Flame-made amorphous solid acids with tunable acidity for the aqueous conversion of glucose to levulinic acid

Article abstract of DOI:10.1039/c9gc02567g

Solid acids of amorphous silica-alumina (a-SA) and amorphous silica-alumina-phosphate (a-SAPO) were prepared by flame spray pyrolysis (FSP). Careful tuning of the acidity of the solid acids was enabled by capitalizing on the advantage of FSP in preserving the metal stoichiometry (i.e., Si, Al, P) in the product nanoparticles. Although the amount of acids on these non-porous solid acids is an order of magnitude lower than the well-recognized strong acidic ZSM-5 zeolite, both exhibit comparable acid strengths. The a-SA and a-SAPO were characterized by a mixture of Br?nsted (B) and Lewis (L) acids, and the B/L ratios were composition-tunable. The highest B/L ratios were recorded for a-SA (Al/(Al + Si) = 0.4) and a-SAPO (Si/(P + Si) = 0.25), giving the highest yields of levulinic acid (≥40% carbon yield) from the conversion of glucose in the aqueous phase without requiring the addition of liquid acids or metal halides. Under the same conditions, the almost exclusive Br?nsted acid ZSM-5 yielded only 17% levulinic acid. The FSP-made solid acid catalyst exhibited good reusability over at least 4 consecutive runs.

Full text of DOI:10.1039/c9gc02567g

Green Chemistry
View Article Online
View Journal
PAPER
Flame-made amorphous solid acids with tunable
acidity for the aqueous conversion of glucose to
levulinic acid†
Cite this: DOI: 10.1039/c9gc02567g
c
Gein Khai Beh,‡a Chang Ting Wang,‡a Kyungduk Kim,‡b Jiangtao Qu,
b,d
Julie Cairney,c Yun Hau Ng,a Alicia Kyoungjin An, a Ryong Ryoo,
Atsushi Urakawa e,f and Wey Yang Teoh
*
a,g
Solid acids of amorphous silica-alumina (a-SA) and amorphous silica-alumina-phosphate (a-SAPO) were
prepared by flame spray pyrolysis (FSP). Careful tuning of the acidity of the solid acids was enabled by
capitalizing on the advantage of FSP in preserving the metal stoichiometry (i.e., Si, Al, P) in the product
nanoparticles. Although the amount of acids on these non-porous solid acids is an order of magnitude
lower than the well-recognized strong acidic ZSM-5 zeolite, both exhibit comparable acid strengths. The
a-SA and a-SAPO were characterized by a mixture of Brønsted (B) and Lewis (L) acids, and the B/L ratios
were composition-tunable. The highest B/L ratios were recorded for a-SA (Al/(Al + Si) = 0.4) and a-SAPO
(Si/(P + Si) = 0.25), giving the highest yields of levulinic acid (≥40% carbon yield) from the conversion of
glucose in the aqueous phase without requiring the addition of liquid acids or metal halides. Under the
same conditions, the almost exclusive Brønsted acid ZSM-5 yielded only 17% levulinic acid. The FSP-
made solid acid catalyst exhibited good reusability over at least 4 consecutive runs.
Received 24th July 2019,
Accepted 13th November 2019
DOI: 10.1039/c9gc02567g
rsc.li/greenchem
forms make it difficult to yield fuels of high calorific values, or
value-added chemicals of high quality. Therefore, they need to
1. Introduction
One of the biggest challenges in the 21st century is to tran- be broken down into monosaccharide units such as glucose
sition from a fossil-based economy to one based on renewable and fructose that can be further converted into a series of
resources that, besides solar and wind, should include robust platform chemicals. Among them, levulinic acid (LA) is
biomass, the most abundant available raw material in listed as one of the top twelve most promising building block
nature.1,2 Agricultural waste products especially those of ligno- chemicals as compiled by the US Department of Energy.5
cellulosic nature from plants such as stalks, wood, kernels and
A wide range of valuable commodity chemicals can indeed
straws are precious renewable biomass feedstocks that are still be derived from LA for use as fuel additives, pharmaceuticals,
underexploited.3,4 The chemical structures in their existing fragrances and plasticizers.6 For example, the hydrogenation
of LA produces γ-valerolactone for use in biofuel-gasoline
blends or as a general solvent for lacquers, adhesives and
insecticides; the conversion to methyl tetrahydrofuran (via 1,4-
pentanediol) for use as fuel blends and as monomers for
polymer fibers; the esterification of LA to levulinic ester as a
aSchool of Energy and Environment, City University of Hong Kong, Kowloon,
Hong Kong S.A.R. E-mail: wy.teoh@unsw.edu.au
bCenter for Nanomaterials and Chemical Reactions, Institute of Basic Science (IBS),
Daejeon 34141, Republic of Korea
cAustralian Centre for Microscopy and Microanalysis, School of Aerospace,
Mechanical and Mechatronic Engineering, The University of Sydney, NSW 2006,
Australia
diesel additive; and the condensation of LA with esters to
produce α-methylene-γ-valerolactone, new acrylic
a
monomer.4,6 Traditionally, the conversion of glucose or fruc-
tose to LA is catalyzed by liquid Brønsted acids such as HCl,
H2SO4, and H3PO4, in combination with Lewis acids in the
form of metal chlorides (e.g., CrCl3, AlCl3, CuCl2, FeCl3), where
the standard has been to use the toxic CrCl3.7 The reaction is
well-established to consist of three sequential steps, namely,
the isomerization of glucose to fructose, the dehydration of
fructose to 5-hydroxymethylfurfural (5-HMF), and the rehydra-
tion of 5-HMF to LA and formic acid (FA) (Scheme 1).7–9 The
decompositions of glucose, fructose and 5-HMF also occur in
dDepartment of Chemistry, Korea Advanced Institute of Science and Technology
(KAIST), Daejeon 34141, Republic of Korea
eCatalysis Engineering, Department of Chemical Engineering, Delft University of
Technology, Van der Maasweg 9, 2629 HZ Delft, The Netherlands
fInstitute of Chemical Research of Catalonia (ICIQ), The Barcelona Institute of
Science and Technology, 43007 Tarragona, Spain
gSchool of Chemical Engineering, The University of New South Wales, Sydney,
NSW 2052, Australia
†Electronic supplementary information (ESI) available. See DOI: 10.1039/
c9gc02567g
‡These authors contributed equally to this work.
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
Scheme 1 Overall reaction pathway for the aqueous conversion of glucose to levulinic acid on solid acids.
parallel, leading to the formation of humins as byproducts. In non-porous flame-made solid acids and the microporous
these sequential reactions, mixed Brønsted and Lewis acids ZSM-5 zeolite, showing a more than 2-fold increase in LA yield
are often necessary to achieve high LA yields. Prior studies in the former. To the best of our knowledge, this work is the
suggested that Lewis acids promote the isomerization of first to demonstrate ready-to-use, FSP-made catalysts for
glucose to fructose,10 while the combination of both types of glucose conversion.
acids is typically required for the fructose dehydration and its
further rehydration to LA.11,12
Compared with liquid acids, solid acid catalysts are advan-
tageous as they are easier to handle, non-corrosive, readily
separable from the end liquid or gas products, and hence easily
2. Experimental
2.1. Flame synthesis and characterization of solid acids
recyclable.13,14 For that reason, a wide range of solid acids and The liquid precursor for the synthesis of a-SA series solid acids
their modifications were developed and shown to be effective was composed of an appropriate amount of tetraethyl orthosi-
in catalysing various biomass-related conversions.15–18 Of par- licate (TEOS, 98%, Aldrich) and aluminium tri-secbutoxide
ticular interest, the ZSM-5 aluminosilicate zeolite is a popular (97%, Acros) codissolved in xylene. The total concentration of
standalone solid acid catalyst for the production of LA, albeit Al + Si was kept constant at 0.5 M. In the case of a-SAPO, the
without requiring the addition of metal halides as homo- concentration of aluminium tri-secbutoxide (97%, Acros) was
geneous phase Lewis acids.19 Other solid acids such as metal fixed at 0.25 M and appropriate amounts of tributyl phosphate
(IV) phosphates have also been demonstrated as standalone (99%, Aldrich) and TEOS were dissolved in xylene to give a
catalysts for the aqueous-phase conversion of glucose to HMF combined Al + Si + P concentration of 0.5 M. All liquid precur-
and LA.20,21 These are much cheaper alternatives compared sors were prepared in a N2-filled glovebox. During FSP,28 the
with ion-exchange polymer resins such as Amberlyst and liquid precursor was fed at a rate of 5 or 1 mL min−1 using a
Nafion, with reported LA yields of 28 and 40% (with addition syringe pump. At the FSP nozzle tip, the liquid precursor was
of 0.2 M H2SO4), respectively.21,22
dispersed by 5 L min−1 O2 (1.5 bar) and ignited by a surround-
In this paper, we report the rapid synthesis of amorphous ing oxy-methane supporting flame (1.5 L min−1 CH4/3.2 L
silica-alumina (a-SA) and amorphous silica-alumina-phosphate min−1 O2) to form the primary aerosol flame. Additional 5 L
(a-SAPO) solid acids by flame spray pyrolysis (FSP) as promis- min−1 of sheath O2 was issued through the outermost stainless
ing alternative catalysts for the aqueous conversion of glucose steel ring consisting of 12 equally spaced outlets (0.8 mm dia-
to LA. FSP is an industrially-relevant and scalable process meter). The product particles leaving the primary flame were
capable of producing high quality, size-tunable nanoparticles collected on a glassfibre filter (Whatmann GF/D) with the aid
in the order of kilograms per hour at the pilot scale.23–25 of a vacuum pump (Edwards). Commercial ZSM-5 zeolite (CBV
Another major advantage of FSP is the ability to fully preserve 2314) and Zeolite X were purchased from Zeolyst International
the metal stoichiometry in the final product catalyst and Grace Davison, respectively.
nanoparticles.26,27 As will be demonstrated in the current
work, such precise control over the stoichiometry of the was measured on
The X-ray diffraction (XRD) of the as-prepared catalysts
PANalytical X’Pert3 diffractometer
a
product is pivotal in the design of solid acids, since the equipped with a PIXCel 1D detector. Cu Kα was used as the
amount of acid is dictated by the amount of Al3+ substitution radiation source operating at 40 kV and 40 mA. The sample
in Si4+ in the case of a-SA, and the Si4+ substitution in P5+ in was scanned at 0.1° s−1 with a step size of 0.013°. The mor-
the case of a-SAPO, respectively. Systematic variation of these phologies of samples were investigated by using a JEM
substituents and thorough characterization allow insights into 2200FS transmission electron microscope (TEM) under
their acidic characteristic-activity relationships of the resultant 200 kV accelerating voltage, and the TEM specimens were pre-
flame-made solid acids. Comparison is also made between the pared by
a
standard particle preparation protocol: the
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
ethanol/particle suspension was dispersed onto
a
TEM was calculated based on the extinction coefficient established
copper grid with a lacey carbon film after sonication for by Emeis:29
30 min. The surface elemental compositions of the as-
prepared samples were characterized by X-ray photoelectron
spectroscopy (XPS) on an ESCALab220Xi (Thermo Scientific)
πR2
W
CðB; LÞ ¼ IMECðB; LÞꢀ1 ꢁ IAðB; LÞ ꢁ
ð1Þ
where C is the concentration of acid sites (μmol gcat−1), IMEC
(B, L) is the integrated molar extinction coefficient (B = 1.67,
L = 2.22 cm μmol−1), IA (B, L) is the integrated absorbance
(cm−1), R is the radius of the catalyst disk (cm) and W is the
weight of the catalyst (mg).
using monochromated Al Kα (20 eV pass energy) radiation
source. Samples were measured under ultrahigh vacuum at
<2 × 10−9 mbar. The specific surface area (SSA) of samples
was evaluated by nitrogen adsorption at 77 K, using the BET
method and measured on a Micromeritics Tristar 3000. Prior
to the measurement, the as-prepared samples were evacuated
in a Micromeritics VacPrep unit heated to 150 °C for at least
2 h to remove adsorbed moisture and volatile organics.
2.2. Catalyst assessment
The hydrothermal conversion of glucose to levulinic acid was
carried out in a Teflon-lined stainless-steel reactor with a total
volume of about 10 ml. D-Glucose (≥99.5%, Aldrich), D-fructose
(99%, Alfa Aesar) or 5-hydroxymethylfurfural (≥99%, Aldrich)
was dissolved in deionized water (18.2 MΩ, Milli-Q). The accu-
rately weighed as-prepared catalyst (5 mg) was suspended in
6 mL of the solution in a reactor and pressurized to 20 bar Ar.
The insulated reactor was heated to the target temperature
using a stirrer-hot plate (IKA) under magnetic stirring at 600
rpm. To quench the reaction, the reactor was immediately
cooled under cold running tap water. Prior to analyses, the
suspension was filtered (Pall syringe filter, 0.2 μm nylon). The
filtered sample was analyzed on a high performance liquid
chromatograph (HPLC, Acquity, Waters) equipped with an
Aminex HPX-87H column (50 °C), a photodiode array (210 nm,
PDA, Waters) and refractive index (RI, Waters) detectors. The
sample (2.5 μL) was eluted using 5 mM H2SO4 (98%, HPLC
The SSA-equivalent diameter was estimated based on dSSA
=
6/(ρ × SSA), where ρ is the mass-weighted density of the
sample assuming elementary oxide forms. Ammonia temp-
erature-programmed desorption (NH3-TPD) studies were
carried out on a Micromeritics AutoChem 2950 HP chemi-
sorption analyzer equipped with a quadrupole Microstar
mass spectrometer. Accurately weighed 0.25 g of the pelle-
tized catalyst was supported on quartz wool plug in a quartz
U-tube. The catalysts were pretreated at 400 °C (10 °C min−1
)
for 1 h under 50 mL min−1 of 5% O2/He to remove surface
carbon residues, and subsequently cooled to 50 °C. The cata-
lyst was surface saturated with NH3 by flowing 50 mL min−1
of 5% NH3/He for 1 h. Physisorbed ammonia was removed by
switching to He at the same flow rate, flowing for at least 1 h
until no change in the thermal conductivity could be
measured at the outlet gas. The TPD commences by heating
the sample at 10 °C min−1 up to 800 °C and the sample was
held at that temperature for an additional 20 min. The de-
sorption products for NH3 (m/z = 16 and 17), H2O (m/z = 18)
and O2 (m/z = 32) were monitored by using the mass
spectrometer.
grade, Aldrich) aqueous solution at a flow rate of 0.6 ml min−1
.
The conversion of reactants and product carbon yields are
defined in eqn (2) and (3), respectively:
½Reactantꢂi ꢀ ½Reactantꢂf
Conversion ¼
ꢁ 100%
ð2Þ
½Reactantꢂi
Pyridine-adsorbed Fourier-transform infrared (pyridine-
FTIR) analyses were carried out in a commercial transmission
IR cell (Dewar Transmission/Reflection Accessory, Harrick
Scientific) mounted on an FT-IR spectrometer (Vertex 70,
Bruker) under N2 purge. To avoid high IR light absorption in
the region of interest by the sample itself and achieve good
sample homogeneity, a two-step pelletization procedure was
employed. First, 200 mg of dried KBr powder was pressed at 5
ton for 5 min and shaped into a round disk of 13 mm dia-
meter. Afterwards 6 mg of a powder sample was homoge-
neously placed over the KBr disk for further pressing at 1.5 ton
for 1 min. In this way, the KBr disk functioned as the support
without interacting chemically with the sample. The pelletized
sample was placed in the cell, heated to 150 °C, and kept at
½Productꢂ ꢁ No: of carbon in the product molecule
Yield ¼
½Reactantꢂi ꢁ No: of carbon in the reactant molecule
ꢁ 100%
ð3Þ
where subscripts i and f refer to the initial and final states of
the reaction, respectively.
3. Results and discussion
3.1. Physicochemical characteristics of a-SA and a-SAPO solid
acids
that temperature for 2 h under vacuum (ca. 5 Pa) and a back- Amorphous silica-alumina (a-SA) solid acids with different Al/
ground spectrum was recorded. Subsequently, 2500 Pa of pyri- (Al + Si) molar ratios, ranging from 0 (pure SiO2), 0.2, 0.4, 0.6,
dine was dosed in the cell and stabilized for 30 min. Finally, 0.8 to 1.0 (pure Al2O3), were synthesized using the one-step
the cell was evacuated to 5 Pa to remove gaseous and physi- FSP. For brevity, the samples are labelled a-SAx, where x refers
sorbed pyridine, and the spectrum of chemisorbed pyridine to the Al/(Al + Si) percentage molar ratio. It is presented in
on the sample was also recorded at 150 °C. Each spectrum was such a way since Al is the source of acid, and the total amount
measured for 15 min at a resolution of 2 cm−1 using a DTGS of Al and Si is kept constant in all a-SA samples. As shown from
detector. The amount of Brønsted (B) and Lewis (L) acid sites the X-ray diffraction (XRD) spectra in Fig. 1a, except for the pure
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
Fig. 1 High resolution XRD spectra of (a) a-SA and (b) a-SAPO series of flame-made solid acids with different compositions. Samples are labelled
a-SAx and a-SAPOy, where x and y represent the percentage molar ratios of Al/(Al + Si) and Si/(Si + P), respectively. The XRD peaks are labelled
α-Al2O3 (open circle, ICSD: 068591), γ-Al2O3 (open square, ICSD: 068771) and AlPO4 (open diamond, ICSD: 71435).
Al2O3, which is composed of mixed α- (2θ = 28.7 and 56.2°) particles. Any surface pores that may form during the flame
and γ-Al2O3 phases (2θ = 32.5, 36.8, 39.5, 45.9, 61.3 and 67°), synthesis would collapse and sinter due to the high tempera-
all Si-containing samples were predominantly amorphous as ture environment.33,34 Small addition of Al in a-SA increased
reflected by the broad amorphous hump (2θ = 17–40°). the specific surface area (SSA) up to 407 m2 gcat (for a-SA20),
−1
Although the FSP generally produces highly crystalline nano- beyond which the SSA gradually decreased (Fig. 2b). This
particles as a result of the high temperature synthesis (with corresponds to an initial decrease in the SSA-equivalent dia-
the highest temperature point reaching up to 2800 K),25 SiO2 meter, dSSA, from 7 nm for a-SA0 to 5 nm for a-SA20 and fol-
and related composites have been one of the few lowed by the systematic increase up to 9 nm for the pure
exceptions,30–32 where the typically short flame residence time Al2O3. The increase in SSA with an initial small addition of the
(in the range of milliseconds) is insufficient to promote the external element, in this case Al, is commonly observed for
35
crystallization to quartz. Furthermore, the presence of SiO2 dis- flame-made catalysts, e.g., Al2O3/Ce0.5Zr0.5O2
and Fe2O3/
rupted the long-range order of the Al2O3 unit cells and thus TiO2.27 This lies within the narrow concentration region at
distorted the overall crystallite structure, as evidenced from the which dopants can stabilize the primary particles under the
broadening and disappearance of Al2O3 diffraction peaks with high temperature flame synthesis, but when excess dopants
increasing SiO2 content for a-SA20 and above. Only trace are present this leads to severe sintering and coalescence. In
α-Al2O3 clusters could be observed under high resolution XRD. the case of a-SAPO, the increase in the Si content systemati-
Similar synthesis was carried out to produce amorphous cally increases the SSA from 58 m2 g−1 for pure AlPO4 to
silica-alumina-phosphate (a-SAPOy) solid acids, where
y
268 m2 g−1 for a-SAPO100, or equivalent to a decrease in dSSA
denotes the Si/(P + Si) percentage molar ratio as analogous to from 34 to 6 nm, respectively (Fig. 2d). A high Si content
the denotation used for a-SA. Here, the total amount of P and appears beneficial in attaining a high SSA for both cases of
Si is kept constant and in equimolar with Al in all a-SAPO a-SA and a-SAPO, inferring suppressed aerosol coalescence
samples. The molar ratios of a-SAPO were varied from Si/(P + during the FSP.
Si) of 0 (AlPO4), 0.25, 0.40, 0.67, 0.80 to 1.0 (equivalent to
By reducing the parametric liquid precursor feed flow
a-SA50). As shown from the XRD in Fig. 1b, in the absence of rate during the FSP synthesis from 5 to 1 ml min−1, this
Si, i.e., a-SAPO0, the monoclinic aluminophosphate (AlPO4) directly decreased the rate of combustion enthalpy resulting
diffraction peaks could be measured in coexistence with a in an aerosol flame that is cooler with shorter residence
largely amorphous phase. As in the case of the a-SA series dis- time and lower metal concentrations. Consequently,
a
cussed above, the addition of Si disrupted the orderly arrange- smaller particle size (4 nm) and much higher surface area
ment of the AlPO4 unit cells thereby resulting in the formation (478 m2 gcat−1) were obtained for a-SA40(HSA – high surface
of fully amorphous solid acids. Because of the high tempera- area) compared with those of the standard a-SA40 (7 nm,
ture and oxygen-rich flame synthesis, all hydrocarbon com- 258 m2 gcat−1) (Fig. 2b). Note: As will be shown below, the
ponents present in the precursors were fully combusted and increase in SSA was able to increase the amount of surface
yielded pristine a-SA and a-SAPO solid acids in their as-pre- acid sites in the former but preserve the surface acid den-
pared forms (see Table S1†).
sities in both cases.
Type II N2 adsorption–desorption isotherms were measured
Fig. 3a and b contain transmission electron microscopy
for all a-SA and a-SAPO samples (Fig. 2), suggesting non- (TEM) images of the a-SA40 as a representative sample of the
porous morphologies that are characteristic of flame-made a-SA series showing the amorphous nature of the nano-
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
Fig. 2 Nitrogen adsorption–desorption isotherms of (a) a-SA and (c) a-SAPO of different compositions, and (b and d) their corresponding specific
surface area (SSA, solid symbols) and SSA-equivalent diameter (dSSA, open symbols). The a-SA40(HSA) is represented as a triangle in (b).
Fig. 3 TEM images of (a and b) a-SA40 and (c and d) a-SAPO25 and their corresponding electron diffraction pattern (inset), which are diffuse, indi-
cating a predominantly amorphous nature. Also shown in (e and f) are the TEM images of the ZSM-5 zeolite with well-defined channel pores.
particles, as reflected from the diffused electron diffraction again displaying a predominantly amorphous nature. In con-
pattern. Some necking occurs between the particles, which is trast, the TEM images of the ZSM-5 zeolite with well-defined
typical of silica-rich particles as a result of the gelation in the channel pores are shown in Fig. 3e and f.
flame. The TEM images of the representative a-SAPO25
Fig. 4a shows the XPS-determined surface Al/(Al + Si) molar
(Fig. 3c and d) show similar morphologies to that of the a-SA, ratio against that of the bulk. A slight decrease in the surface
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
Fig. 4 XPS-determined surface Al/(Al + Si) or Si/(P + Si) compositions on (a) a-SA and (b) a-SAPO, respectively, as a function of their bulk compo-
sitions. The a-SA40(HSA) and ZSM-5 are represented in (a) as an open triangle and square, respectively.
Al for a-SA with mixed compositions can be observed, which is highly ordered arrangement of the SiO4 and AlO4 tetrahedra
consistent with the slight clustering of Al2O3 as seen earlier units and microporous morphology in ZSM-5 afforded an
from the XRD. Clustering of dopants (as oxide) at high concen- extremely high SSA (419 m2 g−1) as well as a high amount of
trations is a way of reducing the overall surface energy of the framework Brønsted acid sites (Fig. 3). In contrast, the a-SA
particles, seen previously for examples in the FSP-made Fe2O3/ solid acids were non-porous and almost completely amor-
TiO2 and Co3O4/ZrO2.36,37 High homogeneity between the phous. For this reason, the increase in the surface area from
27
surface and bulk Si/(P + Si) ratios was measured for the a-SAPO a-SA40 to a-SA40(HSA) enhanced the amount of strong acid
series in the absence of crystallite clusters, as expected for proportionally (Fig. 5b). It is interesting to note from the Tdes
fully amorphous samples (Fig. 4b). Analysis by XPS confirms peaks that the strength of the strong acid components on the
the exclusive presence of Si4+ and Al3+ metal ions for a-SA, and flame-made solid acids appear to match those of ZSM-5, corro-
additional P5+ for a-SAPO. As such, we can expect the for- borating the 1H and 13C MAS NMR measurements by Huang
mation of surface acid sites in accordance with the principles of et al.41
substitutional doping and charge balance as mentioned earlier.
To compare the surface quality of the solid acids, their acid
The amount of surface acid sites was measured using the contents were normalized to their respective SSA values. As
temperature-programmed desorption of ammonia (NH3-TPD). shown in Fig. 5c, the surface density of strong acid sites for
As shown in Fig. 5, the NH3-TPD profiles could be deconvo- a-SA40 (1.1 μmol m−2) remains the highest among the a-SA
luted into a series of Gaussian peaks. For each Gaussian peak, series, consistent with the trend seen earlier based on the
it is assumed that (1) the activation energy of NH3 desorption gravimetric contents. Similar surface acid densities were
(Ed) is constant and independent of NH3 surface coverage, and measured for both the a-SA40 and a-SA40(HSA). This confirms
(2) a normal distribution of Ed to take into account of the het- that the surface acid density can be easily reproduced for FSP-
erogeneity of acid sites distribution.38 It should be noted that made catalysts of different SSA as long as the bulk compo-
m/z 16 (Fig. S1a†) was used to monitor desorption of NH3 sition is retained. The surface density of strong acids on ZSM-5
instead of the stronger signal m/z 17 (Fig. S1b†), which over- remained higher than the flame-made solid acids due to the
laps with fragments from desorbed H2O (m/z 18, Fig. S1c†). To orderly arrangement of the tetrahedra units.
further confirm that the recorded m/z 16 signal was mainly
As shown in Fig. 5d and e, the pristine AlPO4 (i.e., sample
from the desorption of NH3 and not from the fragmentation of a-SAPO0) shows a minimum amount of surface acid, but it
O2 or H2O, the base peaks of H2O and that of O2 (m/z 32, gradually increases with increasing Si content up to Si/(P + Si)
Fig. S1d†) show no correlation with the signal at m/z 16. In = 0.67 (i.e., sample a-SAPO67). This is augmented in part by
Fig. 5a and d, the peaks in the low temperature region (Tdes
<
the increase in the SSA that scales proportionally to the
250 °C) are typically composed of a mixture of very weak acid increase in Si content. However, the optimum composition is
sites irrelevant to chemical reactions and those from desorbed shifted to a-SAPO40 when reported as surface acid density
+
NH3 hydrogen-bonded to NH4 on strong Brønsted acid sites, (Fig. 5f). Unlike the a-SA series, where significant clustering
i.e., NH4 ·nNH3 (n ≥ 1).39,40 As such, only NH3 desorption Al2O3 was measured, the a-SAPO was characterized by a more
+
peaks above 250 °C will be quantified, categorizing them as homogeneous surface and bulk amorphous nature. This par-
weak (250 °C ≤ Tdes < 350 °C) and strong acid sites (Tdes
≥
tially accounts for the multiple-fold increase in acid contents
350 °C). As shown in Fig. 5b, while the amount of weak acid for the a-SAPO catalysts when compared with the a-SA cata-
sites for the a-SA series is similar across various compositions, lysts, based on both the gravimetric content and surface acid
the highest amount of strong acid was measured on a-SA40 densities.
(0.29 mmol g−1) despite being an order of magnitude lower
Fig. 6a and c show the FTIR-pyridine spectra of the a-SA
than that of the reference ZSM-5 zeolite (1.5 mmol g−1). The and a-SAPO, respectively. In the case of a-SA0, the absorbance
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
Fig. 5 NH3-TPD profiles of (a) a-SA and (c) a-SAPO of different compositions, and (b and e) their estimated acid sites normalized to the weight as
well as (c and f) that normalized to the surface area of the solid acids. The a-SA40(HSA) and ZSM-5 (Al/(Si + Al) = 0.08) are represented as an open
triangle and square, respectively. The deconvoluted peaks in (a and d) are labelled in red (weak acid) and blue (strong acid), while the very weakly
adsorbed NH3 irrelevant to reaction is labelled in grey dashed lines (see the text for details).
Fig. 6 FTIR-pyridine spectra of (a) a-SA and (c) a-SAPO of different compositions, and (b and d) their corresponding ratio of Brønsted-to-Lewis acid
sites. The a-SA40(HSA) and ZSM-5 (Al/(Si + Al) = 0.08) are represented in (b) as an open triangle and square, respectively.
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
band at 1445 cm−1 originates from pyridine hydrogen-bonded the presence of hydronium ions and this is predominantly
to silanol, instead of Lewis sites.29 Likewise, the absorbance regulated by Brønsted acid.20,44 The ZSM-5 zeolite, which rep-
band at 1446 cm−1 for a-SAPO0 is attributed to pyridine hydro- resents a sample with an extremely high amount of almost
gen-bonded to surface P–OH. With the exception of these irre- exclusive Brønsted acid, only resulted in 17% LA yield. The
levant bands, the Brønsted and Lewis acid sites can be readily result is consistent with Acharjee and Lee who carried out reac-
distinguished by the unique absorbance bands at 1543 cm−1 tions under similar hydrothermal conditions.12 Further com-
and 1453 cm−1, respectively.42 Quantification of the acid sites parison with commercial Zeolite X, which has the same Si/Al
was carried out based on the extinction coefficients reported composition as a-SA40, found only 21% LA yield (see Fig. S2†).
by Emeis29 and the ratio of Brønsted-to-Lewis acid sites (B/L) The a-SA100 or pristine Al2O3, which is an almost exclusive
presented in Fig. 6b and d. The highest B/L was measured for Lewis acid, resulted in a low LA yield of 19%. All of the results
a-SA40 (B/L = 1.9), which as discussed earlier also has the above reaffirm the importance of mixed Brønsted and Lewis
highest surface density of acid. The presence of mixed acids for the glucose-to-LA conversions. Surface coverage of
Brønsted and Lewis acids in a-SA is due to the retention of more than 30% by carbon residuals (as humins) was detected
Lewis acids on some of the surface alumina sites, i.e., Si–O⋯Al. on both a-SA40 and a-SAPO25 at the end of the reaction, which
The ZSM-5, on the other hand, is structurally dissimilar can be partially removed by calcination at 450 °C (Table S2†).
from a-SA in that the acid sites in the former are predomi-
To probe into the influence of the Brønsted and Lewis acid
nantly formed on the bridged Si–O–Al that give rise to almost content on the overall glucose conversion to LA, we carried out
exclusively Brønsted acids.43 The highest B/L was measured for further assessments on the range of a-SA series catalysts by
a-SAPO25 (B/L = 14, Fig. 6d). The variability in the amount of using the fructose and 5-HMF intermediates as the starting
acids as well as the B/L content in the flame-made solid acids reactants. Doing so allows specific probing of the subsequent
provides a suitable avenue for studying their influence on the reaction after glucose isomerization, i.e., fructose dehydration
glucose conversion as well as the LA selectivity.
and 5-HMF rehydration. As shown in Fig. 8a, when fructose is
used as the starting reactant, the optimum sample for LA yield
was shifted to a-SA80 (36%), whereas the LA yield for a-SA40
decreased to 27%. In other words, the optimum LA yield
shifted from the a-SA catalyst with a Brønsted acid-rich catalyst
(B/L = 1.9) to one richer in Lewis acid (B/L = 0.70). This is unex-
pected since fructose formation from glucose isomerization
has been claimed to rely on Lewis acid.10 Nevertheless, the
result points to the greater influence of Lewis over Brønsted
acid when starting with fructose, and vice versa for glucose. It
should be reiterated that at any time the co-presence of both
Lewis and Brønsted acid sites is a critical feature for attaining
high LA yields. The almost exclusive Lewis acid of a-SA100 (i.e.,
pristine Al2O3) gave a low LA yield. Unlike the isomerization
(of glucose to fructose) and dehydration (of fructose to 5-HMF)
steps, the influence of different acidities of a-SA is less obvious
in the rehydration step (of 5-HMF to LA) (Fig. 8b). Hence, it
can be reasonably deduced that the final LA yield from the
3.2. Glucose conversions over a-SA and a-SAPO solid acids
Fig. 7 shows the assessment of the a-SA and a-SAPO series cat-
alysts for the hydrothermal conversion of glucose at 180 °C
and 24 h, where complete conversion was achieved in all
cases. An optimum LA yield of 40% was measured for a-SA40,
while a similar optimum yield of 42% was measured for
a-SAPO25. The optimum yields coincide with the highest B/L
for these two samples within the a-SA and a-SAPO series,
respectively. It is interesting to note that despite the increased
amount of acid content on the higher surface area analogue of
a-SA40(HSA), the LA yield is not higher than that of the stan-
dard a-SA40. It implies that the amount of acid is not a limit-
ing factor in the current measurement. In fact, while the
mutual existence of both Brønsted and Lewis acids is crucial
in both the dehydration of fructose to 5-HMF and further rehy-
dration to LA (Scheme 1), the latter is strongly influenced by
Fig. 7 Glucose conversion and products yields over (a) a-SA and (b) a-SAPO solid acids at 180 °C. The a-SA40(HSA) and ZSM-5 (Al/(Si + Al) = 0.08)
are represented as open and open cross symbols, respectively.
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
Fig. 8 (a) Fructose and (b) 5-HMF conversion and product yields over a-SA solid acids at 180 °C.
Fig. 9 Reusability of the a-SA40 solid acid catalyst for the conversion of glucose at 180 °C.
overall glucose conversion reaction is more greatly influenced obtain solid acids with SSA as high as 478 m2 g−1 with surface
by the acid properties of the a-SA especially in the early iso- acid density comparable to that produced at a lower SSA. This
merization and dehydration steps. This is consistent with shows the reproducibility of the catalyst surface during the
prior studies that identified these steps as having larger acti- direct scaling of the particle size by FSP. By systematically
vation energies than the final rehydration step.8,9
varying their compositions, it was possible to yield solid acids
Lastly, Fig. 9 shows the good recyclability of the champion with tunable acidity, i.e., the amount of strong and weak acids,
a-SA40 catalyst over 4 consecutive reaction runs. Complete con- as well as the ratio of Brønsted to Lewis acid sites (B/L). The
version of glucose was obtained in all cases, with average solid acids were tested for the aqueous conversion of glucose
5-HMF, FA and LA yields of 10, 19 and 40%, respectively. as standalone catalysts, i.e., without the addition of liquid
Although calcination could partially remove the residue humins acids. Optimum LA yields of 40% or more were achieved for
(see Table S2†), we adopted a less cumbersome approach of a-SA40 (among a-SA series) and a-SAPO25 (for a-SAPO series),
simply rinsing the catalysts with deionized water in between outperforming the protonated ZSM-5 (17% yield) and zeolite X
reaction cycles. This appears to be adequate in making (21% yield) as well as those reported for the ion-exchange
sufficient active sites available for the subsequent reaction polymer resins under similar reaction conditions. While the
run.
mutual existence of Brønsted and Lewis acids is critical, the
optimum glucose-to-LA yield is attributed to the highest B/L
ratio of the flame-made solid acids. The influence of acidity of
the catalysts is more prominent in the isomerization (of
glucose to fructose) and dehydration (of fructose to 5-HMF)
4. Conclusions
We have successfully demonstrated the design of flame-made steps than the final rehydration (of 5-HMF to LA) step. The
amorphous solid acids of silica-alumina (a-SA series) and a-SA40 catalyst was shown to be highly recyclable, with just
silica-alumina-phosphate (a-SAPO series) for the conversion of simple rinsing in between reaction cycles. This work show-
glucose to LA. Despite being non-porous, it was possible to cases the potential of using the economical and scalable FSP-
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
derived solid acids for the wider biomass conversion appli-
cations, especially where regioselectivity as commonly associ-
ated with zeolite catalysts is not essential.
Lewis and Brønsted acids, Chem. Eng. J., 2017, 307, 389–
398.
10 Y. Román-Leshkov, M. Moliner, J. A. Labinger and
M. E. Davis, Mechanism of glucose isomerization using a
solid Lewis acid catalyst in water, Angew. Chem., Int. Ed.,
2010, 49, 8954–8957.
11 Y. J. Pagán-Torres, T. Wang, J. M. R. Gallo, B. H. Shanks
and J. A. Dumesic, Production of 5-Hydroxymethylfurfural
from glucose using a combination of Lewis and Brønsted
acid catalysts in water in a biphasic reactor with an alkyl-
phenol solvent, ACS Catal., 2012, 2, 930–934.
12 T. C. Acharjee and Y. Y. Lee, Production of levulinic acid
from glucose by dual solid-acid catalysts, Environ. Prog.
Sustainable Energy, 2018, 37, 471–480.
Conflicts of interest
There are no conflicts to declare.
Acknowledgements
This work was supported by the Research Grants Council of
Hong Kong through the General Research Funds (Projects
11302714 and 11337116), and the School of Energy and 13 P. Gupta and S. Paul, Solid acids: Green alternatives for
Environment (City University of Hong Kong) through the SEE acid catalysis, Catal. Today, 2014, 236, 153–170.
Collaborative Grant. G. K. B. acknowledges the support from 14 F. Rinaldi and F. Schüth, Design of solid catalysts for the
the Research Grants Council of Hong Kong under the Hong
Kong PhD Fellowship Scheme, and Dr Xincong Lv for technical
conversion of biomass, Energy Environ. Sci., 2009, 2, 610–
626.
assistance. We thank Dr Bill Bin Gong of The University of 15 X. Chu, D. Zhou, D. Li, K. Xia, N. Gan, X. H. Lu, R. F. Nie
New South Wales for the XPS measurements. We are grateful
for the use of the Microscopy Australia TEM facilities at the
University of Sydney.
and Q. H. Xia, Construction of solely Lewis acidic
sites on Zr-MCM-41 and the catalytic activity for the
Prins condensation of β-pinene and paraformaldehyde to
nopol, Microporous Mesoporous Mater., 2016, 230, 166–
117.
16 B. Tang, X. H. Lu, D. Zhou, J. Lei, Z. H. Niu, J. Fan and
Q. H. Xia, Highly efficient epoxidation of styrene and
α-pinene with air over Co2+-exchanged ZSM-5 and Beta zeo-
lites, Catal. Commun., 2012, 21, 68–71.
17 X. T. Ma, D. Zhou, X. Chu, D. Li, J. Wang, W. C. Song,
Y. Shen, X. H. Lu and Q. H. Xia, Highly selective isomer-
ization of biomass β-pinene over the hierarchically acidic
MCM-22 catalyst, Microporous Mesoporous Mater., 2017,
237, 180–188.
18 Q. Q. Zhu, P. Zhuang, L. Y. Wang, Y. Liu and S. B. Wu,
Silicotungstate: An inexpensive and efficient catalyst for
conversion of glucose to levulinic acid in aqueous medium,
J. Biobased Mater. Bioenergy, 2019, 13, 677–682.
References
1 B. R. Caes, R. E. Teixeira, K. G. Knapp and R. T. Raines,
Biomass to furanics: Renewable routes to chemicals and
fuels, ACS Sustainable Chem. Eng., 2015, 3, 2591–2605.
2 A. C. K. Yip, W. Y. Teoh and H. Yang, Advancement in
energy and environmental catalysis in Asia Pacific, Catal.
Today, 2018, 314, 1.
3 P. Gallezot, Conversion of biomass to selected chemical
products, Chem. Soc. Rev., 2012, 41, 1538–1558.
4 A. Corma, S. Iborra and A. Velty, Chemical routes for the
transformation of biomass into chemicals, Chem. Rev.,
2007, 107, 2411–2502.
5 Top value added chemicals from biomass, ed. T. Werpy and 19 F. D. Pileidis and M.-M. Titirici, Levulinic acid biorefi-
G. Petersen, Energy Efficiency and Renewable Energy, U.S.
Department of Energy, 2004.
neries: New challenges for efficient utilization of biomass,
ChemSusChem, 2016, 9, 562–582.
6 D. M. Alonso, J. Q. Bond and J. A. Dumesic, Catalytic con- 20 R. Weingarten, Y. T. Kim, G. A. Tompsett, A. Fernández,
version of biomass to biofuels, Green Chem., 2010, 12,
1493–1513.
7 V. Choudhary, S. H. Mushrif, C. Ho, A. Anderko,
V. Nikolakis, N. S. Marinkovic, A. I. Frenkel, S. I. Sandler
K. S. Han, E. W. Hagaman, W. C. Conner Jr., J. A. Dumesic
and G. W. Huber, Conversion of glucose into levulinic acid
with solid metal(IV) phosphate catalysts, J. Catal., 2013,
304, 123–134.
and D. G. Vlachos, Insights into the interplay of Lewis and 21 R. Weingarten, W. C. Conner Jr. and G. W. Huber,
Brønsted acid catalysts in glucose and fructose conversion
to 5-(hydroxymethyl) furfural and levulinic acid in aqueous
media, J. Am. Chem. Soc., 2013, 135, 3997–4006.
Production of levulinic acid from cellulose by hydrothermal
decomposition combined with aqueous phase dehydration
with a solid acid catalyst, Energy Environ. Sci., 2012, 5,
7559–7574.
8 R. Weingarten, J. Cho, R. Xing, W. C. Conner Jr. and
G. W. Huber, Kinetics and reaction engineering of levulinic 22 Y. Xu, G. Liu, J. Fu, S. Kang, Y. Xiao, P. Yang and
acid production from aqueous glucose solutions,
ChemSusChem, 2012, 5, 1280–1290.
9 W. Wei and S. Wu, Experimental and kinetic study of
glucose conversion to levulinic acid catalyzed by synergy of
W. Liao, Catalytic hydrolysis of cellulose to levulinic
acid by partly replacing sulfuric acid with the Nafion®
NR50 catalyst, Biomass Convers. Biorefin., 2019, 9,
609–616.
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
23 R. Mueller, L. Mädler and S. E. Pratsinis, Nanoparticle syn-
thesis at high production rates by flame spray pyrolysis,
Chem. Eng. Sci., 2003, 58, 1969–1976.
24 K. Wegner, B. Schimmöller, B. Thiebaut, C. Fernandez and
T. N. Rao, Pilot plants for industrial nanoparticle pro-
duction by flame spray pyrolysis, KONA Powder Part. J.,
2011, 29, 251–265.
perties and catalytic behavior in enantioselective hydrogen-
ation, J. Catal., 2003, 213, 296–304.
35 H. Schulz, W. J. Stark, M. Maciejewski, S. E. Pratsinis and
A. Baiker, Flame-made nanocrystalline ceria/zirconia doped
with alumina or silica: structural properties and enhanced
oxygen exchange capacity, J. Mater. Chem., 2003, 13, 2979–
2984.
25 W. Y. Teoh, R. Amal and L. Mädler, Flame spray pyrolysis: 36 W. Y. Teoh, R. Setiawan, L. Mädler, J.-D. Grunwaldt,
an enabling technology for nanoparticles design and fabri-
cation, Nanoscale, 2010, 2, 1324–1347.
26 S. Loher, W. J. Stark, M. Maciejewski, A. Baiker,
R. Amal and S. E. Pratsinis, Ru-doped cobalt-zirconia nano-
composites by flame synthesis: Physicochemical and cata-
lytic properties, Chem. Mater., 2008, 20, 4069–4079.
S. E. Pratsinis, D. Reichardt, F. Maspero, F. Krumeich and 37 W. Y. Teoh, D. E. Doronkin, G. K. Beh, J. A. H. Dreyer and
D. Günther, Fluoro-apatite and calcium phosphate nano-
particles by flame synthesis, Chem. Mater., 2005, 17, 36–42.
27 W. Y. Teoh, R. Amal and S. E. Pratsinis, Flame sprayed
J.-D. Grunwaldt, Methanation of carbon monoxide over
promoted flame-synthesized cobalt clusters stabilized in
zirconia matrix, J. Catal., 2015, 326, 182–193.
visible light-active Fe-TiO2 for photomineralisation of 38 F. Arena, R. Dario and A. Parmaliana, A characterization
oxalic acid, Catal. Today, 2007, 120, 203–213.
study of the surface acidity of solid catalysts by temperature
28 W. Y. Teoh, L. Mädler, D. Beydoun, R. Amal and
programmed methods, Appl. Catal., A, 1998, 170, 127–137.
S. E. Pratsinis, Direct (one-step) synthesis of TiO2 and Pt/ 39 F. Lónyi and J. Valyon, On the interpretation of the NH3-
TiO2 nanoparticles for photocatalytic mineralisation of
sucrose, Chem. Eng. Sci., 2005, 60, 5852–5861.
TPD patterns of H-ZSM-5 and H-modernite, Microporous
Mesoporous Mater., 2001, 47, 293–301.
29 C. A. Emeis, Determination of integrated molar extinction 40 S. A. Skarlis, D. Berthout, A. Nicolle, C. Dujardin and
coefficients for infrared absorption bands of pyridine
adsorbed on solid acid catalysts, J. Catal., 1993, 141, 347–
354.
P. Granger, IR spectroscopy analysis and kinetic modeling
study of NH3 adsorption and desorption on H- and Fe-BEA
catalysts, J. Phys. Chem. C, 2013, 117, 7154–7169.
30 T. Tani, L. Mädler and S. E. Pratsinis, Synthesis of zinc 41 J. Huang, N. Vegten, Y. Jiang, M. Hunger and A. Baiker,
oxide/silica composite nanoparticles by flame spray pyrol-
ysis, J. Mater. Sci., 2002, 37, 4627–4623.
31 T. J. Brunner, R. N. Grass and W. J. Stark, Glass and bio-
Increasing the Brønsted acidity of flame-derived silica/
alumina up to zeolitic strength, Angew. Chem., Int. Ed.,
2010, 49, 7776–7781.
glass nanopowders by flame synthesis, Chem. Commun., 42 G. Crépeau, V. Montouillout, A. Vimont, L. Mariey, T. Cseri
2006, 1384–1386.
32 D. Li, W. Y. Teoh, R. C. Woodward, J. D. Cashion,
C. Selomulya and R. Amal, Evolution of morphology and
and F. Mauge, Nature, structure and strength of acidic sites
of amorphous silica alumina: an IR and NMR study,
J. Phys. Chem. B, 2006, 110, 15172–15185.
magnetic properties in silica/maghemite nanocomposites, 43 G. Busca, The surface acidity of solid oxides and its charac-
J. Phys. Chem. C, 2009, 113, 12040–12047.
33 S. E. Pratsinis, Flame aerosol synthesis of ceramic powders,
Prog. Energy Combust. Sci., 1998, 24, 197–219.
34 R. Strobel, W. J. Stark, L. Mädler, S. E. Pratsinis and
A. Baiker, Flame-made platinum/alumina: structural pro-
terization by IR spectroscopic methods. An attempt at sys-
tematization, Phys. Chem. Chem. Phys., 1999, 1, 723–736.
44 F. Yang, J. Fu, J. Mo and X. Lu, Synergy of Lewis and
Brønsted acids on catalytic hydrothermal decomposition of
hexose to levulinic acid, Energy Fuels, 2013, 27, 6973–6978.
This journal is © The Royal Society of Chemistry 2019
Green Chem.