Separation and purification of 5-hydroxymethylfurfural by metal–organic frameworks

Abstract

5-Hydroxymethylfurfural (HMF) is an important sugar-based platform chemical that can be produced from fructose by acid-catalyzed dehydration. However, the efficient removal of by-products formic acid (FA), levulinic acid (LA) and residual fructose to obtain high purity HMF remains a challenge, especially the separation of LA and HMF. Herein, MOF-808 could selectively adsorb impurity components with high uptake capacity and high selectivity factor (α_imp./HMF), such as fructose (245 mg g_ads⁻¹, 47.4), FA (155 mg g_ads⁻¹, 41.3), and LA (303 mg g_ads⁻¹, 57.8). Moreover, the adsorption kinetic constants of FA and LA in MOF-808 could reach remarkably high values of 58.0 h⁻¹ and 25.9 h⁻¹, respectively. The efficient adsorption separation performance of MOF-808 is attributed to its abundant unsaturated paired sites and hard Lewis acidity. Additionally, MOF-808 showed outstanding separation performance and cycling stability in column separation. This study indicates coordination interactions are effective for the separation and purification of HMF.

---

Introduction

5-Hydroxymethylfurfural (HMF), as a bridge between the biomass feedstock and biorefinery industries, can be used to prepare a wide range of high-value chemicals, such as biofuels, biodegradable plastics, coatings, cosmetics, and pharmaceuticals.^1–5 Therefore, HMF has been listed as a renewable chemical that should be prioritized for development by the U.S. Department of Energy (DOE) and has been hailed as a “sleeping giant”.^6–8 HMF is primarily produced via acid-catalyzed dehydration of fructose; however, by-product impurities such as formic acid (FA) and levulinic acid (LA) are generated during the reaction process, which means the crude product needs to be further purified to obtain high purity HMF. Moreover, the cost of purification accounts for 60–70% of the total production cost, making the development of an efficient HMF purification technology crucial for its widespread application.^9–11

The main methods of HMF separation and purification are extraction, distillation, membrane separation, and adsorption separation.¹² Among these, adsorption separation is particularly advantageous as it can efficiently separate low concentrations of HMF at relatively low temperatures, avoiding the polymerization of HMF. Additionally, it offers low energy consumption and recyclability, making it one of the most advanced separation methods. The key to improving the adsorption separation performance is the development of adsorbents. In recent years, zeolites,¹³ polymers,¹⁴ carbon materials,¹⁵ and Metal–Organic Frameworks (MOFs)¹⁶ have been used for the adsorption separation of HMF. Previous studies showed that improving the hydrophobicity^9,17–19 and microporosity^20,21 of the materials, as well as constructing π-electronic structures,^16,22,23 could improve the adsorption capacity and selectivity of HMF through hydrophobic interactions, extrusion effects, and π–π interactions. However, the efficient separation of HMF from the by-products LA and FA and residual fructose through the aforementioned interactions remains challenging, especially the separation of HMF from LA.^9,10,22,23 The HMF–LA selectivity factor is typically below 9 (Table S1), prompting several studies to address this issue in recent years. Zhang et al. synthesized a hollow-structured porous aromatic polymer (H-PAP), which achieved a highly selective separation of HMF from fructose, FA, and LA. However, the adsorbed HMF could not be desorbed efficiently, resulting in a total HMF recovery rate of only 9.4% after two rounds of desorption.²¹ Hu et al. employed resin BD-11 as an adsorbent and it could effectively separate HMF and LA with a selectivity factor as high as 73.0 by increasing the pH value of the solution above 7.²⁴ However, the adsorption capacity of HMF was only 67.9 mg g⁻¹, and the flow rate of the packed column had to remain below 5 BV per h (BV, bed volume). In addition, HMF would be lost due to a side-reaction when the pH value was increased.¹⁰ Therefore, the development of a more efficient adsorbent for HMF and LA separation is crucial.

MOFs are widely used in adsorption and separation due to their high specific surface area, tunable pore size, replaceable central metal and three-dimensional ordered structures.^25–29 Recently, in the separation of sugar and HMF, all of ZIF-8, NU-1000, and UiO-67–2AS(10) showed a preferred adsorption of HMF with high selectivity factors.^16,17,30 However, the selective adsorption of HMF from LA has not been studied. Due to the carboxyl group (–COOH) in LA (Scheme 1a), it can deprotonate to form a hard base carboxylate (RCOO⁻). Based on the hard/soft acid/base (HSAB) principle,^31–35 we proposed that LA can be separated from HMF by forming a strong interaction with the unsaturated coordination site of the hard acid. Zirconium(IV), one of the most widely used hard acids in MOFs, can form the stable Zr₆O₈ node, which can provide coordination sites for forming coordination interactions with the carboxylic acid impurities in the crude product of HMF when the connectivity number of the Zr₆O₈ node is less than 12. For example, the 6-connected Zr₆O₈ node can provide 12 unsaturated coordination sites (Scheme 1b), which is considered to be the largest number of defect-free coordination sites in common MOFs.³⁶

Scheme 1 (a) The chemical structures of fructose (Fru), formic acid (FA), levulinic acid (LA), and 5-hydroxymethylfurfural (HMF). (b) The illustration of the Zr₆O₈ node adsorbing FA and LA.

In this study, coordination interaction was introduced for the adsorption separation of HMF and impurities. The hard acid Zr-based 6-connected MOF-808 with abundant unsaturated coordination sites was selected as the adsorbent owing to its raw-material being easily available and simply synthesized. We investigated the quaternary competitive adsorption of fructose, FA, LA and HMF, and MOF-808 showed selective adsorption of carboxylic acids (FA and LA). And in all of the binary competitive adsorption experiments involving each of the three impurities with HMF, MOF-808 adsorbed the impurities with extremely high selectivity. The efficient and selective adsorption of carboxylic acid impurities by MOF-808 was attributed to its abundant unsaturated paired sites and high Lewis acid hardness. Moreover, during practical column separation, MOF-808 exhibited remarkable separation performance and cycling stability. The strong coordination interaction could efficiently remove LA and FA from the crude product of HMF, providing valuable insights for the design of advanced adsorbents for HMF purification.

Experimental section

Materials and synthesis

Materials. 1,3,5-Benzenetricarboxylic acid (H₃BTC, 98%), D-fructose (Fru, 99%), zirconium chloride (ZrCl₄, 98%), terephthalic acid (H₂BDC) (99%), levulinic acid (LA, 98%), and 5-hydroxymethylfurfural (HMF, ≥99%) were purchased from Aladdin; aluminum nitrate nonahydrate (Al(NO₃)₃·9H₂O, 99.0%) was purchased from Macklin; trimethyl 1,3,5-benzenetricarboxylate (Me₃BTC, 98%) was purchased from Energy Chemical; zirconium oxychloride octahydrate (ZrOCl₂·8H₂O, 99%) was purchased from Bidepharm; copper nitrate trihydrate (Cu(NO₃)₂·3H₂O, AR), methanol anhydrous (MeOH, AR), N,N-dimethylformamide (DMF, AR), concentrated nitric acid (HNO₃, GR), formic acid (FA, AR), tetrahydrofuran (THF, AR), and 2-methyltetrahydrofuran (2-MeTHF, 99%) were obtained from Sinopharm Chemical Reagent Co., Ltd; and ultrapure water was obtained from Heal Force (SMART-N). All of the materials were used as received without further purification.

Synthesis of MOFs. MOFs were synthesized with reference from published literature. Below are the brief synthesis procedures, and all of the as-synthesized products were collected by centrifugation and washed with fresh solvents. The detailed synthetic procedures and post-treatments are described in the SI.
Synthesis of HKUST-1 (ref. 37). 3.866 g of Cu(NO₃)₂·3H₂O was dissolved in 400 mL of MeOH, and 2.101 g of H₃BTC was dissolved in 100 mL of MeOH. The prepared H₃BTC solution was then poured into the Cu(NO₃)₂ solution, shaken well and heated to 60 °C for 12 h.
Synthesis of MIL-100-Al³⁸. In a Teflon-lined autoclave, 0.230 g of Al(NO₃)₃·9H₂O and 0.104 g of Me₃BTC were dissolved in 0.77 mL of 1 M HNO₃ and 2.80 mL of H₂O and then heated to 210 °C for 3 h.
Synthesis of MOF-808 (ref. 39). A total of 4.85 g of ZrOCl₂·8H₂O and 1.05 g of H₃BTC were dissolved in 225 mL of DMF and 225 mL of FA and then heated to 130 °C for 2 days.
Synthesis of UiO-66-Std⁴⁰. 0.699 g of ZrCl₄, 0.498 g of H₂BDC and 30 mL of acetic acid were dissolved in 200 mL of DMF and heated to 120 °C for 18 h.
Synthesis of UiO-66-Med⁴¹. 0.228 g of ZrOCl₂·8H₂O was dissolved in 32.50 mL of DMF and 0.562 g of H₂BDC was dissolved in 11.25 mL of DMF. Then 10 mL of the prepared ZrOCl₂ solution, 3.33 mL of H₂BDC solution and 4.60 mL of acetic acid were mixed and heated to 90 °C for 18 h.
Synthesis of UiO-66-Def⁴¹. 0.125 g of ZrCl₄ and 1 mL of concentrated HCl were dissolved in 5 mL of DMF. 0.138 g of H₂BDC was dissolved in 11.25 mL of DMF. Then 10 mL of the prepared H₂BDC solution was added to the ZrCl₄ solution and heated to 120 °C for 18 h.

Instrumentation and characterization

Instrumentation. Powder X-ray diffraction (PXRD) patterns were collected on a Bruker diffractometer (D8 Advance) with Cu Kα radiation (λ = 1.5418 Å). Scanning electron microscopy (SEM) images were obtained on a Zeiss microscope (Gemini 450). ¹H NMR spectra were recorded on a Bruker BioSpin spectrometer (AVANCE NEO, 500 MHz). N₂ physisorption isotherms were measured on a Micromeritics specific surface area analyser (3FLEX). The chemical state of Zr was characterized by X-ray photoelectron spectroscopy (XPS, Thermo Scientific, Nexsa G2), and the sample to be measured was first heated at 150 °C for 12 h under vacuum using a degasser. Raman spectra were measured using a WITec spectrometer (WITec Alpha300 RAS). Zr content was characterized using an inductively coupled plasma-atomic emission spectrometer (ICP-AES, Thermo Scientific, iCAP PRO XP). The concentrations of fructose, FA, LA and HMF were characterized by high performance liquid chromatography (HPLC, Agilent, Agilent 1260 Infinity II). Batch adsorption experiments were performed in a Huayuan constant temperature shaker (ZS-AR).

Batch adsorption experiment. 10.0 mg (±0.2 mg) of MOFs were weighed in 5 mL centrifuge tubes. 2 mL of a quaternary, ternary or binary mixture or single-component solution containing fructose, FA, LA and HMF was added to the tubes, and the mass of the added solution was weighed. Then, the tubes were placed in a constant temperature shaker and kept at a specific temperature (30 °C, 40 °C, 50 °C) and a shaking velocity of 140 rpm for 2 h. The suspension was filtered with a 0.22 μm pore size nylon syringe filter, and the changes in the concentrations of fructose, FA, LA and HMF were quantified via HPLC.

Adsorption capacity estimation. The adsorption equilibrium capacity was calculated using eqn (1) and the selectivity factor (SF) for the binary components was obtained using eqn (2):

q_e = (C₀ − C_e) × M/m (1)

α_a/b = (q_ea × C_eb)/(q_eb × C_ea) (2)

where q_e (mg g_ads⁻¹) is the equilibrium adsorption capacity (uptake capacity), C₀ (mg g_sol⁻¹) is the initial solution concentration, C_e (mg g_sol⁻¹) is the equilibrium concentration, M (g_sol) is the mass of the solution, and m (g_ads) is the mass of the adsorbent. α_a/b is the selectivity factor (SF) for component a to component b (a–b); component a and component b are the two components in binary competitive adsorption, respectively. q_ea and q_eb (mg g_ads⁻¹) are the equilibrium adsorption capacities of a and b, while C_ea and C_eb (mg g_sol⁻¹) are the equilibrium concentrations of a and b, respectively. Note: to ensure the accuracy of concentration measurement, mass concentration was used in this study, and dilution was performed via the gravimetric method.

Adsorption kinetics. The adsorption kinetics of the quaternary components (fructose, FA, and LA: 5 mg g_sol⁻¹, HMF: 20 mg g_sol⁻¹) mixture in MOF-808 was characterized. 10.0 mg of MOF-808 was weighed in 5 mL centrifuge tubes, 2.0 mL of the mixture solution was added to the tubes, and the mass of the added solution was weighed. The tubes were then placed in a constant temperature shaker with a shaking velocity of 140 rpm at 30 °C for specific durations: 1 min, 2 min, 5 min, 10 min, 20 min, 30 min, 60 min, 120 min, and 180 min. The suspension was filtered with a 0.22 μm pore size nylon syringe filter, and the concentrations of fructose, FA, LA and HMF were measured via HPLC.

Four adsorption kinetic models were used to evaluate the adsorption kinetics: the pseudo-first-order (PFO, diffusion process), pseudo-second-order (PSO, chemisorption process), intraparticle diffusion (Weber–Morris, intraparticle diffusion process), and the Boyd (intraparticle diffusion or liquid film diffusion) kinetic models. More details can be found in the SI.

Breakthrough experiment. A quaternary-component breakthrough experiment was performed to investigate the performance of the adsorbent for the separation and purification of HMF in practical chromatographic separation (Scheme S1). A total of 0.586 g of the as-synthesized MOF-808 (DMF and H₂O washed for two days each) was packed in an HPLC column (diameter of 4.6 mm, length of 100 mm, volume of 1.66 mL) under a high pressure water flow, and washed with 0.5 M HCl for 1 h at a flow rate of 0.5 mL min⁻¹, and then washed with H₂O for 10 h to activate the MOF-808 column. Using the mass of MOF-808 in the column and the crystal density of MOF-808 (0.86 g cm⁻³),³⁹ the porosity of the column could be calculated as 59%. The breakthrough experiment was performed at 30 °C with a flow rate of 0.5 mL min⁻¹ (u, 0.3 BV per min), and the feed concentrations of fructose, FA and LA were 5 mg g_sol⁻¹, and that of HMF was 20 mg g_sol⁻¹. The effluent was subsequently collected as a sample within a period of time, and the concentration was measured by HPLC. The uptake capacity of the column was calculated using the following equations:where q (mg g_ads⁻¹) stands for the uptake capacity of adsorbent in the column, u (mL min⁻¹) is the flow rate of the feed solution, ρ (g_sol/mL) is the density of the mixture solution, C₀ and C_out (mg g_sol⁻¹) are the inlet and outlet concentrations of the mixture solution, respectively, and m (g_ads) is the mass of adsorbent in the column.

The breakthrough curves for multiple components were fitted using the logistic function, and model parameters were obtained using Origin 2021 software and the Levenberg–Marquardt method.⁴² Here, the breakthrough time is defined as the time at C/C₀ = 0.1.⁴³

The details of the cycling stability test can be found in the SI.

Test conditions for HPLC. The HPX-87H column was selected for the analysis. The column temperature was 40 °C, mobile phase was 5 mM H₂SO₄ and the flow rate was 0.6 mL min⁻¹, and an RID was used to detect the signals at 50 °C. And the concentration was determined using the external standard method.

Results and discussion

MOF-808 preparation and characterization

We synthesized MOF-808 based on the literature reported method³⁹ and removed the FA occupying its unsaturated coordination sites using methanol⁴⁴ (Fig. 1a). This process was monitored by ¹H NMR and Raman spectroscopy, which showed FA was almost completely removed from MOF-808 by methanol washing (Fig. 1b and c). Meanwhile, a significant methoxyl signal was observed in MOF-808–MeOH (Fig. 1b and c). Yang et al. attributed this to the replacement of formate ions by methanol ions.^45–52 And the methanol anion disappeared when MOF-808–MeOH was added to water, and thus the uncoordinated sites of the Zr₆O₈ node were activated (Fig. 1a–c). The binding energy of Zr 3d shifted from 182.8 eV for MOF-808–FA to 182.5 eV for MOF-808–H₂O, which also demonstrated that the chemical environment of the Zr₆O₈ node of MOF-808 was distinctly changed (Fig. 1d). Moreover, the SEM image (Fig. 1e) and PXRD pattern (Fig. 1f) confirmed the high crystallinity and phase purity of the synthesized MOF-808. The N₂ physisorption characterization showed that the pore volume (0.691 mL g⁻¹), Brunauer–Emmett–Teller (BET) surface area (1733 m² g⁻¹), and pore width (1.42 nm) (Fig. 1g, S1 and S2) of the synthesized MOF-808 were comparable to those reported in the literature.^53,54 All the characterization results confirmed the successful synthesis of MOF-808.

Fig. 1 Synthesis and characterization of MOF-808. Schematic of the synthesis process (a). ¹H NMR (b), Raman (c)^# and XPS spectra (Zr 3d) (d) of MOF-808–X. SEM image (e), PXRD pattern (f), and N₂ physisorption isotherm (g) of MOF-808*. The inset of (g) shows the pore size distribution (PSD) of MOF-808 derived from the DFT fitting of the adsorption branch of its N₂ physisorption isotherm at 77 K. ^#The peak at 2868 cm⁻¹ was assigned to ν_CH of the formate ion, while those at 2928 cm⁻¹ and 2822 cm⁻¹ were assigned to ν_CH of the methoxy group. *MOF-808 was activated by degassing after washing with MeOH.

Adsorption performance analysis

To mimic the industrial products for separation, fructose, FA, LA, and HMF were dissolved in water, with the concentration of the first three components being 5 mg g_sol⁻¹, and that of HMF being 20 mg g_sol⁻¹. During the characterization, we observed that the uptake capacity of fructose in the quaternary competitive adsorption was negative under some conditions (Fig. S3), which was attributed to one component adsorption leading to a decrease in the uptake capacity of the other components in the competitive adsorption system.⁵⁵ Therefore, we corrected the uptake capacity according to the method derived from our previous work,⁵⁵ with several modifications to the quaternary competitive adsorption system (SI Section 3), and accessible codes were written (Data availability). In addition, several values in the results remained negative due to their uptake capacity being lower than that of the solvent, and these values were reported directly without zeroing.

The results of the quaternary competitive adsorption study revealed that MOF-808 could selectively adsorb carboxylic acid impurities FA and LA (Fig. 2a). This finding was consistent with the expected result that the hard Lewis acid Zr-sites could form strong coordination interactions with hard base carboxylate (RCOO⁻), thereby adsorbing and separating them from the solution system. To further investigate the adsorption selectivity of MOF-808 on HMF and the three impurity components, we further characterized its adsorption performance in a binary system. MOF-808 exhibited highly selective adsorption of impurities (Fig. 2b). The selectivity factors (α) were all higher than 40 (Fig. 2b), in which the α values of LA–HMF could be as high as 57.8, and the α values of FA–HMF and Fru–HMF were 41.3 and 47.4, respectively. Furthermore, it is noteworthy that the uptake capacities of MOF-808 for a single component of fructose (239 g g_ads⁻¹), FA (154 g g_ads⁻¹) and LA (315 g g_ads⁻¹) at the same concentration are very similar to those in binary competitive adsorption in the presence of HMF (which were 245 g g_ads⁻¹, 155 g g_ads⁻¹, and 303 g g_ads⁻¹, respectively) (Fig. 2b and c). This further demonstrated that the adsorption selectivity of MOF-808 for the aforementioned three impurities was very high, and the uptake capacity was barely affected by HMF. Considering that the separation of HMF from LA is the most difficult part in the purification process of HMF, we further investigated the adsorption selectivity of MOF-808 on LA and HMF at different ratios. In the binary system of LA and HMF, we measured the selective adsorption performance of MOF-808 when the concentration of HMF was constant and the concentration of LA was gradually decreased, which simulated the situation of LA concentration gradually decreasing in the practical adsorption process. The results showed that as the LA concentration gradually decreased, MOF-808 still preferentially adsorbed LA (Fig. 2e), and the selectivity factor gradually increased. Moreover, the uptake capacity was also comparable to that of single-component adsorption of LA at the same concentration (Fig. S4). It is particularly noteworthy that the uptake capacity of MOF-808 for LA was still higher than that of HMF when the initial concentration of HMF was 20 times that of LA, and the selectivity factor was as high as 376 (Fig. 2e). And when the concentration of LA was further decreased to 0.5 mg g_sol⁻¹, after reaching equilibrium, LA was almost completely absent from the solution (Fig. S5), which suggested that MOF-808 possessed the ability to completely remove LA from HMF. These promising findings indicated that the difficulty of LA and HMF separation could be effectively tackled by employing MOF-808 as an adsorbent.

Fig. 2 Adsorption performance characterization of MOF-808. (a) Quaternary competitive adsorption performance of MOF-808 before and after correction. (b) Binary competitive adsorption performance. (c) Single component adsorption performance. (d) Ternary competitive adsorption performance in 2-MeTHF and THF. (e) Binary competitive adsorption performance of MOF-808 at different ratios of LA (10 mg g_sol⁻¹ to 1 mg g_sol⁻¹) and HMF (20 mg g_sol⁻¹). (f) Quaternary competitive adsorption performance at different temperatures (30 °C, 40 °C, 50 °C). (g) Uptake capacity on fructose, FA, LA and HMF at different time intervals. Fitted data of FA (h) and LA (i) derived from the pseudo-second-order kinetic model, t/q_t to t. Note: adsorption experiments were performed at 30 °C and 140 rpm for 2 h, the concentrations of fructose, FA and LA were 5 mg g_sol⁻¹, and that of HMF was 20 mg g_sol⁻¹ (unless otherwise specified).

Effect of solvent on adsorption performance

Water is widely employed as the solvent for preparation of HMF due to its cheap, green and non-toxic properties. However, biphasic solvent systems are another widely used reaction medium for the synthesis of HMF,^56–58 in which the organic phase can extract the generated HMF in the aqueous phase and separate it from the reaction system instantaneously to minimize the side reactions and improve the HMF yield. For example, tetrahydrofuran (THF) and 2-methyltetrahydrofuran (2-MeTHF)^59–62 are two commonly used organic solvents in HMF synthesis. Therefore, we investigated the adsorption performance of MOF-808 in these two solvents. Because of the low solubility of fructose in the two solvents, only the competitive adsorption of FA, LA, and HMF in MOF-808 was investigated. The results showed that MOF-808 could still selectively adsorb FA and LA in these two organic solvents (Fig. 2d); especially in THF, the uptake capacity of HMF was only 9.1 mg g_ads⁻¹, which might be attributed to the strong hydrogen bonding interaction between HMF (hydrogen bond donor) and THF (hydrogen bond acceptor).⁶³ This suggested that MOF-808 could still be used to purify HMF in these two organic solvents.

Effect of temperature on adsorption performance

Variations in temperature change the viscosity of the solution to be separated and consequently the column pressure, while the adsorption selectivity of the adsorbent might also be changed; thus, it is necessary to investigate the effect of temperature. As the temperature increased, the results showed that the uptake capacity of FA decreased slightly, that of LA remained almost unchanged, and that of fructose and HMF changed negligibly at 40 °C and increased slightly at 50 °C (Fig. 2f). However, all the variations were minimal, suggesting that the temperature has a negligible effect on the quaternary components' competitive adsorption in MOF-808.

Adsorption kinetics analysis

Adsorption rate is another critical parameter in evaluating the performance of adsorbents as it directly influences the efficiency of adsorption separations. To ensure the practicality of the adsorption kinetics, we investigated the adsorption kinetics of MOF-808 under quaternary competitive adsorption. The results showed that after only 1 min of adsorption, MOF-808 achieved 89% and 76% of its equilibrium adsorption capacity for FA and LA, respectively (Fig. 2g). Subsequently, the uptake capacity of FA and LA initially decreased slightly before gradually increasing to equilibrium, whereas fructose and HMF showed a slight increase followed by a gradual decline to equilibrium (Fig. 2g). This trend can be attributed to the rapid initial adsorption of FA and LA, which reduced their free concentrations in the pore; consequently, a small number of adsorption sites were replaced by the high concentration of fructose and HMF. However, as FA and LA diffused into the pore, the sites occupied by fructose and HMF were gradually replaced, which also indicated that the interactions of MOF-808 with FA and LA were stronger than those with fructose and HMF. After fitting the kinetic data with different models, the adsorption behaviors of FA and LA conformed to the pseudo-second order (PSO) adsorption kinetic model where the goodness of fit (R²) could reach 0.999 (Fig. 2h and i). This indicated that their adsorption rates were controlled by the adsorption process of their adsorption onto the Zr sites of MOF-808, which was consistent with the speculation that FA and LA were adsorbed through coordination with Zr sites (chemisorption process), and the kinetic constants of FA and LA were as high as 58.0 h⁻¹ and 25.9 h⁻¹, respectively. However, perhaps affected by the molecular size (Fig. S6), LA adsorption was also partially influenced by its diffusion process in the MOF-808 pores (Fig. S7 and S8).

Adsorption mechanism of MOF-808

Our previous study showed that the uncoordinated Zr sites in MOF-808 could effectively adsorb fructose, hence MOF-808 could adsorb fructose with high selectivity in the binary competitive adsorption of fructose and HMF.⁵⁵ However, the adsorption mechanism of FA and LA by MOF-808 remains to be further investigated. We envisioned utilizing the unsaturated coordination Zr sites in MOF-808 to form strong coordination interactions with FA and LA resulting in the separation of FA and LA from HMF. Here, we aimed to confirm this conception by adjusting the defect concentration in the same Zr-based UiO-66 framework and examining the adsorption performance on FA and LA. The concentration of defects in the synthesized UiO-66-X could be regulated by modulating the harshness of the synthesis conditions.⁴¹ The results from ICP-AES (Fig. 3a) showed the Zr content in the synthesized UiO-66-X increased progressively with the severity of the synthesis conditions (from UiO-66-Std to UiO-66-Def), indicating a gradual increase in the number of Zr sites, which had also been verified by ³¹P-TMP SSNMR in our previous work (Fig. 3b).⁵⁵ Subsequently, we examined the quaternary competitive adsorption performance of UiO-66-X and showed that the uptake capacity of UiO-66-X for FA and LA could improve with the increase of Zr sites in its framework (Fig. 3c). This indicated that the Zr sites could indeed adsorb FA and LA efficiently, which was consistent with our hypothesis. Moreover, different from the competitive adsorption, the uptake capacity of MOF-808 on HMF could reach 105 mg g_ads⁻¹ when the concentration of single-component HMF was 20 mg g_sol⁻¹, and its uptake capacity could be as high as 203 mg g_ads⁻¹ when the concentration was increased to 100 mg g_sol⁻¹ (Fig. 3d). This indicated that despite its pore width reaching 1.42 nm (Fig. 1g), MOF-808 could still adsorb HMF through the π-electrons within its structure. However, due to the weaker π–π interaction of HMF with the benzene ring than the coordination interactions of fructose, FA and LA with the Zr sites, MOF-808 showed selective adsorption of the three impurities in the competitive adsorption.

Fig. 3 Adsorption mechanism of MOF-808. (a) Rate of increase of zirconium content in UiO-66-X compared to the theoretical value. (b) Characterization of the concentration of Zr sites in UiO-66-X by ³¹P solid-state NMR, data from our previous work (ref. 55). (c) Quaternary competitive adsorption in UiO-66-X. (d) Uptake capacity of MOF-808 for HMF at different concentrations. (e) Quaternary competitive adsorption performances of HKUST-1, MIL-100-Al, and MOF-808. (f) Schematic structure of the SBU of MIL-100-Al and HKUST-1. Note: adsorption experiments were performed at 30 °C and 140 rpm, the concentrations of fructose, FA and LA were 5 mg g_sol⁻¹, and that of HMF was 20 mg g_sol⁻¹ (unless otherwise specified).

The unique uncoordinated paired sites in MOF-808, which could form a stable bridging coordination structure with carboxylate (RCOO⁻), might also be another significant factor for its efficient adsorption of carboxylic acid impurities.³⁶ To verify this hypothesis, we measured the adsorption performance of MIL-100-Al, which has a similar BET surface area, comparable pore volume, and the same carboxylate linker (1,3,5-benzenetricarboxylate, BTC) as MOF-808 (Fig. S9). According to the hard/soft acid/base (HSAB) principle, aluminum(III), which is comparable to zirconium(IV) in acid hardness, can also form strong coordination interactions with hard base carboxylates (RCOO⁻).^31–35 However, MIL-100-Al was unable to adsorb FA and LA with high selectivity (Fig. 3e), probably due to the fact that the uncoordinated sites of its metal node, Al₃O(COO)₆, were all mono-coordinated sites⁶⁴ (Fig. 3f). Moreover, HKUST-1, also with only mono-coordinated sites⁶⁴ (Fig. 3f), which consists of lower acid-hardness Cu(II) connected with the same carboxylate linker (BTC) and has a similar BET surface area and pore volume to MOF-808 (Fig. S9), showed a further decrease in adsorption of FA and LA (Fig. 3e). And HKUST-1 exhibited complementary adsorption selectivity to MOF-808, namely, HKUST-1 selectively adsorbed HMF but less adsorbed impurities in the binary competitive adsorption of each impurity with HMF (Fig. S10). This suggested that we might be able to regulate the adsorption selectivity of MOFs for HMF and impurities by modulating the hardness of the central metal and the mode of the metal nodes.

Breakthrough experiment and cycle stability test

After systematic characterization, we concluded that MOF-808 could be used to separate and purify HMF. To explore the practical column adsorption separation performance, we conducted a breakthrough experiment of a four-component mixed solution in the MOF-808 column. Consistent with the static adsorption, HMF was the first component to break through (4.5 min, C/C₀ = 0.1), reaching 80% of the feed concentration in the effluent at 6 min, followed by fructose (13.5 min), FA (21.0 min), and LA (33.0 min) in sequence to break through the MOF-808 column (Fig. 4a and S11). And it is noteworthy that the concentration of fructose could reach twice that of the original feed solution at 16 min, which resulted from the replacement of the adsorbed fructose in MOF-808 by FA and LA into the solution (Fig. 4a). This revealed that the interaction of fructose with the Zr sites was lower than the coordination interaction of carboxylate (RCOO⁻) with the paired sites. Moreover, the steep breakthrough curves further demonstrated the fast adsorption kinetics of the components in MOF-808. From the above breakthrough experiment, the effluent of the MOF-808 column contained only HMF before 12 min, which indicated that HMF could be purified with nearly 100% purity by the MOF-808 column (Fig. 4a). This demonstrated that the MOF-808 column had outstanding separation performance of HMF from impurities. We then investigated the cycling stability of the MOF-808 column. An obvious increase in FA uptake capacity was observed in the second cycle (Fig. 4b), which could be attributed to residual FA from the synthesis process that was still coordinated to the starting MOF-808.39 This could be verified by the noticeable desorption peak in the breakthrough curve of FA (Fig. 4a) and the delayed breakthrough time in the subsequent cycles (Fig. S12). After sufficient activation, the uptake capacity of FA was improved. Moreover, the uptake capacity of LA decreased obviously, which might be caused by LA having both carboxylic acid and ketone carbonyl in its structure (Scheme 1), and MOF-808 had a small amount of strong Lewis acid sites in its framework.⁵⁵ Hence we hypothesize that a small amount of LA occupied the strong sites through multiple coordination, forming a stable structure that was difficult to be desorbed. This could be verified in its desorption curve, which indicates that a low but non-negligible concentration of LA still flowed out after 1.5 h of desorption (Fig. S13). And an obvious signal of LA was found in MOF-808 digestion ¹H NMR after 5 cycles (Fig. S14). The occupation of the pore space also led to the decrease of the pore volume (Fig. 4c and S15). In the subsequent cycling test, both the uptake capacity (Fig. 4b) and the breakthrough time of fructose, LA and FA (Fig. 4d and S12) showed good stability. The PXRD pattern of MOF-808 after 5 cycling tests also indicated the stability of its crystal structure (Fig. 4e), suggesting that the MOF-808 column possessed excellent cycling stability. And the recovery rate of HMF could reach 98% after only 18 min (5.42 BV) of desorption (Fig. S13). These results indicated that MOF-808 could not only efficiently separate HMF and impurities (especially LA), but also would not cause obvious loss of HMF, making it a promising adsorbent in the separation and purification of HMF (Fig. 4f).^{10,15,21–23}

Fig. 4 Breakthrough experiment and cycle stability of the MOF-808 column. (a) Breakthrough experiment of a four-component mixed solution in the MOF-808 column. (b) Cyclic performance of the MOF-808 column. (c) N₂ physisorption isotherms of MOF-808 before and after 5 cycling tests. (d) Breakthrough curves of LA in cycling tests. (e) PXRD patterns of MOF-808 before and after 5 cycling tests. (f) Comparison of the FA–HMF selectivity factor (SF_LA) and HMF recovery (R_HMF) of adsorbents. Note: the length and diameter of the MOF-808 column was 100 mm and 4.6 mm, the column volume was 1.66 mL, the flow rate was 0.5 mL min⁻¹ (0.3 BV per min), the concentrations of fructose, FA and LA were 5 mg g_sol⁻¹, the concentration of HMF was 20 mg g_sol⁻¹, and the breakthrough experiments were performed at 30 °C.

Conclusions

The efficient separation and purification of HMF was achieved by introducing coordination interaction. MOF-808 exhibited high-capacity and high-selectivity adsorption performance for fructose (245 mg g_ads⁻¹, 47.4), FA (155 mg g_ads⁻¹, 41.3), and LA (303 mg g_ads⁻¹, 57.8). And the selectivity of MOF-808 for LA was further improved with the decrease of LA concentration until it was completely removed from the solution. In quaternary competitive adsorption, MOF-808 could rapidly and selectively adsorb carboxylic acid impurities (FA and LA), and the adsorption kinetic constants could reach 58.0 h⁻¹ and 25.9 h⁻¹ for FA and LA, respectively. Moreover, the adsorption performance of MOF-808 was almost unchanged at different temperatures (30 °C, 40 °C, and 50 °C). MOF-808 could efficiently separate and purify HMF attributed to the high acid hardness of Zr(IV) and the abundance of uncoordinated paired sites in the 6-connected Zr₆O₈ node. And in practical column separation, MOF-808 demonstrated remarkable separation performance and cycling stability, with HMF recovery as high as 98%. In conclusion, all three impurities can be efficiently separated from HMF by introducing coordination interaction, and this strategy can inform the design of adsorbents for the purification of HMF.

Author contributions

All of the authors have approved the final version of the manuscript.

Conflicts of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The code used to correct the equilibrium concentration has been deposited to Github (https://github.com/Wang-Pu-BEM/Ce_Correction_integrated/tree/master). The data supporting this article have been included as part of the SI.

Supplementary information: experimental details, adsorbent performance comparison, derivation of the concentration correction equation, diagram of molecular structure, details of the adsorption kinetics study, and cycling stability characterization. See DOI: https://doi.org/10.1039/d5ta05030h.

Acknowledgements

This project was supported by the National Natural Science Foundation of China (No. 22378333), and we acknowledge the Westlake Education Foundation and the Research Center for Industries of the Future (RCIF, WU2023A005) at Westlake University for its support. The authors would like to acknowledge the Instrumentation and Service Center for Molecular Sciences and the Instrumentation and Service Center for Physical Sciences at Westlake University for the use of equipment and technical support. We sincerely thank Pu Wang for refining the equilibrium concentration correction code to allow efficient processing of data for ternary and quaternary competitive adsorption. The authors are grateful to Zhen Yang, Wenjing Cao, Min Zhou, Dr Chao Zhang, and Dr Xiaohe Miao for assistance with the measurements.

References

1. Q. Hou, X. Qi, M. Zhen, H. Qian, Y. Nie, C. Bai, S. Zhang, X. Bai and M. Ju, Green Chem., 2021, 23, 119–231.
2. L. Guo, X. X. Zhang, L. Gan, L. Pan, C. X. Shi, Z. F. Huang, X. W. Zhang and J. J. Zou, Adv. Sci., 2023, 10, 2205540.
3. Y. An, T. Lei, W. Jiang and H. Pang, Green Chem., 2024, 26, 10739–10773.
4. C. Chen, M. Lv, H. Hu, L. Huai, B. Zhu, S. Fan, Q. Wang and J. Zhang, Adv. Mater., 2024, 36, 2311464.
5. D. X. Chen, W. L. Li, J. B. Liu and L. C. Sun, Energy Environ. Sci., 2025, 18, 3120–3128.
6. J. J. Bozell and G. R. Petersen, Green Chem., 2010, 12, 539–554.
7. M. Bicker, D. Kaiser, L. Ott and H. Vogel, J. Supercrit. Fluids, 2005, 36, 118–126.
8. K. I. Galkin and V. P. Ananikov, ChemSusChem, 2019, 12, 2976–2982.
9. Y. W. Hsiao, A. Anastasopoulou, M. Ierapetritou and D. G. Vlachos, Green Chem., 2021, 23, 4008–4023.
10. X.-f. Chen, H.-l. Li, X.-r. Ji, Z.-j. Shen, H.-j. Guo, S.-m. Yao, M.-k. Wang, L. Xiong and X.-d. Chen, Sep. Purif. Technol., 2023, 315, 123661.
11. J. Zheng, X. He, C. Cai, J. Xiao, Y. Liu, Z. Chen, B. Pan and X. Lin, Chemosphere, 2020, 239, 124732.
12. L. Hu, Y. Jiang, Z. Wu, X. Wang, A. He, J. Xu and J. Xu, J. Clean. Prod., 2020, 276, 124219.
13. M. León, T. D. Swift, V. Nikolakis and D. G. Vlachos, Langmuir, 2013, 29, 6597–6605.
14. C. Detoni, C. H. Gierlich, M. Rose and R. Palkovits, ACS Sustain. Chem. Eng., 2014, 2, 2407–2415.
15. H. Guo, Y. Isoda, T. Honma, F. Shen and R. L. Smith, Colloids Surf., A, 2022, 637, 128187.
16. M. Yabushita, P. Li, H. Kobayashi, A. Fukuoka, O. K. Farha and A. Katz, Chem. Commun., 2016, 52, 11791–11794.
17. H. Jin, Y. Li, X. Liu, Y. Ban, Y. Peng, W. Jiao and W. Yang, Chem. Eng. Sci., 2015, 124, 170–178.
18. R. Ranjan, S. Thust, C. E. Gounaris, M. Woo, C. A. Floudas, M. v. Keitz, K. J. Valentas, J. Wei and M. Tsapatsis, Microporous Mesoporous Mater., 2009, 122, 143–148.
19. R. Xiong, M. León, V. Nikolakis, S. I. Sandler and D. G. Vlachos, ChemSusChem, 2013, 7, 236–244.
20. W. C. Yoo, N. Rajabbeigi, E. E. Mallon, M. Tsapatsis and M. A. Snyder, Microporous Mesoporous Mater., 2014, 184, 72–82.
21. Y.-B. Zhang, Q.-X. Luo, M.-H. Lu, D. Luo, Z.-W. Liu and Z.-T. Liu, Chem. Eng. J., 2019, 358, 467–479.
22. J. Zheng, L. Hu, X. He, Y. Liu, X. Zheng, S. Tao and X. Lin, Ind. Eng. Chem. Res., 2020, 59, 17575–17586.
23. L. Hu, S. Tao, J. Xian, X. Zhang, Y. Liu, X. Zheng and X. Lin, Chin. J. Chem. Eng., 2022, 43, 230–239.
24. L. Hu, Y. Wu, M. Li, X. Zhang, X. Xian, Y. Mai and X. Lin, Sep. Purif. Technol., 2022, 299, 121756.
25. Z. Hasan and S. H. Jhung, J. Hazard. Mater., 2015, 283, 329–339.
26. S. Zhang, J. Wang, Y. Zhang, J. Ma, L. Huang, S. Yu, L. Chen, G. Song, M. Qiu and X. Wang, Environ. Pollut., 2021, 291, 118076.
27. P. Wu, M. Guo, R.-W. Zhang, Q. Huang, G. Wang and Y.-Q. Lan, Chem. Sci., 2024, 15, 17781–17798.
28. M. Xu, W.-Q. Tang, S.-S. Meng and Z.-Y. Gu, Chem. Soc. Rev., 2025, 54, 1613–1633.
29. J. Pang, W. Jiang, X.-W. Zhang, H.-L. Zhou, Y. Sun, W. Gong, B. Wang, F. Ma, L. He, L. Chen, Q. Chen, H.-Q. Zheng, Y. Cui, W.-J. Shi, C. Zhang, M. Wang, Y.-B. Zhang, J.-P. Zhang, C. Zhong, Y. Cui, S. Wang, J.-R. Li, G. Qian, T.-B. Lu, H. Pang and X.-H. Bu, Sci. China Chem., 2025, 68, 1642–1702.
30. Z. Li, Z. Wang, H. Wang, Y. Shi, Y. Zhao and H. Wang, ACS Sustain. Chem. Eng., 2025, 13, 3985–3993.
31. D. Yang and B. C. Gates, J. Phys. Chem. C, 2024, 128, 8551–8559.
32. Y. An, X. Lv, W. Jiang, L. Wang, Y. Shi, X. Hang and H. Pang, Green Chem. Eng., 2024, 5, 187–204.
33. T. He, X. J. Kong and J. R. Li, Acc. Chem. Res., 2021, 54, 3083–3094.
34. S. Yuan, L. Feng, K. Wang, J. Pang, M. Bosch, C. Lollar, Y. Sun, J. Qin, X. Yang, P. Zhang, Q. Wang, L. Zou, Y. Zhang, L. Zhang, Y. Fang, J. Li and H. C. Zhou, Adv. Mater., 2018, 30, e1704303.
35. L. Feng, K. Y. Wang, G. S. Day, M. R. Ryder and H. C. Zhou, Chem. Rev., 2020, 120, 13087–13133.
36. D. Yang, M. Babucci, W. H. Casey and B. C. Gates, ACS Cent. Sci., 2020, 6, 1523–1533.
37. R. Zhang, L. Hu, S. Bao, R. Li, L. Gao, R. Li and Q. Chen, J. Mater. Chem. A, 2016, 4, 8412–8420.
38. C. Volkringer, D. Popov, T. Loiseau, G. Férey, M. Burghammer, C. Riekel, M. Haouas and F. Taulelle, Chem. Mater., 2009, 21, 5695–5697.
39. H. Furukawa, F. Gandara, Y. B. Zhang, J. Jiang, W. L. Queen, M. R. Hudson and O. M. Yaghi, J. Am. Chem. Soc., 2014, 136, 4369–4381.
40. G. Lu, C. Cui, W. Zhang, Y. Liu and F. Huo, Chem.–Asian J., 2012, 8, 69–72.
41. J. Duncan, D. Sengupta, S. Bose, K. O. Kirlikovali and O. K. Farha, Sustain. Chem. Environ., 2023, 3, 100032.
42. Q. Hu, D. Wang, S. Pang and L. Xu, Arab. J. Chem., 2022, 15, 104034.
43. K.-H. Chen, Y.-R. Lai, N. T. D. Hanh, S. S. S. Wang and Y.-K. Chang, Membranes, 2023, 13, 761.
44. C. M. Jia, F. G. Cirujano, B. Bueken, B. Claes, D. Jonckheere, K. M. Van Geem and D. De Vos, Chemsuschem, 2019, 12, 1256–1266.
45. D. Yang, V. Bernales, T. Islamoglu, O. K. Farha, J. T. Hupp, C. J. Cramer, L. Gagliardi and B. C. Gates, J. Am. Chem. Soc., 2016, 138, 15189–15196.
46. X. Chen, Y. Lyu, Z. Wang, X. Qiao, B. C. Gates and D. Yang, ACS Catal., 2020, 10, 2906–2914.
47. D. Yang, C. A. Gaggioli, D. Ray, M. Babucci, L. Gagliardi and B. C. Gates, J. Am. Chem. Soc., 2020, 142, 8044–8056.
48. Y. Xiao, L. Han, L. Zhang, B. C. Gates and D. Yang, J. Phys. Chem. Lett., 2021, 12, 6085–6089.
49. D. Yang and B. C. Gates, Acc. Chem. Res., 2021, 54, 1982–1991.
50. S. Das, D. Yang, E. T. Conley and B. C. Gates, ACS Catal., 2023, 13, 14173–14188.
51. D. Dittmann, J. Schröder, E. Kaya, J. Mosrati, A. M. Abdel-Mageed and M. Dyballa, J. Phys. Chem. C, 2023, 127, 18962–18970.
52. Y. Xiao, B. C. Gates and D. Yang, ACS Appl. Mater. Interfaces, 2024, 16, 52445–52454.
53. H. Lyu, O. I.-F. Chen, N. Hanikel, M. I. Hossain, R. W. Flaig, X. Pei, A. Amin, M. D. Doherty, R. K. Impastato, T. G. Glover, D. R. Moore and O. M. Yaghi, J. Am. Chem. Soc., 2022, 144, 2387–2396.
54. O. I.-F. Chen, C.-H. Liu, K. Wang, E. Borrego-Marin, H. Li, A. H. Alawadhi, J. A. R. Navarro and O. M. Yaghi, J. Am. Chem. Soc., 2024, 146, 2835–2844.
55. H.-L. He, Z. Liu, F. Liu, J. Chen, P. Wang, X. Yi, A. Zheng and L. Wang, J. Mater. Chem. A, 2024, 12, 26214–26223.
56. Y. Román-Leshkov, J. N. Chheda and J. A. Dumesic, Science, 2006, 312, 1933–1937.
57. G. W. Huber, J. N. Chheda, C. J. Barrett and J. A. Dumesic, Science, 2005, 308, 1446–1450.
58. J. Liang, J. Jiang, T. Cai, C. Liu, J. Ye, X. Zeng and K. Wang, Green Energy Environ., 2024, 9, 1384–1406.
59. K. Wang, A. Rezayan, L. Si, Y. Zhang, R. Nie, T. Lu, J. Wang and C. Xu, ACS Sustain. Chem. Eng., 2021, 9, 11351–11360.
60. A. Rezayan, K. Wang, R. Nie, T. Lu, J. Wang, Y. Zhang and C. C. Xu, Chem. Eng. J., 2022, 429, 132261.
61. W. Guo, E. J. M. Hensen, W. Qi, H. J. Heeres and J. Yue, ACS Sustain. Chem. Eng., 2022, 10, 10157–10168.
62. J. Liang, C. Liu, J. Jiang and K. Wang, Angew Chem. Int. Ed. Engl., 2024, 63, e202410229.
63. H. Wang, J. Cui, Y. Zhao, Z. Li and J. Wang, Green Chem., 2021, 23, 405–411.
64. M. J. Kalmutzki, N. Hanikel and O. M. Yaghi, Sci. Adv., 2018, 4, eaat9180.

---