Optimisation of Electrokinetic Extraction System: Colourimetric Determination of Copper (II) in Sand Using Polymer Inclusion Membrane

ABSTRACT

Electrokinetic extraction (EKE), coupled with polymer inclusion membranes (PIMs) for colourimetric detection, offers a promising approach for extracting metal ions from solid matrices, such as sand, without sample pretreatment. Our EKE system consists of a formic acid gel positioned at the cathode side to neutralise hydroxide ions generated by electrolysis during the EKE of Cu(II), whereas the anode side contains the sample compartment, prepared from a syringe tube into which the sample is placed. The PIM is sandwiched between these compartments to accumulate Cu(II) via EKE through the sand and into the PIM within which it complexes with a colour‐forming reagent, 1‐(2‐pyridylazo)‐2‐naphthol (PAN), enabling quantification via colourimetric analysis. The system has been optimised to expand the quantification range of Cu(II), thus allowing greater applicability to a wider range of samples with different levels of heavy metals. Increasing the column diameter extended the upper limit of the quantification range from 7.5 mg kg⁻¹ (4.5 mm column diameter) to 30.0 mg kg⁻¹ (13.5 mm column diameter). Concurrently, optimisation of the PIM composition, containing 43.3 wt% cellulose triacetate (CTA)–56.0 wt% di‐(2‐ethylhexyl) phosphoric acid (D2EHPA)–0.8 wt% PAN, enhanced the overall recovery of Cu(II) from 64% ± 7% to 90% ± 4%. Overall, the optimised system has achieved a quantification range of 2.8–30.0 mg kg⁻¹ of Cu(II), with an LOD of 0.9 and LOQ of 2.8 mg kg⁻¹ of Cu(II).

Electrokinetic extraction (EKE) allows the direct transfer of heavy metals from soli samples, such as soil and sediment, into a polymer inclusion membrane (PIM) with a colourimetric detection. This avoids length and complex sampling handling and extraction processes.

Abbreviations

ΔGrV

change in grey value

CTA

cellulose triacetate

D2EHPA

di‐(2‐ethylhexyl) phosphoric acid

DCM

dichloromethane

DOP

dioctyl phthalate

EKE

electrokinetic extraction

GrV

average grey values

PAN

1‐(2‐pyridylazo)‐2‐naphthol

PIM

polymer inclusion membrane

pXRF

portable x‐ray fluorescence

1. Introduction

Copper is a metal with diverse applications. It not only serves as an essential micronutrient for living organisms but also plays a vital role in everyday technologies through its use in electronics and appliances [1, 2, 3]. With the increasing use of Cu, waste containing Cu(II) has become a pollutant in the environment [4, 5].

The detection and quantification of Cu(II) in solid matrices, like sand and soil, has become vital as overexposure to Cu(II) can cause lethal effects [1, 6]. To measure Cu(II) in these samples in‐field, multiple methods have been developed. Portable x‐ray fluorescence (pXRF) is rapid and easy to use, while providing quantitative results for a range of elements present in the soil without sample pretreatment [7]. One of the major challenges with pXRF, however, is that it only allows total quantification of metals [8], lacking the ability to provide more environmentally relevant information, for example, the chemical form of metals. Among these, water‐soluble fraction of a metal is normally the most bioavailable and toxic form in the environment, whereas mineralised forms are less mobile and pose a lower immediate risk [5]. Furthermore, the radiation emitted during operation requires proper licensing and specialised training for safe and compliant usage and may require special transport considerations.

Portable colourimetric analytical devices, which can be applied for both qualitative and quantitative analysis, are capable of quantifying Cu(II) without the use of lab‐based instruments [9, 10, 11, 12]. Analysis based on these colourimetric analytical devices is often cheaper compared to lab‐based analysis. Qualitative and semi‐quantitative results can often be obtained by comparing the colour of a solution or test strip to a colour chart. Capturing photographic images and analysing them with phone apps or imageJ software provides a more accurate quantitative determination of the Cu(II) concentration [10, 11, 12, 13]. Currently, these devices are only able to detect and quantify Cu(II) in water and soil extracts using extractants such as ethylenediaminetetraacetic acid [14, 15]. The sensitivity of the colourimetric methods can be strongly affected by the background colour of the aqueous samples.

Polymer inclusion membranes (PIMs) have been used in many applications due to their chemical versatility and the ability to be tailor‐made for specific purposes [16]. The incorporation of the colourimetric dye, 1‐(2‐pyridylazo)‐2‐naphthol (PAN), into a PIM consisting of poly(vinyl chloride) as base polymer and di‐(2‐ethylhexyl) phosphoric acid (D2EHPA) as the PIM extractant or into a polymeric sensing network of a similar composition has allowed sensitive and selective Cu(II) determination in tap and environmental waters [11, 12]. Unlike traditional colourimetric methods, where the sample directly contacts and reacts with PAN, the PIM serves as a selective barrier to minimise or even eliminate potential interferences from other metal ions (i.e., interferences from Zn(II), Fe(III), Al(III) and Cd(II) were eliminated). Additionally, the hydrophobic nature of the PIM mitigates staining from the slurry solution or coloured compounds, yielding more accurate results.

In our recent study, a PIM incorporating D2EHPA, cellulose triacetate (CTA) as its base polymer and dioctyl phthalate (DOP) as plasticiser, was incorporated into an electrokinetic extraction (EKE) system capable of extracting and measuring Cu(II) in solid samples such as sand [17]. The fundamental concept of this approach is the use of an electric field to migrate Cu(II) into the PIM, thereby avoiding sample pretreatment steps and performing extraction and detection simultaneously. All components of the system are in a solid state, making them safe to handle and suitable for field use. Although the feasibility of this approach was previously demonstrated, it exhibited a relatively narrow quantification range of 1.2–9.4 mg kg⁻¹ Cu(II), which is much lower than the normal levels of Cu(II) in sand and soil which are in the range 2–250 mg kg⁻¹ [1, 5]. Although quantifying higher Cu(II) concentrations can be addressed by diluting the sample with a diluent (such as pure silica), this is not ideal for in‐field measurements, nor for long‐term unattended Cu(II) monitoring. In this study, the EKE system, including its PIM and data analysis, was optimised to extend the quantification range of Cu(II), thus making the EKE system suitable for direct analysis of environmental samples.

2. Experimental Procedures

2.1. Materials and Reagents

CTA (Selectophore, Sigma‐Aldrich, MO, USA), D2EHPA (97%, Sigma‐Aldrich, MO, USA), DOP (99%, Sigma‐Aldrich, MO, USA), PAN (indicator grade, Fluka, North Carolina, US) and dichloromethane (DCM) (suitable for HPLC, ≥99.8%, contains amylene as stabiliser; Sigma‐Aldrich, MO, USA) were used for the PIM fabrication.

Anhydrous copper sulphate (CuSO₄; BDH Chemicals Ltd, Poole, England) was used to prepare the Cu(II) stock solution. Formic acid (98%; Sigma‐Aldrich; California, USA) and agarose powder (certified molecular biology agarose; Bio‐Rad, Spain) were used to fabricate neutralising gels containing 0.05% (v/v) formic acid [17] which were used to neutralise the hydroxide ion produced during the EKE process. Deionised water (DI; Milli‐Q IQ Element Purification Unit, Darmstadt, Germany) was used for solution preparation. Universal indicator (pH 3–pH 11; Chem‐Supply, SA, Australia) was used to test the pH of silica sand and neutralising gel after EKE. Method development was performed using silica sand (50–70 mesh particle size, Sigma‐Aldrich, MO, USA). Nitric acid (70%; RCI Labscan, Chem‐Supply, SA, Australia) was used for leaching Cu(II) from PIMs and sand samples.

2.2. PIM Fabrication

The preparation of PIMs was adopted from our previous study [17]. Briefly, the PIMs were prepared using the solvent evaporation method, where all the PIM components were weighed and dissolved in DCM (18 mL DCM for 170 mg of CTA). PIMs were cast either using a glass ring or a Petri dish (PD). For ring casting, 7 mL of the homogenous casting solution was poured into a glass ring (100(OD) × 2.5(T) × 20(H) mm Brandon Scientific Glassblowing, TAS, Australia) placed on top of a glass plate (Figure S1). The glass ring was then covered with a KIM‐wipe (21 × 11 cm; Kimberly Clark Professional, Georgia, USA), watch glass and finally a glass container to slow down the evaporation rate of DCM. The solutions were left undisturbed for a minimum of 18 h in the fume hood. The PIMs were then gently peeled off from the glass plate and glass ring. When using the PD, the PIMs were cast onto a glass PD (60(D) × 15(H) mm; Technosklo, Livingstone, NSW, Australia), following the method described earlier [17]. The PD cast PIM and glass ring (R) cast PIMs were termed PD–PIM and R–PIM, respectively.

2.3. Configuration of the EKE System

The EKE design was also adopted from our earlier study [17]. The set‐up included a sample compartment next to the anode, a PIM and a neutralising gel next to the cathode. The parts for the EKE system were 3D‐printed using the Original Prusa i3 MK3S+ 3D Printer (Prusa Research by Josef Prusa, Prague, Czech Republic), with PLA+ Filament (1.75 mm, 3D‐Fillies, Victoria, Australia). The anode was a 26G × 1″ syringe needle (HSW FINE‐JECT, Tuttlingen, Germany), and the cathode was a 3D‐printed carbon electrode printed from conductive polylactic acid (PLA, 1.75 mm, Proto‐Pasta, Washington, USA). Syringes accommodating 1 mL (Livingstone, Luer Slip Tip, NSW, Australia), 3 mL (Terumo, Luer Slip Tip, Tokyo, Japan) and 5 mL (Terumo, Luer Slip Tip, Tokyo, Japan) were employed as the sample compartment to house both the solid samples and neutralising gels.

2.4. EKE System Optimisation

2.4.1. PIM Casting (PD Casting vs. R Casting)

To compare the PD–PIMs and R–PIMs, EKE of Cu(II) from spiked silica sand was performed using previously used conditions [17]: That is, 20 V for 1 h applied with LabSmith HVS448 High Voltage Sequencer (LabSmith, Livermore, California, USA). Silica sand was spiked to 1.5 mg kg⁻¹ Cu(II) with 37.5 µL of 10 mg L⁻¹ Cu(II) solution to obtain a 15% total moisture level. The sample compartment was a 10 mm length column cut from a 1 mL syringe that held 0.25 g silica sand. A piece of formic acid neutralising gel was inserted between the cathode and PIM. The pH values of the neutralising gel and silica sand after EKE were determined by adding a few drops of universal indicator solution. Experiments were conducted in triplicate. A new PIM and a freshly prepared gel were used for each experimental run to maintain consistent pH conditions while preventing cross‐contamination.

Photographic images of the PIMs were taken before and after EKE using a smartphone within a custom‐designed light box to ensure consistent illumination during image acquisition, thereby minimising the influence of ambient lighting and camera settings (Figure S2). Image analysis of the PIMs was done with the method previously used by us [17]. Briefly, the image of a PIM was split into RGB greyscale images and transformed into average grey values (GrV) with ImageJ [18]. The ‘change in grey value (ΔGrV)’ was calculated by subtracting the GrV of the PIM before and after the EKE.

With the optimal extraction time for R–PIM (3 h), a calibration graph of the EKE of Cu(II) from silica sand spiked to 0–15 mg kg⁻¹ was constructed. Triplicate experiments were conducted for each EKE of Cu(II) using PD–PIMs or R–PIMs.

2.4.2. Column Diameter Optimisation and Quantification Method Optimisation

The effect of the column diameter on the quantification range of Cu(II) with the EKE method was examined using 1‐, 3‐ and 5‐mL syringes (Table S1) corresponding to diameters of 4.5, 10.0 and 13.5 mm, respectively. The EKE conditions were the same as outlined in Section 2.4.1, except that the amount of spiked silica sand was 0.35 g. For a valid comparison of the effectiveness of each column diameter, optimisation of the extraction time was performed for each column diameter.

With the optimal extraction time for each column diameter, EKE was performed on spiked silica sand with 0–150 mg kg⁻¹ Cu(II) to obtain individual calibration graphs. Comparisons of the influence of column diameter were interpreted by using both the grey value (GrV) and the change in grey value (ΔGrV). The influence of the grid size of the analysis area during PIM image analysis was also examined.

2.4.3. PIM Composition Optimisation

The PIM composition was optimised by varying the PAN, D2EHPA and DOP concentrations in a univariate manner. The base level of each component was 37.5 wt% CTA, 48.6 wt% D2EHPA, 13.2 wt% DOP and 0.7 wt% PAN, as used previously [17].

The optimal amount of PAN was determined using PIMs containing 0.2–8.8 mg of PAN through a spot‐test with Cu(II) solution containing 0–10 µg of Cu(II). The spot‐test procedure was described earlier by us [17]. Reflectance was calculated from GrV by the following equation:

$$R=-\mathrm{lo}{\mathrm{g}}_{10}\left(I/I_0\right)$$ (1)

where I is the GrV of the Cu(II) spot, and I ₀ is the colour intensity of a blank spot. The standard deviation (SD) was calculated by the following equation:

$$\mathrm{SD}=\frac{\sqrt{{\left(\frac{\mathrm{\Delta }{I}_0}{I_0}\right)}^2+{\left(\frac{\mathrm{\Delta }I}{I}\right)}^2}}{\ln 10}$$ (2)

The optimal PIM composition of D2EHPA (0–385 mg) and DOP (0–105 mg) was determined as follows: EKE of Cu(II) from silica sand was conducted at 20 V for 2 h, 1.0 mm column length × 13.5 mm column diameter, 0.35 g silica sand spiked with 52.5 µL of 150 mg L⁻¹ Cu(II) solution (15% total moisture level, equivalent to the 22.5 mg kg⁻¹ of Cu(II)), and 0.05% formic acid neutralising gel. The experiments were done in triplicate.

2.5. Validation

With the optimised parameters for the EKE system and PIM composition, EKE was performed on silica sand spiked with 0–75 mg kg⁻¹ Cu(II). A Cu(II) calibration graph was constructed on the basis of the PIM reflectance (Equation 1), and its quantification range was determined. The LOD (Equation 3) and LOQ (Equation 4) were calculated on the basis of the first three points of the calibration graph [19]:

$$\mathrm{LOD}=\left[3.3\times \left(\sigma /\mathrm{s}\right)\right]$$ (3)

$$\mathrm{LOQ}=\left[10\times \left(\sigma /\mathrm{s}\right)\right]$$ (4)

where σ represents the residual SD of the calibration curve, and s is the slope of the calibration curve [20].

The extraction efficiency of the EKE system was determined by quantifying Cu(II) in the PIM and silica sand after EKE. This was done by transferring the PIM and silica sand into plastic bottles containing 5 mL of 1 M HNO₃ and shaking them on an orbital shaker. For silica sand leaching, the suspension was shaken for 1 h, then made up to 10 mL with DI water and filtered with filter paper (Whatman qualitative filter paper, Grade 1; Sigma‐Aldrich, St. Louis, MO, USA). For PIM leaching, the solution was shaken until the PIM lost its reddish colour. The PIM was then removed, and the solution volume was made to 10.0 mL. The quantification of Cu(II) in the solution was performed with flame atomic absorption spectrometry (FAAS, Agilent 55B AA), with calibration of 0–5 mg L⁻¹ Cu(II) solutions.

2.6. Application to Tailings Samples

The EKE system was applied to tailings samples collected in a local mining region: Mount Lyell sample (Sample ML) and the weathered riverine tailings sample (Sample BD). Sample ML was collected from ∼10 cm depth in Princess Creek tailings dam, which houses tailings from the Mount Lyell copper mine, Queenstown, Tasmania (sampling location: 42.0088090° S and 145.504880° E). Sample BD was collected from the tailings‐formed Bank D of the King River (sampling location: 42.196964° S and 145.383940° E), located downstream from the Mount Lyell copper mine [21]. Sample preparation was done as described earlier by us [17]. Cu(II) quantification was done by comparing the EKE results with the calibration graph using silica sand standards.

3. Results and Discussion

Previously, we demonstrated the EKE of Cu(II) from silica sand and tailings samples [17]. A self‐supported and functional PD–PIM with a composition of 37.5 wt% CTA, 48.6 wt% D2EHPA, 13.2 wt% DOP and 0.7 wt% PAN was used, and the colour intensity of the PIM was related to the Cu(II) level in the samples. However, due to the narrow quantification range, dilution with silica sand was required for tailings samples that contained high Cu(II) concentrations. In the present study, optimisation of the PIM and EKE system was carried out to extend the quantification range of Cu(II) to allow direct extraction without dilution.

3.1. PIM Casting Method

The PD–PIM used earlier by us [17] was observed to have coffee‐ring effect when cast [22], causing uneven PIM thickness and limiting the usable PIM area (Table 1). To increase the usable PIM area, casting was attempted using a larger diameter glass ring (R–PIM) placed on a flat glass plate (Figure S3). Although initial attempts led to leakage during overnight evaporation (Figure S3A), doubling all PIM components (with constant DCM volume) increased solution viscosity and prevented leakage (Figure S3B). The resulting R–PIM was visually homogenous with a thickness variation smaller than that of the PD–PIM (Table 1; <10% RSD in PIM thickness across one PIM).

TABLE 1.

EKE results for PD–PIM and R–PIM.

PIM	Thickness (µm)	Applied voltage (V)	Current (µA)	ΔGrV
PD–PIM	11–18	20	96 ± 28	44 ± 6
		0	N/A	6 ± 4
R–PIM	27–32	20	44 ± 4	1.5 ± 1.8
		0	N/A	3.7 ± 0.8

Note: Experimental conditions: extraction for 1 h at 20 V; 0.25 g of silica sand spiked with 1.5 mg kg⁻¹ of Cu(II); 4.5 mm column diameter; 15% moisture level and error bars represent 1 standard deviation (SD), calculated from three independent experiments (n = 3).

Abbreviations: ΔGrV, change in grey value; PD, petri dish; PIM, polymer inclusion membrane.

EKE of Cu(II) was performed using R–PIM and compared with the PD–PIM used earlier by us [17]. The ΔGrV for R–PIM was found to be lower than that for the PD–PIM and similar to the control set without applied voltage (Table 1). This correlated with a lower current observed during EKE using the R–PIM (Table 1), reducing the migration of Cu(II). This was due to the increased thickness of the R–PIM which reduced the electric field distributed over the sample. To test this, the extraction time was extended, and the optimal extraction time was found to be 3 h at which point the ΔGrV reached a plateau of 60 ± 3 (Figure S4).

3.2. EKE System Optimisation

3.2.1. Column Diameter Optimisation

EKE of Cu(II) was conducted with three column diameters: 4.5‐, 10.0‐ and 13.5‐mm (Table S1). Larger diameters were not tested due to insufficient silica sand to fully cover the PIM surface. As the column diameter increased, the corresponding column length decreased. Therefore, the optimum extraction time was determined individually for each configuration. The optimum extraction time, at which ΔGrV reached a plateau despite increasing extraction time, decreased with increasing diameter (Figure S4B; 3 h for 4.5 mm, 2 h for 10.0 mm and 2 h for 13.5 mm).

The reduction in extraction time is attributed to the changes in the overall electrical resistance of the EKE system. The system consists of several conductive components, including the sand matrix, PIM and neutralising gel. Under a fixed applied voltage, the total current is governed by the combined resistance of these components. Although the resistance of the PIM and gel remained constant across experiments due to their fixed lengths, the resistance of the sand matrix varied with column geometry because of the fixed volume of sand used. Increasing the column diameter increased the cross‐sectional area available for charge transport and thereby reduced the effective transport length within the sand, thereby lowering its resistance.

Consequently, this resulted in a higher extraction current (4.5 mm: 44 ± 9 µA; 10.0 mm: 100 ± 13 µA; 13.5 mm: 130 ± 17 µA), which enhanced electrokinetic ion migration and reduced the extraction time required. In addition, the reduction in column length decreased the migration distance required for Cu(II) to reach the PIM, further contributing to the reduction in extraction time (Table S1).

Furthermore, as the diameter increased, the colour formed in the PIM was distributed over a larger area, resulting in a less intense but more widespread colour change. This enabled the PIM to reach its colour saturation at a higher Cu(II) concentration in the sand, thereby extending the upper limit of the calibration range for Cu(II).

To confirm this, experiments with the three column diameters were performed under their optimum conditions at concentrations ranging from 0 to 60 mg kg⁻¹. Figure 1A shows that, as expected, the Cu(II) linear quantification range extended to a higher concentration as the diameter increased, from 7.5 mg kg⁻¹ at 4.5 mm to 22.5 mg kg⁻¹ at 13.5 mm.

FIGURE 1.

Scatter plots (A–C) showing EKE results of Cu(II) extracted from silica sand using different image interpretation methods: (A) change in GrV, GrV with (B) large grid and (C) small grid. Part (D) illustrates the area of quantification of the large and small grids. Experimental conditions: 20 V; 0.35 g of silica sand spiked with 0–60 mg kg⁻¹ of Cu(II); 3 h extraction time for 4.5 mm column diameter, 2 h extraction time for 10.0 and 13.5 mm column diameter; 15% moisture level; SD: n = 3. ΔGrV, change in grey value; GrV, average grey values.

3.2.2. Image Analysis Optimisation

Previously, Cu(II) concentration in silica sand was quantified using ΔGrV, which was the difference in the green‐channel GrV of the PIM before and after EKE. This calculation acted as a baseline correction to minimise interference from the PIM's initial yellow–orange colour. However, this initial background colour of the PIM contributed a signal to the green‐channel (Figure S5; background GrV = 175.7 ± 0.3). Upon exposure to high levels of Cu(II), the PIM colour completely shifted towards red–purple, where the green‐channel signal originated solely from the Cu–PAN complex rather than from the PIM background. Hence, ΔGrV artificially reduced the signal by subtracting the true complex contribution along with the background, leading to underestimation of Cu(II) concentrations.

To resolve this discrepancy, an alternative approach was evaluated on the basis of the unadjusted GrV after EKE, which was without background subtraction. As shown in the calibration graph (Figure 1B), the GrV decreased with increasing Cu(II) concentration, reflecting the increasing red–purple colour intensity as more Cu–PAN complex was formed. This approach produced a calibration curve that better reflected the underlying complexation process and avoided the risk of underestimating Cu(II) concentrations, particularly at higher concentrations.

Another factor considered was the shape of the image that was analysed. When converting the PIM images into GrV, the ImageJ software limited the measurement area to a square shape. Initially, a large square grid was drawn to capture the full colour change on the PIM, aiming to quantify the total amount of Cu(II) extracted (Figure S5). However, because the sample compartment was circular, the grid also included peripheral regions of the PIM that were not in direct contact with the silica sand. This inclusion of irrelevant regions ‘diluted’ the signal, as the quantification was based on the average value across the entire grid (GrV for 4.5 mm: 57 ± 7; 10.0 mm: 65 ± 2; 13.5 mm: 78 ± 5). To reduce this influence, the square grid was resized to internally fit within the circle, thus only containing PIM area in contact with the silica sand (Figure 1C,D). After this adjustment, the GrVs were lower (4.5 mm: 22 ± 1; 10.0 mm: 19 ± 2; 13.5 mm: 20 ± 1), providing clearer differentiation between Cu(II) concentrations and thereby improving the reliability of the measurement.

3.2.3. Quantification Range for Each Column Diameter

A polynomial equation was fitted to each calibration set shown in Figure 1C to determine the Cu(II) quantification range. The range extended up to 7.5 mg kg⁻¹ using a 4.5 mm column, 15.0 mg kg⁻¹ using a 10.0 mm column and 30.0 mg kg⁻¹ using a 13.5 mm column (Figure S6). These results indicated that the EKE system could quantify higher Cu(II) concentrations with increasing column diameter. Therefore, the 13.5 mm diameter column was selected as the optimal column for the subsequent experiments.

3.3. PIM Composition Optimisation

3.3.1. PAN Optimisation

As PAN is the primary colour‐inducing component, its amount in the PIM influences the Cu(II) quantification range. Visual inspection of the PIMs with different amounts of PAN showed slight colour differences. PIMs containing 0.2–2.6 mg of PAN appeared increasingly yellow, whereas those with 5.3 and 8.8 mg of PAN were yellow–orange in colour (Figure S7). These colour variations were reflected in the green‐channel GrV (Figure 2A).

FIGURE 2.

(A) GrV of PIMs with varying amounts of PAN. (B) Maximum response of PIM against the Cu(II) concentration. Reflectance of Cu(II) spot with PIM containing (C) small and (D) large PAN amounts. Experimental conditions: 0.2 µL of solution containing 0–1 µg of Cu(II); 10 min drying time; SD: n = 3. GrV, average grey values; PAN, 1‐(2‐pyridylazo)‐2‐naphthol; PIM, polymer inclusion membrane.

The optimal PAN amount was evaluated using a drop‐test with Cu(II) solution (0.2 µL per drop; 0–0.1 µg of Cu(II)) on PIMs containing 0.2–8.8 mg of PAN. As the colour response continued to increase for PIMs with 1.8–8.8 mg PAN, an additional drop‐test was performed using higher Cu(II) concentrations (up to 1 µg).

The colour of the Cu(II) spots varied depending on the PAN amount (Figure S7). PIMs with lower PAN amounts (0.2–1.8 mg) produced pink–purple coloured spots, and their colour intensity increased at higher Cu(II) concentrations. PIMs with higher PAN amounts (2.6–8.8 mg) showed red spots at low Cu(II) levels, which shifted to dark red‐purple with increasing Cu(II) concentrations.

To allow better comparison between PIMs, the GrVs were converted to reflectance (Equation 1) and plotted on the same scale. Figure 2B shows an increasing trend in the maximum GrV response with increasing PAN amount, but no further improvement was observed beyond 2.6 mg of PAN. At low amounts of PAN in the PIM (0.2–2.6 mg), the availability of PAN to complex Cu(II) was the limiting factor for the quantification range. Once all free PAN had formed complexes, further increases in Cu(II) concentration did not produce additional colour changes (Figure 2C). In contrast, at higher amounts of PAN in the PIMs (5.3–8.8 mg), the quantification range was limited by the capabilities of the camera/detector to capture the colour changes as more Cu–PAN complex was formed (Figure 2D). When a high concentration of Cu–PAN complex was present, the resulting colour intensity saturated the detector, preventing further increases in GrV. As a result, the quantification range did not improve further with additional PAN (Table S2).

In summary, increasing the amount of PAN in the PIM extended the Cu(II) quantification range. However, beyond 2.6 mg of PAN, colour saturation limited further expansion of the range.

3.3.2. D2EHPA Optimisation

PIMs containing 0, 48, 96, 144, 193, 289 and 385 mg of D2EHPA were cast, and their properties evaluated (Figure S8A). All PIMs exhibited sufficient mechanical strength to be peeled from the casting surface. PIMs with 0–96 mg of D2EHPA were transparent, whereas those with 144–385 mg were non‐transparent. All appeared homogeneously orange in colour, except for the PIM containing 385 mg of D2EHPA, which showed visual inhomogeneity.

EKE of Cu(II) from silica sand was performed using each of the PIMs (Figure 3A). The PIM without D2EHPA was non‐conductive, resulting in limited Cu(II) migration and consequently low extraction efficiency. For the remaining PIMs, both the GrV and the average electric current were comparable across experiments, indicating similar Cu(II) extraction. An exception was observed for the PIM with 96 mg of D2EHPA, which exhibited a high average electric current. Despite this, no improvement in Cu(II) extraction was observed. Instead, a colour change appeared in the neutralising gel, indicating the presence of the Cu–PAN complex. This effect can be attributed to the mobilisation of the PIM's liquid phase under high electric current, which reduced the retention of the Cu–PAN complex in the PIM matrix. As a result, the complexes migrated out of the PIM in the same direction as Cu(II). Therefore, the D2EHPA content was maintained at 193 mg per PIM, consistent with the optimal formulation used in the previous section.

FIGURE 3.

The GrV of the PIMs with varying amounts of (A) D2EHPA and (B) DOP after EKE of Cu(II) from spiked silica sand and the average electric current during the EKE. Experimental conditions: extraction for 2 h at 20 V; 0.35 g of silica sand spiked with 22.5 mg kg⁻¹ of Cu(II); 13.5 mm column diameter; 15% moisture level and SD: n = 3. D2EHPA, di‐(2‐ethylhexyl) phosphoric acid; DOP, dioctyl phthalate; GrV, average grey values; PIM, polymer inclusion membrane.

3.3.3. DOP Optimisation

PIMs containing 0, 26.3, 52.5, 78.8 and 105.0 mg of DOP were prepared to assess their mechanical and optical properties (Figure S8B). All PIMs had sufficient mechanical strength to be peeled from the glass casting surface. Among them, only the PIM containing 105.0 mg of DOP was semi‐transparent, whereas the others were non‐transparent.

The extraction efficiency of Cu(II) from silica sand was then evaluated using each PIM formulation (Figure 3B). The extraction efficiency decreased with increasing DOP content, which may be attributed to increased PIM thickness (T) that reduced the overall conductivity of the EKE system (T _0 mg: 33 ± 1.5; T _26.3 mg: 35 ± 1.7; T _52.5 mg: 37 ± 2; T _87.7 mg: 39 ± 3; T _105.0 mg: 45 ± 4 µm). This, in turn, limited the migration of Cu(II) towards the PIM.

Besides, under similarly high‐current conditions (>150 µA), the PIM without DOP exhibited higher extraction efficiency, whereas the PIM containing 193 mg of D2EHPA (52.5 mg DOP) showed lower extraction efficiency and staining of the neutralising gel to pink colouration. These observations suggest that the presence of DOP facilitates the mobility of Cu–PAN complex within the PIM, promoting migration towards the adjacent gel phase under high‐current conditions. In contrast, the DOP‐free PIM appears to favour retention of Cu(II)‐PAN complexes. On the basis of these results, the PIM formulation without DOP was selected for further experiments.

3.4. Performance of the Optimised EKE System for Cu(II) Extraction

3.4.1. Extraction With Optimal PIM Composition

The optimised PIM composition was CTA:D2EHPA:PAN at a weight percentage of 43.3, 56.0, and 0.8 respectively. The optimised PIM demonstrated good storage stability, with no observable leakage of the liquid phase or changes in colour after storage for up to 6 months under laboratory conditions.

As the extraction efficiency of Cu(II) from silica sand is closely related to the electric current passing through the EKE system, the optimal extraction time using the optimised PIM was determined. An extraction time of 2 h was found to be optimal, as no further colour change was observed at longer extraction durations (Figure S9). No observable PIM damage or shrinkage of the gel was observed under the optimised operating conditions, whereas higher voltages (≥40 V) resulted in partial gel shrinkage due to increased heating.

A calibration graph was constructed using the optimised conditions with silica sand spiked with 0–75 mg kg⁻¹ of Cu(II) (Figure S10A). Various regression models were evaluated to fit the calibration data. Using a linear model, the quantification range was determined to be 2.8–15 mg kg⁻¹ of Cu(II) with a high degree of fit (R ² of 0.9998). Lack‐of‐fit analysis indicated no significant deviation from linearity (p = 0.560). However, deviation from linear behaviour was observed at higher concentrations. A second‐order polynomial model was, therefore, applied. The quantification range was extended to a higher concentration, up to 30 mg kg⁻¹ of Cu(II), with a good R ² of 0.9985. The polynomial model also showed no significant lack of fit (p = 0.326). The LOD and LOQ were calculated to be 0.9 and 2.8 mg kg⁻¹ of Cu(II), respectively.

Model suitability was further evaluated by comparing the predicted Cu(II) concentrations obtained from the calibration models with independently prepared spiked silica sand. Within the quantification range, the binomial model provided more accurate predictions compared to the linear model (Figure S10B). Therefore, the binomial model was selected for further experiments.

The extraction efficiency was compared between the initial (unoptimised) and optimised PIMs by quantifying Cu(II) in silica sand using AAS before and after the EKE of Cu(II) from silica sand spiked to 1.5 mg kg⁻¹. With the optimisation of PIM, the extraction efficiency increased from 64% ± 7% (Cu(II)_PIM = 5.1 ± 0.5 µg; Cu(II)_sand = 3.3 ± 0.5 µg) to 90% ± 4% (Cu(II)_PIM = 7.1 ± 0.3 µg; Cu(II)_sand = 0.6 ± 0.1 µg). Although the quantification range did not extend with the optimised PIM, the extraction efficiency increased.

3.4.2. Application to Tailings Samples

To allow for a direct comparison with the previously reported unoptimised EKE system [17], tailings samples were re‐analysed with the optimised system. With the optimised system, the dilution factor was successfully reduced to 50, as the quantification range for Cu(II) extended to higher concentrations (Table 2). Furthermore, the optimised EKE system also demonstrated improved accuracy and precision in predicting water‐extractable Cu(II) in tailings samples.

TABLE 2.

Cu(II) concentrations detected in samples using various extraction methods.

Extraction method	Dilution factor	Cu(II) (mg kg−1)
		Sample BD	Sample ML
HNO3 digestion + AAS (acid extractable)	N/A	2770 ± 40	5900 ± 400
Water extraction + AAS (water extractable)	N/A	268 ± 7	242 ± 3
EKE—unoptimised EKE system a	100	560 ± 70	320 ± 60
EKE—unoptimised PIM	20	285 ± 20	N/A b
EKE—optimised PIM	50	280 ± 30	210 ± 40

Note: The results in electrokinetic extraction (EKE) were scaled back up to their original Cu(II) concentrations. EKE conditions: extraction for 2 h at 20 V; 13.5 mm column diameter; 0.35 g of samples; 15% moisture level and standard deviation (SD): N = 3.

Abbreviation: PIM, polymer inclusion membrane.

^a Result from [17].

^b Experiment was not done on this sample as the method was not optimised.

The EKE results obtained for the sample BD using both optimised and unoptimised PIMs were comparable. Despite the optimised system operating at a higher electric current, the amount of Cu extracted remained the same. This indicates that further increases in the extraction driving force do not recover additional Cu(II) beyond the electrokinetically extractable fraction, even when total acid‐extractable concentrations are significantly higher. This supports the fact that the electrokinetically extractable fraction obtained with the optimised system is suitable to use for the determination of the water‐extractable Cu(II) fraction.

3.4.3. Comparison Against Conventional Methods

When comparing the EKE method to established laboratory‐based instrumentation, such as AAS or ICP‐MS, the latter provides trace‐level detection (<µg kg⁻¹) with high analytical accuracy [8, 23]. In contrast, the EKE system has a much higher detection limit (mg kg⁻¹) with a larger SD. Although portable versions of AAS and ICP‐MS exist, their high cost and limited portability often hinder rapid field‐screening [24].

Although the sensitivity and quantification range of the EKE system are lower than laboratory methods, they are sufficient for screening environmentally relevant contamination levels. Our EKE method is more comparable to portable techniques such as pXRF, as they share a similar detection limit [8, 25].

Nascimento et al. [21] reported Cu concentrations in the sediment samples from the same area, detecting between 500 and 5000 mg kg⁻¹ of Cu downstream from the Mt. Lyell mining site. These values align more closely with acid‐extractable totals than with water‐extractable proportions, indicating pXRF is better at estimating total elemental concentration. In contrast, the EKE system developed in this study specifically targets the environmentally mobile Cu fraction via an applied electric field. This allows for a more targeted assessment of environmental risk that bulk‐analysis techniques like pXRF may overlook.

3.4.4. Future Direction

The EKE system offers several key advantages for on‐site screening applications, including low cost, portability, minimal sample preparation and the potential for visual or smartphone‐based detection. Overall, the proposed system is not intended to replace high‐precision laboratory techniques but rather to serve as a complementary, field‐deployable screening tool.

Although good agreement was observed between electrokinetically extractable and water‐extractable Cu(II), the applied electric field and the presence of neutralising gel may facilitate limited mobilisation of weakly adsorbed or exchangeable Cu species. This can occur through electromigration and localised pH gradients generated during electrolysis or from the gel.

Similar to pXRF, the EKE method remains susceptible to matrix effects [26]. In this study, system development and optimisation were performed using a simplified solid matrix with low loam and organic matter content. Commercially sourced silica sand allowed proof‐of‐concept studies under controlled and repeatable conditions with a fixed moisture content. Application to complex environmental samples may present additional challenges as they can influence Cu(II) adsorption and electroosmotic transport behaviour. In addition, natural variability in soil moisture content may alter the extraction current and, therefore, the electrokinetic behaviour and extraction efficiency of Cu(II).

To address these matrix‐dependent effects, future work will focus on validating the system across soils with diverse physicochemical properties, including contaminated soils and tailings, to assess its robustness under realistic environmental conditions. This will also include the use of certified reference materials to strengthen confidence in the method.

Furthermore, the portability and field deployment of the EKE system will be assessed. This includes the handling and long‐term stability of the system, especially the PIM and neutralising gel, under field conditions. To enhance its utility as a rapid screening tool, further optimisation will aim to reduce the extraction time (2 h) while maintaining stable and reproducible colour responses. Finally, inter‐operator reproducibility will be assessed to ensure the system's suitability for routine environmental monitoring by different users.

4. Conclusions

The optimisation of column diameter and PIM had successfully extended the quantification range of Cu(II) extracted from silica sand, from 1.2–9.4 to 2.8–30.0 mg kg⁻¹. The system achieved a Cu(II) recovery of up to 90%, demonstrating its efficiency when dealing with low loam content samples. The optimised EKE system successfully quantified water‐extractable Cu(II) in tailings samples with a lower dilution factor, while also providing improved accuracy and precision in Cu(II) concentration measurements. Overall, the EKE system enabled Cu(II) quantification in solid samples without requiring complex sample preparation. Additionally, it is a greener alternative to traditional soil analysis methods, as the extraction operates at low voltages. The immobilisation of solvents within solid structures (PIM and neutralising gel) enhances safety during handling. Furthermore, the integration of RGB image analysis allows Cu(II) quantification without a detector, making this method potentially suitable for low‐cost, on‐site screening applications.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: elps70120‐sup‐0001‐SupMat.docx.

Data Availability Statement

Data are available on request from the authors.