Introduction

One of the major challenges of the global society is the lack of potable water or the increasing volume of waste and contaminated water (Warren-Vega et al., 2023). Consequently, the global society is currently stretching environmental, economic, industrial and domestic avenues to overcome the challenges arising from water contamination (Kumar et al., 2021). Some water contaminants have a short life span because they are biodegradable, but most contaminants that are nonbiodegradable or persistent are on the lists of emerging that should be given intensive research concerning remediation approaches because of their established toxicological documentation (Pérez-Lucas et al., 2023; Zahmatkesh et al., 2022)). Some emerging contaminants such as dyes, antibiotics, pesticides and other chemicals are useful in agricultural, scientific laboratories and several industrial sectors (Kaczorowska et al., 2023). This suggests that their production and availability are on the increase and point to the fact that increasing production, demand and supply should be adequately matched by remediation approaches (Silva et al., 2020; 2023).

One of the most vulnerable components of the environment that are worse impacted by the continuous discharg of toxic wastes is the aquatic environment because most contaminants are soluble in water and can easily be transported directly or indirectly to other components of the ecosystem (Caixeta, 2023). Consequently, research scientists around the globe have designed, tested and implemented several water treatment technologies (such as adsorption, ozonolysis, photodegradation, ion exchange, filtration, oxidation) to counter the forces of contamination (Sedighi et al., 2023). Pesticides are useful chemicals in the agricultural sector for the protection of crops against destruction, poor yield, food insecurity and associated economic and industrial problems (Hagner et al., 2019). Regardless of their utility or benefits, most pesticides (including glyphosate) are persistent in the environment and have an inherent tendency to be toxic to aquatic life, mainly when transported in solution to aqueous systems (Maderthaner et al., 2020). Some toxic impacts of glyphosate have been established and reported, including increased salivation, internal burns and cancer (Peillex and Pelletier, 2020). To remediate this impact, some laboratory-scale investigations have confirmed that among the available methods, an adsorption process is an excellent approach for the removal of toxic chemicals (including pesticide residue) from aqueous solutions (Odoemelam et al., 2018). Consequently, several conventional adsorbents have been established to be effective in the decontamination of water such as biochar showing removal efficiency of over 80% for atrazine, chlorothalonil and beta endosulfan from water (Kalsoom et al., 2022). Barbosa et al. (2023) also observed the effectiveness of sugar cane biomass as an adsorbent for the removal of fipronil from water with optimum removal efficiency exceeding 80%. The work of Naghdi et al. (2023) employed a porous metal–organic framework to remove glyphosate from water with an attended adsorption capacity of 440 mg/g, while Diel et al. (2022) observed efficiency approaching 90% with carbon nanotube as an adsorbent. Based on the above-reported work, there are some common challenges facing the selectivity, durability and efficiencies of adsorbents including irreducible particle size to enhance adsorption, the environmental requirement of non-constitution of primary or secondary contaminants, non-thermal stability, lack of structural information for enhanced modelling, inaccessibility, reusability and weak surface properties such as porosity and surface area (Gkika et al., 2022). Some of the listed shortcomings have impeded the economic cost of employing several adsorbents, especially at a commercial scale.

Of significant interest concerning the efficiency of an effective adsorbent is the requirement for minimal particle size and enhanced surface area (Cosgrove et al., 2022). Size modification below the micro-dimension may be impossible within the framework of most of the common adsorbents (Mazari et al., 2021). Therefore, the production of adsorbents with particle sizes of 1 to 100 nm (i.e. nanoparticles) can significantly boost adsorption capacity. Another challenge that concerns the water purification industry is the toxicity of some adsorbents, including some nano-adsorbents (Garg et al., 2021). For example, Hubbe (2022) and An et al. (2015) reported the toxicity of cellulose and jujube-based adsorbents and called for more intensive research attention towards managing the toxicity of the adsorbent. This implies that removing contaminants from water may be accomplished by the introduction of newer and foreign contaminants (Eddy et al., 2022a). As a result, long-term efforts must be made to find adsorbents that will not leach toxic components into the aqueous medium. Consequently, there is a current need to design and apply eco-friendly adsorbents to remove pesticide residue (such as glyphosate) from the aquatic environment (Ghavamifar et al., 2023).

The International Union of Pure and Applied Chemistry (IUPAC) nomenclature for glyphosate is ((3-hydroxy-2-oxopropyl) amino) methyl phosphonic acid. It is an active and standard ingredient in most pesticides, and its commercial value is highly appreciable because of its outstanding effectiveness as a pesticide (Diel et al., 2022). The toxicity of this product has been confirmed, especially its impact on body weight, skeletal composition, gastrointestinal systems and initiation of cancer and death (Tarazona et al., 2017). Unfortunately, this toxic compound can easily be ingested through contaminated food, drinking water and other routes (Da’ana et al., 2021). It can equally affect the aquatic environment through a reduction in primary productivity (and a retardation of the amount of light penetrating the water and subsequent retardation of photosynthesis), nutrient enrichment and the termination of the aquatic population (Peillex & Pelletier, 2020). Given the closeness of the contamination route of glyphosate, its common usage, its toxicity and the need to employ eco-friendly adsorbents for their removal, our current study is aimed at the employment of the precipitation method to synthesize, modify, characterize and apply zinc oxide nanoparticles (ZnO-NP) to eliminate glyphosate from aqueous solution. The chemical structure of glyphosate is shown below.

figure b

The structure suggests some inherent tendency of the compound towards adsorption based on the presence of heteroatoms (O, N and P), adsorption-prone functional groups and pi-electrons (Eddy et al., 2022b).

The success of zinc nanoparticles rather than zinc oxide nanoparticles has also been reported for the removal of glyphosate residue from contaminated water with recorded optimum efficiency of 70%. However, zinc nanoparticles can be very reactive with some contaminants, unlike zinc oxide nanoparticles which are relatively stable (Al-Arjan, 2022). This suggests that the generation of secondary contamination is more feasible with ZnNPs than ZnO-NPs. Some studies have also shown that the adsorption capacity of nanoparticles such as ZnO-NPs can be enhanced through doping or capping with other compounds. For example, Hosseini and Toosi (2019) observed that graphene–titanium oxide nanocomposites yielded removal efficiency extending from 53 to 73% for some herbicides (including 2,4-D, glyphosate compared to TiO2 or graphene that showed efficiency not exceeding 48%. The only literature on the viability of zinc oxide nanoparticles (ZnONPs) as an adsorbent for glyphosate is the one reported by Páez et al. (2019). Their work indicated that ZnO-NPs were synthesized by precipitation methods and applied in the removal of glyphosate-based herbicides from water. The efficiency describing the adsorption removal technology assumes an optimum value of about 90%. The study showed some merit in the application of the nanoparticles and also opens room for further modifications that can thrive towards the enhancement of efficiency; some are the investigation of environmental variables affecting the observed adsorption, the stabilization of the adsorbent by doping and the establishment of the theoretical or computational chemistry descriptors that can be useful in the prediction of future adsorption descriptors and other structure-based predictors. Therefore, in this study, we are consolidating the research need for the enhancement of the adsorption capacity of ZnO-NPs through doping as a predictor for better adsorption removal of the toxic herbicide, which is popularly used in several countries of the world including Nigeria and India. The investigation of environmental variables that can affect the adsorption characteristics of the ZnO-NPs as well as the establishment of theoretical based descriptors is also incorporated into the work and shall form the baseline information for the first time.

Materials and methods

Sample preparation and synthesis

All chemicals were purchased from BDH Limited, Poole, England, and Aldrich Company, Germany. The chemicals were used as delivered and without further purification. The precipitation method described by Mohan and Renjanadevi (2016) was adopted for synthesizing the ZnO-NP using zinc acetate dehydrate and NaOH as precursors. The first stage of the reaction involves the production of sodium acetate and zinc hydroxide acetate (Eq. 1). In contrast, the second stage involves further interaction of the zinc hydroxyl acetate and sodium hydroxide (Eq. 2) to produce ZnO:

$${{{Z}}{{n}}\left({{{C}}{{H}}}_{3}{{C}}{{O}}{{O}}\right)}_{2}+{{N}}{{a}}{{O}}{{H}} \to {{Z}}{{n}}\left({{{C}}{{H}}}_{3}{{C}}{{O}}{{O}}\right)\left({{O}}{{H}}\right)+{{N}}{{a}}{{{C}}{{H}}}_{3}{{C}}{{O}}{{O}}$$
(1)
$${{Z}}{{n}}\left({{{C}}{{H}}}_{3}{{C}}{{O}}{{O}}\right)\left({{O}}{{H}}\right)+{{N}}{{a}}{{O}}{{H}} \to {{Z}}{{n}}{{O}}+ {{N}}{{a}}{{{C}}{{H}}}_{3}{{C}}{{O}}{{O}}+ {{{H}}}_{2}{{O}}$$
(2)

The synthesized zinc oxide nanoparticles were washed with deionized and distilled water severally and filtered with a Whatman No. 1 filter paper. It was washed with 20 ml of 96% ethanol and dried in an oven at 100 °C for 4 h before calcination. A portion of the prepared zinc oxide nanoparticles was modified using oleic acid and xylene, as described by Hong et al. (2006). The remaining portion was also modified with aluminium oxide as described by Yuan et al. (2005).

To prepare stock solutions of glyphosate, 1 g of the pesticide was dissolved in 1000 ml of distilled water to achieve a concentration of 1000 mg/L. Using a UV/VIS 1800 PC spectrophotometer, a calibration curve was developed using various working concentrations (2, 3, 4, 4.5 and 6.5 mg/L) prepared from the stock solutions.

Adsorption removal techniques

According to Yuan et al. (2005), batch adsorption experiments were conducted using 10 mg/L pesticide solutions (except for experiments investigating the effects of initial concentration, in which different concentrations were used: 2.0, 3.0, 4.0, 4.5 and 6.5 mg/L. Each analysis used 0.6 g of adsorbent, except for the study that investigated the effects of adsorbent dosage, which used various adsorbent masses (0.1, 0.2, 0.4 and 0.6 g) and a fixed pesticide concentration (10 m/g).

A UV/VIS 1800 PC spectrophotometer was used to measure absorbance to determine the amount of pesticide adsorbed and the difference between the initial (inlet) concentration and the equilibrium concentration (after adsorption, i.e. \({C}_{0}-{C}_{e}\) was recorded as the amount of pesticide adsorbed. \({C}_{0}\) is the initial concentration, while \({C}_{e}\) is the concentration at equilibrium. The percentage concentration of pesticide removed was calculated using Eq. 3, while the equilibrium amount of pesticide adsorbed was evaluated using Eq. 4 (Ali et al., 2022):

$$\%Removal= \frac{{C}_{0}-{C}_{e}}{{C}_{0}} \times \frac{100}{1}$$
(3)
$${Q}_{e}=\frac{{C}_{0}-{C}_{e}}{{C}_{0}} \times \frac{m}{V}$$
(4)

where m is the mass of the adsorbent used for the study and V is the volume of the solution

Characterization of ZnO-NP

The Fourier transformed infra-red spectrum (FTIR) of the ZnO-NPs was obtained by scanning the sample in a FTIR instrument (Shimadzu FTIR-54005) between the wave number of 4000 and 400 cm−1. The spectrum obtained was presented as peaks of absorbance against wave number. The ZnO-NPs were also characterized using the XRD (model PANalytical X’Pert Pro) which operated at a voltage peak of 45 kV and a regulated current of − 35 mA. The operation also employed a Cu–Kα radiation (K = 1.5406 A˚) and 2θ ranging from 20 to 80° at a continuous speed of 0.045° per min. The morphological analysis of the nanoparticles was implemented using a scanning electron microscope (Model JEOL JSM6100) with an accelerating voltage of 20 kV after super coating within a layer of gold.

Quantum computation

ChemBio version 18 was used to draw chemical structures. Optimization of the drawn chemical structures was done using DFT models in the Materials Studio program, which was also the software used in the computation of the Fukui function through Mulliken and Hirshfeld charges (Q). The DFT/B3LYP/6-311G** level of theory was used for the calculations of semiempirical parameters such as frontier molecular orbital energies, total molecular energy and dipole moment. The adsorption energy and associated energy functions such as rigid adsorption energy, deformation energy and others were obtained through Monte Carlo simulation modelling positioned as an adsorption locator in the Materials Studio program.

Results and discussion

Characterization of the synthesized ZnO-NP

Figure 1 shows the scanning electron micrographs (SEM) of (a) ZnO-NPs, (b) ZnO-NPs modified with alumina and (c) ZnO-NPs modified with oleic acid and xylene. The micrographs were obtained based on the resolution (500 ×) indicated in each figure. The examination of the micrographs (Fig. 1) reveals that the modification changed the morphology of the ZnO-NPs. The unmodified micrograph reveals less clumsiness in the nanoparticles. However, modification with alumina reveals more evenly arranged spherical particles within the exclusive and relatively broad adsorptive sites with different morphological appearances, at the bottom left (largest diameter), centre (average diameter) and top right (most minor diameter).

Fig. 1
figure 1

Scanning electron micrographs of a ZnO-NP, b MAlO and c MOLA

Full size image

Also, the ZnO-NPs modified with oleic acid and xylene gave a relatively better and more ordered morphology than the unmodified ZnO-NPs. From the appearance or morphology of the three micrographs, there is a development of a more ordered and defined structured appearance that can be attributed to modification. Some reports have been documented on the morphology of nanoparticles after modification. For example, Egbuchunam and Balkose (2012) observed that ZnO-NPs consisted of spherical particles and were monodisperse, but modification with inorganic compounds increased aggregation.

In the work of Mohan and Renjanadevi (2016), the morphology of ZnO-NPs synthesized showed a grain-like and spherical appearance, while Kamal et al. (2023) reported a hexagonal shape for ZnO-NPs.

The X-ray diffraction spectra of the ZnO-NPs, ZnO-NPs modified with alumina (MAlO) and ZnO-NPs modified with oleic acid (MOLA) are shown in Fig. 2a to c, respectively. The XRD spectra (Fig. 2) reveal sharp peaks for crystalline ZnO-NPs with a hexagonal wurtzite structure, consistent with the range of values reported elsewhere (Xin, 2022). The observed sharp peaks indicate an excellent crystallinity of the ZnO-NP at 2θ = 8.4874°, 12.6734°, 16.5172°, 28.4438°, 33.0871° and 59.3223°, while the rough surface indicates the amorphous areas with relatively broad peaks.

Fig. 2
figure 2

XRD spectra of (a) ZnO-NP, (b) MAIO (ZnO-NP modified with alumina) and (c) MOLA (ZnO-NP modified with oleic acid and xylene)

Full size image

According to Zhou et al. (2017) and Khoshhesab et al. (2011), the diffraction peaks at 33.0871° and 59.3223° are associated with the hexagonal wurtzite phase of ZnO. Peaks at 8.4874, 12.6734 and 16.5172 indicate the presence of Zn(OH)2 and ethanol in the drying process. A reduction in peak intensity (but an increase in the number of peaks) of ZnO-NPs was observed due to the inclusion of aluminium oxide (Fig. 2b), which contributed to the increase in the amorphous character of the ZnO-NPs (Gogoi et al., 2020). However, the peaks are much broadened, and the crystallinity of ZnO-NPs tends to decrease, while peaks were observed at 2θ = 31.1748° to 67.3585°. On the other hand, the weak crystallinity of the ZnO-NPs after oleic acid inclusion occurred at 2θ = 33.0113° (fig. 2c).

In view of the diffraction peaks observed for the nanoparticles, significant information can be obtained from the XRD spectrum regarding nanoparticle dimension through the application of the Scherrer equation given as Eq. 5 (Hong et al., 2006):

$$Crystalline\;size \left(nm\right)=\frac{k\lambda }{Bcos\theta }$$
(5)

where k is Scherrer’s constant (k = 0.94 for spherical crystallites with cubic symmetry), λ is the wavelength of the X-ray, β is the full width at half maximum of the XRD peak and θ is the diffraction angle, deduced from the XRD peak position, i.e. 2θ.

The crystallite sizes of the unmodified nanoparticles were estimated as 32.25, 18.79, 7.35, 16.39, 8.48 and 7.01 nm, which gave an average of 18.05 nm. Also, the crystallite sizes of the MOLA were 18.27, 14.20, 22.98, 9.65 and 5.90 nm (average = 14.2 nm). A similar trend was observed for the MAlO, whose corresponding crystallite sizes were 46.0, 31.13, 7.04, 21.43, 10.26, 11.84, 4.69, 6.71, 4.35 and 5.53 nm (average = 14.50 nm). The peaks around 33.0871 and 59.3222° are typical for the wurtzite structure of ZnONPs (Pelicano et al., 2017).

The analysed XRD data also confirmed that the modification of ZnONPs with alumina changes the peak at 33.0871 to 33.1362°, while the peak at 59.3222° is absent in the modified ZnO-NPs. Relatively new peaks were formed at 20.24, 29.24, 3.17, 36.39, 40.72, 45.62 and 61.35°. However, modification with oleic acid changes the peak from 33.09° to 33.01°, while the original peak of 12.65° was changed to 12.07°. The increase in the number of peaks (and their broadness) due to alumina and oleic acid modification makes the ZnO nanoparticles less crystalline (and more amorphous) (Zhao et al., 2021). It can be seen from the XRD spectrum that the modification with alumina tends to increase the crystallite size and the number of peaks more than what was found in the ZnO-NPs. In contrast, modification with oleic acid reduces the number of peaks and the particle sizes of the nanoparticles. Also, the alumina tends to make the ZnO-NP more ordered, while the oleic tends to increase the crystallinity of the synthesized nanoproduct. There were also observable changes in peak intensities after the modifications, showing a slight decrease for alumina-modified ZnO-NPs and an increase for the oleic acid-modified ZnO-NPs.

Figure 3 shows the FTIR spectra of the ZnO-NPs, MAlO and MOLA. Figure 3 shows the FTIR spectra of glyphosate, MAlO, MOLA and ZnO-NPs. The spectra reveal a general observation concerning the nanomaterials: enhanced absorption due to modification with alumina (MalO) and oleic acid (MOLA). The FTIR of glyphosate (Fig. 3) reveals OH stretches at 922 and 12,630 cm−1, PO2 stretches at 1198 and 1087 cm−1, CO stretches at 1745 cm−1, and NH2 stretches at 1600 cm−1. Also, the FTIR spectrum of ZnO-NP indicated the presence of vibration due to ZnO at 479 cm−1 and hydroxyl stretches at 3572, 3467, 3549 and 3467 cm−1 (Dangana et al., 2023), possibly due to traces of NaOH and H2SO4 that were used for the preparation (Handore et al., 2014).

Fig. 3
figure 3

FTIR spectra of glyphosate (GLY), ZnO-NPs, ZnO-NPs modified with alumina (MAlO) and ZnO-NP modified with oleic acid and xylene (MOLA)

Full size image

O = C = O stretches were observed at 2348 and 1030 cm−1 (Hong et al., 2021), and C = C stretch was observed at 1598 and 831 cm−1. C-O stretches were observed between 1022 and 1123 cm−1 (Tiwari et al., 2022). These stretches might have resulted from traces of unreacted acetate ions. However, after adsorption of the pesticide, the OH stretches at 3572 and 3439 cm−1 were shifted to 3536 and 2501 cm−1, respectively, while the functional groups at 3467 and 3549 cm−1 were shifted to 3463 and 3422 cm−1, respectively. These changes indicate the interaction between the ZnO-NP and the glyphosate. Also, new OH stretches were formed at 3881, 3866, 3803, 3746 and 3646 cm−1, indicating the involvement of these bonds in adsorption. The C-O, C = C and C-H stretches were not found in the spectra of the adsorbent after adsorption, which also suggests their interaction and subsequent involvement in bond formation.

After modification of the adsorbent with oleic acid, the FTIR of MOLA reveals almost similar functional groups but at different positions, including OH stretch at 3421 cm−1 but not at 3572, 3467, 3549 and 3467 cm−1 that were found in ZnO-NPs. This OH stretch was shifted to 3536 cm−1, contrary to the multiple stretches in ZnO-NPs. The observed difference may be partly attributed to the reducing action of oleic acid. C-O stretch at 1025 cm−1 was missing a functional group. However, new C-O bonds were formed at 1455 and 1404 cm−1 indicating their involvement in adsorption. After the glyphosate adsorption, the C-H bond at 984 cm−1 and the O = C = O stretch were also missing in the MOLA spectra. Some of the newly introduced vibrations might have been introduced by the adsorbed glyphosate, for example, NH2 stretches.

The MAlO FTIR spectrum after the adsorption of glyphosate revealed multiple OH stretching (at 3646, 3463, 3422, 3881, 3866, 3803, 3463 and 3422 cm−1) as shown in the results, which suggest the relative contribution of OH stretching in the adsorption of the glyphosate. The OH stretches must have been due to those from the glyphosate and the adsorbents. However, the C-H stretches (at 1499 and 1429 cm−1) and the C-O stretch (at 11,322 cm−1) were missing in the spectrum of the adsorbent after the adsorption of glyphosate, which also confirmed their involvement in the adsorption process.

It is also evident from the results presented that some functional groups were missing after adsorption, while others experienced changes in the frequency of adsorption. Also, new functional groups were formed. The observed changes and modifications are evidence of interaction and adsorption through the employment of functional groups, as reported elsewhere (Eddy et al., 2022c). For example, the reported stretches at 3572 and 3472 cm−1 in ZnO-NP shifted after adsorption shifted to 3536 and 3501 cm−1, respectively. The O = C = O stretch at 2348 cm−1 in the ZnO-NPs was not found in the FTIR spectrum of the adsorbent (ZnO-NPs) after adsorption. Multiple OH stretches formed new bonds (3881, 3866, 3803, 3746 and 3646 cm−1). These new bonds were probably used in the glyphosate adsorption onto the ZnO-NPs. The employment of MAlO and MOLA as adsorbents for the removal of glyphosate from water also reveals some frequency shifts, the disappearance of bonds and the formation of new bonds.

Batch adsorption study

Results showing the variation in the amount of glyphosate adsorbed with some analytical parameters such as time, concentration and adsorbent dosage are displayed in Fig. 4. The adsorbent dosage was found to exert a positive impact on the adsorption of glyphosate by the three adsorbents (Fig. 4a), such that the equilibrium amount of glyphosate adsorbed experienced a rise with an increase in the mass of the adsorbent. Literature has shown that an increase in the mass of the adsorbent generally results in improved adsorption because there is a corresponding increase in the number of active adsorption sites. Therefore, as more adsorbate molecules diffuse to the surface, the vacancy for further adsorption would be enhanced (Elfeky et al., 2020). These results also align with the findings reported by Naghdi et al. (2023) for the adsorption removal of glyphosate by the metal–organic framework.

Fig. 4
figure 4

Variation of equilibrium amount of glyphosate adsorbed by ZnO-NPs, MAlO and MOLA with adsorbent dosage, pH and initial concentration of glyphosate

Full size image

The best adsorption tendency regarding the period of contact was achieved by the unmodified sample (i.e. ZnO-NPs), while the MAlO exhibited the least. This may be related to the nature of porosity and the tendency of the adsorbent to be relatively weakened as the contact period increases (Fig. 4b). Some studies have given explanations on possibilities that favours the mutual increment in adsorption and time concerning the adsorption of organic molecules. Some attributed such observations to the successful equilibration of the adsorbate–adsorbent system, which is principally dependent on time as explained in this context (Silva et al., 2020). The period of contact of the adsorbate with the adsorbent in the solution was also observed to affect the extent of adsorption of glyphosate pesticides such that as the contact period increases, more of the adsorbate molecules had the opportunity to be adsorbed due to enhanced interaction between the adsorbed species and the penetration power of the adsorbate towards the electric double layer as the contact period increases. A similar trend in variation of the amount of glyphosate adsorbed with time was observed and reported by Sen and Mondal (2022) and Chen et al. (2016) for the adsorption of glyphosate from an aqueous solution by some non-ZnO-NP adsorbents.

Between the pH of 3 and 5 (Fig. 4c), the equilibrium amount of glyphosate adsorbed decreases by 7.67%, 7.87% and 13.77% for ZnO-NPs, MAlO and MOLA, respectively. Consequently, the highest decrease in adsorption was observed for the MOLA adsorbent. The decreasing trend was followed by a sharp increase that resulted in comparable efficiencies for both ZnO-NPs and MAlO. The two adsorbents also seem to display similar adsorption mechanisms (as shown by the parallel pattern recorded for the adsorption profiles of the two adsorbents for glyphosate). However, at a pH of 9, ZnO-NPs had the best adsorption efficiency (% Removal = 93.27%), while MAlO had the lowest, and MOLA had the highest (% Removal = 93.27%). At a pH of 9, the recorded efficiency was 91.54%.

The optimum pH for the removal of the pesticide was 7, 3 and 7 for ZnO-NP, MAlO and MOLA, respectively. That means the removal of MAlO is best favoured at an acidic while profiles the dominance of H+. This may be ascribed to the fact that Al2O3 (though amphoteric) causes the ZnO-NPs to act as a base and thus exert a strong affinity for the adsorption at acidic pH (Pereira et al., 2019). The dependent pH for the other adsorbents witnessed the best conditions at neutral pH. The observed trend is not strange when considering the available literature. For example, Mattos et al. (2017) also reported a pattern similar to the observation in this study while investigating the variation of glyphosate adsorption with pH.

An increase in the initial concentration of the adsorbate (Fig. 4d) was also found to produce an increase in adsorption because more molecules are available to compete for the available adsorption sites (Al-Harby et al., 2021). It is certain that as long as the available adsorption sites have not been saturated, the tendency for an adsorbent to accommodate more adsorbate molecules will increase with an increase in concentration. At lower concentrations, once the adsorbate molecule is unable to diffuse to the adsorbent surface, adsorption will decrease, but at higher concentrations, there would be enough molecules to compete for the adsorption sites (Rápó & Tonk, 2021). The kinetics of the adsorption process was investigated through the fitting of experimental data into pseudo-first-order, pseudo-second-order, Weber–Morris and Elovich kinetic models, which are described by Eqs. 6 to 9, respectively (Babapour et al., 2022):

$$ln\left({Q}_{e}- {Q}_{t}\right)= {lnQ}_{e}-{k}_{1}t$$
(6)
$$\frac{t}{{Q}_{t}}=\frac{1}{{k}_{2}{Q}_{e}^{2}}+\frac{t}{{Q}_{e}}$$
(7)
$${Q}_{t}= {k}_{id}{t}^{1/2}+{C}_{id}$$
(8)
$${Q}_{t}= \frac{1}{\beta }ln\left(1+ \beta \alpha t\right)=\frac{1}{\beta }ln\beta \alpha + \frac{1}{\beta } lnt$$
(9)

The suitability of a pseudo-first-order model to the adsorption of glyphosate by ZnO-NP, MAlO and MOLA was investigated by plotting against time (based on Eq. 6). The linear plot (plots not shown) obtained gave the best R2 value for MALO (Table 1). The values of the theoretical equilibrium amount of the pesticide adsorbed (Qe) were low, but the ZnO-NP gave the closest approximation to the experimental data. The highest pseudo-first-order rate constant was exhibited by MOLA and the least by MAlO, indicating the adsorption becomes faster in that order. However, perfect correlations (R2 = 1) were obtained for the three adsorbents (Table 1) through the pseudo-second-order plots (Fig. 5) compared to the pseudo-first-order model, which gave R2 values (ranging from 0.8316 to 0.9314) that are relatively lower than those calculated for pseudo-second-order. Excellent approximation was also observed for the theoretical equilibrium amount of glyphosate adsorbed (ranging from 82.85 to 82. 97 mg/g) compared to the experimental average of 82.90 mg/g (Table 1). Therefore, the kinetics of the adsorption process is best described by the pseudo-second-order model. Others have published similar deductions concerning the adsorption of glyphosate by various adsorbents and accepted the pseudo-second-order kinetics as best fitted (Jia et al., 2011). Therefore, the adsorption rate is majorly limited by chemisorption and not solely by diffusion (Bazan-Wozniak & Pietrzak, 2023).

Table 1 Evaluated kinetic parameters for the adsorption of glyphosate by ZnO-NPs, MAlO and MOLA
Full size table
Fig. 5
figure 5

Pseudo-first-order plots for the adsorption of glyphosate by ZnO-NP, MAlO and MOLA and Webber–Morris and Elovich isotherms for ZnO-NP, MAlO and MOLA

Full size image

The linear plot describing the intraparticle diffusion model (Fig. 5) did not give a zero intercept, confirming that the adsorption is not limited by diffusion (Xu et al., 2022). The estimated rate constant (Table 1) was the highest for the adsorption of glyphosate by ZnO-NPs and MOLA, which is consistent with the inference drawn from the pseudo-second-order models (Table 1). The application of the intraparticle diffusion model to the adsorption removal of glyphosate by calcined calcium aluminium hydrotalcite has also been affirmed by Peng et al. (2021).

The Elovich model was also accepted since a plot of Qt against t was linear as shown in Fig. 5. From the model, the constants a and b have mgg−1 min−1 and g.mg−1 as their respective units. The initial rate was the highest for MOLA and lowest for ZnO-NPs, which is the inverse of the constant rate values obtained from the other models.

The Elovich model predicts that ZnO-NPs have the least initial adsorption rate but the least desorption rate, while MOLA has the highest initial adsorption rate and desorption rate. These findings are consistent with the experimental data, which showed that ZnO-NPs adsorbed glyphosate better than MOLA. Hnana et al. (2019) also found that the adsorption of glyphosate obeyed the Elovich model onto ghassoul and chitosan. A similar report, on the use of calcite for the adoption removal of glyphosate from an aqueous medium, also affirmed the suitability of the Elovich equation in the explanation of the adsorption characteristics of the system.

Adsorption isotherms are pivotal in the understanding of the adsorption behaviour of ZnO-NPs, MAlO and MOLA adsorbents. We investigated several isotherms and found that the experimental data fit the Temkin, Freundlich, Brunauer–Emmet–Teller and Dubinin–Radushkevich isotherms (shown in Fig. 6d) gave a better degree of fitness. The listed isotherms were developed based on Eqs. 10, 1112 and 13, respectively, which represent the assumptions that established them (Brião et al., 2022):

Fig. 6
figure 6

Freundlich, Temkin, Dubinin–Radushkevich and BET plots for the adsorption of glyphosate by ZnO-NPs, MAlO and MOLA

Full size image
$${lnQ}_{e}= ln{k}_{F}+\frac{1}{n}ln{C}_{e}$$
(10)
$$ln{Q}_{e}=\frac{RT}{{k}_{T}}ln{A}_{T}+\frac{RT}{{k}_{T}}ln{C}_{e}$$
(11)
$$\frac{{C}_{e}}{{Q}_{e}\left({C}_{s}-{C}_{e}\right)}= \frac{1}{{Q}_{m(BET)}{C}_{BET}}+ \frac{\left({C}_{BET}-1\right)}{{Q}_{m(BET)}{C}_{BET}}.\frac{{C}_{e}}{{C}_{s}}$$
(12)
$$ln{Q}_{e}= {lnQ}_{\mathrm{max}(E)}- \beta {\in }^{2}$$
(13)

Equation 6 is the linear equation for the Freundlich isotherm. The consistency of the experimental data with this model was ascertained by the high calculated values of R2 (Table 2). The adsorption constants of the Freundlich system deduced from the plot (Fig. 6a) were 0.7516, 1.6274 and 0.2418 for ZnO-NPs, MAlO and MOLA (Table 2), respectively. The parameter, 1/n (which represents adsorption intensity), energy distribution factor and surface heterogeneity, was less than unity except for MOLA (which had a slight deviation) and therefore confirmed that the adsorption of glyphosate is favourable because \(0<{}^{1}/_{n}<1\) (Mahalakshmi et al., 2016). Furthermore, the 1/n values are less than unity (except for MOLA), suggesting the favouritism of the physisorption mechanism for ZnO-NP and MALO but the chemisorption mechanism for MOLA (Garg et al., 2022). However, based on the fitness of Elovich and pseudo-second-order kinetic models, it is also inferable that the chemisorption mechanism also features in the adsorption of the pesticides onto ZnO-NP and MAlO but with relatively less dominance. The heterogeneity factor connects the free energy of adsorption (\({\Delta G}_{ads}= -RTln{k}_{F})\) with the maximum adsorption surface according to Eq. 14 (Eddy et al., 2022a):

Table 2 Adsorption parameters for the adsorption of glyphosate by ZnO-NPs, MAlO and MOLA
Full size table
$${Q}_{max}= \frac{{\Delta G}_{ads}}{nRT}= \frac{{-lnK}_{F}}{n}$$
(14)

The evaluated values of \({Q}_{max}\) (Table 2) showed that MOLA had the largest unutilized adsorption area, while ZnO-NP had the least, which is also consistent with experimental data.

The Dubinin–Radushkevich plot (a plot of lnQe versus \(\epsilon\), where \(\epsilon =RTln\left(1+1/{C}_{e}\right)\), expressed in Eq. 9) gave a linear plot with reasonable values of R2 (Fig. 6). Evaluated values of the Dubinin–Radushkevich adsorption energies (E = 400 kJ/mol) (Table 2) for the different adsorbents were less than the maximum required for observing a physical adsorption mechanism. Therefore, based on the Dubinin–Radushkevich data, the adsorption of glyphosate on ZnO-NP and modified ZnO-NP occurred through the mechanism of physicochemical adsorption. However, in their study, Pereira et al. (2019) found a physisorption mechanism for the adsorption of glyphosate onto ferrihydrite using the Dubinin–Radushkevich model.

The BET model (Eq. 14) allows for the occupation of the given adsorption sites by more than one molecule, unlike the Langmuir model which is based on the monolayer of adsorption (Langmuir, 1918). The experimental data did not fit the Langmuir adsorption model but fitted the BET model (because a plot of \(\frac{{C}_{e}}{{Q}_{e}\left({C}_{s}-{C}_{e}\right)}\) versus \(\frac{{C}_{e}}{{C}_{s}}\) t was linear) as shown in Fig. 6, indicating that there existed a multimolecular layer of adsorption that operated to create a lower adsorption surface (i.e. BET surface), that was almost the same for all the adsorbent (Table 2), where \({Q}_{m(BET)}\) is the BET monolayer adsorption capacity and \({C}_{BET}\) is the BET adsorption constant (which followed the trend, ZnO-NP > MAlO > MOLA). Al-Arjan (2022) also affirmed that the BET adsorption isotherm was appropriate in the explanation of the adsorption of glyphosate by oil palm front. Also, the fitness of the BET isotherm was affirmed by Rodrigues et al. (2023) for the adsorption of glyphosate by palygorskite. Such levels of fitness is an indication of a multiple adsorption layer which is a descriptive term for the adsorption of more than one molecule of glyphosate by a given adsorption site in the ZnO-NPs.

Regeneration study

The three adsorptions were investigated for their reusability (in the re-adsorption of the glyphosate) after washing with distilled water (severally) and drying in the oven. According to the findings, ZnO-NP had the highest reusability index (94.34%), followed by alumina (92.6%) and MOLA (91.65%).

Computational chemistry consideration

Calculated values of semiempirical parameters for glyphosate were EHOMO =  − 6.84 eV, ELUMO =  − 1.22 eV, the energy gap (ΔΕ = ELUMO – EHOMO = 5.62 eV), dipole moment (7.80 Debye), total energy (− 930.6525 au), area (200.08 Å2), polar surface area (112. 480 Å2), ovality (1.42) and polarizability (53.84 Å3). The HOMO and LUMO parameters can effectively be represented through a molecular surface plot, which is shown in Fig. 7a for glyphosate (Ebenso et al., 2021). The positive charge (red) is seen to concentrate on the heteroatom, while the negative lobes concentrate on the carbon atoms. Figure 6a also shows the electrostatic potential plot (ESP) for glyphosate which reveals different ESP surfaces as shown by differences in colour.

Fig. 7
figure 7

a Surfaces showing the HOMO and LUMO locations in glyphosate. b Adsorption interaction model between glyphosate and ZnO-NPs

Full size image

The fundamental of the semiempirical calculations is that molecular reactivity is addressed as a global property without consideration of the contributions from each of the participating atoms in a molecule. Local reactivity can be addressed by the charges on each atom, but the most formal approach involves the calculation of Fukui functions, which can be defined as a process involving the loss of electron (N − 1), the gain of electron (N + 1) and neutral state (N). The first Fukui function defines electrophilic (Eq. 16), the second nucleophilic (Eq. 15) and the last radical Fukui functions (Eq. 17) (Eddy & Ameh, 2021):

$${f}_{x}^{+}= {q}_{N+1} -{q}_{N}$$
(15)
$${f}_{x}^{-}= {q}_{N} -{q}_{N-1}$$
(16)
$${f}_{x}^{0}= \frac{\left({q}_{N+1}- {q}_{N-1}\right)}{2}$$
(17)

The functionality of the above equations is the assignment of the sites with the highest positive values of the Fukui functions as the appropriate functional atom or bonds. Therefore, based on the calculated Fukui functions, the sites for electrophilic, nucleophilic and radical attacks are on the preference of O(11) atom, C(4)-O(6) bond and O(6) atom, respectively. The Fukui functions were evaluated using Mulliken and Hirshfeld charges for N + 1 \(\left({q}_{N+1}\right)\), N − 1 \(\left({q}_{N-1}\right)\) and N \(\left({q}_{N}\right)\) electronic systems. Quantum chemical investigation was also extended to cover the adsorption-associated interaction energy through Monte Carlo simulation modelling using the adsorption locator program in the Materials Studio software (see Fig. 7b). The adsorption locator function indicated that the insecticide is adsorbed on the surface of zinc oxide nanoparticles, which are also consistent with the findings from the evaluated Fukui functions.

The total energy, adsorption energy (Eads), rigid adsorption energy, deformation energy and \(\frac{{dE}_{ads}}{dN}\) were evaluated to correspond to − 27.9913, − 56.5196, − 62.4737, 5.9541 and − 56.5196 kcal/mol, respectively. The negative value of the Eads confirms the exothermic adsorption process (between the pesticide and the ZnO-NPs) resulting from the decrease in surface energy as the glyphosate molecules approach the surface of the adsorbent. The evaluated adsorption energy (− 56.5196 kJ/mol) is consistent with the range of data expected for the mechanism of chemisorption adsorption (i.e. 200 > Eads (kJ/mol). Therefore, the adsorption of glyphosate on the nanosurface first responded to the physical adsorption and was succeeded by the chemisorption mechanism (Eddy et al., 2018). The sum of the deformation (5.9541 kcal/mol) and rigid adsorption energy (− 62.4737 kcal/mol) is the equivalent of the Eads. The rigid energy component is associated with the energy released when the unrelaxed glyphosate is adsorbed on the surface of the ZnO nanoparticles, while the energy released when the adsorbed molecule is relaxed on the surface is the deformation energy. Also, since the total energy is the sum of the internal and adsorption energy, then, the calculated internal energy of the adsorbed system is equal to 84.5109 kcal/mol (Yuan et al., 2005).

Conclusions

ZnO-NPs with wurtzite crystalline structure have been synthesized by the application of the precipitation method. The produced nanoparticles showed major FTIR peaks at 459 nm and a ZnO stretching are typical ZnO-NPs. Apart from the reduction in crystallite sizes, modification of the adsorbent by alumina and oleic acid (+ xylene) resulted in a more ordered crystal but with an increased amorphous character. The synthesized ZnO-NPs, MAlO and MOLA are good adsorbents for the removal of glyphosate from an aqueous medium. The adsorption tends to be dominated by a chemical adsorption mechanism, showing excellent fitness for pseudo-second-order kinetics. The present results showed that more than 90% glyphosate can be withdrawn from aqueous media through adsorption by the zinc oxide nanoparticles and the modified samples.