Cellulose conversion into lactic acid over supported HPA catalysts

Asimina A. Marianou ab, Chrysoula C. Michailof *a, Dimitrios Ipsakis a, Konstantinos Triantafyllidis *b and Angelos A. Lappas a
aChemical Process & Energy Resources Institute, 6th km Harilaou-Thermi Road, 57001, Thessaloniki, Greece. E-mail: mihailof@cperi.certh.gr
bDepartment of Chemistry, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece. E-mail: ktrianta@chem.auth.gr

Received 26th July 2019 , Accepted 15th October 2019

First published on 28th October 2019

Lactic acid is one of the most important high added value chemicals with various applications in diverse fields, e.g. food industry, pharmaceuticals, and plastics. Recently, it has been gaining even more attention as the starting material for the synthesis of bio-based chemicals such as acrylic acid, 2,3-pentanedione and acetaldehyde. The current industrial process for lactic acid synthesis is based on enzymatic fermentation of carbohydrates (e.g. glucose and sucrose), which is a sensitive process in terms of feed quality/purity, requires strict control of the reaction conditions and produces a considerable amount of waste. Therefore, increasing research effort is being placed on the development of alternative green and sustainable chemocatalytic processes. Lactic acid can be synthesized from glucose via a retro-aldol reaction pathway, which is favored under basic conditions, or via catalysts with pronounced Lewis acidity. On the other hand, glucose is produced from cellulose hydrolysis, which requires Brønsted acidity. In view of developing a one-pot chemocatalytic process for the synthesis of lactic acid from cellulose, the present work investigates the effect of oxides (SiO2, SiO2-Al2O3, Nb2O5, Nb2O5-SiO2, Nb2O5-Al2O3), heteropolyacids (HPAs) (TSA, PTA) and supported HPAs on the oxides as bifunctional catalysts with varying ratios of Lewis to Brønsted acid sites on cellulose conversion into lactic acid. According to the experimental results, the type of acidity played a key role in the reaction pathway and consequently in the product distribution. Among the supported catalysts tested, TSA/SiO2-Al2O3 with the highest Lewis to Brønsted acidity ratio led to the highest lactic acid selectivity (38.4%) and yield (23.5%), at 61.2% cellulose conversion. For this catalyst the effect of reaction conditions, catalyst concentration, cellulose crystallinity and the presence of other biomass components (hemicellulose and lignin) was further evaluated. The stability and reuse of TSA/SiO2-Al2O3 were confirmed for at least 3 reaction cycles, whereas the reaction pathway of cellulose conversion was confirmed and validated via a power law kinetic modeling scheme.


Over the last decade there is a constantly growing consensus regarding the need for reducing the dependence on fossil resources particularly for the synthesis of common chemicals. Lignocellulosic biomass has been identified as one of the most promising alternative, sustainable and renewable carbon sources for the production of chemicals via biotechnological or chemocatalytic processes.1 Biomass consists of two carbohydrate-based biopolymers, cellulose and hemicellulose, and lignin which is a phenolic polymer. Cellulose, the most abundant component of biomass, can be hydrolyzed to yield glucose which can be subsequently converted into a variety of platform chemicals such as fructose,2 HMF,3 levulinic acid,4 lactic acid,5etc. through different reaction pathways.6

Among these, lactic acid, a hydroxy carboxylic acid, has a wide range of applications in the food industry (e.g. as a buffering agent, preservative or flavoring agent), in pharmaceuticals (pills, drug delivery systems, personal and oral care products, etc.) and in plastics (for the synthesis of biodegradable poly-lactic acid used in implants, surgical sutures, etc.). Also, it has been attracting the attention of the chemical industry as a building block for the synthesis of green solvents or other bio based chemicals such as acrylic acid, 2,3-pentanedione, acetaldehyde, etc.7–9 The global lactic acid market was valued at approximately $2.9 billion in 2018 and independent market research foresees that it will grow to around $8 to 10 billion by 2025, considering its multiple applications in the high technology sectors identified above.10 The main feedstocks for its synthesis are glucose, fructose and trioses (dihydroxyacetone and glyceraldehyde); however given the abundance and availability of sugars it is obvious that they should be the feedstock of choice. Presently, lactic acid is produced enzymatically via sugar's fermentation and is accompanied by the inherent drawbacks of enzymatic processes including long reaction time, low tolerance in variations of feedstock's quality and voluminous production of waste, thereby affecting negatively its production volume and final cost.11 Thus, the selective heterogeneously catalyzed synthesis of lactic acid from sugars or preferably directly from hexose-based biopolymers such as cellulose is the key point for the development of a sustainable, cost-effective and “green” industrial process.

The synthesis of lactic acid from cellulose involves three consecutive steps: cellulose hydrolysis to glucose, glucose isomerization to fructose and fructose conversion into lactic acid via retro-aldol reactions.12 The hydrolysis reaction is performed with Brønsted acidity, while the following steps of glucose conversion require a catalyst with Lewis acidity and/or basicity. Hence, inorganic bases such as NaOH, Ca(OH)2, Ba(OH)2 and Sr(OH)2[thin space (1/6-em)]13,14 have been employed as catalysts for the conversion of glucose, cellulose or biomass into lactic acid. The results indicated that bivalent alkalis and alkaline earths may form complexes with lactic acid, thereby shifting the equilibrium. Among them Ba(OH)2 has been identified as the most efficient, resulting in quantitative conversion of glucose into lactic acid within 48 h,14 while 25 wt% yield of lactic acid directly from straw has also been reported.13 Nonetheless, an excessive amount of base is required followed by neutralization of the reaction solution with undesirable strong inorganic acids. Recently, Choudhary et al. reported a 70% yield of lactic acid starting from glucose using the heterogeneous basic catalysts CuO and CuCTAB/MgO (CTAB: cetyltrimethylammonium bromide) but in the presence of NaOH.15

Catalytic conversion of cellulose into lactic acid over Lewis acid soluble salts containing Pb2+, VO2+, Er3+, In2+, Zn2+, Mn2+, Al3+ and Sn2+ has been demonstrated by different groups.16–20 Among them, high catalytic activity was presented by ErCl3[thin space (1/6-em)]18 and Er(OTf)3[thin space (1/6-em)]19 resulting in respective lactic acid yields of 62 and 63%, while the synergistic effect of Al(III) and Sn(II) has also been reported (62% lactic acid).20 Nonetheless, many of these salts are expensive and difficult to recover, while others present high toxicity.

From a process point of view, heterogeneous catalysts are beneficial compared to homogeneous ones, owing to their easy separation, recovery and reuse. Therefore, various types of solid Lewis acids and bifunctional catalysts have been studied such as WO4/Al2O3,21 Al2(WO4)3,22 Er2O3/Al2O3,23 Er/K10,24 ZrO2,25 ZrW,26 ZrO2-Al2O3,27 Zr-SBA-1528 and Cr-Sn-Beta.29 Among them, Er/K10 appeared very promising, offering a lactic acid yield of 68% starting from cellulose.24 However, important drawbacks related to Er leaching and structural changes of the catalyst led to significantly reduced catalytic performance after the first cycle.

Based on the above, the present work focuses on the one-pot conversion of cellulose into lactic acid over bifunctional heterogeneous catalysts. The studied catalytic systems include HPAs (TSA, PTA) having increased Brønsted acidity, oxides (SiO2, SiO2-Al2O3, Nb2O5, Nb2O5-SiO2, Nb2O5-Al2O3) with increased Lewis acidity, and supported HPAs on oxide catalysts with varying ratios of Brønsted to Lewis acid sites. The prepared catalysts were thoroughly characterized with respect to their composition, crystal structure, porosity and acidity. Cellulose conversion and product selectivity were correlated with the synergistic effect of catalysts’ Brønsted and Lewis acidity. Among the materials tested, TSA/SiO2-Al2O3 with the highest Lewis to Brønsted acidity ratio provided the highest lactic acid selectivity (38.4%) and yield (23.5%), while its stability and reuse were confirmed for at least 3 reaction cycles. For this catalyst, an attempt has been made to optimize reaction conditions and investigate the reaction mechanism and related pathway via a kinetic modeling scheme.

Results and discussion

Characterization of catalysts

The composition and structural, textural and acidic characteristics of the oxides and the supported POM catalysts, used in this study, are presented and discussed below.

The ICP analysis (Table 1) of the prepared mixed oxides, Nb2O5-Al2O3 and Nb2O5-SiO2, and the supported POM catalysts indicated that the target loading of Nb (15 wt%) or POM (20 wt% corresponding to 15 wt% W) was achieved.

Table 1 Content and textural and acidic properties of the oxides and supported POM catalysts
Catalyst Nbb (wt%) Wb (wt%) S BET (m2 g−1) Total pore volumed (cm3 g−1) Pore sizee (nm) Total acidity (μmol g−1) Brønsted acidityf (μmol g−1) Lewis acidityf (μmol g−1) L/B acid site ratio
a Si/Al ratio. b Chemical analysis by ICP-OES. c BET surface area determined by the multipoint BET method of the N2 adsorption data at −196 °C. d Total pore volume at P/P0 = 0.99. e Average pore diameter from BJH pore size distribution curves (N2 adsorption data). f From FT-IR/pyridine sorption measurements; pyridine equilibration was done at 150 °C for the meso/macroporous oxide supports and at 100 °C for the POM supported catalysts.
SiO2 329 0.75 8.5 0 0 0
TSA/SiO2 13.40 252 0.59 8.5 112 108 4 0.04
PTA/SiO2 13.41 268 0.59 8.5 120 112 8 0.07
Nb2O5 69.80 35 0.18 32.3 25 1 24 24.0
TSA/Nb2O5 54.90 13.12 30 0.12 32.4 165 142 23 0.16
PTA/Nb2O5 58.87 10.01 27 0.12 33.2 155 136 19 0.14
Nb2O5-SiO2 9.59 319 0.66 9.7 22 1 21 21.0
TSA/Nb2O5-SiO2 8.52 13.60 230 0.47 9.8 146 124 22 0.18
PTA/Nb2O5-SiO2 8.45 13.20 216 0.45 9.7 149 123 26 0.21
γ-Al2O3 168 0.53 14.4 97 0 97
Nb2O5-Al2O3 9.60 135 0.43 14.4 87 3 84 28.0
TSA/Nb2O5-Al2O3 8.27 13.50 125 0.33 14.4 206 83 123 1.48
PTA/Nb2O5-Al2O3 8.01 13.52 124 0.34 14.4 186 80 106 1.33
SiO2-Al2O3 (12.5)a 201 0.46 9.7 98 0 98
TSA/SiO2-Al2O3 13.68 178 0.35 9.7 179 49 130 2.65
PTA/SiO2-Al2O3 14.48 171 0.34 9.7 156 45 111 2.47

The textural properties of the catalysts evaluated in this work are summarized in Table 1, while the respective N2 adsorption/desorption isotherms with the corresponding pore size distribution curves (BJH analysis of adsorption data) are shown in Fig. 1 and S1 of the ESI. The Nb2O5, γ-Al2O3, SiO2 and SiO2-Al2O3 oxides exhibited isotherms of similar shapes (Fig. 1A and C), which are characteristic of meso/macroporous materials. The adsorption isotherms of all oxides were of type IV and the desorption isotherms of Nb2O5, γ-Al2O3 and SiO2 were of type H1, while that of SiO2-Al2O3 was of type H2, according to IUPAC classification.30 H1 refers to pores with relatively narrow pore size distribution, while H2 indicates enhanced heterogeneity in pore size/shape. Among them, SiO2 presented the highest surface area (329 m2 g−1) with an average mesopore size (diameter) of ∼8.5 nm (BJH curves in Fig. 1D, data in Table 1), and Nb2O5 obtained via calcination of an ammonium niobium oxalate salt exhibited the lowest surface area (35 m2 g−1) having a meso/macroporous structure with the highest average pore size of ∼32.3 nm (BJH curves in Fig. 1B, data in Table 1). Addition of Nb2O5 in γ-Al2O3 and SiO2 resulted in a slight decrease of the surface area (from 168 to 135 m2 g−1 and from 329 to 319 m2 g−1 respectively) and total pore volume (from 0.75 to 0.66 cm3 g−1 and 0.53 to 0.43 cm3 g−1 respectively), while the average meso/macropore size remained unaffected.

image file: c9gc02622c-f1.tif
Fig. 1 (A and C) N2 adsorption–desorption isotherms and (B and D) the respective pore size distribution curves (based on the BJH analysis of the adsorption data) of oxides and supported TSA catalysts.

In the case of supported POM catalysts, the shape of the N2 adsorption/desorption isotherms (Fig. 1A and C) and the respective BJH pore size distribution curves (Fig. 1B and D) were similar to those of the corresponding oxide supports, indicating minimum changes in the porous structure of the supports upon the grafting of the two POMs. However, the surface area and the total pore volume of the supported catalysts were reduced, due to partial pore blockage and coverage of their surface by POMs (Table 1). The minimum reduction was observed for the TSA/Nb2O5-Al2O3 and PTA/Nb2O5-Al2O3 supported catalysts, which showed a ∼7% and ∼24% decrease in the surface area and total pore volume for both of them respectively. It should also be noted that there was no differentiation in the effect of the POM type on the porous characteristics of the supported catalysts (Table 1).

The crystal structure of the HPAs (TSA and PTA), the oxides and the supported POM catalysts was investigated by XRD measurements (Fig. 2 and S2 of the ESI). The XRD patterns of highly crystalline HPAs31,32 and amorphous γ-Al2O3, SiO2 and SiO2-Al2O3 oxides33,34 were typical of these types of materials. In the case of the Nb2O5 supported on γ-Al2O3 and SiO2 catalysts, no diffraction peaks were observed, although pure Nb2O5 gave rise to several peaks attributable mostly to the orthorhombic (T) crystalline phase, with some additional peaks attributed to the monoclinic (M and H) phase (Fig. 2A).35 This result implies high dispersion of Nb2O5 on both supports, which has been attributed to enhanced interaction of niobium atoms and niobia crystallites with the surface of γ-Al2O3 and SiO2, thus suppressing the growth of larger crystals at the relatively high calcination temperatures of 800 °C and 600 °C, respectively.34,36,37 The supported POM catalysts presented predominantly the pattern of the respective oxide support. The only broad peaks that could be attributed to the presence of POMs were those with a maximum at 6.5–7.5°, which were absent in the patterns of the bare oxides but were observed in all supported catalysts (Fig. 2 and S2). The absence of clearly defined peaks owing to bulk crystalline POMs indicates the high dispersion of HPAs on the surface of the oxide supports.31

image file: c9gc02622c-f2.tif
Fig. 2 XRD patterns of oxides and supported TSA catalysts: (A) (a) TSA (H4SiW12O40), (b) Nb2O5: (*) orthorhombic (T) and (#) monoclinic (M and H) phases, (c) TSA/Nb2O5, (d) γ-Al2O3, (e) Nb2O5-Al2O3, (f) TSA/Nb2O5-Al2O3, and (B) (a) TSA (H4SiW12O40), (b) SiO2, (c) TSA/SiO2, (d) Nb2O5-SiO2, (e) TSA/Nb2O5-SiO2, (f) SiO2-Al2O3, (g) TSA/SiO2-Al2O3.

The structure of the supported POM catalysts was investigated by FTIR experiments. The respective spectra, along with the spectra of the bare oxide supports and the bulk HPAs, are presented in Fig. 3 and S3 of the ESI. In the case of TSA and PTA, the characteristic Keggin anion structure ([SiW12O40]4− and [PW12O40]3−, respectively) is believed to be a tetrahedron of SiO4 or PO4, respectively, surrounded by four edge-sharing W3O13 sets,38 forming an octahedral structure. This structure has four types of oxygen atoms that are responsible for the fingerprint bands of each TSA and PTA respectively. In the spectrum of TSA the characteristic bands presented at 787, 878, 926, 980 and 1020 cm−1 (Fig. 3) were attributed to W–Oc–W asymmetrical, W–Ob–W asymmetrical, Si–Oa asymmetrical and W[double bond, length as m-dash]Od symmetrical and asymmetrical vibrations, respectively.39 Similarly, in the spectrum of PTA the characteristic bands that appeared at 805, 894, 984, 1000 and 1081 cm−1 (Fig. 3) were assigned to W–Oc–W asymmetrical, W–Ob–W asymmetrical, W[double bond, length as m-dash]Od symmetrical and asymmetrical, and P–Oa stretching vibrations, respectively.40

image file: c9gc02622c-f3.tif
Fig. 3 FT-IR spectra of pure TSA (H4SiW12O40), pure PTA (H3PW12O40) and (a) Nb2O5, TSA/Nb2O5, (b) Nb2O5-SiO2, TSA/Nb2O5-SiO2, (c) SiO2-Al2O3, TSA/SiO2-Al2O3, (d) Nb2O5, PTA/Nb2O5, (e) Nb2O5-SiO2, PTA/Nb2O5-SiO2, (f) SiO2-Al3O2, PTA/SiO2-Al3O2.

The presence of these bands in the spectra of the supported POM catalysts verifies that the Keggin structure of HPAs has been preserved in most of the catalysts. It is notable that these fingerprint bands are broader in the spectra of supported POM catalysts compared to those of the bulk HPAs, indicating the homogeneous dispersion of POM on the surface of the support. In addition, H-bonding interactions should also be considered responsible for the slight broadening of the bands, due to dipole–dipole interactions between the Keggin units and the surface of the support.41 The most clearly defined fingerprint bands were observed in the case of POM supported on SiO2 and Nb2O5 catalysts. In contrast, the FTIR spectra of the SiO2-Al2O3 and Nb2O5-SiO2 based catalysts are ambiguous and only indicate that these catalysts contain polyoxotungstate species, not necessarily with the Keggin structure. Thus, DRS UV-vis spectra of the catalysts were obtained. The characteristic peaks at about 250 and 365 nm were observed for pure PTA and TSA, as previously reported. A broad band centered around 250–275 nm was also observed for the various TSA and PTA supported catalysts, with no significant variations between the SiO2-Al2O3 and Nb2O5-SiO2 supported ones, in accordance with previous studies which have shown the elimination of the 365 nm peak of the pure POM upon heating at ca. 100 °C. Furthermore, there is no concrete indication of the formation of species such as WO42−, as this would entail a peak at 212 nm, nor of WO3 like species, as this would appear as peak-broadening above 350 nm. Nevertheless, regardless the exact structure of the supported polyoxotungstate species, they are capable of providing additional Brønsted acidity compared to bare supports, as discussed below.41–43 The corresponding DRS UV-Vis spectra are presented in Fig. S4 of the ESI.

With regard to acidity, the amount and type (Brønsted and Lewis) of acidic sites of the bare supports and the supported POM catalysts were determined by FT-IR/pyridine adsorption measurements and the respective data along with the FT-IR spectra are shown in Table 1 and in Fig. 4 and S5 of the ESI. The band at the range of 1435–1470 cm−1 is attributed to coordinated pyridine on Lewis acid sites, while the band at the range of 1515–1565 cm−1 is assigned to the pyridinium ion on Brønsted acid sites. The bare oxides γ-Al2O3 and SiO2-Al2O3 presented only Lewis acid sites, Nb2O5 presented predominately Lewis and insignificant Brønsted acidity (a very broad peak at 1541 cm−1) (Fig. 4), while SiO2 had negligible acidity (FT-IR spectrum is not shown).44,45 Although Nb2O5 is known as a water-tolerant Lewis acid with increased acidity offered by NbO4 tetrahedra (Nb[double bond, length as m-dash]O), in the presence of water, it gains also some Brønsted acid sites, attributed to the presence of OH groups formed on the catalyst surface, via the polarization of the Nb–O bonds in the NbO6 polyhedra.46 However, in the present study, the number of Nb2O5 acid sites appeared to be relatively low, due to the relatively high calcination temperature (500 °C) used for its formation.47 Loading Nb2O5 (28 wt%) on γ-Al2O3 resulted in a mixed oxide with very few Brønsted acid sites, while the Lewis acidity was slightly decreased compared to the parent γ-Al2O3 (Fig. 4A and Table 1), in accordance with previous studies by Kitano et al.36 Similarly, the presence of Nb2O5 on neutral SiO2 provided Lewis acid sites on the catalyst comparable with those of the bare Nb2O5, and negligible Brønsted acidity.

image file: c9gc02622c-f4.tif
Fig. 4 FT-IR/sorbed pyridine spectra of the oxide supports and the respective supported POM catalysts: (A) γ-Al2Ο3, Nb2O5, Nb2O5-Al2O3, Nb2O5-SiO2 and the respective supported TSA catalysts and (B) SiO2-Al2O3 and the respective supported POM catalysts (pyridine sorption/equilibration was performed at 150 °C for the bare oxides and at 100 °C for the supported POM catalysts).

In the case of supported POM catalysts, it can be clearly seen that Brønsted acid sites were generated by the grafting of POM on the oxide supports. In addition, the pre-existing Lewis acid sites were preserved in POM/Nb2O5 and POM/Nb2O5-SiO2 catalysts, while they were slightly increased in the case of POM/Nb2O5-Al2O3 and POM/SiO2-Al2O3 catalysts, compared to those of the respective supports (Table 1). This additional Lewis acidity in some of the supported catalysts probably originated from unsaturated coordination of the W species after the dispersion of HPAs on the respective oxide supports.21 Furthermore, it should be noted that no significant variation between the effect of TSA and PTA on the acidic properties of the respective catalysts was observed (Fig. 4, S5 and Table 1).

Based on these results, the studied catalysts could be classified into Brønsted acids (bulk HPAs), Lewis acids (bare oxides) and bifunctional materials with both Brønsted acidity and Lewis acidity (supported POM catalysts). The differences in the number and mostly in the type of acid sites play a key role in the catalytic performance of the materials with regard to the distribution of the products, as will be discussed below.

Cellulose conversion into lactic acid over solid catalysts

Evaluation of catalyst performance under constant reaction conditions. The performance of the various catalysts studied in the present work for the conversion of cellulose into lactic acid was evaluated at 175 °C for 60 min in aqueous medium (Table 2). Initially, a blank experiment was performed, resulting in cellulose conversion of 6.97% with 1.94% yield of glucose and traces of fructose, mannose, HMF and formic acid (yield below 0.65% in any case). The above data suggest a minor effect of the aqueous medium (hot compressed water (HCW) (100–374 °C)) on the hydrolysis reaction, attributed to the dissociation of H2O and the occurrence of H3O+ ions, in accordance with the literature.48,49
Table 2 Catalyst screening for the conversion of cellulose in aqueous medium (reaction conditions: 175 °C, 60 min, cellulose: 6 wt%, homogeneous catalyst: 1 wt%, heterogeneous catalyst: 6 wt%)
Catalyst Cellulose conversion (%) Selectivity (%)
Glucose Fructose Mannose HMF Glycolic acid Levulinic acid Formic acid Lactic acid
TSA 42.54 65.99 5.91 6.02 2.29 4.19
PTA 32.15 70.21 10.78 2.93 3.20 1.75
SiO2 6.17 0.50 0.53 0.13 1.12 0.77 0.29 3.17 0.42
γ-Al2O3 6.59 0.30 0.00 1.05 0.15 0.00 2.00 0.81
SiO2-Al2O3 8.77 1.67 0.49 0.43 1.63 0.13 0.23 1.82 1.80
Nb2O5 9.87 2.04 0.50 4.41 0.97 0.45 2.29 2.14
Nb2O5-SiO2 7.05 2.83 0.65 7.50 0.98 0.95 4.04 3.69
Nb2O5-Al2O3 8.95 1.50 0.19 4.53 0.70 0.36 2.88 6.01
TSA/SiO2 53.48 27.11 0.78 5.93 2.70 0.22 9.84 4.69 0.23
TSA/Nb2O5 50.59 24.05 0.19 1.69 4.75 0.48 9.00 5.38 1.22
TSA/Nb2O5-SiO2 51.86 16.99 0.20 2.42 4.07 0.82 7.41 5.39 2.80
TSA/Nb2O5-Al2O3 32.45 8.78 0.69 0.50 6.63 1.70 1.91 3.54 12.92
TSA/SiO2-Al2O3 17.07 8.57 1.72 1.00 6.04 2.97 1.07 3.27 20.91
PTA/SiO2 43.89 21.01 0.26 4.95 1.25 0.88 2.78 2.43 0.23
PTA/Nb2O5 40.70 20.76 0.27 6.77 3.87 1.53 3.14 3.25 1.21
PTA/Nb2O5-SiO2 42.56 15.47 0.26 1.27 4.25 0.61 3.33 2.99 2.99
PTA/Nb2O5-Al2O3 29.85 4.50 0.39 0.46 4.54 1.41 1.05 2.48 10.80
PTA/SiO2-Al2O3 20.19 4.70 0.54 0.64 2.95 2.44 0.25 2.57 15.11

According to the experimental results, cellulose conversion was catalyzed to different extents, by all materials tested. The main products detected were glucose, fructose, mannose, HMF, glycolic acid, levulinic acid, lactic acid and formic acid (Table 2). The catalytic results are discussed in more detail in the next sections separately for each group of catalysts. However, according to the products obtained, an overall reaction scheme was hypothesized and is presented in Scheme 1. In particular, the first step is cellulose hydrolysis to glucose which is catalyzed by Brønsted acid sites. Afterwards, in the presence of Lewis acidity, glucose may isomerize to fructose, which subsequently either dehydrates to HMF (Brønsted acidity) or converts into lactic acid via retro-aldol reactions (Lewis acidity). In particular, the retro-aldol pathway includes: fructose conversion into the isomers 1,3-dihydroxyacetone and glyceraldehyde, subsequent dehydration and keto–enol rearrangement of glyceraldehyde to pyruvaldehyde followed by conversion of the latter into lactic acid. Formic acid accompanied by acetaldehyde (the latter not detected by our analytical methods) and/or acetic acid accompanied by formaldehyde (the latter not detected by our analytical methods) may also be formed from pyruvaldehyde via α-dicarbonyl cleavage. Further on, lactic acid may be degraded to formic acid and acetaldehyde, while the co-product HMF may be converted into levulinic acid and formic acid (Brønsted acidity). Finally, under alkaline conditions or increased Lewis acidity, glucose may be degraded to glycolic acid via keto–enol rearrangement.44,50

image file: c9gc02622c-s1.tif
Scheme 1 Possible reaction pathway (based on the obtained results with the different catalysts and the reaction mechanisms that have been proposed in ref. 50–53).
Heteropolyacids (TSA, PTA) with increased Brønsted acidity. Heteropolyacids possess increased Brønsted acidity, which renders them suitable for cellulose hydrolysis. In particular, TSA and PTA exhibit stronger Brønsted acidity in aqueous solutions than the typical mineral acids, due to their lower deprotonation enthalpies (DPE), as evaluated by DFT calculations reported in previous studies.54–56 In a study by Shimizu et al.,57 both TSA and PTA proved to be extremely effective for cellobiose hydrolysis to glucose under hydrothermal conditions, with respective selectivities of 86 and 96% and respective yields of 53 and 51% at 120 °C. In the present work, both TSA and PTA catalyzed cellulose hydrolysis to glucose with selectivities of 66 (28.1% yield) and 70% (22.6% yield) respectively (Table 2). The higher selectivity achieved by PTA may be attributed to its stronger Brønsted acid sites. Lactic acid was not detected in either case owing to the absence of Lewis acidity, which is required for the retro-aldol dominated pathway (Scheme 1). On the other hand, due to increased Brønsted acidity, glucose dehydration to 5-HMF, followed by a hydration reaction to levulinic acid and formic acid (Scheme 1) also took place, but with a combined selectivity lower than 3.5% for all reaction products.
Oxides with increased Lewis acidity. Among the oxides tested, neutral SiO2 (Table 2), and γ-Al2O3 and SiO2-Al2O3, which possess only Lewis acidity, demonstrated poor catalytic activity. Cellulose conversion was 6.2–8.8% while the main products detected were traces of glucose, lactic acid and formic acid with respective selectivities and yields below 2.0% and 0.16%, in all cases. The minor hydrolysis of cellulose to glucose in the absence of significant Brønsted acidity could be attributed to water dissociation,48 being further promoted by its interaction with the surface of the catalysts.58,59

In the case of Nb2O5, slightly higher cellulose conversion (9.9%) and glucose selectivity (2.0%) were obtained, while HMF was formed at a slightly higher selectivity (4.4%) compared to lactic acid (2.1%) (Table 2). These results suggest that the water-tolerant Lewis acid sites of Nb2O5 catalyzed the isomerization of glucose into fructose and the evolved fructose was rapidly dehydrated to HMF due to Brønsted acid sites that were probably generated on the surface of the catalyst due to the presence of water, as has been previously reported (paragraph 3.1).46 The presence of levulinic acid and formic acid indicated that some hydration of HMF also occurred. The same performance was obtained in the case of the mixed oxide Nb2O5-SiO2, with comparable acidity to that of pure Nb2O5, where the formation of HMF (7.5% selectivity) over lactic acid (3.7% selectivity) was favored. In contrast, in the presence of the mixed oxide Nb2O5-Al2O3, with significantly higher Lewis acidity compared to both Nb2O5 and Nb2O5-SiO2, the reaction pathway changed towards lactic acid with a selectivity of 6.0%, while the selectivity for HMF was 4.5%.

Supported catalysts with combined Brønsted and Lewis acidity. The combined effect of both types of acidity on the reaction pathway and consequently on product distribution becomes more evident in the case of supported POM catalysts with varying degrees of Brønsted/Lewis acidity. By supporting POMs on oxides via impregnation, the conversion of cellulose and formation of glucose increased remarkably (as compared to those obtained in the presence of the respective bare oxide), accompanied in most cases by a significant increase of HMF and lactic acid selectivity (Table 2). Clearly, increased Brønsted acidity provided by POMs favored cellulose hydrolysis to glucose (Table 2). Meanwhile, the subsequent antagonistic reactions of the produced glucose towards HMF and/or lactic acid were significantly affected by the relative content of Lewis and Brønsted acid sites. Glucose decomposition to glycolic acid also took place under increased Lewis acidity, but at a selectivity lower than 3% in any case. Further information can be revealed by plotting cellulose conversion, glucose selectivity, lactic acid selectivity and the ratio of HMF to lactic acid selectivity versus the ratio of catalysts’ Lewis to Brønsted (L/B) acid sites (Fig. 5). Based on these graphs, it can be concluded that the decrease of the L/B acid site ratio enhanced the selective hydrolysis of cellulose to glucose (Fig. 5a and b). In contrast, increased Lewis acidity over Brønsted acidity favored further glucose conversion, with the predominance of the retro-aldol pathway towards lactic acid over the formation of HMF (Fig. 5c and d). These results are in agreement with the previous published work by our group, where it was observed that starting from glucose, its conversion to HMF over lactic acid increased consistently with the decrease of the L/B acid site ratio and vice versa.44
image file: c9gc02622c-f5.tif
Fig. 5 Correlation of the Lewis to Brønsted (L/B) acid site ratio with (a) cellulose conversion, (b) glucose selectivity, (c) lactic acid selectivity and (d) ratio of HMF to lactic acid selectivity for the supported POM catalysts.

Among the supported TSA catalysts, TSA/SiO2 with the lowest ratio of L/B acid sites (Table 1) led to the highest cellulose conversion, glucose selectivity and yield with values of 53.5%, 27.1% and 14.5% respectively. On the other hand, TSA/SiO2-Al2O3 with the highest L/B acid site ratio, despite being the least active regarding cellulose conversion (17.1%), proved to be the most selective catalyst for lactic acid with the highest value of 20.9%. The antagonistic pathway towards HMF also took place in all studied cases, with the highest selectivity of 6.6% obtained in the presence of TSA/Nb2O5-Al2O3. The same catalytic performance in terms of cellulose conversion and selectivity of glucose, HMF and lactic acid was observed from supported PTA catalysts, but with lower values of selectivity and yields (Table 2) compared to the respective TSA supported catalysts in any case.

Catalyst stability. Another important factor for the catalytic activity is catalyst stability, which is affected by the hydrothermal reaction conditions employed. Therefore, the structure of the POM ion on the used catalysts after the reaction was examined by FTIR measurements, while the potential leaching of W and Nb, in the case of Nb2O5 based catalysts, from the solid materials into the reaction mixture was determined by ICP-OES analysis (Table 3). Before each analysis, the used catalysts were washed with H2O and acetone, dried at 80 °C overnight and heated at 200 °C, as described in the Experimental section. Although the carbonaceous deposits formed during the reaction were not significantly reduced after heating, calcination at higher temperatures (400–500 °C) was not performed due to thermal instability of the POMs’ ionic structure.60
Table 3 Niobium (Nb) and tungsten (W) leaching from the supported POM catalysts in the reaction mixture
Catalyst ICP-OES (ppm) Leachinga (%)
Nb W Nb W
a Calculated on the basis of the metal content in the initial supported POM catalysts prior to reaction.
TSA/SiO2 8280 67.61
PTA/SiO2 7825 61.17
TSA/Nb2O5 13 5170 0.04 63.61
PTA/Nb2O5 51 4350 0.09 47.23
TSA/Nb2O5-SiO2 37 7015 0.55 57.05
PTA/Nb2O5-SiO2 60 7865 0.86 72.30
TSA/Nb2O5-Al2O3 140 3925 1.99 34.18
PTA/Nb2O5-Al2O3 102 4210 1.54 37.58
TSA/SiO2-Al2O3 102 3.84
PTA/SiO2-Al2O3 802 6.50

According to the FTIR spectra of the dried samples (representative results are shown in Fig. S6 of the ESI), the presence of the respective typical bands of pure TSA and PTA verified that the characteristic Keggin ionic structure ([SiW12O40]4− and [PW12O40]3−) has been preserved for both of them in the case of POM/SiO2 and POM/Nb2O5 catalysts. On the other hand, these fingerprint bands or any other bands corresponding to different POM ionic structures could hardly be identified in the case of POM/SiO2-Al2O3, POM/Nb2O5-SiO2 and POM/Nb2O5-Al2O3 catalysts.

It is known that one of the usual problems in the use of supported POM catalysts is related to the leaching of HPAs in the reaction medium.61 In our case and based on the ICP-OES analysis of the liquid products (Table 3), all the supported POM catalysts presented leaching of tungsten (W) to different extents, whereas the leaching of Nb from the catalysts in which Nb2O5 was part of the support was negligible. These results suggest that the type of support has a significant impact on the catalyst (POM) stability. The presence of Al2O3 in POM/Nb2O5-Al2O3 and POM/SiO2-Al2O3 catalysts provided an enhanced stability and very low W leaching. A reasonable explanation for this could be related to the difference in electronegativity between the W of POMs and the respective metal (M) of the support. In particular, the higher the difference in electronegativity, the higher the polarization of the bond formed between the POM and the support (W–O–M) resulting in less stable supported catalysts.61,62 It seems that the bonds formed between Si and Nb with W (Si–O–W and Nb–O–W respectively) were less stable compared to the bond formed between Al and W (Al–O–W), as the catalyst stability (indicated by W leaching) increased in the following order: POM/Nb2O5-SiO2 < POM/SiO2 ≤ POM/Nb2O5 < POM/Nb2O5-Al2O3 < POM/SiO2-Al2O3. Based on this, it can be postulated that the stability of grafted POM in the case of Nb2O5-Al2O3 and SiO2-Al2O3 supports is mostly due to Al2O3. However, additional study is required in order to further support this hypothesis. The most stable catalysts were TSA and PTA supported on SiO2-Al2O3 with less than 3.9 and 6.5% of W leaching respectively. The effect of W leaching was determined by using the liquid filtrate after hydrothermal treatment of fresh supported POM catalysts, which contained only the leached W species as reaction media for cellulose conversion. The results indicated a minimum contribution of the leached species to the overall cellulose conversion as only glucose was detected with 1.5% yield.

In summary, among the supported catalysts tested, TSA/SiO2-Al2O3 led to the highest lactic acid selectivity and yield and proved to be the most stable catalyst. Therefore, it was selected for further study, in order to determine the optimum reaction conditions for maximizing the synthesis of lactic acid from cellulose.

Impact of reaction conditions with the TSA/SiO2-Al2O3 catalyst

Effect of reaction temperature and time. Prolonged reaction times (5–48 h) and a wide range of temperatures (100–240 °C) have been previously reported to favor the production of lactic acid from cellulose over solid catalysts.21,63,64 Therefore, in our case, the effect of temperature was studied at 125, 150 and 175 °C (Fig. 6a) in the presence of TSA/SiO2-Al2O3, as above 200 °C, HPAs are not thermally stable and their characteristic Keggin ionic structure is likely to change. The reaction temperature has a significant effect on the endothermic cellulose hydrolysis and on the subsequent glucose conversion, since it affects the reaction kinetics. According to the experimental results, cellulose conversion progressively increased from 0.3 to 17.1% with temperature, glucose selectivity presented a maximum (28.5%) at 150 °C, while the corresponding values for HMF and organic acids (glycolic acid, levulinic acid, formic acid and lactic acid) increased. Above 150 °C, glucose decomposition reactions were favored with the retro-aldol pathway overcoming the dehydration reaction towards HMF (Fig. 6a). The reduced selectivity of HMF compared to lactic acid can also be attributed to the enhanced formation of levulinic acid and formic acid (Scheme 1) favored by increased temperature. Finally, glycolic acid synthesis was not affected by the reaction temperature, as its selectivity remained almost stable at low values (2.7–2.9%). On the other hand, the lactic acid selectivity and yield reached 20.9% and 3.6% respectively at the temperature of 175 °C, over the corresponding values of 6.0% and 1.0% for HMF, and therefore 175 °C was selected as the optimum reaction temperature.
image file: c9gc02622c-f6.tif
Fig. 6 Effect of reaction (a) temperature (t: 1 h) and (b) time (T: 175 °C) on cellulose conversion and product selectivity with the TSA/SiO2-Al2O3 catalyst in aqueous medium (6 wt% cellulose, catalyst/cellulose 1[thin space (1/6-em)]:[thin space (1/6-em)]1 w/w).

The effect of reaction time at 175 °C was also evaluated in the range of 1–48 h and the experimental results are presented in Fig. 6b. Cellulose conversion increased almost exponentially from 17.1 to 73.7% with time, while the glucose selectivity decreased almost linearly from 8.6 to 0.1%, as the subsequent conversion reactions were promoted. The glycolic acid selectivity increased with time up to 16 h, with a maximum of 4.2%, contrary to the HMF selectivity, which decreased continuously from 6.0 to 0.4%, probably due to the formation of formic acid and levulinic acid. The selectivity of lactic acid presented a maximum of 40.6% after 16 h of reaction, while after 24 and 48 h, it slightly decreased to 38.4 and 37.4% respectively, showing that only minor decomposition of lactic acid took place. In addition, the yield of lactic acid increased remarkably from 3.6 to 27.5% with increasing time, indicating that retro-aldol reactions are the limiting step in its formation. Consequently, it can be noted that extended reaction times favor the retro-aldol pathway towards the production of lactic acid, along with the degradation of HMF, enhancing the selectivity towards the target product (Scheme 1). Considering a compromise among conversion, yield and selectivity, the selected optimum reaction time was 24 h, where the lactic acid selectivity and yield reached 38.4% and 23.5% respectively at 61.2% cellulose conversion.

Effect of the catalyst/cellulose ratio. The effect of the cellulose to catalyst ratio was investigated for the case of TSA/SiO2-Al2O3 under the selected optimum reaction conditions (175 °C, 24 h) and the results are shown in Fig. 7. The cellulose concentration remained constant at 6 wt% while the catalyst concentration varied from 3 to 9 wt%. Cellulose conversion increased with increasing catalyst concentration, whereas the selectivity of glucose, fructose and HMF slightly decreased, while the formation of the organic acids glycolic, levulinic and formic acid remained almost stable. Lactic acid yield increased from 13.9 to 27.2% with increasing catalyst concentration, whereas the selectivity reached a maximum value of 38.4% when TSA/SiO2-Al2O3 was used at a concentration of 6 wt%. This amount of catalyst corresponds to a catalyst/cellulose ratio of 1[thin space (1/6-em)]:[thin space (1/6-em)]1. When the catalyst was in excess of 1.5/1 (respective catalyst concentration: 9 wt%), despite the increase in cellulose conversion from 61.2 to 72.8%, the lactic acid selectivity was slightly decreased (37.3%) while the selectivities of levulinic acid (4.2%) and formic acid (5.1%) were slightly enhanced.
image file: c9gc02622c-f7.tif
Fig. 7 Effect of catalyst concentration on cellulose conversion and product selectivity with the TSA/SiO2-Al2O3 catalyst in aqueous medium (T: 175 °C, t: 24 h, 6 wt% cellulose).

It can be concluded that by tuning the reaction conditions, it is possible to increase the selective conversion of cellulose into lactic acid by influencing the glucose conversion reactions. According to our findings, high reaction temperatures (up to 175 °C), extended reaction time (24 h) and a moderate catalyst to cellulose ratio (1[thin space (1/6-em)]:[thin space (1/6-em)]1) favor the production of lactic acid, contrary to lower temperature and shorter reaction times that favor the first step of cellulose hydrolysis to glucose.

Catalyst reuse – TSA/SiO2-Al2O3

As discussed in the Introduction section, one of the most important advantages of solid catalysts is their recovery and reuse, if possible, without losing their catalytic activity. Therefore, the performance of TSA/SiO2-Al2O3 in terms of catalyst reuse was evaluated in the present study. The solid catalyst was washed with H2O and acetone followed by heating at 200 °C, before being used for a successive cellulose conversion run under the optimum reaction conditions (175 °C, 24 h).

After its first use, the catalyst was loaded with 10.6 wt% carbon in the form of residues including coke, humins, unreacted compounds, etc., which were hardly affected by heating at a low temperature of 200 °C. Complete regeneration of the catalyst may be achieved after calcination for 3 h at 500 °C, but this treatment will have a negative effect on the catalyst structure, as already mentioned. However, despite the relatively high content of residues, the catalyst retained its activity in terms of cellulose conversion and product selectivity (Fig. 8), suggesting that the carbon deposits do not induce significant blockage of the active sites up to three cycles, a conclusion also supported by SEM-EDS mapping indicating that the structure, morphology and dispersion of W-atoms on the catalyst surface were the same before and after the reaction (Fig. S7). ICP-OES analysis performed in the liquid samples after each cycle showed that W presented almost the same leaching after the first two cycles (3.84% and 3.59% respectively), which decreased to 1.59% and 1.51% after the third and fourth cycles respectively, indicating that catalyst composition was not significantly affected. It is assumed that an interaction occurs between the reaction products and the catalyst that causes the leaching, as hydrothermal stabilization of the catalyst prior to its use showed a mere 0.3 wt% leaching of W.

image file: c9gc02622c-f8.tif
Fig. 8 Effect of catalyst recycling and regeneration on cellulose conversion and product selectivity over the TSA/SiO2-Al2O3 catalyst in aqueous medium (T: 175 °C, t: 24 h, 6 wt% cellulose, catalyst/cellulose: 1/1).

Use of pretreated lignocellulosic biomass/cellulose as feedstocks with the TSA/SiO2-Al2O3 catalyst

The complex structure of lignocellulosic biomass consisting of cellulose, hemicellulose and lignin restricts the efficiency of the cellulose hydrolysis reaction.65 In addition, the access of reactants and catalysts to β-1,4-glycosidic bonds in cellulose is significantly hindered by the ordered structure of crystalline cellulose.66 Pretreatment of biomass with the aim to facilitate the access to cellulose (and partially destroy its crystalline structure) by removing/separating hemicellulose and lignin, as well as treatment of microcrystalline cellulose with methods such as ball-milling and sonication, in order to reduce its crystallinity, renders the β-1,4-glycosidic bonds more accessible and consequently increases the efficiency of the hydrolysis reaction.67,68 In the present study, different types of pretreated biomass samples (after the removal of hemicellulose or lignin and hemicellulose) and microcrystalline cellulose samples treated with ball-milling and sonication were used as feedstocks, with the purpose of investigating the influence of pretreatment on cellulose conversion in aqueous medium, over the TSA/SiO2-Al2O3 catalyst.
Characterization of the parent and treated cellulose and biomass. The composition of the parent lignocellulosic biomass (beech wood sawdust, Lignocel) was significantly influenced by both pretreatment methods, as indicated by the results based on NREL analysis methods (Table 4). The lignin content was decreased from 26.1% to 4.1% after the Milox delignification process, leading to a biomass sample (del.Lig.) consisting mainly of cellulose (84.4%), as hemicellulose was also partially removed. In the case of biomass pretreated with 2 wt% H2SO4 (Lig.hem.free), hemicellulose decreased from 20.9% to 4.7%, while a minor removal of cellulose was also observed, with lignin remaining almost unaffected, thus occupying a high relative content in the treated biomass sample.
Table 4 Composition of the parent and pretreated biomass feeds
Biomass Cellulose (%) Hemicellulose (%) Lignin (%)
Lignocel 46.6 20.9 26.1
Lig.hem.free 40.5 4.7 53.1
del.Lig. 84.4 5.6 4.1

According to the XRD patterns of the parent and treated cellulose samples (Fig. S8A of the ESI), it is evident that both pretreatment methods of microcrystalline cellulose (i.e. sonication and ball milling) are capable of reducing its crystallinity. The intensities of peaks due to cellulose crystal planes, especially of the strongest one at 2θ 22.6° corresponding to the crystalline plane 002,69 decreased compared to those of microcrystalline cellulose and this reduction was enhanced with time for both treatments. Comparing the two methods, ball-milling proved to be more efficient, as after 24 h of milling, cellulose became almost amorphous, while ultrasound treatment proved to be less effective probably due to the mild conditions employed. The above observations are reflected by the cellulose crystallinity index (CrI) and crystal size (CrS) of cellulose crystallites.70 Based on the results presented in Table 5, CrI decreased from 86.8% to 79.8% after sonication for 4 h, while the corresponding value after ball-milling for 24 h could not be determined as the XRD pattern did not contain any diffraction peaks. Additionally, CrS was also reduced from 5.0 nm to 4.5 nm after sonication for 4 h.

Table 5 Cellulose crystal size (CrS) and crystallinity index (CrI) of the parent and treated biomass and cellulose samples
Feed CrS (nm) CrI (%)
MCC 5.0 86.8
son.cel._2h 4.6 86.2
son.cel._4h 4.5 79.8
b.m.cel._10h 3.9 72.3
b.m.cel._24h Amorphous
Lignocel 3.2 59.5
Lig.hem.free 3.6 59.0
del.Lig. 3.3 74.2

The crystallinity of biomass samples is attributed to the crystal structure of cellulose, as hemicellulose and lignin are amorphous biopolymers. According to the XRD patterns of the parent and treated biomass (Fig. S8B of the ESI), it can be argued that both pretreatment methods affected biomass/cellulose crystallinity. CrI seemed to increase from 60.9% to 74.2% after the delignification process (Table 5), but this is due to the increased concentration of crystalline cellulose in the dignified sample (del.Lig), after amorphous lignin and hemicellulose removal. In the case of the Lig.hem.free sample, the CrI appeared to be slightly reduced (59.0%) compared to the parent biomass probably due to the minor cellulose subtraction and/or destruction that accompanied hemicellulose removal.

Catalytic conversion of treated cellulose/biomass over TSA/SiO2-Al2O3. Biomass/cellulose conversion was performed under the optimum reaction conditions selected for microcrystalline cellulose (175 °C for 24 h) in the presence of TSA/SiO2-Al2O3 in aqueous medium, whereas the catalyst to cellulose ratio remained stable at 1/1. As shown in Fig. 9A, sonication and ball-milling pretreatment methods led to a significant increase in cellulose reactivity, while the lactic acid yield was also improved, in all cases. In particular, after sonication for 4 h, the conversion of cellulose increased from 61.2 to 76.2% and the lactic acid yield was slightly improved from 23.5 to 24.6%, while after ball-milling for 24 h the respective values were 84.7% and 31.2%. These results suggest that the reduction in cellulose crystallinity enhanced the conversion of cellulose and the production of lactic acid. Previous studies on the hydrolysis of cellulose in water have demonstrated that the amorphous portions are much more reactive than the crystalline ones67,71 and therefore cellulose conversion was enhanced in the case of treated samples compared to microcrystalline cellulose (MCC). Comparing the two pretreatment methods, ball-milling, resulting in samples with less crystallinity (Table 5), proved to be more efficient for the production of lactic acid from cellulose. Between the ball-milled samples, cellulose after milling for 10 h proved to be more selective towards lactic acid, exhibiting 77.2% conversion, 38.6% selectivity and 29.2% yield of lactic acid. Upon increasing the cellulose ball-milling time from 10 h to 24 h, cellulose conversion and lactic acid yield increased, but the lactic acid selectivity slightly decreased to 36.8%, accompanied by a simultaneous minor increase of the selectivity of levulinic acid and formic acid. This observation indicates that the reaction pathway towards HMF (Scheme 1) was also favored by the reduction of cellulose crystallinity.
image file: c9gc02622c-f9.tif
Fig. 9 Effect of pretreatment on (A) cellulose and (B) biomass conversion and product selectivity respectively with the TSA/SiO2-Al2O3 catalyst in aqueous medium (T: 175 °C, t: 24 h, 6 wt% cellulose). The calculations of product selectivity in (B) are biomass based.

Regarding the biomass samples, it is evident from the results presented in Fig. 9B that the presence of hemicellulose and lignin affected considerably the conversion of biomass and product distribution. Biomass conversion reflected the combined conversion of cellulose and hemicellulose components, as lignin cannot be converted under the reaction conditions employed. Unreacted biomass after the reaction could not be separated from the solid catalyst and NREL analysis could not be performed in order to determine the unreacted hemicellulose and cellulose content. Therefore, the calculations for product selectivity and yield were biomass based. When the parent biomass was used as the feed, the conversion was 56.0%, while the lactic acid selectivity reached a value of 23.8%. The minor selectivity towards HMF, levulinic acid and formic acid, with the respective values of 1.2, 1.7 and 3.0%, indicated that the reaction conditions were well chosen to be unfavorable for the glucose dehydration pathway (Scheme 1). Additionally, the formation of acetic acid with a selectivity of 4.2% is mainly attributed to the acetyl units in biomass, which are released upon hemicellulose hydrolysis.72 In the case of the Lig.hem.free sample, the lactic acid selectivity slightly increased (25.2%), compared to the parent Lignocel, despite the presence of the lignin inhibitor. Due to hemicellulose removal, cellulose was the only carbohydrate left to interact with the catalyst active sites and therefore the lactic acid selectivity was increased, while biomass conversion appeared to be slightly reduced (53.7%). In addition, the detection of levulinic acid and formic acid with similar values of selectivity (2.9 and 2.4% respectively) compared to the parent biomass verified that glucose decomposition followed the same reaction pathways. Delignified Lignocel (del.Lig.) with the highest cellulose content (84.4%) resulted in the highest biomass conversion (77.4%) and lactic acid selectivity (27.2%), as the number of accessible β-1,4-glycosidic bonds increased due to lignin removal, while the selectivity of levulinic acid and formic acid presented a minor increase (4.4 and 5.6%).

Reaction kinetics

As observed, TSA/SiO2-Al2O3 proved to be the best performing and stable catalyst for cellulose and biomass conversion to lactic acid. Therefore, it was selected for the investigation of the kinetic modeling scheme with cellulose as the starting material.

The kinetic modeling of the catalytic conversion of cellulose into lactic acid is not a trivial task and to the best of the authors’ knowledge it is still under investigation in the literature. Currently, the most prominent and generalized approach is the implementation of power law kinetic models that are able to quantify the experimental findings.73–75 As discussed in the previous sections, cellulose is hydrolyzed to glucose (1st stage), then glucose reversibly isomerizes to fructose (2nd stage) and/or epimerizes to mannose. Subsequently, fructose dehydrates to HMF and/or is converted into lactic acid (3rd stage). In the respective stages, intermediate components could also be formed but are assumed to react instantly to form the main products. Hence, their presence is excluded from further consideration in this preliminary kinetic study. Similarly, HMF and lactic acid can be converted into acidic products as long as both reaction conditions and catalytic materials can facilitate this route. Specifically, HMF can be hydrated to levulinic and formic acids (4th stage), while lactic acid can be converted into formic acid and formaldehyde (5th stage). Furthermore, glucose, fructose and HMF could also be converted into by-products such as organic acids (acetic, glycolic, etc.) or humins (mainly from HMF). Scheme 2 shows the prominent kinetic modeling scheme of cellulose conversion into lactic acid that is discussed in this study. It is highlighted that Scheme 2 is a simplified, yet accurate representation of the overall reaction Scheme 1, whereas the intermediate (e.g. mannose, glyceraldehyde, and pyruvaldehyde) and final products (e.g. dihydroxyacetone and formaldehyde) that were not detected were excluded from the kinetic evaluation.

image file: c9gc02622c-s2.tif
Scheme 2 Kinetic model for the conversion of cellulose into lactic acid through a series of consecutive reactions.

The accompanying ESI presents in detail the applied kinetic modeling work. Briefly, a power law kinetic scheme was developed and confirmed the simplified, yet accurate, reaction Scheme 2 through a set of dedicated experiments. These experiments were related to the identification of the intermediate products and the underlying reaction routes. As has already been presented by other research studies,73–78 this power law kinetic approach (with regard to the main reactant) is the usual strategy to be applied.

As was found, the kinetic investigation of cellulose conversion indicated that (a) lactic acid is the main reaction product, (b) it does not decompose further under the applied experimental conditions and (c) it is produced in a higher rate when fructose is present in a higher concentration. This valuable research evidence additionally highlights the importance of the glucose to fructose isomerization step. Furthermore, when fructose is used as the starting material (and thus in a higher concentration as compared to glucose), the HMF evolution profile shows a slightly higher rate, whereas it is verified that HMF is converted into humins at a higher rate as compared to formic and levulinic acid. The estimated kinetic parameters of this study are presented in detail in the ESI and have been found to be well within the range (especially for reaction orders) of the postulated kinetic values observed in previous studies.74,75,78,79


In the present work, a series of bifunctional supported POM catalysts with varying ratios of Brønsted to Lewis acid sites were synthesized, characterized and evaluated in the conversion of cellulose into valuable products in aqueous medium. Under the same conditions, the respective bulk HPAs (TSA, PTA) (Brønsted acids) and oxides (Lewis acids) used also as supports were evaluated. It was clearly demonstrated that the type of acidity affected significantly the cellulose depolymerization reaction pathway and as a consequence, the product distribution in the reaction mixture. In the homogeneous aqueous reaction systems of dissolved HPAs with pronounced Brønsted acidity, glucose was the main product. On the other hand, oxides demonstrated poor activity, despite their increased Lewis acidity, as they were unable to catalyze effectively the first step of cellulose hydrolysis. However, the supported HPAs, which combined Brønsted and Lewis acidity, catalyzed both cellulose hydrolysis to glucose and the subsequent dehydration and retro-aldol reaction pathway of glucose conversion mainly into lactic acid. Among them, those with an increased Lewis to Brønsted acidity ratio proved to be more selective towards lactic acid, against HMF and other organic acids.

TSA/SiO2-Al2O3 with the highest Lewis/Brønsted acid site ratio proved to be the most efficient catalyst for cellulose conversion into lactic acid. When the reaction was performed in aqueous solution at 175 °C for 24 h with a catalyst/cellulose ratio of 1[thin space (1/6-em)]:[thin space (1/6-em)]1, the lactic acid selectivity reached a remarkable 38.4%, with a significant yield of 23.5% and high cellulose conversion of 61.2%. To the best of the authors’ knowledge, these results are among the highest values reported in the literature so far in the presence of a heterogeneous stable and “green” catalyst.24,25,27 The study of the effect of reaction conditions (reaction time and temperature, catalyst concentration) conducted with TSA/SiO2-Al2O3 revealed that moderate temperatures (150 °C) and low reaction times (60 min) favored the production of glucose, while higher temperatures (175 °C) and longer reaction times (up to 24 h) induced the production of lactic acid. The important issue of the stability and reuse of the supported catalysts was examined using the above optimum TSA/SiO2-Al2O3 material, by correlating the catalytic performance of the recovered catalysts with the structural integrity of the supported HPAs, in addition to the monitoring of W leaching into the reaction medium. It was demonstrated that the overall catalyst's structure was not significantly affected by the hydrothermal conditions of the cellulose depolymerization reaction. Furthermore, the limited leaching of W (in the order of 40–102 ppm) after each cycle had a minor effect on cellulose conversion and lactic acid selectivity, thus resulting in a stable catalytic performance for at least 3 cycles. Regarding cellulose crystallinity which also affects cellulose decomposition, it was verified that its reduction improved significantly the catalytic activity. When the reaction was performed with ball-milled cellulose (24 h milling) under the optimum reaction conditions in the presence of TSA/SiO2-Al2O3, cellulose conversion and lactic acid selectivity increased up to 84.7% and 31.2% respectively. In addition, as far as the conversion of lignocellulosic biomass is concerned, TSA/SiO2-Al2O3 retained the catalytic activity towards lactic acid, even in the presence of hemicellulose and lignin. Finally, through a power law kinetic modeling approach that was applied to the optimum TSA/SiO2-Al2O3 catalyst, the proposed reaction scheme of cellulose conversion into lactic acid through intermediate stages was quantified and confirmed. Under the reaction conditions employed, the increased stability of lactic acid was revealed, while the formation of other organic acids, such as glycolic acid, formic acid and levulinic acid, was attributed to glucose and fructose degradation reactions.



Microcrystalline cellulose (MCC) was purchased from JRS Pharma with the commercial name VIVAPUR Type 200. A commercial wood biomass feed (Lignocel HBS 150-500) originating from beech wood was also used in the current study. Delignified biomass (del.Lig.) was obtained after applying the Milox process using formic acid (CH2O2) and 2 wt% hydrogen peroxide (H2O2) (80 °C, 60 min),80 while hemicellulose free biomass (Lig.hem.free) was produced after the treatment of biomass with 2 wt% aqueous H2SO4. SiO2 (Sigma-Aldrich), γ-Al2O3 and SiO2-Al2O3 (with Si/Al = 12.5) (Saint-Gobain NorPro) were used as catalysts and also as supports of polyoxometalates (POMs). Nb2O5 was obtained after calcination at 550 °C for 5 h under an air flow of the corresponding ammonium niobate oxalate salt (NH4[NbO(C2O4)2(H2O)2]·(H2O)n, Sigma-Aldrich). The two heteropolyacids (HPAs) (also referred to as POMs) with the Keggin structure used were phosphotungstic acid (PTA) (H3[PW12O40nH2O) and tungstosilicic acid (TSA) (H4[SiW12O40nH2O) (Sigma-Aldrich).

Cellulose pre-treatment methods

The effect of cellulose crystallinity on the cellulose conversion reaction was also studied by using ball-milled and sonicated samples of MCC as feeds.

The milling of microcrystalline cellulose was performed with a 200 ml stainless steel (SS) cylindrical container using SS balls at a rotating speed of 100 rpm for 10 and 24 h. After completion of milling, cellulose was separated from the balls and stored until use.

Ultrasound treatment was performed on an Ultrasonic Processor (Hielscher, UP200Ht). A mixture of cellulose (3 g) and deionized water (47 g) was subjected to sonication at a power of 150 W and a frequency of 50 kHz at room temperature. After the desired time (2 and 4 h), the resulting suspension was used for hydrolysis experiments without further treatment.

Catalyst preparation

Supported niobium pentoxide on γ-Al2O3 and SiO2 by wet impregnation. Supported Nb2O5 on γ-Al2O3 and SiO2 catalysts, with Nb loading of 15 wt%, were prepared by wet impregnation. In particular, an appropriate amount of ammonium niobate oxalate corresponding to 15 wt% Nb loading was dissolved in aqueous solution, followed by the addition of the support. The mixture was stirred at room temperature for 4 h in a rotary evaporator and then dried under vacuum. Finally, the obtained materials denoted as Nb2O5-Al2O3 and Nb2O5-SiO2 were calcined at 800 °C for 3 h and at 600 °C for 4 h respectively under an air flow at a heating rate of 5 °C min−1.
Supported POMs on oxides by incipient wetness. The polyoxometalate (POM) supported catalysts were prepared by incipient wetness (dry impregnation). According to this method, a sufficient quantity of TSA and PTA was dissolved in 0.1 N hydrochloric acid (HCl) aqueous solution respectively in order to achieve 20 wt% of POM loading (corresponding to 15 wt% of the W content), and then each solution (its volume being equal to the pore volume of the support oxide) was added dropwise to the porous support under constant mixing. The obtained catalysts were dried at 100 °C overnight and finally were stabilized by calcination at 200 °C for 4 h under static air.

Materials characterization

The physicochemical properties of the materials used in the present study were determined by means of inductively coupled plasma-optical emission spectroscopy (ICP-OES), N2 adsorption/desorption measurements, X-ray diffraction (XRD), scanning electron microscopy-energy dispersive spectroscopy (SEM-EDS), diffuse reflectance UV-vis spectroscopy (DRS UV-vis) and Fourier transform infrared spectroscopy (FT-IR), as well as FT-IR coupled with in situ pyridine adsorption for acidity characterization.

More specifically, the chemical composition of the catalysts was determined by ICP-OES analysis by using an Optima 4300 DV PerkinElmer spectrometer (USA). The solid samples were dissolved by using H2SO4 and HF aqueous solution, while external standards of the respective metals were used for their quantification.

N2 adsorption/desorption measurements at −196 °C were performed for the evaluation of the textural properties of the catalysts by the use of an Autosorb-1 Quantachrome apparatus. Prior to the analysis, the supported POM and the oxide catalysts were outgassed overnight at 150 and 250 °C respectively under vacuum. The BET surface area of each material was calculated by the Brunauer–Emmett–Teller (BET) equation, the total pore volume was estimated by the absorbed N2 at P/P0 = 0.99 and the pore size was determined by the Barrett–Joyner–Halenda (BJH) method.

Powder XRD measurements were performed for the determination of the crystallinity of feeds and catalysts. Experiments were performed on a SIEMENS D-500 diffractometer equipped with Cu Kα X-ray radiation and a curved crystal graphite monochromator operating at 40 kV and 30 mA; counts were accumulated in the 2θ range of 5–70° every 0.02° (2θ) with a counting time of 1 s per step. The cellulose crystal size in the different feeds was calculated by the method of peak separation with a peak resolution program using Scherrer's equation, and the crystalline index (CrI) was calculated by referring to the diffraction intensities of the crystalline region and amorphous region according to the following eqn (1):

image file: c9gc02622c-t1.tif(1)
where I002 is the maximum intensity of the (002) lattice diffraction at 2θ 22.5° and Iαm is the intensity of diffraction at 2θ 18.3°.81

Scanning Electron Microscopy (SEM) images were obtained on a JEOL 6300 microscope equipped with an X-ray energy dispersive spectroscopy (EDS) system for X-ray microanalysis (OXFORD ISIS 2000). Elemental mapping experiments were conducted on the flat surfaces of catalyst particles coated with carbon.

The diffuse reflectance UV-vis spectra (DRS UV-vis) were recorded under air at room temperature using a UV-3010 spectrometer (Hitachi) with a scanning rate of 60 nm min−1 and a scanning range of 500–190 nm. The sampler window was fused silica.

The FT-IR spectra of the parent heteropolyacids and the oxide supports, as well as of the supported POM catalysts, were obtained on a FT-IR PerkinElmer, model Spectrum 1, with a resolution of 4 cm−1, using KBr diluted powdered samples that were pressed into thin pellets.

The amount and type (Brønsted or Lewis) of the acid sites were evaluated by FT-IR spectroscopy combined with in situ adsorption of pyridine.31 The FT-IR spectra were recorded on a Nicolet 5700 FTIR spectrometer (resolution 4 cm−1) using the OMNIC software and a specially designed, heated, high-vacuum IR cell with CaF2 windows. The samples were finely ground in a mortar and pressed in self-supported wafers (15 mg cm−2). The wafers were outgassed in situ at 250 °C (supported POM catalysts in order to preserve the structure of POMs) or 450 °C (oxide supports) for 1 h under high vacuum (10−6 mbar) and a background spectrum was recorded at 150 and 100 °C, respectively. Adsorption/equilibration with pyridine vapors was then conducted at 150 °C (oxide supports) or 100 °C (supported POM catalysts) by adding pulses of pyridine for 1 h at a total cell pressure of 1 mbar. Spectra were recorded at 150 °C (or 100 °C), and at higher temperatures, i.e. in the range of 250–450 °C for oxide catalysts and 150–250 °C for supported POM catalysts, after further outgassing for 30 min at each temperature. The bands at 1515–1565 cm−1 (pyridinium ions) and 1435–1470 cm−1 (coordinated pyridine) were used to identify and quantify the Brønsted and Lewis acid sites, respectively, by adopting the molar extinction coefficients provided by Emeis.82

Cellulose conversion and product analysis

The catalytic conversion of cellulose was performed in a batch, stirred, autoclave reactor (C-276 Parr Inst., USA), under a N2 gas. A dispersion of cellulose or biomass in water and the catalyst were charged into the reactor. Once the desired temperature was reached, zero time was recorded and the reaction was allowed to proceed for a given time under continuous stirring. After completion of the reaction, the reactor was cooled rapidly and the liquid product was separated by vacuum filtration.

The reaction products were analyzed by ion chromatography (ICS-5000, Dionex, USA). The quantification was based on external calibration, using standard solutions of sugars and sugar alcohols (glucose, mannose, xylose, fructose, galactose, arabinose, rhamnose, sorbitol and mannitol), hydroxymethylfurfural (HMF) and organic acids (formic, acetic, glycolic, lactic, levulinic, propionic and butyric acid). The analysis of sugars was performed using a CarboPac PA1 (10 μm, 4 × 250 mm) column and precolumn (10 μm, 4 × 30 mm) connected to a pulsed amperometric detector (PAD). The eluent was 20 mM NaOH at a 0.6 ml min−1 flow rate and the total analysis time was 75 min. The analysis of the organic acids was performed on an AS-15 (9 μm, 4 × 250 mm) column and pre-column (9 μm, 4 × 30 mm) connected to a conductivity detector (CD). The eluent was 8 mM NaOH at a 1 ml min−1 flow rate and the total analysis time was 75 min.

The conversion of cellulose, the yields and the selectivity of the products (weight based) were calculated according to the following eqn (2), (3) and (4):

image file: c9gc02622c-t2.tif(2)
image file: c9gc02622c-t3.tif(3)
image file: c9gc02622c-t4.tif(4)

The hemicellulose, α-cellulose and lignin content of the biomass samples was determined by the well-known NREL methods (NREL/TP-510-42618 and NREL/TP-510-42619).83,84

The stability of the POM supported catalysts in terms of metal leaching was investigated by analyzing the liquid filtrate after the reaction for the presence of W (and Nb for Nb2O5 based catalysts) by means of ICP-OES analysis. Furthermore, the potential changes in the characteristic Keggin structure of POM ions were examined by performing FT-IR measurements on the recovered solid after washing with water and acetone, drying at 80 °C overnight and heating at 200 °C for 4 h under air. With regard to catalyst reuse, the dried catalyst after heating was tested by using a fresh cellulose mixture with water following the experimental procedure described above. This process was repeated at least three consecutive times, corresponding to four catalytic cycles.

Conflicts of interest

There are no conflicts to declare.


This project has received funding from the Hellenic Foundation for Research and Innovation (HFRI) and the General Secretariat for Research and Technology (GSRT), under grant agreement no. 287-128035/I2. The authors thank Professor D. Achilias and Dr L. Nalbandian for their assistance in conducting the FTIR and DRS UV-Vis analysis.


  1. P. Gallezot, Chem. Soc. Rev., 2012, 41, 1538–1558 RSC .
  2. A. A. Marianou, C. M. Michailof, A. Pineda, E. F. Iliopoulou, K. S. Triantafyllidis and A. A. Lappas, ChemCatChem, 2016, 8, 1100–1110 CrossRef CAS .
  3. R.-J. van Putten, J. G. de Vries, J. C. van der Waal, E. de Jong, C. B. Rasrendra and H. J. Heeres, Chem. Rev., 2013, 113, 1499–1597 CrossRef CAS PubMed .
  4. W. C. Conner, G. A. Tompsett, A. Fernández, G. W. Huber, R. Weingarten, Y. T. Kim, E. W. Hagaman, K. S. Han and J. A. Dumesic, J. Catal., 2013, 304, 123–134 CrossRef .
  5. Y. Román-Leshkov and M. E. Davis, ACS Catal., 2011, 1, 1566–1580 CrossRef .
  6. X. Tong, Y. Ma and Y. Li, Appl. Catal., A, 2010, 385, 1–13 CrossRef CAS .
  7. J. C. Serrano-Ruiz, R. Luque and A. Sepúlveda-Escribano, Chem. Soc. Rev., 2011, 40, 5266–5281 RSC .
  8. M. Dusselier, P. Van Wouwe, A. Dewaele, E. Makshina and B. F. Sels, Energy Environ. Sci., 2013, 6, 1415–1442 RSC .
  9. W. Deng, Y. Wang, Y. Wang, K. M. Gupta, H. Asakura, M. J. Hülsey, N. Yan, S. Zhang, P. J. Dyson, L. Liu, E. M. Karp, T. Tanaka, G. T. Beckham, J. Jiang and Y. Han, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, 5093–5098 CrossRef CAS PubMed .
  10. https://www.zionmarketresearch.com/sample/lactic-acid-market .
  11. F. H. Isikgor and C. R. Becer, Polym. Chem., 2015, 6, 4497–4559 RSC .
  12. S. Li, W. Deng, Y. Li, Q. Zhang and Y. Wang, J. Energy Chem., 2019, 32, 138–151 CrossRef .
  13. D. Esposito and M. Antonietti, ChemSusChem, 2013, 6, 989–992 CrossRef CAS PubMed .
  14. L. Li, F. Shen, R. L. Smith and X. Qi, Green Chem., 2017, 19, 76–81 RSC .
  15. H. Choudhary, S. Nishimura and K. Ebitani, Appl. Catal., B, 2015, 162, 1–10 CrossRef CAS .
  16. Y. Wang, W. Deng, B. Wang, Q. Zhang, X. Wan, Z. Tang, Y. Wang, C. Zhu, Z. Cao, G. Wang and H. Wan, Nat. Commun., 2013, 4, 1–7 Search PubMed .
  17. Z. Tang, W. Deng, Y. Wang, E. Zhu, X. Wan, Q. Zhang and Y. Wang, ChemSusChem, 2014, 7, 1557–1567 CrossRef CAS PubMed .
  18. X. Lei, F. F. Wang, C. L. Liu, R. Z. Yang and W. S. Dong, Appl. Catal., A, 2014, 482, 78–83 CrossRef CAS .
  19. F. F. Wang, C. L. Liu and W. S. Dong, Green Chem., 2013, 15, 2091–2095 RSC .
  20. W. Deng, P. Wang, B. Wang, Y. Wang, L. Yan, Y. Li, Q. Zhang, Z. Cao and Y. Wang, Green Chem., 2018, 20, 735–744 RSC .
  21. F. Chambon, F. Rataboul, C. Pinel, A. Cabiac, E. Guillon and N. Essayem, Appl. Catal., B, 2011, 105, 171–181 CrossRef CAS .
  22. N. Shi, Q. Liu, X. He, H. Cen, R. Ju, Y. Zhang and L. Ma, Bioresour. Technol. Rep., 2019, 5, 66–73 CrossRef .
  23. R. Yang, B. Zhao, C. Liu, H. Ren, W. Dong and H. Li, Res. Chem. Intermed., 2016, 42, 7199–7211 CrossRef .
  24. F. F. Wang, J. Liu, H. Li, C. L. Liu, R. Z. Yang and W. S. Dong, Green Chem., 2015, 17, 2455–2463 RSC .
  25. P. Wattanapaphawong, P. Reubroycharoen and A. Yamaguchi, RSC Adv., 2017, 7, 18561–18568 RSC .
  26. Y. Swesi, C. Nguyen, T. T. Ha Vu, F. Rataboul, M. Eternot, P. Fongarland and N. Essayem, ChemCatChem, 2017, 9, 2377–2382 CrossRef CAS .
  27. P. Wattanapaphawong, N. Mimura, A. Yamaguchi, O. Sato, P. Reubroycharoen and K. Sato, Catalysts, 2017, 7, 221 CrossRef .
  28. L. Yang, X. Yang, E. Tian, V. Vattipalli, W. Fan and H. Lin, J. Catal., 2016, 333, 207–216 CrossRef CAS .
  29. M. Xia, Z. Shen, S. Xiao, B. Peng, M. Gu, W. Dong and Y. Zhang, Appl. Catal., A, 2019, 583, 117126 CrossRef CAS .
  30. M. Thommes, K. Kaneko, A. V. Neimark, J. P. Olivier, F. Rodriguez-Reinoso, J. Rouquerol and K. S. W. Sing, Physisorption of gases, with special reference to the evaluation of surface area and pore size distribution (IUPAC Technical Report), 2015, vol. 87 Search PubMed .
  31. K. V. Avramidou, F. Zaccheria, S. A. Karakoulia, K. S. Triantafyllidis and N. Ravasio, Mol. Catal., 2017, 439, 60–71 CrossRef CAS .
  32. Y. Ren, B. Liu, Z. Zhang and J. Lin, J. Ind. Eng. Chem., 2015, 21, 1127–1131 CrossRef CAS .
  33. H. Jiao, X. Zhao, C. Lv, Y. Wang, D. Yang, Z. Li and X. Yao, Sci. Rep., 2016, 6, 1–9 CrossRef PubMed .
  34. J. He, Q. J. Li and Y. N. Fan, J. Solid State Chem., 2013, 202, 121–127 CrossRef CAS .
  35. V. S. Braga, I. C. L. Barros, F. A. C. Garcia, S. C. L. Dias and J. A. Dias, Catal. Today, 2008, 133–135, 106–112 CrossRef CAS .
  36. T. Kitano, T. Shishido, K. Teramura and T. Tanaka, J. Phys. Chem. C, 2012, 116, 11615–11625 CrossRef CAS .
  37. V. S. Braga, J. A. Dias, S. C. L. Dias and J. L. De Macedo, Chem. Mater., 2005, 17, 690–695 CrossRef CAS .
  38. I. Colicchio, F. Wen, H. Keul, U. Simon and M. Moeller, J. Membr. Sci., 2009, 326, 45–57 CrossRef CAS .
  39. V. Brahmkhatri and A. Patel, Ind. Eng. Chem. Res., 2011, 50, 13693–13702 CrossRef CAS .
  40. M. Jose Da Silva, A. L. Cardoso, F. De Lima Menezes, A. Mendes De Andrade, M. Gonzalo and H. Terrones.
  41. G. Marcì, E. García-López, M. Bellardita, F. Parisi, C. Colbeau-Justin, S. Sorgues, L. F. Liotta and L. Palmisano, Phys. Chem. Chem. Phys., 2013, 15, 13329–13342 RSC .
  42. L. R. Pizzio, C. V. Cáceres and M. N. Blanco, Appl. Catal., A, 1998, 167, 283–294 CrossRef CAS .
  43. E. I. Ross-Medgaarden and I. E. Wachs, J. Phys. Chem. C, 2007, 111, 15089–15099 CrossRef CAS .
  44. A. A. Marianou, C. M. Michailof, A. Pineda, E. F. Iliopoulou, K. S. Triantafyllidis and A. A. Lappas, Appl. Catal., A, 2018, 555, 75–87 CrossRef CAS .
  45. X. Li, K. Peng, Q. Xia, X. Liu and Y. Wang, Chem. Eng. J., 2018, 332, 528–536 CrossRef CAS .
  46. K. Nakajima, Y. Baba, R. Noma, M. Kitano, J. N. Kondo, S. Hayashi and M. Hara, J. Am. Chem. Soc., 2011, 133, 4224–4227 CrossRef CAS PubMed .
  47. K. Tanabe, Catal. Today, 2003, 78, 65–77 CrossRef CAS .
  48. M. Möller, P. Nilges, F. Harnisch and U. Schröder, ChemSusChem, 2011, 4, 566–579 CrossRef PubMed .
  49. C. Pinel, F. Rataboul, E. Guillon, V. Jollet, A. Cabiac, N. Essayem and F. Chambon, Green Chem., 2009, 11, 2052 RSC .
  50. Z. Srokol, A. G. Bouche, A. Van Estrik, R. C. J. Strik, T. Maschmeyer and J. A. Peters, Carbohydr. Res., 2004, 339, 1717–1726 CrossRef CAS PubMed .
  51. M. S. Holm, S. Saravanamurugan and E. Taarning, Science, 2010, 328, 602–605 CrossRef CAS PubMed .
  52. S. Wang and S. H. Liu, Appl. Mech. Mater., 2013, 411–414, 2141–2147 Search PubMed .
  53. S. Dutta, S. De and B. Saha, Biomass Bioenergy, 2013, 55, 355–369 CrossRef CAS .
  54. J. Macht, M. J. Janik, M. Neurock and E. Iglesia, Angew. Chem., Int. Ed., 2007, 46, 7864–7868 CrossRef CAS PubMed .
  55. J. Macht, M. J. Janik, M. Neurock and E. Iglesia, J. Am. Chem. Soc., 2008, 130, 10369–10379 CrossRef CAS PubMed .
  56. W. Deng, Q. Zhang and Y. Wang, Dalton Trans., 2012, 41, 9817–9831 RSC .
  57. K. I. Shimizu, H. Furukawa, N. Kobayashi, Y. Itaya and A. Satsuma, Green Chem., 2009, 11, 1627–1632 RSC .
  58. R. Réocreux, T. Jiang, M. Iannuzzi, C. Michel and P. Sautet, ACS Appl. Nano Mater., 2018, 1, 191–199 CrossRef .
  59. T. S. Mahadevan and S. H. Garofalini, J. Phys. Chem. C, 2008, 112, 1507–1515 CrossRef CAS .
  60. N. Legagneux, J. M. Basset, A. Thomas, F. Lefebvre, A. Goguet, J. Sá and C. Hardacre, J. Chem. Soc., Dalton Trans., 2009, 2235–2240 RSC .
  61. M. Kuzminska, T. V. Kovalchuk, R. Backov and E. M. Gaigneaux, J. Catal., 2014, 320, 1–8 CrossRef CAS .
  62. F. Su, L. Ma, Y. Guo and W. Li, Catal. Sci. Technol., 2012, 2, 2367–2374 RSC .
  63. J. Tian, C. Fang, M. Cheng and X. Wang, Chem. Eng. Technol., 2011, 34, 482–486 CrossRef CAS .
  64. S. Li, W. Deng, Y. Li, Q. Zhang and Y. Wang, J. Energy Chem., 2018, 6, 1–15 Search PubMed .
  65. A. Verardi, I. De, E. Ricca and V. Calabr, in Bioethanol, 2012, pp. 95–122 Search PubMed .
  66. M. Jarvis, Nature, 2003, 426, 611–612 CrossRef CAS PubMed .
  67. H. Zhao, J. H. Kwak, Y. Wang, J. A. Franz, J. M. White and J. E. Holladay, Energy Fuels, 2006, 20, 807–811 CrossRef CAS .
  68. Q. Zhang, M. Benoit, K. De Oliveira Vigier, J. Barrault, G. Jégou, M. Philippe and F. Jérôme, Green Chem., 2013, 15, 963–969 RSC .
  69. R. H. Newman, Solid State Nucl. Magn. Reson., 1999, 15, 21–29 CrossRef CAS PubMed .
  70. A. D. French and M. Santiago Cintrón, Cellulose, 2013, 20, 583–588 CrossRef CAS .
  71. Y. Yu and H. Wu, AIChE J., 2011, 57, 793–800 CrossRef CAS .
  72. C. K. Nitsos, K. A. Matis and K. S. Triantafyllidis, ChemSusChem, 2013, 6, 110–122 CrossRef CAS PubMed .
  73. Q. Jing and X. Lü, Chin. J. Chem. Eng., 2008, 16, 890–894 CrossRef CAS .
  74. B. Girisuta, L. P. B. M. Janssen and H. J. Heeres, Ind. Eng. Chem. Res., 2007, 46, 1696–1708 CrossRef CAS .
  75. Y. Matsumura, S. Yanachi and T. Yoshida, Ind. Eng. Chem. Res., 2006, 45, 1875–1879 CrossRef CAS .
  76. M. Paivi, L. S. Irina, S. Tapio and Y. M. Dmitry, Chem. Rev., 2014, 114, 1909–1971 CrossRef PubMed .
  77. F. S. Asghari and H. Yoshida, Ind. Eng. Chem. Res., 2006, 45, 2163–2173 CrossRef CAS .
  78. P. Calvini, A. Gorassini and A. L. Merlani, Cellulose, 2008, 15, 193–203 CrossRef CAS .
  79. B. Girisuta, L. P. B. M. Janssen and H. J. Heeres, Green Chem., 2006, 8, 701–709 RSC .
  80. K. G. Kalogiannis, S. Stefanidis, A. Marianou, C. Michailof, A. Kalogianni and A. Lappas, Waste Biomass Valorization, 2015, 6, 781–790 CrossRef CAS .
  81. S. Park, J. O. Baker, M. E. Himmel, P. A. Parilla and D. K. Johnson, Biotechnol. Biofuels, 2010, 3, 1–10 CrossRef PubMed .
  82. C. A. Emeis, J. Catal., 1993, 141, 347–354 CrossRef CAS .
  83. A. Sluiter, B. Hames, R. Ruiz, C. Scarlata, J. Sluiter, D. Templeton and D. C. Nrel, Determination of Structural Carbohydrates and Lignin in Biomass Determination of Structural Carbohydrates and Lignin in Biomass, 2012, vol. 2011 Search PubMed .
  84. A. A. Sluiter, R. Ruiz, C. Scarlata, J. Sluiter and D. Templeton, Determination of Extractives in Biomass: Laboratory Analytical Procedure (LAP), 2008 Search PubMed .


Electronic supplementary information (ESI) available. See DOI: 10.1039/c9gc02622c

This journal is © The Royal Society of Chemistry 2019