Carbon fibers (diameter: 12 μm, length: 5 mm) were purchased from Cytec Engineered Materials (West Paterson, NJ, USA) and used for CFME elaboration. Standards of FNT (97%), MNP (98%), potassium ferrocyanide (99%), potassium mono- and dihydrogen phosphate (99%), sulfuric acid (98%), and sodium hydroxide (98%) were purchased from Sigma-Aldrich. Absolute ethanol (99.9%) and acetonitrile (99.9%) were purchased from AnalaR NORMAPUR. All solutions were prepared using deionized water (pH 6.5, conductivity < 0.1 μS/cm, and DOC < 0.1 mg/L).

A helical steel auger with 99% screw-on tools was employed for neutral soil sampling. A RETSCH brand analytical sieve with a 75 µm mesh, certified according to ISO 3310/1 standards, was utilized for sieving neutral soil samples. An OHAUS brand analytical balance with a precision of 10⁻⁴ g was employed to weigh the samples. For the separation of liquid and solid phases following extraction, a CU-5000 IEC brand centrifuge was used. Extracts were concentrated using a BüCHI brand R-200 and B-490 rotary evaporator.

A Rotating Tubular Divider TT-RSD200 brand was used for sampling replicates of blank soil and spiked soil samples.

An Agilent Technologies chromatograph equipped with an Agilent J&W HP-5 column (30 m × 0.32 mm × 0.25 µm) and a Nitrogen Phosphorus Detector was utilized for reproducibility assessment. GC-NPD conditions:

Injection volume 1 µL, splitless mode at 150˚C (held for 1 min), ramped at 10˚C/min to 260˚C; temperature of the detector 310˚C, carrier gas He at 1.5 mL/min. FNT retention time: 13.348 min.

A PalmSens Instrument PC233 Potentiostat (Netherlands), controlled using PalmSensVs1.60 software, was used to carry out voltammetric measurements.

PSLITE 1.7.3 software was used to control data management.

Stock solution (1 g·L⁻¹) of FNT was prepared by dissolving the equivalent mass of FNT in absolute ethanol. Working solutions of FNT at 10 mg/L were then prepared by suitable dilution. 0.1 M PBS (pH 7.2) was used as the supporting electrolyte.

Electrochemical experiments were performed using square wave voltammetry (SWV), which offers higher sensitivity and faster scan rates than differential pulse voltammetry, making it suitable for trace-level FNT detection in complex matrices [15].

Measurements were performed using a PalmSens 3 type potentiostat (Netherlands) connected to a personal computer using Ivium PC and PSLite software. A three-electrode configuration was employed, consisting of a naked carbon fiber microelectrode as the working electrode (CFME) (diameter Φ = 12 μm), a saturated calomel electrode (SCE) as the reference electrode, and a platinum wire as the auxiliary electrode. The carbon electrode surface was renewed using a homemade electrochemical treatment. Indeed, the carbon fiber microelectrode was first pretreated electrochemically in a mixture of sulfuric acid H₂SO₄ (0.5 M)/ethanol (50/50 w/w) followed by a treatment in 0.1 M Phosphate Buffered Saline (PBS) pH = 7.2 using the following conditions: potential scanning rate: 100 mVs⁻¹ in the potential range −1.8 to +1 V for 20 cycles. Quality control of the CFME cleanliness was performed using visual and electrochemical tests. Electrochemical experiments were conducted in a 50 mL glass voltammetric cell at room temperature.

The appropriate solutions were transferred into the electrochemical cell. The analytical procedure for SWV was then optimized by systematically studying the experimental parameters that affect the responses, such as the pulse potential frequency related to the total pulse duration, amplitude of the pulse, and height of the potential step.

Scanning was performed from −1.8 to +1 V vs. SCE, with a step potential of 10 mV, an amplitude of 60 mV, and a pulse potential frequency of 60 Hz.

Before each experiment, the solutions were deaerated by bubbling nitrogen, and the electrochemical cell was maintained under a nitrogen atmosphere throughout the experiments.

For example, to record the SWV voltammograms of FNT, 50 mL of the supporting electrolyte was introduced into the electrochemical cell. Next, the supporting electrolyte was bubbled with nitrogen and stirred simultaneously for ten (10) minutes. The SWV voltammogram of the blank, represented by PBS, was then recorded ten seconds after stopping the stirring and bubbling of the solution.

Analytical curves were obtained for the pure electrolyte using the standard addition method.

This was carried out in 0.1 M PBS at pH 7.2 and at varying concentrations of FNT.

LOD and LOQ were calculated using formula (1):

whereS₀ is the standard deviation.

Recovery (%) was calculated using formula (2):

FNT residue levels in spiked soil were assessed using formula (3) and (4):

C_soil: concentration of FNT residues in soil;C_cell (µg·L⁻¹): concentration of residues in the cell, evaluated using the calibration curve;C_theor: theoretical value expected from spiked soil;m_soil: mass of soil used for analysis.

V_stand: volume of FNT standard solution used to spike soil.

A 20 g soil sample with a predetermined moisture content was placed in a 250 mL Erlenmeyer flask. Acetonitrile was selected based on its high extraction efficiency for organophosphates in soil [8]. The mixture was stirred for one hour (optimized in preliminary tests to maximize recovery) and centrifuged at 500 rpm for 15 minutes to separate the solid and liquid phases. At the end of centrifugation, two phases were separated: the solid phase consisting of the soil and a liquid phase or aliquot, represented by the extracts. The aliquot was collected, filtered, and evaporated to dryness using a rotary evaporator at a temperature of approximately 50˚C. The dried residues were then solubilized with a minimum amount of absolute ethanol and transferred to a 50 mL flask. The flask was then filled with phosphate buffer at pH 7.2. The solution thus obtained was then transferred into the electrochemical cell for analysis.

The electrochemical behavior of FNT was studied using square wave voltammetry, and the obtained signal is shown in Figure 3.

Figure 3 shows that no peak was observed in the SWV voltammogram of the PBS. However, in the presence of FNT (100 μg·L⁻¹), a well-resolved peak is observed at −1.02 V versus SCE. This peak is characteristic of the reduction of the nitro group (-NO₂) of FNT to a hydroxylamine group (-NHOH) according to the electrochemical equation below [13].

Figure 2. Electrochemical reduction of nitro group to hydroxylamine group [13].

Figure 3. Electrochemical behavior of PBS and FNT 100 µg·L⁻¹; pH = 7.2; frequency = 10 Hz; amplitude = 100 mV; increments = 20 mV; reference electrode: ECS.

The validation of the reported electrochemical method for the determination of fenitrothion (FNT) in soil was conducted comprehensively in accordance with EURACHEM/CITAC guidelines [20] and ISO/IEC 17025 [19] requirements, ensuring its fitness for purpose for quantitative analysis. The method underwent rigorous evaluation for key performance parameters, including ruggedness, specificity/selectivity, matrix effects, linearity, accuracy, precision, limits of detection (LOD) and quantification (LOQ), and measurement uncertainty.

According to IUPAC recommendations relating to harmonized guidelines for single-laboratory validation of methods of analysis [21], the key parameters of a method must be validated according to relevant criteria. Thus, this method will be validated when the following are met for the method to be adequate for its intended use:

The calibration function must describe an optimal line of best fit, minimizing the sum of squared residuals.Precision < 5% RSD, at a 95% confidence interval.No detectable bias at 95% confidence interval.Limit of detection < 10 μg∙kg⁻¹, with RSD < 10% at the 95% confidence interval.Accuracy ranges from 80% to 120%, with RSD < 10% at the 95% confidence interval.No significant interference with the matrix.Any potential interferences identified and, if needed, compensated for.Critical reagents used for method validation must be stable during the period covering the tests.Measurement uncertainty < 10%, at a 95% confidence interval.

3.2.1. Ruggedness Test

Ruggedness evaluation was carried out using the Plackett-Burman experimental design (Table S1, Table 1, and Table 2). The standard deviation (S) of the results for spiked soil samples containing FNT at 60 μg∙kg⁻¹ was previously estimated under repeatability conditions as 1.07 μg∙kg⁻¹ based on eight determinations. The critical difference ( Δ crit ) is calculated according to the following formula (5):

The results showed that factors such as pH, pulse potential frequency, pulse amplitude, and increment height significantly influenced the analytical response. These critical factors were subsequently optimized. The optimal values were: pH = 7.2; pulse potential frequency of 10 Hz; pulse amplitude of 100 mV; and an increment of 20 mV (Figure 4).

Table 1.Parameters evaluated in the ruggedness test.

Table 2. Evaluation of results from a ruggedness study of the analytical process.

Figure 4.Optimal values of pH, amplitude, frequency, and increments.

3.2.2. Specificity/Selectivity Assessment

FNT is detected based on the reduction potential of its nitro group. To ensure the specificity of the method, the voltammogram of FNT 100 µg∙L⁻¹ was recorded in the presence of electroactive compounds that could potentially interfere, such as profenofos, p-nitrophenol, p-aminophenol, 3-methyl-4-nitrophenol (MNP), dimethoate, S O 4 2 − , N O 3 − , Na⁺, K⁺ ions. The results in Table 3 show that, apart from MNP, none of the studied interferents has a significant effect on the detection of FNT.

MNP, a major FNT metabolite in soil, strongly interferes with FNT detection, as both compounds exhibit similar reduction potentials. This limits the method’s selectivity in aged or heavily contaminated soils where MNP may coexist with FNT. Future work should explore chromatographic pre-separation or chemometric deconvolution to resolve this interference.

The interference of these two molecules has already been reported in many previous works and could be explained by the fact that the reduction potential of MNP is very close to the reduction potential of FNT [22][23].

Table 3. Effect of interferents on the detection of FNT.

3.2.3. Matrix Effects

The matrix effect was evaluated based on the contribution of the residual current from the matrix to the faradaic current. Matrix effect affects the intercept and the slope of a calibration function and then sample quantification by increasing or decreasing the signal attributed to the measurand. To overcome matrix effect, standard addition has been used for calibration, and the value of the intercept has been evaluated to be equal to zero (Table 5).

However, a formal matrix-matched calibration study (comparing slopes in solvent vs. soil extract) was not performed; future work should confirm the absence of signal suppression or enhancement.

3.2.4. Assessment of the Calibration Function

Linear regression was used to assess the expected relationship between the analytical response and the concentration of FNT. The assays were carried out by the standard addition method, adding increasing volumes of the standard stock solution of FNT. A fresh stock solution of FNT was prepared each day, and all assays were performed under controlled light conditions to prevent photodegradation.

To confirm the reliability of the linear regression model, three sets of independent quality control samples at 15 µg∙L⁻¹, 30 µg∙L⁻¹, 50 µg∙L⁻¹, 90 µg∙L⁻¹, and 150 µg∙L⁻¹ were prepared by spiking from the standard stock solution of FNT. Control was carried out over a three-day period to improve precision and provide additional checks on the calibration solution.

In the absence of an independent standard solution, the same FNT standard solution was used to prepare the quality control solutions.

The plot of peak current versus the respective concentration of FNT was found to be linear in the working range (10 - 200 µg∙L⁻¹) and was represented by the linear equation: I p = 0.0022 × C ( μg ⋅ L − 1 ) + 0.0065 with R² = 0.9997 (Figure 5). The statistical analysis of the data, based on the assumption that there is no relationship between variables at a 95% confidence interval, showed that the slope coefficient is significantly different from zero (Tables S2-S4, Table 4 and Table 5).

Furthermore, data analysis showed that the optimal line of best fit minimizes the sum of squared residuals, indicating that at least one predictor variable has a non-zero coefficient. That assumes linear regression is a useful model.

Moreover, the statistical analysis of quality control data showed that there is no significant difference between variances of QC data collected over three-day periods (Table S5 and Table 6).

Figure 5. Square wave voltammetry of different concentrations of FNT in PBS 0.01 M, pH = 7.2, frequency = 10 Hz, amplitude = 100 mV, increments = 20 mV, FNT concentration range: 10 - 200 µg∙L⁻¹.

Table 4. ANOVA calculations for least-squares linear regression.

Table 5. Statistical analysis of linear regression data.

Table 6. ANOVA table for Levene’s test for QC data.

3.2.5. Assessment of Method Accuracy

The bias of the method was evaluated using independently spiked blank soil samples and expressed as the relative recovery rate of the spiking. Specifically, five concentration levels—30 μg∙kg⁻¹, 60 μg∙kg⁻¹, 90 μg∙kg⁻¹, 120 μg∙kg⁻¹, and 150 μg∙kg⁻¹—were prepared from 500 g of blank soil.

To do this, an appropriate volume of the FNT standard solution was added to 100 g of blank soil to obtain the concentration corresponding to each level. Subsequently, each of the 100 g of fortified soil corresponding to each level was used to prepare three replicates using a Rotating Tubular Divider TT-RSD200.

All 15 independent spiked samples were subjected to the extraction procedure in the presence of five soil blank samples, as described in subsection 2.4 on “Extraction and Clean-up of FNT Residues”.

The aliquots were analyzed, and C_soil (μg∙kg⁻¹)—corresponding to the difference between the mean concentration of the spiked sample and the unspiked sample—was then calculated. Recovery (%) was finally calculated against matrix-matched standards using formula (2) above.

Experimental data and results of statistical analysis of the data are shown in Table S6, Table S7, and Table 7.

The results showed that there is no bias inherent in the method.

Recovery rates of 89.12% - 94.76% (Table 7) suggest that the extraction and dilution procedure minimizes matrix effects for the tested soil. However, validation was performed on a single neutral soil type; applicability to acidic, alkaline, or high-organic-matter soils requires further investigation.

Table 7. Statistical assessment of method accuracy.

3.2.6. Assessment of Method Precision

The precision of the method was evaluated by means of a nested design. Four batches of spiked blank soil samples were used to prepare four groups of eight independent samples. Samples were prepared from soil samples spiked to 60 μg·kg⁻¹ by adding an appropriate volume of FNT standard solution. Each group of samples was analyzed under repeatability conditions.

Three series of eight samples were analyzed over three different days using three different carbon fiber microelectrodes.

The fourth series of eight samples was used for reproducibility assessment, carried out by GC-NPD analysis using an Agilent J&W HP-5 column (30 m × 0.32 mm × 0.25 µm). The FNT peak was found at 13.348 min.

The results of the paired t-test showed that there is no significant difference between both analytical methods (Table S9), indicating that the method offers good reproducibility.

Statistical analyses of data (Table S8) are given in Table 8, Table S10, and TableS11.

Table 8. Statistical analysis of precision data.

3.2.7. Assessment of Limit of Detection (LOD) and Limit of Quantification (LOQ)

Due to the lack of a measurable signal obtained from soil blank samples, test samples with concentrations of analyte were prepared by spiking soil blank samples with a small volume of the FNT stock solution.

So, ten (10) independent measurements of the test samples were prepared and analyzed under repeatability conditions according to the whole measurement procedure, including any sample preparation steps. The standard deviation (S₀) was then calculated. No baseline correction was carried out on the test results.

The statistical analysis of the results is summarized in Table 9.

The LOQ was then validated by analyzing three spiked samples at the quantification limit, yielding recoveries of 91.43 to 93.87 μg∙kg⁻¹.

Table 9. Experimental data for LOD and LOQ assessments.

3.2.8. Assessment of the Stability of the FNT Stock Standard Solution

The stability of the standard stock solution of FNT (10 mg∙L⁻¹) was evaluated by measuring the current of a fresh 100 μg∙L⁻¹ quality control sample prepared from the standard stock solution of 100 mg/L FNT using the standard addition method. The stock solution was conserved at 7˚C for three months. Measurements were carried out once a week over a period of three months.

The stability of FNT in ethanolic stock solution was confirmed over three months at 7˚C (Table 10). However, FNT stability in spiked soil samples during storage was not assessed. To ensure accurate results, soil samples should be analyzed promptly after collection, or a stability study should be conducted to establish acceptable storage conditions.

Statistical analysis of the data in Table 10, Tables S12-S14 confirms the homogeneity of the variances of the data collected over three-month periods.

Table 10.Statistical analysis of data for stock solution stability assessment.

3.2.9. Measurement Uncertainty

Measurement uncertainty was estimated using the GUM approach [24].

Measurement uncertainty was estimated by integrating contributions from weighing, volumetric measurements, calibration, and reagent purity, resulting in an expanded uncertainty of U = 0.04399 × C_soil (Table S15).

The uncertainties associated with each component were combined accordingly:

The expanded uncertainty U ( x ) is obtained by applying a coverage factor of 2, based on a 95% confidence interval.

The results of the uncertainty assessment, which include both random and systematic errors, show that U ( x ) × 100 C soil ( μ g kg ) = 4.399 % . This value, being less than

10%, suggests that the true value lies within a smaller range. This suggests that the validated method is accurate and precise.