Selective solid-phase extraction of atrazine from agricultural environmental water samples using high permeability nanoporous carbon derived from melamine-based polybenzoxazine followed by HPLC-UV

ABSTRACT For the first time, a highly permeable nanoporous carbon derived from melamine-based polybenzoxazine (NPC-PBZ-m) was used as the solid-phase extraction sorbent for trace analysis of atrazine (ATZ) herbicide in environmental water samples as a pre-treatment to enhance detection for high-performance liquid chromatography. Extraction performance was investigated using an agricultural field matrix-water sample. The parameters affecting the solid-phase extraction efficiency were carefully investigated. The optimisation conditions were determined as 200.0 mg of sorbent with an adsorption flow rate of 10.0 mL min−1, an elution flow rate of 1.0 mL min−1, and an elution volume of 3.0 mL. Moreover, it was not necessary to adjust the pH of the sample. Under these optimised extraction conditions, good linearity between 5.00 and 30.0 µg L−1 was achieved along with a limit of detection (LOD) of 1.25 µg L−1 and a limit of quantification (LOQ) of 3.79 µg L−1. Furthermore, the reusability of the NPC-PBZ-m column was found up to 8 cycles making it a new cost-effective material for herbicide enrichment. Finally, this developed method was successfully applied to analyse ATZ in water samples from rice fields, fish farms, and natural canal water, with 98–117% relative recoveries. These results indicate that NPC-PBZ-m with our proposed method is feasible and demonstrates its potential application as an effective adsorbent for herbicide enrichment.


Introduction
Atrazine (ATZ) is a herbicide used to control pre-and post-emergence broad-leaved weeds.It is widely used in several countries such as the US, Canada, Africa, and Asia-Pacific includes Thailand [1,2].Because of its persistence, ATZ residue can be found in soils and water resources as a common environmental pollutant [3][4][5][6][7].Previous work reported that the contamination has increased by up to 30 µg L −1 [8], and many studies have addressed the harmfulness of ATZ [4,9,10].The US Environmental Protection Agency (USEPA) and the Water Environment Partnership in Asia (WEPA) have established the maximum contaminant level (MCL) for ATZ in water at 3.0 µg L −1 [5,11].However, ATZ concentrations in stream water and sediment samples were reported in some regions of Thailand to be above this level as high as 4.7 µg L −1 and 27.42 µg kg −1 [12].For this reason, the quantitative determination of ATZ is necessary for human health and environmental monitoring.
The determination of trace ATZ in agricultural water samples is particularly hard because of the complex matrices and low concentration of ATZ, which often requires a sample preparation technique to achieve an acceptable level of preconcentration and clean-up before instrumental analysis.There have been several methods reported for the extraction and determination of ATZ.They include liquid-liquid microextraction [13], dispersive liquid-liquid microextraction [14], dispersive solid-phase extraction [15][16][17], continuous sample drop flow microextraction [18], and supercritical fluid extraction [19].Although these microextraction techniques are fast and require less solvent and sample volume [20], they involve many working steps for extraction such as centrifugation, evaporation, and re-dissolving before analysis [18,20,21].Some of these techniques require a ternary solvent system [22,23] and many reagents such as surfactant, ion-pairing, or ionic liquid are combined with heating for effective extraction [21,[24][25][26].
Solid-phase extraction (SPE) has been one of the most widely used methods for sample clean-up and enrichment of ATZ.In addition, it can provide a high sample volume, can be combined with pre-existing automation techniques and has little to no problem with emulsion formation during the extraction process.Notably, the selectivity of target extraction can be improved by changing the solid material [27][28][29][30][31][32][33][34].The SPE column can be packed with various sorbent types, such as molecularly imprinted polymers [28,[35][36][37][38][39][40][41][42][43] or modified natural substances [44][45][46].Carbonaceous materials deliver excellent effectiveness among these sorbents due to the considerably higher surface area and adsorption capacity [27,47,48].However, previously reported materials are expensive, require an evaporation step and a vacuum manifold unit to conduct the extraction.Most of the reported SPE adsorbents are single-use materials, which are more costly and wasteful.Therefore, it is necessary to develop a reusable sorbent material with a high surface area and high pore accessibility to extract and enrich ATZ from water samples.
Nanoporous carbon derived from melamine-based polybenzoxazine (NPC-PBZ-m) is a new promising adsorbent with a high surface-to-volume ratio [49] and high pore accessibility due to its 3D nanoporous properties.It can be synthesised from polybenzoxazine (PBZ) via the sol-gel process [50,51].Phenol, amine, and formaldehyde are the starting reagents for benzoxazine monomers which then undergo ring-opening polymerisation during heat treatment to produce PBZ.PBZ exhibits excellent mechanicalchemical properties such as high thermal stability and high char yields with customisable molecular structures and properties which can be tailor-made by changing the starting reagents (either amine or phenol derivatives) [52,53].Furthermore, since ATZ can selectively interact with melamine and nylon6-pyrrole via hydrogen bonds and π-π interactions [54,55], as a result the NPC-PBZ-m can be an effective adsorbent for ATZ extraction.It is important to highlight here that the application of NPC-PBZ-m for the extraction of either ATZ or other analytes has not been previously reported.
In this work, high permeability NPC-PBZ-m sorbent was synthesised according to our previous work [49], packed inside an SPE column, and connected to a small syringe pump for a semi-automated extraction procedure.This new sorbent SPE approach was applied to extract ATZ before detection and quantification using HPLC-UV, and it was shown that the NPC-PBZ-m sorbent was reusable for up to 8 cycles.The influential factors of the SPE procedure were optimised, and after getting the optimal conditions, the developed method was successfully used to detect ATZ in environmental water samples.Thus, our proposed semi-automated SPE extraction approach has the potential as an in-field sampling that minimises the risk of degradation, contamination, and sample transportation.
The stock solutions of ATZ (100 and 500 mg L −1 ) and chlorpyrifos (500 mg L −1 ) were prepared in methanol.Paraquat and glyphosate stock solutions (500 mg L −1 ) were prepared in Milli-Q water (Millipore, USA).2,4-D stock solution (500 mg L −1 ) was prepared in methanol before making up the volume with Milli-Q water.All standard solutions were stored at 4°C.Working standard solutions were freshly prepared by diluting each standard solution with agricultural water collected from a rice field with no ATZ usage (Lopburi province, Thailand) as a matrix water sample which are used for the optimisation study.

Synthesis and characterisation of NPC-PBZ-m sorbent
The NPC-PBZ-m sorbent was synthesised based on our previous work [49].The synthesis procedure can be found in detail in the Supplementary Material file.The final obtained material (denoted as NPC-PBZ-m) was characterised again by FTIR, BET, XPS, and FE-SEM to confirm that the same material as in our previous work was obtained.The textural properties of NPC-PBZ-m are listed in Table S1.The NPC-PBZ-m sorbent was prepared by grinding the material with a household blender and sieving by test sieves No. 40-60 purchased from Rung Arun Machinery, Thailand.The particle size of NPC-PBZ-m sorbent was in the range 253-400 µm.

Instrumentation
The infrared spectra were measured with a Nicolet 6700 FTIR spectrometer (Waltham, MA, USA).The textural characteristics of NPC-PBZ-m were investigated by conducting the N 2 adsorption-desorption isotherm (Quantachrome-Autosorb-1MP).The sample was outgassed at 250°C for 8 h under vacuum before analysis.The specific surface area (S BET ) was determined by the Brunauer-Emmett-Teller (BET) algorithm [56].The total pore volume (V tot ) was calculated at a relative pressure of 0.992.Barrett-Joyner-Halenda (BJH) [57] and density functional theory (DFT) [53] were used to determine the pore size distributions of mesopores and micropores, respectively.The t-plot method was conducted to analyse micropore volume (V micro ) [58].The functionalities of nitrogencontaining carbons were analysed using an X-ray photoelectron spectroscope (XPS, Kratos Axis Ultra DLD).A field emission scanning electron microscope (FE-SEM, JSM-7610 F) was used to investigate the surface morphology of NPC-PBZ-m.The sample was coated with platinum under vacuum conditions before investigation.
The chromatographic determination of ATZ was conducted in an HPLC system from Agilent technology (Agilent 1260 Infinity II Quaternary pump) equipped with an automatic sampler, a degasser, a quaternary pump, a thermostated column compartment, and a UV-Vis diode array detector.Separation was performed using a ZORBAX Eclipse Plus C18 (150 mm × 4.6 mm, 5 µm particle size, Agilent Technologies Inc., USA).The sample injection volume and detection wavelengths were 20.0 µL and 222 nm, respectively [59].The flow rate was 1.5 mL min −1 with a column temperature of 30°C.An isocratic elution method was utilised using a mixture of MeOH/H 2 O (60:40, v/v) as a mobile phase.

SPE procedure
To assess the applicability of the synthesised material, 200 mg of the dry sorbent was packed into a polypropylene column (Bond Elut Reservoir, 1 mL) with a polyethylene frit at the bottom.Before loading the ATZ standard/sample solution, the SPE column was conditioned and equilibrated at a flow rate of 1.0 mL min −1 with 5.0 mL of methanol and 10.0 mL of methanol/water (1:1, v/v), respectively.Then, the standard/sample solution (300.0 mL) was loaded at a flow rate of 10.0 mL min −1 using a syringe pump (ProSense BV, model NE-1000, USA).After that, the column was washed with 3.0 mL of deionised water, and then air was flown through.Finally, 3.0 mL of methanol was flushed through the column at a flow rate of 1.0 mL min −1 to elute the ATZ before HPLC-UV analysis.The procedure of the proposed method is illustrated in Fig. S1.The laboratory practices, the use of personal protective equipment, and the method of waste disposal complied with the protocols provided by the Department of Chemistry and the Centre for Safety, Health and Environment of Chulalongkorn University.

Reusability of NPC-PBZ-m sorbent (SPE)
The reusability of a sorbent is significant merit to evaluate when determining the efficiency of sorbent materials for extraction.For this purpose, 300.0 mL of ATZ standard solution (30.0 µg L −1 ) prepared in water was used in this experiment.The adsorptiondesorption steps were conducted as described above (section 2.4) using the optimised parameters.After each cycle, the column was conditioned and re-equilibrated before performing the next cycle.It is worth clarifying here that a new sample was applied for each cycle.

The samples collection
Surface water samples from Nakhon Pathom agricultural field (rice field), aquatic fish farm, and natural canal waters in Saraburi province were collected in high-density polyethylene (HDPE) gallons and immediately transported to the laboratory.Suspended particles were removed by filtering through filter paper (Thomas baker Grade 1010, 125 mm diameter) and 0.45 µm cellulose acetate membrane filters (Vertical, 47 mm diameter) and then stored in HDPE bottles at 4°C.Since these samples did not show detectable levels of ATZ, they were spiked with various levels of the ATZ.

Sample preparation for LC-MS/MS validation method
The collected samples were filtered with 0.20 µm cellulose acetate membrane filters.Samples were analysed directly without any further dilution.In order to examine the relative recovery percentage, the samples were spiked with different concentrations of ATZ standards.

Optimisation of NPC-PBZ-m SPE protocol
To obtain the highest sensitivity towards ATZ determination, the parameters that could affect the extraction and preconcentration of ATZ, such as sample pH, the amount of NPC-PBZ-m, sample loading rate, and volume and flow rate of methanol, were investigated.For this purpose, 300.0 mL of the ATZ spiked water sample with a final concentration of 25.0 µg L −1 was pumped through the SPE column packed with NPC-PBZ-m (unless otherwise stated).All experiments were performed in triplicate, and the mean values of the resulting data were used for analysis.

Effect of sample pH on ATZ adsorption
Since the pH of a sample could affect the extraction efficiency and sensitivity; a series of pH values (3.0, 7.0, 9.0) were investigated.Herein, 100.0 mL of ATZ solution (5.0 µg mL −1 ) was prepared from 100.0 µg mL −1 ATZ stock solution and diluted with Milli-Q water.The pH of Milli-Q water was adjusted using either 0.1 M of HCl or NaOH.Then, 10.0 mL of the solution was added to a polypropylene tube containing 0.0100 ± 0.0009 g of NPC-PBZ-m.Adsorption was allowed to occur for 24 h using a rotating shaker in order to reach extraction equilibrium.The ATZ concentration in the filtrate was measured by a UV-Vis spectrophotometer at 222 nm.The adsorption percentage was calculated using the following formula where C 0 is the initial concentration and C t is the solution concentration after adsorption for 24 h.
As shown in Figure 1, there was no significant difference in the ATZ adsorption capability between these pH values indicating that the ATZ adsorption process is pHindependent.Typical pH values of the surface water samples were between 6.5 and 8.5, and therefore, pH did not need to be adjusted, and water samples could be analysed directly.Under a batch adsorption study at pH 7, the NPC-PBZ-m sorbent was shown to adsorb ATZ as much as 4.80 ± 0.05 mg/g.

Effect of NPC-PBZ-m amount
The amount of sorbent plays a vital role in adsorption performance since an increase in the amount of sorbent can supply more adsorption sites.Therefore, different masses of NPC-PBZ-m from 100.0 to 400.0 mg packed into the column were examined.The procedure was the same as that described above (section 2.4).As shown in Figure 2(a), the adsorption efficiency increased with the increasing sorbent, peaking at 200.0 mg.The proposed structure of sorbent (Fig. S2) has electron-donating functional groups, the  amine, and hydroxyl groups, and an aromatic ring promoting the adsorption.The readsorption phenomenon of ATZ could be the reason for the decrease in the adsorption efficiency of NPC-PBZ-m at the higher sorbent amounts (300.0 and 400.0 mg) [60,61].Consequently, 200.0 mg of NPC-PBZ-m was chosen as the optimal value for subsequent optimisations.

Effect of the sample loading flow rate
The impact of the sample loading flow rate on the adsorption efficiency and the extraction time was evaluated at 5.0, 10.0, and 12.0 mL min −1 .Herein, two levels of ATZ concentrations, 10.0 and 25.0 µg L −1 , were employed.Figure 2(b) demonstrates similar trends with the loading rates at both concentration levels.However, the highest ATZ adsorption efficiency was with a flow rate of 5.0 mL min −1 .This is perhaps due to the long contact time between NPC-PBZ-m sorbents and ATZ.Unfortunately, the total analysis time was long.On the other hand, the lowest adsorption efficiency was observed when the flow rate was as high as 12.0 mL min −1 because of the inadequate contact time between ATZ and NPC-PBZ-m sorbent leading to the loss of ATZ molecules from the sorbent during the sample loading.To compromise between the total analysis time and the adsorption efficiency, a sample loading flow rate of 10.0 mL min −1 was chosen for further studies.

Effect of volume and flow rate of methanol
The eluent solvent volume and its flow rate through the column can affect the desorption efficiency of ATZ from the sorbent.Therefore, these two parameters were examined.
First, the volume of methanol was varied from 3.0-5.0mL at a constant flow rate (1.0 mL min −1 ).The results showed that the increase in the volume of methanol led to a decrease in extraction efficiency (ATZ concentration in the eluate) due to the dilution effect.Thus, 3.0 mL was selected as the best volume.The flow rates of methanol were studied at 0.5 and 1.0 mL min −1 , revealing only slight differences in adsorption efficiency between these two flow rates (Data not shown).Therefore, a flow rate of 1.0 mL min −1 was chosen as the optimum flow rate.A summary of the selected conditions for the preconcentration procedure with the NPC-PBZ-m column is shown in Table S2.

Reusability of NPC-PBZ-m unit
The reuse of a sorbent can significantly reduce analysis costs.Therefore, the cycles number is a critical index to evaluate when determining the extraction performance of any newly developed extraction sorbent.In this work, the content of ATZ in the eluant after each preconcentration cycle was compared with that of the first cycle.The percentage difference (% difference) was calculated as (P cycle1 -P cycleN )/P cycle1 × 100, where P cycleN is the peak areas obtained after N times (N = 2 to 8) application of the column for preconcentration of ATZ and P cycle1 is that obtained from the first cycle.As shown in Figure 3, the NPC-PBZ-m column can be reused for ATZ preconcentration in a real matrix sample up to 8 cycles without a significant loss (the difference between the first and eighth cycle was less than 10%), which is an advantage compared to disposable adsorbents [62,63].

Adsorption selectivity
An adsorption selectivity study is essential to assess the feasibility of any developed method towards ATZ extraction.The effect of other pesticides, which are commonly used with ATZ, on the extraction efficiency of ATZ was investigated [64,65].To this end, a 0.0150 mg L −1 ATZ standard was mixed with 1.50 and 7.50 mg L −1 of paraquat, glyphosate, 2,4-D, and chlorpyrifos (concentration ratio of 1:100 and 1:500).Each solution was then preconcentrated using the NPC-PBZ-m column under the optimum conditions (n = 3).The concentration levels of either paraquat, glyphosate, 2,4-D, or chlorpyrifos added over ATZ were higher than the reported values of their MCLs [66][67][68][69][70]. Therefore, a percent difference of ±5% was used as the maximum allowance value to evaluate the influence of other herbicides on the adsorption selectivity [71].The percent difference was calculated as (C a -C x )/C a × 100, where C x and C a are the concentrations of ATZ after the preconcentration with and without the other pesticides, respectively.Our results showed that the concentration of paraquat and glyphosate could coexist as high as 500 times (7.50 mg L −1 ) while 2,4-D and chlorpyrifos concentrations can coexist up to 100 times (1.50 mg L −1 ) over ATZ concentration without interfering with the measurement of ATZ (Table 1).These results could indicate that our sorbent has sufficient adsorption selectivity to ATZ, and the developed method is feasible for ATZ extraction from real water samples.The chromatograms of interference testing are shown in Fig. S3.

Analytical features
The developed method's analytical features were examined and summarised in Table 2. Based on the calibration plot, residuals plot, response factors plot, and percent relative error of back-calculated concentrations (%RE) (Fig. S4), the linear working range was between 5.00 and 30.0 μg L −1 with a coefficient of determination equal to 0.9988.The limit of detection (LOD) and limit of quantification (LOQ) were estimated as follows: LOD = 3.3* sd intercept /slope and LOQ = 10* sd intercept /slope [72].The detection limit was 1.25 µg L −1 , which is lower than the MCL value (3 µg L −1 ) established by USEPA and WEPA for ATZ in water.A LOQ of 3.79 µg L −1 was observed.A sample spiked at the LOQ level was prepared, preconcentrated, and detected by the proposed method to estimate the relative standard deviation and relative recovery.The relative standard deviation (% RSD) was 9 (n = 6), and relative recovery (%RR) was 114.
Moreover, the repeatability of the developed method was evaluated by spiking water samples at 10.0, 15.0, and 30.0 µg L −1 .The intra-day precision (n = 9), as evaluated by the %RSD and %RR was between 2% to 4% and 97% to 111%, respectively.While inter-day precision (n = 3), as evaluated by the %RSD and %RR was between 3% to 6% and 99 to 106%, respectively, which are acceptable and considered practical.
Two surface agricultural water samples were collected from different provinces (Lopburi and Nakhon Pathom) in Thailand to assess the matrix effects on the proposed methods.These surface samples were spiked with ATZ and analysed with the NPC-PBZ -m/HPLC-UV method.The matrix effects of the substance were estimated by comparing the slope of each calibration curve and are reported as a slope ratio.The calculated ratio value should ideally be between 0.8 and 1.2 to indicate that the matrix effect for ATZ detection in various samples can be ignored [73,74].As shown in Table S3, the proposed method provided a slope ratio within the acceptable range.Therefore, it can be implied that the method's performance is not affected by the matrix of each water sample.
Moreover, the calibration curves were also obtained with a good coefficient of determination (>0.99) for all samples.The LOD values were <3 µg L −1 regardless of the sample matrix.These results demonstrated that the NPC-PBZ-m/HPLC-UV method has good sensitivity and reproducibility and can be reliably applied for ATZ analysis in different agricultural water samples.

Analysis of surface water samples
To evaluate the accuracy and feasibility of the proposed method for practical analysis, additional relative recovery tests were conducted using surface water collected from a rice field (in the Nakhon Pathom province of Thailand), a fish farm, and from natural canal water (both from the Saraburi province of Thailand).The results are shown in Table 3. ATZ was not detectable in these samples using the developed method.Therefore, these samples were spiked with different concentrations of standards.Relative recoveries of 98-115%, 104-107%, and 115-117% were obtained using our developed method for the rice fields, fish farms, and from the natural canal water samples, respectively, with %RSD values less than 6.
The water sample analysis results of our proposed method were compared with LC-MS /MS determination.The LC-MS/MS procedure was based on EPA Method 536.0 discussed in detail in the Supplementary Material file (Table S4).The results indicated that data values of the three water samples determined by our method are in satisfactory agreement with the results obtained from LC-MS/MS measurement (t exp = 0.73-4.25 < t crit = 4.30, n = 3 at 95% of confidence level).Suggesting that our developed method provides good precision and accuracy; i.e. our developed method will be applicable to the determination of ATZ in these water matrices.Figure 4(a-c) show the chromatograms of rice field water, fish farm, and natural canal water samples, respectively, before and after spiking with ATZ standards.

Comparison of the developed method with other reported methods
The developed adsorbent and analysis method was compared with previously reported methods that determine ATZ concentration in water matrices.As shown in Table 4, comparable relative recovery was achieved, verifying the accuracy and sensitivity of the method.The LOD of our method was as sensitive as other analytical methods for ATZ determination in this field.Moreover, the LOD achieved by this method is lower than the MCL value [5].Our developed method has a great advantage over other methods in that the NPC-PBZ-m sorbent column can be used for 8 cycles with a minimum percentage error of less than 10%.A reusable material would perfectly fit the context of on-line SPE approaches.Furthermore, the semi-automated extraction procedure presented in this work reduced the labour requirements.Overall, detecting target analyte using a simple sample preparation is of interest to readers in this field.

Conclusions
A selective and efficient approach was developed for sensitive analysis of ATZ in real environmental water samples based on the combination of a newly developed SPE sorbent and HPLC-UV analysis.NPC-PBZ-m sorbent was synthesised using a green solvent to develop an adsorbent with high permeability, suitable surface area, and good ATZ extraction performance.The procedural conditions for extraction were optimised using an experimental extraction setup that simulated the conditions in the field by using water collected from an agricultural field without atrazine usage.Moreover, our SPE unit was designed to operate with a small syringe pump, which was easy to handle, movable for on-site preconcentration application, and cost-effective.The NPC-PBZ-m sorbent could be reused for up to 8 adsorption/desorption cycles.The calibration curve was linear between 5.00 and 30.0 μg L −1 , while the LOD was 1.25 μg L −1 .It is worth mentioning here that the LOD obtained by our method is lower than the authorised maximum contaminant level (3.0 µg L −1 ) of ATZ in waters.Furthermore, the developed method can be applied for ATZ analysis in different water samples without pH adjustment, providing good recoveries.To the best of our knowledge, this is the first study of the use of NPC-PBZ-m sorbent for ATZ extraction and enrichment in environmental samples.

Figure 1 .
Figure 1.Effect of the sample pH on the adsorption efficiency.

Figure 2 .
Figure 2. Effect of different factors on the peak areas of the ATZ investigated by using HPLC-UV at 222 nm; (a) the amount of NPC-PBZ-m; (b) the sample loading rate.

Figure 3 .
Figure 3.The reusability of NPC-PBZ-m column for ATZ adsorption under the optimised conditions, investigated by using HPLC-UV at 222 nm.
The linear dynamic range was 5.00-30.0µg L −1 b The relative standard deviation was evaluated by spiking water sample with three concentration levels (µg L −1 )

Figure 4 .
Figure 4.The chromatograms of before and after, 10.0 and 20.0 µg L −1 , of ATZ spiking in rice field water (a); before and after, 8.00 and 25.0 µg L −1 , of ATZ spiking in fish farm water (b); before and after, 10.0 and 20.0 µg L −1 , of ATZ spiking in natural canal water (c).

Table 1 .
Effects of the potential pesticide interferences on the ATZ preconcentration using the NPC-PBZ-m SPE column under the optimum conditions.The concentration of ATZ after the preconcentration step followed by HPLC-UV detection. b

Table 2 .
Analytical features of the proposed method.

Table 3 .
Application of the developed method to determine ATZ in surface water samples.