Yanjun
Guo
and
Richard G.
Compton
*
Department of Chemistry, Physical and Theoretical Chemistry Laboratory, Oxford University, South Parks Road, Oxford, OX1 3QZ, UK. E-mail: Richard.Compton@chem.ox.ac.uk
First published on 1st July 2021
Chloride quantification is important in drinking water quality control. A bespoke, rapid and reagent free electrochemical method is reported for a simple and accurate chloride sensor specifically for mineral water without the need for added electrolyte. The voltammetry used embraces first the reduction of oxygen to clean and activate the electrode surface and ensure reproducibility without the requirement for any mechanical polishing, followed by silver chloride formation and stripping. A linear correlation was found with silver chloride stripping peak currents and chloride concentrations within the range of 0.4 mM to 3.2 mM on a silver macro disc electrode. The chloride concentrations in two different mineral water samples were measured giving excellent agreement with independent analysis.
mg L−1 | mM | |
---|---|---|
WHO limit | 2503 | 7 |
Mineral waters | ||
Highland | 6.1 | 0.17 |
Evian | 10 | 0.28 |
Tesco | 14 | 0.39 |
Aqua Pura | 15 | 0.42 |
Volvic | 15 | 0.42 |
Nestle Pure Life | 18 | 0.51 |
Buxton | 37 | 1.04 |
Natural mineral waters | ||
Voluvesi | 7.3 | 0.2 |
Saaremaa Vesi | 13.4 | 0.4 |
Saku | 27.1 | 0.8 |
Nabeghlavi | 42–95 | 1.2–2.7 |
Narzan Kislovodsk | 100–150 | 2.8–4.2 |
Smironovskaya | 250–300 | 7.1–8.5 |
Borjomi (sparkling) | 380 | 10.7 |
Jernuk | 300–400 | 8.5–11.3 |
Varska | 444 | 12.5 |
Vichy | 601.5 | 17.0 |
Varska original | 1338 | 37.7 |
Current chloride detection and quantification techniques for application in drinking water include titration, electrochemistry and chromatography,6,8,9 where ion chromatography is specifically recommended by CEN and ISO10 for a large number of samples. Each sample is added to a carbonate–bicarbonate eluant. After ion-exchange the separated chloride ions are directed into a suppressor where the eluant is neutralised. Chloride is identified by retention time standards and quantified via conductivity measurements.8 Apart from facile analysis for multiple samples, ion chromatography also avoids toxic reagents such as silver nitrate or mercury salts as used in titrimetric analysis and a low detection limit of 4 μg L−1 chloride can be achieved if required using ion chromatography. However, one drawback to the method are the costs of consumables, notably the expensive columns and eluants required.
Meanwhile, electrochemical methods have been applied in chloride detection including potentiometry via the chloride ion-selective electrode (Cl-ISE) based on a silver chloride membrane and the coulometric chloridometer. The Cl-ISE can detect chloride in the range 10 μM to 100 μM accurately, but the addition of nitrate reagent is required to maintain stable performance in very dilute solutions.11 In the coulometric method applied to chloride its concentration is determined by monitoring the conductivity of the analyte solution whilst it is titrated with in situ generated silver ions.12 However, the equipment requires trained personnel to operate and the addition of reagents is required. Ideally a direct measure of chloride in drinking water is required which simply and exclusively only needs the introduction of a probe into the solution.
Amperometric quantification via voltammetry opens up a possible alternative inexpensive and robust, reagent free method for halides.13–16 In particular, silver electrodes are analytically useful for chloride in seawater and sweat,17,18 utilizing the voltammetric response of metallic silver oxidation in chloride media. Notice that these literature studies were done with excess supporting electrolyte or high ionic strength media (‘full support’ conditions) in contrast to the low ionic strength solutions found in drinking water (‘low support’ conditions). The challenge in low support conditions is the distortion of voltammetric signals due to significant attraction/repulsion from the charged electrode surface (migration effect).19 Recently, the feasibility of making quantitative electrochemical reactions in aqueous media without supporting electrolyte has been demonstrated using microelectrodes by Li et al.19 with ultra-low conductivity H2O (60 nS cm−1). Silver deposition and stripping voltammetry has also been studied in low support aqueous solutions on gold microelectrodes,20,21 suggesting that the detection of chloride with silver electrodes in low support aqueous systems may be possible. This paper explores the possibility of simple voltammetric analysis directly in bottled mineral and tap water without the addition of any reagents, including the supporting electrolyte commonly added to such experiments.
Herein we present a chloride sensor for bespoke use in drinking water employing a silver macro electrode and utilizing the formation and stripping of silver chloride (eqn (1)) without any added supporting electrolyte or the need for microelectrode instrumentation. In addition, to avoid the need for the cleaning of the electrode the reduction of oxygen naturally present in the water via air saturation (eqn (2) and (3)) is used as an ‘in situ’ activation process18 to activate the electrode surface in a reproducible manner prior to the analysis.22
AgCl(s) + e− ⇌ Ag + Cl−; E° = 0.2223V | (1) |
O2 + e− ⇌ O2˙−; E° = −0.284 V | (2) |
O2˙− + H2O + e− ⇌ HO2− + OH−; E° = −0.0649 V (in alkaline solutions) | (3) |
Generically the work suggests the feasibility of electro-analysis beyond the constraints imposed by the usual demands of operating with a fully supported solution phase together with rigorous outgassing of the solution which, whilst essential for quantitative physical electrochemistry is seen to be unnecessary in the context of at least some bespoke electroanalytical contexts even when using macro-electrodes.
A glassy carbon macro disc electrode (GC, radius 1.49 mm, BAS, Technical, UK) or a silver macro-disc electrode (Ag, homemade, radius 1.13 mm) calibrated as reported in previous papers18,23 were used as working electrodes. Both electrodes were polished using a sequence of 1.0, 0.3 and 0.05 μm alumina lapping compounds (Bucher, Germany).
Second, the detection of chloride using a silver electrode was undertaken in pure potassium chloride solutions and Tesco minerial water of chloride concentrations in the range 0.4 mM to 3.2 mM comparable to that in found in typical mineral waters (see Table 1). Linear correlations between the voltammetric peak current of silver chloride formation and chloride concentration were obtained.
Third, detections of chloride concentration in Tesco brand mineral water with varying added chloride concentrations were performed using a silver nanoparticles modified glassy carbon electrode via stripping voltammetry and the results compared with those from the silver macro disc electrode.
Finally, we performed chloride detections of two real samples using the calibration curve obtained in Tesco water with the silver macro disc electrode to validate the analytical procedure for the measurement of the chloride content of authentic samples of mineral waters under self-support conditions and without the addition of electrolyte or other reagents.
Five peaks can be observed in Fig. 1, as labelled schematically in the inset, depending on the concentration of KNO3 in the solution. First, the large diffusional peak 1 in the anodic scan corresponds to the two-electron transfer reduction of oxygen24 (eqn (2) and (3)). Peak 1 was observed at −0.84 V vs. Ag/AgNO3 under ‘self-support’, which gradually shifted to a more positive potential at −0.68 V with its peak height more than doubled when 4.0 mM KNO3 was added. The oxygen reduction on silver with 0.1 M NaClO4 at 0.01 V s−1 was reported at −0.3 V vs. SCE. by Neumann et al.24 which is equivalent to −0.74 V vs. Ag/AgNO3. The distortion of the ORR peak under the ‘low support’ condition as compared to that in the ‘full support’ case is expected and the low conductivity of the medium and additionally may reflect the sensitivity of the formation of peroxide to the more extended and dilute double layer.
In peak 2, silver was oxidized to form silver chloride with a corresponding stripping feature (peak 5) seen on the reverse, cathodic scan. Significantly broadened peaks were observed under ‘self-support’ conditions regarding silver chloride formation and stripping at −0.04 V and −0.45 V respectively. In the KCl solution with 4.0 mM KNO3, peak 2 and peak 5 shifted to −0.13 V and −0.30 V respectively and both of their shapes narrowed. The peak potentials are consistent with the values reported by De Mele et al.25 in 0.09 M NaCl, 0.91 M NaClO4 solution at 0.02 V s−1. The formation and stripping of silver chloride on a silver electrode were seen at ca. 0.37 V and ca. 0.15 V vs. NHE respectively, which are −0.11 V and −0.33 V vs. the Ag/Ag+ reference electrode used in the present study.
Peak 3 and peak 4 were only clearly observed in the potential range studied in ‘low support’ experiments under conditions where the local chloride ions were significantly depleted via conversion into AgCl in peak 2. This suggests that peak 3 represents the onset of the silver oxidation signal leading to Ag+ and peak 4 represents the Ag+ reduction (eqn (4)). The onset of the silver oxidation is consistent with the formal potential of the Ag/Ag+ couple estimated to be 0.12 V vs. the reference electrode used in this study. Comparison of the voltammograms shown in Fig. 1 in the absence of added electrolyte and those with low levels of support indicates a more sustained silver oxidation as the level of electrolyte support increases. Notice that when the concentration of KNO3 increases from 0 to 4.0 mM, the continuous shift in peak 3 causes difficulties in the signal subtraction from the AgCl formation process, thus the stripping peak of AgCl is used in chloride determinations as reported below.22
Ag+(aq) + e− ⇌ Ag; E° = 0.8 V vs. NHE | (4) |
Next, we question whether we can realistically expect to be able to apply the silver chloride stripping peak current under a certain ‘support level’ to quantify chloride ions in common drinking waters with varying ionic strength. The compositions of typical commercially available mineral waters were used to generate the conductivity data shown in Table 2. Estimations were made via the labelled compositions and the literature values for the equivalent molar conductivities.26 Overall support levels in each sample are listed along with the equivalent potassium nitrate concentrations with the same conductivity. Most branded mineral waters have ionic strengths corresponding to ‘low support’ conditions and can be approximated as solutions with concentrations in the range of 1 mM to 4 mM KNO3. As shown in Fig. 1, the AgCl stripping peak current is almost unchanged from solutions with 1 mM to 4 mM KNO3, suggesting, on the basis of Table 2, similar voltammetric parameters can permit direct measurements in most drinking waters, though dilutions to reach this range are required in (surprisingly) high salt mineral waters such as Borjomi and Varska waters.
Conductivity/μS cm−1 | Equivalent KNO3 concentration/mM | |
---|---|---|
a Conductivity values estimated from compositions labelled on the mineral water package via equation ![]() ![]() |
||
Mineral waters | ||
Highland Spring | 313a (33627) | 2.16 (2.32) |
Evian | 442a (600) | 3.05 (4.14) |
Tesco | 159a | 1.10 |
Aqua Pura | 218a | 1.50 |
Volvic | 218a (220) | 1.50 (1.52) |
Nestle Pure Life | 437a (160–960) | 3.01 (1.10–6.63) |
Buxton | 582a | 4.02 |
Natural mineral waters | ||
Voluvesi | 5961 | 4.11 |
Saaremaa Vesi | 5881 | 4.06 |
Saku | 1751 | 1.21 |
Borjomi Sparkling | 4114a | 52.27 |
Varska | 21301 | 14.70 |
Varska original | 43701 | 30.16 |
The electroanalytical responses of the silver/silver chloride stripping signal measured at a scan rate of 0.01 V s−1 were explored in air-saturated pure KCl solutions and Tesco–KCl solutions with chloride concentration ranges from 0.4 mM to 3.2 mM. Fig. 2 shows the voltammograms which consist of oxygen reduction (to activate the electrode and ensure reproducibility of the analytical signal) and silver chloride deposition/reduction, scanned from −0.50 V to −0.90 V then scanned positively to −0.05 V and finally swept back to −0.50 V. Three peaks (peak 1, peak 2 and peak 5 as indicated in Fig. 2) were observed for each scan in all the concentration range studied.
First, the ORR signals (peak 1) in pure KCl solutions were compared to those in Tesco–KCl solutions. In the former medium in the concentration range 0.4 mM to 0.8 mM KCl solution, the peroxide feature shifted from −0.85 V to ca. −0.75 V and the peak current increased from 3 μA to 4 μA, reflecting the increased ionic concentration. In contrast, the ORR signal in Tesco–KCl water was fixed at ca. −0.70 V plausibly due to the higher conductivity in the mineral water. As the ORR signal reflects more the ionic strength instead of the real chloride levels in the solution, it has little analytical value in the present study beyond and very importantly creating a reproducible surface for silver chloride formation and stripping.
Next, silver chloride formation signals (peak 2) were analysed. In pure KCl solutions due to the extreme low ionic strength, the peaks were broad and the current steadily increased within the potential range studies creating a voltammetric ‘loop’ to be was observed with the highest current at −0.05 V. In the Tesco–KCl solutions, clear silver chloride formation peaks were observed at −0.1 V. In both cases the formation peak currents increased as the chloride concentration increased.
On the cathodic scan, silver chloride stripping peaks (peak 5) were observed during the reverse scans for both pure KCl and Tesco–KCl solutions. As more silver chloride formed on the electrode surface, the peak shifted to more negative potentials and its height increased. Linear relationships were observed between AgCl stripping peak currents and chloride concentrations (Fig. 2c) for both solutions. The correlation ranges were 0.4 mM to 2.4 mM and 0.4 mM to 3.2 mM for pure KCl solutions and Tesco–KCl solutions respectively with the different intercepts reflecting the different amounts of AgCl formation within the potential window. The extreme data point of 3.2 mM KCl solution deviated from linearity likely due to the increased ionic strength. Most importantly, the calibration curve slope in pure KCl solutions is 8.99 μA mM−1, almost identical to that in Tesco–KCl solutions (8.76 μA mM−1), indicating that the detection sensitivity is not affected by mineral water interferences, validating the use of Tesco mineral water as the solvent in the following sections.
Two broad peaks were observed in the oxidative scan in chloride solutions. As shown in Fig. 3, peak a was observed at −0.10 V, partially overlapping with peak b at a higher potential between 0.05 V and 0.25 V. With a constant amount of AgNPs modified on the electrode, peak a increased in magnitude as the chloride concentration increased in contrast to peak b which current was gradually decreased and potential shifted to a lower value. A further experiment was performed in 1 mM KNO3 to exclude the presence of chloride and which has a conductivity identical to that of the Tesco mineral water (see Table 2). As only peak b was observed in the blank data, we infer peak a correlates to the formation of AgCl whilst peak b is associated with the oxidation of the AgNPs to Ag+. Similar observations were reported by Toh et al.17 who performed the stripping of AgNPs modified GC in ‘full support’ conditions using 0.1 M NaNO3 solution with an addition of 2 mM to 40 mM KCl content. Two broad peaks were seen each at −0.2 V (peak a) and 0.2 V (peak b) vs. Ag/AgNO3 for the silver chloride formation and the silver cations formation respectively. Compared to results reported in ‘full support’ conditions, both peaks shifted to more positive potentials in this study probably due to the low ionic strength.
A linear correlation was observed between the current of peak a and the concentration of chloride ions with a slope of 8.57 μA mM−1, which is comparable to that observed with the silver macro disc electrode (8.76 μA mM−1). With sufficient silver deposited (27.8 nmol of Ag) to ensure the sensitivity with AgNPs,17 the similarity in slopes indicates that the reaction of AgCl formation is chloride concentration limited, and the use of a silver macro disc electrode can achieve the same sensitivity as with AgNPs; the analytical signal is controlled in both cases by the diffusion of the chloride to the geometric area of the electrode, either pure silver or the silver nanoparticle modified glassy carbon electrode. Accordingly, the use of the silver electrode was adopted preferentially for reasons of simplicity and speed. In this manner chloride concentrations in real samples were estimated on a bulk silver electrode using the calibration curve evaluated from Tesco–KCl solutions as reported in the next section.
A cyclic voltammogram was recorded from −0.50 V to −0.90 V before sweeping positively to −0.05 V and returned to −0.50 V at a scan rate of 0.01 V s−1. Each experiment was repeated three times. Fig. 4 shows the voltammograms recorded from the mineral water samples. Three peaks were observed. First, the ORR feature (peak 1) was seen at −0.75 V and −0.70 V for sample 1 and 2 respectively, reflecting the conductivity increased from 218 μS cm−1 to 411 μS cm−1 (recall Table 2). Second, peak 2 was observed at −0.12 V and −0.15 V corresponding to the silver chloride formation in sample 1 and sample 2, followed by the generation of Ag+ at −0.05 V. Finally, silver chloride was removed in the reversed scan, for each solution, a shoulder was observed at ca. −0.32 V before the main peak appeared at ca. −0.36 V. The shoulder in silver chloride stripping peak has been reported,18,28,29 possibly reflecting instantaneous AgCl nucleation or silver dissolution.
By measuring the current in peak 3, we estimated the chloride concentration in samples. The experimentally measured values were 0.44 ± 0.01 mM and 1.10 ± 0.02 mM for Volvic water and 1/10 Borjomi mineral water respectively and the corresponding manufacturer reported chloride concentrations were 0.42 mM (Volvic) and 1.07 mM (Borjomi). The inset of Fig. 4 displays the samples at the expected chloride concentration, showing the measured chloride concentration in examined samples fitted well to the expected values. In this way, the developed method was validated in and applied to authentic mineral waters.
This journal is © The Royal Society of Chemistry 2021 |