Acid–Base Property of Tetragonal YNbO4 with Phosphate Groups and Its Catalysis for the Dehydration of Glucose to 5‐Hydroxymethylfurfural

Acid–base properties of amphoteric crystalline YNbO4, which is synthesized by co‐precipitation method using a water‐soluble Nb peroxo complex, are adjusted by the impregnation of phosphate groups. The resulting phosphate–YNbO4 materials are tested for glucose dehydration to 5‐hydroxymethylfurfural (HMF). Both activity and selectivity increase by the introduction of phosphate groups, giving the optimal catalyst with five phosphate groups per nm2 of surface (5P‐YNbO4). 5P‐YNbO4 shows ≈50% selectivity at 75% conversion. Pyridine adsorption experiments show that 5P‐YNbO4 exclusively possesses Lewis acidity. NH3‐ and CO2‐temperature programmed desorption measurements confirm that phosphoric acid treatment increases acid strength and acid density with concurrent loss of basicity. Experiments using isotope‐labeled glucose clarify that HMF is formed on 5P‐YNbO4 through the stepwise dehydration reaction mechanism. Recyclability experiments show an accumulation of carbon deposits at the end of each reaction causing deactivation, but the activity can be completely recovered by successive washing with water and subsequent calcination at 400 °C in air.


Introduction
Climate emergency and fast depletion of fossil fuels are signaling that a shift to sustainable resources must become the fundaments to the sustainable development of our future society. Great impulse has then been given in the past decade to biomass resources, due to their renewable nature, worldwide availability, and low cost. Lignocellulosic biomass has attracted the most attention among the scientific communities. Cellulose, the main component of lignocellulose biomass, is a linear polymer of D-glucose units linked by β(1!4) glycosidic bonds. Depolymerization of cellulose is typically achieved by acid catalysis producing glucose and its oligomers. [1,2] Glucose can be transformed into several platform molecules that can be used by industry to deliver commodities such as polymers and liquid fuels. One of the attractive products obtainable from glucose is 5-hydroxymethylfurfural (HMF) (Scheme 1).
The US department of energy identified HMF as one of the top 10 biobased chemicals. [3] The reason behind its great importance relies on its unique structure that allows a huge variety of further transformations, making it possible to cover several industrial sectors. One of the most promising compounds synthesized from HMF is 2,5-furandicarboxylic acid (FDCA). Polymerization of FDCA with renewable ethylene glycol produces polyethylene furanoate (PEF), a 100% renewable polymer alternative to fossil fuel-derived polyethylene terephthalate (PET). [4][5][6] HMF can be synthesized from glucose by Lewis acid and Brønsted acid catalysts (Scheme 1). [7,8] Strong Brønsted acids such as HCl and H 2 SO 4 can facilitate the dehydration of glucose to HMF, but the highly corrosive nature makes their utilization extremely limited for their practical application. Homogeneous Lewis acids such as AlCl 3 , CrCl 3 , and rare earth triflates also catalyze the reaction, but difficulties in their separation and recovery remain a major drawback as well as homogeneous Brønsted acids. [9,10] Water-tolerant Lewis acid sites on Nb 2 O 5 ·nH 2 O [11][12][13][14] and TiO 2 [15][16][17][18][19][20] were found to be effective for glucose dehydration in water and HMF yields were significantly improved after the introduction of phosphate groups on their surfaces. Water is supposed to deactivate or decompose Lewis acid sites on metal oxides by the formation of very stable Lewis acid-water adducts or metal hydroxide species, respectively. However, Lewis acidic NbO 4 tetrahedra on amorphous Nb 2 O 5 .nH 2 O were reported to catalyze the dehydration of glucose even in the presence of water. Nb 2 O 5 .nH 2 O also performed well the dehydration of xylose to furfural in a biphasic system (H 2 O-toluene) leading to 77% selectivity at nearly complete conversion. [21] Lewis acidic TiO 4 tetrahedra on low-crystalline and anatase TiO 2 work together with Lewis basic lattice oxygens to promote the dehydration of glucose to HMF. [15,17] Such concerted catalysis by Lewis acid and base pair sites is specific for amphoteric oxides and not observed with a typical acidic oxide without an effective Lewis base site such as Nb 2 O 5 .nH 2 O.
A serious drawback of these oxide catalysts in the application of sugar conversion is the loss of original activity after thermal treatment. The dehydration of sugars always involves complex side reactions to form polymerized byproducts so-called humins, which are deposited on the surface of solid catalysts. Calcination treatment (>500°C) in the presence of oxygen is widely applicable to the used catalysts to remove detrimental organic deposits from the surface, which recovers original activity in extensive reaction cycles. However, the well-established procedure to regenerate the activity is strictly limited only to thermally stable solid catalysts. Nb 2 O 5 .nH 2 O retains its intrinsic acidity only in its amorphous phase and becomes less performant in many acidcatalyzed reactions after crystallization upon thermal treatment. [22,23] Calcination of low crystalline anatase TiO 2 also loses its initial activity due to the decrease in surface area as well as active Lewis acid and base sites. Therefore, the reusability of such amorphous or low-crystalline materials in the dehydration of sugars is greatly hampered by the need for calcination treatment, even if these are tentatively stable within several reaction cycles as reported in previous papers. Accordingly, crystalline and thermally stable catalysts are preferable, as they would better withstand structural changes due to thermal post-treatment.
Recently, we have studied Lewis acid and base catalysis of an Nb-based complex oxide, YNbO 4 , in the condensation of benzaldehyde with acetylacetone, the retro-aldol reaction of diacetone alcohol, and lactic acid formation from glucose or a triose sugar (1,3-dihydroxyacetone). [24] The amphoteric nature of this material was conferred by Lewis acidic Nb centers derived from NbO 4 tetrahedra and Lewis basic lattice oxygens due to vicinal and more electropositive Y atoms. We found that the acid-base property of YNbO 4 changes by calcination treatment. Amorphous YNbO 4 calcined at 80°C has both Lewis acid and base sites and converts glucose to lactic acid (19.6% yield). In contrast, tetragonal YNbO 4 calcined at 700°C has mainly Lewis acid sites and facilitates the dehydration of glucose to HMF (11.6% yield). The main difference in product selectivity between these two YNbO 4 catalysts is most likely attributed to Lewis base sites.
In this work, we explore tetragonal YNbO 4 in the dehydration of glucose to HMF. Phosphoric acid treatment was conducted to enhance HMF selectivity, because TiO 2 and Nb 2 O 5 .nH 2 O increased HMF selectivity from 9% to 81% and 12% to 52%, respectively, after the introduction of phosphate groups on their surfaces. [15] However, the increase in HMF selectivities of TiO 2 and Nb 2 O 5 .nH 2 O after the introduction of the phosphate group is not fully understood: the phosphate group itself is not active for HMF formation. The activity of phosphate-YNbO 4 was compared systematically with reference catalysts including amorphous Nb 2 O 5 .nH 2 O and crystallized Nb 2 O 5 . We have studied the structure-activity relationship of tetragonal YNbO 4 before and after phosphoric acid treatment to reveal the influence of the phosphate group on acid-base property as well as HMF selectivity. The crystalline structure of tetragonal YNbO 4 is suitable for the regeneration of the used catalyst by calcination treatment in air. We made profitable use of the potential stability and conducted recycling experiments of the used phosphate-YNbO 4 .  [25,26] Phosphorylation of tetra-YNbO 4 was conducted by impregnating a phosphoric acid solution and subsequent thermal condensation at 700°C in the air to assure a thermally stable material. The content of the phosphate group varies based on surface concentration. Based on the surface area of tetra-YNbO 4 (35 m 2 g À1 ), we immobilized two, five, and ten phosphate groups per 1 nm 2 . Three different phosphate catalysts are nominally denoted as 2P-YNbO 4 , 5P-YNbO 4 , and 10P-YNbO 4 . The phosphate loadings of P-YNbO 4 , 5P-YNbO 4 , and 10P-YNbO 4 correspond to 0.9, 2.4, and 4.8 wt %, assuming that phosphate moiety is present in the form of M─O─P ¼ O(OH) 2 (M ¼ Nb and/or Y). X-ray diffraction (XRD) patterns showed no change in the original diffraction features of the tetragonal YNbO 4 phase after phosphoric acid impregnation ( Figure S1, Supporting Information), while a small but gradual decrease of surface area was observed with the increase in the phosphate loading ( Table 1).

Results and Discussion
We explored these YNbO 4 samples in glucose dehydration reaction (Scheme 2) using a biphasic system (Table 1), since continuous extraction of HMF from water to organic phase minimizes by-product formation, improving HMF selectivity and catalyst lifetime. [27] The biphasic reaction system was designed with water, methyl tetrahydropyran (MTHP) as an extraction solvent, and NaCl as a promoter to enhance HMF extraction from the water phase. Phosphorylation of tetra-YNbO 4 led to more performant materials, increasing both glucose conversion and HMF yield, while carbon balance was %60% in all runs (entries [1][2][3][4]. Moderate values of carbon balance can be largely attributed to the formation of insoluble humins, which are visible as brown deposits. 5P-YNbO 4 and 10P-YNbO 4 were more active than bare YNbO 4 and 2P-YNbO 4 , giving %80% and 50% in glucose conversion and HMF selectivity, respectively. Almost no difference in the activity between the two phosphate catalysts suggests that the optimal concentration of phosphate moiety on the tetra-YNbO 4 surface is obtained in 5P-YNbO 4 . 5P-YNbO 4 was compared with niobium oxide, yttrium oxide, and some conventional Lewis and Brønsted acid catalysts. Amorphous Nb 2 O 5 ·nH 2 O and its phosphate form (5P-Nb 2 O 5 ·nH 2 O) were used as reference materials for comparison (entries 5, 6). 5P-Nb 2 O 5 .nH 2 O was synthesized by similar procedures to 5P-YNbO 4 , but the final calcination treatment at 700°C in air was eliminated to retain the amorphous nature of the resulting material. Nb 2 O 5 ·nH 2 O also improved both HMF yield and selectivity after the introduction of the phosphate group, but their activities were lower than that of 5P-YNbO 4 . Crystallized Nb 2 O 5 , which was synthesized by calcination of amorphous Nb 2 O 5 ·nH 2 O at 700°C for 5 h ( Figure S2, Supporting Information), and its phosphate form (5P-Nb 2 O 5 ) were also examined in glucose dehydration. The phosphate group was impregnated on crystallized Nb 2 O 5 by the same procedures as 5P-YNbO 4 and finally stabilized by thermal condensation at 700°C for 1 h. Crystallization of amorphous Nb 2 O 5 ·nH 2 O significantly decreased surface area from 197 to 9 m 2 g À1 as well as the activity, giving poor HMF yield (entry 7). The poor activity obtained also with 5P-Nb 2 O 5 (entry 8) demonstrates that crystallized Nb 2 O 5 is ineffective for glucose dehydration even after phosphorylation. The loss of activity after crystallization is well documented in previous papers and can be attributed to the loss of Brønsted and Lewis acid sites accompanied by a severe reduction of surface area. [28,29] Y 2 O 3 and 5P-Y 2 O 3 were unselective for HMF formation (entries 9 and 10), which is most likely attributed to high basicity that largely promotes the formation of humins, as evidenced by very low values of carbon balance (<10%). These results suggest that acid-base catalysis of tetra-YNbO 4 in glucose dehydration is largely distinct from those of amorphous Nb 2 O 5 ·nH 2 O, crystallized Nb 2 O 5 , and Y 2 O 3 .
We examined a series of conventional Brønsted and Lewis acid catalysts under the given reaction conditions. Amberlyst-15, a Reaction conditions: 100 mg of catalyst, 2 mL of 3 wt% glucose solution (substrate to catalyst ratio ¼ 0.6 wt%/wt%), 7 mL of MTHP (organic phase), 240 mg of NaCl, 140°C, 4 h; b) 10 mol% catalyst; c) 20 mol% catalyst; d) formic acid; e) levulinic acid. Scheme 2. Glucose dehydration to HMF together with a side reaction to two organic acids. sulfonated polystyrene resin, produced formic acid and levulinic acid (entry 11). This is due to the hydrolytic decomposition of HMF in water, which cannot be suppressed even in the presence of an extraction solvent. H 2 SO 4 and H 3 PO 4 showed poor activities with <25% in glucose conversion and <10% in HMF yield (entries 12 and 13). These results suggest that any Brønsted acids are ineffective for HMF formation. Sc(OTf ) 3 and Y(OTf ) 3 , which are well-known as homogeneous Lewis acid catalysts workable in water, 10 showed better activities than Brønsted acids in terms of glucose conversion and HMF yield/selectivity (entries 14 and 15). These differences between Brønsted acid and Lewis acid imply that the activity of 5P-YNbO 4 is most likely related to the watertolerant Lewis acid site of tetra-YNbO 4 . [24] Some reaction parameters such as the content of NaCl, the volume of organic solvent, and the weight ratio of substrate to catalyst (S/C) were investigated for their influence on the HMF yield of 5P-YNbO 4 . Glucose conversion at the same reaction time (4 h) increased as expected when the S/C decreased ( Figure S3, Supporting Information). Almost no changes in HMF selectivity were observed at S/C ¼ 1.2 and 0.6, while the selectivity slightly dropped at S/C ¼ 0.3. This could be simply due to the side reactions enhanced at the higher level of glucose conversion. Varying the volume of organic solvent had a major impact on the reaction ( Figure S4, Supporting Information). There was no significant difference in activity between 5 mL and 7 mL of MTHP. However, the decrease in MTHP below 5 mL resulted in lowering of both glucose conversion and HMF selectivity. Decreasing the activity can be attributed to the formation of humin-like compounds and their deposition on the catalyst surface, which leads to catalyst deactivation during the reaction. The effect of NaCl content was also investigated ( Figure S5, Supporting Information) to improve HMF yield and selectivity because NaCl enhances the extraction of HMF with the organic solvents, which prevents HMF degradation by side reactions. [30] Unexpectedly, there was no substantial difference in HMF yield and selectivity among NaCl ¼ 120, 240, and 480 mg, suggesting the addition of NaCl does not largely improve HMF extraction with MTHP as evidenced by the unchanged selectivity. Next, kinetic studies with bare and three tetra-YNbO 4 catalyst (X ¼ 1, 5, 10) were performed to obtain better insights into the effect of phosphorylation ( Figure 1A).
5P-and 10P-YNbO 4 showed nearly identical performance and were superior to 2P-and bare YNbO 4 ( Figure 1A). We attributed this improvement to the promotional effect of the phosphate group impregnated on the YNbO 4 surface. However, no difference between 5P-and 10P-YNbO 4 implies that the density of the phosphate group is maximized at 5P-YNbO 4 (2.4 wt%). Further kinetic results of 5P-YNbO 4 were shown in Figure 1B to understand the reaction pathway. The only products detected were fructose and HMF. Fructose yield reached the maximum (5%) at 30 min and then decreased gradually with time, while HMF yield steadily increased by 4 h and stayed almost constant for an additional 4 h. The calculated byproducts selectivity, based on the missing carbons, was kept almost constant at %50% during the reaction. This constant selectivity throughout the reaction can be ascribed to the formation of humin-type byproducts from glucose, rather than those from fructose and HMF, which accounts for the decrease in carbon balance when glucose conversion increases. The formation of insoluble humins was evidenced by the suspension of insoluble brown solid particles in the reaction mixtures. In addition, the re-hydration reaction of HMF was absent, as formic acid and levulinic acid were detected only in traces even at 8 h.
Two possible reaction routes have been reported for HMF formation from glucose. Isomerization-dehydration is the widely accepted mechanism in which glucose is first isomerized to fructose and then dehydrated to HMF. [31] Another mechanism is the stepwise dehydration mechanism where glucose undergoes direct dehydration to HMF via the formation of 3-deoxyglucosone (3DG) intermediate. [32] The fructose formation in Figure 1B suggests that HMF can be formed via fructose by isomerization and subsequent dehydration to HMF. To clarify whether or not the formation of HMF via fructose isomerization is predominant, we conducted tracer experiments with deuterium-labeled glucose (glucose-2-d) in reference to our previous works. [16,21] In the isomerization and dehydration mechanism, the hydrogen at C(2) will be shifted  to C(1) in the first isomerization step (Scheme 3A). When the reactant is glucose-2-d, the deuterium atom at C(2) will be always shifted to C(1), resulting in fructose-1-d. If dehydration between OH at C(1) and H or D at C(2) in fructose-1-d proceeds equally, the content of deuterium atom at C(1) of the resulting HMF should be %50%. In the stepwise dehydration mechanism (Scheme 3B), the first dehydration starts between OH at C (3) and H at C(2). The deuterium atom in glucose-2-d will be expelled in this step, leading to the formation of deuteriumfree HMF only. Figure 2 shows 2 H nuclear magnetic resonance (NMR) spectra of the reaction mixtures using Nb 2 O 5 , YNbO 4 , and 5P-YNbO 4 when glucose-2-d was used as the substrate. The activities of these three catalysts in the tracer experiments are almost the same as those in Table 1 using unlabeled glucose (Table S1, Supporting Information). Two resonance bands at 3.1 and 3.4 ppm are attributed to unreacted glucose-2-d, while a resonance band at 4.5 ppm is derived from HDO. These reaction mixtures contain HMF but there is no band at %9.6 ppm due to a deuterium atom bonded to aldehyde carbon, meaning that the deuterium atom present in glucose-2-d disappears during the reaction. These results are well consistent with those of anatase TiO 2 and its phosphate form in our previous works. [16,21] Consequently, we concluded that HMF formation with these Nb-based catalysts occurs via 3-deoxyglucosone (3DG) as the main intermediate through stepwise dehydration mechanism as shown in Scheme 3B. No difference in reaction mechanism among the three catalysts (Nb 2 O 5 , YNbO 4 , and 5P-YNbO 4 ) implies that phosphate groups do not participate in the stepwise dehydration of glucose with 5P-YNbO 4 and their role does not appear in the mechanistic diagram. In contrast, a control experiment confirmed that a typical Lewis acid catalyst, Sc(OTf ) 3 , forms HMF from glucose through the isomerizationdehydration mechanism ( Figure S6, Supporting Information).
Control experiments were conducted for the formation of HMF from fructose with 5P-YNbO 4 under the same reaction conditions as those from glucose. Complex side reactions to form humin-type byproducts are more facilitated in fructose dehydration ( Figure S7, Supporting Information), giving lower HMF yields/selectivities throughout the reaction than those in glucose dehydration ( Figure 1B). This difference is another evidence to support that HMF is formed via 3-deoxyglucosone through stepwise dehydration. Fructose appeared in Figure 1B is most likely formed by Lewis acid-catalyzed isomerization with A B Scheme 3. Presumable reaction pathways on HMF formation from isotopically labeled glucose (glucose-2-d) through: A) isomerization/dehydration mechanism and B) stepwise dehydration mechanism. Lewis acid is denoted as LA.  5P-YNbO 4 , but it is a minor reaction path against HMF formation.

Acid-Base Properties of YNbO 4 and 5P-YNbO 4
Next, we studied the acid-base properties of these catalysts to reveal catalytically active sites as well as the role of phosphate moiety in the improvement of HMF yield and selectivity. Pyridine adsorption experiments with Fourier-transform-infrared (FTIR) spectroscopy were performed to investigate the nature of acid sites on YNbO 4 , 5P-YNbO 4 , and Nb 2 O 5 . [33][34][35] Self-support disks of the catalysts were placed in an IR cell and heated at 150°C for 1 h under vacuum to remove physisorbed water. The dehydrated samples were exposed to saturated pyridine vapors at room temperature and subsequently evacuated for 1 h under vacuum to remove weakly physisorbed pyridine. Figure 3 shows the difference FTIR spectra of pyridine-adsorbed catalysts, where the spectra of bare catalysts before pyridine adsorption were used as background spectra. There are characteristic vibrational bands for coordinated pyridine with Lewis acid centers at around 1444 and 1606 cm À1 in three spectra. , respectively. The surface density of the Lewis acid site for 5P-YNbO 4 (0.62 μmol m À2 ) is approximately twice that of YNbO 4 (0.31 μmol m À2 ). This increase is most likely due to the reaction of H 3 PO 4 with the YNbO 4 surface during phosphoric acid treatment, producing phosphate moiety as well as new Lewis acid sites.
We conducted NH 3 -temperature-programmed desorption (TPD) experiments to probe the acid strength of YNbO 4 and 5P-YNbO 4 . Two profiles in Figure 4A have a poor S/N ratio as a consequence of their low surface areas and are suitable only for qualitative analysis. Owing to the absence of Brønsted acidity from the pyridine adsorption experiments, we can ascribe the ammonia desorption peaks almost exclusively to Lewis acid strength. YNbO 4 has an intense desorption peak at approximately 250°C and a weak and broad peak at %470°C, indicating the presence of two different Lewis acid sites. The area of the former peak accounts for 70% of the total area of the two peaks. The NH 3 -TPD profile of 5P-YNbO 4 is largely different from that of YNbO 4 . Indeed, the two distinct desorption peaks of bare YNbO 4 disappeared, while a broad desorption peak from 150 to 600°C is simply observed, suggesting the formation of strong Lewis acid sites in phosphoric acid treatment. The change in acid strength influenced HMF yield and selectivity ( We also conducted CO 2 -TPD measurements of YNbO 4 and 5P-YNbO 4 to probe their basic properties ( Figure 4B). Amorphous YNbO 4 possesses basic character as evidenced by its activity in base-catalyzed test reactions. [24] The CO 2 -TPD profile of YNbO 4 has an intense signal at around 240°C, indicating the presence of weakly basic sites. The absence of this desorption peak in the 5P-YNbO 4 profile indicates that YNbO 4 lost its basic character after phosphoric acid treatment. We hypothesized that the increase in HMF selectivity most probably comes from the decrease in the basic character. In fact, basic Y 2 O 3 exhibited high glucose conversion but no HMF formation ( Table 1, entry 9).
Lewis acid-base pairs on amphoteric TiO 2 were identified by FTIR measurements of CHCl 3 -adsorbed samples and were found to work concertedly for stepwise dehydration of glucose to HMF through theoretical calculations. [36] We examined Lewis acid-base pair sites of YNbO 4 and 5P-YNbO 4 with CHCl 3 as an acidic molecular probe. The self-support disks of YNbO 4 and 5P-YNbO 4 were placed in an IR cell and heated at 150°C for 1 h under vacuum to remove physisorbed water before the introduction of CHCl 3 . Figure 5 depicts the difference FTIR spectra of CHCl 3 -adsorbed samples, where FTIR spectra of the dehydrated samples before CHCl 3 adsorption were used as background spectra. CHCl 3 has a weakly acidic H atom and its interaction with the base site leads to the downshift of the CH stretching vibration band. A C─H stretching vibration band for adsorbed CHCl 3 is observed at 3006 and 3021 cm À1 for YNbO 4 and 5P-YNbO 4 , respectively ( Figure 5A), while the original vibration band of gaseous CHCl 3 appears at 3020 cm À1 . [24] No difference between adsorbed CHCl 3 on 5P-YNbO 4 and gaseous CHCl 3 suggests that the original base sites of YNbO 4 are fully decomposed or neutralized with phosphoric acid.
In addition to C─H stretching mode, the change of the H─C─Cl bending mode is an indicator to confirm the presence of the Lewis acid-base pair site. Original bending vibration bands of gaseous CHCl 3 and simply physisorbed CHCl 3 via hydrogen bonding of the acidic C─H bond are present at 1219 cm À1 ( Figure 5B). In contrast, a new band appears at %1235 cm À1  when CHCl 3 is adsorbed on a solid surface through an acidic H with a base site and a Cl atom coordinated with a Lewis acid site simultaneously (see the inset of Figure 5B). [37] Based on the molecular size of CHCl 3 , the presence of this "new bending vibration" suggests that the Lewis acid site and its conjugated base site can activate potentially and concertedly a reactant

Discussion
Phosphorylation of metal oxide catalysts is a widespread technique used frequently to increase HMF yield/selectivity in glucose dehydration but its effect is still under debate. We speculated that phosphate moiety deactivates the conjugated bases of Lewis acidic NbO 4 on Nb 2 O 5 ·nH 2 O or TiO 4 on anatase TiO 2 , resulting in both a decrease in byproducts formation and an increase in HMF yield and selectivity. [11,15] Atanda et al. reported that phosphate groups on TiO 2 simply introduce Brønsted acidity to the parent Lewis acid oxide, thus increasing the efficiency of the dehydration step during glucose dehydration. [18,19] Acikgoz et al. proposed that phosphate groups modify the electronic structure of the γ-Al 2 O 3 oxide surface by inducing dipoles and specifically increasing the electronic density of the Lewis acid centers. [38] The strong dipoles of parent YNbO 4 , due to the presence of two metals (Nb and Y) with different electronegativity, provide unique amphoteric nature as confirmed by FTIR and TPD measurements using acidic and basic probe molecules. We observed a drastic change in the original amphoteric property of the parent YNbO 4 after the introduction of phosphate groups. Two important consequences obtained from structural characterization studies are: 1) total loss of basicity confirmed by CO 2 -TPD and FTIR measurement with CHCl 3 and 2) an increase in the number of Lewis acid sites obtained from NH 3 -TPD and FTIR measurement with pyridine. These changes are closely related to the reactions of H 3 PO 4 with the basic components of the YNbO 4 surface, such as hydroxyls (Nb-OH and Y-OH) and lattice oxygens (Nb-O-Nb, Y-O-Y, and Nb-O-Y), because anionic species, which is an oxygen atom in the case of a metal oxide, is inevitably responsible for the basic property. No significant improvement in the activity of crystalline Nb 2 O 5 and Y 2 O 3 after phosphorylation as shown in entries 7-10 of Table 1 suggests that the reactions of H 3 PO 4 with two hydroxyls (Nb-OH and Y-OH) and two lattice oxygens (Nb-O-Nb and Y-O-Y) are of no particular importance for YNbO 4 . Based on the experimental results and this assumption, we attribute the change in the acid-base property of YNbO 4 after phosphorylation mainly to the reaction of H 3 PO 4 with Nb-O-Y on the YNbO 4 surface (Scheme 4).
We conducted 31 P MAS NMR measurements of 5P-Nb 2 O 5 , 5P-YNbO 4, and 5P-Y 2 O 3 to analyze the structure of the phosphate group immobilized on the surfaces ( Figure S8, Supporting Information). Phosphate groups on 5P-Nb 2 O 5 give rise to a broad symmetric band at À15 ppm, which is in accordance with the values reported in the literature for 31 P MAS NMR of phosphate niobia and other phosphated metal oxides. [39,40] 5P-Y 2 O 3 generates a signal at À2 ppm. The lower chemical shift indicates a more de-shielded P center, probably due to the higher basicity of the lattice oxygen compared to the oxygen atoms on Nb 2 O 5 . A single symmetric resonance band indicates that the chemical environment of the phosphate group on these reference samples is homogeneous. In the case of 5P-YNbO 4 , there are two different signals at À3 ppm and at À10 ppm. Based on the chemical shifts observed on the two reference oxides, we assign a broad band at À3 ppm to Y─O─P ¼ O(OH) 2 and a sharp band at À10 ppm to Nb─O─P ¼ O(OH) 2 , respectively, as schematically illustrated in Scheme 4. The formation of Y─O─P ¼ O(OH) 2 and/or Nb─O─P ¼ O(OH) 2 causes the increase in HMF yield/selectivity as well as the change in the number and strength of the Lewis acid site. The total loss of Lewis basicity by phosphorylation contributes mainly to the increase in HMF yield/selectivity. However, acid-base catalysis of Nb─O─P ¼ O(OH) 2 or Y─O─P ¼ O(OH) 2 toward HMF formation from glucose has not been yet clarified. The changes in the structure and the activity of amphoteric metal oxides including YNbO 4 before and after phosphorylation is further under investigation with theoretical calculations.

Recyclability
Catalyst reusability is one of the most important characteristics of heterogeneous catalysis. We first conducted several reactions of 5P-YNbO 4 without any post-treatment. The used catalyst was recovered by filtration, dried in an oven overnight at 110°C, and utilized for the consecutive reaction ( Figure 6A). In the second run, the activity dropped to %70% of the fresh catalyst, while HMF selectivity remained unchanged. This tendency was also observed in the third and the fourth reuses. The loss of the activity while retaining the HMF selectivity in the second run suggests the number of active sites decreases by the deposition of humin-type insoluble species on the catalyst surface. The formation of carbonous deposits was confirmed by thermogravimetrydifferential thermal analysis (TG-DTA) as shown in Figure S9, Supporting Information. The used catalysts after the first run had two exothermic peaks at 310 and 405°C, which is in line with previous works for the decomposition of carbon deposits formed in the dehydration of sugars. [41,42] Therefore, the catalyst recovered after the fourth run was subjected to thermal treatment at 700°C in air to remove carbonaceous deposition and regenerate the original surface structure of 5P-YNbO 4 . Despite a modest gain in glucose conversion, the HMF selectivity dramatically dropped from 49% to 18%. XRD analysis revealed the formation of the monoclinic phase in the regenerated catalyst while still retaining a part of the original tetragonal phase ( Figure S10a,b, Supporting Information). The formation of the monoclinic phase after calcination at 700°C was unexpected, as 5P-YNbO 4 underwent already the same calcination treatments twice in the synthesis of 5P-YNbO 4 . Control experiments using 5P-YNbO 4 and NaCl indicated that a modest amount of NaCl promotes phase transition even at 700°C and produces less active monoclinic YNbO 4 ( Figure S10c-e, Supporting Information). We modified the procedure of the recycling experiment by introducing a thorough washing treatment of the used catalyst with hot water between reaction cycles to attempt the removal of soluble organic and inorganic deposits. This procedure improved the recyclability of the 5P-YNbO 4 and reduced the cumulative loss of HMF yield even after the fourth cycle ( Figure 6B). The original activity of the 5P-YNbO 4 was fully restored when the used catalyst after the fourth run was calcined at 400°C for 5 h. Neither crystal growth nor phase transition proceeded by this calcination treatment ( Figure S11, Supporting Information), while a small and weak diffraction peak for NaCl is observed at 31.8°. The recovery can be explained by the results of TG-DTA measurement. Figure S12, Supporting Information, shows TG-DTA profiles of the spent catalysts with washing treatment using hot water after each run. Total weight Figure 6. Recycle experiments of 5P-YNbO 4 in glucose dehydration. A) the used catalyst was recovered by filtration, dried in an oven at 110°C overnight, and examined in the next run. The used catalyst after the fourth run was calcined at 700°C for 5 h in air and then used in the fifth run. B) The used catalyst was recovered by filtration, rinsed thoroughly with hot water, dried in an oven at 110°C for 16 h, and then examined in a consecutive run. The used catalyst after the fourth run was calcined at 400°C for 5 h in air and then used in the fifth run. C) The used catalyst was recovered by filtration, rinsed thoroughly with hot water, calcined at 400°C for 5 h in air, and then examined in the consecutive run.
www.advancedsciencenews.com www.small-structures.com loss, as well as two exothermic peaks at 325 and 405°C, gradually increase with the number of recycle experiments, suggesting that the building up of insoluble deposits on the catalyst surface was not avoidable by the washing treatment, as indicated by the gradual loss of activity after each reuse. The TG-DTA profile of the regenerated catalyst after the fourth reuse, which was obtained by the calcination of the used catalyst at 400°C for 5 h, was almost identical to that of the fresh 5P-YNbO 4 . Therefore, the calcination treatment is a simple and promising approach to regenerating the activity of 5P-YNbO 4 . Based on these results, we further designed additional recycling experiments in which the used catalyst recovered was rinsed thoroughly with water and calcined at 400°C for 5 h before the consecutive reaction. There was no significant decrease in HMF yield and selectivity even after three reuses (4th run), indicating that carbonaceous species formed on the catalyst surface is removed efficiently without structural change of 5P-YNbO 4 . These results suggest that Lewis acid sites as well as phosphate groups are stable during the catalytic reaction and regeneration treatment. We also examined the reaction liquor of the first run by inductively coupled plasma atomic emission spectroscopy (ICP-AES) and detect no phosphorus species (lowering the detection limit), suggesting that phosphate groups immobilized on the YNbO 4 surface remain intact after the reaction.

Conclusion
We found 5P-YNbO 4 as a capable catalyst for glucose isomerization to HMF in a biphasic system. Phosphorylation was a precious tool to enhance both activity and selectivity of the tetra-YNbO 4 . Acid-base characterization using TPD and FTIR revealed that phosphorylation neutralizes basic sites induces side reactions, and also produces new Lewis acid sites. We propose new Lewis acid sites are Nb─O─P ¼ O(OH) 2 and/or Y─O─P ¼ O(OH) 2 formed by the reaction of H 3 PO 4 with basic lattice oxygen atoms, which is most likely attributable to Nb─O─Y species. Tracer experiments using glucose-2-d demonstrated that glucose is converted to HMF through not the most common isomerization/dehydration mechanism but stepwise dehydration mechanism. 5P-YNbO 4 was confirmed to be reused without loss of original activity by thorough washing treatment of the used catalyst and subsequent calcination in air, while several low crystalline or amorphous metal oxides cannot be subjected to such thermal treatment for the removal of carbon deposits from the catalyst surface. This feature is solely attributed to a stable crystalline structure with effective Lewis acid sites for glucose dehydration. The strategy developed in this study enables precise control of Lewis acidity and the basicity of amphoteric mixed metal oxides, which leads to a highly profitable solid catalyst for practical and scalable HMF production.

Experimental Section
Materials Synthesis: YNbO 4 was synthesized by co-precipitation of peroxo-niobium complex (NH 4 ) 3 [Nb(O 2 ) 4 ] and Y(NO 3 ) 3 .6H 2 O. The peroxo-niobium complex was synthesized by following a previously reported procedure. [25] Briefly, (NH 4 ) 3 3 .6H 2 O aqueous solution after the addition of an ammonia solution (30%) and subsequent calcination of the precipitate at 700°C for 5 h in a muffle oven under static air. Phosphate YNbO 4 , Nb 2 O 5 , and Y 2 O 3 were obtained by conventional wetness impregnation using bare oxides and aqueous phosphoric acid solutions with different phosphoric acid contents. After impregnation, the resulting oxides were transferred in a muffle oven and calcined at 700°C for 5 h. Phosphate Nb 2 O 5 .nH 2 O was obtained by the same impregnation procedures but without calcination to avoid crystallization as previously reported.
Characterizations: XRD measurements were performed using Rigaku Ultima IV with Cu Kα radiation (40 kV, 40 mA). Nitrogen adsorption experiments were conducted on Micromeritics 3-Flex at 77 K. The samples were outgassed at 100°C for 1 h prior to the measurement. Brunauer-Emmet-Teller (BET) surface areas were estimated over the relative pressure (P/P 0 ) range of 0.05-0.30. FTIR spectroscopy measurements were conducted by Shimadzu IRspirit equipped with a Tryglycine sulfate detector. Self-supported disks (20 mm, %70-100 mg) were placed inside an IR cell attached to a closed circulation system. The disks were outgassed under vacuum at 200°C for 1 h before the measurements. Pyridine and chloroform were used as probe molecules for acid and base sites, respectively. Temperature-programmed desorption experiments were performed by using MicrotracBel Belcat II equipped with a thermo conductivity detector and a mass spectrometer. Catalysts (150-300 mg) were placed in the quartz cell and outgassed at 500°C for 1 h prior to the measurements. Ammonia and CO 2 were used as acidic and basic molecular probes, respectively. The final spectra were obtained by subtracting a blank experiment to cancel the baseline drift and exclude the background signal contribution. Thermogravimetry analysis (TGA) was conducted on a Rigaku Themo plus Evo2 instrument. Samples (%10 mg) were heated in a temperature interval from 50 to 1000°C at a ramping rate of 10°C min À1 under a constant air flow of 50 mL min À1 . High-resolution 31 P magic angle spinning (MAS) NMR spectra of the phosphorylated samples were obtained on JEOL ECA-600 (14.1T) equipped with an additional 1 kW power amplifier using a ZrO 2 rotor (4 mm in size) under ambient conditions. The relaxation delay was 15 ms. The 31 P MAS NMR shifts were referenced to ammonium dihydrogen phosphate at þ1.6 ppm. The samples were spun at 15 kHz.
Glucose Dehydration: Catalytic reactions were performed in a 50 mL pressure-resistant glass tube. Typically, the mixture of a 3 wt% glucose solution (2 mL), MTHP (7 mL), catalyst (100 mg), and NaCl (240 mg) was heated at 140°C for 4 h in an oil bath. After cooling the reactor in an ice bath, the reaction mixture was isolated by centrifugation. Aliquots of the aqueous phase were analyzed by a Shimadzu highperformance liquid chromatography (HPLC, Nexera X2) equipped with a refractive index detector and an ultraviolet (245 nm) detector. A Biorad Aminex HPX-87H column was used for products separation and quantification against calibration curves made by standard compounds. Aliquots of the organic phase were analyzed by a Shimadzu gas chromatograph mass spectrometer (GCMS, GC-2010 Plus) equipped with a DB-5MS column.
Isotopic Labeled Reactions: Reactions were performed by the same procedures described earlier using deuterium-labeled glucose. After the reaction, the aqueous phase was qualitatively analyzed by 2 H NMR (JEOL, 600 MHz). Aliquots of the organic phase were injected in a Shimadzu GC-2010 Plus gas chromatograph equipped with a DB-5MS column and the analytes were detected by means of a mass spectrometer detector.
Recycle Experiments: The mixture of a 3 wt% glucose solution (2 mL), MTHP (7 mL), catalyst (100 mg), and NaCl (240 mg) was stirred at 140°C for 4 h in an oil bath. In one set of experiments, the catalyst after www.advancedsciencenews.com www.small-structures.com Small Struct. 2023, 4, 2200224 the reaction was separated by centrifugation and dried at 110°C overnight, while the reaction liquors after appropriate dilution were analyzed with HPLC. At the end of the fourth run, the used catalyst was calcined at 700°C for 5 h prior to the fifth run. In another set of experiments, the used catalyst was thoroughly washed with hot water on filter paper and then dried at 110°C overnight. After the fourth run, the catalyst was washed with water and calcined at 400°C before the fifth run.

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