Enhancing Sensitivity of Manganese Detection in Drinking Water Using Nanomaterial AuNPs/GP

Manganese (Mn) was previously considered a mere aesthetic concern that causes colored water and stained surfaces; however, recent epidemiological research found that excessive exposure to Mn has neurotoxic effects on humans, especially in children. In response to the health concerns, Health Canada and the World Health Organization moved towards stricter standards on Mn to protect public health. Currently, the standard analytical methods for Mn2+ are spectroscopic. Although they are highly sensitive, they are not cost effective or portable for high frequency analysis in the field. In this article, the sensitivity of electrochemical techniques, chronoamperometry (CA) and cathodic stripping voltammetry (CSV), are compared as well as the sensitivity of a non-modified glassy carbon screen-printed electrode (GCE SPE) vs a gold nanoparticle modified graphene (AuNPs/GP) coated GCE SPE for Mn2+ detection and quantification. Regarding the coating of the GCE SPE, detection performed with AuNPs/GP modified GCE SPE shows a wider linear range from 0–520 μM and an improved LOD of 0.75 μM. Application of the sensors was tested using drinking water samples returning high recovery rates from 92.9 to 106.8% depending on material and method used for Mn2+ detection and quantification.

Manganese (Mn) is an essential nutrient for plants, animals, and humans, necessary for enzyme activation, maintenance of immune and nerve function, and blood sugar regulation. Previously, the primary concern associated with Mn in drinking water was its aesthetic impact, such as discoloration and staining of laundry. [1][2][3][4] However, in response to the recent epidemiological evidence associating excess Mn to neurotoxicological effects in children, the concern has shifted from mere aesthetic issue to a public health concern. 4,5 New health-based guideline values were created as maximum acceptable concentration (MAC) of Mn in drinking water at 120 μg l −1 (2.18 μM) by Health Canada and 80 μg l −1 (1.4 μM) by the World Health Organization (WHO). 6,7 To ensure the management of Mn in drinking water, it is essential to utilize reliable and sufficient analytical tools for monitoring Mn levels. 6,7 Currently, laboratory-based spectroscopic and mass spectrometry methods (such as inductively coupled plasma mass spectrometry, ICP-MS) are used to measure Mn in source water, during water treatment processes, as well as in the drinking water distribution system. 8 These laboratory-based methods exhibit high sensitivity and offer many advantages, including the ability to analyze Mn in various species and sample matrices. 9,10 However, where laboratorybased methods fall short is their lack of portability, cost-effectiveness, and necessity of trained personnel. 11 Consequently, the samples require transportation to a laboratory for analysis using bench-sized spectroscopic instruments. Turnaround times for sample analysis can vary, with lab results returned to the user after several hours or days, which is limiting for timely Mn management since the nature of Mn release event is sporadic and difficult to predict. 6,7 Electrochemical field-testing methods can overcome the drawbacks of laboratory methods as they are more cost effective and portable than spectroscopic methods. 12,13 There are a variety of electrochemical techniques that can be used for Mn 2+ detection, herein cathodic stripping voltammetry (CSV) and chronoamperometry (CA) were investigated to compare their sensitivity for Mn 2+ detection. CSV is an electrochemical technique that can be separated into two steps: pre-concentration by CA and stripping by linear sweep voltammetry. 14 When a constant positive potential is applied in a Mn 2+ solution by CA, insoluble MnO 2 forms at the electrode surface constituting the pre-concentration step. Immediately after, a negative linear sweep is applied reducing MnO 2 back to Mn 2+ . The reduction occurring during the stripping step produces a voltammetric peak as the analytical signal representative of the Mn 2+ in solution. CSV has been studied intensively for Mn 2+ detection, due to its resilience to interference; however, CA at a positive potential also shows promise in terms of interference resistance, as it accumulates analyte deposition to amplify the detection signal, similar to CSV. 15 In our previous work, CA was integrated with quartz crystal microbalance as a detection technique for Mn 2+ quantification and found to be more resistant to matrix interference (except for Fe 2+ ) than other conventional voltammetric techniques. 16 To overcome the peak suppression due to Fe 2+ interference, we developed a pattern recognition approach using multiplex CA analysis. 17 Other than Mn 2+ detection, CA also has a long history of combining high-performance liquid chromatography as an electrochemical detector to analyze oxidation-reduction (redox) active compounds. 18,19 CA has shown great promise to detect analytes qualitatively or quantitatively, thus the performance of CA for Mn 2+ detection is worth comparing with an established electrochemical technique (i.e., CSV).
An alternative approach to improve electrochemical sensing is to enhance sensitivity. Nanomaterials are materials possessing at least one dimension around 1-100 nm. 20 Nanostructured electrodes utilizing nanomaterials offer a promising approach to enhance electrochemical sensing by significantly boosting sensitivity through their unique properties and increased surface area, thereby enabling more efficient detection of analytes. 20 A screen-printed electrode (SPE) integrates electrodes in a small paper or ceramic chip, making it suitable for microvolume measurement and easy to manufacture into a portable device. Rocha et al. modified a SPE surface with nanographite in effort to detect Mn 2+ . 21 Their results showed that the modified SPE had an improved sensitivity and wider linear range compared to the unmodified SPE, thereby indicating the great promise that carbon nanomaterials (such as graphene) have for electrochemical sensing of Mn 2+ . 21,22 In addition to carbon nanomaterials, Liu et al., reviewed the latest advancements of SPEs z E-mail: sarahjane.payne@queensu.ca; zhe.she@queensu.ca modified with nanomaterials and highlighted Au-modified sensors due to their electric conductivity, good biocompatibility and uniform particle size. 23 In this study, an AuNPs modified graphene (AuNPs/GP) material was synthesized by an electrochemical method and was used for detecting Mn 2+ in drinking water. Current literature employs the use of nanomaterial modified electrodes for heavy metal detection; however, comparative results of graphene modified glassy carbon (GCE) SPE and bare GCE SPE for Mn 2+ detection are limited. Additionally, the difference between two electrochemical techniques, CSV and CA, is worth exploring. The combination of a graphene layer with an additional layer of AuNP through electrochemical gold deposition to increase detection sensitivity is a novel spin on an established method. We anticipate that nanomaterial modified SPE with the right selection of electrochemical technique will be capable of monitoring Mn 2+ in drinking water systems.
Nanostructured electrode fabrication process.-The GCE SPEs purchased from Metrohm DropSens (Mississauga, Canada) were used as the substrate for making the AuNPs/GP sensor. Fabrication was performed in two steps: drop casting of the graphene solution followed by electrochemical deposition of AuNPs. Figure 1 shows the fabrication procedure for the modified sensor where the graphene solution was prepared by dispersing 10 mg of graphene nanoplatelets into 1 ml of DMF and 25 μl of Nafion TM , and the resulting solution was then sonicated for 20 min. About 10 μl of the graphene solution was drop cast onto the working electrode of an GCE SPE using a micropipette. The coated GCE SPE was left to air dry in the fume hood for 2 h.
The deposition of AuNP was performed using a CHI 420C potentiostat (CH Instruments, Texas, US). The SPE was mounted in an electrochemical quartz crystal microbalance (EQCM) cell as seen in Fig. S1, which was purchased from CH Instruments (Texas, US).
An Ag/AgCl in 3 M KCl reference electrode was purchased from CH Instruments (Texas, US) with an in-house prepared gold counter electrode to complete the cell for deposition, allowing the electroplating of gold on the SPE without consuming Au 3+ in the solution. The deposition was performed by CA at -0.2 V for 3 min in 3.0 mM HAuCl 4 /0.5 M H 3 PO 4 electrolyte solution.
Electrochemical measurement.-Electrochemical experiments were performed and analyzed by the same potentiostat used for the gold deposition. The general set-up employed in the electrochemical measurements consisted of a SPE as the working electrode, a platinum wire counter electrode, and an Ag/AgCl reference electrode in 3 M KCl. The working and counter electrode cell was connected to the reference electrode cell via a salt bridge. CVs were taken for each material, bare GCE SPE, graphene coated GCE SPE, and AuNPs/GP coated GCE SPE in a ferri/ferrocyanide solution (5 mM ferricyanide, 5 mM ferrocyanide and 1 M NaClO 4 ) with a scan rate of 0.1 V s −1 .
Electrochemical impedance spectroscopy (EIS) was performed within a frequency range of 100 kHz to 0.1 Hz with an amplitude of 5 mV. The measurement was performed in the same ferri/ferrocyanide solution as CV. The open circuit potential was employed as the detection potential. The acquired Nyquist curves were subsequently simulated using ZSimpWin V3.60 software by AMTEK (Michigan, USA).
CSV was carried out by applying a potential of 1.2 V for 3 min to pre-concentrate Mn on the SPE, followed by stripping by ramping the potential from 1.2 V to −0.3 V at a rate of -2 V s −1 . The reactions occurring during the measurements are demonstrated in Fig. 2. Measurement of the Mn 2+ samples at increasing concentrations (0-1 mM) were performed in triplicate. CSV peaks were plotted as the absolute value of current.
Sample preparation for the interference study.-A 6 μM Mn 2+ solution was prepared in Milli-Q water and used to prepare 5 individual samples with different interfering metals by serial dilution. Each sample employed in the interference study was prepared to contain a concentration of 600 μM for one of the interfering reagents: K + , Cd 2+ , Cu 2+ , Fe 3+ , and Fe 2+ .
Preparation of a drinking water sample.-One drinking water sample was obtained from the lab tap (Kingston, ON). The drinking water sample was left to sit overnight, opened to the atmosphere to remove chorine residues. The pH of the sample was adjusted to pH 6 by 20% nitrate acid to be consistent with the pH of Milli-Q water. Mn 2+ was spiked into the pre-treated drinking water sample in the form of MnSO 4 and was fully dissolved before serial dilution. Recovery test was performed for Mn 2+ concentrations of 4 and 6 μM.
where the standard deviation (σ) was calculated from the 7 repeated measurements and the slope was taken from the calibration curve of the method.
Scanning electron microscopy (SEM).-The SEM images of the materials were taken with a Themo Fisher Scientific Quanta 250 eSEM with a 20 kV accelerating voltage and 0.9 Torr water vapor chamber pressure. Energy dispersive X-ray spectroscopy (EDS) was performed with an EDAX Element detector. The bare GCE SPE, graphene coated GCE SPE, AuNPs/GP coated GCE SPE and AuNPs/ GP coated GCE SPE after MnO 2 deposition were analyzed by EDS without coating. When taking SEM images, only graphene coated GCE SPE were coated with Pt to improve contrast. All the other samples were imaged without coating. The Pt coating was performed on a Leica ACE 600 sputter coater. The MnO 2 deposition was performed in 1 mM Mn 2+ Milli-Q solution by 1.2 V CA for 10 min.
Atomic force microscopy (AFM).-The same SPE samples from SEM without coating (bare GCE SPE, graphene coated SPE, and AuNPs/GP coated SPE before and after MnO 2 deposition) were further characterized by atomic force microscopy (AFM). AFM images were obtained with a Dimension Icon+ atomic force microscope (Bruker, Germany) working in air in contact mode. Scanasyst tips made from silicon nitride cantilevers (Bruker, Germany) with a spring constant of 0.4 Nm −1 were used for all the measurements. AFM images were analyzed using Gwyddion 2.59 software. For roughness analysis of the AFM topography images (5 μm 2 × 5 μm 2 ) were line corrected by matching to the height polynomials, and horizontal scars were removed. The root mean square (RMS) values of the surface roughness were measured by performing Gwyddion statistical analysis.

Results and Discussion
CV and EIS measurements of AuNPs/GP.-CVs were taken for each individual material (bare, graphene, AuNPs/GP) in a ferri/ ferrocyanide solution to compare the electron transfer rates across materials. A higher and sharper peak is preferred, as it indicates faster electron transfer and more well-defined electrochemical reaction, allowing for enhanced sensitivity and accurate determination of peak potential in detecting and quantifying analytes. 20,23 As shown in Fig. 3a, the oxidation peak current of bare GCE SPE, graphene GCE SPE, and AuNPs/GP GCE SPE were 0.1170, 0.1709 and 0.1716 mA, respectively. The addition of the graphene improves the current density and the electron transfer rate due to the increase in surface area and more binding sites on surface. 24 Additionally, the ferrocyanide oxidation peak of AuNPs/GP was much sharper than on graphene coated and bare GCE SPEs. The AuNPs are of uniform size and distributed evenly across the graphene sheet which provides many sites for electrochemical reactions to take place, thereby increasing electron transfer reflected in the sharper peaks. 24 Figure 3b shows the Nyquist plots of bare GCE SPE, graphene coated GCE SPE and AuNPs/GP coated GCE SPE. The circuit models used to simulate the plots are shown in Fig. S2. The graphene coated and AuNPs/GP GCE SPEs had an additional pore resistance element in their circuit models due to the porous nature of the materials. The charge resistance of AuNPs/GP was 2.54 mΩ, which is smaller than graphene (15.41 mΩ). Bare GCE SPE had the largest charge resistance of 75.08 mΩ, which results in the lowest CV peak density as shown in Fig. 3.
Surface characterization by SEM.-The characterization of nanomaterials on the surface of the GCE SPE was achieved using a SEM. Figure 4a shows the unmodified smooth surface of a bare GCE SPE. The first surface modification was achieved through dropcasting graphene. Graphene nanostructures can be visualized on the SPE surface as the white cloud like structures seen in Fig. 4b. Graphene usually creates a nanosheet on electrode surfaces; however, the sheet structure of the graphene is more clearly seen in Fig. 4c after the deposition of the AuNPs. The small white dots dispersed over the sheets are electrochemically deposited AuNPs. The AuNPs exhibit a predominantly uniformed size and spherical shape, which are typical characteristics of evenly distributed AuNPs. The spherical AuNPs have a large surface area that offer active sites for electrochemical reactions affording increased detection sensitivity of the AuNPs/GP GCE SPE compared to the graphene coated GCE SPE and bare GCE SPE. The GCE SPE surface was successfully and evenly coated with the AuNPs/GP material without any major agglomeration and can be classified as a nanomaterial. After MnO 2 deposition (Fig. 4d, a second population of nanoparticles are observed on the electrode indicating that the deposited MnO 2 particles are also in nanoscale. The elemental composition of the nanomaterial was examined by EDS (See supplementary data for full elemental results table and spectra). All samples, bare GCE SPE, graphene coated GCE SPE, AuNPs/GP coated GCE SPE, and MnO 2 deposited AuNPs/GP coated GCE SPE, exhibited trace amounts of Al and Si, and small amounts of Cl due to the materials used for the fabrication of the GCE SPE which include ceramic and insulator. The GCE SPE and graphene materials are made of carbon and the EDS data also shows predominantly carbon compositions of 95.78 weight % (wt%) and 95.62 wt%, respectively. Upon addition of AuNPs to the electrode the carbon content decreases to 62.56 wt% carbon and 22.16 wt% of gold is observed. After MnO 2 deposition, 1.56 wt% of Mn was observed indicating Mn was deposited via CA.
Surface characterization by AFM.-AFM characterization of the electrode surfaces is reported in Fig. S11, with similar surface structures observed by AFM as by SEM. The average RMS of the surface roughness was obtained as 31.8 ± 7.9 nm for the GCE SPE  surface. However, when the GCE SPE surface is coated with graphene and Nafion TM , the larger grains are observed, and surface roughness increases to 179.2 ± 51.9 nm. After electrochemically depositing AuNPs on a graphene coated GCE SPE, although it is difficult to visualize individual AuNP due to the rough surface topography, the addition of AuNPs appear to contribute to the surface roughness between GCE SPE and graphene, with the average surface roughness of 87.3 ± 22.6 nm. After MnO 2 deposition, the average surface roughness further decreased to 37.0 ± 11.2 nm, due to the formation of evenly distributed MnO 2 nanoparticles on the surface.  6 The electrochemical measurements of Mn 2+ at different concentrations (0-1 mM) were conducted to determine the applicability of the two methods to detect Mn 2+ . Figure 5a shows the overlayed CSV voltammograms from 0 to 12 μM using AuNPs, where the concentration of Mn 2+ is most likely to occur data from concentrations above 12 μM are reported in Fig. S4a. Samples with a higher concentration of Mn 2+ in the solution will generate a thicker layer of MnO 2 , this in turn will produce a higher current in the CSV, as more Mn 2+ is being solubilized as the solid MnO 2 is stripped off the electrode surface. To emphasize the peak signal and remove the interference of background, background correction was conducted with the results shown in Fig. 5b. The background corrected reduction peak of Mn 2+ seen in Fig. 5b lies from 0.43 to 0.47 V, consistent with other examples in the literature. 25 In Fig. 5b, a second peak at 0.68 V was observed. The peak could be from the detection background, as a small peak was observed at 0 μM of Mn 2+ . Fortunately, the Mn 2+ peak potential is consistent at different concentrations, and the linearity of the calibration curve was not affected by the peak at 0.68 V.
Mn 2+ detection using AuNPs/GP allows better sensitivity and wider linear range compared to bare GCE SPE. Compared with bare GCE SPE, AuNPs/GP electrodes demonstrate a higher current than GCE SPE by background corrected CSV voltammogram seen in Fig. 6a. The original CSV voltammograms using bare GCE SPE are shown in Fig. S3. The calibration curves between bare GCE SPE and AuNPs/GP coated GCE SPE were compared as shown in Fig. 6b. The current passed on the AuNPs/GP electrode for the varying concentrations of Mn 2+ is higher than for the GCE SPE, indicative of a better sensitivity to Mn 2+ for the AuNPs/GP electrode.
Measurement of Mn 2+ by CA.-CSV of Mn 2+ detection follows 2 steps: preconcentration by CA and stripping by linear sweep voltammetry. In CSV analysis, the reduction peak from the stripping step is used to detect and quantify Mn 2+ . The CA data collected during preconcentration could also be used to analyze Mn 2+ . CA applies a constant potential for a pre-determined time. While applying a potential to the analyte, the current passed correlates to the amount of redox reaction happening on the surface of the electrode and should be proportional to the concentration of the analyte in solution. An optimal electrode material would accelerate this reaction, resulting in high current even when the concentration of Mn 2+ in solution is low.
The CA i-t curves (Fig. 7a) for the AuNPs/GP and for the bare GCE SPE (Fig. S5) both demonstrate the association between higher concentration of Mn 2+ in solution and higher current recorded. Integrating the current over the deposition time gives the total charge   transfer on the surface. Unlike the calibration curves obtained from CSV (Fig. 5b), both AuNPs/GP electrodes and GCE SPE (Fig. 7b) show a linear range over all the low concentration samples measured with strong linearity reflected in their R values, 0.993 and 0.999, respectively.
Comparison between techniques and materials.-The linear range and LOD were determined for both AuNPs/GP and bare GCE SPEs are shown in Table I. When using CSV as the detection technique, GCE SPE coated with AuNPs/GP offers lower LOD and wider linear range. In Canada, the maximum concentration of Mn 2+ reported (1991-2014) was 26200 μg l −1 (476 μM) in British Columbia. 6 The linear range of both materials are wide enough to cover the maximum concentration of Mn 2+ when using CSV, but only AuNPs/GP has the linear range below the health-based MAC value (2.18 μM) by Health Canada. 6 Compared to CA, CSV can be deemed the more sensitive method by 15%-30%; however, the linear range of CA was wider than CSV.
According to the literature, our methods can potentially enhance their detection sensitivity by employing longer deposition times. 21 As shown in Table I, our linear range offers the advantage of being capable of detecting Mn 2+ within the range encompassing the Health Canada MAC value and the maximum observed Mn 2+ level in field data collected from various locations across Canada. This feasibility makes our method well-suited for practical Mn 2+ detection in realworld scenarios.
Interference study with common drinking water components.-When using CSV, an oxidative potential is first applied to accumulate Mn 2+ as MnO 2 through an oxidation reaction (i.e., CA). During this process, only certain metal ions can undergo oxidation and be accumulated within the pre-determined timeframe, which makes CSV and CA more selective to those metal ions. Additionally, under the oxidative potential of CSV, the oxidation products are metal oxides (such as MnO 2 ), which avoids overlapping peaks from intermetallic species during the stripping process. 29 Berg et al. used stencil-printed carbon ink electrode for Mn 2+ detection and found that CSV is resistant to Cr 6+ , Fe 3+ , Mg 2+ , Ni 2+ , and Zn 2+ , but it was susceptible to Al 3+ , Cu 2+ , Fe 2+ at concentration ratios at or below one. 28 Rusinek et al. found that no interference was observed from the metal ions tested (Zn 2+ , Cd 2+ , Pb 2+ , In 3+ , Sb 3+ , Al 3+ , Ba 2+ , Co 2+ , Cu 2+ , Ni 3+ , Bi 3+ , and Sn 2+ ) except for Fe 2+ , when using indium tin oxide electrode. 25 Boselli et al. reported that when using platinum miniature sensor, no significant alternation to the CSV voltammetric response was observed when 100 times of Zn 2+ , Mg 2+ , Cu 2+ , Pb 2+ , and Fe 2+ were spiked into the Mn 2+ sample. 27 Those studies demonstrate that metallic electrodes appear to be more selective towards the detection of Mn 2+ . Cd 2+ , Cu 2+ , Fe 2+ , and Fe 3+ were chosen to investigate their interference to CSV and CA as common drinking water metal components that strip at similar potentials to Mn 2+ . K + was also chosen as a common ion in drinking water. Figure 8 shows the current observed by CSV in the presence of common drinking water metal cations, no significant difference was seen for the Mn 2+ signal in a 1:100 ratio with K + , Cd 2+ , Cu 2+ , Fe 3+ , except for Fe 2+ . Interference by Fe 2+ is a common issue for CSV as it suppresses the signal of Mn 2+ , causing an underestimation of Mn 2+ concentrations. 25 The peak suppression of Mn 2+ was due to the competing oxidation reaction of Fe 2+ . 17 When the Fe 2+ concentration is 100 times the concentration Mn 2+ , the Mn 2+ CSV peak current decreased by 99.3%. Mn 2+ detection using CA is also affected by the presence of Fe 2+ , due to the oxidation reaction of Fe 2+ at the oxidative potential. The oxidation reaction of Fe 2+ contributed to the overall CA signal, which results in a 17-fold increase in the CA signal when a 1:100 Mn 2+ :Fe 2+ solution was used.
Measurement of Mn 2+ in a drinking water sample.-The application to sensing in complex matrices was tested using a tap water sample. The performance of the sensor herein will be measured by the recovery rate of Mn 2+ . Overall, the sensor   performed well and provided accurate recovery rates, as shown in Table II. Between bare GCE SPE and AuNPs/GP, both materials yielded excellent average recovery rates of 106 and 101% using CSV at 4 μM Mn 2+ , respectively. The recovery rates between CSV and CA are similar as well, such as CSV had an average recovery rate of 92.9% and CA had an average of 93.8% using bare GCE SPE at 6 μM Mn 2+ . The recovery rates that were over or under 100% could be due to several possible reasons, such as random errors, the interference of the drinking water matrix and the residual Mn 2+ in the drinking water sample. The high recovery rates recorded in Table II show promising results for the future development of a portable point-of-use sensor for Mn 2+ in drinking water.

Conclusions
A comparative study was conducted to determine the suitable electrode materials and methods for Mn 2+ , in consideration with sensitivity, linear range, and signal resistance to interference reagents and drinking water matrix. The addition of AuNPs/GP improved the LOD of Mn 2+ detection and allowed a wider linear range than bare GCE SPE. Between CSV and CA, CSV can be deemed the more sensitive method for Mn 2+ detection as it yields a lower LOD; however, CA provides an improved linearity by offering a wider linear range. No significant difference was observed of Mn 2+ signal in a 1:100 ratio with K + , Cd 2+ , Cu 2+ , Fe 3+ for both CA and CSV. A drinking water sample spiked with Mn 2+ received excellent recovery rates of Mn 2+ detection using the sensor. The superior sensitivity of AuNPs/GP and CSV are highlighted within this study to detect Mn 2+ in accordance with drinking water regulations and guidelines. CA also shows promises as a Mn 2+ detection technique due to its robustness, good linearity, and simplicity to integrate with other techniques. Sensor materials and techniques should be chosen accordingly for the detection goals. The methods could be further integrated into a portable device for in-line or off-line drinking water analysis for point-of-use sample analysis. The development of CSV or CA with AuNPs/GP are expected to enable faster, more reliable, and user-friendly devices that allow more frequent Mn monitoring to provide early-warning on Mn release events and offer a tool to help the utilities to create a more comprehensive water treatment strategy.