You are on page 1of 18

Catalysis Today 404 (2022) 201–218

Contents lists available at ScienceDirect

Catalysis Today
journal homepage: www.elsevier.com/locate/cattod

Optimization of reaction parameters by using response surface


methodology (RSM) for the selective dehydration of glucose to 5-hydrox-
ymethylfurfural (5-HMF), a valuable platform chemical over a mesoporous
TiO2 catalyst in dimethylsulfoxide (DMSO) medium
Richa Tomer , Prakash Biswas *
Department of Chemical Engineering, Indian Institute of Technology Roorkee, Roorkee 247667, Uttarakhand, India

A R T I C L E I N F O A B S T R A C T

Keywords: A highly active and selective mesoporous TiO2 was developed by the sol-gel hydrolysis technique for the selective
Glucose dehydration transformation of glucose to 5-hydroxymethylfurfural (5-HMF). The catalytic performance was tested in a glass
5-HMF tube reactor, and the reaction parameters were optimized using response surface methodology (RSM). The best-
Optimization
optimized reaction conditions were obtained using four-factors, four-responses, and the 5-level central composite
Catalyst recycle
Separation
design (CCD) integrated with the desirability approach. At the optimum reaction condition, RSM predicted
Power-Law kinetics complete conversion of glucose and a very high 5-HMF yield of ~56% in the dimethylsulfoxide (DMSO) solvent.
The predicted yield of other major products, including formic acid (FA) and levulinic acid (LA), was ~18% and
~2%, respectively. The empirical process model developed by RSM was validated by the experimental data
generated. Experimental results demonstrated that the TiO2 catalyst developed can easily be recycled with a
constant glucose conversion and 5-HMF yield. Further, a new liquid-liquid extraction method (LLE) is proposed
to extract 5-HMF from the product. The results depicted that 5-HMF can easily be extracted from the product
mixture with very high purity (~96%). Additionally, the Power-Law kinetic study suggested that the conversion
of glucose over TiO2 followed the first-order kinetics, and the calculated activation energy was 67.5 kJ mol−1.

1. Introduction β-1,4-glucosidic bonds [7]. This biomass-derived glucose has been


regarded as a feedstock for many valuable building block chemicals like
Global demand for energy and chemicals is increasing day by day 5-hydroxymethylfurfural, fructose, and organic acids like formic acid,
due to the growing population and urbanization. Energy Information levulinic acid, acetic acid (AA), lactic acid, etc. [8]. Among all the
Administration (EIA, U.S.) projected ~50% increase in world energy biomass-derived chemicals, 5-HMF comes in the top ten building block
consumption by 2050 [1]. Till today, nonrenewable sources such as chemical list given by the US department of energy [9] because of its
crude oil (~29%), coal (~27%), and natural gas (~23%) are the primary versatility and outstanding chemical reactivity [10–12]. 5-HMF can be
contributor to global energy demand and chemical supply [2,3]. Be- used as an intermediate for many industries such as bulk chemicals,
sides, nuclear (~4%) and renewable sources (~14%), including wind, biofuel, pharmaceuticals, polymer/rubber, food packaging, agrochemi-
solar, hydro, and biomass, also contribute significantly. Future global cals, cosmetic, and textile industries. It could also be widely used to
energy utilization projection also shows an increasing trend to utilize produce various high-value compounds like 2,5-furandicarboxylic acid
more renewables due to the depletion of fossils as well as environmental (2,5-FDCA), 2, 5-dimethylfuran, caprolactam, caprolactone, adipic acid,
issues [4,5]. For the last two decades, renewable lignocellulose biomass etc. [13].
has gained significant attention as a potential substitute for the pro- 5-HMF is generally synthesized via the conversion of biomass-
duction of energy and valuable chemicals [6]. Lignocellulosic biomass derived carbohydrates, i.e., cellulose, glucose, and fructose, in the
contains approximately 40–50% cellulose, 25–35% hemicellulose, and presence of an acid catalyst. A higher 5-HMF yield (~90%) can easily be
5–30% lignin [3], where cellulose is a D-glucose unit polymer linked by achieved from fructose [14]. However, fructose availability and its high

* Corresponding author.
E-mail addresses: prakash.biswas@ch.iitr.ac.in, prakashbiswas@gmail.com (P. Biswas).

https://doi.org/10.1016/j.cattod.2022.03.019
Received 31 August 2021; Received in revised form 9 March 2022; Accepted 24 March 2022
Available online 30 March 2022
0920-5861/© 2022 Elsevier B.V. All rights reserved.
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

cost limit its industrial applications. Alternatively, glucose can be 105 ◦ C after 2.7 h with 39.7 mg of catalyst. The information regarding
considered a sustainable feedstock due to its more accessibility and the catalyst reusability and stability is not reported. Similarly, Sarwono
lower cost [2]. The dehydration of glucose follows a tandem process, et al. [34] optimized the reaction process in [BMIM]Cl with the CrCl3
including the formation of fructose via glucose isomerization in the catalyst using an RSM IV optimal design with three variables, i.e.,
presence of Lewis acid sites (L-sites) accompanied by fructose dehy- temperature, time, and catalyst amount, respectively. Although the
dration to 5-HMF in the presence of Brønsted acid sites (B-sites) of a obtained 5-HMF yield was high, however, the cost of ionic liquid (IL)
catalyst [15]. Various heterogeneous solid acid catalysts such as zeolites, and the product separation difficulties make it less attractive. In the
sulfated zirconia, phosphate-titania, chromium catalysts, ion-exchanged organic DMSO medium, only one research paper was reported by Mo-
resins, heteropolyacids, and metal-organic frameworks (MOFs) had rales et al. [11]. A 3-level CCD model was used to optimize the glucose
been synthesized and used for glucose dehydration reaction [3,16–19]. dehydration reaction by considering run time, temperature, and catalyst
Among all these catalysts, various phosphated and sulfated metal oxide amount as the primary factors. The maximum 5-HMF yield of ~33% was
catalysts were widely used due to their tunable properties and acidity. A obtained at 147 ◦ C with 23 wt% catalyst loading after 24 h of reaction
higher (53–81%) yield of 5-HMF was reported in a biphasic medium time. However, the catalyst activity was reduced by 25% after the sec-
(THF: water) in the presence of phosphated TiO2 catalyst synthesized by ond cycle due to a significant loss of acidic sites, and their reaction time
sol-gel method [18,20,21]. WO3-TiO2 catalyst synthesized by was too long.
microwave-assisted hydrothermal method and demonstrated 58% yield In this study, the TiO2 was developed by the sol-gel hydrolysis
of 5-HMF in MIBK: H2O medium at 175 ◦ C after 2 h [22]. The phosphate method and examined for glucose dehydration in a DMSO medium. The
group and WO3 enhanced the B-sites. It has been shown that the anatase reaction parameters were optimized using RSM coupled with a 5-level
phase of TiO2 was very active for glucose dehydration because of the central composite design. The effect of variable reaction variables, i.e.,
presence of both L-sites and B-sites [18,23]. The electron pair accepting time, temperature, catalyst quantity, glucose loading on the conversion,
nature of Ti+4 corresponded to the Lewis acidic sites, and the proton and products (5-HMF, LA, and FA) yield, was evaluated. The signifi-
donor sites (-O-H) available corresponded to Brønsted acidic sites. The cance of independent variables and the interaction effect between in-
ratio of these two different acidic sites played a vital role in the selective dependent factors were also examined using RSM contour plots. The
transformation of glucose or fructose to 5-HMF. optimization of the process was performed based on the desirability
The glucose dehydration reaction is a solid acid-catalyzed reaction in approach. The model was validated with the experimental data gener-
various reaction mediums/solvents. Various reaction mediums such as ated at the optimum condition suggested by the RSM. In addition to that,
aqueous, ionic liquid ([BMIM]Cl, [EMIM]Cl), biphasic medium (THF: a catalyst regeneration and recycle strategy was developed, which
H2O, DMSO: H2O, THF/H2O/NaCl), and organic solvents (DMSO, MIBK, demonstrated that the TiO2 catalyst synthesized can easily be reused in
THF) have been explored [10,24–26]. Among all these solvents, water is several cycles without much deactivation and loss of 5-HMF yield.
known as a green and polar protic solvent that provides protons and high Furthermore, a new liquid-liquid extraction method is proposed to
polarity [27]. However, in a water medium, a lower yield of 5-HMF was separate the 5-HMF, and the results revealed that 5-HMF could be
reported because of more side reactions [10]. It has been reported that a separated from the product mixture with very high purity. Finally, the
higher yield of 5-HMF could be obtained in ionic liquids. However, the experimental data were fitted with a Power-Law model, and the kinetic
higher cost and separation inefficiency made the ionic liquids parameters were calculated for the dehydration of glucose over the TiO2
commercially less attractive. Instead, the biphasic medium often gave a catalyst.
better performance than the monophasic system. However, solvent
recycling and product separation are the main issues with the biphasic 2. Experimental
medium [28]. Among all other solvents, DMSO, as the organic solvent,
have been reported [16,18,25] as a promising dehydration medium due 2.1. Catalyst synthesis
to its advantageous properties such as strong dehydration capacity,
ability to prevent rehydration reaction, and favorable solvation property The TiO2 catalyst was developed by the sol-gel hydrolysis method
of the substrate, etc. [29]. [18]. In the typical synthesis, 20 mL of titanium isopropoxide (97%,
Previous literature mainly discussed the experimental optimization Aldrich, India) was added to 40 mL of 2-propanol (99%, Thomas Baker,
of the glucose dehydration process based on one factor at a time. This is India) and stirred for 5 min to make the solution homogeneous. Further,
not a very appropriate approach to determine the true optimum region. distilled water was added drop by drop to the titanium isopropoxide
This technique does not account for the interaction effect of different homogeneous solution for hydrolysis. The resulting mixture was stirred
parameters. Therefore, few statistical techniques such as response sur- for two hours, followed by the slow addition of 100 mL of 0.5 M HCl
face methodology and the Taguchi method have been proposed to (35–38%, Thomas Baker, India) for complete hydrolysis of titanium
determine the optimum region mathematically. These mathematical isopropoxide. The solution was stirred for 3 h and aged overnight at
techniques reduce the time of analysis, chemical consumption, and the room temperature. The white color slurry obtained was diluted and
number of experiments considerably. RSM is the statistical tool that decanted a number of times with deionized water till the pH of the so-
helps to find out the appropriate relationship between multiple variables lution reached ~2–3. The white-colored suspension obtained was
factors that affect the reaction with their response in the working heated at 70 ℃ for 3 h to make a white gel. Afterward, the prepared gel
domain by using statistics, optimization, and regression techniques [30]. was dried at 120 ℃ overnight in a conventional oven. The shiny white
RSM provides a large amount of information by performing a minimum agglomerated particles obtained were grounded and calcined at 200 ℃
set of runs to elucidate the experimental performance [31]. Very few for 2 h in a muffle furnace in the presence of air. The resulting solid
literature information is available on the RSM optimizations for trans- power catalyst was designated as TiO2.
forming glucose to 5-HMF in various reaction mediums [11,32–34]. In
the aqueous medium, a simplex-centroid (SC) method was used by 2.2. Catalyst characterization
Catrinck et al. [33] for the optimization of 5-HMF yield in the presence
of niobium acid (NbO) and niobium phosphate (NbP). A 55% glucose The specific surface area (Brunauer-Emmett-teller) pore dimensions
conversion with 56% selectivity of 5-HMF was obtained at 151.8 ◦ C in of the TiO2 catalyst were determined by the N2 adsorption-desorption
120 min. The actual 5-HMF yield and the information regarding catalyst technique in an ASAP 2060 (Mircomertics, USA) instrument. The BET
stability are missing. In a [BMIM]Cl (ionic liquid) medium, Utami et al. area and the catalyst’s pore size distributions were investigated using
[32] optimized the glucose dehydration process in the presence of Yb the Brunauer-Emmett-teller (BET) and Barret-Joyner-Halenda (BJH)
(OTf)3. A 3-level CCD was used. A 52% yield of 5-HMF was attained at methods. Before the analysis, the powdered sample was outgassed by

202
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

flowing helium (20 mL min−1) gas under vacuum at 150 ℃ for 2 h. with a binary pump (Waters, 515, USA), RI detector (Waters 2414, USA),
The powdered X-ray diffraction (P-XRD) pattern was recorded in a a column (HPX 87-H+, 4.6 mm × 350 mm, Bio-Rad, USA). The column-
D8-Advance diffractometer (Bruker, Germany) equipped with Ni filtered heater was used to maintain the desired temperature of the column. The
radiation (Cu-Kα, λ = 1.5406 Å, voltage: 40 kV, current: 40 mA). The temperature of the column and the detector were kept at 65 ◦ C and
diffraction pattern was recorded at a 2θ value of 5–90◦ . The step size and 55 ◦ C, respectively. For HPLC analysis, degassed 5 mmol H2SO4 was
scan speed were 0.02◦ and 4◦ min−1, respectively. The XRD peaks were used as a mobile phase. The mobile phase flow rate, as well as the in-
identified with the help of X′ pert-pro software and a joint committee of jection volume, were 0.6 mL min−1, and 20 µL, respectively. In every
powder diffraction standards (JCPDS). The line width of the XRD peaks analysis, prior to the injection of the liquid product into the HPLC for
corresponding to different crystal planes was used in the Scherrer for- analysis, it was filtered using a 0.22 µm syringe filter. The concentration
mula to determine the average crystallite size of the TiO2. of reactants and products was determined with the help of a standard
The acidity of the catalyst sample was determined by NH3-TPD calibration curve of all known compounds. The following formula was
(ammonia temperature-programmed desorption) technique on a used for the calculation of glucose conversion and products yield.
Micromeritics instrument (pulse chemisorb 2750, USA) coupled with a
TCD (thermal conductivity detector). In each experiment, a small % Conversion = [((Moles fed − Moles remaining)/Moles fed) ×100] (1)
amount of catalyst (~40 mg) was pretreated for 2 h under helium flow % Yield = [(Product moles formed/ Glucose moles fed) ×100] (2)
(20 mL min−1) at 150 ℃ followed by cooling at ambient conditions.
After the pretreatment, the catalyst sample was flooded with ammonia
by flowing a mixture of gas (10% NH3-He) (20 mL min−1) for 45 min at
room temperature. Further, the catalyst was purged with helium for 1 h 3. Results and discussion
to remove the physically adsorbed ammonia on the catalyst’s surface.
The saturated sample was further heated from 25 to 900 ℃ at 10 ℃ 3.1. Catalyst characterization
min−1, and the amounts of NH3 desorbed were continuously recorded
with a TCD. In the nitrogen adsorption-desorption isotherm (Fig. 1(a)) of TiO2, a
The Brønsted (B) and Lewis (L) acidity of the catalyst were analyzed hysteresis loop was detected at the relative pressure range of 0.4 < P/Po
by the pyridine-FTIR (Py-FTIR) in a spectrophotometer (Bruker Tensor < 0.7. These results represented the type-IV isotherm and H2 type hys-
II, Germany). In the distinctive process, initially, an appropriate amount teresis loop, which demonstrated the mesoporous characteristics of TiO2
of catalyst dried for 1 h under vacuum at 120 ◦ C followed by cooling to [36]. Moreover, this hysteresis loop also confirmed the
room temperature. Further, 50 mg of dried sample was saturated with ink-bottled-shaped pore structure of the TiO2 [37]. For TiO2 catalyst,
pyridine (100 µL) in a glass vial for 1 h. Then pyridine saturated sample Nui et al. [38] reported a similar type of loop at a P/P0 of 0.4–1. The BET
was heated at 120 ◦ C in the oven to evaporate the excess amounts of surface area, cumulative pore volume, and pore diameter of the TiO2
pyridine adsorbed. The prepared catalyst sample (5 wt%) was mixed catalyst were 272.66 m2 g−1, 0.189 cm3 g−1, and 2.3 nm, respectively.
with KBr (95:5 wt%) and pressed to a pellet of 13 mm diameter. Before The distributions of the pore size of catalyst are shown in Fig. 1(a) (inset
the analysis, the pellet weight was noted, and then the prepared pellet image). Raj et al. [39] reported a similar surfaced area of 277 m2 g−1 for
was fixed in the IR cell of the FTIR instrument. After that, the FTIR TiO2 catalyst without calcination. After calcination in the presence of air
spectra of the pyridine adsorbed catalyst sample were collected at the at 500 ◦ C for 2 h, the BET area of TiO2 was found to be decreased to 100
wavenumber of 500–4000 cm−1. The resolution and scan were 4 cm−1 m2 g−1.
and 64, respectively. The B/L ratio was obtained following the method In the XRD pattern (Fig. 1(b)) of TiO2, the major crystalline peaks
described in our previous work [35]. were identified at the two theta values of 25.4◦ , 37.8◦ , 48.0◦ , 55.3◦ ,
The catalyst morphology was analyzed by scanning electron micro- 62.8, 69.0◦ , 75.3◦ , and 82.7◦ corresponding to the (101), (112), (200),
scopy (FE-SEM, Carl-Zeiss-Ultra-Plus, Germany) armed with energy- (105), (204), (116), (215), and (303), crystal planes of tetragonal TiO2,
dispersive X-ray (EDX). EDX was used to determine the elements pre- respectively (JCPDS: 04–0477) [18]. These crystalline peaks of TiO2
sent in the sample and their percentage. Before the FE-SEM analysis, the were well matched with the anatase phase with the tetragonal structure
crushed sample spread on the specimen holder and coated with gold of TiO2. Similar two theta values corresponding to anatase TiO2 were
with the help of a sputter coater (Edwards S150, USA). The gold coating also reported by Muniandy et al. [40]. The Scherrer formula was applied
was done under the flow of argon to eliminate the charging of particles. to estimate the average crystallite size of the TiO2. The calculated
The analysis was performed under vacuum at an accelerated voltage (20 average value based on the line width of their respective XRD peaks was
kV). ~22.4 nm.
The NH3-TPD profile (Fig. 1(c)) of TiO2 displayed the desorption
2.3. Catalytic reaction test peaks at three different temperature regions, i.e., at 45–200 ℃,
200–300 ℃, and > 600 ℃, respectively. The desorption peaks repre-
The catalytic activity of TiO2 was carried out in a 15 mL borosilicate sented weak, moderate, and strong-strength acidic sites on the TiO2
glass tube connected with a tight screw cap. The appropriate amount of surface [41]. The total acidity of the TiO2 was calculated based on the
glucose was dissolved in 5 mL DMSO (99.7%, Loba Chemie, India) and desorbed volume of NH3. For TiO2, the estimated total acidity was
added to the glass tube reactor (Borosil, India). After that, a pre- 0.76 mmol g−1, comparable with the value (0.87 mmol g−1) reported
determined amount of catalyst was added to the reaction mixture. The earlier for anatase titania [20].
glass reactor was tightly closed and kept in a preheated oil bath equip- Pyridine FTIR study was performed to identify the nature of acidic
ped with a magnetic stirrer (Spinot digital model MC 02, Tarsons, India). sites of TiO2 synthesized. The FTIR spectra of the pyridine adsorbed TiO2
The dehydration reaction was done in the temperature range of catalyst demonstrated four peaks in the wavenumber range of
100–180 ◦ C. The glucose weight percentage in the feed, catalyst amount, ~1420–1705 cm−1, as shown in Fig. 1(d). The peak at the wavenumber
and reaction time varied in the range of 10–50 wt%, 0.15–0.965 g, and of ~1443.5 cm−1 and ~1536.5 cm−1 was associated with L-sites and B-
2.5–8.5 h, respectively. After completion of the reaction, the glass tube sites, respectively [42,43]. While the peak allied at ~1484.9 cm−1
reactor was shifted from the oil bath and quenched with iced water to represented both (B+L) sites of TiO2 catalyst [10]. These peaks
end the reaction instantly. Further, the catalyst and products were (1443.5 cm−1, ~1536.5 cm−1, and ~1484.9 cm−1) were attributed to
separated in a centrifuge (R-8C, Remi, India) at 4000 rpm, and the the coordinated pyridine molecule with L-sites and B-sites, respectively
composition of the product was measured in high-performance liquid [2,22]. The peak observed at ~1616 cm−1 was assigned to the physi-
chromatography (HPLC, Waters, USA). The HPLC system was equipped cally adsorbed pyridine ring on the catalytic surface [19,44]. The B/L

203
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Fig. 1. (a) N2 isotherm with pore-size variation, (b) XRD pattern, (c) NH3-TPD pattern, (d) Pyridine FTIR spectra, (e) FESEM images at different magnification, (f)
Particle size histogram, (g) EDX Plot.

204
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

ratio was calculated based on the absorbance area of the peaks detected temperature (A), reaction time (B), catalyst amount (C), and glucose
at 1484.9 cm−1 and 1536.5 cm−1, respectively. The obtained B/L ratio amount (D). These independent factors were symbolized as X1(A), X2(B),
was 2.73. Pyridine-FTIR study demonstrated the total acidity (B+L) of X3(C), and X4(D), respectively. Besides, ß0, ßi, ßii and ßij, and e repre-
0.031 mmol g−1, which was much lower than the total acidity value sented the model constant, linear, quadratic and interaction coefficients,
(0.76 mmol g−1) obtained from the NH3-TPD analysis. and the error term, respectively. To understand the effect of independent
The surface morphology of TiO2 was determined by FE-SEM equip- factors on responses with real and coded dimensionless factors, the
ped with EDX. Results dictated small and uniform spherical-shaped empirical model equations developed in CCD are shown in Table TS1
particles of TiO2 at different magnifications (Fig. 1(e)). Smaller and (supplementary information).
uniform particles were formed due to the stable gel formation during the To determine the suitability of the selected model, the statistical
synthesis via the sol-gel method [45]. For TiO2, Narabe et al. [46] re- analysis of the experimental data obtained for each response was eval-
ported the agglomerated morphology at lower magnification with a uated by the analysis of variance (ANOVA) [33]. ANOVA also measures
scale of ~5 µm and smaller spherical particles at higher magnification of the significance of each independent factor, the goodness of fit, lack of
the scale size 200 nm. In this study, the average particle size of fit, Fisher’s test (F-value), and probability (P-value) [50]. The goodness
10–20 nm was observed (Fig. 1(f)) for TiO2, which was very similar to of fit was judged by the determination coefficient (R2), and correlation
the value obtained in XRD (22.4 nm). Furthermore, the EDX equipped coefficient values obtained [51]. F-value was used to analyze the sta-
with FE-SEM analysis was done to obtain the elemental composition of tistical importance of the proposed model with the help of a P-value with
TiO2. The presence of titania, oxygen, and trace amounts of chlorine was a 95% confidence interval. Further, the lack of fit approach was used to
confirmed by the results reported in Fig. 1(g). In the EDX plot, the peak verify the acceptability and data variation. The acceptability of the
detected at 4.5 eV corresponded to bulk titania, and the peak at 5 eV was model was also verified by conducting the required runs at the antici-
due to the surface titania. Similar findings were also reported in the pated conditions given by the CCD. The goodness of fit of the model was
previous literatures [35,47]. improved by the RSM by tuning the respective P-value and F-values. The
interaction terms with a high P-value (> 0.05) were considered insig-
nificant in the model. Therefore, insignificant terms (P > 0.05) were
3.2. Experimental design by using response surface methodology (RSM)
neglected, which enhanced the model’s fitness. The modified empirical
model for glucose conversion and yield of products (FA, LA, 5-HMF)
Among the various RSM methodology known, CCD is better because
yield is shown in Table TS2, TS3, TS4, and TS5 (supplementary infor-
the region of operability is more than the region of interest [31,48,49].
mation). The significance of the modified models for all the responses
The CCD gives excellent prediction on the center point and is insensitive
was confirmed with the help of the F-value (48.4, 35.8, 21.8, and 28.6
to any missing data point. Therefore, the CCD method was chosen to
for conversion of glucose, 5-HMF yield, FA yield, LA yield, respectively),
design the glucose dehydration reaction in this study. From the previous
and p-value (< 0.0001).
literature, it has been observed that reaction temperature (A), reaction
The goodness of the model’s fit was also determined with the help of
time (B), catalyst amount (C), and glucose amount (D) was the primary
the coefficient of determination (R2), which represented the nearness of
reaction parameter that affected the glucose conversion as well product
the experimental, and the anticipated response [49]. A higher R2 value
yield. Thus, these reaction parameters were considered as the main
(> 0.80) obtained for the glucose conversion, the yield of 5-HMF, and
factors, and the corresponding decided responses were the conversion of
the yield of LA indicated the higher importance of the model and better
glucose, the yield of 5-HMF, the yield of FA, and the yield of LA,
predictability of the responses. The R2 value (0.94) for glucose conver-
respectively. For experimental design, five-level coded factors (−α, −1,
sion revealed that 94% of experimental data were well explained by the
0, +1, +α) were considered, the corresponding values are reported in
model. Similarly, 92% of experimental data were explained by the
Table 1. The ranges of factors for the initial design of the experiments
model for 5-HMF and LA yield with the R2 value of 0.92. However, a
were decided based on a few preliminary experiments performed. The
comparatively lower R2 value (0.61) was obtained for FA yield because
design matrix of the experiments included 31 runs containing 16
of the fluctuation in the concentration of FA throughout the design
factorial points, 6 center points, and 9 axial points, respectively. The
matrix. It might be because of the decomposition of FA to other com-
obtained design matrix and the actual value of all the responses are
pounds such as furfural and CO2, H2, CO, and H2O with the variation of
presented in Table 2. To design the experiment runs and to evaluate the
temperature [52,53]. According to the ANOVA fit statistics, the differ-
R2 value, Design-Expert (Version 11) was used.
ence between predicted R2 and adjusted R2 should be < 0.2 for a better
The quadratic (second-order) polynomial model equation developed
fitting of the model. The observed difference was much lower
by RSM correlates the responses (Y1, Y2, Y3, Y4) as a function of the
(0.06–0.09) for the glucose conversion, 5-HMF yield, the yield of LA,
factors (X1, X2, X3, X4) as Eq. (3) [2,49].
and the yield of FA, respectively, indicating a better fit. The adequate
n
∑ n
∑ n
∑ precision should be > 4 for a better prediction of the model. Adequate
Y = ß0 + ßi Xi + ßii X 2ii + ßij Xi Xj + e (3) precision suggests the domain of anticipated response with respect to its
i=1 i=1 i=1
connected error, i.e., signal/noise ratio. In this study, the adequate
i<j
precision value was in the range of 16.3–28.4, which was much higher
than the limiting value for all the responses. These results also suggested
Where Y (Y1, Y2, Y3, and Y4) represent the predicted responses of glucose better model prediction. The obtained statistical parameters for all
conversion, yield of 5-HMF, yield of FA, and yield of LA, respectively. Xi models are shown in Table 3, and the experimental and model-predicted
and Xj are the real values of the independent factors, i.e., reaction values are compared in Fig. 2.

Table 1 3.3. Catalytic performance


Selected factors with their corresponding levels.
Coded value Factor-A [h] Factor-B [℃] Factor-C [g] Factor-D [wt%] 3.3.1. Temperature and time interaction effect
− α (−2) 2.5 100 0.105 10
The combined effect of temperature (100–180 ℃) and time
−1 4.0 120 0.32 20 (2.5–8.5 h) on the conversion of glucose and yield of the product were
0 5.5 140 0.535 30 determined by keeping the other reaction parameters such as glucose
1 6.5 160 0.75 40 amount (30 wt%) and catalyst amount (0.535 g) constant. The volume
+ α(-2) 7.0 180 0.965 50
of DMSO was also kept constant at 5 mL for all the runs. The obtained
A
Reaction time, Breaction temperature, Ccatalyst amount, Dglucose amount conversion of glucose and yield of the product is displayed in Fig. 3(a)-

205
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Table 2
Experiment design in the presence of TiO2 catalyst.
Run Space type Factor Response

Time (h) Temperature ( C)



Catalyst Amount (g) Glucose % Glucose conversion 5-HMF yield FA yield LA yield

1 Factorial 7.0 160 0.320 40 94.8 36.8 6.4 2.3


2 Factorial 4.0 160 0.320 20 93.2 40.6 16.7 1.4
3 Center 5.5 140 0.535 30 83.5 25.2 10.7 0.6
4 Center 5.5 140 0.535 30 82.4 33.8 10.6 1.1
5 Factorial 7.0 160 0.320 20 93.7 46.5 17.9 2.6
6 Axial 8.5 140 0.535 30 88.8 31.5 11.1 1.9
7 Factorial 4.0 120 0.320 20 64.3 13.0 5.5 0.9
8 Center 5.5 140 0.535 30 84.2 27.6 10.1 0.3
9 Factorial 7.0 120 0.320 20 75.2 26.0 12.5 2.4
10 Axial 5.5 180 0.535 30 100.0 35.6 60.2 4.2
11 Axial 5.5 140 0.535 50 89.2 33.3 21.9 1.4
12 Factorial 4.0 160 0.320 40 90.6 32.1 14.4 1.8
13 Axial 6.5 140 0.535 30 83.7 26.3 7.1 1.6
14 Factorial 4.0 120 0.750 40 77.0 15.7 10.0 0.2
15 Factorial 4.0 160 0.750 40 94.1 34.5 14.3 2.5
16 Factorial 4.0 120 0.750 20 71.0 27.7 8.6 0.3
17 Axial 5.5 140 0.535 10 92.1 52.3 7.6 2.8
18 Center 5.5 140 0.535 30 76.6 32 21.2 0.0
19 Factorial 4.0 120 0.320 40 74.4 7.6 3.7 0.2
20 Factorial 7.0 120 0.320 40 83.0 10.4 6.5 0.0
21 Factorial 7.0 160 0.750 40 99.1 32.2 16.3 3.4
22 Center 5.5 140 0.535 30 83.7 26.4 9.1 0.6
23 Factorial 4.0 160 0.750 20 97.9 40.7 53.3 3.0
24 Factorial 7.0 120 0.750 40 79.1 17.2 10.0 0.7
25 Axial 2.5 140 0.535 30 82.3 18.5 11.7 0.0
26 Axial 5.5 140 0.965 30 89.6 33.1 11.8 1.0
27 Factorial 7.0 160 0.750 20 97.8 42.5 35.0 3.4
28 Axial 5.5 140 0.105 30 79.3 14.1 8.7 0.5
29 Factorial 7.0 120 0.750 20 74.7 30.9 11.8 1.8
30 Axial 5.5 100 0.535 30 48.6 8.7 1.1 0.2
31 Center 5.5 140 0.535 30 82.5 26.8 11.4 0.7

low temperature (< 140 ◦ C) and at < 4 h (Fig. 3(c)). However, at a


Table 3
higher temperature (> 160 ◦ C) and after a longer reaction time (> 6 h),
Statistical parameter for the glucose conversion to 5-HMF.
the LA was detected. The maximum LA yield of ~5% was observed at
Response R2 Adjusted R2 Predicted R2 Adequate precision > 160 ◦ C after a reaction time of more than 7 h. Morales et al. [11]
Glucose conversion 0.94 0.92 0.84 28.4 reported the maximum LA yield of < 4% at 150 ◦ C with Amberlyst-70
5-HMF yield 0.92 0.89 0.83 25.8 catalyst after the reaction time of 24 h in the DMSO medium. The
LA yield 0.92 0.89 0.83 20.3 lower yield of LA was observed because the DMSO restricted the rehy-
FA yield 0.61 0.58 0.49 16.3
dration of 5-HMF. Previous reported density functional theory (DFT)
calculation on the stability of 5-HMF in DMSO/water mixture dictated
(d), respectively. As expected, glucose conversion was found to be that DMSO increased the LUMO energy of 5-HMF, which reduced the
increased with temperature (Fig. 3(a)) [10,52]. The conversion was very probability of the nucleophilic attack and stabilized the 5-HMF in the
high (~80%) even at low temperatures (~140 ℃), and it was achieved DMSO medium. The molecular simulation study also suggested that the
~100% at > 160 ℃. Conversion also displayed an increasing trend with carbonyl group of 5-HMF solvated in DMSO prevented the rehydration
increasing reaction time. After 8.5 h, glucose conversion was ~80–86% of 5-HMF to other side products, including levulinic acid, formic acid,
in the temperature range of 120–140 ℃, and it reached ~90% at etc. [55,56]. The yield observed for FA is shown in Fig. 3(d). Initially, at
~160 ℃ in less than 3 h of reaction. These results revealed that the rate low temperature (< 140 ◦ C), in the entire reaction time domain
of glucose conversion was ~1.4 times (3.9 × 10−3 mmol mL−1 min−1) (2.5–8.5 h), the FA yield was very low (< 14%). However, FA yields
higher at 160 ◦ C concerning the rate (2.7 × 10−3 mmol mL−1 min−1) increased significantly (~20–40%) at 160–180 ◦ C. The formation rate of
observed at 120 ◦ C. At higher temperatures, the favorable electron ex- FA was also calculated, and it was observed that formation rate (7.2 ×
change rate facilitated, which enhanced the glucose conversion rate [2]. 10−4 mmol mL−1 min−1 at 160 ◦ C) was ~3 fold higher at high tem-
The time and temperature interaction results suggested that, in the perature (160–180 ◦ C) with respect to the rate (2.3 × 10−4 mmol mL−1
presence of the TiO2 catalyst, very high glucose conversion (~80–100%) min−1 at 120 ◦ C) calculated at low temperature (100–140 ◦ C). These
can be achieved in the temperature range of 140–180 ◦ C, and after a results revealed that higher reaction temperature with prolonged reac-
reaction time of 4–8 h. The variation of the yield of 5-HMF with tem- tion time facilitated the formation of FA and LA, which was formed due
perature and time is manifested in Fig. 3(b). The yield of 5-HMF was to the rehydration/degradation of 5-HMF/glucose. A similar type of
found to be very low (< 20%) at low temperature (< 140 ◦ C) and in the rehydration product was also observed mainly at high temperatures in
entire time domain (2.5–8.5 h). At this low-temperature range, the various previous studies [11,12,57]. The contour plots reported in Fig. 3
calculated 5-HMF formation rate with time was also low (~5.5 × 10−4 (a)-(d) suggested that the influence of the reaction temperature was
mmol mL−1 min−1 at 120 ◦ C). However, at high temperatures (> more significant with respect to the reaction time on glucose conversion
160 ◦ C), the calculated rate of formation increased to approximately 3.6 along with the products (5-HMF, LA, and FA) yield.
times (2 × 10−3 mmol mL−1 min−1 at 160 ◦ C). Similar results were re-
ported previously over several solid acid catalysts and reaction solvents 3.3.2. Glucose amount and temperature interaction effect
[22,54]. A maximum ~43% yield of 5-HMF was achieved at As discussed in Section 3.2, the interaction terms with a high P-value
160–180 ◦ C, and after a reaction time of 5.5 h. LA was not observed at (> 0.05) were considered insignificant in the model. Therefore,

206
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Fig. 2. Experimental vs. predicted values (a) conversion of glucose, (b) yield of 5-HMF, (c) yield of LA, and (d) yield of FA.

negligence of insignificant terms (P > 0.05) from the model improved (< 20 wt%) at the temperature of < 160 ◦ C. Also, the high reaction
the model’s fitness. The ANOVA statistics suggested that the combined temperature facilitated the rehydration of the desired 5-HMF to side
effect of glucose amount and temperature on 5-HMF and FA yield was products such as LA.
negligible. Therefore, the obtained interaction effect of glucose amount
(10–50 wt%) and reaction temperature (100–180 ◦ C) on glucose con- 3.3.3. Glucose amount and time interaction effect
version and LA yield with a fixed amount of catalyst (0.535 g) and re- To determine the combined interaction effect, glucose amount and
action time (5.5 h) is shown in Fig. 4. It was found that the conversion of time were varied in the range of 10–50 wt%, and 2.5–8.5 h, respectively.
glucose was ~60% at a low temperature (< 120 ◦ C) in the entire region The catalyst amount (0.535 g) and the reaction temperature (140 ◦ C)
of glucose amount (10–50 wt%) studied. However, a slight increase were kept constant. As shown in Fig. 5(a), glucose conversion was
(~72%) in conversion was observed with 50 wt% glucose at < 120 ◦ C observed to be increased with time, and it decreased with glucose con-
because of higher accessibility of glucose molecules [10,20]. The cur- centration. It has been seen that in the reaction time domain of 7–8.5 h,
vature of contours lines was found to be increased at a high temperature glucose conversion decreased from ~92% at < 20 wt% glucose to ~87%
(> 140 ◦ C), which indicated the enhanced interaction of glucose amount at 20–40 wt% glucose. Further, at a higher glucose amount (> 40 wt%),
with temperature [2]. The complete conversion (100%) of glucose was the conversion was found to be increased marginally because the con-
observed in the temperature region of ~160–180 ◦ C at a low glucose centration (> 40 wt%) was beyond the design space. Conversion
amount (< 20 wt%). However, at higher glucose concentration (> 30 wt decreased with increasing glucose amount was caused by the less
%), the maximum conversion was ~95% at 160–180 ◦ C. Glucose con- availability of active centers in the presence of a constant amount of
version decreased because of the presence of an insufficient amount of catalyst. The comparable observations were stated by Sarwono et al.
active sites since the catalyst amount was constant [58]. LA was not [58]. The glucose conversion rate was calculated with the variation of
detected at higher glucose concentration (> 30 wt%) and lower tem- reaction time as well as glucose amount. The rate was found to be
perature region (~100–140 ◦ C). However, the formation of LA was decreased with reaction time and increased with glucose amount. At
found to be favorable at high temperatures (~140–180 ◦ C) in the entire 140 ◦ C in the presence of 30 wt% of glucose, the conversion rate
glucose amount domain, as shown in Fig. 4(b) [11]. The formation of LA decreased from 5.9 × 10−3 to 4.4 × 10−3 mmol mL−1 min−1 in the re-
via 5-HMF rehydration was favorable at high reaction temperatures. The action time domain of 2.5–8.5 h. Moreover, after 5.5 h, the calculated
maximum yield of LA (~5%) was obtained at the temperature region of glucose conversion rate was observed to be increased from 9.1 × 10−4 to
160–180 ◦ C. This interaction study suggested that complete glucose 3.9 × 10−3 mmol mL−1 min−1 in the glucose amount range of 10–50 wt
conversion can be achieved in the presence of a lower glucose amount %. The decreasing trend of the yield of 5-HMF with increasing glucose

207
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Fig. 3. 3-D surface and contour plots represent the temperature-time interaction effect on (a) Glucose conversion, (c) 5-HMF yield, (d) LA yield, (d) FA yield.
Reaction condition: 0.535 g catalyst, 30 wt% glucose in the solvent, 5 mL DMSO solvent.

208
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Fig. 4. 3-D surface and contour plots represent the temperature-glucose amount interaction effect on (a) Glucose conversion (b) LA yield. Reaction condition: 0.535 g
catalyst, 5.5 h, 5 mL DMSO solvent.

amount in the reaction mixture was observed (Fig. 5(b)). The value suggested that glucose concentration of < 20 wt% and the reaction time
decreased from ~52% to ~32% on increasing the glucose concentration of 4–8 h was beneficial for higher glucose conversion and 5-HMF yield.
from 10 wt% to 50 wt%. It has been observed from previous literature
that, in the presence of a constant amount of catalyst, beyond a certain 3.3.4. Temperature and catalyst amount interaction effect
glucose concentration, conversion of glucose/5-HMF to other undesired The combined effect of catalyst amount and temperature on the
products such as humin facilitated [58,59]. Because of this reason, the glucose conversion and FA yield was insignificant as the P-value was
5-HMF yield decreased at a high concentration of glucose. However, > 0.5. Therefore, the obtained interaction effect of reaction temperature
with increasing reaction time, the yield of 5-HMF increased marginally (100–180 ◦ C) and catalyst amount (0.105–0.965 g) on the 5-HMF and
due to the availability of sufficient time to obtain the equilibrium con- LA yields are shown in Fig. 6. The other reaction factors, such as glucose
version. In the presence of a limited amount of catalyst, prolonged re- concentration (30 wt%) and reaction time (5.5 h), were constant. It was
action time helps achieve equilibrium conversion [2]. The highest observed from Fig. 6(a) that the yield of 5-HMF was negligible (< 10%)
(~52%) yield of 5-HMF was detected within 4–8 h of reaction time at at a lower amount of catalyst (0.105–0.32 g) and a low reaction tem-
lower glucose concentration (< 20 wt%). The curvature of lines in Fig. 5 perature (< 120 ◦ C). The 5-HMF yield was lower at low temperatures
(c) indicated the interaction of glucose amount and reaction time on LA due to low activation energy. However, in the low-temperature region
yield. The overall yield of LA was very low (< 4%). LA was formed due to (< 120 ◦ C), at high catalyst loading (> 0.75 g), ~19% yield of 5-HMF
the rehydration of 5-HMF; very low LA yield suggested that the rehy- was obtained because of the accessibility of more active centers [25].
dration reaction was negligible. However, the results reported in Fig. 5 It was observed that in the high-temperature region (> 140 ◦ C), the
(c) suggested that LA formation was facilitated at a higher reaction time 5-HMF yield decreased with the increasing catalyst amount (> 0.535 g).
(> 7 h) and in the presence of a lower amount of glucose (< 20 wt%) At > 160 ◦ C, in the low catalyst amount region (0.105–0.32 g), the
[19]. The interaction effect of glucose amount and reaction time maximum 5-HMF yield of ~41% was obtained. Moreover, the 5-HMF

209
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Fig. 5. 3-D surface and contour plots represent the glucose amount-reaction time interaction effect on (a) glucose conversion, (b) 5-HMF yield, (c) LA yield. Reaction
condition: 140 ◦ C, 0.535 g catalyst, 5 mL DMSO solvent.

210
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Fig. 6. 3-D surface and contour plots represent the temperature-catalyst amount effect on (a) 5-HMF yield (b) LA yield. Reaction condition: 5.5 h, 30 wt% glucose,
5 mL DMSO solvent.

yield decreased to ~34% at higher catalyst loading (> 0.75 g) in the increasing reaction time from 2.5 to 8.5 h. Conversion also increased
same temperature region. These results suggested that increasing the from ~77–90%, with the catalyst loading increasing from 0.105 to
catalyst amount and temperature facilitated the formation of 5-HMF up 0.965 g (Fig. 7(a)). These results suggested that both factors had an
to an optimum value. Beyond that, excess catalyst amount and high effective linear outcome on glucose conversion. A similar outcome of
temperature facilitated the side reactions, i.e., rehydration of 5-HMF. time and the amount of catalyst was also observed on the yield of 5-
The variation of the yield of LA is shown in Fig. 6(b). LA was not HMF, as manifested in Fig. 7(b). The yield of 5-HMF increased from
formed at a low-temperature region (< 140 ◦ C) in the complete domain 17% to 25%, increasing the time from 2.5 h to 8.5 h at low catalyst
of catalyst amount (0.105–0.965 g) studied. However, LA formation was loading (0.105–0.32 g). A maximum yield of 5-HMF of ~18–35% was
observed predominantly at the high-temperature region (> 160 ◦ C). The achieved using the catalyst amount of 0.105–0.965 g. High catalyst
maximum percentage yield of LA was ~5.5% at a high catalyst amount loading (> 0.75 g) and reaction time (> 5.5 h) enhanced the yield of LA
(> 0.535 g) and a high temperature (160–180 ◦ C) because of the rehy- (2%) and FA (18%), as depicted in Fig. 7(c) and (d), respectively. These
dration of 5-HMF [60]. The combined effect of temperature and catalyst results suggested that glucose conversion and 5-HMF yield increased
amount study suggested that a high 5-HMF yield (~41%) can be ach- with time and catalyst amount because of the accessibility of more active
ieved in the reaction temperature range of 140–160 ◦ C and the presence centers at higher catalyst loading for a more extended period. However,
of an optimum amount of catalyst (0.32–0.535 g), respectively. the higher accessibility of active centers also enhanced the yield of
rehydrated products such as LA and FA, respectively [61,62]. The time
3.3.5. Time and catalyst amount interaction effect and catalyst amount interaction effect suggested that the optimum
The combined effect of reaction time (2.5–8.5 h) and catalyst catalyst amount of < 0.75 g and the reaction time of < 7 h were favor-
amount (0.11–0.97 g) on conversion and product yield at a constant able for a higher 5-HMF yield.
glucose concentration (30 wt%) and reaction temperature (140 ◦ C) is As discussed earlier, the B/L ratio of the catalyst played an important
shown in Fig. 7. The obtained straight contour lines in all Fig. (7(a)-(d)) role in product distribution. The B-L acidic sites of TiO2 were identified
indicated a more negligible interaction effect between these two factors based on the functional group obtained in FTIR spectroscopy, as re-
on all the responses. Glucose conversion increased from ~77–84%, ported in Fig. 9(b). The FTIR spectrum of TiO2 catalyst demonstrated

211
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Fig. 7. 3-D surface contour plots represent the catalyst amount-reaction time interaction effect on (a) the conversion of glucose, (b) the yield of 5-HMF, (c) the yield
of LA, and (d) the yield of FA. Reaction condition: 140 ◦ C, 30 wt% glucose, 5 mL DMSO as solvent.

212
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

three significant bands at the wavenumber of 3374.1, 1619.5, and value for Yi response. All individual desirability (di(Yi)) functions for all
1021–500 cm−1, corresponding to the H-O-H stretching of adsorbed the responses were combined to obtain the maxima, a unique function
water molecules, Ti-O-H stretching, and Ti-O fingerprints, respectively. named overall desirability (DI), which was found by using the following
An O-Ti group was responsible for Lewis (L) acidic sites, while the Ti-O- equation:
H group was responsible for Brønsted (B) acidic sites. The proposed /
∑n ( )
structure of the TiO2 catalyst is shown in Fig. S1 (Supplementary file). 1 si n

Similar types of B and L sites are also reported earlier [63]. Based on the
s1 s2
DI = (d1 d2 .dn )sn i=1
= di si (5)
product distribution obtained over the TiO2 catalyst, a probable reaction i=1

pathway is proposed (Fig. 8), which depicts the different steps of


where ’n′ was the total number for responses, ’si’ was the relative
product formation via the interaction of molecules with the different
importance assigned to each response, which might be varied from 1 for
acidic sites of the catalyst [64]. DMSO acted as a reaction medium and
the least important response to 5 for a most important response. DI also
prevented the rehydration of 5-HMF via the solvation of its carbonyl
varied from zero to one. DI worked on a direct search method to
group [56].
maximize the desirability function [65]. The direct search method was
advantageous because, in this method, any information regarding the
3.4. Optimizations of reaction parameter and model validation gradient of the objective function is not required.
The individual desirability functions were defined for all the re-
The optimal reaction conditions for high glucose conversion and sponses as below:
high yield of 5-HMF were determined using the optimization (numeri- ⎧ ⎫
⎪ 0 if Yi ≤ 80 ⎪
cal) features of Design Expert-11 (software). The goal of the optimiza- ⎪





tion process was to maximize the glucose conversion and 5-HMF yield. Y1 − 80
d1 (Y1 ) = if 80 < Yi < 100 (6)
⎪ 100 − 80
The constraints and importance levels decided to achieve the selected ⎪
⎪ ⎪

⎩ ⎪

goals are shown in Table 4. The reaction parameter study demonstrated 1 if Yi ≥ 100
that the high conversion and the desired product (5-HMF) yield could be ⎧ ⎫
obtained at a higher temperature, reaction time, catalyst amount, and ⎪
⎪ 0 if Y2 ≤ 10 ⎪


⎨ ⎪

low glucose concentration. As a result, temperature, time, and amount of Y2 − 10
d2 (Y2 ) = if 10 < Y2 < 60 (7)
catalyst were set a goal as "in range", and the glucose amount was ⎪
⎪ 60 − 10 ⎪


⎩ ⎪

assigned with a goal of 10 wt% to achieve the maximum response for 1 if Y2 ≥ 60
feed (glucose) conversion and desired product (5-HMF) yield. The so-
lutions acquired with these four variables, including the desirability ⎧ ⎫

⎪ 1 if Y3 ≤ 1.5 ⎪

index, were generated by the software, and the results obtained are ⎪
⎨ ⎪

Y3 − 1.5
shown in Table 5. The individual desirability index (’di’) was set as an d3 (Y3 ) = if 1.5 < Y3 < 4.5 (8)

⎪ 4.5 − 1.5 ⎪

objective function ranging from 0 to 1 (0 ≤ di ≥ 1). When the response ⎪
⎩ ⎪

0 if Y3 ≥ 4.5
was at the target, ’di’ was equal to one (1), and when the response was
exterior of the desired region, then ’di’ was equal to zero (0). Then the ⎧ ⎫
⎪ 1 if Y4 ≤ 20 ⎪
individual desirability function was described as follows: ⎪





Y4 − 20
⎧ ⎫ d4 (Y4 ) = if 20 < Y4 < 30 (9)

⎪ 0 if Yi ≤ Yi min ⎪
⎪ ⎪
⎪ 30 − 20 ⎪


⎪ ⎪
⎪ ⎪
⎩ ⎪

⎨ min ⎬ 0 if Y4 ≥ 30
Yi − Yi
di(Yi ) = max min if Yi
min
< Y i < Yi
max (4)

⎪ Yi − Yi ⎪






⎭ To verify the model’s prediction ability, a glucose dehydration re-
1 if Yi ≥ Yi max action was performed at the optimal reaction conditions suggested by
the software. The comparison of model-predicted and experimental
Where di(Yi) was the desirability of Yi response, Yimin was the lower values for a particular solution is shown in Table 6. The results suggested
acceptable value for the Yi response, Yimax was the acceptable upper

Fig. 8. Reaction pathway of glucose dehydration in the presence of a TiO2 catalyst.

213
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

Table 4
Constraints for numerical optimization by using the CCD method in RSM.
Factor Goal Lower limit Upper limit Lower weight Upper weight Importance

Time (h) In range 4 7 1 1 3


Temperature (◦ C) In range 130 160 1 1 3
Catalyst amount (g) In range 0.32 0.75 1 1 3
Glucose amount (wt%) Target= 10 10 30 1 1 3
Glucose conversion (%) Maximize 80 100 1 1 5
5-HMF yield (%) Maximize 10 60 1 1 5
LA yield (%) Minimize 1.5 4.2 1 1 3
FA yield (%) Minimize 20 30 1 1 4

Table 5
Possible solutions for the optimum conditions to maximize the 5-HMF yield and glucose conversion.
No. A (h) B (◦ C) C (g) D (wt%) Y1 (%) Y2 (%) Y3 (%) Y4 (%) Desirability index (DI)

1 4.0 159.3 0.320 10.0 100.0 56.4 18.4 2.4 0.921


2 4.0 158.0 0.320 10.0 99.2 55.8 17.6 2.4 0.913
3 4.0 155.8 0.320 10.0 97.9 54.8 16.5 2.3 0.899
4 5.6 155.7 0.320 10.0 100.0 57.3 16.5 3.1 0.857
5 5.8 160.0 0.320 12.8 100.1 54.0 18.7 3.0 0.838
6 5.4 149.6 0.321 10.0 95.9 53.9 13.7 2.9 0.825
7 5.8 160.0 0.385 14.5 99.3 51.1 20.0 2.9 0.811

4. Catalyst stability, reusability, and regeneration study


Table 6
Comparison between the experimental and model-predicted values at the opti-
The catalyst stability and the reusability were examined at the
mum condition.
optimized reaction conditions. After each experiment, the used catalyst
Response Experimental value Predicted Value % error
was separated via centrifuge from the reaction mixture. The catalyst was
(%) (%)
washed with distilled water and acetone several times to separate all the
Glucose conversion 98.2 100.0 1.8 deposited byproducts. After washing, it was filtered and heated at 80 ◦ C
(%)
overnight in a conventional oven for drying. The dried catalyst was
5-HMF yield (%) 56.1 56.4 0.54
LA yield (%) 2.38 2.4 0.84 further used in the subsequent cycle of the reaction. The results sug-
FA yield (%) 19.58 18.4 6.03 gested that the catalyst activity decreased in the subsequent cycle (Fig. 9
(a)), and conversion decreased from 98% to 64% in cycle-3. Glucose
Optimized reaction condition: 159.3 ◦ C, 4 h, 10 wt% glucose, 0.32 g catalyst,
5 mL DMSO solvent, and DI = 0.92. conversion decreased due to the deposition of humin or other organic
molecules formed on the catalyst active surface, which blocked the
active catalyst sites in every cycle. Therefore, after cycle-3, the catalyst
that the experimental values for reactant (glucose) conversion and
was regenerated by heating at 400 ◦ C in a furnace for 1 h in air presence.
product (5-HMF, LA, and FA) yield were very near the model’s predicted
Further, the activity of this regenerated catalyst was evaluated in the
value. The relative error between predicted and experimental values was
next run under the same reaction condition. Results dictated that con-
less than 2% for reactant (glucose) conversion and desired product (5-
version and yield further increased to 87% and 54%, respectively (Fig. 9
HMF) yield. It can be concluded that the developed model for conversion
(a)). These values were comparable with the conversion (98%) and yield
(glucose) and the main product (5-HMF) yield demonstrated very good
(56%) obtained over fresh catalysts. These results revealed that the TiO2
predictability with adequate accuracy within the experimental domain.
catalyst developed was very active, selective to 5-HMF, and easily

Fig. 9. (a) Recycle study of the TiO2 catalyst, (b) FTIR spectra of fresh, spent, and regenerated catalyst. Reaction condition: 160 ◦ C, 4 h, 0.32 g catalyst, 10 wt%
glucose, 5 mL DMSO solvent.

214
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

recycled by simply heating in the presence of air. The removal of process proposed in this study may be very attractive for the production
deposited organics from the active sites of TiO2 catalyst was found to be and separation of 5-HMF on a larger scale.
very crucial for its regeneration. In Table 7, the result obtained in this study is compared with the
To understand the reason for the catalyst deactivation, the FTIR previously reported results in the DMSO solvent. In the previous studies
analysis of fresh, used, and regenerated catalysts were performed, and [5,10,16,17,26,44,69–73], the glucose conversion and 5-HMF yield
the obtained spectra were compared in Fig. 9(b). FTIR spectrum of fresh were reported in the span of 97–100% and ~26–67%, respectively. In
TiO2 catalyst demonstrated three significant bands at the wavenumber this study, a maximum of 56% yield was reached for 5-HMF with 98%
of 3374.1, 1619.5, and 1021–500 cm−1, assigned to the H-O-H stretch- glucose conversion at 160 ◦ C after the 4 h of reaction. Comparatively,
ing of water molecules adsorbed, Ti-O-H stretching, and Ti-O finger- Qu et al. [17] reported a higher yield (~67%) of 5-HMF in the presence
prints [38,66], respectively. In the used catalyst, besides these bands, of ionic liquid as a catalyst. However, the isolation of products and the
some other wide-stretching bands were observed at the wavenumber cost of ionic liquids were critical issues. The marginally higher yield
range of 2094.8–1795.2 and 1478.0–1278.5 cm−1. The broad spectrum (57%) of 5-HMF was also reported by Lu et al. [69]. Nevertheless, in
detected at 2094.8–1795.2 cm−1 represented the C– – O stretching, and their study, it was observed that the reaction time was higher, and most
at 1478.0–1278.5 cm−1 corresponded to C-O, C-O-C stretching of the importantly, the 5-HMF yield was found to be decreased approximately
furan ring, separately. These distinct results confirmed the existence of 21% after recycling. None of these studies discussed the extraction of
humin and/or other deposited organic molecules on the used catalyst 5-HMF. In the present study, ~54% 5-HMF yield was obtained even after
surface [35]. Furthermore, in the FTIR spectra of the regenerated cata- cycle-3 and 5-HMF were separated from the reaction mixture with very
lyst, the band (2094.8–1795.2 cm−1) corresponding to the humin high purity. Therefore, the results reported in this study are very
and/or other organic molecules was absent. These results confirmed that promising and may be very attractive commercially for the
the humin and/or other organics deposited on the catalyst surface were industrial-scale manufacturing of 5-HMF from glucose in the DMSO
removed entirely after regeneration. The peak intensity corresponding medium.
to the deposited organic species at the wavenumber of 5-HMF is an important platform chemical and is used as a feedstock
1478.0–1278.5 cm−1 was reduced significantly, and the FTIR spectra in various chemical industries, including bulk chemicals, biofuel,
resumed like a fresh catalyst (Fig. 9(b)-inset). pharmaceuticals, polymer/rubber, food packaging, agrochemicals,
The Power-Law kinetics of glucose conversion was developed cosmetic and textile industries, etc. Various important chemicals, i.e.,
following the Levenspiel method. The glucose concentration at different 2,5-furandicarboxylic acid (2,5-FDCA), adipic acid, 2,5-Dimethylfuran,
temperatures (120–140 ◦ C) was analyzed at the time interval of 10 min caprolactam, and caprolactone 5-HMF, can also be synthesized from
up to 1 h, as shown in Fig FS2(a) (supplementary file). The obtained 5-HMF. A high 5-HMF yield can easily be acquired from fructose.
results suggested first-order kinetics of glucose dehydration reaction However, the availability and the cost of fructose are major concerns.
with the Ea (activation energy) and A0 (Arrhenius factor) of Alternatively, the production of 5-HMF from the low cost and available
67.5 kJ mol−1 and 1.2 × 107 min−1, respectively (Fig. FS2(b)) (supple- glucose is very attractive. However, the pyranose structure of glucose
mentary file). The value of Ea obtained was less with respect to the hindered the formation of 5-HMF from glucose. The development of a
previously reported value for the same reaction systems demonstrating special type of catalyst, solvent, and catalytic process is highly desirable
excellent activity of the TiO2 catalyst in the DMSO medium. The detailed for the development of a commercial process. In addition to that, the
kinetics study is shown in Section S1 in the supplementary information. separation and purification of 5-HMF from the reaction mixture is also a
big challenge. In this study, a new promising, easily available, and
5. 5-HMF extraction from the product mixture recyclable TiO2 was synthesized for the manufacturing of 5-HMF from
glucose. Most importantly, a new product separation strategy is also
Previous literature discussed the formation of 5-HMF in different proposed. Therefore, the information reported in this study may be
reaction mediums in the presence of various types of catalysts. However, helpful for the large-scale production of 5-HMF from glucose in the near
none of the studies discussed the separation and purification of 5-HMF,
which is the major hurdle in the industrial manufacture of 5-HMF.
Therefore, in the present study, a new liquid-liquid extraction process Table 7
is proposed for the separation of 5-HMF. The process flow diagram of the Comparison of results obtained with previously reported literature in the DMSO
separation process proposed is displayed in Fig. FS3 (supplementary medium.
file). In the typical procedure, the TiO2 catalyst was segregated from the S. Temperature, Catalyst Glucose 5- Reference
product mixture by centrifuge. Further, the NaCl saturated water was no. Time Conversion HMF
added into the remaining product mixture to separate the humin/ yield
polymeric byproducts. Then the 5-HMF was removed from the super- 1. 140 ◦ C, 12 h UiO-66 – 49.7 [5]
natant by the Liquid-liquid extraction (LLE) in the separating funnel. 2. 130 ◦ C, 1 h PIL-Sn 99 51.1 [10]
Tetrahydrofuran (THF) was used as a low boiling solvent for the 3. 130 ◦ C, 4 h SO42−/ZrO2/ 99 47.6 [16]
Al2O3
extraction of 5-HMF from the product mixture containing DMSO. The 4. 180 ℃, 1 h [CO2HMIM]BF4 – 67.4 [17]
oxygen heterocyclic structure of THF facilitated the 5-HMF extraction 5. 170 ◦ C, 2 h Al-KCC-1 Silica 97 39.0 [26]
efficiency [67]. The addition of NaCl in the aqueous phase increased the 6. 140 ◦ C, 2 h Al2O3-B2O3(Al/ 98 41.5 [44]
partition coefficient due to the "salting-out effect" between the organic B=5:5)
7. 140 ◦ C, 10 h Hydroxylated 97 57.0 [69]
and aqueous layer [68], which enhanced the separation. After the
AlF3
addition of THF, two distinct aqueous-organic layers were visible, as 8. 150 C, 8 h

APS/SiO2-ZrO2/ 100 34.0 a
[70]
shown in Fig FS4 (supplementary file). An evident color change, i.e., PAI-HF
from clear solution to the dark-brown color of the THF layer, was visible, 9. 195 ◦ C, 1.5 h SO42−/ZrO2/ – 25.8 [71]
which indicated the transport of 5-HMF to the THF layer. 5-HMF was ZSM-5
10. 80 ◦ C, 3 h Hydrotalcite/ – 25.0 [72]
extracted multiple times from the aqueous layer, and the HPLC results Amberlyst-15
demonstrated that ~95.7% of 5-HMF was extracted in the THF layer. 11. 120 C, 5 min

Mesoporous TiO2 – 34.3 [73]
The extracted 5-HMF was separated from THF in a rotary evaporator. nanaopartclies
The purity (Fig. FS5) (supplementary file) of 5-HMF was also verified by 12. 160 ◦ C, 4 h TiO2 98 56.1 This
study
HPLC equipped with a Diode-array detector (DAD) and observed
~96.1% purity. The results depicted that the liquid-liquid extraction a
selectivity

215
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

future. References

6. Conclusions [1] A. Kahan, EIA, Today in Energy - U.S. Energy Information Administration (EIA),
(2019) 2018–2021. 〈https://www.eia.gov/todayinenergy/detail.php?id=41433〉
(accessed March 31, 2021).
The mesoporous TiO2 was synthesized by sol-gel hydrolysis tech- [2] N.A.S. Ramli, N.A.S. Amin, Fe/HY zeolite as an effective catalyst for levulinic acid
nique, and the properties of the catalyst were characterized by N2- production from glucose: characterization and catalytic performance, Appl. Catal.
B Environ. 163 (2015) 487–498, https://doi.org/10.1016/j.apcatb.2014.08.031.
physisorption, XRD, FESEM, FTIR, NH3-TPD, Py-FTIR, etc. XRD results [3] Y. Te Liao, B.M. Matsagar, K.C.W. Wu, Metal-organic framework (MOF)-Derived
revealed the anatase phase of the tetragonal TiO2. The Py-FTIR effective solid catalysts for valorization of lignocellulosic biomass, ACS Sustain.
confirmed the existence of both the B-sites and L-sites in the TiO2 Chem. Eng. 6 (2018) 13628–13643, https://doi.org/10.1021/
acssuschemeng.8b03683.
catalyst, and the obtained B/L ratio was ~2.73. It was observed that the [4] C. García-Sancho, I. Fúnez-Núñez, R. Moreno-Tost, J. Santamaría-González,
B/L ratio played a crucial role in the transformation of glucose to 5- E. Pérez-Inestrosa, J.L.G. Fierro, P. Maireles-Torres, Beneficial effects of calcium
HMF. The reaction parameters, including catalyst amount, tempera- chloride on glucose dehydration to 5-hydroxymethylfurfural in the presence of
alumina as catalyst, Appl. Catal. B Environ. 206 (2017) 617–625, https://doi.org/
ture, time, and glucose concentration, were optimized using RSM with a
10.1016/j.apcatb.2017.01.065.
5-level CCD. The empirical model equations for all the responses were [5] Y. Zhang, B. Li, Y. Wei, C. Yan, M. Meng, Y. Yan, Direct synthesis of metal-organic
developed by RSM software, and the models were found to be statisti- frameworks catalysts with tunable acid–base strength for glucose dehydration to 5-
cally well fitted with an R2 value of ~0.9 for the primary reaction hydroxymethylfurfural, J. Taiwan Inst. Chem. Eng. 96 (2019) 93–103, https://doi.
org/10.1016/j.jtice.2018.12.020.
products. At the optimum reaction condition suggested by RSM, a higher [6] S. Mérida-Morales, C. García-Sancho, M. Oregui-Bengoechea, M.J. Ginés-Molina, J.
yield (~56%) of 5-HMF and glucose conversion (~98%) was achieved. A. Cecilia, P.L. Arias, R. Moreno-Tost, P. Maireles-Torres, Influence of morphology
The relative error between the experimental and predicted response for of zirconium-doped mesoporous silicas on 5-hydroxymethylfurfural production
from mono-, di- and polysaccharides, Catal. Today 367 (2021) 297–309, https://
the glucose conversion and product (5-HMF and LA) yield was found to doi.org/10.1016/j.cattod.2020.02.029.
be in the acceptable range (< 2%). However, a little higher relative error [7] Z. Wen, L. Yu, F. Mai, Z. Ma, H. Chen, Y. Li, Catalytic conversion of
(~6%) was observed for FA yield because of the decomposition of FA at microcrystalline cellulose to glucose and 5-hydroxymethylfurfural over a niobic
acid catalyst, Ind. Eng. Chem. Res. 58 (2019) 17675–17681, https://doi.org/
high temperatures. The catalysts recycle study revealed that humin 10.1021/acs.iecr.9b03824.
removal from the active centers of the titania catalyst was a very crucial [8] J. Song, H. Fan, J. Ma, B. Han, Conversion of glucose and cellulose into value-
step. The regenerated catalyst showed a similar yield of 5-HMF even added products in water and ionic liquids, Green Chem. 15 (2013) 2619–2635,
https://doi.org/10.1039/c3gc41141a.
after the 3rd cycle. The Power-law kinetic study suggested the first-order [9] A. Rusanen, R. Lahti, K. Lappalainen, J. Kärkkäinen, T. Hu, H. Romar, U. Lassi,
dependency concerning glucose with a low activation energy of Catalytic conversion of glucose to 5-hydroxymethylfurfural over biomass-based
67.5 kJ mol−1, which demonstrated the promising activity of the TiO2 activated carbon catalyst, Catal. Today 357 (2020) 94–101, https://doi.org/
10.1016/j.cattod.2019.02.040.
catalyst. Finally, the new liquid-liquid extraction process proposed was
[10] G. Qiu, B. Chen, C. Huang, N. Liu, X. Sun, Tin-modified ionic liquid polymer: a
very effective in extracting the 5-HMF with high purity. novel and efficient catalyst for synthesis of 5-hydroxymethylfurfural from glucose,
Fuel 268 (2020), 117136, https://doi.org/10.1016/j.fuel.2020.117136.
Funding [11] G. Morales, J.A. Melero, M. Paniagua, J. Iglesias, B. Hernández, M. Sanz, Sulfonic
acid heterogeneous catalysts for dehydration of C6-monosaccharides to 5-hydrox-
ymethylfurfural in dimethyl sulfoxide, Cuihua Xuebao/Chin. J. Catal. 35 (2014)
Indian Institute of Technology Roorkee, Uttarakhand, India, sup- 644–655, https://doi.org/10.1016/s1872-2067(14)60020-6.
ported this work, and the Ministry of Education, Government of India, [12] F. Yang, J. Weng, J. Ding, Z. Zhao, L. Qin, F. Xia, Effective conversion of
saccharides into hydroxymethylfurfural catalyzed by a natural clay, attapulgite,
provided the student fellowship. Renew. Energy 151 (2020) 829–836, https://doi.org/10.1016/j.
renene.2019.11.084.
CRediT authorship contribution statement [13] G. Delgado Martin, C.E. Bounoukta, F. Ammari, M.I. Domínguez, A. Monzón,
S. Ivanova, M.Á. Centeno, Fructose dehydration reaction over functionalized
nanographitic catalysts in MIBK/H2O biphasic system, Catal. Today 366 (2021)
Richa Tomer: Conceptualization, Data curation, Formal analysis, 68–76, https://doi.org/10.1016/j.cattod.2020.03.016.
Investigation, Methodology, Validation, Visualization, Writing – orig- [14] G. Portillo Perez, M.J. Dumont, Production of HMF in high yield using a low cost
and recyclable carbonaceous catalyst, Chem. Eng. J. 382 (2020), 122766, https://
inal draft, Writing – review & editing. Prakash Biswas: Conceptuali- doi.org/10.1016/j.cej.2019.122766.
zation, Formal analysis, Funding acquisition, Investigation, [15] B. Das, K. Mohanty, Microwave induced one-pot conversion of D-glucose to 5-
Methodology, Project administration, Resources, Supervision, Writing – hydroxymethylfurfural by a novel sulfate-functionalized Sn-red mud catalyst,
Sustain. Energy Fuels 4 (2020) 6030–6044, https://doi.org/10.1039/
review & editing.
D0SE01476A.
[16] H. Yan, Y. Yang, D. Tong, X. Xiang, C. Hu, Catalytic conversion of glucose to 5-
Declaration of Competing Interest hydroxymethylfurfural over SO42-/ZrO2 and SO42-/ZrO2-Al2O3 solid acid catalysts,
Catal. Commun. 10 (2009) 1558–1563, https://doi.org/10.1016/j.
catcom.2009.04.020.
The authors declare that they have no known competing financial [17] Y. Qu, C. Huang, Y. Song, J. Zhang, B. Chen, Efficient dehydration of glucose to 5-
interests or personal relationships that could have appeared to influence hydroxymethylfurfural catalyzed by the ionic liquid,1-hydroxyethyl-3-
the work reported in this paper. methylimidazolium tetrafluoroborate, Bioresour. Technol. 121 (2012) 462–466,
https://doi.org/10.1016/j.biortech.2012.06.081.
[18] K. Nakajima, R. Noma, M. Kitano, M. Hara, Selective glucose transformation by
Acknowledgments titania as a heterogeneous Lewis acid catalyst, J. Mol. Catal. A Chem. 388–389
(2014) 100–105, https://doi.org/10.1016/j.molcata.2013.09.012.
[19] D. Chen, F. Liang, D. Feng, M. Xian, H. Zhang, H. Liu, F. Du, An efficient route from
The authors generously acknowledge the Ministry of Education, reproducible glucose to 5-hydroxymethylfurfural catalyzed by porous coordination
Government of India, for providing student fellowship and the Indian polymer heterogeneous catalysts, Chem. Eng. J. 300 (2016) 177–184, https://doi.
Institute of Technology Roorkee, Uttarakhand, for supporting the org/10.1016/j.cej.2016.04.039.
[20] K.T.V. Rao, S. Souzanchi, Z. Yuan, C. Xu, One-pot sol-gel synthesis of a phosphated
essential resources to conduct the work. TiO2 catalyst for conversion of monosaccharide, disaccharides, and
polysaccharides to 5-hydroxymethylfurfural, New J. Chem. 43 (2019)
Appendix A. Supporting information 12483–12493, https://doi.org/10.1039/c9nj01677e.
[21] S. Songtawee, B. Rungtaweevoranit, C. Klaysom, K. Faungnawakij, Tuning
Brønsted and Lewis acidity on phosphated titanium dioxides for efficient
Supplementary data associated with this article can be found in the conversion of glucose to 5-hydroxymethylfurfural, RSC Adv. 11 (2021)
online version at doi:10.1016/j.cattod.2022.03.019. 29196–29206, https://doi.org/10.1039/d1ra06002c.
[22] P. Ganji, S. Roy, Trade-off between acidic sites and crystallinity of the WO3-TiO2
catalyst toward dehydration of glucose to 5-hydroxymethylfurfural, Energy Fuels
33 (2019) 5293–5303, https://doi.org/10.1021/acs.energyfuels.9b00461.

216
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

[23] H. Li, M. Vrinat, G. Berhault, D. Li, H. Nie, P. Afanasiev, Hydrothermal synthesis [47] K. Nagaveni, M.S. Hegde, G. Madras, Structure and photocatalytic activity of
and acidity characterization of TiO2 polymorphs, Mater. Res. Bull. 48 (2013) Ti1−xMx O2±δ (M = W, V, Ce, Zr, Fe, and Cu) synthesized by solution combustion
3374–3382, https://doi.org/10.1016/j.materresbull.2013.05.017. method, J. Phys. Chem. B 108 (2004) 20204–20212, https://doi.org/10.1021/
[24] J. Zhao, C. Zhou, C. He, Y. Dai, X. Jia, Y. Yang, Efficient dehydration of fructose to jp047917v.
5-hydroxymethylfurfural over sulfonated carbon sphere solid acid catalysts, Catal. [48] L. Vera Candioti, M.M. De Zan, M.S. Cámara, H.C. Goicoechea, Experimental
Today 264 (2016) 123–130, https://doi.org/10.1016/j.cattod.2015.07.005. design and multiple response optimization using the desirability function in
[25] M. Lopes, K. Dussan, J.J. Leahy, V.T. da Silva, Conversion of D-glucose to 5- analytical methods development, Talanta 124 (2014) 123–138, https://doi.org/
hydroxymethylfurfural using Al2O3-promoted sulphated tin oxide as catalyst, 10.1016/j.talanta.2014.01.034.
Catal. Today 279 (2017) 233–243, https://doi.org/10.1016/j.cattod.2016.05.030. [49] S. Madadi, L. Charbonneau, J.Y. Bergeron, S. Kaliaguine, Aerobic epoxidation of
[26] F. Shahangi, A. Najafi Chermahini, M. Saraji, Dehydration of fructose and glucose limonene using cobalt substituted mesoporous SBA-16 Part 1: optimization via
to 5-hydroxymethylfurfural over Al-KCC-1 silica, J. Energy Chem. 27 (2018) Response Surface Methodology (RSM), Appl. Catal. B Environ. 260 (2020),
769–780, https://doi.org/10.1016/j.jechem.2017.06.004. 118049, https://doi.org/10.1016/j.apcatb.2019.118049.
[27] I.K.M. Yu, D.C.W. Tsang, A.C.K. Yip, Z. Su, K. De Oliveira Vigier, F. Jérôme, C. [50] N. Bala, M. Napiah, I. Kamaruddin, Nanosilica composite asphalt mixtures
S. Poon, Y.S. Ok, Organic acid-regulated Lewis acidity for selective catalytic performance-based design and optimisation using response surface methodology,
hydroxymethylfurfural production from rice waste: an experimental- Int. J. Pavement Eng. 21 (2020) 29–40, https://doi.org/10.1080/
computational study, ACS Sustain. Chem. Eng. 7 (2019) 1437–1446, https://doi. 10298436.2018.1435881.
org/10.1021/acssuschemeng.8b05141. [51] M.P.J. Kizhakedathil, S.D. Chandrasekaran, Media optimization for extracellular
[28] S.P. Teong, G. Yi, Y. Zhang, Hydroxymethylfurfural production from bioresources: amylase production by Pseudomonas balearica vitps19 using response surface
past, present and future, Green Chem. 16 (2014) 2015, https://doi.org/10.1039/ methodology, Front. Biol. 13 (2018) 123–129, https://doi.org/10.1007/s11515-
c3gc42018c. 018-1485-3.
[29] J.B. Binder, R.T. Raines, Simple chemical transformation of lignocellulosic biomass [52] N.A.S. Ramli, N.A.S. Amin, Thermo-kinetic assessment of glucose decomposition to
into furans for fuels and chemicals, J. Am. Chem. Soc. 131 (2009) 1979–1985, 5-hydroxymethyl furfural and levulinic acid over acidic functionalized ionic liquid,
https://doi.org/10.1021/ja808537j. Chem. Eng. J. 335 (2018) 221–230, https://doi.org/10.1016/j.cej.2017.10.112.
[30] P. Arun, S.M. Pudi, P. Biswas, Acetylation of glycerol over sulfated alumina: [53] B. Girisuta, L.P.B.M. Janssen, H.J. Heeres, Green chemicals: a kinetic study on the
reaction parameter study and optimization using response surface methodology, conversion of glucose to levulinic acid, Chem. Eng. Res. Des. 84 (2006) 339–349,
Energy Fuels 30 (2016) 584–593, https://doi.org/10.1021/acs. https://doi.org/10.1205/cherd05038.
energyfuels.5b01901. [54] J. Cao, M. Ma, J. Liu, Y. Yang, H. Liu, X. Xu, J. Huang, H. Yue, G. Tian, S. Feng,
[31] A. Al Ameen, S. Mondal, S.M. Pudi, N.N. Pandhare, P. Biswas, Liquid phase Highly effective transformation of carbohydrates to 5-Hydroxymethylfurfural with
hydrogenolysis of glycerol over highly active 50%Cu-Zn(8:2)/MgO catalyst: Al- montmorillonite as catalyst, Appl. Catal. A Gen. 571 (2019) 96–101, https://
reaction parameter optimization by using response surface methodology, Energy doi.org/10.1016/j.apcata.2018.12.011.
Fuels 31 (2017) 8521–8533, https://doi.org/10.1021/acs.energyfuels.7b00766. [55] G. Tsilomelekis, T.R. Josephson, V. Nikolakis, S. Caratzoulas, Origin of 5-hydrox-
[32] S.P. Utami, N.S. Amin, Optimization of glucose conversion to 5-hydroxymethyl- ymethylfurfural stability in water/dimethyl sulfoxide mixtures, ChemSusChem 7
fulfural using [BMIM]Cl with ytterbium triflate, Ind. Crop. Prod. 41 (2013) 64–70, (2014) 117–126, https://doi.org/10.1002/cssc.201300786.
https://doi.org/10.1016/j.indcrop.2012.03.036. [56] M.R. Whitaker, A. Parulkar, P. Ranadive, R. Joshi, N.A. Brunelli, Examining acid
[33] M.N. Catrinck, E.S. Ribeiro, R.S. Monteiro, R.M. Ribas, M.H.P. Barbosa, R. formation during the selective dehydration of fructose to 5-hydroxymethylfurfural
F. Teófilo, Direct conversion of glucose to 5-hydroxymethylfurfural using a mixture in dimethyl sulfoxide and water, ChemSusChem 12 (2019) 2211–2219, https://
of niobic acid and niobium phosphate as a solid acid catalyst, Fuel 210 (2017) doi.org/10.1002/cssc.201803013.
67–74, https://doi.org/10.1016/j.fuel.2017.08.035. [57] N.R. Peela, S.K. Yedla, B. Velaga, A. Kumar, A.K. Golder, Choline chloride
[34] A. Sarwono, Z. Man, A. Idris, A.S. Khan, N. Muhammad, C.D. Wilfred, Optimization functionalized zeolites for the conversion of biomass derivatives to 5-
of ionic liquid assisted sugar conversion and nanofiltration membrane separation hydroxymethylfurfural, Appl. Catal. A Gen. 580 (2019) 59–70, https://doi.org/
for 5-hydroxymethylfurfural, J. Ind. Eng. Chem. 69 (2019) 171–178, https://doi. 10.1016/j.apcata.2019.05.005.
org/10.1016/j.jiec.2018.09.020. [58] A. Sarwono, Z. Man, N. Muhammad, A.S. Khan, W.S.W. Hamzah, A.H.A. Rahim,
[35] R. Tomer, P. Biswas, Dehydration of glucose/fructose to 5-hydroxymethylfurfural Z. Ullah, C.D. Wilfred, A new approach of probe sonication assisted ionic liquid
(5-HMF) over an easily recyclable sulfated titania (SO42-/TiO2) catalyst, New J. conversion of glucose, cellulose and biomass into 5-hydroxymethylfurfural,
Chem. 44 (2020) 20734–20750, https://doi.org/10.1039/d0nj04151c. Ultrason. Sonochem. 37 (2017) 310–319, https://doi.org/10.1016/j.
[36] M. Faycal Atitar, A.A. Ismail, S.A. Al-Sayari, D. Bahnemann, D. Afanasev, A. ultsonch.2017.01.028.
V. Emeline, Mesoporous TiO2 nanocrystals as efficient photocatalysts: impact of [59] F.R. Tao, C. Zhuang, Y.Z. Cui, J. Xu, Dehydration of glucose into 5-hydroxyme-
calcination temperature and phase transformation on photocatalytic performance, thylfurfural in SO3H-functionalized ionic liquids, Chin. Chem. Lett. 25 (2014)
Chem. Eng. J. 264 (2015) 417–424, https://doi.org/10.1016/j.cej.2014.11.075. 757–761, https://doi.org/10.1016/j.cclet.2014.01.044.
[37] K.S.W. Sing, R.T. Williams, Physisorption hysteresis loops and the characterization [60] B. Velaga, N.R. Peela, Seed-assisted and OSDA-free synthesis of H-mordenite
of nanoporous materials, Adsorpt. Sci. Technol. 22 (2004) 773–782, https://doi. zeolites for efficient production of 5-hydroxymethylfurfural from glucose,
org/10.1260/0263617053499032. Microporous Mesoporous Mater. 279 (2019) 211–219, https://doi.org/10.1016/j.
[38] Y. Niu, F. Li, K. Yang, T. Qiu, R. Wang, C. Lin, Preparation and characterization of micromeso.2018.12.028.
sulfated TiO2 with rhodium modification used in esterification reaction and [61] Z. Tang, J. Su, Direct conversion of cellulose to 5-hydroxymethylfurfural (HMF)
decomposition of methyl orange, Chin. J. Chem. Eng. 24 (2016) 767–774, https:// using an efficient and inexpensive boehmite catalyst, Carbohydr. Res. 481 (2019)
doi.org/10.1016/j.cjche.2015.12.016. 52–59, https://doi.org/10.1016/j.carres.2019.06.010.
[39] K.J.A. Raj, B. Viswanathan, Single-step synthesis and structural study of [62] Y. Nie, Q. Hou, C. Bai, H. Qian, X. Bai, M. Ju, Transformation of carbohydrates to 5-
mesoporous sulfated titania nanopowder by a controlled hydrolysis process, ACS hydroxymethylfurfural with high efficiency by tandem catalysis, J. Clean. Prod.
Appl. Mater. Interfaces 1 (2009) 2462–2469, https://doi.org/10.1021/ 274 (2020), 123023, https://doi.org/10.1016/j.jclepro.2020.123023.
am900437u. [63] Y. Shao, W. Du, Z. Gao, K. Sun, Z. Zhang, Q. Li, L. Zhang, S. Zhang, Q. Liu, X. Hu,
[40] S.S. Muniandy, N.H. Mohd Kaus, Z.T. Jiang, M. Altarawneh, H.L. Lee, Green Sulfated TiO2 nanosheets catalyzing conversion of biomass derivatives: influences
synthesis of mesoporous anatase TiO2 nanoparticles and their photocatalytic of the sulfation on distribution of Brønsted and Lewis acidic sites, J. Chem.
activities, RSC Adv. 7 (2017) 48083–48094, https://doi.org/10.1039/c7ra08187a. Technol. Biotechnol. 95 (2020) 1337–1347, https://doi.org/10.1002/jctb.6318.
[41] D.K. Pandey, P. Biswas, Production of propylene glycol (1,2-propanediol) by the [64] K. Saravanan, K.S. Park, S. Jeon, J.W. Bae, Aqueous phase synthesis of
hydrogenolysis of glycerol in a fixed-bed downflow tubular reactor over a highly 5–hydroxymethylfurfural from glucose over large pore mesoporous zirconium
effective Cu-Zn bifunctional catalyst: effect of an acidic/basic support, New J. phosphates: effect of calcination temperature, ACS Omega 3 (2018) 808–820,
Chem. 43 (2019) 10073–10086, https://doi.org/10.1039/c9nj01180c. https://doi.org/10.1021/acsomega.7b01357.
[42] F.F. Oloye, I.A. Ololade, Influence of molybdenum loadings on the properties of [65] R.H. Myers, D.C. Montgomery, C.M. Anderson-Cook, Response surface
MoO3/zirconia catalysts, Chem 1 (2018) 119–126, https://doi.org/10.1007/ methodology: process and product optimization using designed experiments.
s42250-018-0016-6. Third, Wiley, 2009.
[43] J. Guo, S. Zhu, Y. Cen, Z. Qin, J. Wang, W. Fan, Ordered mesoporous Nb–W oxides [66] S.M. Patil, S.P. Deshmukh, K.V. More, V.B. Shevale, S.B. Mullani, A.G. Dhodamani,
for the conversion of glucose to fructose, mannose and 5-hydroxymethylfurfural, S.D. Delekar, Sulfated TiO2/WO3 nanocomposite:an efficient photocatalyst for
Appl. Catal. B Environ. 200 (2017) 611–619, https://doi.org/10.1016/j. degradation of Congo red and methyl red dyes under visible light irradiation,
apcatb.2016.07.051. Mater. Chem. Phys. 225 (2019) 247–255, https://doi.org/10.1016/j.
[44] J. Liu, H. Li, Y.C. Liu, Y.M. Lu, J. He, X.F. Liu, Z.B. Wu, S. Yang, Catalytic matchemphys.2018.12.041.
conversion of glucose to 5-hydroxymethylfurfural over nano-sized mesoporous [67] Z. Cao, Z. Fan, Y. Chen, M. Li, T. Shen, C. Zhu, H. Ying, Efficient preparation of 5-
Al2O3-B2O3 solid acids, Catal. Commun. 62 (2015) 19–23, https://doi.org/ hydroxymethylfurfural from cellulose in a biphasic system over hafnyl phosphates,
10.1016/j.catcom.2015.01.008. Appl. Catal. B Environ. 244 (2019) 170–177, https://doi.org/10.1016/j.
[45] F. Kılıç Dokan, M. Kuru, A new approach to optimize the synthesis parameters of apcatb.2018.11.019.
TiO2 microsphere and development of photocatalytic performance, J. Mater. Sci. [68] D. Gupta, E. Ahmad, K.K. Pant, B. Saha, Efficient utilization of potash alum as a
Mater. Electron. 32 (2021) 640–655, https://doi.org/10.1007/s10854-020-04845- green catalyst for production of furfural, 5-hydroxymethylfurfural and levulinic
y. acid from mono-sugars, RSC Adv. 7 (2017) 41973–41979, https://doi.org/
[46] T. Narabe, M. Hagiwara, S. Fujihara, A biphasic sol–gel route to synthesize anatase 10.1039/c7ra07147g.
TiO2 particles under controlled conditions and their DSSC application, J. Asian [69] Y.M. Lu, H. Li, J. He, Y.X. Liu, Z.B. Wu, D.Y. Hu, S. Yang, Efficient conversion of
Ceram. Soc. 5 (2017) 427–435, https://doi.org/10.1016/j.jascer.2017.09.005. glucose to 5-hydroxymethylfurfural using bifunctional partially hydroxylated AlF3,
RSC Adv. 6 (2016) 12782–12787, https://doi.org/10.1039/c5ra24013a.

217
R. Tomer and P. Biswas Catalysis Today 404 (2022) 201–218

[70] Y. He, A.K. Itta, A.A. Alwakwak, M. Huang, F. Rezaei, A.A. Rownaghi, [72] M. Ohara, A. Takagaki, S. Nishimura, K. Ebitani, Syntheses of 5-hydroxymethyl-
Aminosilane-grafted SiO2-ZrO2 polymer hollow fibers as bifunctional microfluidic furfural and levoglucosan by selective dehydration of glucose using solid acid and
reactor for tandem reaction of glucose and fructose to 5-hydroxymethylfurfural, base catalysts, Appl. Catal. A Gen. 383 (2010) 149–155, https://doi.org/10.1016/j.
ACS Sustain, Chem. Eng. 6 (2018) 17211–17219, https://doi.org/10.1021/ apcata.2010.05.040.
acssuschemeng.8b04555. [73] S. Dutta, S. De, A.K. Patra, M. Sasidharan, A. Bhaumik, B. Saha, Microwave assisted
[71] Y. Feng, M. Li, Z. Gao, X. Zhang, X. Zeng, Y. Sun, X. Tang, T. Lei, L. Lin, rapid conversion of carbohydrates into 5-hydroxymethylfurfural catalyzed by
Development of betaine-based sustainable catalysts for green conversion of mesoporous TiO2 nanoparticles, Appl. Catal. A Gen. 409–410 (2011) 133–139,
carbohydrates and biomass into 5-hydroxymethylfurfural, ChemSusChem 12 https://doi.org/10.1016/j.apcata.2011.09.037.
(2019) 495–502, https://doi.org/10.1002/cssc.201802342.

218

You might also like