Glyphosate Adsorption from Water Using Hierarchically Porous Metal–Organic Frameworks

The selective removal of one ligand in mixed‐ligand MOFs upon thermolysis provides a powerful strategy to introduce additional mesopores without affecting the overall MOF structure. By varying the initial ligand ratio, MOFs of the MIL‐125‐Ti family with two distinct hierarchical pore architectures are synthesized, resembling either large cavities or branching fractures. The performance of the resulting hierarchically porous MOFs is evaluated toward the adsorptive removal of glyphosate (N‐(phosphonomethyl)glycine) from water, and the adsorption kinetics and mechanism are examined. Due to their strong affinity for phosphoric groups, the numerous Ti–OH groups resulting from the selective ligand removal act as natural anchor points for effective glyphosate uptake. The relationships between contact duration, glyphosate concentration, and adsorbent dosage are investigated, and the impact of these parameters on the effectiveness of glyphosate removal from contaminated water samples is examined. The introduction of additional mesopores has increased the adsorption capacities by nearly 3 times with record values exceeding 440.9 mg g−1, which ranks these MOFs among the best‐reported adsorbents.

be introduced into the environment at various stages of its production and usage. [10,11] Incidents of glyphosate toxicity in humans have raised concerns about its health effects, including eye and skin irritation, contact dermatitis, eczema, cardiac and respiratory issues, and allergic responses. Buffin and Topsy provide a detailed assessment of glyphosate's acute hazardous effects in humans. [2] There is no recommended value for glyphosate residue in drinking water; however, the EU standard for any pesticide in drinking water is 0.1 µg L −1 . [12] This is, without a doubt, a significant challenge for a portable water treatment facility. In light of the rising reports of glyphosate in the aquatic environment, it has been claimed that the cost of building the required equipment to remove herbicides from drinking water may be £1.0 billion, with annual operating expenses of £50-100 million in the UK. [13][14][15][16] As a result, glyphosate-related water contamination must be strictly controlled. Although various traditional techniques, including the use of activated carbons, oxidation, ozonation, and photocatalytic degradation, may be used to remove glyphosate, an efficient and cost-effective approach is still desired. [17][18][19] The removal of glyphosate from mineral surfaces and soils has been studied using biodegradation, [20] adsorption, [21] oxidation, [22] and photocatalysis degradation processes. [23] Some have concentrated on quantitative removal, [24][25][26] while others have attempted to comprehend the interactions at the molecular level. [27] Many researchers are interested in adsorption because of its advantages of convenience, low cost, and environmentally friendly operation. [28,29] Activated carbon, zeolite, and carbon-based composites are the most commonly reported adsorbents for glyphosate. However, their adsorption properties currently remain unimpressive, since the majority of them have poor selective recognition ability and adsorption capacity, limiting their use in organic pesticide removal. [8,[30][31][32][33][34][35][36][37] As a result, the parameters influencing the adsorption process between glyphosate and the adsorbent must be investigated.
Metal-organic frameworks (MOFs) are porous materials with record surface areas and tunable chemistries that render them highly promising candidates for the adsorptive elimination of toxic inorganic and organic substances as well as photocatalysis. [37][38][39][40][41][42] The absorption capacity of MOFs can be varied by adjusting pore size, shape, and chemistry through different synthetic techniques or post-synthetic modification. [43] Besides sufficient stability in aqueous solutions, suitable MOFs thus need appropriate pore size and connectivity that allow access to active adsorption sites for effective glyphosate adsorption. [44] In our previous study, we developed a highly selective ligand removal technique (aka SeLiRe strategy), building on the work by Feng et al., [46] which enables the construction of MOFs with dual porosity from their related mixed-ligand MOFs. [47] Through careful tuning of synthetic conditions and heating parameters, we were able to design MIL-125-Ti either with isolated cavitybased mesopores of uniform diameter or branching narrow fracture-type pores without noticeable collapse of the parent micropore structure. The process for removing ligands was investigated through comprehensive ex situ and in situ studies combined with Density Functional Theory (DFT) simulations. The ligand removal is very selective for aminoterephthalic acid (BDC-NH 2 ), and it occurs in two phases, each of which may be finetuned by changing the temperature and duration. [47] In this study, we investigate the effect of type, connectivity, and size of the added mesopores on the glyphosate adsorption. In addition, we also investigate single-ligands MIL-125-Ti and NH 2 -MIL-125-Ti as well as the corresponding pristine mixedligand MOFs prior to SeLiRe to identify the mechanism of glyphosate adsorption. Our results demonstrate that the introduction of large cavity-type mesopores significantly improves both the capacity and efficiency of glyphosate adsorption due to improved accessibility of the interior surface and increased number of Ti sites created by the SeLiRe process. Therefore, our study offers a fascinating example of how rationalized pore engineering can improve the adsorptive properties of MOFs for larger compounds.
The proportion of the amino ligand in the mixed-ligand structure strongly influences the geometry of the introduced mesopores. Based on our previous work, [46] we chose two ligand ratios that are representative of two distinct pore geometries: 2%BDC-NH 2 , which demonstrates cavity-type pores, and 50%BDC-NH 2 , which shows fracture-type pores in the final MOF structures. XRD analysis confirms that all samples are crystalline, with no impurities or second phases ( Figure S2a, Supporting Information). Additionally, the TEM images demonstrate that the particles have a disk-like shape ( Figure S2b,c, Supporting Information). The selected area electron diffraction patterns (SAED) confirm the expected structure and absence of extra phases ( Figure S2d,e, Supporting Information). As seen from Figure 1b,c, the 2%-SeLiRe MOF contains blocks of cavity-type pores while the 50%-SeLiRe MOF contains narrow fracture-type pores. Both of these newly introduced pores serve as indicators of linker removal. Physisorption with Ar at 87 K was used to assess the development of new porosity in SeLiRe samples ( Figure S3, Supporting Information). The results for both samples prior to SeLiRe show type I(a) isotherms, typical of micropore samples. After ligand removal, the isotherms change to a hybrid of I(a)/ IV(a), indicating the generation of additional mesoporosity. The shape of the isotherm further indicates that in the 2%-SeLiRe sample the pores are isolated from one another and partially blocked. In contrast, the mesopores in the 50%-SeLiRe sample appear to be accessible from the surface with openings <4-5 nm.

Adsorption Studies of Glyphosate
The adsorption of glyphosate was evaluated using aqueous solutions comprising various adsorbent concentrations www.afm-journal.de www.advancedsciencenews.com

(3 of 8)
© 2023 The Authors. Advanced Functional Materials published by Wiley-VCH GmbH (0.5-10 mg L −1 ) at room temperature with 5 mg L −1 glyphosate. The amount of residual glyphosate in each sample was determined during a period ranging from 30 min to 24 h (more details in Table S1, Supporting Information). The residual glyphosate concentration from the amount of phosphorous in different samples was determined by ICP-OES. The amount of adsorbed glyphosate at equilibrium, q e (mg g −1 ) and removal efficiency (percentage of adsorption), E%, were calculated based on Equations S3 and S4 (Supporting Information). Figure S4 (Supporting Information) depicts the effect of adsorbent dosage on adsorption efficiency for single-ligands, 2%-mixed-ligand, and the corresponding SeLiRe sample after adsorption of glyphosate. Based on the results, 5 mg L −1 of adsorbents was selected for further experiments. Figure 1a which is obtained from the initial rate of the adsorption per hour shows the comparison of samples with cavity-type pores (2%, 5%, and 10%) to samples with fracture-type pores (50%, 80%) indicates that the samples with cavity-type pores have a higher adsorption efficiency. Since there are no remaining NH 2 groups in the structure of the SeLiRe MOFs (as revealed by Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS) and Proton Nuclear Magnetic Resonance Spectroscopy ( 1 H NMR), [47] the increase must be attributed to increased mesoporosity. Thus, the pore geometry has a significant impact given the considerably jump in relative adsorption. Therefore, a higher number of adsorption sites and their increased accessibility are key to the observed performance enhancement of the SeLiRe-treated samples compared to the purely microporous single-ligand and mixed-ligand MOFs.

Adsorption Kinetics
The kinetic order and associated rate constants for glyphosate adsorption were determined to investigate the adsorption kinetics. Figure 2a shows that q t rises sharply in all MOFs within the first few hours. The SeLiRe MOF exhibits the highest rate of q t /time, which is attributed to the additional mesopores that likely make it easier for glyphosate to be transported to the MOF's adsorption sites. The other samples exhibit different times at which the adsorption approaches saturation, with the NH 2 -containing MOF taking the longest. The adsorption kinetics according to pseudo-first-order (PFO) and pseudosecond-order (PSO) models (Equations S5 and S6, Supporting Information) are shown in Figure S5a,b (Supporting Information). The linear correlation coefficients (R 2 ) obtained from the PSO are 0.997, 0.990, 0.992, and 0.993 for SeLiRe MOF, mixed-ligand MOF, MIL, and NH 2 -MIL, respectively, while PFO only shows lower values; hence, the time dependence of the glyphosate adsorption is significantly better fitted with the PSO. Furthermore, the calculated q e is close to the experimental data, indicating that the PSO model effectively characterizes the adsorption process (Table S4, Supporting Information). The reaction has three essential components that affect the kinetics, namely: i) the adsorption sites on MOFs, ii) the adsorbate, glyphosate, and iii) the solvent, water. Since water is abundant, its concentration can be neglected. Additional kinetic parameters are listed in Tables S2 and S3 (Supporting Information).

Adsorption Isotherms
The isotherms of glyphosate adsorption on MOFs are shown in Figure 2b. Note that the adsorption capacity at equilibrium (q e ) increases in all cases with increasing glyphosate concentration at equilibrium (C e ). However, saturation is reached at far lower concentrations with the SeLiRe MOFs (plateauing at C e = 5 mg L −1 ) compared with MIL and the mixed-ligand MOFs (C e ≈20 mg L −1 ) and with NH 2 -MIL yieldings a C e of 23 mg L −1 .
The q e to C e ratios ( Figure S6a, Supporting Information) may be indicative of the accessibility of adsorption sites.
The maximum adsorption capacity (q max ) for all samples is shown in Figure 2c. As expected, the SeLiRe sample exhibits considerably higher values than the single-ligand and  mixed-ligand samples. Note that the highest value was observed for 2%-SeLiRe MOF with 440.9 mg g −1 , which is ≈3 times higher than the other samples. This value also ranks among the highest reported adsorption capacities among various adsorbents summarized in Table 1.
A closer look reveals a direct correlation between q max and the respective BET surface area for all single-ligand and mixedligand MOFs; thus, the number of adsorption sites on the interior pore surface is identified as directing parameter. In contrast, the values of q max for the SeLiRe samples are significantly higher than expected from the respective surface areas (Table S5, Supporting Information). This may be due to the additional space for glyphosate upon introduction of larger mesopores. Another con-tribution may stem from uncoordinated Ti-centers on the secondary building units (SBUs) created by the SeLiRe process that act as additional adsorption sites. Consequently, the SeLiRe process has not only facilitated reactant access through mesopores, but also increased the total number of adsorption sites.
In order to confirm the contribution of additional Ti sites, we analyzed the adsorption data with both Langmuir and Freundlich models. Accordingly, the single-ligand MOFs and the mixed-ligand sample fit the Langmuir model as expected ( Figure S6a, Supporting Information), indicating that there is only one type of adsorption site. In contrast, the isotherms for the SeLiRe MOF are clearly Freundlich-type ( Figure S6b, Supporting Information), with K F 3 times higher than other MOFs (more details in Table S6, Supporting Information). This indicates that there are two distinct adsorption sites with different interaction strength.

Proposed Mechanism of Glyphosate Adsorption
The adsorption of glyphosate was further investigated with FTIR spectroscopy; the spectra of the 2%-SeLiRe sample before and after glyphosate adsorption are shown in Figure 3a, and the results of the other samples are shown in Figure S9 and summarized in Table S8 (Supporting Information).
The spectrum of the MOF before glyphosate adsorption contains several bands that are described in our previous work. [47] The main feature important for the present study are the dominant bands at 1545 and 1386 cm −1 , which are commonly attributed to asymmetric and symmetric νCOO coordination of the ligand to the Ti-SBU, respectively. Note that upon glyphosate adsorption both bands are down-shifted to 1429 and 1289 cm −1 , respectively, indicating a weakening of the carboxyl coordination.  The spectrum after glyphosate adsorption contains some prominent new bands. The band at 1682 cm −1 corresponds to the carbonyl vibration of the carboxylic function in glyphosate, [55] which suggests that the carboxyl function of glyphosate does not significantly interact with the MOF network. The most important features are the doublet bands at 1132 and 1110 cm −1 as well as at 932 and 880 cm −1 , which are generally attributed to asymmetric and symmetric νP-O and νP-OH of the phosphonate function in glyphosate, respectively. [56] Note that the first doublet is slightly red-shifted compared to free phosphonates (i.e., 1155, 1075 cm −1 ), [56] which is indicative of a Ti-phosphonate complex whose νP-O is lower in the comparatively stronger Ti-O-P than in P-OH. In a similar manner, P-OH is higher compared to free phosphonate, since Ti-O-P-OH is weaker than P-OH. [56] These observations strongly suggest that glyphosate is adsorbed through a Ti-phosphonate coordination. Note that the νP-OH vibration at 932 cm −1 is broader than the others, which suggests that the residual OH group in the coordinated phosphonate interacts with the nearby carboxylic coordination through H-bonds.
To support these finding, we performed DFT simulation studies (details in Section S7, Supporting Information). Figure 3b shows the most stable configuration of glyphosate within the MOF framework. Note that glyphosate is coordinated to one Ti from the SBU through one O from the phosphonate function (encircled area). In addition, the remaining hydroxyl group from the phosphonic function interacts with the neighboring carboxylic function of the BDC ligand though H-bond formation, as indicated by the broadened νP-OH. The higher electron density in Ti upon phosphonate coordination (in contrast to hydroxyl groups in the initial MOF) and the trapping of electron density in the aforementioned H-bonds both induce a weakening of the carboxylic coordination, which is reflected by the significant red-shift of the COO bands in FTIR. Moreover, we calculated the TiOP bond length of the new phosphonate coordination to be 2.22-2.23 Å. The negative adsorption energy values (Table S9, Supporting Information) support the formation of a phosphonate coordination as the most favorable adsorption mode. Also note that neither the NH group nor the carboxyl function of glyphosate seem to interact significantly with the MOF network, which is in line with the FTIR results.
The combined results confirm that glyphosate is adsorbed through monodentate phosphonate coordination and supported by H-bond formation. Ti sites are Lewis acids with significant  affinity toward phosphates. One possibility is that glyphosate adsorbs via condensation with the OH groups of the Ti sites ( Figure 3c). Another is an ion-exchange of hydroxyl groups with deprotonated glyphosate molecules. The latter mechanism should not be neglected, considering that the adsorption studies were performed at pH 5.4, at which glyphosate exists in deprotonated form (pKa = 2.3 for phosphonate function). [57] In both cases, glyphosate adsorption is facilitated by the selective ligand removal strategy, which in addition to enhancing access to the Ti clusters also introduces additional TiOH sites.
We also studied the adsorption of glyphosate at various pH values (2.5, 5.4, 9.2, and 10.8). The results show similar adsorption capacities for the first three pH values ( Figure S8a, Supporting Information), despite considerable changes in Zetapotential of the MOFs ( Figure S8b, Supporting Information), which indicates that additional contributions by electrostatic interactions to the adsorption process can be neglected. Note that the adsorption capacity drops considerably at pH 10.8, at which point the amine function deprotonates (pKa = 10.6). This likely alters the electron distribution across the glyphosate molecule and hence its ionic strength, consequently disfavoring ion exchange with the hydroxyl groups at the Ti-SBUs.

Recyclability and Stability of SeLiRe MOFs
Real-world applications require stable and reusable adsorbents. Therefore, we tested the cycling performance of the SeLiRetreated samples and evaluated the stability with XRD. In this case, glyphosate-loaded adsorbents were added to a sodium chloride saturated solution, the solution was stirred for 12 h, centrifuged, washed with deionized water and ethanol, and the MOFs particles were recovered followed by drying under vacuum. This process was repeated 4 times. Figure 4a shows similar values of q e for each cycling step (19.7-19.3 mg g −1 ), which indicates that the adsorption process is reversible. Moreover, the adsorption efficiency shows only a small decrease from 98% to 96% after 4 cycles, which further confirms the excellent reversibility of the MOFs. According to literature, sodium chloride facilitates the release of adsorbed glyphosate through competitive ion-exchange facilitated by softening of H-bonds, albeit the actual mechanism remains a subject of discussion. [57] Finally, XRD studies after glyphosate adsorption also confirm the high stability of this MOF in aqueous conditions, as they revealed no changes in peak position or relative intensity, nor any evidence of impurities or additional phases after 4 times recycling (Figure 4b).

Conclusion
In this work, we investigated various MIL-125-Ti based MOFs toward the adsorption of glyphosate from water. In particular, we demonstrate that the selective removal of ligands (SeLiRe) greatly enhances the adsorption performance in various ways: i) by providing additional Ti adsorption sites that interact with glyphosate more strongly than the intrinsic Ti sites (as shown by the change of isotherm from Langmuir to Freundlich) and ii) by introducing additional mesopores that facilitate the accessibility of these adsorption sites.
The structure of the mesopores has a significant impact on the adsorption capacity: the best performance was achieved with cavity-type pores, which resulted in a 3 times increase compared to fracture-type pores. This suggests that the adsorption performance is dominated by the mesopore diameter, which is bigger in cavities than in fracture-type pores. The observed adsorption capacity of 440 mg g −1 ranks these materials among the best-known adsorbents. Moreover, these MOFs are highly stable in water and show an excellent reusability.
We further investigated the glyphosate adsorption mechanism. FTIR studies revealed that glyphosate adsorbs via phosphonate coordination with Ti from the SBU likely through an ion-exchange mechanism, supported by the formation of H-bond with the neighboring carboxylic coordination of the BDC ligand with the Ti-SBU. DFT simulations confirmed that such phosphonate coordination leads to the energetically most favorable system.  In conclusion, the selective ligand removal strategy provides a fascinating way to engineer adsorption sites in MOFs and to enhance their performance for adsorption and separation technologies. Future studies will transfer this concept to Zr-based MOFs, due to the higher affinity of Zr to glyphosate compared with Ti.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.