Ion-Selective Electrodes in the Food Industry: Development Trends in the Potentiometric Determination of Ionic Pollutants

Abstract

Food quality assessment is becoming a global priority due to population growth and the rise of ionic pollutants derived from anthropogenic sources. However, the current methods used to quantify toxic ions are expensive and their operation is complex. Consequently, there is a need for affordable and accessible methods for the accurate determination of ion concentrations in food. Electrochemical sensors based on potentiometry represent a promising approach in this field, with the potential to overcome limitations of the currently available systems. This review summarizes the current advances in the electrochemical quantification of heavy metals and toxic anions in the food industry using potentiometric sensors. The healthcare impact of common heavy metal contaminants (Cd²⁺, Hg²⁺, Pb²⁺, As³⁺) and anions (ClO₄⁻, F⁻, HPO₄⁻, SO₄²⁻, NO₃⁻, NO₂⁻) is discussed, alongside current regulations, and gold standard methods for analysis. Sensor performances are compared to current benchmarks in terms of selectivity and the limit of detection. Given the complexity of food samples, the percentage recovery values (%) and the methodologies employed for ion extraction are also described. Finally, a summary of the challenges and future directions of the field is provided. An overview of technologies that can overcome the limitations of current electrochemical sensors is shown, including new extraction methods for ions in food.

1. Introduction

Current projections estimate that a 56% increase in food production will be required to feed the global population by 2050 [1]. This trend in food production must be followed by an improvement in food safety, given the increase in ionic pollutants in the environment. Ionic pollutants are charged compounds present in the environment that can pose a risk to human health. These include heavy metal cations [2], or anions such as fluoride or arsenate [3]. Industrial processes including fertilizer use from agriculture [4], mining [5], pesticide use [6], or even car traffic [7] contribute to the contamination of food sources. Ionic pollutants can accumulate in the body, leading to long-term conditions with a subsequent increase in human and economic losses. It is estimated that the economic burden of foodborne illnesses can be as high as USD 90 billion annually [8]. Consequently, food safety and security is becoming a global concern, which requires affordable and accurate sensing systems to enable the assessment of food quality by a large population of people.

One of the main limitations of food monitoring is the lack of low-cost and portable sensing systems used commercially that can be used to maintain strong quality control. Within the past few years, multiple biosensors that can be used for the detection of biological agents such as food pathogens [9] or persistent organic pollutants [10] have been developed. Hydroquinone represents one of these persistent organic pollutants, which has a detrimental impact on bones and cartilage [11]. This molecule is often used as a food additive, in the form of tert-butyl hydroquinone [12], and it can also be found in cosmetics, dyes, and pharmaceuticals [13]. However, long-term exposure has been correlated to cytotoxicity and cancer [12]. Multiple electrochemical approaches have been reported to enable the detection of this compound, using inorganic materials such as FeWO₄ [14], palladium nanoparticles [15], and N-rGO/SrZrO₃ [13].

However, the accurate detection of small analytes such as ions and oxyanions still represents a challenge given the low ion size and the lack of appropriate receptors. Moreover, food samples are complex media, containing multiple compounds that can interfere with and limit the accuracy of sensors.

The current standards for ion quantification in food include inductively coupled plasma optical emission spectrometry [16] and atomic absorption spectrometry [17]. These methods are accurate, showing low detection limits. However, they are expensive, and they require lengthy preparation since they are sensitive to matrix effects [18]. Consequently, they are restricted to laboratory analysis, and cannot be used in situ. Within recent years, electrochemical devices that can determine the concentrations of ions in biological samples, with low detection limits and high selectivity, have been reported. These sensors can meet the requirements of cost and accuracy to be used in food samples within real-world scenarios. Different electrochemical approaches can be applied to the quantification of ions in food, including voltammetry, amperometry, and potentiometry [19].

In voltametric sensors, a potential is applied to the working electrode, leading to electrochemical reactions at the surface of the electrode that change the measured current. When a potential is applied in ion-selective electrodes operating under voltammetry, ions from the sample are transferred into a selective membrane. This ion transfer triggers a redox reaction ion-to-electron transducer layer, commonly made with conductive polymers [20]. Voltametric methods can be used for the multiplexed detection of ions, such as Li⁺ and Ca²⁺ [20], which could enable a complete assessment of food quality. This technique has been employed in the detection of multiple ionic pollutants in food, including heavy metals and anions. Li et al. [21] developed a voltametric sensor, based on Fe₂O₃/Bi₂O₃ nanocomposites for the simultaneous determination of Cd²⁺ and Pb²⁺ concentrations in water and milk samples. The final device showed a limit of detection within the nanomolar level for both heavy metals. Voltammetry has also been applied to anions such as perchlorate [22], and other common food contaminants such as hydroquinone [23]. Despite their low limit of detection and high sensitivity, these devices exhibit a short lifespan due to the delamination and leaching of membrane components. The use of thick membranes compared to ultra-thin films has been demonstrated to expand the lifespan of devices to 1 month [24]. However, ion-selective electrodes operating under potentiometry can typically be operated for several months [24]. Moreover, voltametric sensors can be sensitive to interference, especially in complex media such as food samples [25].

Amperometric sensors can determine analyte concentrations through changes in current, driven by specific redox reactions. The most common example of amperometric devices are glucose sensors, that make use of glucose oxidase, or other non-enzymatic materials such as copper oxide to oxidize glucose molecules [26]. The need for specific redox materials has limited the application of this technique to ion-selective electrodes. In most cases, this technique has been applied to the detection of oxyanions, such as sulphate, nitrate, and nitrite [27]. The detection of these oxyanions is possible through the use of enzymes, such as nitrate reductase, combined with electron carriers and mediators. Although these devices show a high selectivity and low detection limits, they are highly sensitive to interference, including that from environmental oxygen and the presence of bacteria. Moreover, the lifetime of sensors is short, restricted to 1–2 weeks of continuous operation [27].

Potentiometry offers several advantages compared to other electrochemical techniques, including high lifetimes, low cost, short response time, and wide detection ranges [28]. This method also leads to low power consumption, since no current flows between electrodes, and no external voltage is applied [29,30]. Moreover, electrodes can be designed to have low detection limits, enabling the ultra-trace detection of pollutants [31]. Consequently, they show promise in the development of advanced sensing platforms for food quality analysis, and can be easily used by consumers.

This review provides an overview of the current developments in the potentiometric detection of ionic pollutants in food, including novel materials and techniques used for measuring heavy metals and key anions. The performance of these sensors is evaluated against existing standards, along with their limitations. Sensor performances are evaluated in terms of the limit of detection, selectivity, and percentage recovery values. The manuscript is organized as follows. In Section 2, we provide an overview of ion-selective electrode technology, including an introduction to the structure and fabrication processes. In Section 3, we describe current advances in the detection of heavy metals (Cd²⁺, Hg²⁺, Pb²⁺, As³⁺), including the current landscape of each ion in terms of its impact on the food industry and its detection methods. Section 4 summarizes the reported work in the detection of key anions. Finally, current challenges in the field as well as future opportunities and trends are discussed.

2. Ion-Selective Electrodes

Ion-selective electrodes constitute a specific family of sensors used in the determination of ion concentrations. These sensors typically operate by potentiometry, and involve a permselective membrane, typically made of plasticized poly vinyl chloride (PVC), an ionic site, and an ionophore. The structure of an ion-selective electrode also includes a working and a reference electrode to enable the detection of ions. Two main configurations are used nowadays as ion-selective electrodes: liquid membrane and all-solid state (Figure 1a–c).

In liquid membrane ion-selective electrodes, a permselective membrane separates the sample from an inner solution. Ionophores selectively convey electrolytes from the sample phase into the inner solution, triggering a reaction between Cl⁻ from the inner solution and a Ag⁺ electrode, changing the measured potential. This approach constitutes one of the first technologies for ion detection that were developed, and, in general terms, shows high reproducibility and long-term stability. Consequently, it is one of the preferred options in commercially available devices. However, these sensors are expensive, with reduced portability, and require relatively large sample sizes to operate.

Within the past few years, all-solid-state configurations where a permselective membrane is directly placed onto a conductive electrode with an ion-transducer interlayer have been designed. Ion transducer films are intermediate layers that transform the changes in ion concentration into an electrical signal. This film also reduces the electrochemical noise due to the diffusion of water between the electrode and sensing film [33]. By avoiding the need for an inner solution, all-solid-state ion-selective electrodes can be fabricated within smaller dimensions when compared to liquid membrane devices, and they can also reduce costs by simplifying the fabrication process.

The deposition of plasticized films can be carried out using different low-cost methods, including drop-casting [34], aerosol deposition [35], and spin coating [36]. However, these sensors still have unsolved limitations regarding electrochemical stability, due to the diffusion of a thin water layer between the electrode and sensitive films [37]. Despite the limitations of all-solid-state configurations, devices can be fabricated at a relatively low cost, and the use of potentiometry can lower the detection limits by up to eight orders of magnitude [38]. The sensitivity of these devices is governed by the Nernst equation (Equation (1)):

$$E=E^0-\frac{RT}{zF}\mathrm{l}\mathrm{o}\mathrm{g}(a)$$

(1)

where E is the cell potential, E⁰ is the standard potential of the electrode, R is the universal gas constant, T is the temperature, z is the ion charge, F is the Faraday constant, and a is the activity of the ion in solution. Given the logarithmic nature of this equation, it is deduced that ion-selective electrodes are particularly susceptible to electrochemical noise. A 1 mV change in signal can lead to a 4% error in monovalent ions and over 8% for divalent ions [36]. Moreover, the standard sensitivity for ion-selective electrodes is in the range of 59 mV for positive monovalent ions at room temperature, and it decreases for cations or anions with higher valence. As such, keeping low potentiometric noise rates is key to ensuring accurate measurements and percentage recovery values close to 100%.

Both liquid membrane and all-solid-state ion-selective electrodes can be applied to the detection of a wide range of charged compounds, from cations and anions, up to small molecules such as histamines [39], and even pharmaceuticals like salbutamol [40]. To date, there are over 100 different ionic species that can be analysed using ion-selective electrodes, including potentially toxic compounds, such as heavy metals and oxyanions.

3. Ion-Selective Electrodes for Heavy Metal Measurement

Cations are ubiquitously present in all food sources, being crucial for the correct functioning of physiological conditions. However, to maintain a healthy diet, only a group of cations is required, and the excess of some others, such as heavy metals, can result in harmful effects on the organism. Sodium, potassium, calcium, and magnesium are amongst the most common cations involved in biological processes, while others such as zinc, iron, and copper are also required but at lower doses. On the contrary, there are some cations whose presence at any concentration can pose a risk to consumers. Consequently, their detection in food is critical to avoid economic and human losses worldwide. For example, in communities where there is a high daily intake of heavy metals, increased blood pressure and developmental abnormalities have been observed [41].

The definition of heavy metals includes those metals and metalloids with a density above 5 g/cm³ [42]. These elements often show carcinogenic and mutagenic effects [43], which can pose a risk to human health. Contamination with heavy metals represents one of the key concerns in fruit production. As part of their natural lifecycle, plants accumulate ions from the soil inside their organs, which leads to higher concentrations in fruits and leaves. Mawari et al. [44] analysed 24 different vegetables and fruits in India, showing alarmingly high levels of ions above the health risk index. Although this accumulation typically takes place near industrial and mining areas, as a consequence of pollution [45,46], it can also take place in greenhouses [47,48]. Consequently, contamination by heavy metals represents the fourth most frequent hazard notified in the Rapid Alert System for Food and Feed (RASFF) [49].

The heavy metals found in contaminated food most associated with human poisoning are cadmium, lead, and mercury [50]. The presence of these metals in food occurs naturally due to their presence in the environment and their intake by animals and plants. Therefore, it is expected that most food products contain them in variable concentrations. The maximum concentration limits to commercialize food containing these elements are regulated by the European Commission No. 1881/2006 [51]. Standards for some of these heavy metals in fruits, fish, and meat are shown in

Table 1. Maximum allowed concentrations for key heavy metals in vegetables, fish, and meat products as set by the European Commission. Limits for mercury content in meat is not specified by the European Regulation No 1881/2006.

Element	Maximum Concentration (mg/kg)
	Vegetables	Fish	Meat
Cadmium	0.05	0.05	0.05
Lead	0.1	0.3	0.1
Mercury	0.1	0.5	-
Arsenic	0.3	0.3	0.3

.

3.1. Lead Detection

Lead is a toxic heavy metal, with known long-term accumulation in the human body. The half-life of lead in blood and bones is 30 days and up to 20 years, respectively [52,53]. Within the past few years, increasing evidence of the toxic effects of Pb²⁺ ions has been reported, with an impact on the kidneys and bones and the reproductive, cardiovascular, and nervous systems [54]. Given the harmful effects on the organism, the imposed legal levels of lead are 0.1 mg kg⁻¹ in meat and vegetables [51]. Lead contamination can take place due to a wide variety of sources. Lead is a naturally occurring heavy metal, which can be present in the environment, leading to bioabsorption by plants. High lead contamination has been observed in mining areas [55], volcanoes [56], and also as a result of anthropogenic activities, such as waste from glass industries [57], and waste waters [58].

The most widespread methods for lead quantification involve the use of bulky and expensive equipment such as surface-enhanced Raman spectroscopy [59]. Using Raman spectroscopy, a limit of detection in the nanomolar range (~5 nM) can be achieved, with a good selectivity when compared to other heavy metals such as Cu²⁺, Cd²⁺, or Hg²⁺. Multiple ion-selective electrodes operating under potentiometry have also been reported. First, Pb²⁺-selective membranes were involved the use of soluble salts, mostly sulphides. However, these ion-selective electrodes suffered from interference from other heavy metals, such as Hg²⁺ and Cu²⁺ [60]. Since then, multiple neutral ionophores for the electrochemical and optical detection of Pb²⁺ have been developed. The lowest detection limits so far have been reported in lead-selective electrodes incorporating optodes, within the range of 10⁻¹² M [31].

The use of neutral ionophores inside a plasticized polymeric film represents the most established approach in the development of Pb²⁺-selective sensors. Lead ionophore IV represents one of the most widely used ionophores used in Pb²⁺-selective electrodes. Zeng et al. [61] developed a specific electrode based on lead ionophore IV embedded inside a PVC film with 2-nitrophenyl octyl ether plasticizer. The device showed a sub-nanomolar limit of detection (~0.6 nM), combined with high reproducibility and selectivity. Within recent years, additional ionophores have been designed. Özbek et al. [62] developed a sulfonyl hydrazone derivative for the detection of Pb²⁺ ions embedded inside a PVC membrane with dibutyl phosphate plasticizer. The device could measure Pb²⁺ concentrations in drinking water down to 3 µM over a wide pH range (2–12). However, the selectivity of the final sensor, especially towards common electrolytes such as sodium, was poor. Alternative ionophores such as N,N′-tetrabutyldipicolinamide [63], and (E)-2-(1-(4-(3-(4-chlorophenyl)ureido)phenyl)ethyli-dene)hydrazinecarbothioamide [64] embedded inside plasticized PVC films have been reported. However, these electrodes still lack enough selectivity for their use in complex samples that might contain moderate concentrations of common electrolytes such as Na⁺ or K⁺.

Despite the good selectivity of ionophores embedded inside plasticized PVC films, alternative materials have been developed to overcome the limitations of traditional electrodes, especially regarding long-term stability. In particular, polyaniline-derived materials have shown suitability as Pb²⁺-specific sensors with long lifetimes. Polyaniline-zirconium(IV) iodate membranes deposited by drop-casting have been tested in the detection of Pb²⁺, achieving a limit of detection of 3.3 nM [65] (Figure 2a), and reusability for at least 6 months. The final device could be used in the quantification of lead ions in drinking water with high percentage recovery values (~98%). Despite low detection limits, selectivity coefficients were high, especially towards common cations such as Na⁺ and K⁺. In both cases, coefficients were $log{K}_{{Pb}^{2+}, X}^{Pot}>-1$. Li et al. [66] overcame the challenges in selectivity through the incorporation of poly(aniline-co-2-hydroxy-5-sulfonic aniline) nanoparticles inside a dioctyl phthalate-plasticized PVC membrane, achieving a sub-nanomolar detection limit of 2.2 × 10⁻¹¹ M, and a lifetime of 15 months (Figure 2b). The device could be used for the determination of Pb²⁺ concentrations in both water and food samples, after extraction using HNO₃ and H₂O₂. In this case, the highest interference was K⁺, with an improved selectivity coefficient of $log{K}_{{Pb}^{2+}, K^{+}}^{Pot}~-3$. However, H⁺ also represented a significant interference, with a selectivity coefficient in the range of $log{K}_{{Pb}^{2+}, H^{+}}^{Pot}~-3$. As such, samples with a low pH, including those treated with acidic solutions for the extraction, have an increased level of error when compared to gold standards. This led to a significant error (>5%) when quantifying Pb²⁺ from green gram. Moreover, the synthesis of co-polymer nanomaterials required additional steps that could limit the commercialisation of devices, and the reproducibility of different sensor batches was not tested.

Polymeric nanoparticles in combination with multiwalled carbon nanotubes have also been developed through ionic imprinting [67]. These nanoparticles were embedded inside a PVC membrane and could be used for quantification in drinking water, showing lower selectivity coefficients compared to previous approaches. Ion-selective electrodes where an ionophore is embedded inside a plasticized film have also been tested in this field. Abraham et al. [68] developed a Pb²⁺-selective sensor using 1,2-Bis(N′benzoylthioureido)benzene as an ionophore, and reduced graphene oxide (Figure 2c,d). In this case, the detection limit was higher than polymer-based approaches (25 nM).

Figure 2. (a). SEM imaging of polyaniline-zirconium (IV) iodate nanocomposite, used for the detection of Pb²⁺ in drinking water. Figure reused with permissions from [65], Copyright, 2024 Elsevier B.V. (b). Mechanism of deposition and detection of Pb²⁺ using poly(aniline-co-2-hydroxy-5-sulfonic aniline) nanoparticles. Figure adapted with permissions from [66], Copyright 2011 American Chemical Society. (c). SEM imaging of reduced graphene oxide deposited onto glassy carbon as electrode materials for Pb²⁺-sensing devices. Figure reused with permissions from [68], Copyright 2015 Elsevier Ltd. (d). Representation of ionically imprinted polymer, and its application in sensing of lead ions in water. Figure reused with permissions from [67], Copyright 2024 Elsevier B.V.

Figure 2. (a). SEM imaging of polyaniline-zirconium (IV) iodate nanocomposite, used for the detection of Pb²⁺ in drinking water. Figure reused with permissions from [65], Copyright, 2024 Elsevier B.V. (b). Mechanism of deposition and detection of Pb²⁺ using poly(aniline-co-2-hydroxy-5-sulfonic aniline) nanoparticles. Figure adapted with permissions from [66], Copyright 2011 American Chemical Society. (c). SEM imaging of reduced graphene oxide deposited onto glassy carbon as electrode materials for Pb²⁺-sensing devices. Figure reused with permissions from [68], Copyright 2015 Elsevier Ltd. (d). Representation of ionically imprinted polymer, and its application in sensing of lead ions in water. Figure reused with permissions from [67], Copyright 2024 Elsevier B.V.

Pb²⁺-selective electrodes based on the traditional configuration, with ionophores embedded inside a plasticized PVC film, represent one of the most widespread methods for quantification in food and beverage samples. However, most of the ionophores that have been reported lack the necessary selectivity to allow for an accurate assessment of food quality. This has led to most reported manuscripts being focused on liquid samples with low concentration of interferences such as drinking water. A high selectivity is required given the low quantities at which Pb²⁺ can be present compared to other common electrolytes. For example, it has been estimated that the average concentration of Na⁺ in meat and fish products is close to 200 mM kg⁻¹ of product [69], while the maximum allowed concentration of Pb²⁺ is ~1.5 mM kg⁻¹, over two orders of magnitude higher. As such, selectivity coefficients lower than $log{K}_{{Pb}^{2+}, X}^{Pot}<-2$ are preferable.

The use of ionically imprinted polymers has raised interest within the past few years, given the ability to develop specific materials towards virtually any analyte [70]. Using this method, selective Pb²⁺ with tailored properties for detection in food has been designed. Silica-based nanomaterials show promise in this field, given the versatility of tetraethyl orthosilicate precursors, which can be used to synthesise nanoparticles with selected porosity. This material has been used in colorimetric and electrochemical configurations. Lou et al. [71] developed a lateral flow test for Pb²⁺ detection in tea leaves using silica nanoparticles functionalised with aminopropyl triethoxysilane. Amine groups combined with the carboxyl group from a poly acrylic acid solution could selectively bind Pb²⁺, showing a limit of detection of 7 µM [71]. The interactions between silica particles and Pb²⁺ can be further improved by the use of (2-aminophenyl)-1Hbenzimidazole and 4-vinyl pyridine [72]. By using a voltametric approach, the limit of detection of imprinted sensors could be reduced to 0.2 pM. However, ionically imprinted devices still show poor selectivity, especially towards those ions with similar charges and sizes [73]. Moreover, the synthesis conditions can greatly impact the performance of the final materials, which could limit the reproducibility of the electrodes. Thus, new low-cost ionophores for the direct and fast recognition of Pb²⁺ are needed. A summary of ion-selective electrodes applied to lead analysis in food is shown in

Table 2. Summary of ion-selective electrodes developed for lead detection in foodstuff and beverages, and performance. Performance parameters, including sensitivity, limit of detection (LOD), and percentage recovery values (%), are specified.

Material	Ion	Sensitivity (mV/Log[Pb2+])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
Lead ionophore IV, /Ag@PANI solid contact	Pb2+	29.1	0.6 nM	3.0–9.0	97–109	$log{K}_{{Pb}^{2+},{Cu}^{2+}}^{SSM}=-4.9$ $log{K}_{{Pb}^{2+},{Cd}^{2+}}^{SSM}=-6.2$ $log{K}_{{Pb}^{2+},H^{+}}^{SSM}=-3.9$ $log{K}_{{Pb}^{2+},{Zn}^{+}}^{SSM}=-7.8$	Drinking water	[61]
Synthesis of 2-(2-formylphenoxy)acetic acid	Pb2+	27.7	2.9 µM	2.0–12.0	97–98	$log{K}_{{Pb}^{2+},{Li}^{+}}^{SSM}=-1.1$ $log{K}_{{Pb}^{2+},{Ni}^{2+}}^{SSM}=-1.2$ $log{K}_{{Pb}^{2+},{Na}^{+}}^{SSM}=-1.3$ $log{K}_{{Pb}^{2+},{Cd}^{2+}}^{SSM}=-1.3$	Drinking water	[62]
N,N′-tetrabutyldipicolinamide	Pb2+, Hg2+, Cd2+	25.6	70 nM	-	-	$log{K}_{{Pb}^{2+},{Na}^{+}}^{SSM}=-1.7$ $log{K}_{{Pb}^{2+},K^{+}}^{SSM}=-1.7$ $log{K}_{{Pb}^{2+},{Fe}^{2+}}^{SSM}=-1.7$ $log{K}_{{Pb}^{2+},{Sn}^{2+}}^{SSM}=-1.2$	Fish	[63]
(E)-2-(1-(4-(3-(4-chlorophenyl)ureido)phenyl)ethylidene)hydrazinecarbothioamide	Pb2+	28.0	1.7 µM	5.0–11.0	94–98.5	$log{K}_{{Pb}^{2+},{Cd}^{2+}}^{MSM}=-1.7$ $log{K}_{{Pb}^{2+},{Cu}^{2+}}^{MSM}=-$1.2 $log{K}_{{Pb}^{2+},{Ca}^{2+}}^{MSM}=-$2.0 $log{K}_{{Pb}^{2+},{Na}^{+}}^{SSM}=-1.8$	Drinking water	[64]
Polyaniline-zirconium (IV) iodate	Pb2+	29.4	3.3 nM	2.5–6.5	90–97	$log{K}_{{Pb}^{2+},K^{+}}^{MSM}=-0.9$ $log{K}_{{Pb}^{2+},{Na}^{+}}^{MSM}=-0.8$ $log{K}_{{Pb}^{2+},{Mg}^{2+}}^{MSM}=-0.8$ $log{K}_{{Pb}^{2+},{Mn}^{2+}}^{MSM}=-1.5$	Drinking water	[65]
Poly(aniline-co-2-hydroxy-5-sulfonic aniline)	Pb2+	29.3	16 pM	3.5–7.0	-	$log{K}_{{Pb}^{2+},H^{+}}^{FIM}=-3.2$ $log{K}_{{Pb}^{2+},{Na}^{+}}^{FIM}=-3.1$ $log{K}_{{Pb}^{2+},K^{+}}^{FIM}=-3$ $log{K}_{{Pb}^{2+},{Li}^{+}}^{FIM}=-3.2$	Drinking water (pH = 7.3), green gram (pH = 4.1)	[66]
2,2′:6′,6″- terpyridine (terpy)-based Pb2+ -imprinted polymer	Pb2+	28.6	0.3 nM	4.5–8.0	98–102	$log{K}_{{Pb}^{2+},{Co}^{2+}}^{SSM}=-2.8$ $log{K}_{{Pb}^{2+},{Zn}^{2+}}^{SSM}=-2$ $log{K}_{{Pb}^{2+},{Ag}^{+}}^{SSM}=-2.1$ $log{K}_{{Pb}^{2+},{Hg}^{2+}}^{SSM}=-2.1$	Drinking water	[67]
1,2-Bis(N’- Benzoylthioureido)benzene/reduced graphene oxide	Pb2+	30.4	25 nM	4.0–8.0	84–101	$log{K}_{{Pb}^{2+},{Fe}^{3+}}^{SSM}=-1.5$ $log{K}_{{Pb}^{2+},{Na}^{+}}^{SSM}=-1.8$ $log{K}_{{Pb}^{2+},{NH4}^{+}}^{SSM}=-1.8$ $log{K}_{{Pb}^{2+},{Cs}^{+}}^{SSM}=-1.8$	Milk	[68]

.

3.2. Arsenic Detection

Arsenic is classified as a metalloid and belongs to Group 15 in the periodic system. Until recently, this element had been used as a pesticide and a fungicidal agent [72]. However, the increasing evidence of its high carcinogenic potential led to the ban of this component in the EU and USA in 2004 [74]. There is increasing evidence of the harmful effects of arsenic. This ion has been linked to lung, skin, and bladder cancer, as well as skin lesions in doses between 0.3 and 8 µg/kg body weight. However, it has been estimated that inorganic arsenic consumption in certain populations, such as toddlers and infants, could be between 0.3 and 1.2 µg/kg body weight, which is within the maximum recommended [75]. Among all the potential sources of arsenic, rice has been described as the largest contributor to exposure in humans. Rice can bioaccumulate arsenic, due to the use of phosphate fertilisers and soil amendments [76]. This trend represents a worldwide concern, since rice is the most important food source, feeding over 50% of the population, and accounting for 21% of the global caloric intake [77].

One of the limitations of arsenic detection in solution is the lack of suitable ionophores that can readily be used. While arsenic is a cation, with multiple potential valences (mostly As³⁺ and As⁵⁺ in solution), it is commonly found in the form of oxyanions, bound to three or four oxygen atoms. Depending on the pH, arsenic oxyanions can also be bound to multiple hydrogen atoms, with a subsequent change in the binding dynamics towards ionophores. Moreover, the presence of redox components [78] or other inorganic elements such as iron [79] can change the oxidation state, leading to a wide variety of arsenic ion forms in solution. Different forms of arsenic in solution can lead to variable interactions with electrode surfaces. Moreover, As³⁺ can oxidise to As⁵⁺ in the presence of catalytic materials. Hasnat et al. [80] characterised the kinetics of arsenite oxidation on a platinum surface. Different oxidation kinetics were observed when the media was acidic, compared to higher pH values. This mechanism suggests that, for accurate arsenic quantification in food, samples must be prepared to reach a standardised pH value.

The pH dependence of arsenic oxidation states makes the development of effective ionophores challenging, since additional information from the sample (i.e., pH) might be required to conduct a total assessment. To date, no arsenic ionophores are available commercially. Initial tests for the determination of arsenate in solution involved the precipitation of arsenate using lanthanum nitrate, and the excess of lanthanum was then titrated using fluoride and an ion-selective electrode [81]. However, this method could not be easily applied to fast monitoring in real samples, given the time-consuming elaborations required.

Polymeric and inorganic materials have been used for the voltametric detection of arsenic. Khan et al. incorporated silica gel inside polyurethane fibres [82], leading to a Nernstian value across a wide concentration range (10⁻⁸–1 M). However, these fibres showed poor selectivity, especially towards other common oxyanions such as NO₃⁻ and PO₄³⁻, with coefficients $log{K}_{{AsO}_4^{3-},x}^{Pot}>-1$. Poly(diallyl dimethylammonium chloride) reinforced with graphene oxide nanosheets and functionalized with acrylic acid has also been implemented in arsenic detection [83]. The device showed a detection limit within the micromolar range (30 µM). However, selectivity coefficients were not provided in this case.

Given the low availability of ionophores, their use in food and beverage detection has been very limited. So far, sulphide salts have been employed in this field, showing a high selectivity towards arsenite when compared to common ions such as Na⁺ and other heavy metals including Cu²⁺. This electrode could be used in the analysis of beer [84]. A concentration of 26.5 µg L⁻¹ was determined, below the legal limit of 100 µg L⁻¹. Other ionophores employed in this field include chromone-based compounds, which combine a good selectivity towards anions in solution [85] with a low detection limit of 5 × 10⁻⁷ M. However, the synthesis of ionophores represents a challenging step, since the presence of impurities or different molecule conformations could impact the sensitivity and selectivity of final devices. Recently, alternative approaches using molecularly imprinted polymers have also been developed and applied to arsenate detection in water [86]. However, in this case, only the selectivity towards cations was determined, while the selectivity towards potential oxyanion interferences such as SO₄²⁻ was not tested.

As mentioned, other oxyanions such as nitrate and phosphate could represent significant interferences, given the oxidised state of arsenic ions in solutions. As such, more research is needed to determine the potential of ionically imprinted materials in food and beverage analysis. A summary of ion-selective electrodes applied to food analysis is provided in

Table 3. Summary of ion-selective electrodes developed for arsenic detection in food and beverages and performance.

Material	Ion	Sensitivity (mV/Log[As])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
Ag3AsO4	AsO43−	19.0	1 µM	6.0–10.0	-	$log{K}_{S^{2-},\mathrm{A}\mathrm{s}{O}_4^{3-}}^{MSM}=-1.7$ $log{K}_{{SO}_3^{2-},\mathrm{A}\mathrm{s}{O}_4^{3-}}^{MSM}=-4$ $log{K}_{{SO}_4^{2-},\mathrm{A}\mathrm{s}{O}_4^{3-}}^{MSM}=-4$ $log{K}_{{Ag}^{+},\mathrm{A}\mathrm{s}{O}_4^{3-}}^{MSM}=-2$	Beer (pH = 10.2 in buffer)	[84]
4H-1-benzopyran-3-carboxaldehyde, 4-oxo-, 3-[2-(2,4-dinitrophenyl)hydrazone]	AsO2−	61.6	80 nM	3.0–9.0	96–102	$log{K}_{C_6{H}_5{O}_7^{3-},{AsO}_2^{-}}^{FIM}=-2.3$ $log{K}_{{SO}_3^{2-},{AsO}_2^{-}}^{FIM}=-2.4$ $log{K}_{{Br}^{-},{AsO}_2^{-}}^{FIM}=-2.5$ $log{K}_{N_3^{-},{AsO}_2^{-}}^{FIM}=-2.2$	Drinking water (pH = 5.5 in buffer)	[85]
Methacrylic acid	As3+	20.5	0.5 µM	4.0–8.0	96–107	$log{K}_{{Ag}^{+},{As}^{3+}}^{MPM}=-2.1$ $log{K}_{{Cd}^{2+},{As}^{3+}}^{MPM}=-2.1$ $log{K}_{{Zn}^{2+},{As}^{3+}}^{MPM}=-2.1$ $log{K}_{{Mn}^{2+},{As}^{3+}}^{MPM}=-2.0$	Drinking water (pH = 6.5 in buffer)	[86]
Clinoptilolite	HAsO42−	−28.3	30 nM	7.0–11.0	101–105	$log{K}_{{NO}_2^{-},{HAsO}_4^{2-}}^{FIM}=-1.2$ $log{K}_{{IO}_3^{-},{HAsO}_4^{2-}}^{FIM}=-1.4$ $log{K}_{{SO}_4^{2-},{HAsO}_4^{2-}}^{FIM}=-1.6$ $log{K}_{S_2{O}_3^{2-},{HAsO}_4^{2-}}^{FIM}=-1.7$	Drinking water	[87]

, including information about the sensitivity, pH detection range, limit of detection, and selectivity coefficients.

3.3. Mercury Detection

Mercury exposure represents a healthcare concern, given its presence in the environment. Among all the forms of mercury, organic forms exhibit higher toxicity and are the most common [88]. The reaction of mercury to form methylmercury is primarily carried out by bacteria in aquatic environments [88]. In humans, the largest contributor to dietary mercury exposure is fish [89]. The bioaccumulation of methylmercury in the food chain has been further enhanced due to climate change due to changes in bacterial and plankton communities and ocean acidification [90,91].

Methylmercury is known for its high toxicity, being responsible for large outbreaks of poisoning in Iraq [92] and Japan [93], among others. Symptoms from mercury exposure can take weeks to months to appear [94], and clinical diagnosis is often challenging [95]. Mercury exposure is typically diagnosed through indirect tests, such as blood cell count, electrocardiography, electrolyte assays, and renal and hepatic function tests [95]. The main methods employed nowadays for mercury quantification in food are inductively coupled plasma mass spectrometry and atomic absorption spectrometry [96]. These methods can provide accurate measurements, with a limit of quantification below 1 µg kg⁻¹ [96]. However, they require strict preparation procedures which, in most cases, involve the use of potentially dangerous chemicals, limiting their use by non-specialists.

The main limitation of mercury detection in biological samples is the presence of its organic form. Organic mercury is naturally synthesised by bacteria. Although methylmercury is charged positively, there is a lack of reported ionophores that can be used for the potentiometric detection of this compound. Recently, a fluorescent approach based on tryptophan and a six-membered cyclic amine as an electron-donating group have been used as a fluorescent sensor for methylmercury [97]. However, portable and low-cost methods that do not require expensive detection systems are desirable. Within the past few years, nanomaterials, including sulphur-doped graphitic carbon nitride [98] and ion-imprinted polymers combined with multi-walled carbon nanotubes [99], have been developed for the specific detection of methylmercury through voltammetry with good selectivity. However, the potentiometric detection of this ion through potentiometric ion-selective electrodes has not been achieved yet.

Ion-selective electrodes that can detect inorganic Hg²⁺ have been widely reported in the literature using neutral ionophores (

Table 4. Summary of reported approaches in mercury quantification using ion-selective electrodes. Performance metrics include limit of detection, sensitivity, pH working range, and selectivity.

Material	Ion	Sensitivity (mV Log[Hg2+])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
1-methyl-2-butylthioimidazolium bis(trifluoromethane sulphonyl)imide	Hg2+	6.7	41 pM	2.0–3.0	-	$log{K}_{{Ca}^{2+},{Hg}^{2+}}^{FIM}=-3.8$ $log{K}_{{Mg}^{2+},{Hg}^{2+}}^{FIM}=-4.4$ $log{K}_{{Na}^{+},{Hg}^{2+}}^{FIM}=-4.7$ $log{K}_{{Cu}^{2+},{Hg}^{2+}}^{FIM}=-4.7$	Seafood (pH = 3 using NaOH)	[102]
1,3-bis [2-(N-morpholino)acetamidothiophenoxy]propane	Hg2+	30.0	5 nM	2.2–4.5	96–104	$log{K}_{K^{+},{Hg}^{2+}}^{MPM}=-5.3$ $log{K}_{{Na}^{+},{Hg}^{2+}}^{MPM}=-5.3$ $log{K}_{{Cd}^{2+},{Hg}^{2+}}^{MPM}=-3.1$ $log{K}_{{Pb}^{2+},{Hg}^{2+}}^{MPM}=-3.2$	Drinking water	[103]
Zirconium antimonate	Hg2+	30.0	50 nM	2.5–8.5	98–99	$log{K}_{K^{+},{Hg}^{2+}}^{FIM}=-2.0$ $log{K}_{K^{+},{Hg}^{2+}}^{FIM}=-1.0$ $log{K}_{{Ca}^{2+},{Hg}^{2+}}^{FIM}=-2.1$ $log{K}_{{Co}^{2+},{Hg}^{2+}}^{FIM}=-2.2$	Tap water, fish	[104]
4-(2- Thiazolylazo) resorcinol	Hg2+	29.7	2 nM	3.0–4.5	98–101	$log{K}_{K^{+},{Hg}^{2+}}^{MPM}=-5.3$ $log{K}_{{Na}^{+},{Hg}^{2+}}^{MPM}=-5.3$ $log{K}_{{Cd}^{2+},{Hg}^{2+}}^{MPM}=-3.1$ $log{K}_{{Pb}^{2+},{Hg}^{2+}}^{MPM}=-3.2$	Fish, shrimp (pH adjusted to 3)	[105]
Multi-walled carbon nanotube (MWCNT)-grafted 2, 6-bis [2-(amino methyl)phenol]pyridine	Hg2+	29.8	0.8 nM	3.0–4.5	88–112	$log{K}_{K^{+},{Hg}^{2+}}^{MPM}=-5.0$ $log{K}_{{Na}^{+},{Hg}^{2+}}^{MPM}=-4.7$ $log{K}_{{Fe}^{3+},{Hg}^{2+}}^{MPM}=-4.1$ $log{K}_{{Pb}^{2+},{Hg}^{2+}}^{MPM}=-3.0$	Drinking water (pH = 4.2)	[106]

). The first neutral ionophore for Hg²⁺ was reported by Lai et al. [100] in 1986 using crown-based ionophores. This electrode could detect the presence of Hg²⁺ ions down to the micromolar range. Since then, multiple ionophores with varied structures have been developed. Mahajan et al. [101] designed an imidazole-based ionophore embedded inside a plasticized PVC matrix for the detection of mercury, with a detection limit within the sub-micromolar level (~8 µM). Similar imidazole-based ionophores have been employed in the detection of mercury in food combined with carbon nanomaterials. Miao et al. [102] employed an ionic liquid-based matrix for the characterisation of seafood. This device combined an ultra-low detection limit, below the nanomolar level (41 pM), and good selectivity when compared to other common electrolytes, including Na⁺, Ca²⁺, and K⁺. However, the device showed a sub-Nernst sensitivity of 6.7 mV Log[Hg]²⁺, and experienced a further decrease in sensitivity after only 1 month of use. Moreover, the working pH range of this sensor was narrow, between 2 and 3. Carbon paste electrodes, based on conductive graphite and modified with Hg²⁺ ionophores, have also been developed [103]. These devices also showed a low limit of detection (5 nM), and good selectivity coefficients when compared to similar heavy metals (Cd²⁺, Pb²⁺), and common electrolytes (Na⁺, K⁺). However, these devices could only be operated within a pH range of 2.2–4.5.

Long-term stability challenges and resilience to pH changes of mercury-selective electrodes can be circumnavigated by the use of inorganic sensing materials such as zirconium antimoniate. Aglan et al. [104] reported a screen-printed electrode that produced a Nernstian response over a wide concentration range (10⁻¹–5 × 10⁻⁸ M). The final sensor could be used in the detection of Hg²⁺ in seafood with high percentage recovery values, and the working range for pH increased up to 2.5–8.5. However, the device showed a lower selectivity, especially towards common electrolytes, including K⁺ ($log{K}_{{Hg}^{2+},K^{+}}^{Pot}~-1$), and other metal cations ($log{K}_{{Hg}^{2+},{Fe}^{3+}}^{Pot}~-2$).

Ionically imprinted polymers have also been used in this field. 4-(2-Thiazolylazo) resorcinol ligands imprinted onto graphene nanosheets and alumina nanoparticles have been used to detect Hg²⁺ ions from fish and shrimp samples [105]. The devices showed a low detection limit of 2 nM, and high selectivity when compared to common electrolytes and other common heavy metals. In this case, samples were pre-digested using HNO₃ and H₂O₂, followed by a heating treatment at 180^o. However, in this case, the pH working range was narrow (3.5–4), being smaller than the one found in neutral ionophore-based devices. This limitation could be a consequence of the reaction of 4-(2-Thiazolylazo) resorcinol molecules with H⁺. These result suggest that, although ion imprinting represents a promising technology in food analysis, with high selectivity and stability, the ligands chosen must be stable for the application in this field.

3.4. Cadmium Detection

Cadmium is a heavy metal with proven toxicity in humans. The primary source of cadmium exposure in humans is the diet, through food and drinking water. In particular, vegetable-based foods are a major contributor to cadmium intake [107]. As such, certain population subgroups such as vegetarians and people living within contaminated areas are at a high risk of toxicity [108]. Cd²⁺ overexposure has been associated with multiple conditions including kidney disease [109], bone damage [110], and cancer [111]. According to EFSA, the tolerable weekly intake of cadmium is 2.5 µg/kg body weight [112]. The average intake in European populations is close to this limit (2.04 µg/kg body weight, and a potential 95th percentile of 3.66 µg/kg body weight) [113].

Dissolved cadmium is commonly found as a divalent cation. Cd²⁺ is typically assessed in food samples using ICP-MS. Ion-selective electrodes have been reported for the quantification of cadmium in food, using ionophores embedded inside a plasticized PVC film. Bioactive compounds such as salophen have been used for the detection of Cd²⁺ in drinking water [114]. While the selectivity achieved towards other heavy metal ions was relatively low, being $log{K}_{M^{2+},{Cd}^{2+}}^{MPM}>$−2 for Hg²⁺ and Cu²⁺, interference from common electrolytes including Na⁺, Ca^2+, and Mg²⁺ was small. Similar results have been obtained by other neutral ionophores such as N,N-(4-methyl-1,2-phenyle-ne)diquinoline-2-carboxamide [115], Benzyl bis(carbohydrazide) [116], and 1, 2-bis(quinoline-2-Carboxamido)-4-chlorobenzene [117]. However, in all cases, a high pH working range was obtained.

A key limitation of current ion-selective electrodes is the need for low-cost and flexible materials for electronic component manufacturing. Jiang et al. [118] incorporated a writing ink using graphene nanosheets as electrodes to improve the mechanical resilience of devices. Electrodes were deposited using a brush pen-based method, and a commercially available cadmium ionophore (N, N,N′, N′-Tetrabutyl-3,6-dioxaoctanedi(thioamide)) was incorporated inside the silver ink mix. In this case, the selectivity towards heavy metals improved. However, the limit of detection increased by two orders of magnitude, being in the range of 1 µM, while salophen-based sensors achieved a limit of detection of 16 nM. Other ionophores such as (E)-2-benzylidenehydrazinecarbothioamide have also been incorporated in water analysis using an all-solid-state configuration [119]. However, the performance of these devices did not match the values reported by commercially available ionophores in terms of the selectivity and limit of detection (

Table 5. Summary of reported approaches in cadmium quantification using ion-selective electrodes. Performance metrics include limit of detection, sensitivity, pH working range, and selectivity.

Material	Ion	Sensitivity (mV Log[Cd2+])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
Salophen	Cd2+	29.2	16 nM	2.5–7.5	99–102	$log{K}_{{Hg}^{2+},{Cd}^{2+}}^{MPM}=-$1.9 $log{K}_{{Cu}^{2+},{Cd}^{2+}}^{MPM}=-1.8$ $log{K}_{{Zn}^{2+},{Cd}^{2+}}^{MPM}=-2.6$ $log{K}_{{Pb}^{2+},{Cd}^{2+}}^{MPM}=-2.6$	Drinking water	[114]
N,N-(4-methyl-1,2-phenylene)diquinoline-2-carboxamide	Cd2+		0.8 µM	4.0–9.0	97–101	$log{K}_{{Fe}^{3+},{Cd}^{2+}}^{FIM}=-1.2$ $log{K}_{{Ag}^{+},{Cd}^{2+}}^{FIM}=-1.0$ $log{K}_{{Cu}^{2+},{Cd}^{2+}}^{FIM}=-0.7$ $log{K}_{{Pb}^{2+},{Cd}^{2+}}^{FIM}=-1.6$	Drinking water	[115]
Benzyl bis(carbohydrazone)	Cd2+	29.7	32 nM	2.0–9.0	-	$log{K}_{{Cr}^{2+},{Cd}^{2+}}^{SSM}=-$2.1 $log{K}_{{N{H}_4}^{+},{Cd}^{2+}}^{SSM}=-2.0$ $log{K}_{K^{+},{Cd}^{2+}}^{SSM}=-3.0$ $log{K}_{{Cu}^{2+},{Cd}^{2+}}^{SSM}=-3.0$	Chocolate	[116]
1, 2-bis(quinoline-2-Carboxamido)-4-chlorobenzene	Cd2+	30.3	0.8 µM	2.4–9.0	98–103	$log{K}_{{Cu}^{2+},{Cd}^{2+}}^{SSM}=-$2 $log{K}_{{Hg}^{2+},{Cd}^{2+}}^{SSM}=-2$ $log{K}_{{Zn}^{2+},{Cd}^{2+}}^{SSM}=-2.4$ $log{K}_{{Ag}^{+},{Cd}^{2+}}^{SSM}=-3$	Drinking water	[117]
N,N,N′,N′-Tetrabutyl-3,6-dioxaoctanedi(thioamide)	Cd2+	28.6	1 µM	-	94–100	$log{K}_{{Mg}^{2+},{Cd}^{2+}}^{SSM}=-$2.5 $log{K}_{{Zn}^{2+},{Cd}^{2+}}^{SSM}=-$3.7 $log{K}_{{Ni}^{2+},{Cd}^{2+}}^{SSM}=-$4.1 $log{K}_{{Ca}^{2+},{Cd}^{2+}}^{SSM}=-$4.1	Drinking water	[118]
(E)-2-benzylidenehydrazinecarbothioamide	Cd2+	22.9	1.8 µM	5.0–10.0	95–97	$log{K}_{K^{+},{Cd}^{2+}}^{SSM}=-$2.5 $log{K}_{{Zn}^{2+},{Cd}^{2+}}^{SSM}=-$2.6 $log{K}_{{Sr}^{2+},{Cd}^{2+}}^{SSM}=-2.7$ $log{K}_{{Co}^{2+},{Cd}^{2+}}^{SSM}=-2.9$	Drinking water	[119]
Methyl methacrylate	Cd2+	29.9	100 nM	4.0–7.0	92	$log{K}_{{Cu}^{2+},{Cd}^{2+}}^{MPM}=-$2 $log{K}_{{Ni}^{2+},{Cd}^{2+}}^{MPM}=-2$ $log{K}_{{Zn}^{2+},{Cd}^{2+}}^{MPM}=-2.4$ $log{K}_{{Ca}^{2+},{Cd}^{2+}}^{MPM}=-3$	Drinking water	[120]

).

Ionic imprinting materials have also been employed in this field, using methyl methacrylate monomers bound with ethyleneglycoldimethacrylate crosslinkers [120]. This material was embedded inside a plasticized PVC film and achieved a detection limit of 1.6 µM. However, this device also showed a high pH sensitivity, only being usable in the pH range between 4 and 7.

4. Anion Detection

The development of anionic ionophores represents a key challenge in supramolecular chemistry [121], which is reflected in a lack of ionophores towards anions. Unlike cations, anions can show an increased variety of geometries, and have a low charge-to-radii ratio and low solvation energy. Some anions are also pH-sensitive [122,123], restricting the design of new ionophores. In the case of hydrophilic anions, such as phosphates and fluoride, this task becomes increasingly difficult, due to the low solvation energy, which restricts the diffusion of ions from the solvent towards membrane electrodes. Although certain insoluble metals such as Ag and their oxidised forms have traditionally been used for the detection of halides, they show poor selectivity. Despite these limitations, the use of ion-selective electrodes in this area still represents one of the most widespread methods for the quantification of anions in real-world settings. Some electrodes, such as LaF₃ for F⁻ are well established and have been used in food characterisation for decades. This is a consequence of the complexity of other methods, such as ion chromatography, which, in some cases, leads to a similar performance to the one achieved by ion-selective electrodes [124] while requiring lengthy preparation steps.

4.1. Fluoride Ion Sensors

Fluoride represents one of the most common contaminations in food (Figure 3a). In humans, the main source of exposure to F⁻ is toothpaste and fluorinated water [125]. Although fluoride ions are not essential for correct physiological functioning, it has been reported to have beneficial effects on oral health, reducing the incidence of caries [126] when the daily intake is around 0.05 mg/kg body weight [127]. However, higher intakes can also lead to dental fluorosis, bone damage, and fatigue [128]. In children, the consumption of fluoride above 0.1 mg/kg body weight/d increases the risk of fluorosis. Therefore, there is a narrow gap between the beneficial and harmful dose of fluoride, which requires close monitoring. High fluoride concentrations are common in tea leaves [129], and are less frequent in fish [130] and rice [131] due to intake from contaminated soils.

To date, the most common material used in the detection of fluoride is lanthanum fluoride (LaF₃). This compound has been used in commercially available models such as the probe-based sensors (Figure 3b) and can be used for the detection of F⁻ down to 10⁻⁶ M [132]. Organic ionophores have also been developed such as Bis(fluorodioctylstannyl)methane. This molecule is commercially available as “Fluoride ionophore I”, and its use was first reported by Perdikaki et al. [133] in 2002. When incorporated into a PVC matrix with bis(2-ethylhexyl) sebacate as a plasticizer, this ionophore can selectively detect fluoride, discriminating some of the most common ionic interferences found in food samples such as bromide and chloride, with selectivity coefficients below $log{K}_{F,x}^{Pot}<-2$ for both ions. However, these devices lack selectivity towards phosphates and thiocyanates [133,134]. Although they are minor components in tea infusions, phosphates and thiocyanate can be found in tropic foods, including sweet potato, corn, and lima beans [134]. Seafood and legumes among others are also rich in phosphates [135,136], and both phosphate and thiocyanate are often used as additives [137,138]. Consequently, the usability of this device is restricted to tea samples, with low phosphate and thiocyanate concentrations.

More recently, alternative fluoride ionophores have been developed, which can circumnavigate the challenges of poor selectivity towards common additives. In most cases, these ionophores involve a metallic complex, that interacts selectively with the target anion. These ionophores are typically used, as in the case of cation electrodes, embedded inside a plasticized PVC film. Mitchell-Koch et al. [139] explored the use of aluminium ions with porphyrins inside a PVC film, leading to a selectivity of $log{K}_{F,SCN}^{Pot}<-3$. Moreover, Li et al. developed multiple organo-antimony(V) compounds, which could achieve a selectivity up to $log{K}_{F,SCN}^{Pot}<-2.7$. However, in both cases, the selectivity towards phosphate, which is a significant interference in most food samples, was not reported. Schiff base complexes such as N, N’-Ethylene-bis(salicylideneaminato) nickel(II) [140] have also been employed in the detection of fluoride. This device showed a low detection limit of 3.6 µM while maintaining good selectivity, especially towards phosphate and chlorate ions. A comparison between ion-selective electrodes developed for fluoride detection in food is shown in

Table 6. Summary of applications of fluoride-selective electrodes. Performance metrics include limit of detection, sensitivity, pH working range, and selectivity.

Material	Ion	Sensitivity (mV Log[F−])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
N, N’-Ethylene-bis(salicylideneaminato) nickel(II)	F−	−57.2	3.6 µM	2.1–9.1	-	$log{K}_{{{CO}_3}^{-},F^{-}}^{SSM}=-1.3$ $log{K}_{{SCN}^{-},F^{-}}^{SSM}=-1.2$ $log{K}_{{HPO}_4^{-},F^{-}}^{SSM}=-2.1$ $log{K}_{{{BrO}_3}^{-},F^{-}}^{SSM}=-1.1$	Drinking water, tea	[140]
ThermoOrion model 96-09	F−	-	-	-	-	-	Chicken (pH = 5.5 in buffer)	[141]
ThermoOrion model 96-09	F−	−60.0	-	-	81–105	-	Seafood (pH = 7.2 using HCl)	[142]
ThermoOrion model 96-09	F−	−57.8	-	-	99–103	-	Potato chips	[143]
ThermoOrion model 96-09	F−	-	27 µM	-	81–89	-	Cocoa powder (pH = 5.2 in buffer)	[144]
Fluoride ion-selective electrode (FISE) DC219 from Mettler Toledo	F−	−57.5	500 nM		-	-	Pomegranate, Mint, Green tea (pH = 5.3 using buffer)	[145]
Thermo Scientific, model 9609BNWP	F−	−54.7	160 nM	5–7	86–110	-	Soft drinks, juice, tea (pH ~6 using buffer)	[146]
Tetrachloro-substituted organoantimony(V) compound	F−	−59.5	5 µM	3	-	$log{K}_{{CN}^{-},F^{-}}^{SSM}=-1.3$ $log{K}_{{SCN}^{-},F^{-}}^{SSM}=-2.7$ $log{K}_{{ClO}_4^{-},F^{-}}^{SSM}=-2.8$ $log{K}_{I^{-},F^{-}}^{SSM}=-3.1$	Tap water	[147]
Fluoride ion-selective electrode (FISE) DC219 from Mettler Toledo	F−	−57.9	100 µM	-	84–101	-	Fish, seafood, vegetables (pH = 7.0 using NaOH)	[148]
Ion-selective electrode (Model 6.0502.150, Metrohm)	F−	-	40 nmol g−1	-	92	-	Medicinal plants (pH = 5.5 using NaOH)	[149]
pH/ion meter devices (Metrohm types 713 and 744)	F−	-	-	-	-	-	Anchovies (pH = 5 using NaOH)	[150]
Digital Ion-meter (Philips PW 9414)	F−	−55.0	5.6 µM	3.5–8.0	-	-	Cheese	[151]

.

Despite the development of neutral ionophores for F⁻ detection, the use of LaF₃ represents the most common approach for food analysis. Consequently, most reported work in the field has been focused on the use of this material applied to different samples [141,142,143,144]. However, the limit of detection achievable by this material is typically within the micromolar range, while neutral ionophore-based devices can often determine concentrations within the nanomolar range.

One of the limitations of ion-selective electrodes is the need for a liquid media in contact homogeneously with the sensitive membrane. In the case of beverages, including tea or juice, sensors can directly be applied to the fluid media. However, in these cases, additional devices, such as pH sensors, must be incorporated to estimate the fraction of fluoride found in the form of hydrofluoric acid [146], or the sample pH can be buffered to ensure a consistent pH. However, in the case of the determination of fluoride in solid foods, such as meat or fish, ions must first be extracted from the samples. They are then diluted in liquid media. One of the most common methods for ion extraction in food is dry ashing, where water and organic components are subjected to high temperatures, leaving only the ions contained in the sample, and removing organic interferences [152]. However, this process is time-consuming, requiring lengthy treatments to maximise the extraction of ions. Ion extraction can be sped up through the use of a microwave oven and nitric acid, to allow for the fast digestion of food samples. By using this method, Rocha et al. [148] could quantify the amount of fluoride in food down to 0.13 mg/kg food. This method was tested in seafood products, vegetables, legumes, cereals, fruits, and tubers (Figure 3c).

Figure 3. (a). Summary of healthcare benefits and potential toxicity of fluoride. Figure reused from [153], distributed under Creative Commons Attribution (CC BY 4.0). (b). Picture of commonly used probe-based sensors, commercially available for fluoride detection in liquid samples [154], copyright 2023, distributed under Creative Commons Attribution (CC BY). (c). Schematic representation of microwave-assisted digestion of food samples for fluoride detection. Figure reused with permissions from [148], Copyright 2013 American Chemical Society.

Figure 3. (a). Summary of healthcare benefits and potential toxicity of fluoride. Figure reused from [153], distributed under Creative Commons Attribution (CC BY 4.0). (b). Picture of commonly used probe-based sensors, commercially available for fluoride detection in liquid samples [154], copyright 2023, distributed under Creative Commons Attribution (CC BY). (c). Schematic representation of microwave-assisted digestion of food samples for fluoride detection. Figure reused with permissions from [148], Copyright 2013 American Chemical Society.

4.2. Phosphate

Inorganic phosphate is a common additive used by the food industry to preserve food and improve palatability and texture [137]. Phosphates are naturally present in the form of organic esters within a wide range of foods, including potatoes, bread, and meat [155]. However, only between 40 and 60% of organically bound phosphate is absorbed in the gastrointestinal tract [156]. On the contrary, when used in the inorganic form as a food additive, most of the phosphate is absorbed. Within the past few years, the potentially harmful effects of high phosphate have been reported, including cardiovascular risks [157], renal damage [158], and even cancer [159]. Thus, quantifying phosphate in food is becoming a key priority, especially for patients undergoing certain conditions such as chronic kidney disease [160].

Although there is a demand for electrodes to determine phosphate levels in food, the research in this area has been limited so far. The standard method for phosphate detection is the molybdenum blue formation coupled with spectrophotometric detection [161]. Using this method, phosphates react with molybdenum oxide to generate a blue-coloured solution with an absorbance peak of around 800 nm. However, the spectrophotometric detection of phosphate lacks selectivity, being sensitive to different types of ions, including orthophosphate, arsenate, or silicates, among others that can also be present in the food sample [162,163].

The development of ionophores has been challenging due to the relatively large size of phosphate molecules, requiring complex host molecules, with pre-arranged functional groups to allow for selective and reversible binding [164]. Consequently, phosphate ion-selective electrodes are often made by using a combination of solid inorganic compounds. Kalayci et al. [165] developed a phosphate-specific sensor using Ba₃PO₄, Cu₂S, and Ag₂S as sensitive materials. This device showed selectivity below $log{K}_{{PO}_4^{3-},X}^{Pot}<-3$ when tested against halides such as Cl⁻ and other anions such as NO₃⁻ or SO₄²⁻. Using this method, authors reported a concentration of 423 mg/100 g phosphate in beans, being similar to the results obtained by UV spectrophotometry. Surface-modified tungsten electrodes have also been employed [166]. While this device could achieve good selectivity, with no significant interferences from common electrolytes (i.e., Cl⁻, HCO₃⁻), the detection limit showed a pH dependency. Copper nanoparticles electrodeposited onto a copper wire have also been applied for the characterisation of drinking water samples. In this case, the fabrication costs were considerably lower compared to previous inorganic approaches, given the simplicity of the device’s composition [167]. Moreover, the sensors showed a good selectivity when compared to most oxyanions. However, it showed significant interferences from OH⁻, which suggested a high sensitivity to pH changes.

Within recent years, organic ionophores that can provide higher selectivity towards phosphates have been designed. In most cases, ionophores combine -CH or -NH groups that can form hydrogen bonds with this anion, with a size that optimises these interactions. These ionophores are embedded inside a plasticized PVC film. To date, compounds such as nicotianamide have been incorporated into PVC films with 2-nitrophenyl octyl ether plasticizer for the detection of phosphate [168]. Other molecules that can form such hydrogen bonds include thiourea [169] and Acyl-Hydrazine functionalized Calix [4]arenes [170]. However, the incorporation of these sensors in food analysis is still in its infancy, and additional work is required to develop sensing devices that can meet the current standards for practical applications. A summary of recently reported approaches is provided in

Table 7. Summary of applications of phosphate-selective electrodes. Performance metrics include limit of detection, sensitivity, pH working range, and selectivity.

Material	Ion	Sensitivity (mV Log[HPO4−])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
Ba3PO4, Cu2S and Ag2S	HPO4−	−57.0	1 µM	7.0–9.0	-	$log{K}_{{Cl}^{-},{HP{O}_4}^{-}}^{SSM}=-3.7$ $log{K}_{S^{2-},{HP{O}_4}^{-}}^{SSM}=-3.9$ $log{K}_{{NO}_3^{-},{HP{O}_4}^{-}}^{SSM}=-3.5$ $log{K}_{{{SO}_4}^{-},{HP{O}_4}^{-}}^{SSM}=-4.2$	Beef, beans, garlic, dried apricots	[165]
Surface-modified tungsten	HPO42−	−36.9	1 µM	7.0–10.0	95–98	$log{K}_{{Cl}^{-},{HP{O}_4}^{-}}^{SSM}=-4.2$ $log{K}_{{{HCO}_4}^{2-},{HP{O}_4}^{-}}^{SSM}=-4.5$ $log{K}_{{NO}_3^{-},{HP{O}_4}^{-}}^{SSM}=-2.8$ $log{K}_{{{SO}_4}^{-},{HP{O}_4}^{-}}^{SSM}=-4.5$	Juice, Coca-Cola, milk, tap water	[166]
Copper nanoparticles	HPO42−	−27.8	1 µM	6.0–11.0	97–102	$log{K}_{{Cl}^{-},{HP{O}_4}^{-}}^{SSM}=-2.0$ $log{K}_{{{HCO}_4}^{2-},{HP{O}_4}^{-}}^{SSM}=-3.3$ $log{K}_{{NO}_3^{-},{HP{O}_4}^{-}}^{SSM}=-4.8$ $log{K}_{{OH}^{-},{HP{O}_4}^{-}}^{SSM}=-0.9$	Drinking water (pH = 7.6)	[167]
Nicotinamide	HPO4−	−53.3	0.9 µM	-	-	$log{K}_{{Cl}^{-},{HP{O}_4}^{-}}^{SSM}=-1.3$ $log{K}_{{C{H}_{3}COO}^{-},{HP{O}_4}^{-}}^{SSM}=-16.3$ $log{K}_{{NO}_3^{-},{HP{O}_4}^{-}}^{SSM}=-0.1$ $log{K}_{{S{O}_4}^{-},{HP{O}_4}^{-}}^{SSM}=-0.9$	Water	[168]
Acyl-Hydrazine Functionalized Calix [4]arene	HPO42−	−29.3	10 µM	6.5–9.5	-	$log{K}_{{Cl}^{-},{HP{O}_4}^{-}}^{SSM}=-2.7$ $log{K}_{{SCN}^{-},{HP{O}_4}^{-}}^{SSM}=+1.1$ $log{K}_{{NO}_3^{-},{HP{O}_4}^{-}}^{SSM}=-1.0$ $log{K}_{{Cl{O}_4}^{-},{HP{O}_4}^{-}}^{SSM}=+2.4$	Water	[170]

.

4.3. Nitrite and Nitrate

Nitrites and nitrates are similar components, often found together in food. Nitrate is an oxyanion molecule with a trigonal planar structure, and π-bonds between the central nitrogen and oxygen atoms. Dietary nitrate is converted to nitrite by oral bacteria, or in the stomach [171]. Nitrites represent one of the most common anions found in food, being found in fruits and a wide range of vegetables [172]. However, the concentration of nitrates in vegetables is highly variable, and its presence is influenced by environmental factors such as the soil composition, weather, growth period [173], or the use of fertilisers [174,175]. This molecule is also used as a curing agent for meat, providing its characteristic pink colour and savoury flavour and inhibiting the oxidation and growth of harmful bacteria [176,177].

Nitrate-based diets have been used to treat patients with chronic kidney disease due to their cardiovascular benefits [178,179]. However, overexposure to nitrates and nitrites can lead to an increased risk of cancer and methemoglobinemia [180]. A positive association has been found between high nitrite intakes and gastrointestinal [174] and breast cancer [181]. Moreover, it has been reported to increase the risk of type 2 diabetes [182]. As such, there is an unmet need for devices that can be used for the monitoring of nitrates and nitrites in food. The European Food Safety Authority recommends an acceptable daily intake of nitrites of 0.07 mg/kg bw/day [183]. However, meat products can have concentrations as high as 147.4 mg/kg [173]. The standard method for nitrite determination in food is the Griess test [184], which has a typical detection limit of 1–2 µM [185]. Unlike ion-selective electrodes, the components involved in this assay cannot be reused, and the food components require extensive treatments for the detection. Voltametric approaches for nitrate detection in acidic media have been reported. Abit et al. [186], developed a method involving Cu/Au electrocatalysis by electrodeposition, that could determine nitrate concentrations as low as 0.52 µM. The device could be used for the characterisation of tap water samples. Similar approaches have been implemented for nitrite detection. Cancelliere et al. [187] developed graphite-screen-printed electrodes based on post-consumer polyethene terephthalate. This sensor could quantify nitrites down to 3.3 nM. However, the number of reported studies on potentiometric sensors for nitrate and nitrite detection is low.

Nitrate-selective sensors have extensively been reported in agriculture applications, given the increasing use of fertilisers, and the key role of nitrate in plant growth. Some examples of ion-selective electrodes applied to this field include soil contamination monitoring [188], and the monitoring of nutrient solutions in hydroponic cultures [189]. However, there is a shortage of ionophores for nitrate sensing. One of the earliest works using these devices for the characterisation of vegetable food samples was described by Consalter et al. [190]. Carrots, wild endive, celery, chicory, spinach, and parsley were characterised using a commercially available Orion sensor. The highest concentration of nitrate was measured for celery, with 163 mg/kg fresh product. Pérez-Olmo et al. [191] employed a similar device for the study of nitrite in meat products. The team determined the performance of the device, which could detect nitride down to 4 µM, and showed good selectivity towards common interferences, including nitrate ($log{K}_{{PO}_4^{3-},X}^{Pot}=-2.4$). Using this method, the team determined that the concentration of nitrites in some of the meat products employed, such as pork paté and corned beef, have a value higher than the maximum allowed concentration of 100 mg kg⁻¹. A key challenge in the determination of nitrites in liquid samples is the high concentration of interference ions, which could lead to inaccurate results, especially when the selectivity of the measuring device is low. Pankratova et al. [192] enhanced the performance of devices through the incorporation of a device for an in-line acidification system. This sensor incorporated a cation-exchange membrane to reduce the pH of samples to 5. This process reduced the limit of detection by 2 orders of magnitude using nitrite ionophore VI as a sensing molecule inside a plasticized PVC film. This sensor allowed for the assessment of nitrites even in highly concentrated environments such as salt water. The final device configuration is depicted in Figure 4a,b. Metal complex ionophores embedded in a nitrophenyl octyl ether have also been developed as potential selective molecules for nitrite [193].

Within the past few years, the development of carbon-based nanomaterials has enabled the design of selective potentiometric sensors that can tackle the challenges of the lack of ionophores for nitrate and nitrite detection. Using laser-induced graphene, Soares et al. [194] developed a nitrate-selective electrode that showed a comparable performance to standard methods, with a Nernstian sensitivity close to 60 mV dec⁻¹ and a detection limit of 7.2 µM. The final device could be incorporated in the measurement of nitrates in meat samples, including sausages, ham, and bacon after the extraction of nitrates using DI water. A summary of the reported work in nitrite and nitrate detection in food is provided in

Table 8. Summary of applications of nitrate and nitrite-selective electrodes. Performance metrics include limit of detection, sensitivity, pH working range, and selectivity.

Material	Ion	Sensitivity (mV Log[NOx−])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
Orion 93-07 nitrate ion-selective electrode	NO3−	-	-	-	100–110	-	Carrot, wild endive, chicory, spinach, parsley, and celery	[190]
Orion 93-46 nitrite ion-selective electrode	NO2−	−54.7	10 µM	4.0–5.5	-	$log{K}_{I^{-},{NO}_2^{-}}^{SSM}=-1.2$ $log{K}_{{Cl}^{-},{NO}_2^{-}}^{SSM}=-2.3$ $log{K}_{{NO}_3^{-},{NO}_2^{-}}^{SSM}=-2.4$ $log{K}_{{ClO}_4^{-},{NO}_2^{-}}^{SSM}=-2.2$	Meat products (pH = 5.5 in buffer)	[191]
FKL cation exchange membrane	NO2−	−57.1	0.55 µM	4.0–6.0	-	$log{K}_{{OH}^{-},{NO}_2^{-}}^{SSM}=+2$ $log{K}_{{Cl}^{-},{NO}_2^{-}}^{SSM}=-3.7$ $log{K}_{{NO}_3^{-},{NO}_2^{-}}^{SSM}=-3.3$ $log{K}_{{SO}_4^{-},{NO}_2^{-}}^{SSM}=-3.9$	Drinking water (pH = 5.0 by exchange matrix)	[192]
Silver (I) bisdiethyldithiocarbamate	NO2−	−56.3	6.7 µM	4.0–9.0	92–97	$log{K}_{{SO}_4^{2-},{NO}_2^{-}}^{SSM}=-3.0$ $log{K}_{{Cl}^{-},{NO}_2^{-}}^{SSM}=-4.1$ $log{K}_{P{O}_4^{3-},{NO}_2^{-}}^{SSM}=-2.7$ $log{K}_{{CO}_3^{2-},{NO}_2^{-}}^{SSM}=-3.0$	Cherry, apricot, drinking water	[193]
Laser-induced graphene	NO2−	−59.5	7.2 µM	-	-	$log{K}_{{{NO}_3^{-}}^{-},{NO}_2^{-}}^{SSM}=-1.39$ $log{K}_{{Cl}^{-},{NO}_2^{-}}^{SSM}=-1.74$ $log{K}_{{H_2{PO}_4^{-}}^{ },{NO}_2^{-}}^{SSM}=-3.17$	Meat products	[194]

.

4.4. Sulphide Detection

Sulphites are a family of compounds that contain sulphate IV ions. These molecules are naturally present in multiple foods and drinks such as wine [195]. Moreover, they are used as preservatives in dried fruits [196], dried potatoes [196], meat [197], and seafood [198], given its antioxidant activity. The European Food Safety Authority (EFSA) recommends an intake no higher than 0.7 mg per kg of body weight. While it is indicated that more evidence is required, some preliminary toxicity effects, including a delayed response of nerve cells, have been observed when exposing the tissue to high concentrations of sulphites [113]. Moreover, the prevalence of sulphite intolerance is relatively high within the global population, especially among patients with asthmatic conditions (3–10%) [199].

Typically, sulphite concentrations are determined using ion exclusion chromatography and titration methods, such as titrimetric and Monier–Williams distillation [200]. However, these methods are destructive and require lengthy treatments for quantification. Early approaches in the potentiometric detection of sulphites include the use of inorganic materials such as titanium phosphate embedded inside an epoxy matrix [201]. Despite the Nernstian response in the presence of sulphites, the sensors showed poor selectivity, with high values for selectivity coefficients for all the studied anions (i.e., I⁻, NO₂⁻, S²⁻, NO₃⁻), and $log{K}_{X,{SO}_2^{3-}}^{Pot}>-1$ for SO₄²⁻, F⁻, and HPO₄²⁻ only. Zeolitic materials have also been used as selective materials, improving the selectivity of the final devices [202]. However, the selectivity coefficients are relatively high, only being $log{K}_{{SO}_2^{3-},X}^{Pot}>-2$ in the case of Br⁻, SO₄²⁻, C₂O₄²⁻, and ClO₃⁻.

Similar to previous anion ion-selective electrodes, the presence of sulphites can be detected through the incorporation of ionophores inside a PVC film containing ionic sites. However, to date, there is a shortage of commercially available sulphite sensors [203]. The topology of ionophores can impact on the sensitivity and selectivity for sulphites, as well as the lifetime and response times. Generally, neutral ionophores with high lipophilicity show a long lifetime by minimising the diffusion into the sample media. Jeon et al. [204] developed a sulphite ionophore based on different Calix [4] diquinones that could quantify sulphite ions with a Nernst sensitivity and a limit of detection in the range of 10⁻⁵ M. While the final devices showed a lifetime in the range of 1 month, the selectivity was poor when compared against perchlorate and thiocyanates. Cobalt(II) phthalocyanine has also been incorporated, with low detection limits, within the micromolar range [205], and high selectivity with a coefficient $log{K}_{{PO}_4^{3-},X}^{Pot}<-5$ when compared against fluoride, sulphate, and phosphates. However, selectivity was not determined in this case against common interferences such as thiocyanate and perchlorate. These ions are likely to represent significant interference, following the work of Hassan et al. [206], who also developed ion-selective electrodes based on Cobalt(II) phthalocyanine embedded inside plasticized PVC films, but with low thiocyanate selectivity. Although thiocyanate is commonly found in cruciferous vegetables, including mustard, bamboo shoots, and kale [207], the concentration in wine is typically low. As such, these devices were restricted to the use in low thiocyanate samples. These devices were also limited by the narrow working pH. Given the relatively high acidity of wine, samples had to be mixed with a buffer to increase the pH to 5, limiting its use. A summary of recently reported approaches for sulphite detection in food analysis is provided in

Table 9. Summary of applications of sulphate and sulphite-selective electrodes. Performance metrics include limit of detection, sensitivity, pH working range, and selectivity.

Material	Ion	Sensitivity (mV Log[SOx−])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
Cobalt(II) Phthalocyanine	SO32−	−29.8	1.1 µM	5.0–7.2	-	$log{K}_{{NO}_3^{-},{SO}_3^{2-}}^{SSM}=-4.2$ $log{K}_{{NO}_2^{-},{SO}_3^{2-}}^{SSM}=-4.7$ $log{K}_{{Cl}^{-},{SO}_3^{2-}}^{SSM}=-4.9$ $log{K}_{F^{-},{SO}_3^{2-}}^{SSM}=-5.7$	Malt beverage, juice (pH = 6.0 in buffer)	[205]
Cobalt(II) Phthalocyanine	SO32−	−27.4	1 µM	5.0–7.0	-	-	Beer, malt beverage, vinegar, sugar lumps, grape drink (pH = 5.0 in buffer)	[206]
5,10,15,20-tetraphenyl(porphyrin)zinc(II)	SO2−	−59.5	3.7 µM	-	95–104	-	Wine (pH = 1.6 in buffer)	[208]
Fe3O4 nanoparticles	SO42−	−29.7	3.1 µM	-	-	$log{K}_{{Cl}^{-},{SO}_4^{2-}}^{SSM}=+1.4$	Drinking water	[209]

.

4.5. Perchlorate Detection

Perchlorate ions consist of a tetrahedral molecule containing four oxygen atoms and a central chlorine. This compound is typically found associated with cations, such as potassium or ammonium, to form salts that disassociate in solution. Perchlorates are strong oxidising agents that can be ubiquitously found in nature [210]. Given its non-volatile and fast diffusion, it has been found in drinking water aquifers [211] and soil samples [212]. The origin of this compound in food is diverse and can derive from naturally occurring reactions and human pollution. In nature, perchlorate is generated in the atmosphere, through the exposure of chlorine to an electrical discharge [213]. Perchloride can then diffuse into the rain and be incorporated into the food chain. Moreover, this component has been used in the development of solid fuel for rockets [214], and it can be found as a common contaminant in some fertilizers [215]. The increasing amount of perchlorate in drinking water [216], suggests that the use of sodium hypochlorite, a common treatment for degrading bacteria and toxins in water, could represent a major contributor to the widespread contamination with perchlorates [217].

Perchlorates have been shown to exhibit endocrine-disrupting effects, interfering with the correct functioning of the thyroid gland [218], due to the inhibition of iodine uptake [219]. It has also been suggested that early exposure during pregnancy to perchlorate can impact brain development [220]. Thus, it is crucial to monitor the levels of perchlorates in food to avoid adverse health effects, especially within vegetable-rich diets. The current legal limit of perchlorate in food is 0.05 mg kg ⁻¹, although it is still allowed as additives in packaging for food contact materials [221]. Given the prevalence of perchlorate in water and soil, plants can accumulate this component in leaves and fruits. A test conducted by the US FDA across 27 different types of food identified the high presence of perchlorate in lettuce and spinach, with concentrations of 10 µg kg⁻¹ and 115 µg kg⁻¹, respectively [222]. Moreover, it has been reported in infant formulas and breast milk [223] among other sources.

The standard method for perchlorate detection in food is LC-MS/MS [224]. Ion-selective electrodes have also been developed for the continuous monitoring of this compound in solution. Typically, ionophores containing a central metallic atom such as gold [225] are used embedded in a plasticized PVC film. However, these electrodes have poor selectivity towards similar complex and bulky ions such as IO₃⁻ and MnO₄⁻. Rezaei et al. [226] developed a cobalt-based electrode that could determine the concentration of perchlorate within a sub-micromolar level of 5.6 × 10⁻⁷ M, and showed good selectivity even in the presence of other complex anions, with coefficients below $log{K}_{{ClO}_4^{-},X}^{Pot}<-2$. However, this limit of detection could not meet the current regulatory requirements. The current limit imposed for perchlorate in drinks by the US Environmental Protection Agency is around 1.8 × 10⁻⁷ M. Neutral ionophores, incorporating a dodecabenzylbambus [6]uril with a central cavity for perchlorate binding, have also been developed [227]. This device could achieve a sub-micromolar detection limit of 4 × 10⁻⁸ M while showing good selectivity. The performance of devices can be further improved by the incorporation of conductive components such as multi-walled carbon nanotubes [228]. However, the use of these devices in food and beverage quantification is still limited.

The most commonly reported use of perchlorate ion-selective electrodes is in the characterisation of beverages and drinking water samples. As mentioned, one of the limitations in the development of perchlorate-selective electrodes is the low number of ionophores, as well as the strict detection limits required. Multiple ionophore-based ion-selective electrodes have been developed for water quality characterisation. A potentiometric device based on neocuproine-Cu(II) has been developed with good selectivity and a detection limit below the micromolar level. In this case, the lowest selectivity coefficient achieved was obtained for thiocyanate, $log{K}_{{ClO}_4^{-},{SCN}^{-}}^{Pot}=-1.3$ [229], and the detection limit was 0.1 µM. Ionophores based on bis(dibenzoylmethanato)nickel(II) have also been developed, enabling the quantification of perchlorates even in rainwater. However, the limit of detection in this case was relatively high, in the range of 7 × 10⁻⁷ M [230], being higher than the regulatory limits.

While the use of potentiometric devices in drink characterisation has been widely reported, the use of solid food for perchlorate analysis is scarce. The detection of perchlorate in vegetables, including potato, lettuce, pepper, beans, and tomato, has been demonstrated by the use of a multi-commutated flow system, achieving sub-micromolar-level detection and low selectivity coefficients below $log{K}_{{ClO}_4^{-},X^{-}}^{Pot}<-9.2$. In none of the studied cases, the observed concentration surpassed the current regulatory limits [231]. In this case, the extraction of perchlorates was conducted by blending samples and using a sodium sulphate solution. A summary of materials employed in the detection of perchlorides in food is provided in

Table 10. Summary of applications of perchlorate-selective electrodes. Performance metrics include limit of detection, sensitivity, pH working range, and selectivity.

Material	Ion	Sensitivity (mV Log[ClO4−])	LOD	pH Range	Recovery (%)	Selectivity	Sample	Ref.
Neocuproine	ClO4−	−53.0	0.1 µM	3.0–11.0	-	$log{K}_{{BF}_4^{-},{ClO}_4^{-}}^{MPM}=-1.5$ $log{K}_{{SCN}^{-},{ClO}_4^{-}}^{MPM}=-1.3$ $log{K}_{{CO}_3^{2-},{ClO}_4^{-}}^{MPM}=-2.4$ $log{K}_{I^{-},{ClO}_4^{-}}^{MPM}=-2.8$	Drinking water	[229]
bis(dibenzoylmethanato) nickel(II)	ClO4−	−58.5	0.7 µM	3.0–9.0	92–128	$log{K}_{{NO}_3^{-},{ClO}_4^{-}}^{MPM}=-1.6$ $log{K}_{{HCO}_3^{-},{ClO}_4^{-}}^{MPM}=-2.9$ $log{K}_{{SO}_4^{2-},{ClO}_4^{-}}^{MPM}=-4.1$ $log{K}_{{Cl}^{-},{ClO}_4^{-}}^{MPM}=-3.0$	Drinking water	[230]
Bisnaphthalimidopropyl 4,4′-diaminodiphenyl methane	ClO4−	-	0.3 nM	-	95–110	$log{K}_{{NO}_3^{-},{ClO}_4^{-}}^{SSM}=-10.2$ $log{K}_{{ClO}_3^{-},{ClO}_4^{-}}^{SSM}=-10.8$ $log{K}_{{SO}_4^{2-},{ClO}_4^{-}}^{SSM}=-9.31$ $log{K}_{{Br}^{-},{ClO}_4^{-}}^{SSM}=-9.61$	Potato, lettuce, tomato, red pepper, beans	[231]

.

5. Limitations of Ion-Selective Electrodes in Food and Future Trends

Food and beverages are complex media, containing a high number of different ions and organic components that can act as interferences for ion-selective electrodes. However, The low limit of detection requirements for certain toxic ions imposed by regulatory organisms demand highly accurate and low-cost electrochemical sensors that can be used for the detection of toxic ions. Established methods, such as ICP-MS, can meet current analytical standards, but they are limited by operational costs and the need for specialized staff for the analysis of the results. Potentiometric sensors have the potential to overcome these challenges, by offering an accurate and cost-effective assessment of specific ions in real time. However, the current landscape of electrodes for food analysis reveals a shortage of materials that can be applied to the selective detection of toxic ions in food. In particular, the range of anion-selective ionophores available commercially is limited, as well as ligands for complex ions, including methylmercury.

Traditionally, ionophores based on Schiff bases, porphyrin-based compounds, and crown ethers have been employed as specific ligands for heavy metals and anions embedded inside plasticized PVC films. Sensors based on plasticized PVC films typically show low detection limits, which can reach the picomolar level, and high selectivity. Moreover, the number of preparation steps required to characterise food and beverage samples using these electrodes is significantly lower compared to more complex methods such as chromatography. However, ionophores can be toxic [232], and the lifetime can be limited due to the sustained diffusion of ionophores into the sample phase [233]. This diffusion can limit their direct use in food, since the presence of toxic ionophores could represent a risk for consumers.

Some inorganic materials such as copper nanoparticles and metallic silver can be used for the monitoring of simple anions with high long-term resilience. However, in most cases, they cannot meet the selectivity requirements for the trace detection of ions. In the case of F⁻ detection, LaF₃ is the most commonly used material. However, the detection limit for these devices is often high, within the micromolar range. Moreover, the selectivity of these materials is lower than membranes based on neutral ionophores. Some engineering approaches, such as changing the pH of solutions to reduce the detection limit, have been employed [192]. However, this technology is only applicable to liquid samples.

Ionically imprinted materials represent a promising approach in the field that has already been applied to certain heavy metals and metalloids where ionophores are not readily available, including arsenic and lead. These materials are synthesised using the target analyte as a template, allowing for the detection of virtually any compound by optimising fabrication conditions [73]. Using this approach, sub-nanomolar level detection has been achieved, even in complex media such as fruit juice, and the material composition can be tailored to overcome the current limitations in ion sensors. This technology has been recently applied to simple anions such as chloride, with a material composition able to overcome standard ion-selective electrodes, such as the Nernst sensitivity limit [234]. Moreover, they can be applied to complex ions where ligands have not been reported yet, including methylmercury [99]. As such, the implementation of this technology could be key in the field.

A key challenge in the development of sensors for food analysis is the operational pH of devices. Food and beverages have highly variable pH values, which can range between acidic, in the case of juices, to alkaline. Moreover, extraction methods used in solid foods often involve the use of strong acids, such as HNO₃. Changes in the sample pH can accelerate ionophore leaching in the case of neutral ionophore-based sensors [140], and can degrade inorganic films. Devices with a wide acidity working range are desirable. Ion-selective electrodes based on neutral ionophores embedded inside plasticized PVC films typically show wide pH ranges, since the polymeric film can hinder the diffusion of ions. However, in the case of electrodes based on metallic compounds, or ion-imprinted polymers, changes in pH can lead to a change in the conformation of materials. Some materials used for sensing, such as tungsten, implemented in HPO₄²⁻ sensors, can only be operated within a pH of 7–8. Consequently, samples must be pre-treated using buffers [166]. In the case of imprinted materials, sensors typically involve polar groups such as carboxylic acid or amines, which can interact with H⁺, changing their affinity towards ions.

A comparison between the different materials that can be used in the selective detection of ions in food is provided in Figure 5.

The lack of materials for the detection of heavy metals could potentially be solved through the use of multiple sensing devices in the form of electronic tongues, combined with advanced algorithms for data analysis, including AI. This technology can combine information from multiple potentiometric sensors, and provide an accurate evaluation of food (Figure 6a). Although the concept of electronic tongues using multiple ion-selective electrodes is relatively recent, this technology has already been applied to food quality analysis with low detection limits. Da Costa et al. [236] incorporated three potentiometric sensors for the detection of Cu²⁺, Pb²⁺, Cd²⁺, and Fe³⁺ in wine, achieving a performance that can comply with current regulations. Similarly, Wilson et al. [237] developed a system for the simultaneous detection of chlorate and sulphides using five potentiometric sensors, enabling detection within the micromolar level. As such, the use of AI in sensing can enhance the accuracy of devices, and potentially be used for the detection of ions that cannot be easily characterised using existing materials. Moreover, the information from multiple ion sensors could be incorporated for the assessment of other pollutants in food such as hydroquinone. This concept was tested by Molinara et al. [238], who combined multiple carbon-based nanomaterials to enable the simultaneous detection of hydroquinone and benzoquinone. The device could determine the concentration of both pollutants down to the micromolar level. However, the use of AI in sensing requires high computational power, and, in some cases, they cannot be easily integrated into existing systems [239].

Besides the lack of specific materials for ion detection, the need, in the case of solid foods, for pre-treatments to accurately measure ions limits the usability of devices. This pre-treatment often involves the use of hazardous chemicals, such as strong acids and high temperatures (Figure 6b). However, direct methods of integrating sensors directly inside food samples or blending them carry the risk of underestimating the measured quantities of ions. For example, dietary fibres have been observed to adsorb ions, decreasing their activity [240]. As such, new green methods for analyte extraction from samples are desirable.

The use of microfluidics has also been established as a potential alternative for accurate food analysis with limited product manipulation. However, this approach is, in most cases, limited to liquid foods or emulsions [241]. New technologies in smart farming to directly extract ions from leaves could potentially be applied in the food field. Reverse iontophoresis is a method where charged compounds are extracted through an applied voltage. This method has been used in the non-invasive measurement of pH and K⁺ ions directly from tomato leaves [242], as well as the extraction of metabolites from basil leaves [243], and they could be incorporated in the analysis of vegetable-derived foods (Figure 6c).

Figure 6. (a). Schematic representation of electronic noses. These systems mimic the processes that take place in the brain, utilising the information from multiple sensor sources to generate a unique profile that can be used to evaluate food quality. Figure reused from [236], Copyright 2014 Elsevier B.V. (b). Picture summary of pre-treatment options for food processing, including dry ashing and wet ashing. After treatment, samples are analysed using gold-standard methods such as ICP-MS. Figure reused from [244], copyright 2023, distributed under Creative Commons Attribution (CC BY). (c). Reverse iontophoresis represents an alternative for the extraction of ions from food samples. Using this method, charged compounds can be extracted from samples directly, potentially avoiding preparation steps. Figure reused from [242], copyright 2023, distributed under Creative Commons Attribution (CC BY).

Figure 6. (a). Schematic representation of electronic noses. These systems mimic the processes that take place in the brain, utilising the information from multiple sensor sources to generate a unique profile that can be used to evaluate food quality. Figure reused from [236], Copyright 2014 Elsevier B.V. (b). Picture summary of pre-treatment options for food processing, including dry ashing and wet ashing. After treatment, samples are analysed using gold-standard methods such as ICP-MS. Figure reused from [244], copyright 2023, distributed under Creative Commons Attribution (CC BY). (c). Reverse iontophoresis represents an alternative for the extraction of ions from food samples. Using this method, charged compounds can be extracted from samples directly, potentially avoiding preparation steps. Figure reused from [242], copyright 2023, distributed under Creative Commons Attribution (CC BY).

Despite these limitations, ion-selective electrodes are still positioned as a promising technology in the analysis of food samples. Compared to voltametric-based approaches, ion-selective electrodes can be used in the continuous monitoring of analytes while achieving low detection limits and requiring low operational costs. Although the commercialisation of materials that can selectively bind specific ions such as methylmercury or Pb²⁺ is limited, the development of alternative approaches such as ionic imprinting shows promise for the design of accurate and accessible technologies for food quality monitoring.

6. Conclusions

Food samples represent complex media, containing a wealth of organic and inorganic components that could interfere with the measurements. Multiple potentiometric sensors have been developed to enable the fast and accurate detection of ionic pollutants. However, the uptake of this technology in the food industry has been slow. This is a consequence of the lack of ionophores that can meet current standards, in terms of the limit of detection and selectivity. Moreover, changes in food pH still represent a challenge, since most devices can only be operated within a specific pH range.

Advanced materials, including graphene oxide nanosheets and inorganic crystals (LaF₃), have been used, with good selectivity and long lifespans. Moreover, ion imprinting can be used to synthesise selective materials that can be tailored to any compound at a low cost. However, more research is required to expand this approach to challenging analytes such as arsenic and oxyanions.

Besides the variable pH commonly found in liquid samples, the detection of ions in solid foods also represents a challenge, given the need for complex preparation procedures to extract ions. These preparation steps involve strong acids or high temperatures. As such, new devices that combine non-invasive methods for extraction, and the latest advances in AI are needed. Urgent collaborative efforts are required to accelerate the adoption of these innovations and ensure the safety and integrity of our food supply.