Template LAAR 2021

Abstract-- The hydrothermal conversion of a multiple sugars solution (xylose, arabinose and glucose) was studied using high surface ZrO₂-based catalysts to produce lactic acid. The catalysts were synthesized by template-assisted sol-gel method and alternative template removal procedures. Samples obtained by template extraction with solvent reported a specific surface area of 323 m²/g, a pore volume of 0.21 cm³/g and a total amount of acid sites of 359 µmol/g. An experimental design was performed to model lactic acid and by-products obtaining around the selected catalyst. The response surface methodology (RSM) was used to optimize the main operating conditions. A mass selectivity of 35.1 % towards lactic acid was found at 183 °C and 212 min. In addition, the reaction mechanisms involved were proposed based on intermediates and degradation compounds detected by HPLC analysis.

Keywords-- experimental design, biomass, xylose, arabinose, glucose

I. INTRODUCTION

Recently, the lignocelullosic biomass has become one of the most prominent alternatives for replacing the fossil resources in fuels and chemicals manufacture (Davis et al., 2020). Lignocellulose is an abundant and low-cost resource mainly obtained from agricultural and forest wastes. However, its valorization is a complex process since it is composed of different amounts of high-molecular weight polymers such as lignin, cellulose and hemicellulose (Tursi, 2019). Thus, the first step consist on depolymerization through acid, alkaline or enzymatic hydrolysis to obtain their monomeric water-soluble compounds (Aspromonte et al., 2019). These monomers are 5-carbon sugars such as xylose and arabinose, and 6-carbon sugars such as glucose, mannose and galactose. Subsequently, the catalyzed hydrothermal conversion of sugars is the most suitable process to produce high-value added molecules such as sorbitol, xylitol, furfural, hydroxymethylfurfural (HMF), lactic, levulinic and succinic acids (Biddy et al., 2016). Among them, lactic acid is an important bio-product highly demanded in cosmetic, pharmaceutical and food industry. In addition, it is the chemical precursor for bio-plastics, bio-solvents and acrylic acid synthesis (Corma Canos et al., 2007).

While about the 90 % of world production of lactic acid is performed by biological process (Alexandri et al., 2019), the research found in literature demonstrated that catalyzed hydrothermal conversion of sugars might be an effective way to obtain reasonable yields. The majority of process uses alkaline homogeneous catalysts like NaOH, Ca(OH)₂(Yan et al., 2010) and Ba(OH)₂ (Esposito and Antonietti, 2013) at high temperatures (250 – 300 °C) to reach yields in the range of 20 – 50 %. Regarding heterogeneous catalysts, various Lewis acid solids such as γ-Al₂O₃ (Kosri et al., 2021), Nb₂O₅ (Cao et al., 2017) and Sn-Beta zeolites (Xia et al., 2018), and amphoteric metal oxides such as ZrO₂ (Yang et al., 2015), TiO₂ (Pattnaik et al., 2021) and ZnO (Paulino et al., 2018), were tested in a wide range of operating conditions, demonstrating high water-tolerance and satisfactory selectivities.

In this work, high surface zirconia-based catalysts were synthesized by means of controlled sol-gel synthesis to be evaluated in the hydrothermal conversion of sugars towards lactic acid. A multiple sugars solution comprising xylose, arabinose and glucose, was used as feedstock to represent a real lignocellulosic biomass hydrolysate. The textural, crystalline and acidic properties of synthesized catalysts were determined to establish a relation with their catalytic performance. Finally, an experimental design was performed and analyzed with response surface methodology (RSM) around the most promising catalyst to found the optimal operating conditions for lactic acid production.

II. METHODS

A. Synthesis of catalysts

Zirconia-based catalysts were synthesized by the template assisted sol-gel method. Zirconium (IV) 1-propoxide (C₁₂H₂₈O₄Zr) was used as zirconia precursor, isopropanol as synthesis medium and non-ionic copolymer Pluronic P123 (PEG₂₀PPG₇₀PEG₂₀) as template.

First, 10 g of template were dissolved in 50 ml of isopropanol under vigorous stirring at room temperature. After complete dissolution, the vessel was sealed and 8 g of zirconium 1-propoxide were added with a syringe. After stirring for 1 h, hydrochloric acid was added dropwise under vigorous stirring until the mass ratios HCl/Zr = 0.04 and H₂O/Zr = 0.80 were reached. A few minutes later, gelation took place and wet gel was dried at 60 °C for 72 h and then at 110 °C for 24 h.

Template was removed from fresh catalysts by three different techniques: i) calcination in airflow at 400 °C for 4 h (1 °C/min); ii) refluxing extraction in ethanol at 80 °C for 8 h and recovering by centrifugation (three repetitions); and iii) refluxing extraction and subsequent calcination. Samples were denoted as ZC, ZE and ZEC, respectively.

B. Physicochemical characterization

Physicochemical properties of synthesized catalysts were determined by N₂adsorption-desorption at -196 °C, X-Ray Diffraction (XRD), Fourier Transform Infrared Spectroscopy (FTIR) and potentiometric titration (PT).

N₂ adsorption-desorption isotherms were obtained in a Micromeritics ASAP 2020 analyzer. Samples were previously degassed at 150 °C during 8 h under vacuum. Specific surface area (S_BET) was calculated by BET theory and total pore volume (V_P) by adsorption single point at a relative pressure of 0.975. Pore size distribution (PSD) and pore diameter (D_P) were determined with BJH method on adsorption branch using Harkins-Jura thickness curve.

XRD was performed in a Shimadzu XD-D1 instrument with CuKα monochromatic radiation (1.5406 Å) in 2θ range of 5° - 70°. Crystallite average size (D_hkl) was approximated by Debye-Scherrer equation with a shape factor of 0.9. FTIR was carried out with a Shimadzu 8101M IR Prestige-21 system in the wavenumber range 400 – 4000 cm^-1.

Acidity assessment was carried out with PT. A 0.01 N n-butylamine solution in acetonitrile was used as titrant at a flow rate of 0.1 mL/min. The potentiograms were recorded at 10 s intervals with a Metrohm 913 pH-meter, using a reference electrode of LiCl saturated in ethanol. The total amount of acid sites was calculated by the first derivative method.

C. Catalytic evaluation

Reaction tests were performed in an AISI 304 stainless steel batch reactor with a capacity of 45 mL. Reactant solution consisted of glucose 0.01 M, arabinose 0.02 M and xylose 0.04 M. In a typical experiment, a desired amount of catalyst and 18 mL of the solution were added into reactor. Then, it was sealed, purged several times and pressurized with high purity N₂. Next, it was heated up to the selected temperature (5 °C/min) under magnetic stirring (1500 rpm). Once reaction time concluded, it was quickly cooled down in a water bath. Wasted catalyst was separated from products mixture by vacuum filtration.

Products and remaining reactants were analyzed by High Performance Liquid Chromatography (HPLC) using a Perkin Elmer system provided with an Aminex HPX-87H column (Bio-Rad). Mobile phase was a 5 mM sulfuric acid solution pumped at 0.6 mL/min. Column temperature was maintained at 65 °C. Samples were previously filtered (0.45 µm) and diluted 1/5 in mobile phase. Sugars (xylose, arabinose and glucose) were quantified using the refractive index detector (RI), while organic acids (lactic, glycolic, formic, acetic and levulinic) and furans (HMF and furfural) were quantified using the UV-Vis detector set at 210 and 260 nm, respectively. Conversion, mass yields and selectivities were calculated as follows:

, (1)

, (2)

, (3)

where, X: sugars conversion, C concentration (g/L) of remaining glucose (G), arabinose (A) and xylose (X), C⁰: initial sugar concentration (g/L), Y_P: mass yield of product P, C_P: concentration (g/L) of product P, S_P: mass selectivity of product P = lactic (LA), glycolic (GA), acetic (AA), formic (FA) and levulinic acids (LeA), HMF and furfural (F).

D. Response surface methodology (RSM)

The response surface methodology (RSM) was employed to optimize the most relevant operating conditions and maximize selectivity towards lactic acid (Montgomery, 2013). Experimental data was collected through an extended Box-Behnken design of experiments around 4 variable factors, with 30 runs and 3 replications at central point. The selected factors were temperature (A), reaction time (B), catalyst mass (C) and initial pressure (D). Central design point was fixed to A = 180 °C, B = 90 min, C = 70 mg and D = 30 bar. The designated levels for factorial points settings were A = 140, 160, 180 and 200 °C; B = 30, 60, 90, 120, 150, 240 min; C = 35, 70 and 100 mg; and D = 10, 30 and 50 bar.

Data processing was done with open-source software R studio. The function model coefficients were estimated with multiple linear regression fit (MLRF) by least squares method. The statistical significance of the model was decided with an Analysis of Variance (ANOVA) adopting a probability value (p-value) lower than 0.05. The fitting reliability was determined with the multiple factors determination coefficient (R²) and the lack of fit (LOF). Finally, the stationary point was located by solving the partial derivatives of the empirical model equation.

III. RESULTS

A. Physicochemical properties of catalysts

Physicochemical characterization results of synthesized catalysts are summarized in Table 1. Textural properties, such as specific surface area (S_BET), total pore volume (V_P) and average pore diameter (D_P) were determined from N₂ adsorption-desorption isotherms (Fig. 1). Crystalline properties, such as crystalline phase and average crystallite size (D_hkl) were obtained from X-Ray diffractograms (Fig. 2). In addition, acidic properties, such as acid sites total amount (AS_A), density (AS_D) and strength (E_max), were calculated from n-butylamine titration potentiograms (Fig. 3).

As shown in Fig. 1, the three materials exhibited a IV-type isotherm, according to the IUPAC classification with a wide pore size distribution (Sing, 1985). For extracted (ZE) and extracted-calcined (ZEC) catalysts, the

Table 1. Physicochemical properties of synthesized catalysts.

Catalyst	Template removal method	Textural^a			Crystallinity^b		Acidity^c
Catalyst	Template removal method	S_BET (m²/g)	V_P (cm³/g)	D_P (nm)	Phase	D_hkl (nm)	AS_A (µmol/g)	AS_D (µmol/m²)	E_max (mV)
ZE	Extraction	323	0.21	1.9	-	-	359	1.111	183
ZC	Calcination	108	0.09	3.7	Tetragonal	6.8	49	0.454	82
ZEC	Combined	118	0.07	2.3	Tetragonal	9.7	34	0.288	90

a Determined from N₂ adsorption-desorption isotherms at -196 °C.

b Determined from XRD measures.

c Determined from potentiometric titration with n-butylamine curves

Figure 1. N₂ adsorption-desorption isotherms at -196 °C of A) extracted, B) calcined and C) extracted/calcined catalysts and respective pore size distributions (inset plots).

PSD is centered on 2 nm and covers the range of < 1 to 5 nm, suggesting the contribution of both micropores and mesopores. On the other hand, the calcined catalyst (ZC) shows a higher contribution of mesopores with a larger H2-type hysteresis loop, and a PSD center at 3.7 nm. ZE catalyst reported the highest specific surface area and total pore volume as seen in Table 1. Whereas, the thermal treatment on ZC and ZEC catalysts reduced the S_BETto about one third and the V_P to less than a half.

Figure 2 presents the XRD results of synthesized materials. Calcined catalysts (ZC and ZEC) show patterns coincident with the tetragonal phase of zirconia (ZrO₂). The average crystallite size (D_hkl) was calculated by Scherrer method from the most intense peak located at 2θ = 30.3° (Table 1). Other authors have also reported that sol-gel synthesis allows obtaining a metastable tetragonal ZrO₂ phase with D_hkl below 10 nm, at temperatures approaching 400 °C (da Silva and Vasconcelos, 2019). Instead, the ZE catalyst show a weak broad peak at 2θ range from 20° to 40°, probably attributed to very small polydispersed nanocrystallites formed on hydrated zirconia (Deshmane and Adewuyi, 2012)

PT curves presented in Fig. 3 show a similarity between both calcined materials and a notably different behavior for ZE catalyst. The maximum potential difference reached (E_max) is considered an acidic strength parameter. Some authors classified the acid sites corresponding to a value of E_max > 100 mV as very strong acid sites (Gorsd et al., 2018). As observed in Table 1, ZE reached a E_max of 183 mV, double than calcined materials. In the same way, the total amount of acid sites in ZE resulted more than 7 times higher.

It is highlighted that the solvent extraction method of template allows preserving the porous structure and high specific surface area of synthesized hydrated zirconia, as well as the enhanced acidic properties. On the other hand, the calcination at 400 °C cause the water elimination and crystallization, reducing the surface area and removing most of the acid sites. Figure 4 shows the FTIR spectra of synthesized zirconia-based material without template removal The disappearance of the intense signals in the regions 2800 - 3000 cm^-1 and 1000 - 1500 cm^-1, corresponding to the C-H, C-C and C-O bonds of ethylene and propylene oxides, indicate that both methods were suitable for complete template removal. It can be noted that ZE catalyst preserves the signals corresponding to –OH vibrations at around 3430 and 1630 cm^-1, while in ZC catalyst they are significantly reduced.

B. Catalytic performance

The synthesized zirconia-based catalysts were evaluated in the hydrothermal conversion of a xylose, arabinose and glucose solution with a representative composition of a wheat bran hydrolysate. Moderate operating conditions (180 °C, 90 min, 70 mg of catalyst and 10 bar of N₂) were adopted to select the best performance catalyst and subsequently carry out a design of experiment to maximize the lactic acid production.

Figure 5 presents the possible reaction pathways for hydrothermal conversion of xylose, the majority sugar in feedstock solution. Direct dehydration produces furfural, and subsequent rehydration yields formic and levulinic acids (Oefner et al., 1992). In addition, furans such as furfural can undergo polycondensation to form solid

Figure 2. X-Ray diffractograms from synthesized catalysts.

Figure 3. Potentiometric titration with n-butylamine curves of synthesized catalysts.

Figure 4. FTIR spectra of synthesized catalysts a) without template removal, b) extracted and c) calcined.

humins. On the other hand, xylose isomerizes to xylulose, and can undergo retro-aldol condensation, strongly promoted by Lewis acid sites or alkaline mediums (Holm et al., 2012). The retro-aldol reaction cleaves the ketose leaving the formation of glyceraldehyde and glycolaldehyde. Glycoladehyde may oxidize to obtain glycolic acid, while glyceraldehyde can interconvert to dihydroxyacetone (DHA). Finally, DHA can be more easily converted

Table 2. Catalytic performance comparison at 180 °C.

Catalyst^a	X^b	S_LA^b	S_GA^b	S_HMF^b	S_F^b
ZE	90.2	30.9	10.6	1.3	13.1
ZC	65.6	15.1	8.3	2.3	22.9
ZEC	58.8	15.3	9.1	1.9	24.2

^a 70 mg catalyst/18 ml reactant solution, 90 min, 10 bar.

^b Sugars conversion and main products mass selectivities (%). LA: lactic acid, GA: glycolic acid, HMF: hydroxymethylfurfural, F: furfural.

to lactic acid, by means of a sequence of reactions. First, 2-hydroxypropenal is obtained by dehydration and subsequent tautomerization leaves the formation of pyruvaldehyde. Last, lactic acid is produced through internal Canizzaro reaction. Arabinose, another pentose like xylose, can follow the same reaction pathways. Analogously, glucose can dehydrate directly to the furan 5-hydroxymehtylfurfural (HMF) or undergo isomerization to the 6-carbon ketose, fructose. Fructose can be cleaved to glyceraldehyde and DHA.

Table 2 shows the sugars conversion and main products selectivities obtained in catalytic evaluation. Lactic, glycolic acids, furfural and HMF were the main products detected by HPLC analysis. Also, acetic, formic and levulinic acids were detected as minor by-products, but quantities are not shown in Table 2. In accordance with physicochemical characterization, ZE catalysts stood out over the calcined catalysts, reaching the higher conversion (90.2 %). ZE also resulted the most selective towards lactic acid production with 30.9 % over a theoretical mass selectivity of 66.5 %. On the other hand, ZC and ZEC catalysts attained furans selectivities, i.e. the sum of HMF and furfural, of 25.2 and 26.1 %, respectively.

C. Optimization by RSM

The ZE catalyst was selected according to the catalytic evaluation results to carry out a design of experiments and find the maximum selectivity towards lactic acid with the response surface methodology (RSM). Starting from the central operating conditions in which all synthesized catalysts were compared, 30 experiments were performed varying the temperature (A), reaction time (B), catalyst mass (C) and initial pressure (D).

Lactic acid selectivity (S_LA) and dehydration products selectivity (S_DH) were the responses modeled to study the target and side reactions behavior with the ZE catalyst. S_DH was calculated from Eq. 4 and involves furans and related degradation compounds.

. (4)

where, S_NI is the selectivity towards no identified compounds and it is assumed to include humins, water generated from dehydration reactions and other furans derived from oxidative reactions. S_NI was calculated as 100 less the summation of all identified products selectivities, including glycolic (S_GA) and acetic acids (S_AA).

The response fit models acceptability was decided based on three statistical parameters as shown in Table 3. On the other hand, lack of fit (LOF) resulted non-influential with a p-value above 0.05. The goodness of fit was defined with R² coefficient and resulted greater than 0.9

Figure 5. Hydrothermal conversion of xylose, simplified reaction pathways. 1) Dehydration, 2) Rehydration, 3) Isomerization, 4) Retro-aldol condensation, 5) Oxidation, 6) Dehydration, 7) Tautomerization and 8) Canizzaro.

Table 3. ANOVA and statistical parameters for S_LA fit model.

for both models. The following are the empirical equations obtained for fitting models:

. (5)

. (6)

It can be observed that both models depend on temperature (A), reaction time (B) and catalyst mass (C). Linear, interactional, quadratic and an additional cubic term in S_LA model were sufficient to explain the variability of data. However, initial pressure (D) was not an influential parameter in either case.

Figure 6 shows the response surface plot for lactic acid selectivity fitting model as a function of temperature and reaction time, and at constant catalyst mass. The optimum operating conditions corresponding to the maximum selectivity were mathematically determined by solving the partial derivatives of Equation 5. The stationary point was located at A = 183 °C, B = 212 min and C = 71 mg, corresponding to a S_LA= 35.1 %.

Regarding the dehydration products selectivity, a stationary point was also calculated from Equation 6. However, the coordinates found correspond to a minimum selectivity value, as shown in contour lines presented in Fig. 7. The respective conditions were located at A = 151 °C, B = 142 min and C = 100 mg, and corresponding to a S_DH = 44.3 %.

Figure 6. Response surface of S_LA fit model as a function of temperature (A) and reaction time (B), at constant catalyst mass C = 71 mg.

Figure 7. Contour lines of S_DH (%) as a function of temperature (A) and reaction time (B), slice at catalyst mass C = 100 mg.

The highest lactic acid selectivity was found at a temperature above 180 °C but the by-products obtained by the competitive reaction (direct dehydration) were relevant at these conditions. Since furans and their degradation products are not intermediates in the formation of lactic acid, there is not possible to improve the production by operating at this temperature, even if the reaction time is extended beyond the design limits. To enhance the selectivity towards lactic acid, it is desired to work at lower temperatures (150 – 170 °C) and reaction times longer than 240 min.

IV. CONCLUSIONS

ZrO₂ based catalysts were synthesized by template assisted sol-gel method. Extraction with ethanol, calcination at 400 °C and their combination were studied as alternative template removal methods. It was demonstrated that extraction avoids zirconia crystallization, which causes porous structure collapse and hydroxyl groups elimination. Thus, the extracted catalyst exhibited the highest specific surface area, pore volume, strength and amount of acid sites.

The superior physicochemical properties of the extracted catalyst allowed reaching the highest conversion and selectivity towards lactic acid in sugars hydrothermal conversion. The optimal operating conditions determined by an experimental design were 183 °C, 212 min and 3.94 mg/ml of catalyst, and reached the maximum lactic acid mass selectivity of 35.1 % over a theoretical of 66.5 %.

REFERENCES

Alexandri, M., Schneider, R., Mehlmann, K. and Venus, J. (2019) Recent advances in d-lactic acid production from renewable resources. Food Technology and Biotechnology. 57, 293–304.

Aspromonte, S.G., Romero, A., Boix, A.V. and Alonso, E. (2019) Hydrolysis of cellulose to glucose by supercritical water and silver mesoporous zeolite catalysts. Cellulose. 26, 2471–2485.

Biddy, M.J., Scarlata, C. and Kinchin, C. (2016) Chemicals from Biomass: A Market Assessment of Bioproducts with Near-Term Potential. National Renewable Energy Laboratory.

Cao, D., Cai, W., Tao, W., Zhang, S., Wang, D. and Huang, D. (2017) Lactic Acid Production from Glucose Over a Novel Nb2O5 Nanorod Catalyst. Catalysis Letters. 147, 926–933.

Corma Canos, A., Iborra, S. and Velty, A. (2007) Chemical routes for the transformation of biomass into chemicals. Chemical Reviews. 107, 2411–2502.

da Silva, D.G. and Vasconcelos, W.L. (2019) Effect of sol-gel processing parameters on structure of zirconia. Ceramica. 65, 17–21.

Davis, R., Bartling, A. and Tao, L. (2020) Biochemical Conversion of Lignocellulosic Biomass to Hydrocarbon Fuels and Products: 2019. National Renewable Energy Laboratory.

Deshmane, V.G. and Adewuyi, Y.G. (2012) Synthesis of thermally stable, high surface area, nanocrystalline mesoporous tetragonal zirconium dioxide. Microporous and Mesoporous Materials. 148, 88–100.

Esposito, D. and Antonietti, M. (2013) Chemical conversion of sugars to lactic acid by alkaline hydrothermal processes. ChemSusChem. 6, 989–992.

Gorsd, M.N., Sosa, A.A., Frenzel, R.A. and Pizzio, L.R. (2018) Synthesis and characterization of tungstophosphoric acid-modified mesoporous sponge-like TUD-1 materials. Journal of Sol-Gel Science and Technology. 87, 204–215.

Holm, M.S., Pagán-Torres, Y.J., Saravanamurugan, S., Riisager, A. and Dumesic, J.A. (2012) Sn-Beta catalysed conversion of hemicellulosic sugars. Green Chemistry. 14, 702–706.

Kosri, C., Kiatphuengporn, S., Butburee, T., Youngjun, S. and Thongratkaew, S. (2021) Selective conversion of xylose to lactic acid over metal-based Lewis acid supported on γ-Al2O3 catalysts. Catalysis Today. 367, 205–212.

Montgomery, D.C. (2013) Design and analysis of experiments (8th ed.). Wiley.

Oefner, P.J., Lanziner, A.H., Bonn, G. and Bobleter, O. (1992) Quantitative Studies on Furfural and Organic Acid Formation during Hydrothermal, Acidic and Alkaline Degradation of D-Xylose. Monatshefte Für Chemie. 123, 547–556.

Pattnaik, F., Nanda, S., Kumar, V., Naik, S. and Dalai, A. K. (2021) Subcritical water hydrolysis of Phragmites for sugar extraction and catalytic conversion to platform chemicals. Biomass and Bioenergy. 145, 105965.

Paulino, P.N., Reis, O.C., Licea, Y.E., Albuquerque, E. M. and Fraga, M.A. (2018) Valorisation of xylose to lactic acid on morphology-controlled ZnO catalysts. Catalysis Science and Technology. 8, 4945–4956.

Sing, K.S.W. (1985) Reporting physisorption data for gas/solid systems with special reference to the determination of surface area and porosity. Pure and Applied Chemistry. 57, 603–619.

Tursi, A. (2019) A review on biomass: Importance, chemistry, classification, and conversion. Biofuel Research Journal. 6, 962–979.

Xia, M., Dong, W., Gu, M., Chang, C., Shen, Z. and Zhang, Y. (2018) Synergetic effects of bimetals in modified beta zeolite for lactic acid synthesis from biomass-derived carbohydrates. RSC Advances. 8, 8965–8975.

Yang, L., Su, J., Carl, S., Lynam, J.G., Yang, X. and Lin, H. (2015) Catalytic conversion of hemicellulosic biomass to lactic acid in pH neutral aqueous phase media. Applied Catalysis B: Environmental. 162, 149–157.

Yan, X., Jin, F., Tohji, K., Kishita, A. and Enomoto, H. (2010). Hydrothermal conversion of carbohydrate biomass to lactic acid. AIChE Journal. 56, 2727–2733.

Received: August 8, 2022

Sent to Subject Editor: August 29, 2022

Accepted: September 19, 2022

Recommended by Subject Editor Laura Briand