chemosensors Article Determination of Chemical Oxygen Demand (COD) Using Nanoparticle-Modified Voltammetric Sensors and Electronic Tongue Principles Qing Wang and Manel del Valle * Sensors and Biosensors Group, Department of Chemistry, Universitat Autònoma de Barcelona, Bellaterra, 08193 Barcelona, Spain; qing.wang2@e-campus.uab.cat * Correspondence: manel.delvalle@uab.cat; Tel.: +34-93-581-32-35 Abstract: This manuscript reports the use of nanoparticle-modified voltammetric sensors for the rapid and green determination of chemical oxygen demand in river waters and waters from agricul- tural waste. Four different variants of modified electrodes have been prepared: CuO nanoparticles electrogenerated over Cu and covered with Nafion film (CuO/Cu-Nf), and graphite–epoxy compos- ites modified with Cu, CuO, and Cu–Ni alloy nanoparticles. The response features of these electrodes were assessed by calibrating them vs. glucose, glycine, ethyleneglycol, and hydrogenphtalate in alkaline media, as samples providing different difficulty in their (bio)degradation characteristics. The most sensitive electrode was demonstrated to be the (CuO/Cu-Nf) electrode, with an LOD of 12.3 mg O2 ·L−1 . The joint information provided by the sensor array showed the ability of estimating both the organic load and the type of sample in terms of difficulty of degradation, in what can be   named an intelligent sensor assembly. Citation: Wang, Q.; del Valle, M. Determination of Chemical Oxygen Keywords: chemical oxygen demand; voltammetry; copper oxide electrode; copper nanoparti- Demand (COD) Using cles electrode; copper oxide nanoparticles electrode; nickel copper alloy nanoparticles electrode; Nanoparticle-Modified Voltammetric electronic tongue Sensors and Electronic Tongue Principles. Chemosensors 2021, 9, 46. https://doi.org/10.3390/ chemosensors9030046 1. Introduction Accompanied with the development of the modern industry, the protection of water Academic Editor: Núria Serrano resources from being polluted by waste effluents is becoming more relevant. Organic compounds are one of the most common pollutants, either in urban waste, agricultural and Received: 22 January 2021 cattle production, or industrial runoffs. Chemical oxygen demand (COD) is an important Accepted: 23 February 2021 parameter for evaluating the degree of water pollution by generic organic compounds, Published: 27 February 2021 being a faster variant to ascertain the actual requirements for its biodegradation, the biochemical oxygen demand (BOD), which has precise long estimation times. COD is Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in defined as the equivalent amount of molecular oxygen (in milligrams of O2 ) required to published maps and institutional affil- decompose by chemical oxidation all the organic compounds contained in 1 L of aqueous iations. solution to carbon dioxide and water. There is a well-established and standardized method in many countries for measuring COD using potassium dichromate as the oxidizer in sulfuric acid media. Nevertheless, this method is very time-consuming, has high costs, and it is environmentally hazardous, as there is a reflux (2–4 h) in the process, and some expensive (Ag), corrosive (H2 SO4 and Cr2 O7 2− ), and highly toxic (Hg and Cr (VI)) reagents Copyright: © 2021 by the authors. are needed throughout the procedure [1]. Licensee MDPI, Basel, Switzerland. In order to circumvent the drawbacks of the conventional method, some new trials This article is an open access article distributed under the terms and and approaches have been studied, some with the use of sensors. Teixeira developed a conditions of the Creative Commons novel photochemiresistor sensor using m-bismuth vanadate material for the determination Attribution (CC BY) license (https:// of COD [2]. Kong and Wu tried to utilize ultraviolet spectroscopy to detect COD values creativecommons.org/licenses/by/ in printing and dyeing waste waters to get a rapid method [3]. Cheng and coworkers 4.0/). developed a high-throughput chemiluminescence (CL) method for the determination based Chemosensors 2021, 9, 46. https://doi.org/10.3390/chemosensors9030046 https://www.mdpi.com/journal/chemosensors Chemosensors 2021, 9, 46 2 of 14 on a potassium permanganate–glutaraldehyde CL system [4]. Photocatalytic and photo- electrocatalytic methods were also studied to determine the COD values of wastewater samples [5,6]. Recently, a portable photoelectrochromic visualization sensor for COD detection was also developed by Wang et al. [7]. Unfortunately, photogenerated elec- tron pairs recombined easily, and using UV light increased the costs associated to COD measurements [8]. In recent years, increasing attention has been paid to electrochemical methods utilizing differently modified electrodes and looking for the ones with the highest catalytic activity. Electrochemical methods are among the preferred ones because these kinds of methods are highly sensitive, time-saving, low cost, and easy to operate. As reported, many materials, such as TiO2 [9], Rh2 O3 [10], IrO2 [11], CoO [12], Ni [13], or Ni Cu alloy [14], were used as electrocatalysts to fabricate sensors for COD measurements based on amperometric detection. Nevertheless, it is difficult to oxidize some organic substances directly with metal or carbon sensors, in which high potentials are needed. However, the potentials required are sometimes outside the potential windows of electrochemical devices reported, whose positive potentials are limited by the oxidation of water [15]. Copper electrodes have wide acceptance because of the fact that copper is a powerful and efficient electrocatalyst in alkaline media facilitating the degradation of carbohydrate and aminoacids by oxidation; these two types of compounds are common organic substances in wastewater and are considered as the main sources for COD. Up to now, different electrode variants to estimate organic load based on copper or copper oxide have been reported [16–18]. Nevertheless, it is nearly impossible to achieve the full oxidation of all organic contaminants to carbon dioxide and water by electrochemical methods performing at a high enough voltage. In spite of the fact that it is almost impossible to decompose all the organic contaminants, getting close to it is a necessary step to improve the accuracy and precision in the COD electrochemical determination. In addition to this information, whether a certain organic compound is easily decomposed or this is difficult contributes some important additional information to the organic load of a sample. Recalcitrant organic compounds are normally a challenge when their decontamination is sought, and detecting their presence can help in the management and operation of wastewater treatment plants. Winquist and coworkers utilized the voltammetric electronic tongue technique for multicomponent analysis of drinking water [19]. Inspired by this research, the combination of voltammetric sensors and an electronic tongue is a good choice to try to get a quantitative and qualitative analysis of a water sample simultaneously. According to the IUPAC technical report, the electronic tongue is defined as “a multisensory system, which consists of a number of low-selective sensors and uses advanced mathematical procedures for signal processing based on the pattern recognition and/or multivariate data analysis” [20,21]. Therefore, the electronic tongue, consisting of an array of non-specific, poorly selective sensors, supported by chemometric processing, can be used for analyzing the composition of multispecies solutions qualitatively and quantitatively. Principal component analysis (PCA) is commonly used in the data-processing algorithms to identify or classify the components [22]. The PCA technique, through data processing and rearranging, calculates a large number of variables to get fewer new variables, which not only reduces the number of variables but also retains the useful information of the original data. In the work here presented, four electrodes based on copper, copper oxide, and nickel copper alloy were fabricated to compose an electronic tongue array. This voltammetric array, consisting of a Nafion film-covered electrodeposited CuO/Cu electrode as well as copper, copper oxide, and nickel copper alloy nanoparticle–graphite–epoxy composite elec- trodes, was employed for analyzing river water samples quantitatively and qualitatively. The COD values were calculated based on the calibration equation obtained based on the linear relationship of COD values and current intensity. The PCA technique applied to the profiles shown in the voltammograms was used to evaluate the main component of a polluted water sample, which was able to indicate whether this is easy or difficult to be degraded. Chemosensors 2021, 9, 46 3 of 14 2. Experimental 2.1. Chemicals and Reagents Reagents in this work were used as received from the manufacturers without any purification and the purities were all of analytical grade. All solutions were prepared using ultrapure water purified by a Milli-Q system (Millipore, Billerica, MA, USA). Nafion (5%, w/w) was purchased from Aldrich (St. Louis, MO, USA); copper (Cu) nanoparticles (40–60 nm) and copper (II) oxide (CuO) nanoparticles (<50 nm), which were used as modifiers, from Sigma-Aldrich (St. Louis, MO, USA); gamma alumina powder, which was used for polishing the surface of copper block, was purchased from the company CH Instruments (Austin, TX, USA). Nickel copper alloy powder (Ni-Cu alloy) nanoparticles (<80 nm) were purchased from Nanografi Nano Technology (Germany). Glucose, ethylene glycol, potassium hydrogen phthalate, glycine, and sodium hydroxide (NaOH) were purchased from Sigma-Aldrich (St. Louis, MO, USA); graphite powder (<50 µm) used for the construction of the electrodes was purchased from BDH (BDH Laboratory supplies, Poole, UK) and epoxy for composite electrodes, the corresponding H77 resin, and hardener were purchased from Epoxy Technologies (Billerica, MA, USA). 2.2. Apparatus All the cyclic voltammetric measurements were performed with use of a PGSTAT 30 Autolab potentiostat (EcoChemie, Netherlands) with GPES 4.7 version software (Eco- Chemie). Electrochemical measurements were carried out in a conventional three-electrode system composed of fabricated electrodes as working electrodes, a metallic platinum elec- trode, and an Ag/AgCl electrode, as the auxiliary and reference electrode, respectively. The working electrodes were fabricated by us, and the auxiliary and reference electrodes were bought from Fisher Scientific, which is a combination ORP electrode. The sealed tube for COD detection was purchased from the HACH company, Spain. 2.3. Electrodes Fabrication 2.3.1. Fabrication of Electrode E1 Electrode E1 is a Nafion-covered electrodeposited CuO/Cu electrode. To fabricate E1, as shown in Figure 1A, a copper block, 5 mm in diameter and 4 mm long, was soldered to an electrical connector. Then, this couple was inserted into a 20 mm long PVC tube, 6 mm in diameter, wrapped with H77 epoxy resin and cured for 24 h in an oven at 80 ◦ C. Then, the surface of the electrode was polished with emery papers of decreasing grain size and alumina powder (1.0, 0.3 and 0.05 micron). Then, the final glossy and smooth surface was obtained and thoroughly rinsed with deionized water. After thoroughly rinsing with deionized water and drying, then, 10 microliters of Nafion was dropped on the electrode surface and keep it in air for 48 h, and then all the solvents were volatilized. The electrochemical deposition was conducted in a solution of 0.05 mol·L−1 NaOH solution with cycling potential from −0.1 to +0.8 V for 50 times, and a dark film formed on the surface, which consists of copper oxide (CuO) nanoparticles. 2.3.2. Fabrication of Electrodes E2, E3, and E4 Electrodes E2, E3, and E4 are graphite-modifier (metal nanoparticles)–epoxy resin composite electrodes, and the modifiers are copper (Cu), copper oxide (CuO), and nickel copper alloy (Ni Cu alloy) nanoparticles, respectively. To fabricate the electrodes, as shown in Figure 1B, a copper disk (5 mm diameter) was soldered to an electrical connector. Then, the couple was inserted into a 20 mm long PVC tube, and a 4 mm deep cylinder cavity was formed. The cavity was filled with different paste composites, consisting of 54% graphite, 5% modifier, the corresponding epoxy resin, and hardener. This ratio of the components in the composite and the amounts of modifiers were inspired by the reported works in our laboratory [23,24]. Then, these electrodes were cured at 40 °C for 48 h. Electrode surfaces were polished with emery paper of decreasing sizes and rinsed with deionized water. Finally, electrodes E2, E3, and E4 with shiny and smooth surfaces were obtained. ing of 54% graphite, 5% modifier, the corresponding epoxy resin, and hardener. This ratio of the components in the composite and the amounts of modifiers were inspired by the reported works in our laboratory [23,24]. Then, these electrodes were cured at 40 ℃ for 48 h. Electrode surfaces were polished with emery paper of decreasing sizes and rinsed with Chemosensors 2021, 9, 46 deionized water. Finally, electrodes E2, E3, and E4 with shiny and smooth surfaces were 4 of 14 obtained. Figure 1. Diagram of construction of the Cu/CuO electrode (A) and the graphite–epoxy electrodes Figure 1. Diagram modified of construction with metal nanoparticlesof(B). the Cu/CuO electrode (A) and the graphite–epoxy electrodes modified with metal nanoparticles (B). 2.4. Electrochemical Measurements 2.4. Electrochemical 2.4.1. Measurements Cyclic Voltammetric Response 2.4.1. Cyclic Voltammetric Response Four different organic compounds with different reducibility, glucose, glycine, ethy- leneFour different glycol, organic compounds and potassium with different hydrogen phthalate were reducibility, used as CODglucose, standard glycine, eth- substances ylene glycol, and potassium hydrogen phthalate were used as COD standard substances to evaluate the performance of the prepared sensors. The voltammetric responses of to evaluate the performance of the prepared sensors. The voltammetric responses of elec- electrodes to those standard substances were conducted individually in 0.05 M NaOH trodes to thoseelectrolyte background standard substances solution, sincewerethis conducted media isindividually needed for thein 0.05 M NaOH catalytic back- copper op- ground electrolyte solution, since this media is needed for the catalytic copper operation. eration. This electrolyte was inspired by reported works [15,16]; we reduced slightly the NaOH This concentration electrolyte to 0.05 mol was inspired ·L−1 . Theworks by reported reason[15,16]; for that was we to enhance reduced thethe slightly stability NaOH of the sensors and concentration also mol∙L to 0.05 enlarge −1. the Thelifetime of the reason for thatdevices was toin a milder enhance condition. the Each stability of thecyclic sen- voltammogram was recorded after a specific volume of standard solution was sors and also enlarge the lifetime of the devices in a milder condition. Each cyclic voltam- added into the electrolyte solution mogram was recorded −after and stirred a specific volume of standard solution was added into theat for 20 s. The cyclic voltammograms were recorded a scan rate of 50 mV · s 1 by cycling the potential between 0 and +1.0 V vs. Ag/AgCl in electrolyte solution and stirred for 20 s. The cyclic voltammograms were recorded at a room scan temperature. rate of 50 mV·s−1These standard by cycling solutions between the potential of glucose, ethylene 0 and +1.0 Vglycol, and glycine vs. Ag/AgCl in roomare 0.1 mol·L−1 (0.1 M) in concentration. The standard potassium hydrogen phthalate (KHP) temperature. These standard solutions of glucose, ethylene glycol, and glycine are 0.1 solution is 0.05 M in concentration because of the lower solubility. All of these solutions mol∙L−1 (0.1 M) in concentration. The standard potassium hydrogen phthalate (KHP) so- are prepared using ultrapure water purified by a Milli-Q system. lution is 0.05 M in concentration because of the lower solubility. All of these solutions are prepared using ultrapure 2.4.2. Reproducibility water purified by a Milli-Q system. Study To check the stability of the electrodes, the reproducibility tests were carried out in a 2.4.2. Reproducibility Study 2.15 mmol·L− 1 (2.15 mM) glucose in 0.05 M NaOH solution and a blank electrolyte solution (0.05ToMcheck NaOH).the stability The cyclicof the electrodes, thewere voltammograms reproducibility recorded attests were a scan carried rate of 50 out mVin ·s−a1 2.15 mmol∙Lthe by cycling −1 (2.15 mM) glucose potential between in 0.05 0 andM+1.0 NaOH solution V vs. and ain Ag/AgCl blank roomelectrolyte solution temperature. Ten (0.05 M NaOH). The cyclic voltammograms were recorded at a scan rate of cyclic voltammetric responses of these electrodes to the glucose and blank electrolyte 50 mV·s −1 by cycling thewere solutions potential between carried 0 and +1.0 out repeatedly V vs. Ag/AgCl in parallel. in roomus This permitted temperature. Tenstability to evaluate the cyclic voltammetric responses of these electrodes to the glucose and blank electrolyte of the responses, and if any hysteresis or memory effect was present. Before each sample solutions were carried outthese measurement, repeatedly in parallel. electrodes Thiswith were rinsed permitted us towater. deionized evaluate the stability of the 2.5. Measurements of Real Water Samples Sample 1: This sample was originally taken from an effluent of an agricultural field in Gavà, Spain. The original pesticide concentration was really low (less than 0.5 ppm). Then, two pesticides (diuron and bentazon) were spiked to 10 ppm. Then, the sample was obtained after being treated by a fungal treatment process (Trametes versicolor fungus) for 93 days. The pesticide concentrations should be in the range of 5–10 ppm. The COD value of the sample was detected as 1312 mg O2 ·L−1 using the sealed tube method. Sample 2: This sample was obtained from effluent generated in typical post-treatments to wash agricultural equipment and machines after their use and without any other process. Chemosensors 2021, 9, 46 5 of 14 In this case, the initial concentration of pesticides is much higher, 1–100 ppm. The COD value of the sample was detected as 2088 mg O2 ·L−1 using the sealed tube method. Sample 3: This sample was originally taken from a stream very near to a farm in Lleida, Spain and without any other process. The COD value of its clean, surface water was unknown. To conduct the measurements and analysis of real water samples, 100 µL of each real water sample was added into 20 mL 0.05 mol·L−1 NaOH solution to be a test solution. Electrodes were submerged in the test solution and submitted to a cycling potential between 0 and +1 V vs. Ag/AgCl. The cyclic voltammograms were recorded at a scan rate of 50 mV·s−1 under room temperature. Then, the test solutions of Sample 1 and 2 were spiked with 40 µL 0.1 mol·L−1 glucose solution and were measured in the same way after being stirred for 30 s, respectively. 3. Results and Discussion 3.1. Voltammetric Array Response As the departure point shows, our aim in this study is to estimate the organic load of water samples with electronic tongue tools. In this sense, the first step was to check if the electronic tongue sensor array was properly built. For this, the voltammetric response of each electrode toward the different COD reference substances was analyzed as the first step to check the diversity of the generated signals and whether the obtained data were informative and rich enough to build a multivariate multiway calibration model. In this work, four organic substances of different reducibility, glucose, glycine, ethylene glycol, and potassium hydrogen phthalate (KHP) were used as COD standard substances. Glucose and KHP are the representatives of substances that are easy and hard to be oxidized in water samples, respectively. The voltammetric electronic tongue array consists of four sensors, and the three- electrode system was built with a Pt auxiliary and an Ag/AgCl (KCl 3M) reference elec- trode for cyclic voltammetric measurements. The array consists of one electrodeposited CuO/Cu electrode E1 and three composite electrodes; the latter were bulk-modified with Cu nanoparticles (E2), CuO nanoparticles (E3), and NiCu alloy nanoparticles (E4), respec- tively, and were prepared as described in Section 2.3. This choice was aimed at maximizing the observed differences in the obtained voltammograms for the different sensors, taking advantage of the catalytic effects shown by the used metals and the reactive properties boosted by the nanoparticulate material. In the supplementary information, as shown in Figure S1, the scanning electron microscopy image of electrode E1 shows the particles elec- trogenerated on the Cu surface, with electron dispersive spectroscopy (EDS) demonstrating ca. 80% of the nanostructure seen in Cu (Figure S2 and Table S1). Compared with the SEM image of bare copper, we can see that the CuO nanoparticles spread over the electrode, and the surface looks rough and very different from the bare copper surface. The modifiers of electrodes E2, E3, and E4 are also perceptible from the SEM images shown in Figure S3 (supporting information). Cyclic voltammetric responses of fabricated electrodes to standard substances were measured as described in Section 2.4.1. The 0.05 M NaOH solution was chosen to be the background electrolyte to provide an alkaline environment for electrochemical measure- ments. The obtained oxidation parts of cyclic voltammograms are shown in Figures 2 and 3. The figures show the voltammogram shapes of the four substances tested, with the four electrodes E1 to E4, where we can see that the different signals originated by the reference substances are different. Figure 3 represents the voltammograms but groups them in terms of the electrodes used. As we can see from these figures, the oxidation curves of the four electrodes show different shapes toward different substances, which is very useful for the following electronic tongue analysis. In addition, the different electrodes employed also display different signals for the compounds assayed; in there, electrode E1 gets higher cur- rent signals than the other electrodes, and it can make many organic compounds oxidized. ments. The obtained oxidation parts of cyclic voltammograms are shown in Figures 2 and 3. The figures show the voltammogram shapes of the four substances tested, with the four electrodes E1 to E4, where we can see that the different signals originated by the reference substances are different. Figure 3 represents the voltammograms but groups them in terms of the electrodes used. As we can see from these figures, the oxidation curves of the Chemosensors 2021, 9, 46 four electrodes show different shapes toward different substances, which is very useful 6 of 14 for the following electronic tongue analysis. In addition, the different electrodes employed also display different signals for the compounds assayed; in there, electrode E1 gets higher current signals than the other electrodes, and it can make many organic compounds oxi- Unfortunately, the KHP dized. Unfortunately, thesubstance is very KHP substance hardhard is very to be to oxidized with be oxidized withthese thesefour electrodes, four elec- buttrodes, this isbut a fact that can be also assessed from these measurements. this is a fact that can be also assessed from these measurements. Chemosensors 2021, 9, x FOR PEER REVIEW 7 of 14 demonstrates the good function of deposited CuO nanoparticles for improving the elec- trode performance. Figure 2. Oxidation 2. Oxidation curves curves of cyclic of cyclic voltammetricresponses voltammetric responses of of electrodes electrodesE1E1 (A), E2E2 (A), (B),(B), E3 (C), and 𝑛𝑛𝑛𝑛𝑛𝑛E4 (D)E4to(D) fourtostandard Figure chemical oxygen demand (COD) substances: 2.15 mM glucose; 𝐼𝐼p =2.15 0.446 mM∙ 𝑛𝑛 ∙ 𝐹𝐹 ∙ 𝑐𝑐2.15 glycine; ∙ 𝐴𝐴 ∙mM �((C), and E3 √𝑣𝑣 ∙ethylene four standard )glycol; 2.15 mM KHP. (1) chemical oxygen demand (COD) substances: 2.15 mM glucose; 2.15 mM glycine; 2.15 mM ethylene 𝑅𝑅𝑅𝑅 glycol; 2.15 mM KHP. To demonstrate the enhanced performance of the deposited electrode E1, the active surface areas of the four electrodes were calculated according to the Randles–Sevcik Equa- tion (1) [25,26], in which n is the number of electrons involved in the redox reaction, F is the Faraday’s constant (96,485 C∙mol−1), A is the electrode active area in cm−2, c is the con- centration of electroactive substance (mol∙cm−3), 𝑣𝑣 is the scan rate (V∙s−1 ), R is the gas constant (8.314 J∙mol−1·K−1), T is the temperature in K, and D is the analyte diffusion coef- ficient (6.32∙10−6∙cm−2∙s−1). Seven cyclic voltammograms of each electrode at different scan rates (10, 25, 50, 100, 200, 300, and 500 mV∙s−1) were recorded in the solution of 5 mmol∙L−1 [Fe(CN) 6 ]3−/4−. The values of the active area of the electrodes could be calculated from the slope of the fitted line of 𝑣𝑣 1/2 vs. Ip/c graph. The calculated active areas were 139.23 mm2 for E1, 29.10 mm2 for E2, 31.09 mm2 for E3, and 35.65 mm2 for E4, whereas the geometric area was 28.27 mm2 (Ø = 6 mm). These results confirm the correct modification of the electrodes. The active area of E1 is about five times that of the other electrodes, which FigureFigure 3. Oxidation 3. Oxidation curvesofofcyclic curves cyclic voltammetric voltammetric responses of electrodes responses E1, E2, of electrodes E1,E3,E2, andE3, E4 and to four E4standard to four COD sub- COD standard stances: (A) 2.15 mM glucose; (B) 2.15 mM glycine; (C) 2.15 mM ethylene glycol; (D) 2.15 mM KHP. Scan rate: 50 mV·s−1. substances: (A) 2.15 mM glucose; (B) 2.15 mM glycine; (C) 2.15 mM ethylene glycol; (D) 2.15 mM KHP. Scan rate: 50 mV·s−1 . However, it is tedious to compare each cyclic voltammogram to assess the respond- To demonstrate the enhanced performance of the deposited electrode E1, the ac- ing patterns of the electrodes, and multivariate analysis can be of help. In order to show tive surface the areas ofathe result visually, four electrodes complementary werecomponent principal calculatedanalysis according towas (PCA) the conducted. Randles–Sevcik Equation [25,26], in which n To do this study, different replicates at the above specified concentrations were the (1) is the number of electrons involved in redox reac- measured C·mol −1 ), A is the electrode active area in cm−2 , andF inspected tion, by PCA constant is the Faraday’s to observe(96,485 similitudes between samples. In Figure 4, we present the simplified representation of samples through the scores corresponding to the two first principal components for the array response; this has been prepared for the four standard substances. That is, we composed an unfolded vector formed by the four measured volt- ammograms in order to perform the calculation. In there, each compound sample is dif- ferentiated and clearly clustered. From the figure, we can see that glucose, which is easy Chemosensors 2021, 9, 46 7 of 14 c is the concentration of electroactive substance (mol·cm−3 ), v is the scan rate (V·s−1 ), R is the gas constant (8.314 J·mol−1 ·K−1 ), T is the temperature in K, and D is the analyte diffusion coefficient (6.32·10−6 ·cm−2 ·s−1 ). Seven cyclic voltammograms of each electrode at different scan rates (10, 25, 50, 100, 200, 300, and 500 mV·s−1 ) were recorded in the solution of 5 mmol·L−1 [Fe(CN)6 ]3−/4− . The values of the active area of the electrodes could be calculated from the slope of the fitted line of v1/2 vs. Ip/c graph. The calculated active areas were 139.23 mm2 for E1, 29.10 mm2 for E2, 31.09 mm2 for E3, and 35.65 mm2 for E4, whereas the geometric area was 28.27 mm2 (Ø = 6 mm). These results confirm the correct modification of the electrodes. The active area of E1 is about five times that of the other electrodes, which demonstrates the good function of deposited CuO nanoparticles for improving the electrode performance. s √  nDF I p = 0.446 ·n· F · c· A· v· (1) RT However, it is tedious to compare each cyclic voltammogram to assess the responding patterns of the electrodes, and multivariate analysis can be of help. In order to show the result visually, a complementary principal component analysis (PCA) was conducted. To do this study, different replicates at the above specified concentrations were measured and inspected by PCA to observe similitudes between samples. In Figure 4, we present the simplified representation of samples through the scores corresponding to the two first principal components for the array response; this has been prepared for the four standard substances. That is, we composed an unfolded vector formed by the four measured voltammograms in order to perform the calculation. In there, each compound sample is differentiated and clearly clustered. From the figure, we can see that glucose, which is easy to be oxidized, is much more left than other substances. Oppositely, KHP, which Chemosensors 2021, 9, x FOR PEER REVIEW is hard to be degraded, is much more right. In this sense, PCA is informing the organic load that is detected (the PC2 axis is mainly related to this), and the type of compound, i.e., the difficulty of degradation can be also differentiated (PC1 axis). Figure Figure 4. Score 4. Score plot plot of theof the first twofirst two components components obtained obtained after principalafter principal component component analysis (PCA). analy A(PCA). Asamples total of 20 total ofwere 20 samples were analyzed analyzed corresponding corresponding to quadruplicate to quadruplicate determinations determinatio of glucose (blue), glucose glycine (blue), (green), glycine ethylene (green), glycol ethylenehydrogen (red), potassium glycol (red), potassium phthalate hydrogen (purple), and phthalate (pur blank (orange). blank (orange). 3.2. Calibration Curves To evaluate the behavior of each electrode toward each standard substan calibration curves were studied based on the cyclic voltammetry technique. The Chemosensors 2021, 9, 46 8 of 14 3.2. Calibration Curves To evaluate the behavior of each electrode toward each standard substance, some calibration curves were studied based on the cyclic voltammetry technique. The voltam- mograms of the four electrodes responding to the four standard substances in increasing concentration are shown in Figures S3–S6, supplementary information. As shown in these figures, these electrodes show different oxidation capabilities to different substances. The corresponding theoretical COD values were obtained based on the consumed oxygen calculated according to chemical equations. Based on the increasing tendency of current intensity, some calibration equations of current intensity at +0.70 V vs. COD values were obtained as shown in Table 1. These calibration equations are different because of the different capabilities of different electrodes toward different substances. The electrode E1 shows higher current signals to these substances, especially to glucose. As can be seen in Figure 5, the current intensity at +0.70 V can be used as the magnitude proportional to the concentration of glucose; corresponding currents were plotted (see Figure 6) and fitted by least squares (y = m·x + b) with a linear range from 19.20 to 905.5 mg O2 ·L−1 and a regression equation I (µA) = 1.5281·COD+112.75 (R2 = 0.9999). Limit of detection values were calculated as 12.3 mg O2 ·L−1 (using the formula of 3.3Sy/x /m), which improves data from some of the reported electrodes in the literature. Table 1. Calibration lines of the different electrodes employed in this work vs. the different reference organic compounds tested. Measured current vs. organic load. E1 E2 E3 E4 y = 1.5281x +112.75, y = 0.2044x + 14.23, y = 0.1504x + 10.21, y = 0.1908x + 59.18, Glucose R2 = 0.9999 R2 = 0.9934 R2 = 0.9946 R2 = 0.9720 y = 1.8878x + 121.02, y = 0.1067x + 10.49, Glycine – – R2 = 0.9941 R2 = 0.9925 y = 2.6844x + 122.71, Ethylene Chemosensors Glycol 2021, 9, x FOR PEER REVIEW – – – 9 of 14 R2 = 0.9945 KHP – – – – x = COD values (mg O2 ·L−1 ); y = IE=+0.70V (µA). x = COD values (mg O 2 ·L−1); y = I E=+0.70V (µA). Figure5.5.Oxidation Figure Oxidationcurves curvesof ofcyclic cyclicvoltammetric voltammetricresponses responsesofofelectrode electrodeE1 E1totoglucose glucoseofofincreasing increas- ing concentration (from 0.10 to 6.90 mM). Scan rate: 50 concentration (from 0.10 to 6.90 mM). Scan rate: 50 mV·s .mV·s − 1 −1. Chemosensors 2021, 9, 46 Figure 5. Oxidation curves of cyclic voltammetric responses of electrode E1 to 9glucose of 14 of in ing concentration (from 0.10 to 6.90 mM). Scan rate: 50 mV·s−1. Figure 6. Calibration graph corresponding to glucose standards, with current measured at +0.70 V, Figure electrode6.E1. Calibration graph corresponding to glucose standards, with current measured at electrode E1. In comparison with other reported electrodes for COD determination appearing in the literature, as shown in Table 2, the electrode E1 in this work shows a good limit of In comparison with other reported electrodes for COD determination appea detection, linear range, and high response signals. the literature, as shown in Table 2, the electrode E1 in this work shows a good detection, Table 2. Comparison linear of the range, proposed COD and sensorhigh response E1 with signals. others reported in the literature. LOD Linear Range Reproducibility Sensor Electrode Modifier Standard Substance Table 2. Comparison of the proposed COD sensor(mg E1 Owith 2 ·L )others (mg −1 reported O2 ·L−1 ) in the literature. (RSD%) CuO/AgO CuO/AgO sensor [1] Glucose 28 106–1292 Not mentioned nanoparticle Standard Sub- LOD Linear Range Reproduci Sensor Boron Doped DiamondElectrode [8] Modifier- KHP, phenol 7.5 20–9000 1.9 Cobalt oxide/GCE [12] Ni Cu alloy/GCE [14] Cobalt oxide film Ni Cu alloy film stance Glycine Glucose (mg1.1 O 2 ·L−1) 1.0 (mg 5.7–155.8 10–1533 O 2 ·L−1) 5.7 1.0 (RSD% CuO/AgO sensor CuO/Cu electrode CuO/AgO [15] CuOnano- nanoparticle Glucose 20.3 53.0–28014 Not mentioned Ti/Sb-SnO2 /PbO2 [27] Sb-SnO2 /PbO2 Glucose Glucose 0.328 0.5–200 106–1292 <5% Not menti [1] Nano-Ni/GCE [28] F-PbO2 sensor [29] particleNano-Ni Pb (II) particles Glucose Glucose 1.1 15.5 10–1533 100–1200 4.7 Not mentioned Boron Doped This work,Dia- electrode E1 Cu/CuO-Nf Glucose 12.3 19.2–905.5 2.48% - KHP, phenol 7.5 20–9000 1.9 mond [8] 3.3. Reproducibility Study Cobalt oxide/GCE Cobalt oxide Reproducibility film Glycine 1.1electrode used for5.7–155.8 is a basic demand for an the electronic tongue 5.7 [12] technique and for any other methodology. To check the stability of the four electrode types, 10 measurements of 2.15 mM glucose in 0.05 M NaOH solution and a blank electrolyte solution (0.05 M NaOH) were carried out in series by following the experimental conditions described in Section 2.4.2. The current intensity at +0.70 V of the measurements toward electrodes E1, E2, E3, and E4 are displayed in Figure 7. As we can see from these figures, the differences of current intensity of measurements to both glucose and blank solutions are very small. The relative standard deviation (RSD) values based on such measurements were calculated as shown in Table 3. The calculated RSD values of these electrodes are all smaller than 5%, which indicates a good reproducibility of the electronic tongue array composed of such four electrodes. Moreover, these electrodes show remarkable good reproducibility values, especially electrode E1, compared with some of the reported ones (Table 2). The excellent values obtained permit being confident on a series of measurements, and they also assure there are no hysteresis or memory effects. solutions are very small. The relative standard deviation (RSD) values based on such measurements were calculated as shown in Table 3. The calculated RSD values of these electrodes are all smaller than 5%, which indicates a good reproducibility of the electronic tongue array composed of such four electrodes. Moreover, these electrodes show remark- able good reproducibility values, especially electrode E1, compared with some of the re- Chemosensors 2021, 9, 46 10 of 14 ported ones (Table 2). The excellent values obtained permit being confident on a series of measurements, and they also assure there are no hysteresis or memory effects. Figure Figure 7. 7. Current Current intensities intensities at at +0.70 +0.70 V V of ofcyclic cyclic voltammograms voltammograms of of electrodes electrodes E1 E1 (A), (A), E2 E2 (B), (B), E3 E3 (C), (C), and and E4 E4 (D) (D)responding responding to to the the repeated repeated cycle cycle ofof blank blank (blue) (blue) and and 2.15 2.15 mM mM glucose (orange). Scan glucose (orange). Scan rate: rate: 50 mV·s−1−.1 . 50 mV·s Table 3. Relative standard deviation (RSD) values of electrodes E1, E2, E3, and E4 in blank and glucose solutions. RSD E1 E2 E3 E4 Blank 1 4.04% 3.07% 3.90% 3.44% Glucose 2 2.48% 3.57% 3.87% 4.14% 1 0.05 M NaOH solution; 2 2.15 mM glucose solution. 3.4. Real Samples Detection The COD detection of real samples was conducted using the proposed cyclic voltam- metry technique with electrode E1 as described in Section 2.5 and the sealed tube method. The latter method was used to check the accuracy of the former one. COD values found for sample 1 and 2, which were estimated as currents at +0.70V, were lower than the intercept of the calibration line, suggesting that the values were below the LODs. This was mainly due to the large dilution performed (100 µL of these two samples were added into 20 mL NaOH individually). Afterwards, sample 1 and 2 were spiked with glucose, and their voltammo- grams were also recorded under the same condition. The net changes of current intensity at +0.70 V were calculated as 64.11 and 63.57 µA, respectively (see Figure 8). These are near the theoretical current intensity change of 58.27 µA according to the aforementioned equation in Section 3.2 (y = 1.5281x + 112.75). The spiking tests of these three real samples with fewer dilution were further con- ducted; hereby, 10 mL of each sample was added into 10 mL of NaOH individually and then spiked with 100 µL glucose solution. These results have been summarized in Table 4, where the recovery yields of spiked glucose in these samples were very near 100%. These indicate that the electrode E1 has good performance in quantifying the organic compounds that are easy to be degraded. aforementioned equation in Section 3.2 (y = 1.5281x + 112.75). The spiking tests of these three real samples with fewer dilution were further con- ducted; hereby, 10 mL of each sample was added into 10 mL of NaOH individually and then spiked with 100 μL glucose solution. These results have been summarized in Table Chemosensors 2021, 9, 46 4, where the recovery yields of spiked glucose in these samples were very near 100%.11 of 14 These indicate that the electrode E1 has good performance in quantifying the organic com- pounds that are easy to be degraded. Figure Figure 8. 8. Oxidation Oxidation curves curves of of cyclic cyclic voltammetric voltammetricresponses responsesof ofelectrode electrodeE1 E1to toSample Sample11(A) (A)and and22 (B) (B) before before (black (black lines) lines) and after (red lines) being spiked with 0.2 mM glucose. Scan rate: 50 mV·s−1−.1 and after (red lines) being spiked with 0.2 mM glucose. Scan rate: 50 mV·s . Table 4. Recovery yield of spiked glucose detected by electrode E1 toward Sample 1 and Sample2. Table 4. Recovery yield of spiked glucose detected by electrode E1 toward Sample 1 and Sample2. Real Sample COD Measur- Spiked Sample COD Values Added Glucose Recovery Yield Real Samples Real COD Values Found Real Sample Found(mg ing CODVessel Measuring (mg Added Glucose Found Spiked(mgSample Recovery Samples (mg O2(mg ·L−1 )O 2 ·L−1) (mg O2 ·L−1 ) −1 Vessel (mg O · L −1 ) (mg (mgO 2 ·L O · L−1−)1 ) Found (mg O · L −1 ) (%) Yield (%) O 2 ·L ) O 2 ·L ) O 2 ·L ) −12 2 −1 2 Sample 1 1312