Direct conversion of cellulose to α-hydroxy acids (AHAs) over Nb₂O₅-SiO₂-coated magnetic nanoparticles

1 Introduction

Currently, both the academe and the industry are focused on finding new sustainable technologies that can lower the environmental pollution caused by the extended utilization of the fossil resources. In this context, the lignocellulosic biomass, which originates mainly from forest and agricultural residues, is considered as one of the most promising renewable carbon resources, along with glucose, which is a potential precursor to useful compounds, such as the bio-based platform chemicals [1], [2]. Cellulose is the most abundant biopolymer from the lignocellulosic materials (40%–50% [3]) and can be converted to glucose, which produces bioethanol or platform chemicals by fermentation and dehydration, respectively [4].

Lactic acid is one of the most important platform molecules obtained from biomass, and is extensively used in different industrial sectors, such as food, cosmetics, pharmaceuticals, and bulk chemicals [5]. Moreover, its polymerization to biodegradable poly(lactic) acid is one of the most promising and rapidly growing application [6]. However, its current relatively high price precludes many of the applications listed above, because it is produced almost exclusively through the fermentation of sugars, a process that generates a significant amount of wastes [7]. Therefore, the identification of more efficient and less costly production methods compared with the current fermentation procedure is one of the most important tasks of current researchers in the field of biomass transformation.

Meanwhile, glycolic acid is an extremely versatile liquid that is safer to handle, easier to use, and offers many additional benefits, including low corrosion to most common metals, negligible odor, low toxicity, and non-flammability. Glycolic acid is also compatible with many additives, is soluble in water, easy to rinse, environmentally friendly, and has an 89% rate of biodegradation in 7 days. Owing to these important properties, glycolic acid is used especially for household cleaning products and medical sutures.

Nowadays, industrial glycolic acid production is carried out through the catalytic carbonylation of formaldehyde [8]. However, in light of the depletion of fossil resources, novel catalytic paths for glycolic acid production from renewable raw materials should be developed. In this context, the well-known advantages of the heterogeneous catalysis have encouraged the development of some acid or base solid materials that are able to catalyze the one-pot conversion of cellulose to lactic acid. Chambon et al. [9] showed, for instance, that solid Lewis acid catalysts, such as tungstated zirconia and tungstated alumina, exhibited a remarkable promoting effect on the cellulose depolymerisation, leading to a 27% lactic acid yield. Meanwhile, Liu et al. [10] used a solid-base MgO catalyst to achieve similar yields in methanol. For the case of the synthesis of the glycolic acid, the best selectivities (52%–70%) were reported in the glycerol transformation on gold-based catalysts [11].

Tolerance to water is another noteworthy aspect of evaluating the performances of the proposed heterogeneous catalysts. In this context, niobium compounds exhibiting both high stability and high acidity (H₀=−8.2) in an aqueous medium have attracted great research attention in the past few years [12]. Niobium compounds have been included in investigations of different transformations of the biomass derivatives [13], [14], [15], [16]. However, commercial bulk niobia suffers from poor hydrothermal stability under high-water environment and high temperatures [17]. Interestingly, its deposition over oxide supports (e.g. silica, titania, or alumina) enhances both hydrothermal stability and acidity in aqueous phase reactions; thus, it has potential catalytic applications [18].

In the specific context of the one-pot conversion of cellulose to platform molecules, we recently reported the synthesis and development of highly efficient Nb-containing hydroxylated AlF₃ catalysts, which can facilitate the depolymerisation of cellulose to α-hydroxy acids (AHAs) as the main products, where the lactic acid is preponderant [16]. The catalytic results confirmed that the active sites responsible for the lactic acid formation are the Nb(V)/Nb(IV) species from the amorphous/crystalline phases, diluted in the aluminum oxide fluoride matrix.

In addition, the consideration of materials containing magnetically active nanoparticles cores, such as magnetite (Fe₃O₄), may generate an efficient separation from different mixtures by simply applying of an external magnetic field. Few examples of such economically and efficient catalytic systems were recently developed in our group for the direct production of sorbitol and glycerol from cellulose (Ru(III)-SiO₂-Fe₃O₄) [19] and for the lignin fragmentation (magnetically recyclable composite Co@Nb₂O₅@Fe₃O₄ catalysts) [20]. Very important, in the presence of oxygen, the ionic Ru(III)-SiO₂-Fe₃O₄ became an efficient catalyst also for the synthesis of succinic acid from levulinic acid [21] and glucose [22].

Based on these considerations and with the aim of improving the former Nb-based inorganic fluorides catalysts for the cellulose transformation into α-hydroxy acids (AHAs), namely lactic and glycolic acids, we focused on the design of new niobium-based catalysts. Thus, we report herein the preparation of a series of core-shell catalysts based on SiO₂-Nb₂O₅ (30 wt% and 60 wt% Nb₂O₅, respectively) composites (shell) and magnetic nanoparticles (core) prepared via co-precipitation (CP) and sol-gel, followed by precipitation (SGP) methods. The new designed catalysts were tested in the one-pot conversion of cellulose affording α-hydroxy acids (AHAs). Correlating the catalytic properties with the catalytic performances allowed us to determine the most important catalytic features that are necessary for a highest productivity to lactic acid or glycolic acid.

2 Materials and methods

All synthesis reagents are analytically pure and used as received from Sigma-Aldrich (Schnelldorf, Germany). α-Cellulose (product no. C8002) contained 92.2% glucan and 7.8% xylan, on dry basis, and a broad distribution of the polymerization degree (PD) with the maximum value of 1.5×10³ [23], [24]. All used materials (i.e. FeCl₂*4H₂O, [Fe(NO₃)₃*9H₂O], NH₄OH (25%), TEOS, HCl, and ANBO) were purchased from Sigma-Aldrich, Schnelldorf, Germany and used as received.

3 Results and discussion

XRD patterns corresponding to the 30Nb-Si@MNP-SGP samples (Figure 1) displayed only the diffraction lines characteristic to magnetite (2θ=30.1°, 35.4°, 43.1°, 53.4°, 57.1°, and 62.6°, indexed to the cubic spinel phase of magnetite [(220), (311), (400), (422), (511), and (440), JCPDS 19-629] and a broad and low intensity diffraction line characteristic to the amorphous silica, which coated the surface of the MNP. No defined XRD reflexes from niobium oxides were observed, showing a high dispersion onto silica with no phase separation.

Figure 1:

X-ray diffraction patterns of the 30Nb-Si@MNP-SGP samples: (A) non-calcined and (B) calcined at 500°C.

However, the increase of the niobium content from 30 wt% to 60 wt%, led to a Nb aggregation resulting in niobium oxide phases that did not arrange in a well-ordered way (Figure 2). The XRD pattern display showed two weak broad lines centered at 2θ=22.4° and 50.0°, assigned to the amorphous silica (2θ=22.4°) and a niobium oxide phase (2θ=50.0°) (Figure 2).

Figure 2:

X-ray diffraction patterns of 60Nb-Si@MNP-SGP samples: (A) non-calcined and (B) calcined at 500°C.

Meanwhile, the X-ray diffraction patterns of the Nb-Si@MNP-CP samples indicated a completely amorphous structure with a high dispersion of the niobium oxide phase in the silica matrix, irrespective of the niobium oxide content (30 wt% or 60 wt%) (Figure 3). The only visible diffraction lines from the XRD patterns belonged to the magnetite core.

Figure 3:

X-ray diffraction patterns of (A) 30Nb-Si@MNP-CP and (B) 60Nb-Si@MNP-CP samples.

The magnetite nanoparticle sizes for both calcined and un-calcined samples, which were determined using the Scherrer formula [26], were around 8.8–9.0 nm.

DRIFT spectra were collected with the scope of the identification of the functional groups (Figures 4–7). As expected, the spectra of non-calcined samples (Figures 4 and 6) presented typical vibrations of the surface -OH groups and adsorbed water at 1660 and 3000–3400 cm⁻¹ irrespective of the niobium content or the preparation methodology. Further, the asymmetric stretching vibration of the Si-O-Si bond located at 1280 cm⁻¹ was present in all samples, indicating the silica matrix formation. After the ANBO complex deposition, a new band, located at 1.440 cm⁻¹, was visible in the corresponding IR spectra of the non-calcined samples (Figures 4 and 6), irrespective of the preparation method. Moreover, the higher the niobium content, the higher the absorption band intensity. Most probably, this band was associated with the ν (C=O) vibrations of the oxalate ligand from ANBO complex. After the sample calcination at 500°C, this band totally disappeared, indicating a total decomposition of the complex (Figures 5 and 7).

Figure 4:

IR spectra of the (A) SiO₂@MNP; (B) non-calcined 30Nb-Si@MNP-SGP, (C) non-calcined 60Nb-Si@MNP-SGP samples.

Figure 5:

IR spectra of the (A) calcined 30Nb-Si@MNP-SGP and (B) calcined 60Nb-Si@MNP-SGP samples.

Figure 6:

IR spectra of the (A) non-calcined 30Nb-Si@MNP-CP, (B) non-calcined 60Nb-Si@MNP-CP samples.

Figure 7:

IR spectra of the (A) calcined 30Nb-Si@MNP-CP and (B) calcined 60Nb-Si@MNP- CP samples.

The -OH vibrational frequencies of the calcined samples provided very important information on the nature of the niobia species. DRIFT spectra of SGP samples presented only bands at 3657 cm⁻¹ that were characteristic of the SiO-H groups (Figure 5). For the CP sample bands, the additional band at 3743 cm⁻¹ was characteristic of the NbO-H [Nb (V)] groups (Figure 7). However, irrespective of the preparation route, no absorption bands at 3555–3602 cm⁻¹ were observed, confirming the rarity or non-presence of Nb (IV) in these materials. This finding is in accordance with a previous report of Tielens et al. [27].

The most critical information provided by the IR spectra is that from the 500–1.250 cm⁻¹ domain on the framework bonds. For the SGP samples, the decrease in the intensity of the band at 1.280 cm⁻¹ (corresponding to the asymmetric stretching vibration of Si-O-Si), the shift of the bands was located at 800–820 cm⁻¹ (attributed to the symmetric stretching vibrations of Si-O-Si) towards the lower values, and the appearance of new bands at 884 cm⁻¹ and 600–500 cm⁻¹ reflected the structural effects induced by the presence of niobia. Shifted bands and the novel bands from 600–500 cm⁻¹ were assigned to the Nb-O-Nb stretching mode vs. ([-O-Nb-O-]n). The decrease of the intensity of Si-O-Si band (from 1.280 cm⁻¹) indicated the interaction of Nb with these bonds, whereas the novel absorption band located at 884 cm⁻¹ can be attributed to the -Nb=O bonds [27]. For the CP samples, the band from 1.280 cm⁻¹ shifted at 1.209 and 1.147 cm⁻¹, respectively, depending on the niobium content, concomitant with the apparition of new bands at 750–720 cm⁻¹. In this case, new Si-O-Nb bonds were formed, with the final catalysts being characterized by a higher dispersion of the small niobia particles in the silica matrix.

In conclusion, (i) all the investigated samples contained Nb(V) sites but not Nb(IV) sites; (ii) samples prepared via the SGP route displayed large but not well-ordered niobia phases (XRD results) containing niobyl -Nb=O and NbO-H species (DRIFT results); (iii) samples prepared via the CP route were characterized by the formation of a small well dispersed niobia phase (XRD results), which contained high amounts of NbO-H species (DRIFT results).

The acid-base properties of the Nb-Si@MNP samples were investigated by CO₂- and NH₃-TPD measurements. The TPD profiles are given in Figures 8 and 9, respectively. The CO₂-TPD profiles (Figure 8) revealed a high density of basic sites for Nb-Si@MNP-CP samples, whereas Nb-Si@MNP-SGP samples were characterized by a lower content in such sites. A quantification of the basic site density and their strength is given in Table 1.

Figure 8:

CO₂-TPD profiles for 30Nb-Si@MNP-SGP (A), 60Nb- Si@MNP-SGP (B), 30Nb-Si@MNP-CP (C) and 60Nb-Si@MNP-CP (d) samples.

Figure 9:

NH3-TPD profiles of 30Nb-Si@MNP-SGP (A), 60Nb- Si@MNP-SGP (B), 30Nb-Si@MNP-CP (C), and 60Nb-Si@MNP-CP (D) samples.

As shown in Table 1, the density of basic sites increased with the niobium content irrespective of the preparation route. The highest density was determined for the CP samples, irrespective of the niobium content. Meanwhile, the NH₃-TPD profiles are displayed in Figure 9, whereas the density of the acid sites and their strengths are given in Table 2.

CO₂-TPD results indicated a correlation between the content of niobium and the population of the basic sites but no correlation with the acid site density. This is not at all surprising when we consider that only NbO-H species are basic in nature. The acidity in this case can be generated by different species, such as the Nb=O, NbO-H_(acid) or SiO-H groups. Nevertheless, the acid/base ratios may offer valuable qualitative information on the acid-base characteristics of the synthesized catalysts (Table 3).

As shown in Table 3, the acid/base ratio is highly influenced by the preparation method. For the CP samples, basic sites predominated and the content of the niobium did not influence the acid/base ratio, which also represented a confirmation of the high dispersion of the niobium phase in the silica matrix. Contrarily, for the SGP samples, the higher niobium content led to lower acid/base ratios. In this case, an increased density of the basic sites for increased niobium content corresponded to the formation of large agglomerations of niobium phases on the outer silica shell. The existence of Nb-OH sites with a basic character rather than an acidic one has also been suggested by several authors [27], [28].

The catalyst characteristics influence their performances in terms of conversion, yields towards the reaction products, and carbon efficiency (E_c) (Table 4). E_c was calculated from the ratio of the weight of hydroxyacids in the product mixture to the weight of cellulose transformed during the reaction; it expressed the overall selectivity to the desired liquid phase products. Herein, one considered the products dissolved in the water phase as useful, whereas water-insoluble oligomers/humans, i.e. those that cannot be analyzed by GC-FID/GC-MS, as non-useful.

Different research groups, including ours, already demonstrated that the hydrolysis of cellulose to oligomers is promoted under hot compressed water (i.e. where H₃O⁺ species exist), without the addition of any catalyst [29]. However, under these conditions, the conversion of cellulose is low (<12%). The very high conversions (78%–95%, Table 4) in the presence of the Nb-Si@MNP catalysts can be attributed to the high density of active sites displayed by the external surface of these materials, which can interact with the oligomers produced in the initial hydrolysis stage of the cellulose. Moreover, the CP catalysts were slightly more active than the SGP ones, and this can be attributed to the higher dispersion of the niobium phase on the silica shell.

Also noteworthy are the high E_c levels, which indicate a high selectivity of these catalytic systems to water-soluble products (i.e. LA and GA, respectively). According to the literature, Nb-Si@MNP-CP samples led to unprecedented 97.7%–100% values of E_c.

Another very important green characteristic of this chemical process is the easy recoverability of the catalyst from the solid residues and its reuse without a significant loss of activity (around 10%) or selectivity for several runs (an example is given in Figure 10, for 60Nb-Si@MNP-CP sample). This behavior is not necessarily a consequence of the catalyst deactivation but of a catalyst blocking with the small carboxylic acids which, usually strongly adsorb on the catalytic active sites. Indeed, the ICP-OES measurement in the separated reaction solutions showed an insignificant leaching of Nb species in a range of 1–3 ppm, indicating a high hydrothermal stability of the investigated catalysts.

Figure 10:

Catalytic results obtained in the 60Nb-Si@MNP-CP recycling tests.

Along the hydrothermal stability of the generated structure, the existence of multiple functionalities (i.e. acid-base sites, in this case) and the strength of the active sites are crucial for the catalytic process. LA can be generated by both acid and base catalysis. However, the base sites seem to be effective for obtaining high yields in LA (Figure 11). On the contrary, higher GA yield was favored by a decrease in the base site density for both series of catalysts (i.e. CP or SGP) (Figure 12). However, a comparison of the behaviors of the CP and SGP catalysts indicated that the CP samples also produced the highest yields of GA (Figure 11).

Figure 11:

The variation of the yield to lactic acid as a function of the acid and base site densities in the Nb-Si@MNP catalysts.

Figure 12:

The variation of the yield to glycolic acid as a function of the acid and base site densities in the Nb-Si@MNP catalysts.

For the production of lactic acid, the mechanism commonly claimed [16] involves the glucose isomerisation to fructose and retro-aldol condensation of fructose, followed by the triose isomerization to lactic acid. All these reactions are catalyzed by either Lewis acid or base sites. Glycolic acid, on the other hand, seems to be formed through the cleavage of 1,3-dihydroxyacetone to glycol aldehyde and formaldehyde, which claims Brønsted acid sites, with the subsequent oxidation of carbonyl group to GA. However, to the best of our knowledge, this is the first example of a catalyst affording the selective production of water soluble products, especially leading only to LA and GA. Clearly, this involves a synergetic effect of the active sites corroborated to their dispersion and accessibility to the reactants. Thus, these results are superior to those reported in the literature for the one-pot synthesis of lactic acid [9], [10], [16] from cellulose, and of glycolic acid [11], [30] from glycerol.

4 Conclusions

Two different series of magnetite nanoparticles (MNP) coated niobia-silica shell (Nb-Si@MNP) were prepared via co-precipitation (CP) and sol-gel followed by precipitation (SGP) preparation routes. Catalysts with 30 wt% and 60 wt% were prepared in this way. CP route led to materials with high density of NbO-H species that were homogeneously distributed, whereas the SGP procedure led to materials, in which niobium was distributed in between niobyl -Nb=O and NbO-H species on the outer shell of SiO₂@MNP particles.

CO₂- and NH₃-TPD measurements showed that NbO-H species may exhibit both acidic and basic properties. For CP catalysts, they predominated the basic sites and the variation of the niobium content did not influence the acid/base ratios, thereby confirming the high dispersion of niobium phase in the silica matrix. For the SGP samples, higher niobium content corresponded to a lower the acid/base ratio, which also indicated the formation of large agglomerations of the niobium phase on the outer silica shell.

In an un-precedent way, these catalysts were able to selectively transform cellulose into two important water-soluble molecules, i.e. lactic and glycolic acids. The catalytic results confirmed that the active sites responsible for the lactic acid formation from glucose (i.e. through reaction sequences involving the glucose isomerization to fructose, retro-aldol condensation of fructose followed by the triose isomerization) are the basic Nb(V)-OH species. In addition, the unexpected high levels of E_c were observed as a result of the synergetic effect of the active site nature as well as their dispersion and accessibility to the reactants.

In conclusion, the new Nb-Si@MNP catalysts provided an optimum combination of active sites for the one-pot transformation of cellulose to lactic and glycolic acids. The system presented here is still under investigation by our group for further optimization and elucidation of a detailed reaction mechanism.