced emission sensor was based on red-emissive sulfur quantum dots, synthesized using a two-step oxidation technique, and exhibiting efficient electrochemiluminescent features. The red-emissive sulfur quantum dots were used as modifiers for the electrode, functioning as luminophores. Furthermore, the aptamer was modified in the form of a double-helix structure with the complementary DNA chain. The electrochemiluminescent signal was lowered, given the scarce electrical conductivity of the biomolecules and associated poor electron transfer. In the presence of the analyte, the aptamer double helix structure was untangled, and the analyte interacted with the electrode surface, restoring the electrochemiluminescent signal, leading to outstanding sensitivity, given by a detection limit as low as 0.219 fM [221].

A comparative analysis focused on optical methods applied to food, shows that UV-VIS molecular absorption spectrophotometry is characterized by simplicity of procedure and selectivity. Nevertheless, the sample pre-processing methods may be elaborated. Techniques including data processing algorithms are accurate, fast, but a high number of samples is required for calibration. Colorimetric techniques based on nanozymes are able to detect the analytes fast and sensitively. Nevertheless, the nonspecific absorption can affect the results’ accuracy. Techniques based on nucleic acid amplification have facility of procedure, benefit from low equipment requirements, and facility of sensor development. Fluorescent methods based on nanozymes are regarded as highly selective, sensitive and simple. The potential cytotoxicity and the impact on food safety should be considered. Sensors based on DNA probes have high stability and are viable for small molecule detection. Nevertheless, the preparation procedure is complex, and the costs involved may be higher, when compared to general food analysis [277].

The advantages of luminescence-based biosensors encompass high sensitivity, contactless detection ability and signal enhancement without major interferences. Due to the labeling of biomolecules, a broad range of analytes can be assayed. The potential drawbacks referred to potential matrix interferences, and a low intensity of the analytical signals [278].

Surface-enhanced Raman methods incorporating genetic tools yield a powerful signal and are highly sensitive and specific. Nevertheless, the substrate may involve high costs and can give interferences in complex samples. Immune techniques are specific and sensitive. Nevertheless, the synthesis process is complex and antigens can be broken down during food processing. Vibrational spectroscopic techniques are non-destructive, and the presence of wide bands that can severely overlap can be amended by inclusion of machine algorithm models, but further progress needs to be done in interpretability [277].

The integration of chemometrics in biosensing platforms is important to achieve signal processing and analysis [279]. The growing interest in machine learning algorithms provides improved signal processing in optical sensors, with automated analysis of spectrums, promoting accuracy and efficacy [244].

Portability represents an essential characteristic, to be exploited in contaminant level assay. Novel one-platform-two-channels detection techniques integrating smartphone-based reading and nanozymes, are evidenced by outstanding sensitivity and selectivity, mainly in colorimetric assay [75].

Human error in colorimetric signal analysis can be lowered by image processing algorithms integrated with spectrometers or smartphones. Machine learning can improve sensitivity by mitigation of fluctuations in lighting, or by hampering errors originating from handling by users, and background interferences [280].

Learning algorithms represent sound tools, that enable the detection and processing of rich, complex data sets, proving scalability. Machine learning facilitates accuracy and interpretation of data provided by wearable sensors [281].

Discussing and tackling the challenges that the field of nanosensors faces, is a key issue. The transition from lab scale to commercial use may be difficult, but the successful application in agriculture promotes productivity and sustainability [282].

Wearable integrated sensors, that received ongoing increasing attention, are prone to improvement in sensitivity and stability. Recently, a portable glyphosate photoelectrochemical sensor was developed, with C₆₀ encapsulated in a porphyrin-derived molecular organic framework. Given the delocalized conjugated structure, C₆₀ is endowed with outstanding electron buffering capacity, behaving both as acceptor and donor, and possesses distinctive light absorption ability. The encapsulation of nanofullerene tunes the charge distribution of Copper(II) meso-tetra(4-carboxyphenyl)porphine and responds to the Cu²⁺/Cu⁺ conversion, to enhance the redox peak current. Under illumination, fullerene enables charge transfer from the copper complex to fullerene, which promotes glyphosate adsorption. The photoelectrochemical sensor was characterized by reproducibility, stability, a broad linear range from 10 to 10⁸ pM and a low detection limit of 10⁻² pM. The practical ability of the portable sensor was proved at the assay of real samples, including real-time and in-field assay [283].

Regenerability, stability, and the performances at multiplex assay necessitate ongoing optimization. In antibody- and nucleic acid-based sensors, the biorecognition element binding to the target can affect regeneration, reproducibility and stability during the recognition step. These are key aspects in the functioning of wearable devices, alongside providing high quality data despite variability, ensuring data security and interoperability [284].

Machine learning and high-throughput computational methods enable accurate optimization of active-site configurations, prediction of structure–activity interplay, insight into the catalytic mechanism. The systematic study of nanozyme catalytic efficacy, stability, and substrate specificity enable the assay of complex media containing proteins, lipids, and nucleic acids. Future developments in the application of cascade nanozymes are based on artificial intelligence integration [285].

Cross-reactivity, as a sensor’s ability to respond also to non-targeted substances is regarded as a limitation, but can be exploited in sensor arrays (like “electronic noses”) to detect analytes rather via pattern recognition, than relying on specificity. Nanomaterials are characterized by versatility, facility of fabrication, and can improve sensitivity to target analytes. Carbon nanotubes, molecularly modified metal nanoparticles, metal oxide nanoparticles, and semiconducting nanowires can be integrated in sensor arrays [286].

Avoiding nanosensor fouling, can be attained by developing reproducible calibration methods, or by using preconcentration and separation techniques to reach an adequate analyte concentration that hampers saturation. It has been reported that graphene oxide/silver nanoparticles nanocomposites used in sensor coatings possess antifouling properties that can be exploited in sensor development [287].

One discussed issue is the nanomaterials’ toxicity, associated with health risks. It has been reported that smaller-size nanoparticles possess higher penetrability and reactivity in living tissues, when compared to bigger-size nanoparticles. Also, smaller nanoparticles can more easily avoid phagocytosis by immune cells. Other factors are the dose and the route of exposure: inhalation, parenteral, or intra-tracheal routes induce higher nanotoxicity than the oral or topical ones. The physiological barriers such as stratum corneum or mucous membrane hamper trans-membrane diffusion and penetrability of nanoparticles, diminishing their absorption into systemic circulation. Aggregated nanoparticles induce higher toxicity, when compared to de-aggregated or primary nanostructures. Surface functionalization with sodium citrate, polyvinylpyrrolidone, or surfactants proved their potential to suppress toxicity and immunogenicity. The diminished immunogenicity and toxicity of nanoparticles were assigned to the hampering of metal ions’ leaching-out. The complete delineation of these mechanisms can reduce health risks [288].

Batch-to-batch variability in nanomaterials encompasses discrepancies in key features (size, surface area, purity, shape) that can negatively impact reproducibility and reliability in research and commercialization. Several strategies aim to minimize the impact of batch-to-batch variability: monitoring physicochemical properties (by Transmission Electron Microscopy, X-ray diffraction, X-ray Photoelectron Spectroscopy, Brunauer-Emmett-Teller analysis); batch-to-batch variability inclusion in kinetic models to exert control on quality; moving from batch to continuous processes to improve control; using standardized materials like OECD reference nanomaterials for comparison. The OECD offers guidelines for testing nanoparticles, focusing on harmonized techniques for safety assessment. Nevertheless, to exploit their full abilities, they must be accurately developed, and checked in relevant reference models. Microfluidic systems can simulate conditions in particular microenvironments, becoming a viable strategy for both nanoparticles’ preparation and characterization. Microfluidic technologies enhance batch-to-batch reproducibility, by exerting control on the conditions of nanoparticle synthesis [289].

A new fully portable microfluidic Point-of-Source-Testing chemiluminometer, incorporating a semi-automated microfluidic valve to exert control on the reagent, and employing a microfluidic disposable paper-based analytical device, enabled sensitive assay of fruits and tomatoes, with a limit of detection of 0.016 ppm [217].

The expenses that an analysis involves includes the cost of skilled personnel, the cost of materials, instrumentation and reagents, with the possible integration into complex data assay systems. UV-Vis spectrometers used are often expensive. A SERS sensor with low-cost (less than 10 cents) was reported as appropriate for fast screening of triazophos [50,290].

Paper-based analytical techniques have attracted increased interest for pesticides detection, given their simplicity, portability, disposability, and cost-effectiveness [291].

Machine learning offers an improved insight on the analytical performance, enabling optimization of the materials’ structure. Paper-based microfluidics may further encompass developing prototypes of 3D devices operating on capillary flow, the design of techniques for mixing fluids and their inclusion in other microfluidic platforms. Novel developments enabled outstandingly sensitive assays, with limits of detection within the femtomolar level. Paper substrates associated with data analysis systems include microcontrollers and wireless systems. Hence, the features of paper-based substrates, such as real-time data transfer and at-distance monitoring can be improved. It is thought that emphasis should be placed on catalytic features and stability enhancement, miniaturization of electronic components, as in the case of lateral flow assay devices [292].

Novel materials are explored, like 2D materials, metal–organic frameworks, and nanocomposites, that impart outstanding sensing properties and enhanced stability. Advances in nanofabrication are 3D printing and nanoscale assembly, promoting the design of complex and highly specialized nanosensor structures. These methods enable the inclusion of multiple sensing modalities, such as optical, electrical, and mechanical sensing into a single device, promoting accuracy and extending the range of measurements [293].

The necessary transition to commercialization is still subject to rigorous validation, regulatory approvals, and cost optimization. Fluorescent and colorimetric chemical sensors are characterized by elevated sensitivity, being prone to applications in detecting contaminants. The cost-effectiveness of chemosensors is exploitable in sustainable monitoring of both food processing steps and quality. Algorithms facilitate data interpretation, particularly by discriminating low signals from noise. It is thought that chemical and biochemical sensors address limitations of conventional analytical methods, such as complexity of the procedure, high costs, and long time [294].

In conformity to the Optical Chemical Sensors—Global Strategic Business Report, the global market for optical chemical sensors was valued at 3.9 Billion US in 2024 and is projected to reach 9.4 Billion US by 2030 [295].

A series of aspects are essential in the field of nanosensors’ future application. The innovation encompasses the quest for materials with improved features and functionalities: stability, bio- and environmental compatibility. Integration with novel technologies such as artificial intelligence, Internet of Things, and blockchain technology can promote efficiency in data processing, security, and sensor design. Future works should investigate how these technologies can result in efficient and interconnected sensing devices, promoting transparent, straightforward and reliable mechanisms and networks for collecting, checking and storing data provided by sensors.

Priorities are cost-effectiveness, facility and efficiency in manufacturing, scalability, alongside high analytical performances, sensitivity, selectivity and functionality in peculiar applications like monitoring pollutants [296].

Green synthesis-based nanosensors are broadly employed, given the easiness of synthesis, availability, lack of toxicity, cost-effectiveness, but also precision and accuracy. Hampering the use of both toxic chemicals and high energy expenditure, as well as the integration of natural sources, significantly lowers the environmental impact. Green synthesis diminishes manufacturing costs and promotes the process’s scalability. Green and sustainable technologies can enhance integration of green nanomaterials in the industrial field. Emphasis should be placed on scalability of eco-friendly methods, promotion of functional features, lack of toxicity and safety.

Nanotechnology, namely where green materials are employed for nano-biosensing, has the potential to reshape material science, promoting safety and efficiency. Cooperation between researchers, policymakers, and industry leaders is necessary to successfully apply these technologies [297].