Coal Fly Ash Modified Graphite-Polyurethane Composite Electrodes for the Electrochemical Determination of Cadmium(II) in Batteries and Water

Abstract Coal fly ash (FA), an aluminum silicate by-product and environmental pollutant which is generated during the combustion of coal in coal-fired power stations, was used as an electrode modifier for the determination of Cd(II) in aqueous solutions. In this work, graphite/polyurethane-based composites containing different amounts of FA were prepared and characterized by scanning electron microscopy (SEM), energy dispersive X-ray spectroscopy (EDX) thermogravimetry (TGA), and derivative thermogravimetry (DTG). The graphite/polyurethane composite electrodes (GPUE), with and without FA modifier, were evaluated with regard to their performance as voltammetric electrodes in the determination of metallic cations, using Cd(II) as a probe. After optimizing solution and instrumental parameters affecting the peak current, a differential pulse anode stripping voltammetry (DPASV) procedure was developed for GPUE modified with 5% FA (m/m), resulting in a linear response for Cd(II) from 2.0 x 10−7 to 1.0 x 10−6 mol L−1 with a detection limit (LOD) of 6.6 x 10−8 mol L−1. Cd(II) was added to natural water samples and determined at the 10−7 mol L−1 level with a mean recovery of 99%. It was also extracted from exhausted rechargeable Ni-Cd batteries and diluted to ca. 0.2 µmol L−1 and determined with the same electrode, with recoveries of 98.7% when compared to FAAS. These results serve as a proof of concept that FA is a useful electrode modifier.


Introduction
Coal fly ash (FA) is a by-product generated during the combustion of pulverized coal in thermoelectric power plants.It originates from the lighter particles that rise with the flue gases during combustion and is collected using electrostatic precipitators or bag filtering systems (van der Merwe et al. 2014).As a residue with high pollutant potential, with millions of tonnes being produced annually (Ge, Yoon, and Choi 2018), alternatives for the use of FA have been sought as a way to remove it from the environment.The most common include its incorporation into cement, concrete and bricks, i.e., in civil construction (Wu, Chi, and Huang 2014;Woszuk, Bandura, and Franus 2019;Zhang et al. 2020).Other uses include improvement of soils for agriculture (Yao et al. 2015), synthesis of zeolites (Ameh et al. 2016), and geopolymers (Blissett and Rowson 2012) and the extraction of valuable elements such as aluminum (van der Merwe et al. 2017).
FA particles present interesting physical features such as a spherical morphology, relatively low particle size distribution, smooth surface topography and high thermal stability, leading to a pursuit of highly valuable applications, e.g., its potential application as polymer filler (van der Merwe, Mathebula, and Prinsloo 2014), as an absorbent for air and water pollutants (Ge, Yoon, and Choi 2018) and extraction of valuable metals (Vilakazi et al. 2022).The only reference found in literature regarding the electrochemical application of FA relates to the use of FA as binder in supercapacitor electrodes, based on activated carbon, for the determination of specific capacitance (Martinovic et al. 2017).According to our knowledge, the use of FA as electrode modifier intended for electroanalysis, as done with other silicates (Cesarino et al. 2007), has not been reported.
The graphite-polyurethane composite electrode (GPUE) used in this work offers advantages such as high hydrophobicity, resistance to non-aqueous solvents and a wide useful potential window in both acidic and basic media.In addition, it is easy to incorporate different types of modifiers since it is prepared from a liquid polymer (Mattioli, Cervini, and Cavalheiro 2020).Our group has published several papers using this electrode, without (Mendes, Calero-Neto, and Cavalheiro 2002;Cervini and Cavalheiro 2008;Calixto, Cervini, and Cavalheiro 2012) and with modification (Santos and Cavalheiro 2014;Baccarin, Cervini, and Cavalheiro 2018), but this is the first attempt of using FA as a modifier.
In industry, cadmium is used in conjunction with nickel in rechargeable batteries, in pigments, deposited on various metals due to its resistance to corrosion, in alloys with other metals, and as impurity in the extraction of zinc (Pacer, Ellis, and Peng 1999).There are many papers describing the use of different electrodes for the determination of Cd(II) in water (Hu et al. 2019;Hassanpoor and Rouhi 2021), the simultaneous determination of Cd(II) and Pb(II) in water (Oularbi, Turmine, and el Rhazi 2019;van Staden and Arnold Tatu 2019;Guenang et al. 2020), herbal food supplements (Palisoc, Vitto, and Natividad 2019), decorative materials (Hou et al. 2020), hair (Manihandan and Narayanan 2019), bivalve molluscs (Pizarro et al. 2019) and paint (Cui and Li 2019), among others.Other examples describe the simultaneous determination of Cd(II), Pb(II) and Cu(II) using modified glassy carbon in water (Hassan, Elhaddad, and AbdelAzzem 2019) and parts of cigarettes (Walcarius 1999).Only one reference was found concerning the voltammetric determination of Cd in batteries (Aglan, Hamed, and Saleh 2019).In that paper, Aglan, Hamed, and Saleh (2019) used a carbon paste electrode modified with lanthanum tungstate ion exchanger to determine Cd, reaching a limit of detection of 8.0 Â 10 À8 mol L À1 .
It is well known that silicas can be used to improve electrode sensitivity in electroanalytical procedures by preconcentrating cationic species (Walcarius 1999).Our group successfully used modified amorphous SBA-15 silica as a graphite polyurethane composite electrode modifier in the determination of several cationic species in water and ethanol media (Mendes, Claro-Neto, and Cavalheiro 2002;Cesarino et al. 2007).
Considering that FA consists mostly of aluminum silicates, this work proposes the use of this material to act as electrode modifier by pre-concentration of cationic species.
The main objective of this work was to demonstrate the possibility of using FA as an electrode modifier and its features for metal ion analysis.Therefore, it was necessary prepare, characterize, and apply a graphite-polyurethane composite electrode modified with FA (GPUE-FA) in the determination of a metallic cation, as a "proof of concept" of its viability.In this case Cd(II) in Ni-Cd rechargeable batteries and natural water samples were used as a probe.Up to our knowledge this is the first attempt of applying FA as an electrode modifier as alternative for the reuse of this relevant environmental pollutant.Due to its remarkable environmental importance and toxicity, in addition to its wide industrial use and well-known voltammetric behavior, Cd (II) was chosen as the electroanalytical probe.

Reagents and materials
A commercial-grade FA sample was obtained from Ash Resources, South Africa.This FA sample is marketed as an ultrafine spherical alumina-silicate polymer filler with a low carbon content and is specified to have a mean particle size of 4.5 lm, with more than 99% of the volume distribution of its particles having a diameter smaller than 25 lm.Detailed bulk and surface characterization of this sample was reported elsewhere (van der Merwe et al. 2014).Briefly, it consisted of an amorphous alumina-silicate glass phase (62.1%), which co-existed with two primary crystalline phases, mullite (Al 6 Si 2 O 13 ; 31.8%) and quartz (SiO 2 ; 6.1%).XRF analysis indicated the presence of six major chemical constituents (SiO 2 , 49.3%; Al 2 O 3 , 34.0%; Fe 2 O 3 , 5.8%; CaO, 5.1%; TiO 2 , 2.0% and MgO, 1.0%) and a low loss on ignition (LOI <0.6%).
All chemicals used in this work were of analytical grade and was used without further purification.Graphite powder (particle size < 20 lm), methylene diphenyl isocyanate (MDI) and castor oil (both from Univar, Brazil).
All solutions were prepared using water treated in a reverse osmosis system (Gehaka OS10 LZ, Brazil) and then purified in a Barnstead D13321 EasyPure RoDi system (Thermo Scientific), with resistivity !18 MX cm À1 .A 1.0 Â 10 À3 mol L À1 Cd(II) standard stock solution was prepared from a 1000 mg L À1 standard solution (SpecSol).Working solutions were prepared daily by appropriate dilutions of the Cd(II) stock solution in 0.10 mol L À1 KCl at different pHs, adjusted with HCl (Mattioli, Cervini, and Cavalheiro 2020).

Instrumentation
Electrochemical data were obtained using an Autolab PGSTAT 204 potentiostat/galvanostat, equipped with a FRA32 module for electrochemical impedance spectroscopy (EIS) measurements, controlled by NOVA software (version 2.1.1)all from Metrohm.
Electrochemical measurements were performed in a three-electrode cell.The counter electrode was a platinum wire (1.0 cm length and diameter of 1.0 mm), and a saturated calomel electrode (SCE, Hg/Hg 2 Cl 2 ) was used as the reference electrode.GPUE and GPUE-FA were used as the working electrodes, both with a geometric area of 0.070 cm 2 (diameter of 3.0 mm).
A Zeiss model LEO (440 kV), with 63 kV resolution was used to collect scanning electron microscopy (SEM) -energy dispersive spectroscopy (SEM-EDS) data at room temperature.SEM images were collected to evaluate the mean size and distribution of FA on the electrode surface of GPUE-FA while EDS was performed to evaluate the compositions of FA and GPUE-FA, respectively.The samples were sputter-coated with gold in a Bal-Tec Med 020 high vacuum coating system.Thermogravimetric (TG) and derivative thermogravimetric (DTG) curves were obtained in an SDT-Q600 module controlled by Universal Analysis software (both from TA Instruments), using sample masses of 6.5 ± 0.1 mg in open a-alumina sample holders at atmospheric pressure.The heating rate was 10 C min À1 under N 2 until 600 C and under dynamic air atmosphere to 1000 C. The gas flow was 100 mL min À1 in both cases.
Flame atomic absorption spectrometric (FAAS) analysis were performed in a PINAACLE 900 T (Perkin Elmer) to determine the Cd(II) concentration in the water and battery samples, to compare the results with those obtained using the GPUE-FA electrode.The concentrations of concomitants present in the natural water samples were also estimated by FAAS.
Wettability of the electrodes were evaluated by measuring the contact angles in a C201 optical tensiometer (Attension) equipped with a Navitar digital camera (Navitar) and controlled by One Attension software (Attension).A drop of water was added to the surface of the composites and 50 optical scans were performed to verify the contact angle formed between the solution and the surface.The left and right contact angles of water with the composite surfaces were determined and the average was calculated.
Preparation of the graphite-polyurethane composite electrode (GPUE) and the graphite-polyurethane composite electrode modified with FA (GPUE-FA) Preparation and best composition of the unmodified graphite-polyurethane composite (GPU) and its electrode was previously established by Mendes, Claro-Neto, and Cavalheiro (2002).Briefly, a 2.85:1.1:0.8 graphite, MDI and castor oil mixture was homogenized for 5 min in a glass mortar, pressed in a manual press, extruded as 3.0 mm diameter rods, and allowed to cure for 24 h at room temperature, after which the rods were cut into 1.0 cm sections.The rods were then connected to copper wires using a conductive silver epoxy kit (Electron Microscopy Sciences).After 24 h, the composite/copper wire assembly was inserted into a glass tube (6.0 mm i.d., 8.5 cm length), sealed with SQ 3024 epoxy resin (Silaex) and allowed to cure for 24 h.Mechanical polishing with 2000 grit sandpaper was performed in an APL 02 motorized polisher (Arotec) to remove excess of epoxy resin from the surface and expose the composite.The electrode was sonicated in isopropanol for 5 min and then in water for additional 5 min before each working day.The GPU composites modified with FA (GPU-FA) were prepared using the compositions described in Table 1.The same preparation, assembly and activation process used for GPUE was adopted for the FA modified electrodes (GPUE-FA).

Procedures
Solution and instrumental parameters affecting the voltammetric peak current were optimized for GPUE-FA 5% at room temperature before a differential pulse anode stripping voltammetry (DPASV) procedure was developed.Factors that could potentially influence the analytical signal were the following: modifier loading, electrolyte medium, pH, pulse amplitude (a), scan rate (), accumulation time (t acc ) and accumulation potential (E acc ).
EIS measurements were performed using 5.0 Â 10 À3 mol L À1 potassium ferricyanide in 1.0 mol L À1 KCl at pH 3 by varying the frequency from 50000 to 1 Hz and applying 0.22 V (vs.SCE).
Analytical curves were obtained in triplicate for Cd(II) solutions with concentrations ranging between 1.0 Â 10 À7 and 1.0 Â 10 À6 mol L À1 , using GPUE and GPUE-FA 5% under the optimized DPASV conditions (KCl 0. For the determination of Cd(II) in batteries, an exhausted Ni-Cd battery was unassembled and the different layers were carefully separated, according to the STA-eletrônica user's manual (Systems and Applied Tecnology 2023).The anode, made of Cd(OH) 2 , was chosen for analysis.The anode material was separated, 1.0 g weighed (with precision of ± 0.1 mg) and kept under reflux in 0.5 mol L À1 sulfuric acid for 30 min, according to Freitas and Rosal em (2005).After the dissolution of Cd(OH) 2 , the suspension was hot filtered on a sintered glass filter under reduced pressure.
According to the STA-eletrônica user's manual (Systems and Applied Technology 2023, the concentration of Cd(II)in the leaching solution should be ca.2.0 Â 10 À2 mol L À1 .After diluting this stock solution to 2.0 Â 10 À4 mol L À1 , 10 mL were added to the cell containing 20.0 mL of a 0.10 mol L À1 KCl solution at pH 3, resulting in a 2.0 Â 10 À7 mol L À1 Cd 2þ estimated working concentration.After that, three 10.0 mL standard additions from the 1000 mg L À1 standard Cd(II) stock solution were performed in order to reach a 2.0 Â 10 À7 mol L À1 Cd(II) and DPASV measurements were performed in triplicate at GPUE-FA 5% after each standard addition.
Bottled mineral water was purchased at a local market.A tap water sample was collected straight from the public supply at USP Campus at São Carlos/SP, Brazil, while the artesian well water sample was collected from the Production and Distribution Center of the public water supply company of São Carlos at 22.021(W); 47.887(S).Tap and artesian well waters were collected in polypropylene vessels and kept in a refrigerator until use.All water samples were spiked with a Cd(II) standard solution to be 2.0 Â 10 À7 mol L À1 and a suitable mass of solid KCl was dissolved in each to obtain a KCl concentration of 0.10 mol L À1 , while the pH was adjusted to 3.0 with HCl.Finally, 20.0 mL aliquots were inserted in the voltammetric cell and DPASV voltammograms were taken in triplicate, under the optimized parameters.After that, three standard additions from the 1000 mg L À1 standard Cd(II) stock solution were performed in order to reach 2.0; 4.0 and 6.0 Â 10 À7 mol L À1 Cd(II), respectively and the voltammograms were recorded under the optimized conditions for each.The procedure was repeated in triplicate for each sample.

Results and discussion
Characterization of the GPU-FA composite SEM images of the fractured and polished surfaces of the GPU-FA 5% composite were obtained at different magnifications (Figure 1).FA particles with particle sizes ranging between 2 and 10 mm were widely distributed throughout the composite, without any evidence of agglomeration (Figures 1 c and 1d).The two images in Figures 1b and 1d show the same frame, presented with different contrasts to better observe the details of insertion of the FA particles into the composite matrix.Figure 1d indicated tight bonding between FA and the composite matrix.Spherical shaped cavities could also be observed (Figure 1c), probably as a result of FA spheres being removed during surface polishing.The images of the fractured surfaces (Figures 1 a and 1b) showed that the FA particles were inserted in the composite, suggesting that they were embedded in the bulk of the composite matrix.EDS analysis of GPU-FA 5% performed on the areas presented in Figure 1 and focused on the GPU matrix, i.e., the area outside the FA particles, presented 98.5% carbon and traces of iron, potassium and calcium impurities as well as small amounts of silicon and aluminum, probably from the FA modifier (Table S1, Supplementary File).This result was anticipated, since most of the composite consisted of graphite and polyurethane, which mainly consists of carbon.Gold from the sputter-coating process could also be detected.When focusing on the FA spheres of the composite surface (Table S1, Supplementary File), several other elements were detected.O, Al and Si were the major elements, while Mg, K, Ca and Fe were observed in smaller quantities.These observations agree with the previously reported composition of FA (van der Merwe et al. 2014).
The wettability of the composites was evaluated by contact angle measurements.The contact angle for the unmodified GPU composite was (77.5 ± 0.4) o , while that of GPU-FA 5% composite was (84.8 ± 0.2) o , suggesting that both composites presented a certain degree of wettability.The small difference in the results obtained for GPU and GPU-FA 5% was attributed to the fact that, due to the manual nature of the polishing process, neither electrode was perfectly flat.The reason for the relative hydrophobicity in both cases was ascribed to the hydrophobic nature of polyurethane, which comprises almost 40% of the composite's mass.
Thermogravimetric curves were obtained for the GPU-FA composites to confirm that the modifier loadings were indeed 2.5, 5 and 10% (m/m).The thermogravimetry/derivative thermogravimetry (TGA/DTG) data obtained for the GPU and GPU-FA composites is summarized in Table S2 (Supplementary File).The FA modifier consisted of inorganic oxides (mainly silicon and aluminum) and had a loss on ignition below 0.6%, indicating that the sample contained a very low percentage of moisture, sulfur, unburned carbon, carbonates, and hydroxides (van der Merwe et al. 2014).Thermal treatment of the FA sample did therefore not contribute substantially to mass loss events occurring up to 900 C, thus the observed mass loss events were ascribed to the decomposition of polyurethane and the combustion of graphite (Table S2).TGA/DTG analysis of the unmodified GPU composite resulted in a 0.53% residue at 900 C.This quantity was subtracted from the residues generated by the GPU-FA composites to obtain the net residue, which was taken as the percentage of FA modifier in the GPU-FA composites.TGA/DTG analysis confirmed that the actual loadings of FA modifier in the GPU-FA composites compared well to the expected values (Table S2).

Determining the best modifier loading
The best modifier loading for the composite electrodes, GPUE and GPUE-FA, was evaluated using cyclic voltammetry and a 1.0 Â 10 À3 mol L À1 potassium ferricyanide solution in 0.10 mol L À1 KCl at pH 3 as a probe (Figure 2a).The lowest current intensities were observed at GPUE and GPUE-FA 10%, while at GPUE-FA 2.5% and GPUE-FA 5% the current intensities were almost the same, and 11% higher than the unmodified electrode.The explanation for this observation is two-fold.FA is a non-conductive material, while graphite is often used as electrically conductive filler in composite electrodes (Mendes, Claro-Neto, and Cavalheiro 2002).When conductive graphite active sites are replaced by non-conductive FA particles on the electrode surface, the electrochemical response of the composite is reduced.On the other hand, FA can interact with the analyte cations in solution, pre-concentrating them on the electrode surface, which in turn increases the electrochemical response of the device.Incorporation of 2.5% and 5% (m/m) FA modifier in the composite increased the voltammetric response of GPUE, indicating that the pre-concentration effect is more significant than the substitution of conductive graphite on the electrode surface when modifier loadings up to 5% (m/m) are used.However, at higher FA loadings (c.a.10% (m/m)) the interaction with the metallic species is less significant than the occupation of the conductive graphite sites, resulting in a decreased current (Martinovic et al. 2017;Mattioli, Cervini, and Cavalheiro 2020).
The same evaluation was performed for a 1.0 Â 10 À5 mol L À1 Cd(II) solution, using GPUE, GPUE-FA 2.5% and GPUE-FA 5%.GPUE-FA 10% was not used once it presented a lower current and a non-defined peak shape, probably due to the presence of a higher amount of non-conductive modifier, as discussed above and was not presented in Figure 2b.
Better results regarding shape and an increase of 23% in current intensity were obtained for the Cd(II) peak around À0.78 V (vs.SCE) using GPUE-FA 5%, when compared with GPUE-FA 2.5%, therefore it was subsequently chosen for further studies (Figure 2b).

Electrochemical impedance spectroscopy (EIS)
EIS measurements were used to compare the equivalent circuits of GPUE and GPUE-FA 5% and to determine the influence of the modifier on the electrical response of the electrodes.The equivalent circuits proposed for GPUE and GPUE-FA 5%, and the EIS Nyquist graphs which were obtained from the proposed equivalent circuits, are presented in Figure 3. Values of the equivalent circuit parameters obtained from the EIS spectra of GPUE and GPUE-FA 5% are summarized in Table S3 (Supplementary Material).
Figure 3a shows the equivalent circuit proposed for GPUE, in which two Randles circuits are observed in parallel, one of them having a Gerischer impedance (G) which could indicate that there are two sites where the electrochemical reaction is occurring.Considering that GPUE is a porous electrode and that the Gerischer impedance can be used to model such type of electrodes, it is likely that the two reaction sites occur at the electrode surface and the interior of the pores.The Gerischer element can be interpreted as an equilibrium between the outer and the inner part of the pores.Once inside the pore, the probe can transfer charge simultaneously at two different sites, which is represented by the two parallel Randles circuits.The Randles circuit in series with the Gerischer element represents charge transfer at the inner part of the pore, while the other Randles circuit represents charge transfer at the electrode outer surface.This assumption can be made since both the R 1 resistance (91.8 X) and W 1 Warburg Impedance (0.82 Â 10 À3 Mho s 1/2 ) are greater than R 2 (27.1 X) and W 2 (0.29 Â 10 À3 Mho s 1/2 ), which corroborates the hypothesis.Lastly, v 2 for this circuit was determined to be 5 Â 10 À4 when fitted against the Bode modulus, Bode Phase and Nyquist graph (Table S3 and Figure 3c).
In Figure 3b, the equivalent circuit proposed for GPUE-FA 5% presents only one resistance associated to charge transfer on the graphitic surface.However, there are two constant phase elements (CPE), one close to the uniformity of a capacitor, shown by its N of 0.62 which can be associated to the electric double layer formation on the graphitic surface and another much more non-uniform, with an N of 0.95, which can be associated to FA modification of GPUE.The roughness and porosity presented by the GPUE base electrode creates a deviation from linearity as the distance between the polarized surface and the ions at the outer plane of Helmholtz varies throughout the entire surface.The elevated capacitance observed at CPE 2 (N ¼ 0.95) compared to that of CPE 1 (N ¼ 0.62) sustains that hypothesis.
The fact that the GPUE-FA equivalent circuit does not present a Gerischer element can be explained by the fact that FA contains alumina acidic sites (Gafurov et al. 2015), which when in its basic form will repel the anionic probe, preventing it to enter the pores.In addition, since FA is a non-conducting aluminum-silicate species, it does not appear in the equivalent circuit as a separated charge transfer resistance, but only contributes to the surface capacitance.FA is a harder material than graphite and therefore the polishing procedure and the sandpaper used to polish the graphite did not affect the morphology of the FA particles.This may explain why the capacitive behavior of FA is close to that of an ideal capacitor, indicating low surface roughness for the electrode.The equivalent circuit for GPUE-FA 5% showed a v 2 of 4 Â 10 À4 when fitted to the Nyquist graph (Table S3 and Figure 3d), indicating that it is an accurate interpretation of the electrode/solution interface.

Analytical evaluation
The effect of varying the supporting electrolyte on the voltammetric response of GPUE-FA 5% during the determination of Cd(II) was studied by cyclic voltammetry of 1.0 Â 10 À4 mol L À1 CdCl 2 in 0.10 mol L À1 solutions of LiCl, phosphate and KCl at pH 3 (Figure 4).A higher current intensity (35 lA) was observed at c.a. À0.81 V (vs.SCE) for the oxidation of Cd(II) in the KCl solution.Finally, 0.10 mol L À1 KCl pH 3 was therefore chosen as supporting electrolyte in subsequent experiments, once higher peak current intensities and well defined voltammetric profiles were found in this medium.Best resolution in the voltammetric response could found extending the cathodic range to À1.2 V (vs.SCE), in cyclic voltammetric experiments, when compared with the voltammograms in Figure 2.
Figure 5 presents the results for optimization of pH, pulse amplitude, scan rate, accumulation time and accumulation potential in the voltammetric response of GPUE-FA 5%, evaluated by DPV and DPASV of 1.0 Â 10 À4 mol L À1 CdCl 2 in 0.10 mol L À1 KCl.The analytical curve obtained for the determination of Cd(II) under the optimized conditions is also shown.The effect of the hydrogen ion concentration of the medium on the voltammetric response was evaluated at different pH values (Figure 5a).The obtained differential pulse voltammograms presented increased current intensity (0.59 lA at c a. À0.87 V (vs.SCE)) and a better peak profile for DPV measurements performed at pH 3.These results suggest a competition between the active sites of FA that seem to be protonated at pH 2.0 and chloride complexes that are present in relevant concentrations at pH !4.0 (Vanderzee and Dawson 1953;Zirino and Yamamoto 1972).
The pulse amplitude and scan rate were optimized by DPASV according to 2 n factorial experimental planning.Pulse amplitude of 50 mV and scan rate of 10 mV s À1 were chosen to present the best voltammetric profile and peak current intensity (Figure 5b).Optimization of the accumulation potential (Figure 5c) displayed lower peak currents at lower accumulation potentials, especially for measurements performed at À0.80 and À0.85 V (vs.SCE).The stripping voltammogram obtained after accumulation at À0.85 V (vs.SCE) presented a peak displacement toward a more positive potential and a shoulder at the same potential as the voltammogram obtained after accumulation at À0.80 V (vs.SCE).This is probably due to an over deposit of Cd on a first Cd layer or a deposition in two different sites e.g. on both graphite and FA contained in GPUE-FA 5%.The highest peak currents (measured at c.a. À0.8 V (vs.SCE)) were obtained after applying À0.95 and À1.0 V (vs.SCE) as accumulation potentials, for which single peaks were observed in both instances.An accumulation potential of À1.0 V (vs.SCE) was chosen for further experiments.
The accumulation time (t acc ) was evaluated between 30 and 300 s (Figure 5d).For t acc more than 180 s, the peak current tended toward a constant value.Therefore, t acc ¼ 180 s was chosen for further studies once above this accumulation time we conclude that the current increase was not enough to justify the lowering in the analytical frequency.
Analytical curves (Figure 5f) were obtained using the optimized parameters for DPASV at GPUE-FA 5% (Figure 5e), using Cd(II) solutions with concentrations ranging from 1.0 Â 10 À7 to 1.0 Â 10 À6 mol L À1 .For comparison, the same set of experiments was performed using the unmodified GPUE, which presented a similar response, however with lower peak currents.
From Figure 5f it is clear that there is an exponential grow in Cd(II) stripping current for concentrations higher than 4 mmol L À1 , suggesting that initial nucleation of the metal on the surface facilitates the deposition and increases the mass of metal on the electrode, especially in higher concentrations.This over deposition was proposed based on results in Fig. 5c and based on the displacement of peak potential.The effect is more accentuated in the electrode containing fly ash, demonstrating the effect of this modifier in the current enhancing and in the increase in the sensitivity of the composite when modified with FA.
At the unmodified GPUE, a linear dynamic range between 4.00 Â 10 À7 and 1.00 Â 10 À6 mol L À1 was observed, with a limit of detection (LOD) of 1.49 Â 10 À7 mol L À1 and R ¼ 0.9990, for n ¼ 4. The LOD was calculated as three times the standard deviation of the blank, divided by the slope, according to Long and Winefordner (1983).At GPUE-FA 5% a linear range (Figure 5f) was obtained between 2.0 Â 10 À7 and 1.0 Â 10 À6 mol L À1 , with an LOD of 6.61 Â 10 À8 mol L À1 and R ¼ 0.9989, for n ¼ 5.This LOD was comparable to recent studies of other electrode compositions (Hu et al. 2019), but lower than that observed for some more expensive or non-usual materials (Hu et al. 2019;Oularbi, Turmine, and el Rhazi 2019;van Staden and Arnold Tatu 2019;Guenang et al. 2020;Hassanpoor and Rouhi 2021).Hou et al. (2021) used a carbon paste electrode containing modified fly ash on top of which Cd(II) was electrochemically co-deposited with Sb or Bi films.The FA was modified by extensive treatment in NaOH 3 mol L À1 , during 24 h at 75 C.The treatment changed the spherical shape of the particles while change its chemical composition.Despite the much more time consuming process and the use of toxic metal films the authors reported limit of detections close to those reached in the present work.
However, GPUE-FA has the advantage of being composed of inexpensive and recycled materials, making it more financially and environmentally friendly.Above 1.0 Â 10 À6 mol L À1 , a deviation from the linear response at both GPUE and GPUE-FA was observed, probably due to the response of graphite active sites at the electrode surface with the increasing in Cd(II) concentration.Both the sensitivity and linear range obtained were higher for GPUE-FA 5% in comparison with GPUE.

Determination of Cd(II) in battery samples
Aliquots of the solutions containing Cd(II) extracted from the anodes of exhausted Cd-Ni batteries were added to the cell containing 0.1 mol L À1 KCl solution at pH 3 in order to obtain a working concentration of ca.2.0 Â 10 À7 mol L À1 .After that, three additions of 2.0 Â 10 À7 mol L À1 Cd(II) from standard solution were performed, and DPASV voltammograms were registered in triplicate using GPUE-FA 5% under the optimized conditions.Table 2 presents the Cd(II) concentration determined by triplicate DPASV measurements for three extractions of Cd from the batteries in comparison to the concentration determined by atomic absorption spectrometry (FAAS).The results obtained with GPUE-FA 5% agreed with those by FAAS at the 95% confidence level according to the Student's t-test.When comparing the results from DPASV via the standard addition method, with those from FAAS, the relative error (E r ) was determined to be less than 1.5% for each individual extraction experiment.This result demonstrates the effectiveness of GPUE-FA 5% for the determination of Cd(II).

Determination of Cd(II) in water samples
Water samples, collected from three natural sources (artesian well, tap water and non-sparkling mineral bottled water) were spiked with Cd(II) standard solutions until a Cd(II) concentration from 2.0 to 3.0 10 À7 mol L À1 was reached.The spiked water samples containing 0.10 mol L À1 KCl solution at pH 3 were added to a voltammetric cell.After that, three additions of 2.0 Â 10 À7 mol L À1 Cd(II) solution were performed and DPASV voltammograms were recorded in triplicate using GPUE-FA 5%.The obtained results were compared with those from FAAS (Table 2).
The results obtained with GPUE-FA 5% agreed with those from FAAS in the 95% confidence level according to the t-test.When comparing the results from DPASV via the standard addition with those from FAAS, the relative error was less than 2% for each water sample tested.The effectiveness of GPUE-FA 5% for the determination of Cd(II) in water samples was therefore confirmed.These results are particularly noteworthy since each water sample contained different quantities and types of minerals due to their different origins.Furthermore, the initial Cd(II) concentrations of the water samples varied, showing the ability of the GPUE-FA 5% electrode to determine Cd(II) concentrations in different parts of the linear range.
The concentration of concomitant cations in the natural water samples investigated in this study was determined by FAAS (Table S4, Supplementary Material).The presence of mainly Li, Na, Ca, Mg, Fe and Zn cations, up to the levels in which they naturally occur, did not interfere with the DPASV determination of Cd(II) using GPUE-FA 5%.
Additionally, the long-lasting useful life of the GPUE-FA 5% electrode was demonstrated by the fact that only one electrode was used for all the voltammetric determinations presented in this paper.These determinations were performed over a period of 6 months.
Proposal for the action of FA in the pre-concentration of metal ions Figure 6 proposes a scheme to explain the effect of FA embedded in the GPU composite matrix with relation to the preconcentration of metallic cations, according to observations made in the current study.
Functional groups on the FA surface may be reversibly protonated in a sufficiently acidic medium.This statement is justified by the decrease of the Cd(II) voltammetric signal in a medium of pH < 3 (Figure 5a).Therefore, this parameter requires optimization for DPASV measurements.Following optimization of the pH (i.e., acidity), the FA will have deprotonated surface functional groups which are free to interact with the metallic cation, thereby promoting pre-concentration of the cation on the FA surface.This effect will lead to increased sensitivity in the voltammetric signal of the FA modified electrode, when compared to that of the unmodified electrode.
At adequate potential, Cd 2þ is reduced and accumulated in the device, being oxidized in the redissolution (or stripping) step.Since a voltammetric response was also observed for the unmodified electrode, it is inferred that the GPU matrix also allows Cd 2þ deposition on its surface.However, the presence of FA with its surface functional groups intensifies this process.
All steps involved proved to be reversible, allowing for the electrode to be reused without the need for surface renewal between successive measurements.This model can be applied to other metallic cations, under optimized conditions of electrolytic medium, pH, accumulation time and accumulation potential.

Conclusions
This work demonstrated that FA can be incorporated to a graphite-polyurethane composite to prepare modified voltammetric electrodes for determination of metallic cations in natural waters, as demonstrated using Cd 2þ as an electroanalytical probe.
The presence of FA in the composite matrix was demonstrated by SEM, EDS and TG/DTG.The FA modifier was firmly supported in the GPU composite and it was distributed throughout the bulk of the material.
Electrochemical measurements demonstrated that FA improved the sensitivity of the voltammetric response to potassium ferricyanide.It has furthermore shown that preconcentration of Cd(II) on the FA surface may facilitate reaching limits of detection at sub-mmol L À1 level, when DPASV is used for detection.
EIS showed that the addition of FA to the GPU composite influenced its equivalent circuit, due to the FA modifier occupying the empty pores of the GPU composite to repel the analyte, preventing it from entering the pores with subsequent accumulation of the analyte at the electrode surface.
Cd(II) was successfully determined by DPASV using the GPUE-FA 5% electrode in the anodes of exhausted rechargeable batteries and in natural water samples, following spiking with standard solutions of CdCl 2 .The results for determination of Cd(II) by DPASV agreed well with FAAS data.The presence of naturally occurring concomitants in the water samples had no influence on the accuracy of Cd(II) determinations.
The GPUE-FA electrodes were easy to construct, can be built at low cost, presented easiness of surface renovation once the FA modifier was dispersed in the whole bulk and had a sensitivity on par with other electrodes.The proposed FA modified GPU electrodes therefore presented a viable option for the reuse of a by-product an important environmental waste which is generated during combustion of coal in coal-fired power stations.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Figure 1 .
Figure 1.SEM micrographs obtained for the fractured GPU-FA 5% composite at (a) 3 000X and (b) 8 000X magnifications and polished GPU-FA 5% at (c) 3 000X and (d) 15 000X magnifications.In images (b) and (d) the same frame is presented at different contrasts to better observe the details of insertion of the FA particle into the composite matrix.

Figure 4 .
Figure 4. -Cyclic voltammograms obtained at v ¼ 50 mV s À1 for GPUE-FA 5% in the presence of 1.0 x 10 À4 mol L À1 Cd(II) in 0.10 mol L À1 solutions of LiCl, phosphate and KCl at pH 3.

Figure 6 .
Figure 6.Schematic representation of the action of FA in the pre-concentration of Cd 2þ (as a model), on the surface of GPUE-FA 5% (m/m).

Table 1 .
Composition of the GPU-FA composites.

Table 2 .
Results of Cd content in battery extracts and spiked water samples using DPASV at GPUE-FA 5% and FAAS.