Differential pulse voltammetric assessment of phthalate molecular blocking effect on the copper electrode modified by multi-walled carbon nanotubes: Statistical optimization by Box-Behnken experimental design

2.1 Chemicals and reagents

MWNTs with a purity of >95 wt% (carbon nanotubes), outer diameter of 20–30 nm and the length of 10–30 μm were obtained from Nanosany corporation (Mashhad, Iran). Phosphoric acid 85% (w/w), iron (II) chloride tetrahydrate (FeCl₂·4H₂O) 99% (w/w), iron (III) chloride hexahydrate (FeCl₃·6H₂O) 99% (w/w), potassium chloride 99% (w/w), sodium chloride 99% (w/w), were purchased with high purity from Merck (Darmstadt, Germany). Sodium hydroxide, DBP, DMP, DCHP and DEHP were purchased from Sigma-Aldrich (Milwaukee, WI, USA).

The electroencephalography gel was purchased from Abzar darman corporation (Tehran, Iran). Based on EDAX results, the MWCNTs have C -more than 97% (w/w) and Co, Cl, S and Al elements lower than 3% (w/w).

Stock solutions of PAEs (1000 mg L⁻¹) were prepared by dissolution of known amount of each chemical in methanol. Moreover, a stock solution of Fe²⁺ ions (10⁻³ mol L⁻¹) was prepared by dissolving a certain amount of FeCl₂·4H₂O in doubly distilled water. Phosphate buffer (0.01 mol L⁻¹, pH = 7) was utilized for pH adjustment of the solutions, and KCl, NaCl, KNO₃ and Na₂SO₄ solutions (1 mol L⁻¹) were prepared and used as the supporting electrolyte.

2.2 Instruments

All the electrochemical experiments were performed with a Potentiostat-Galvanostat instrument (µAutolab, Netherlands) equipped with a three-electrode cell consisting of nanosensor, Ag/AgCl, and platinum rod as the working, reference and auxiliary electrodes, respectively. FT-IR spectrometer (Shimadzu, Model 8900, Japan) was used to characterize the functional groups of the MWCNTs before and after injection of PAEs. Crystal structure of the MWCNTs was determined by an X-ray diffractometer (XRD, Model PW1730, Philips, the Netherlands) using Cu Kα radiation (λ = 0.1541 nm) at 2θ angle in the range of 10^◦ to 80^◦, and energy dispersive X-ray spectroscopy (EDS, Model MIRAII, TESCAN, Czech Republic) was employed for the elemental analysis. Porosity of MWCNTs was determined by Brunauer–Emmett–Teller analysis (Model BELSORP MINI II, BEL, Japan) via heating up to 450 °C in vacuum. Moreover, the morphology and size of the MWCNTs were characterized by field emission scanning electron microscopy (FESEM, Model MIRAII, TESCAN, and Czech Republic). A Metrohm pH meter (model 827, Switzerland) was utilized for pH adjustments.

2.3 Modification of copper electrode surface with MWCNTs

In this work, a cylindrical copper electrode (I.D. = 7 mm, O.D. = 10 mm) was modified by MWCNTs and used as the working electrode for trace determination of PAEs by DPV. To modify the copper electrode, a certain amount of MWCNTs was mixed with the electroencephalography gel and the copper electrode was uniformly coated by the smooth and adhesive paste. The modified electrode was then air-dried for 10 min and the fabricated nanosensor was used for PAEs determinations via blocking the surface of the electrode and reduction in the DPV signal of the Fe²⁺/Fe³⁺ anodic peak.

2.4 Analytical procedure for determination of PAEs

Electrochemical determination of the PAEs was performed based on their surface blocking effect on the fabricated nanosensor for oxidation of ferrous ions. In each experiment, 5 mL Fe²⁺ solution (10⁻³ mol L⁻¹), 1 mL phosphate buffer (0.01 mol L⁻¹, pH = 7), and 4 mL supporting electrolyte solution (0.1 mol L⁻¹) were added to the electrochemical cell. Then, the differential pulse voltammogram was recorded in the potential range of −1.0 to 1.0 V with a scan rate of 0.1 V s⁻¹ and the anodic peak current was measured for Fe²⁺/Fe³⁺ as the redox probe. The same procedure was repeated after injecting 20 µL of each PAEs solution into the surface of the modified electrode by a microsyringe, which resulted in a decrease in the anodic peak current for Fe²⁺ ions. Finally, the difference between the measured peak currents (Δi) was recorded as the DPV response. After each experiment, the modified working electrode was rinsed with pure methanol, sanded and washed again with pure methanol to ensure removing the modifier from the electrode surface. Finally, the electrode was placed in doubly distilled water. Surface modification of copper electrode and the electrochemical system are schematically indicated in Fig. 1.

Close

Fig. 1

Schematic image of modification of Cu electrode by MWCNTs.

Export to PPT

2.5 Optimization of the conditions

In the study, the univariate method was first employed to determine the optimum value of the solution pH, type of the supporting electrolyte, and volume of the DBP solutions. In the experiments, the anodic peak current of the Fe²⁺/Fe³⁺ pair was recorded before and after injection of DBP into the electrode surface to obtain the DPV response (Δi). Further studies were then conducted through designing the experiment and subsequent statistical analyses using Design-Expert software (version 11.0). A Box-Behnken design with four factors (modifier to gel ratio, solution pH, concentrations of Fe²⁺ and the supporting electrolyte) was used for statistical optimization of the electrochemical process. The studied factors and their levels can be seen in Table 1 and the BBD matrix is given in the Table S1.

3.1 Characterization of MWCNTs modified copper electrode

BET analysis was used to measure the specific surface area (SSA) and the porosity of the MWCNTs. The bigger the surface area, the higher chance of Fe²⁺ electrooxidation on the modified surface of electrode. The results were: SSA = 62.21 m² g⁻¹, total volume of pores = 0.473 cm³ g⁻¹ and specific diameter of each pore = 30.4 nm. These findings are so promising for on-going modification steps.

Fig. 2 shows the XRD pattern of MWCNTs which displays a peak at 2θ ≈ 26° corresponding to the 002 plane of the graphitic structure of the modifier. Also, the peaks at 43° and 54° correspond to the 100 and 004 reflections (Atchudan et al., 2015). These characteristic peaks confirm the structure of MWCNTs.

Close

Fig. 2

XRD pattern of MWCNTs (a) and ref pattern (b).

Export to PPT

To visualize very fine topographic properties of the modified surface of electrode, before and after PAEs injection, low-kinetic-energy electrons penetrated to the immediate material surface employed via FESEM. The images of MWCNTs and MWCNTs-PAEs were shown in Fig. 3. The nano-scale dimensions of the MWCNTs are completely obvious in the first image (a). Comparison of the images suggests that no significant change occurs in the surface structure of the MWCNTs after spiking by PAEs.

Close

Fig. 3

FESEM images of MWCNTs before (a) and after spiking by DEHP (b), DBP (c), DCHP (d), and DMP (e).

Export to PPT

Adequate mechanistic information regarding nanosensor-blocker interaction is so complicated to obtain. In spite of this obstacle, FT-IR was used to study the interaction. The spectra of raw MWCNTs, nanosensor and PAEs-injected nanosensor were shown in Fig. 4. In Fig. 4a, several peaks at 1725 cm⁻¹, 2353 cm⁻¹ and 3400–3600 cm⁻¹ are good indications for the carbonyl, carboxylic and hydroxide groups of MWCNTs electrocatalyst, respectively. After blending the MWCNTs with EEG, the fabricated nanosensor was analysed as shown in Fig. 4b. The carbonyl stretching vibration appeared at 1725 cm⁻¹. In the other region, a decrease was observed on the COOH peak around 2352 cm⁻¹. Moreover, additional peaks in the range of 3400–3700 cm⁻¹ were detected. Here is the start of a challenging part of mechanistic insight into interactions. The EEG gel is mainly based on polyacrylate and sodium chloride. The sodium acrylate ingredient is a suitable candidate for Na⁺exchange with strong H-bonded (COOH) on MWCNTs electrocatalyst. The proposed reactions are summarized below:

Close

Export to PPT

Close

Fig. 4

FT-IR spectra of MWCNTs before spiking (a), MWCNTs-EEG (b) and MWCNTs after spiking by DEHP (c), DBP (d), DCHP (e) and DMP (f).

Export to PPT

It seems that the above-mentioned reactions and high coverage of the surface (because of the high weight fraction of EEG in the nanosensor) are responsible for the remarkable signal decrease around 2355 cm⁻¹. Also, since the proposed reactions are in equilibrium, new peaks were created in the OH range (3400–3700 cm⁻¹). Fig. 4c-d correspond to the blocker-injection step. It’s clear that the injection of various blockers carries the peaks, somewhat.

The carbonyl stretching vibration was observed at 1628 cm⁻¹ (DCHP), 1626 cm⁻¹ (DMP), 1718 cm⁻¹ (DEHP) and 1720 cm⁻¹ (DBP). The conjugated carbonyl in cases of DCHP and DMP tend to lower FT-IR wavenumbers. The major alteration occurred in the OH region (3400–3700 cm⁻¹). These changes in OH signals may be related to an oxygen-hydrogen interaction of the phthalate carbonyl group and MWCNTs surface hydroxyl. This idea is in agreement with previously reported one (Baghayeri et al., 2014; Yu et al., 2014; Gupta et al., 2013; Lehman et al., 2011; Rausch et al., 2010; Ren et al., 2011; Abdel-Ghani et al., 2015). At a pH higher than 6, the surface of the MWCNTs electrocatalyst is negatively charged. This negative surface attracts Fe²⁺ ions via immigration. As a result, this immigration amplifies the total oxidation current of Fe²⁺. After blocker injection, the negative sites on MWCNTs electrocatalyst surface (O⁻), fall in electrostatic attraction with partially positive carbon of carbonyl groups of PAEs. This phenomenon not only blocks the negative sites (O⁻) of the electrocatalyst (MWCNTs), but also creates spatial hindrance on interface of the nanosensor.

3.2 Optimization strategies

As previously mentioned, in preliminary experiments, the optimum levels of the solution pH, type of the supporting electrolyte and volume of the DBP solution were determined by the univariate method. Fig. 5 displays the effect of the solution pH on the value of Δi showing the maximum value of Δi at pH = 8. Therefore, this pH was chosen as the optimal pH level. The effect of the supporting electrolyte type on the value of Δi in DBP solution is illustrated in Fig. 5. As indicated, KCl (0.1 mol L⁻¹) resulted in the best response and was chosen as the supporting electrolyte. Additionally, Fig. 5 illustrates the effect of the DBP solution volume on the Δi value. As shown, the best response belonged to the 20 µL of DBP solution and then, this volume was selected as the optimum volume.

Close

Fig. 5

The results of optimizing the effects of solution pH, the volume of DBP solution, and type of the supporting electrolyte by the univariate method.

Export to PPT

In this study, BBD as a response surface design was utilized to obtain the optimum levels of the parameters affecting the electrochemical response of the modified electrode. The effects of the experimental variables including the modifier to gel mass ratio (0.05–0.09), solution pH (7.2–8.2), Fe²⁺ concentration (250–550 μmol L⁻¹), and supporting electrolyte concentration (0.05–0.15 mol L⁻¹) on the response (Δi) were investigated to determine the best combination of variables for the analyses. The statistical parameters of the polynomial models developed for the determination of DBP are shown in Table 2. As shown, the highest values of the correlation coefficient (R²), adjusted correlation coefficient (Adj R²) and predicted correlation coefficient (Pred R²) were observed for the quadratic model developed for the electrochemical response. A value close to 1 for R² indicates a strong relationship between the experimental and predicted responses. Furthermore, adjusted R² (0.9187) indicated an improvement in fitting with the use of a quadratic model. The predicted R² value (0.8537) matches with the adjusted R². The quadratic response surface model in terms of actual factors is as follows:

The statistical significance of the developed quadratic model was evaluated by ANOVA. Table 3 shows the results of ANOVA for the quadratic model. According to the results, p-value lower than 0.0001 for AD, BD, CD, and A² indicates significant effects of the factors on the response. Moreover, the parameters C, D and C² showed significant effects at a certainty of >99%.

Normal probability plot of the residuals and the plot of the predicted versus actual responses are shown in Fig. 6. As shown in Fig. 6a, the normal probability plot of residuals is linear that confirms the model adequacy. Also, because of the small residuals, it can be concluded that there are no possible irregularities. According to Fig. 6b, the model predicted Δi values are relatively close to the diagonal line showing goodness of fit of the developed model for predicting the electrochemical responses and confirmed that the obtained responses are very close to the predicted responses.

Close

Fig. 6

Normal probability plot of residuals (a) and plot of the residuals versus predicted responses (b).

Export to PPT

Fig. 7 shows the perturbation plot exhibiting the effect of each of the independent variables on the response. This figure is useful to compare the effect of all factors in the experiment and each response was plotted by changing one factor over its entire range with other factors as constants (Maran and Manikandan; 2012). As shown in Fig. 7, parameter A has more deviation than B, C and D from the reference point. Also, B and C have small deviations comparatively. A steep slope or curvature in the plots suggests the sensitivity of the response factor on the variable (Oh et al., 1995).

Close

Fig. 7

Perturbation plot showing the effects of experimental variables on the response. A: Fe²⁺ concentration, B: pH, C: supporting electrolyte, D: modifier/gel ratio.

Export to PPT

The three-dimensional response surface plots for the significant two-factor interactions of AD, CD, and BD are presented in Fig. 8. The interaction effect of the modifier/gel ratio and solution pH on the response (C_Fe²⁺ = 400 µM and C_KCl = 0.1 M) is shown in Fig. 8a. The plot shows that with increasing the modifier/gel ratio, the response of the electrode increased. Also, the figure indicates that pH effect on the response depends on the modifier/gel ratio. Fig. 8b shows the combined effects of modifier/gel ratio and the supporting electrolyte concentration at pH = 7.7 and C_Fe²⁺ = 400 µM. As it can be seen, the response increases by increasing the supporting electrolyte concentration and the effect of modifier/gel ratio is similar to plot a. Fig. 8c indicates the interaction effect of modifier/gel ratio and Fe²⁺ concentration on Δi value at pH = 7.7 and 0.1 M KCl. Obviously, the response decreases by increasing the Fe²⁺concentration. Also, above 430 µM of Fe²⁺, the response increases with increasing the concentration.

Close

Fig. 8

3D response surface plots for two-factor interaction effects of modifier/gel ratio and pH (a), modifier/gel ratio and supporting electrolyte concentration (b), modifier/gel ratio and Fe²⁺ concentration (c).

Export to PPT

The main purpose of the response surface technique was to obtain the optimal values for each factor that cause maximum response (Δi). To obtain the optimal conditions, four model predicted experiments were conducted, and the experiment for which the response was close to the predicted answer by the design was selected. According to the percentage of error obtained, one experiment was selected as the optimal condition and was used for further experiments. Comparison of the error percentage, predicted and experimental values were shown in Table 4. Therefore, Fe²⁺ concentration of 402 µM, pH = 8.2; and concentration of 0.13 M for KCl as the supporting electrolyte was selected as the optimal condition.

3.3 Electrochemical impedance spectroscopy

Surface modification of the nanosensor and injection of PAEs to its modified surface will change the resistive nature of the interface. The electrochemical impedance spectroscopy was used to touch these changes during above-mentioned steps.

Fig. 9(a) shows the Nyquist plot of the nanosensor. Based on the circuit fitting and simulation, the constructed circuit can be drawn as Fig. 9(b). Analysis of the proposed circuit resulted to these figures: total resistance (R_T) = 654.8 KΩ and charge transfer resistance (R_CT) = 549.4 KΩ. The R_sei assigned for solid Fe(OH)₃ and electrolye interface layer due to trace formation of Fe(OH)₃ in just vicinity of the bare copper surface at pH = 7.0.

Close

Fig. 9

Nyquist plot of the bare copper electrode (a) and equivalent circuit model (b).

Export to PPT

From the Nyquist plot in Fig. 10(a), it is obvious that a new element should be added to the equivalent circuit of the modified surface of a nanosensor as drawn in Fig. 10(b). The R_nanosensor element comprises the modification step and an MWCNTs/EEG paste layer. In this case, R_T = 571.2 KΩ and R_CT = 489.3 KΩ. The R_SEI rises 1.5 times, which means well-built coverage of the bare surface of the copper electrode during the modification step. The modification step uses two materials; first, MWCNTs powder that will increase conductance as an electrocatalyst and second, EEG that amplifies not only charge transfer process but also paste conductance. These ingredients determine the resistance of prepared paste. According to above-mentioned data, the modified surface shows a −12.8% decrease in R_T and −10.9% decline in R_CT. The recent findings reveal the most important role of bare copper surface modification by MWCNTs/EEG, which makes electron transfer easier. This may be related to the large specific surface area, high porosity and electrocatalytic character of MWCNTs.

Close

Fig. 10

Nyquist plot of the modified electrode (a) and equivalent circuit model (b).

Export to PPT

In Fig. 11(a), the impedance spectrum of nanosensor after blocker injection could be seen. Injection of PAEs as blockers enlarges charge transfer resistance because of their blocking effect on the nanosensor’s surface. This remarkable increase caused the R_SEI to be lost inside of the semicircle part. Blocker injection not only alters the resistance (R_T = 665.8 KΩ and R_CT = 612.2 KΩ), but also eliminates one part of circuit as seen in Fig. 11(b). Both of surface modification and injection of blocker increase the resistance. Comparing with modified surface, these growths are equal to 16.6% in R_T and 25.1% in R_CT. For that reason (25.1% magnification of R_CT), blocker effectively interacts with the surface of nanosensor and blocks active sites of oxidation and consequently, oxidation peak appears on the lower position in DPV. Fig. 11 and its equivalent circuit lead us to the most interesting reason of this research.

Close

Fig. 11

Impedance spectra of nanosensor after injection of PAEs (a) and equivalent circuit model (b).

Export to PPT

The various R_CT values for PAEs were calculated and summarized in Table 5. The results in table, confirm the existence of correlation between PAEs nature and changes in resistance (ΔR_CT). In the other words, chemical structure of PAEs is a key factor affecting ΔR_CT. So, this can be used in characterization of various PAEs based on ΔR_CT at the same concentration level. By taking a closer look, spatial hindrance of PAEs plays major role in R_CT shifts. The smaller blocker provides the better point-by-point coverage of surface that concludes the bigger shift in charge transfer resistance.

The above-mentioned results, give us exciting sign of future opportunity for developing the impedimetric nanosensor of PAEs. This nanosensor will work based on the charge transfer resistance shifts versus changes of PAEs concentration. It is clear that the small changes in carbonyl group polarity will affect ΔR_CT and bring selectivity for impedimetric nanosensor of PAEs. Further research is needed to confirm these novel findings.

3.4 Electroanalytical performance of nanosensor

3.4.1

3.4.1 Figures of merit

The figures of merit provide valuable information about the sensitivity, precision and linear range of nanosensor. So, the figures of merit of the proposed nanosensor were determined under the optimum conditions for determination of PAEs. In this study, separate standard solutions of PAEs in the concentration range of 0.1 to 5000 µg L⁻¹ were used as blocker and the resulted DPV signals were measured. Correlation between the response (Δi) and the concentration of PAEs via DPV used as the calibration curve. A linear relation between response and C_PAEs were obtained in this range under the optimum conditions. The oxidation current was extracted from the differential pulse voltammogram and was given in Table 6 along with the concentrations of PAEs that show a good linear relationship between the oxidation current and the concentration. As shown in Fig. 12, different solutions of PAEs were used as blocker and the resulted DPV signals were measured. It was concluded that with increasing in the concentration of PAEs, the current decreases and the potential shifts to the right and positive values.

Table 6 Data obtained from differential pulsed voltammograms at different concentrations.

	−Δi (A) × 10+5
C (μg L−1)	DBP	DMP	DCHP	DEHP
0.1	1.52	4.68	2.52	1.28
0.5	1.7	4.72	2.58	1.37
1	1.85	4.77	2.64	1.45
3	2.38	4.89	2.83	1.68
5	2.73	5.10	3.14	1.86
10	2.92	5.17	3.17	2.00
15	3.07	5.35	3.22	2.13
25	3.22	5.52	3.47	2.31
50	3.45	5.75	3.64	2.55
75	3.69	5.92	3.87	2.75
100	3.86	6.14	4.01	2.99
250	4.05	6.33	4.25	3.17
500	4.33	6.61	4.37	3.28
750	4.47	6.78	4.66	3.53
1000	5.05	6.95	4.95	3.74
5000	5.87	7.16	5.12	3.97

Close

Fig. 12

Anodic peak current of Fe²⁺/Fe³⁺ pairs after addition of PAEs as blocker to the MWCNTs modified Cu electrode. (Fe²⁺ concentration of 402 µM, 0.10 M KCl, pH = 8.0, modifier to gel ratio = 0.09, and 20 µL of PAEs solution).

Export to PPT

Limit of detection and limit of quantification were obtained based on the following equations:

(2)

$$\mathrm{L}\mathrm{O}\mathrm{D}=\frac{3\mathrm{\sigma }}{\mathrm{S}}$$

(3)

$$\mathrm{L}\mathrm{O}\mathrm{Q}=\frac{10\sigma }{S}$$ where σ is standard deviation of five replicates of blank measurements and S is the slope of the calibration curve. Moreover, the linear dynamic range (LDR) was obtained from the linear correlations between the response and PAEs concentrations. Table 7 lists the LOD, LOQ, LDR and the linear regression equations.

Table 7 Results of the PAEs determination by the proposed method.

PAEs	LOD (µg L−1)	LOQ (µg L−1)	LDR (µg L−1)	Linear equation	Correlation coefficient (R2)
			0.1–5	y = −2 × 10−6x − 2 × 10−5	0.9836
DBP	0.004	0.01	10–100	y = −1 × 10−7x − 3 × 10−5	0.9747
			250–1000	y = −1 × 10−8x − 4 × 10−5	0.9252
			0.1–5	y = −8 × 10−7x − 5 × 10−5	0.9875
DMP	0.009	0.03	10–100	y = −1 × 10−7x − 5 × 10−5	0.96
			250–1000	y = −8 × 10−9x − 6 × 10−5	0.9813
			0.1–5	y = −1 × 10−6x − 3 × 10−5	0.9888
DCHP	0.008	0.03	10–100	y = −9 × 10−8x − 3 × 10−5	0.9594
			250–1000	y = −1 × 10−8x − 4 × 10−5	0.9707
			0.1–5	y = −1 × 10−6x − 1 × 10−5	0.9847
DEHP	0.008	0.03	10–100	y = −1 × 10−7x − 2 × 10−5	0.9779
			250–1000	y = −8 × 10−9x − 3 × 10−5	0.9800

3.5 Real samples

The real samples create a real challenge for the nanosensor. To evaluate the suitability of the proposed modified electrode for analysis of real samples, PAEs were measured in different samples including well water (Lakani Street, Rasht, Guilan, Iran), tap water (Rasht, Iran), sea water (Caspian Sea, Guilan, Iran) and river water (Shafarood River, Masal, Guilan, Iran). The samples were first filtered to remove any particulates and then were analyzed by the standard addition method. Based on the method, the water samples were separately spiked by 20 μL of each PAEs (1 µg L⁻¹) and the DPV responses were obtained at the optimum electrochemical conditions. Table 8 indicates the results of five replicated measurements. As shown, the recovery percentages were in the range of 98.0–100.0 with RSD values ranging from 1.4 to 6.3%, which confirmed the high accuracy and precision of the proposed method for determination of PAEs in real water samples.

3.6 The comparative performance

In order to show the comparative efficiency of nanosensor, the obtained results were tabulated with those reported by other researchers for PAEs determination (Table 9). As shown in Table 9, the detection limits of this work stay in the range of 0.004–0.009 µg L⁻¹ which is better than other studies. As is evident, the proposed nanosensor possesses high sensitivity and satisfactory performance for determination of PAEs in the aqueous solutions.