1 Introduction

The depletion of high-quality fossil fuels, shortage of food raw materials, and increasing emission of CO2 make it of urgent importance to develop new catalytic technologies for production of organic compounds from alternative raw materials [1–5]. Glucose is an industrially important valuable carbohydrate produced traditionally by enzymatic or acid hydrolysis of starch or sucrose. 5-Hydroximethylfurfural (5-HMF) known as a platform molecule is a promising agent for chemical industry, manufacturing plastics and fuel production [1, 5–13]. 5-HMF can be quite easily prepared from hexoses such as glucose and fructose [1, 8, 11, 12, 14, 15] which are the products of hydrolytic processing of di- and/or polysaccharides (sucrose, starch, inulin) derived from edible sources (crops and vegetables). A high selectivity (up to 89%) to 5-HMF can be archived by treating fructose in biphasic systems comprising dilute aqueous solution of mineral acids, inorganic salts and organic solvents (DMSO, MIBK, THF, alcohols, ketones etc.) [16,17,18,19]. Substitution of solid acid catalyst (silicoaluminophosphate SAPO-44) for mineral acids provides high yields of 5-HMF from fructose (78%), glucose (67%), maltose (57%), cellobiose (56%) and starch (68%) [20]. Different catalytic systems with ionic liquids as solvents were used to produce 5-HMF at high yields from fructose (75–92%) [19, 21, 22], glucose (up to 88%) [23] and sucrose (48%) [21]. An only commercial technology is currently on stream for the production of 5-HMF from fructose using a modified hydrothermal carbonization process [24]. AVA Biochem, Switzerland, has started the industrial manufacturing of 5-HMF since January 2014 [25]. The involved compounds are important for the food industry and, hence, there may occur undesirable competition for agricultural lands and resources. Cellulose is the main component of plant biomass and the most abundant natural non-food polysaccharide [2, 26]. On the other hand, organic solvents or ionic liquids are more expensive and ecologically unfavorable than water as the solvent. Therefore transformation of cellulose in pure hot water over a solid acid catalyst seems the most attractive alternative for one-pot catalytic production of glucose and 5-HMF without isolation of intermediates [8, 9, 27].

A number of catalysts such as soluble mineral acids and bases [28], heteropolyacids [29], carbons with different morphology [30], supported noble metals [31], polymer resins [32], alumina and silica [32,33,34,35], zeolites [34, 36] were suggested for cellulose depolymerization in aqueous solutions. However, there are some drawbacks of these catalysts. For example, soluble catalysts (mineral acids, bases, heteropolyacids) are corrosion-active and needs separation from soluble products. Noble metals are quite expensive. The catalysts based on carbons and silica may be unstable in aggressive hot water. It is important to emphasize that the authors of works cited above usually do not report explicit data on 5-HMF yield but pay only attention to the glucose formation. Hence, the development of highly active and stable solid acid catalyst for the one-pot hydrolysis-dehydration of cellulose to glucose and 5-HMF is still a great challenge [37].

Oxides of niobium(V), titanium(IV), zirconium(IV), tungsten(VI), aluminum(III) possessing acid or acid-base properties are stable in hot water and seem the most promising solid catalysts. These oxides were successfully applied for hydrolysis of some disaccharides (cellobiose, sucrose, maltose) [38,39,40] and dehydration of monosaccharides (fructose, glucose) to 5-HMF [41,42,43,44]. ZrO2 also was tested as a catalyst for transformation of saccharides into 5-HMF. Zirconia was active to isomerization of glucose to fructose [43, 45] and to glucose dehydration into 5-HMF [45, 46]. The highest yields of 5-HMF from fructose were 38% using microwave heating and only 12% using sand bath heating [43].

However, the application of the said metal oxides as catalysts for direct hydrolytic-dehydration of cellulose is only discussed in few papers (see, e.g., reviews [8, 12, 27, 47]). The possibility of direct processing biomass or cellulose into 5-HMF over modified zirconia catalysts such as ZrO2–TiO2 and silica-zirconia cogel SiO4–ZrO2 was shown elsewhere [46, 48]. Chareonlimkun and co-authors reported 8.6% yield of 5-HMF from different biomass species (sugarcane bagasse, rice husk and corncob) over ZrO2–TiO2 at 573 K [46]. Gliozzi et al. produced 5-HMF and furfural from softwood dust at the total yield of 8.2% over silica-zirconia at 423 K [48]. However, pure oxides ZrO2, SiO2 are low active to cellulose hydrolysis [32, 33]. Hydrolysis can be improved by anchoring sulfuric acid species onto ZrO2 [32, 49], TiO2 [50], SiO2 [51]. Niobium oxide is another promising catalyst for hydrolysis processes [52] including hydrolysis of soluble disaccharides [39, 53] and for dehydration of monosaccharides into 5-HMF [42, 44, 53]. Niobium oxide bears a great array of acidic centers on the surface and seems very promising additive to ZrO2. Niobium oxide supported on zirconia and zirconium-niobium mixed oxides catalyzed successfully dehydration of 2-prapanol [54], glycerol [55] and 1,2-propanediol [56] due to Brønsted acidity of polymeric niobium oxide species.

The aim of the present work was to develop a solid acid catalytic system containing niobium oxide supported on ZrO2 for one-pot hydrothermal processing of cellulose to produce glucose and especially 5-HMF in pure hot water as well as to explore a promoting effect of niobium on catalyst activity.

2 Experimental

2.1 Chemicals and Materials

Isopropoxide of zirconium Zr(OiPr)4 (70 vol% solution in iPrOH, Acrosorganics) and hydrate of niobium(V) oxalate С10Н5NbO20·xH2O (Alfa Aesar) were used for catalyst synthesis. Microcrystalline cellulose (fraction <0.10 mm, Vekton, Saint-Petersburg, Russia) was used as the substrate. 5-Hydoxymethylfurfural (Sigma-Aldrich), d-fructose (Sigma-Aldrich), d-mannose (Sigma-Aldrich), d-glucose (Reahim, Moscow, Russia), d-cellobiose (Alfa Aesar), levulinic acid (Acros Organics) were purchased as HPLC standards. Milli-Q water (Millipore, France) was used for preparing all the solutions.

2.2 Catalyst Preparation and Characterization

Zirconia was prepared by diluting 5 mL of Zr(OiPr)4 solution (70% solution in iPrOH, Acros organics) with 5 mL of iPrOH. The reaction solution was hydrolyzed by water steam and calcined for 8 h at 873 K (heating rate 2.5 K min−1).

For supporting niobium oxide, zirconia powder was mounted into a thermostated reactor and vacuumized at 323 K for 15 min. The calculated amount of niobium(V) oxalate was dissolved in a small volume of distilled water and placed into the thermostated reactor containing ZrO2. The catalyst was kept in the reactor for 30 min, dried in air for 24 h, and calcined at 723 K for 4 h (heating rate 3.6 K min−1).

Textures catalysts and the zirconia support were characterized using low-temperature nitrogen absorption at 77 K on ASAP-2400 device (Micromeritics, USA). All the catalysts were vacuumized to reach pressure of 6 Torr at 473–573 K before the analysis. The BET specific surface areas (SBET) were calculated from the adsorption curves of the isotherms. Pore volumes were determined by the αS-method, and pore size distribution by the BJH method.

The contents of Nb in the catalysts before and after the experiments as well as in the reaction media were measured by ICP-OES with Optima 4300 DV spectrophotometer (PerkinElmer Inc., USA).

XRD studies of the catalysts were conducted using BrukerD8 Advabced (Germany) diffractometer (step 2Ө = 0.05°, signal accumulation time 1–3 s per point). Monochromatic CuKα-radiation (λ = 1.5418 Å) was used for recording spectra. The phase composition was analyzed using ICSD data base (2009). Theoretical XRD patterns were calculated and phase compositions estimated using the PowderCell 2.3 program package.

UV–Vis diffuse reflectance spectra (UV–Vis DRS) of the catalysts without any pretreatment were acquired at ambient temperature using UV-2501 PC Shimadzu spectrometer with a IRS-250A diffusion reflection attachment at the 11,000–54,000 cm−1 range using BaSO4 as the reference. The UV–Vis edge energy (Eg) was determined from the intercept of the straight line of the low-energy rise at the plot of [F(R∞)hν]2 versus hν, where hν is the incident photon energy [57].

2.3 Cellulose Mechanical Activation and Characterization

A planetary-type mill Pulverizette 5 (Fristch, Germany) was used for mechanical activation of cellulose. Activation conditions: drum volume 250 mL, cellulose mill feeding 15 g, ball diameter 20 mm, ball feeding 7 pcs., activation time 40 min.

Cellulose particle size was determined by optical microscope Zeiss—Axiostar plus (Germany) equipped with a photocamera. The average diameter of cellulose particles was calculated as the mean diameters of 100 particles.

XRD studies of cellulose samples were carried out with BrukerD8 Advabced (Germany) diffractometer with monochromatic CuKα-radiation (λ = 1.5418 Å) at the 2θ range of 10°–40°. All the diffraction patterns were distributed into individual peaks (101, 10ī, 021, 002, 040 and amorphous cellulose) with 2θ equal to 15.2, 16.8, 20.6, 22.7, 34.1 and 21.5, respectively [58]. Crystallinity index (CI) was evaluated as the relation of the sum of 101, 10ī, 021, 002, 040 peak squares subtracting the square of amorphous cellulose halo (2θ ~ 21.5°) to the total square of diffraction pattern (2θ = 10°–40°) [59]. The coherent scattering region (CSR) was determined by the Seliakov–Sherrer formula [60].

2.4 Catalytic Tests and Analytic Techniques

Cellulose hydrolysis-dehydration was carried out in an autoclave (Autoclave Engineers, USA) at 453 K, argon pressure 1 MPa under intensive stirring at 1500 rpm. Cellulose and a catalyst (10 g L−1 each) and 45 mL of distilled water were put into the reactor. The reactor was purged with flowing argon, heated up to 453 K, and the reaction was started. In 0, 1, 2, 3 and 5 h of the reaction, aliquots of the reaction mixture (~1 mL) were drawn from the reactor to monitor the concentrations by HPLC; these aliquot quantities did not influence the process in the autoclave.

The samples were analyzed using Shimadzu Prominence LC-20 HPLC equipped with a RI detector. Saccharides and acids was analyzed using Rezex RPM-Monosaccharide Pb2+ or Rezex ROA-Organic Acids columns (Phenomenex, 300 × 5.0 mm) termostated at 343 and 313 K, respectively. Deionized water or 1.25 mM H2SO4 solution were used at the flow rate of 0.6 mL min−1 for analysis of sugar and acid, respectively.

The total organic carbon (TOC) content was determined after the reaction using Multi N/C 2100S TOC Analyzer (Analytik Jena, Germany). A 500 μL aliquot of the reaction mixture was added into an analyzer injector. Time of analysis was min. The amount of organic carbon (g L−1) was calculated based on the calibration curves plotted before.

pH of the reaction solutions was measured before and after the experiment with a pH tester Anion 4100 (TD Anion, Russia).

3 Results and Discussion

3.1 Cellulose Pretreatment

Cellulose is a biopolymer which consists of chains with glucose monomer residues linked by 1,4-glycosidic bonds. Side hydroxyl groups in the polymer chains form intramolecular (between residues of a chain) and intermolecular (between chains) hydrogen bonds. The hydrogen bonding leads to the ordering of cellulose chains and to the formation of crystal structure [61]. While crystalline cellulose is resistant to chemical transformations [62], its activation is an indispensable procedure.

In this study cellulose was activated by grinding in a planetary mill. This method seems quite promising because of amorphizating crystal domains and cleavage of glycosidic and hydrogen bonds [61, 63]. Optical spectroscopic studies of cellulose particles before and after the activation revealed that the average length decreased from 114 ± 35 (Fig. 1a) to 13 ± 6 μm (Fig. 1b). Comparison of optical and SEM images of particles related to initial and activated cellulose showed changes in the particle shape from oblong to almost spherical crystals (Fig. 1). XRD pattern with reflexes typical of the crystalline cellulose, CSR ranging from 4 to 7 nm and calculated CI equal to 80–90% were observed for the initial microcrystalline cellulose. Grinding cellulose during 40 min led to a considerable increase in the amorphous halo (2θ ~ 21.5°) on the XRD pattern with simultaneous decrease in CI (35–55%) and CSR (2–4 nm) (Table 1).

Fig. 1
figure 1

Microphotographs and particle size distribution (insets) (a, b), SEM images and XRD pattern (insets) (c, d) for the non-treated microcrystalline cellulose (a, c) and the activated 40 min in planetary-mill cellulose (b, d)

Full size image
Table 1 The parameters of cellulose samples before and after 2400 s (τ) ball-milling: average particle length <d l >, crystalline index (CI), coherent-scattering region (CSR)
Full size table

3.2 Catalysts Preparation and Characterization

Zirconia (ZrO2) was prepared by hydrolysis of Zr(OiPr)4 in water stream followed by calcination at 873 K. Niobium oxide (0.5, 1.1 and 2.8 wt.%) was supported on zirconia (see Sect. 2.2). The loading of Nb impregnated into the samples was determined by elemental ICP-OES analysis of the fresh catalysts. Analysis of the spent catalysts and solutions after the reaction indicated no leaching of Nb during the reaction in the hydrothermal reaction media.

Texture properties of the catalysts were revealed by low-temperature nitrogen adsorption (Table 2). Specific surface area (S BET ) of non-modified ZrO2 was 16 m2 g−1, and total pore volume (V PORE ) was extremely low. The type III isotherms have been identified which are typical of non-porous and/or macroporous materials free of micro/mesopores. While S BET for this kind materials often are no more than 10–20 m2 g−1 because of agglomeration during the calcination [64], low S BET of ZrO2 was expected. Addition of niobium oxide to the catalyst makes S BET decreased by ~25% to 11–12 m2 g−1 and the average pore diameter (<d PORE >) increased by 50% from 20 to 30 nm at the constant V PORE . S BET and <d PORE > do not depend on the Nb loading in the catalysts in the range of 0.5–2.8 wt.%.

Table 2 Texture parameters of NbOx/ZrO2 catalysts determined by nitrogen adsorption and ZrO2 phase composition revealed by XRD
Full size table

The nominal monolayer coverage was calculated using the textural data in the assumption that the cross-section of a Nb2O5 unit is 0.32 nm2 [54, 65].

From XRD data, the ZrO2 and NbOx/ZrO2 catalysts contain mainly monoclinic phase of zirconia (99% for pure ZrO2 and 97% for NbOx/ZrO2) (Table 2). The monoclinic structure is more thermally stable then the tetragonal one. A decrease in the proportion of the monoclinic phase during deposition of niobium oxide on the ZrO2 was reported elsewhere [56, 66]. We observed a similar trend. This observation allows us to conclude that the immobilization of Nb-containing particles stabilizes the tetragonal modification of zirconia. The obtained XRD data indicate the amorphous state of the active component but not the formation of crystalline niobium pentooxide (Fig. 2). The absence of crystalline phases of niobium pentooxide after its immobilization on the surface of other oxide supports was shown earlier [56].

Fig. 2
figure 2

XRD diffraction patterns of NbOx/ZrO2 catalysts

Full size image

The coordination geometry and the chemical arrangement of the Nb species in the catalysts were studied by UV–Vis DRS spectroscopy (Fig. 3). It was shown [67] that the traditional method used to study Brønsted acid sites by NH3 or pyridine adsorption followed by IR spectroscopy was low sensitive to the catalysts with the loading of niobium <5 wt.%. Therefore, we chose the UV–vis DRS spectroscopic technique the relationship between the absorption edge energy and the degree of condensation of niobium oxide species shown before [54, 68]. The zirconia absorption edge energy (Eg = 5.13 eV) is consistent with the reported data [54] due to the charge transfer O2− → Zr4+. The absorption edge energy for Nb-containig species in NbOx/ZrO2 for 0.5%Nb/ZrO2 (Eg = 4.45 eV) is close to that reported for isolated NbO4 species in Nb-MCM-41 [68] and to that of niobium oxalate [54] that indicates the predominant formation of isolated NbO4. The UV–vis absorption edge energy for 1.1%Nb/ZrO2 and 2.8%Nb/ZrO2 decreases to 4.20 and 4.03 eV (Fig. 3, inset) toward the value typical of polymer Nb structures (for example, niobic acid, Nb2O5·nH2O, Eg = 3.9 eV [54]). However, these values remain higher than that of niobium oxide (Eg = 3.42–3.2 eV). The red shift of edge energy at an increase of the Nb loading indicates a higher degree of polymerization: the Nb cations in the 0.5%Nb/ZrO2 sample are predominantly isolated NbO4 species, while polymerized surface niobium oxide NbOx (NbO6) species are formed in the 1.1–2.8%Nb/ZrO2 catalysts [54, 68]. These results are consistent with the nominal monolayer coverage assumed from the texture data (40% of monolayer for 0.5%Nb/ZrO2; 100 and 240% for 1.1 and 2.8%Nb/ZrO2, respectively). The acidity of our catalysts was confirmed indirectly by pH measuring.

Fig. 3
figure 3

Diffuse reflectance UV–Vis spectra of ZrO2 (1) and NbO x /ZrO2 catalysts: 0.5 wt.% Nb (2), 1.1 wt.% Nb (3), 2.8 wt.% Nb (4); inset shows the Kubelka–Munk function versus energy of the excitation source

Full size image

3.3 Cellulose One-Pot Hydrothermal Hydrolysis-Dehydration

The results of the catalytic hydrothermal hydrolysis-dehydration of mechanically-activated cellulose in the autoclave at 453 K and inert argon atmosphere are shown in Table 3 and Fig. 4. The main reaction products were glucose and 5-HMF. The formation of low amounts of mannose and fructose (via glucose isomerization), formic and levulinic acids (caused by 5-HMF decomposition), and furfural (the side product of fructose transformation) were observed in this study and reported in literature [69, 70]. Oligosaccharides were formed during the 1st hour of the reaction, then the yields of soluble oligomers went down to zero. The formation of humins (brown colored side products of monosaccharide and 5-HMF condensation) also was observed under the reaction conditions. While both the formation of humans and cellulose transformations over solid catalysts need high temperature, and former are unavoidably produced. It has been shown before [71] that amorphous cellulose hydrolysis starts at 423 K and crystalline cellulose at 453 K. With our catalytic system comprising the solid substrate and the solid catalyst, the critical step is depolimerization of cellulose to soluble oligosaccharides to be transformed to glucose and 5-HMF over a catalyst. It is well known that dissociation constant of water (K w ) increases with temperature elevation. At 473 K −lgKw becomes equal to 11 that leads to an increase in the concentrations of Н+ and ОН− ions [72]. The increase in −lgKw allows water to behave as the co-catalyst during cellulose hydrolysis-dehydration [73].

Table 3 Cellulose hydrolysis-dehydration in the presence of NbOx/ZrO2
Full size table
Fig. 4
figure 4

The glucose and 5-HMF yields derived from cellulose versus reaction time in the presence of a catalyst (a ZrO2; b 0.5%Nb/ZrO2; c 1.1%Nb/ZrO2; d 2.8%Nb/ZrO2). Conditions: see Table 3

Full size image

The yields of the reaction products were calculated according to formula [74]:

$$Y=\frac{{C \cdot V \cdot {N_C}}}{{{n_{Cell}}}} \cdot 100,$$

where Y is the product yield (%); C is the product concentration (mol L−1); V is the reaction volume (L); M C is molar mass of carbon (g mol−1); N C is the number of carbon atoms in one product molecule (six in glucose, fructose and mannose, five in 5-HMF, furfural etc.); \({n_{Cell}}\) is moles of carbon in cellulose sample (mol).

Turnover frequency, TOF (s−1), i.e. the total amount of glucose and 5-HMF (mol) produced by Nb atoms and ZrO2 of the catalysts (mol) per second was calculated according to formula:

$$TOF=\frac{{R \cdot V}}{{{m_{cat}} \cdot \left( {\frac{{{\omega _{Nb}}}}{{{M_{Nb}}}}+\frac{{{\omega _{Zr{O_2}}}}}{{{M_{Zr{O_2}}}}}} \right)}},$$

where R is the initial rate of glucose and 5-HMF formation (mol L−1 s−1); V is the volume of the reaction mixture (L); m cat is the catalyst mass (g); M Nb is the molar mass of niobium (g mol−1); \({M_{Zr{O_2}}}\) is the molar mass of zirconia (g mol−1); \({\omega _{Nb}}\) and \({\omega _{Zr{O_2}}}\) are mass fractions of niobium and zirconia, respectively.

The initial reaction rate (R) was determined from the kinetic data as follows:

$$R=\frac{{({C_{Glu}}+{C_{5-HMF}}) - ({C^0}_{Glu}+{C^0}_{5-HMF})}}{\tau },$$

where R is the rate of glucose and 5-HMF formation (mol L−1 s−1); C Glu and C 5−HMF are concentrations of glucose and 5-HMF (mol L−1), respectively, achieved at the second kinetic point (1 h); \(C_{Glu}^0\) and \(C_{5-HMF}^0\) are the concentrations of glucose and 5-HMF (mol L−1), respectively, at the first kinetic point (0 h); τ is the reaction time measured in seconds (3600 s).

Maximum yields of all the side products usually were no more than 1–1.5 mol% except the yield of levulinic acid that reached 1.8 and 2.5 mol% in 5 h of the reaction over 1.1%Nb/ZrO2 and 2.8%Nb/ZrO2, respectively (Table 3).

All the catalysts were highly active to the formation of 5-HMF, while it was not formed in the blank experiment at all. In the blank experiment, the maximal glucose yield was no more than 7.0% in 5 h of the reaction. In the presence of all the solid catalyst under study the cellulose conversion was around 50% and glucose yields were 12–22%. Hence, our catalysts interact indeed with cellulose and catalyze its depolymerization probably by the acid mediated mechanism. Inspection of pH of the reaction solution also showed the increasing activity. pH varied in the range of 3.6–4.5 in the presence of the catalysts. An increase in the Nb amount leads to a rise of pH compared to pure ZrO2.

Initial reaction rates (R) calculated from combined kinetics of glucose and 5-HMF formation are 5–20 times higher in the presence of the catalysts compared to the blank experiment. The activity of pure zirconia (without Nb) was high enough. Maximum yields of glucose and 5-HMF over ZrO2 were 12.7 and 13.3 mol%, respectively. The observed activity of this catalyst to cellulose treatment was higher than that described in literature. Wang and co-authors reported 9 mol% glucose yield and 15.8 mol% total yield from cellulose over pure ZrO2 after 12 h at 433 K [33]. The activity of pure ZrO2 prepared by us also is comparable to the activity of mixed ZrO2–TiO2 (8.6 mol% 5-HMF yield at 573 K [46]) and sulfated ZrO2 (14 mol% of glucose at 423 K and 24 h [32]).

However, almost two-fold decrease in the catalytic activity (initial reaction rate R and TOF) was observed upon addition of a small amount of NbOx into the catalyst. The close yield of glucose (12.9 mol%) was only reached at 5 h of the reaction (Fig. 4a, b). The activity of 0.5%Nb/ZrO2 decreased even though the acidity of the reaction solution increased compared to that of ZrO2. The further increase in the Nb loading upto 1.1 and 2.8 wt.% led to ca. two times increase in the activity against that observed with ZrO2. Close activities of 1.1%Nb/ZrO2 and 2.8%Nb/ZrO2 and close yields of glucose and 5-HMF over these catalysts (21–22 and 16 mol%, respectively, see Table 3; Fig. 4c, d) were revealed. However the yield of levulinic acid formed via 5-HMF decomposition was somewhat higher over 2.8%Nb/ZrO2 than over 1.1%Nb/ZrO2 indicating that the catalytic activity of 2.8%Nb/ZrO2 was a little bit higher. These similar results can be accounted for by close pH of the reaction media in spite of more than twofold difference in the Nb loading.

To our knowledge, the NbOx/ZrO2 catalytic system was not been used before for direct cellulose transformation to glucose and 5-HMF in pure water. The application of niobium for cellulose depolymerization was studied by Takagaki et al. [75]; they used the HNbMoO6 catalyst and observed the 1 mol% of glucose and a total 8.5 mol% yield of the main products (cellobiose and glucose). The authors assign the negligible yields of monomers and the absence of 5-HMF in the reaction products to the reaction temperature (373 K).

The activity changes can be interpreted in the following manner. Zirconia is known to be an acid-base bifunctional catalyst, and the cooperation of weak Lewis acid and weak base sites (acid-base pairs are suitably oriented on ZrO2) plays an important role in the achievement of high selectivity to some reactions [76]. The concentration of Lewis acid sites on the zirconia surface was shown not to decrease when niobium oxide was supported on the surface [54]. We referred to the literature data in assuming that, in spite of slightly increasing total acidity (according to pH) of the reaction medium, the activity of 0.5%Nb/ZrO2 decreases against that of ZrO2 due to the formation of dispersed isolated NbO4 species (40% of monolayer) to isolate spatially the acid and base centers (according to UV–vis DRS). This also allowed us to suppose that the catalyst surface plays an important role in the hydrolysis of cellulose to glucose. Unfortunately, as mentioned in Sect. 3.2, the direct techniques for measuring acidity (NH3 or pyridine adsorption) are unsuitable for our catalysts because of the low Nb content and low surface area of the catalysts [67]. The formation of additional Brønsted sites was observed upon supporting niobium at the concentrations higher then 0.69, 3.4 and 5 wt.% on zirconia [54], titania [77], and alumina [78], respectively. Analysis of the Nb phase structure showed that the Brønsted acidity was caused by polymer NbOx species, while the isolated species had no acidity [54]. In the catalysts under study, the formation of polymerized NbOx (NbO6) species was observed (see UV–Vis DRS data above) in the samples containing 1.1 and 2.8 wt.% Nb at a more than monolayer Nb coverage. Obviously, additional acid sites are responsible for the increasing activity to cellulose depolymerization.

After completing the reaction, the TOC concentration in the reaction solutions was analyzed (Table 3). TOC values appeared quite close to one another and increased as follows: 2.01 < 2.16 < 2.26 ~ 2.29 g L−1. This series is similar to the series of the catalyst activity 0.5%Nb/ZrO2 < ZrO2 < < 1.1%Nb/ZrO2 ~ 2.8%Nb/ZrO2. The cellulose conversion (X) was calculated from the TOC data by the following formula (Table 3):

$${X_{Cell}}=\frac{{{C_C} \cdot V}}{{{m_{Cell}}}} \cdot 100,$$

where X Cell is cellulose conversion (%); C C is the concentration of carbon atoms (g L−1); V is the reaction volume (L); \({m_{Cell}}\) is the amount of carbon in cellulose sample (g).

The cellulose conversion was 14.1% in the blank experiment but close to 50% in all catalytic experiments. The highest conversion equal to 52.3% was observed over 1.1%Nb/ZrO2 and 2.8%Nb/ZrO2.

3.4 Hydrothermal Transformation of Monosaccharides and 5-HMF

The most active catalyst 2.8%Nb/ZrO2 was tested in dehydration of main intermediates of cellulose transformation (glucose and fructose) and in degradation of 5-HMF under conditions identical to the cellulose treatment (Table 4; Fig. 5).

Table 4 Dehydration of the main cellulose transformation intermediates in the presence of 2.8%Nb/ZrO2
Full size table
Fig. 5
figure 5

The yields of glucose, mannose, fructose, and 5-HMF derived from glucose (a), fructose (b), and 5-HMF (c) versus time over 2.8%Nb/ZrO2. Conditions: see Table 4

Full size image

The initial reaction rate (R) was calculated as the rate of the substrate consumption:

$$R= - (C - {C^0}) \cdot {\tau ^{ - 1}},$$

where R is the rate of glucose and 5-HMF formation (mol L−1 s−1); C and \(C_{}^0\) are actual and initial concentrations of the substrate, respectively (mol L−1); τ is the reaction time in seconds (3600 s).

The highest yields of 5-HMF obtained from glucose and fructose equal to 22.6 and 31.8 mol%, respectively were achieved in 3 and 0.5 h of the reaction. The fructose dehydration is a fast reaction (X of fructose was 100% in 1 h). Glucose transformation is much slower because it limited by isomerization of glucose to fructose. We did not achieved a very high yield of 5-HMF from monosaccharides due to its degradation (X = 56.7 mol% in 5 h) at 453 K. Lowering of the reaction temperature is expected to result in an increase in the yields of 5-HMF from the monosaccharides. The yields of levulinic and formic acids obtained from 5-HMF and furfural were very low in all the experiments.

4 Conclusions

A series of new Nb-containing zirconia catalysts were prepared, characterized (XRD, N2 adsorption, ICP-OES, UV–Vis DRS) and tested for the first time in pure hot water in one–pot hydrolysis-dehydration of activated microcrystal cellulose to produce glucose and 5-hydroxymethylfurfural. ICP-OES analysis of the spent catalysts and reaction solutions showed the excellent stability of the catalysts under hydrothermal conditions. The activity of the catalysts increased as follows: 0.5%Nb/ZrO2 < ZrO2 < < 1.1%Nb/ZrO2 ~ 2.8%Nb/ZrO2. The maximum yields of glucose and 5-HMF were 22 and 16 mol%, respectively. All the catalysts were non-macroporous bulk zirconia containing mainly monoclinic phase of ZrO2 (97–99%). The formation of Nb2O5 crystal phase was not observed. The state of NbOx species was studied by UV–Vis DRS spectroscopy. 0.5%Nb/ZrO2 contained niobium as inactive isolated NbO4 species, while polymerized niobium oxide NbOx (NbO6) resided on the surface of 1.1%Nb/ZrO2 and 2.8%Nb/ZrO2. Isolated NbO4 structures separated acid and base sites of ZrO2 to decrease its own catalyst activity. Polymeric niobium oxide bore Brønsted acid sites, thus increasing essentially the acid properties and the catalytic activity.