Glycerol utilization: solvent-free acetalisation over niobia catalysts

Abstract

With increasing biodiesel production, availability of glycerol is expected to increase. New processes are needed for converting this surplus glycerol to value-added chemicals. In this work, we used niobia catalysts for the liquid-phase acetalisation of glycerol without using any solvent. High conversions were achieved (∼80%), though water was present in the reaction system. The calcination temperature changed the strength as well as the nature of acidity of the samples. Samples with higher Brønsted acid strength exhibited higher catalytic performance. The results show that niobia is a water tolerant, reusable catalyst for glycerol acetalisation.

---

1. Introduction

Biodiesel production is accompanied by the formation of glycerol as a by-product. Its availability has tripled within the last ten years due to the increased biodiesel production.¹ Several countries target increased biofuel usage in future, hence the glycerol availability will still increase.² With the oversupply and the consequent drop in price, glycerol is expected to become a major platform chemical.³ The use of glycerol is limited now, being mainly confined to pharmaceuticals and cosmetics. Hence new processes are needed that can convert the surplus glycerol to value-added chemicals. In addition to taking advantage of the relatively low price of glycerol, this will also improve the economic viability of biodiesel manufacture.

Research efforts are gaining momentum for the synthesis of value added products from glycerol. A few studies on glycerol reactions have been reported that involve catalytic processes such as reforming, oxidation, hydrogenolysis, etherification, dehydration and esterification reactions.^1,4,5 A possible way of utilizing glycerol is through its condensation with aldehydes and ketones to acetals and ketals respectively.^6–12 Glycerol acetals and ketals may be used as fuel additives, surfactants and flavours.^6–12 When incorporated into standard diesel fuel, they have led to a decrease in particles, hydrocarbons, carbon monoxide and unregulated aldehyde emissions.^12,13 These products can act as cold flow improvers for use in biodiesel, also reducing its viscosity. This is important because there is a growing demand for new additives specifically for biodiesel that are biodegradable, non-toxic and renewable. The addition of these compounds to biodiesel improved the viscosity and also met the established requirements for flash point and oxidation stability.¹⁴

The main drawback of the glycerol acetalisation is the production of water. This weakens the acid strength of the catalyst and limits the glycerol conversion. The use of solvents such as benzene, toluene, petroleum ether or chloroform to increase the glycerol conversion into acetals or ketals has been described.⁷ However, this method is not very efficient and presents environmental problems. We report solvent-free liquid-phase glycerol acetalisation with acetone using niobia catalysts. Niobia (Nb₂O₅) has been used as a water-tolerant solid acid catalyst for various water-involving reactions.^15–17 The effect of calcination/pretreatment temperature of hydrated niobium oxide, acidity and other reaction parameters on the catalytic performance was investigated. The results of this study show that niobia acts as a stable active catalyst for glycerol acetalisation (Scheme 1), forming 2,2-dimethyl 4-hydroxymethyl-1,3-dioxolane (solketal) predominantly.

Scheme 1 Acetalisation of glycerol.

2. Experimental

Niobia sample (Niobia HY-340) was kindly provided by CBMM, Brazil. The composition of this batch is as follows: Nb₂O₅ 80.5 ± 5%, Fe 200 ppm max., free Cl 400 ppm max. and loss on ignition 20.5 ± 5%. The samples were calcined at different temperatures for 4 h in a muffle furnace in static air before use. The catalysts are represented as Nb₂O₅-xxx, where xxx denotes the calcination temperature in K (Table 1). The reactions were carried out in the liquid phase in a 50 ml glass reactor equipped with a condenser and a magnetic stirrer. Required amounts of glycerol (99+%, Alfa Aesar) and acetone (99+%, Alfa Aesar) were stirred with a specific mass of the catalyst in powder form. The molar ratio of glycerol to acetone was 1 : 1.5 unless otherwise stated. To monitor the reaction, samples of the reaction mixture were taken periodically and analysed by gas chromatography (Perkin Elmer Clarus 500) using a 50 m BP5 capillary column and an FID detector. In a typical reaction, 34 mmol of glycerol was reacted with 50 mmol acetone and 0.2 g catalyst. The N₂ adsorption–desorption isotherms were measured at 77 K on a Micromeritics ASAP-2000 after evacuation at 473 K for 5 h. Surface areas were calculated by the BET method. Powder X-ray diffraction patterns were collected on a Bruker powder X-ray Diffractometer (D5000) operated at 40 kV and 20 mA using Nickel filtered Cu Kα radiation (1.5406 Å). The system used for ammonia adsorption flow calorimetry has been described previously.^18,19 It is based on a Setaram 111 DSC with an automated gas flow and switching system, with a mass spectrometer (Hiden HPR20) to sample the downstream gas flow. The sample (20–30 mg) was held on a glass frit in a vertical silica sample tube and activated at 423 K under a dried nitrogen flow (5 ml min⁻¹) for five hours. After activation, the sample temperature was maintained at 423 K and 1 ml pulses of the probe gas (1% ammonia in nitrogen) at atmospheric pressure were injected at regular intervals into the carrier gas stream from a gas sampling valve. The ammonia concentration downstream of the sample was monitored continuously by mass spectroscopy. The pulse interval was chosen to ensure that the ammonia concentration in the carrier gas returned to zero to allow the DSC baseline to stabilise. The net amount of ammonia irreversibly adsorbed from each pulse was determined by comparing the MS signal with that recorded through a control experiment with a blank sample tube. The net heat released by each pulse was calculated from the thermal DSC curve (Table 1). Diffuse reflectance infrared Fourier transform (DRIFT) spectra of adsorbed pyridine were obtained on a Nicolet NEXUS FTIR spectrometer equipped with a controlled environment chamber (Spectra-Tech Inc., model 0030-101). Catalyst samples were diluted with KBr powder (10 wt% in KBr) and pre-treated at 150 °C/0.1 Torr for 1 h. The samples were then exposed to pyridine vapour at room temperature for 1 h, followed by pumping out at 150 °C/0.1 Torr for 1 h to remove the physisorbed pyridine. Then the DRIFT spectra of the adsorbed pyridine were recorded at room temperature.

Table 1 Textural properties of niobia samples

Catalyst calcination temperature/K	BET surface area/m^2 g^−1	Pore volume/cm^3 g^−1	Average pore diameter^a/nm	Average pore diameter^b/nm	Acid site concentration^c/mmol g^−1	Acid strength^d/kJ mol^−1
a Calculated from BJH adsorption. b From BJH desorption. c NH3 adsorbed with −ΔH^oads ≥80 kJ mol^−1. d Average heat of adsorption above 80 kJ mol^−1.
473	121.0	0.18	5.8	5.5	0.28	92.8
573	105.3	0.17	6.5	6.3	0.20	98.3
773	59.4	0.16	9.0	8.2	0.13	87.3
973	10.6	0.05	—	—	—	—

---

Atomistic simulation techniques used in this work are formulated within the framework of the Born model of the solids.²⁰ This comprises of long-range electrostatic and short-range interactions to describe the attractive van der Waals interactions and the dispersive London forces acting between the ions. The long-range Coulombic interactions converge slowly as a function of increasing ion separation. This is overcome by using the mathematical transformation of Ewald²¹ when the system is periodic in three dimensions and that of Parry²² when the system is only periodic in two dimensions (for example when studying surfaces). The short-range forces are described using simple, parameterised, pairwise potentials such as the Buckingham potential of the form Φ(r_ij) = A_ijexp(−r_ij/ρ_ij) − (C_ij/r⁶_ij), where Φ(r_ij) is the short range interaction energy and A_ij, ρ_ij and C_ij are parameters specific to each pairwise interaction and r_ij is the separation between atom i and atom j. Additionally, the effect of polarisability on the anion is incorporated using the shell model of Dick and Overhauser.²³ Here the ion is divided into a core containing the mass and the core electron charge and a shell containing the valence electron charge. These are connected by a harmonic spring. The polarisability of the model ion is then determined by the spring constant (k) and the charges of the core and the shell. We used the values of the potential parameters reported by Jackson et al.²⁴ and are shown in Table 2.

Table 2 Potential parameters used in this work

	A/eV	Ρ/Å	C/eV Å^6	C/eV Å^−2	Charges, e
Nb^5+(core)⋯O^2−(shell)	3023.184	0.300	0.0	—	Nb	5+
O^2−(shell)⋯O^2−(shell)	22 764	0.149	27.89	—	O (core)	0.077
O^2−(core)⋯O^2− (shell)	—	—	—	27.29	O (shell)	−2.077

---

All simulations were performed using the METADISE (minimum energy techniques applied to dislocation interface and surface energies) code which enables us to perform static, energy minimisations on a solid-state system.²⁵ In the METADISE program the crystal is regarded as a series of charged planes parallel to the surface and periodic in two dimensions. The basis of this approach is to calculate the total interaction energy, often called the lattice energy.

3. Results and discussion

Fig. 1 shows the XRD patterns of the Nb₂O₅ catalysts calcined at different temperatures. The samples heated to 473 and 573 K were essentially amorphous, with very broad peaks. For the sample calcined at 773 K, peaks corresponding to the orthorhombic (T) phase of niobium pentoxide were observed. Some amorphous material was still present in this sample (Fig. 1). At 973 K, well defined peaks corresponding to the characteristic reflections of T-Nb₂O₅ were observed, indicating complete crystallization. The niobia structure was simulated starting from the bulk T-Nb₂O₅ previously described by Kato and Tamura.²⁶ This structure has a space group symmetry of PBAM. From this, a pseudo cell was generated with no symmetry constraints where Nb lattice sites with less than 10% occupancy were removed. The lattice disorder of the remaining Nb, described by 50% occupied sites less than 1 Å were replaced by a singular occupied site representing an average of the two sites. The structure of the crystal was then optimised with no applied pressure in order to find the minimum energy configuration. During the optimisation, no symmetry constraints were applied. The relaxed structure has lattice parameters of a = 6.00, b = 27.58, c = 3.63 and α = β = γ = 90°, which differs from the original experimental structure by 3–8%. This is in good agreement, especially considering that the structure of T-Nb₂O₅ was not used in the derivation of the potential model and that no symmetry restraints were applied during the simulation. Fig. 2 shows the relaxed T-Nb₂O₅ structure, in which 8 Nb ions are present in distorted octahedra and the remaining 8 Nb ions occupy pentagonal bipyramidal sites. The oxide ions form a large open channel structure throughout the unit cell. Using this relaxed structure, a simulated X-ray powder diffraction pattern (Fig. 2) was generated using the methodology of Gaussian function as implemented in the GDIS-Package.²⁷ This pattern matches well with the experimental pattern, demonstrating that the experimental sample and the simulated niobia lattice have the same structure.

Fig. 1 X-Ray diffraction patterns of niobia samples as a function of calcination temperature.

Fig. 2 Simulated X-ray diffraction pattern (left) using the relaxed T-Nb₂O₅ structure (right).

Table 1 shows the textural properties of the Nb₂O₅ samples derived from the nitrogen physisorption isotherms. Both surface area (from 121 to 10.6 m² g⁻¹) and pore volume (from 0.18 to 0.06 cm³ g⁻¹) decreased with increasing calcination temperature. The average pore diameter increased with calcination. These changes may be correlated with amorphous to crystalline transformation. In general, surface area decreased with increasing crystallinity. The most crystalline orthorhombic (T) phase obtained by calcination at 973 K had the lowest surface area. The low angle XRD pattern of Nb₂O₅-473 (Fig. 3a) shows only one peak at around a 2θ value of 1.5°. The pattern resembles more that of disordered mesoporous materials than that of regular mesoporous materials.^28–31 The mesoporosity in these samples is due to the extra framework void space resulting from the intergrowth of small primary particles. The mesoporosity is confirmed by the N₂ adsorption isotherm (Fig. 3b), which can be classified as type IV.

Fig. 3 (a) Low angle XRD of Nb₂O₅-473 and (b) N₂ adsorption–desorption isotherms of Nb₂O₅-473 and Nb₂O₅-573.

The total acidity of the samples was evaluated by ammonia adsorption microcalorimetry (Fig. 4). The initial differential heat of NH₃ adsorption was the highest for Nb₂O₅-573 (the sample calcined at 573 K), indicating stronger acid sites than other samples. The heat of adsorption decreased with increasing calcination temperature. We have quantified acidity by estimating the total ammonia adsorbed with heat greater than 80 kJ mol⁻¹ and averaged the enthalpies of points above 80 kJ mol⁻¹ for average acid strength (Table 1). It is an arbitrary method, but provides an indication of concentration and strength.^32,33 Nb₂O₅-973 not only had weaker acidity but also the concentration of sites was lower (Fig. 4). Note that this sample is highly crystalline with a low surface area. So, as calcination temperature was increased, the sample transformed from the amorphous to the crystalline phase, accompanied by a loss in surface area and a decrease in number and strength of acid sites.

Fig. 4 NH₃ adsorption microcalorimetry data of niobia samples calcined at different temperatures.

The nature of acid sites (Brønsted or Lewis) was determined by DRIFT spectroscopy of adsorbed pyridine (Fig. 5). The presence of Brønsted acid sites is indicated by the band at ∼1540 cm⁻¹ characteristic of the pyridinium ion.³⁴ Lewis acid sites are detected by observing the band at 1450–1460 cm⁻¹ attributed to the complex between pyridine and the Lewis site. At low calcination temperatures, samples had both Lewis and Brønsted acid sites. In contrast, Nb₂O₅-973 exhibited only Lewis acidity.

Fig. 5 DRIFT spectra of adsorbed pyridine: (a) Nb₂O₅-973, (b) Nb₂O₅-473 and (c) Nb₂O₅-773.

Fig. 6 shows the catalytic performance of niobia samples. These acetalisation reactions were conducted at 343 K, with a glycerol : acetone molar ratio of 1 : 1.5 using 6.4 wt% of catalyst (wrt glycerol) without any solvent. The activities were influenced significantly by the calcination temperature. The samples pre-treated at lower temperatures were most active while the sample calcined at 973 K was least active. For the samples pre-treated at 473 and 573 K, more than 70% glycerol was converted after 2 h. For the samples calcined at higher temperatures, initial conversions were lower but increased over extended reaction time. Initial activities (TOF, s⁻¹) were calculated from the glycerol kinetic curve and taking ammonia adsorbed with heat greater than 80 kJ mol⁻¹ for acid site concentration. The values are 4.7, 8.0 and 5.7 for niobia calcined at 473, 573 and 773 K. Comparison with NH₃ adsorption microcalorimetry data (Fig. 4) shows that the activity of the catalysts is in accordance with their acidity, indicating that higher acidity is favourable for higher catalytic activity. A reaction without any catalyst gave ∼15% glycerol conversion at a temperature of 343 K after 6 h.

Fig. 6 Glycerol conversion (left) and selectivity to 1a (right) using niobia samples calcined at different temperatures. 6.4 wt% catalyst wrt glycerol; reaction temperature = 343 K.

Glycerol acetalisation can produce two isomers, 5- and 6-membered cyclic acetals 1a and 1b, respectively (Scheme 1), with the simultaneous liberation of water. In our experiments, all catalysts exhibited higher selectivity to the 5-membered cyclic acetal 1a. Selectivity preference towards 5-membered acetal was observed previously.⁶ It is proposed that dehydration of the acetone/glycerol hemiketal yields a tertiary carbenium ion, followed by a rapid nucleophilic attack of the secondary hydroxyl group to form the five-membered ring ketal (Scheme 2).⁶ The acid strength had an obvious effect on the selectivity. The selectivity to 5-membered cyclic acetal was higher when the catalyst acidity was stronger. The sample pre-treated at 573 K, which possesses stronger acid sites, converted around 80% of glycerol with a selectivity towards 1a of ∼92% after 6 h at 343 K, indicating the usefulness of this material for glycerol acetalisation (Fig. 6). Though water is produced during the reaction, high conversions were achieved indicating the tolerance of niobia acid sites towards water. Based on the present results, it is not so straightforward to correlate the nature of acid sites (Brønsted/Lewis) with the activity and selectivity.

Scheme 2 Tentative mechanism for the formation of 5-membered cyclic acetal from glycerol and acetone.

The reaction occurs significantly during the first few hours itself as illustrated by the data obtained for Nb₂O₅-473 at 323 K (Fig. 7). Glycerol conversion was ∼15% after 15 minutes of reaction, increased rapidly for around 2 h, and then more or less remained constant with reaction time. The selectivity to 5-membered acetal was high from the beginning of the reaction and remained so with reaction time.

Fig. 7 Glycerol conversion and selectivity to 1a during initial period of reaction; catalyst = Nb₂O₅-473; reaction temperature = 323 K; glycerol : acetone = 1 : 3 (molar); 6.4 wt% catalyst wrt glycerol.

The catalytic performance was influenced by the reaction temperature (Fig. 8). 20% glycerol was converted after 1 h at 303 K using sample calcined at 573 K, which increased to ∼70% after 20 h of reaction. Upon increasing the reaction temperature to 323 K, the rate increased and ∼68% glycerol conversion was observed at 1 h. However, further increase in reaction temperature to 343 K did not seem to increase the rate a great deal. If the higher reaction times are affordable, the reaction may be conducted at temperatures even close to room temperature. Higher temperatures favour 5-membered acetal (1a) (Fig. 8). Nevertheless, 1a was the major product even at 303 K.

Fig. 8 Glycerol conversion at different reaction temperatures. Catalyst = Nb₂O₅-573; 6.4 wt% catalyst wrt glycerol. The dark and open symbols indicate conversion of glycerol and selectivity to 1a respectively.

Fig. 9 shows the catalytic performance as a function of catalyst amount. Higher catalyst amounts increased the conversion, as expected. It is particularly interesting that low catalyst amounts are effective for the reaction even at low temperature, indicating the promising performance of niobia for glycerol acetalisation.

Fig. 9 Catalytic performance as a function of catalyst amount. The percentage weights of catalyst wrt glycerol are indicated in the figure. The dark and open symbols indicate conversion of glycerol and selectivity to 1a respectively. Catalyst = Nb₂O₅-573; reaction temperature = 323 K.

To check the reusability, four consecutive reaction cycles were performed using Nb₂O₅-573 K. After each cycle, the catalyst was filtered, washed with acetone and water and calcined at 573 K. The results were similar to the original reaction and no significant deactivation was observed, confirming the reusability of the catalyst (Fig. 10).

Fig. 10 Catalyst reuse studies. After each cycle, catalyst was filtered, washed with acetone, water and calcined at 573 K. Catalyst = Nb₂O₅-573; reaction temperature = 343 K, time = 6 h.

4. Conclusions

In summary, our results show that high conversions can be achieved for glycerol acetalisation using niobia catalysts in relatively short times under suitable experimental conditions. High selectivity to solketal was observed. No solvent is necessary which makes the system environmentally benign. Acid strength of niobia depends on the calcination temperature and higher acid strength favours higher catalytic performance. The catalyst can be reused without deactivation, hence making it an ideal choice for the acetalisation.

Acknowledgements

We acknowledge CBMM, Brazil for providing niobium oxide.

References

1. C.-H. Zhou, J. N. Beltramini, Y.-X. Fana and G. Q. Lu, Chem. Soc. Rev., 2008, 37, 527; A. Behr, J. Eilting, K. Irawadi, J. Leschinski and F. Lindner, Green Chem., 2008, 10, 13.
2. Official Journal of the European Union, L140/6, 05-06-2009.
3. Top value added chemicals from biomass, ed. T. Werpy and G. Petersen, US Department of Energy (USDOE), vol. 1, 2004.
4. M. Pagliaro, R. Ciriminna, H. Kimura, M. Rossi and C. D. Pina, Angew. Chem., Int. Ed., 2007, 46, 4434.
5. N. R. Shiju, D. R. Brown, K. Wilson and G. Rothenberg, Top. Catal., 2010, 53, 1217.
6. C. X. A. da Silva, V. L. C. Gonçalves and C. J. A. Mota, Green Chem., 2009, 11, 38.
7. J. Deutsch, A. Martin and H. Lieske, J. Catal., 2007, 245, 428.
8. (a) B. Delfort, I. Durand, A. Jaecker, T. Lacome, X. Montagne and F. Paille, Fr Patent, 2 833 607 A1, 2003; (b) G. Hillion, B. Delfort and I. Durand, Int Patent, 2005/093 015 A1, 2005.
9. A. Piasecki, A. Sokolowski, B. Burczyk and U. Kotlewska, J. Am. Oil Chem. Soc., 1997, 74, 33.
10. (a) M. J. Climent, A. Veltry and A. Corma, Green Chem., 2002, 4, 565; (b) M. J. Climent, A. Corma and A. Veltry, Appl. Catal., A, 2004, 263, 155.
11. P. Ferreira, I. M. Fonseca, A. M. Ramos, J. Vital and J. E. Castanheiro, Appl. Catal., B, 2010, 98, 94.
12. G. Vicente, J. A. Melero, G. Morales, M. Paniagua and E. Martín, Green Chem., 2010, 12, 899.
13. B. Delfort, I. Durand, A. Jaecker, T. Lacome, X. Montagne and F. Paille, US Patent, 6890364, 2005.
14. E. García, M. Laca, E. Pérez, Á. Garrido and J. Peinado, Energy Fuels, 2008, 22, 4274–4280.
15. I. Nowak and M. Ziolek, Chem. Rev., 1999, 99, 3603.
16. K. Tanabe, Catal. Today, 2003, 78, 65.
17. M. Ziolek, Catal. Today, 2003, 78, 47.
18. N. R. Shiju, M. AnilKumar, W. F. Hoelderich and D. R. Brown, J. Phys. Chem. C, 2009, 113, 7735.
19. N. R. Shiju, H. M. Williams and D. R. Brown, Appl. Catal., B, 2009, 90, 451.
20. M. Born and K. Huang, Dynamical Theory of Crystal Lattices, Oxford University Press, 1954.
21. P. Ewald, Ann. Phys. (Leipzig), 1921, 369, 253.
22. D. E. Parry, Surf. Sci., 1975, 49, 433.
23. B. G. Dick and A. W. Overhauser, Phys. Rev., 1958, 112, 90.
24. M. Kilo, R. A. Jackson and G. Borchardt, Philos. Mag., 2003, 83, 3309.
25. G. W. Watson, E. T. Kelsey, N. H. de Leeuw, D. J. Harris and S. C. Parker, J. Chem. Soc., Faraday Trans., 1996, 92, 433.
26. K. Kato and S. Tamura, Acta Crystallogr., Sect. B: Struct. Crystallogr. Cryst. Chem., 1975, 31, 673.
27. S. Fleming and A. Rohl, Z. Kristallogr., 2004, 220, 580.
28. E. Prouzet and T. J. Pinnavaia, Angew. Chem., Int. Ed. Engl., 1997, 36, 516.
29. Z. Zhang and T. Pinnavaia, J. Am. Chem. Soc., 2002, 124, 12294.
30. K. Sivaranjani and C. S. Gopinath, J. Mater. Chem., 2011, 21, 2639.
31. D. G. Kulkarni, A. V. Murugan, A. K. Viswanath and C. S. Gopinath, J. Nanosci. Nanotechnol., 2009, 9, 371.
32. P. F. Siril, N. R. Shiju, D. R. Brown and K. Wilson, Appl. Catal., A, 2009, 364, 95.
33. A. Auroux, Top. Catal., 1997, 4, 71.
34. K. Tanabe, in Catalysis Science and Technology, ed. J. R. Anderson and M. Boudart, Springer, 1981, ch. 5.

---