Conversion of glucose into levulinic acid with solid metal(IV) phosphate catalysts

Article abstract of DOI:10.1016/j.jcat.2013.03.023

We have prepared a series of well-characterized solid acid metal(IV) phosphate catalysts and tested them for the two-step dehydration/rehydration reaction to produce levulinic acid from glucose. The catalysts include zirconium (ZrP) and tin (SnP) phosphates with varying ratios of phosphorus to metal(IV). The structural, surface and bulk properties have been investigated using XRD, BET, XPS and ³¹P solid-state MAS NMR spectroscopy. ZrP is distinguished by a high concentration of polyphosphate species in the bulk phase, as well as increased hydroxyl groups on the surface. ZrP also shows a higher concentration of total acid sites and Bronsted acid sites compared to SnP, as determined by TPD measurements using gas-phase NH₃ and isopropylamine. The catalyst selectivity is a function of the Bronsted to Lewis acid site ratio using either heterogeneous or homogeneous catalysts. Catalytic activity increases with increased Lewis acid sites. The Lewis sites mainly produce fructose via isomerization reactions and undesired degradation products (humins). HMF is produced on both Bronsted and Lewis sites, whereas levulinic acid is exclusively produced on Bronsted sites. Zirconium phosphate with a P/Zr molar ratio of 2 is favorable for levulinic acid production due to its inherently high surface area and enhanced Bronsted acidity. This study lays the grounds for further design of improved solid acid catalysts for aqueous phase production of HMF and levulinic from carbohydrates.

Full text of DOI:10.1016/j.jcat.2013.03.023

Journal of Catalysis 304 (2013) 123–134
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Conversion of glucose into levulinic acid with solid metal(IV) phosphate
catalysts
Ronen Weingarten a, Yong Tae Kim a, Geoffrey A. Tompsett b, Alejandro Fernández b, Kee Sung Han c,
Edward W. Hagaman c, Wm. Curt Conner Jr. b, James A. Dumesic a, George W. Huber a,
⇑
a Univeirsty of Wisconsin–Madison, Department of Chemical and Biological Engineering, 1415 Engineering Dr., Madison, WI 53706, USA
b University of Massachusetts–Amherst, Department of Chemical Engineering, 159 Goessmann Lab, 686 North Pleasant St., Amherst, MA 01003, USA
c Oak Ridge National Laboratory, Chemical Sciences Division, Building 4100, Room A123, 1 Bethel Valley Rd., Oak Ridge, TN 37831, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
We have prepared a series of well-characterized solid acid metal(IV) phosphate catalysts and tested them
for the two-step dehydration/rehydration reaction to produce levulinic acid from glucose. The catalysts
include zirconium (ZrP) and tin (SnP) phosphates with varying ratios of phosphorus to metal(IV). The
structural, surface and bulk properties have been investigated using XRD, BET, XPS and 31P solid-state
MAS NMR spectroscopy. ZrP is distinguished by a high concentration of polyphosphate species in the bulk
phase, as well as increased hydroxyl groups on the surface. ZrP also shows a higher concentration of total
acid sites and Brønsted acid sites compared to SnP, as determined by TPD measurements using gas-phase
NH3 and isopropylamine. The catalyst selectivity is a function of the Brønsted to Lewis acid site ratio
using either heterogeneous or homogeneous catalysts. Catalytic activity increases with increased Lewis
acid sites. The Lewis sites mainly produce fructose via isomerization reactions and undesired degradation
products (humins). HMF is produced on both Brønsted and Lewis sites, whereas levulinic acid is exclu-
sively produced on Brønsted sites. Zirconium phosphate with a P/Zr molar ratio of 2 is favorable for lev-
ulinic acid production due to its inherently high surface area and enhanced Brønsted acidity. This study
lays the grounds for further design of improved solid acid catalysts for aqueous phase production of HMF
and levulinic from carbohydrates.
Received 27 September 2012
Revised 22 January 2013
Accepted 26 March 2013
Available online 25 May 2013
Keywords:
Metal(IV) phosphate
Solid acid catalyst
Glucose
Levulinic acid
Brønsted
Lewis
Dehydration
Aqueous phase
Ó 2013 Elsevier Inc. All rights reserved.
1. Introduction
groups provides this compound with interesting reactivity
pathways.
Decline of petroleum supplies and increased levels of green-
house gases in the atmosphere have stimulated a worldwide initia-
tive to develop clean technologies that utilize a sustainably
produced feedstock. Lignocellulosic biomass is a renewable feed-
stock that has received considerable attention as a sustainable
alternative for the production of fuels and chemicals [1–10]. In
contrast to other renewable energies, biomass is the only renew-
able source of fixed carbon, which is essential for the production
of liquid hydrocarbon fuels and chemicals [11–13]. One of the
key sugar-based building blocks that can be obtained from ligno-
cellulosic biomass is levulinic acid, which has been identified by
the US Department of Energy as a top platform chemical [14,15].
Levulinic acid is a five carbon molecule with carboxylic acid and
ketone functionalities, and the existence of these two functional
As a top platform chemical, levulinic acid is the building
block to produce a great number of bio-chemicals such as succi-
nic acid, resins, polymers, herbicides, pharmaceuticals, flavoring
agents, solvents, plasticizers, and anti-freeze agents. In particular,
some levulinic acid derivatives can be used for fuels and oxygen-
ated fuel additives. Esterification of levulinic acid with C1–C2
alcohols produces levulinic esters which can be used as diesel
additives [16]. Elliott and Frye have also shown that levulinic
acid can be hydrogenated in the presence of a bifunctional cata-
lyst to produce methyl-tetrahydrofuran (MTHF) [17], which can
directly serve as a gasoline blendstock [18]. Levulinic acid can
also undergo hydrogenation to produce
c-valerolactone (GVL)
[19–21], which has been shown to be a sustainable liquid trans-
portation fuel suitable of replacing ethanol in gasoline–ethanol
blends [22]. Lange et al. have shown that continued hydrogena-
tion of GVL produces valeric acid which can be esterified with
alcohols to produce a new class of cellulosic transportation fuels,
‘‘valeric biofuels’’ [23]. Bond et al. developed an integrated cata-
lytic process to convert GVL to liquid alkenes (ranging from C8 to
⇑
Corresponding author. Address: 4631 Engineering Hall, 1415 Engineering Dr.,
Madison, WI 53706, USA. Fax: +1 608 262 5434.
E-mail addresses: rweingarten@wisc.edu (R. Weingarten), exkimyt@gmail.com
(Y.T. Kim), tompsett@ecs.umass.edu (G.A. Tompsett), jandro.fs@hotmail.com (A.
Fernández), hanks@ornl.gov (K.S. Han), hagamanew@ornl.gov (E.W. Hagaman),
wconner@ecs.umass.edu (W.C. Conner Jr.), dumesic@engr.wisc.edu (J.A. Dumesic),
huber@engr.wisc.edu (G.W. Huber).
C24) which could be blended with gasoline, diesel or jet fuels
[24].
0021-9517/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2013.03.023

124
R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
Fig. 1. Reaction pathway scheme for the conversion of glucose to levulinic acid.
Levulinic acid is produced through acid-catalyzed dehydration
In the field of biomass conversion, various studies have been re-
ported on the selective production of HMF from carbohydrates in
the aqueous phase using metal phosphates as solid acid catalysts
[55–58]. More recently, we found that amorphous zirconium phos-
phate containing a high Brønsted to Lewis acid site ratio exhibits
high selectivity for furfural production from xylose [40]. The activ-
ity and selectivity for furfural production of the zirconium phos-
phate were comparable to those obtained from dehydration
reactions with HCl. The dehydration of sorbitol to produce isosor-
bide with metal(IV) phosphates has also been reported [59]. Li
et al. have found that platinum supported on zirconium phosphate
serves as a stable, selective, and active catalyst for aqueous phase
hydrodeoxygenation of aqueous sugar solutions to produce high-
octane gasoline [60,61]. They found the catalyst to be stable in
aqueous media at high temperatures (245 °C), and no deactivation
occurred after 200 h time-on-stream. ICP studies confirmed that no
leaching occurred. This coincides with a study by Asghari et al. who
also found zirconium phosphate to be stable under subcritical
water conditions [62]. Niobium-based catalysts have also received
much consideration as solid acid catalysts due to their hydrother-
mal stability and high activity for aqueous phase dehydration of
alcohols to olefins [63,64].
Zirconium phosphate has unique properties as a solid acid cat-
alyst in that it is compatible in aqueous media. Various studies
have investigated the source of the active sites on this class of cat-
alysts. In their study of crystalline zirconium phosphate, Hattori
et al. concluded that the catalyst possesses weak and strong acid
sites, both of which are derived from P(OH) groups [65]. Clearfield
and Thakur observed a decrease in the activity of zirconium phos-
phate after poisoning the catalyst with quinoline or following pro-
ton exchange with Cs+ on the catalyst surface [44,48]. They
correlated the active sites to surface hydroxyl groups but also asso-
ciated some of the activity to Lewis-type sites. La Ginestra et al.
used a similar surface poisoning technique to conclude that the
catalytic activity of zirconium phosphate was solely from Brønsted
acid sites on the surface of the catalyst [45]. Sinhamahapatra et al.
observed both Brønsted and Lewis acid sites in their study on mes-
oporous zirconium phosphate [66]. They postulated that the
Brønsted acid sites possibly arise from geminal P(OH) groups.
The Lewis acid centers could be attributed to Zr4+, as suggested
by Spielbauer et al. [67].
of hexose sugars. Among these, glucose is the most abundant in
nature, available in the polysaccharide form as starch or cellulose
in biomass [25]. Levulinic acid is produced from glucose by way
of a two-step acid-catalyzed reaction (Fig. 1). We have studied
the kinetics of levulinic acid production from aqueous glucose
solutions with a homogeneous acid catalyst in a continuous regime
[26]. Overall, there are four parallel pathways in which glucose can
react: (1) Reversion reactions can lead to the formation of cellobi-
ose and levoglucosan [27–29]; (2) degradation reactions to form
highly polymerized carbonaceous species (i.e. humins); (3) epi-
merization reactions to form fructose and mannose [28], and (4)
dehydration to produce 5-hydroxymethylfurfural (HMF). It has
been reported that reversion and epimerization products can also
decompose to form humins [30]. The humins are both water solu-
ble and water insoluble. Water-soluble humins polymerize with
time to form water-insoluble compounds. The isomerization of
glucose into fructose has been reported to be favored by the pres-
ence of Lewis acid sites [31–33]. Recent studies have reported the
efficient conversion of glucose to HMF using a combination of Le-
wis and Brønsted acid catalysts in a biphasic reactor system with
2-s-butylphenol as the extracting solvent [34,35]. The reaction pro-
ceeds through a tandem pathway including isomerization of glu-
cose to fructose followed by dehydration of fructose to HMF.
Subsequently, HMF can rehydrate to produce levulinic acid and
formic acid. HMF can also decompose to produce humins and pro-
duce furfural via loss of formaldehyde [36–39]. Furfural can also
undergo degradation reactions to form humins [40]. Formic acid
is also a by-product of furfural degradation by way of hydrolytic
fission of the furfural aldehyde group [41,42].
Attempts have been made to utilize recyclable solid acid cat-
alysts to produce levulinic acid from biomass feedstocks. These
range from zeolites and novel metal oxides to acidic ion-ex-
change polymer resins [43]. In particular, metal(IV) phosphate
catalysts have shown to be selective and active for aqueous
phase acid-catalyzed dehydration and isomerization reactions
[44–47]. Extensive investigations have been carried out to deter-
mine the physicochemical and acidic properties of these catalysts
[48,49]. Studies have shown that the amorphous form of these
catalysts demonstrate higher activity due to increased overall
acidity and surface area compared to their crystalline analog
[50,51]. In addition, metal(IV) phosphates are beneficial as cata-
lyst supports for metal oxides due to their textural and acid
properties [52–54].
The objective of this study is to characterize the different cata-
lytic sites on supported metal(IV) phosphate solid acid catalysts
and identify their roles in catalyzing the aqueous phase conversion

R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
125
of glucose to produce levulinic acid. The nature of the metal(IV)
and phosphorus loading will be investigated, as well as the role
of Lewis and Brønsted acid sites. Common gas-phase characteriza-
tion techniques will be used to determine the concentration and
type of the acid sites, including temperature-programmed desorp-
tion using ammonia and isopropylamine as probe molecules. Addi-
tional techniques such as XPS, ICP, and 31P solid-state NMR will be
employed to study the surface and bulk properties of the solid acid
catalysts.
with thermogravimetric analysis–mass spectrometry (TA instru-
ments SDT Q600 system), as reported by Gorte et al. [68–70]Approx-
imately 50–60 mg of sample was loaded and degassed at 673 K for
1.5 h under a constant helium flow of 100 mL minÀ1 (Airgas, UHP).
The sample was then cooled to 318 K, at which point isopropylamine
was bubbled with helium and adsorbed for 15–20 min. After satura-
tion, the isopropylamine supply line was shut off and helium was
purged at 100 mL minÀ1 to remove any physically adsorbed isopro-
pylamine. The sample was then heated linearly at a rate of
10 K minÀ1 from 318 to 973 K (773 K for the tin phosphate samples)
under a constant helium flow of 100 mL minÀ1. During the TPD anal-
ysis, the signal m/z 39 (propylene) was monitored by the mass spec-
trometer. The mass spectrometer was calibrated for propylene
(Sigma Aldrich, P99%) for quantification purposes.
2. Experimental
2.1. Catalyst preparation
Adsorption and desorption isotherms (40 points each) of nitro-
gen were obtained for all of the samples at 77 K using a Quanta-
chrome AutosorbÒ iQ2 automated gas sorption system. The
surface areas were calculated using BET analysis ranging from
0.05 to 0.3 relative pressure P/Po on the adsorption branch, and
the BET ‘‘C’’ constants were recorded. The samples were degassed
before each experiment at 523 K for 12 h.
The zirconium phosphate catalysts (ZrP) were prepared follow-
ing procedures previously reported [51], which consisted of precip-
itation of ZrCl2OÁ8H2O (Sigma Aldrich, 1 mol LÀ1, 140 mL) and
NH4H2PO4 (Sigma Aldrich, 1 mol LÀ1, 280 mL) at a molar ratio of
P/Zr = 2 (ZrP2). The solution was stirred at room temperature and
then filtered, washed with de-ionized (DI) water until pH 5 and
dried overnight at 373 K. The catalyst was powdered and calcined
at 673 K for 4 h in air prior to reaction. The other two zirconium
phosphate catalysts (ZrP1 and ZrP3) were prepared following the
same procedure but varying the molar ratio of P/Zr to 1 and 3
respectively.
The tin phosphate catalyst (SnP1) was obtained following a pro-
cedure reported by Patel et al. [47]. A 0.1 M aqueous solution of
disodium hydrogen phosphate (Sigma Aldrich, ReagentPlus™
P99%) was added drop wise to an equimolar, equivolume, stirred
aqueous solution of tin tetrachloride (SnCl4Á5H2O, Fisher Scien-
tific). The resulting gel was stirred at 50 °C for 2 h and at room tem-
perature for 24 h. It was then filtered and dried at room
temperature. The dried material was powdered and converted to
the hydrogen form by treatment with a 1 M aqueous solution of
HNO3 (Fisher Scientific) for 3 h, the acid being intermittently re-
placed with a fresh batch. It was then washed several times with
de-ionized water to remove the excess acid until pH of 3 and finally
dried at room temperature. Tin phosphate (SnP2) was prepared fol-
lowing the same procedure but varying the molar ratio of P/Sn to 2
and then washed thoroughly until pH 3.
XPS measurements were performed using a Physical Electronics
Quantum 2000 Scanning ESCA Microprobe equipped with a mono-
chromatic AlK
a anode (1486.6 eV). Samples were prepared by
dusting them on double sided adhesive tape. They were then evac-
uated and placed on the stage of the instrument. Charging was
neutralized using the turnkey electron/ion system, so that the
C1s peak was fixed at 284.0 eV. Analysis was performed by taking
sequential spectra from a 5° take-off angle increasing the angle by
3° per spectrum until the C1s line increased. Since this indicated
that the substrate tape was detected, the angle was then decreased
to that before the tape was detected. This proved to be a 15°
take-off angle which provided excellent signal to noise for quality
spectra, acquired using a 200 lm spot at 50 W. All quantitative
calculations used atomic sensitivity factors tailored to the
instrument by the manufacturer.
Solid-state 31P magic angle spinning (MAS) nuclear magnetic res-
onance (NMR) spectra were obtained on a 9.4 T Bruker AvanceÒ
spectrometer at a resonance frequency of 161.97 MHz. Powdered
samples were spun at 12 kHz in a 4 mm CP/MAS probe under ambi-
ent temperature and humidity conditions. Samples were not dried
prior to examination. Spectra were acquired using a single pulse
Zirconium oxide was obtained by calcining zirconium hydrox-
ide (supplied by MEL chemicals-(XZO 880/01)) at 673 K for 2 h in
air. Hydrochloric acid was supplied by Fischer Scientific. Ytterbium
(III) trifluoromethanesulfonate hydrate, Yb (OTf)3, was supplied by
Strem Chemicals.
excitation pulse sequence (zg) with a 5 ls (90°) pulse length and
120 s recycle delay. The 31P spin–lattice relaxation time estimated
from progressive saturation experiments is ca. 30 s and does not
show marked differences between the partially resolved resonance
bands. 31P chemical shifts were referenced to external 85% H3PO4
(0 ppm).
2.2. Catalyst characterization
The structures of the catalysts were determined by X-ray dif-
fraction (XRD). The XRD patterns were obtained with a Philips
X’Pert Pro diffractometer equipped with a X’Celerator detector
operated at 45 kV and 40 mA using Cu ja radiation
Total acid sites were determined by temperature-programmed
desorption of ammonia (NH3-TPD) with a Quantachrome ChemBET
Pulsar™ TPR/TPD Automatic Chemisorption Analyzer coupled with
a thermal conductivity detector (TCD) to quantify the ammonia
desorbed from the sample. A sample of approximately 300 mg
was initially degassed at 673 K for 1.5 h under a constant helium
flow of 12 mL minÀ1 (Airgas, UHP). The sample was cooled and
ammonia (Airgas, electronic grade) was adsorbed at 373 K for
30 min to reach saturation. Afterward, the ammonia supply line
was shut off and helium was purged at 12 mL minÀ1 for 2 h to re-
move any physically adsorbed ammonia. The sample was then
heated linearly at a rate of 10 K minÀ1 from 373 to 973 K (773 K
for the tin phosphate samples) under a constant helium flow of
12 mL minÀ1. The sample was held at the temperature set point
for an additional 2 h.
(k = 0.15406 nm) at a scan rate of 0.1° (2h) sÀ1
.
The amounts of phosphorus, zirconium, and tin present in the
solid acid catalysts were determined by elemental analysis by
way of inductively coupled plasma (ICP) analysis. The samples
were not degassed prior to analysis. The analyses were performed
by Galbraith Laboratories, Inc.
2.3. Catalyst activity
Batch reactions were carried out in a 100 mL reactor vessel pro-
vided by Parr Instrument Company, series 4560. Glucose (Fisher
Scientific) solutions were prepared with de-ionized water at the
specified concentration. Throughout all of the experiments, the
amount of loaded reaction solution was kept constant at 70 g.
The concentration of Brønsted acid sites was determined by tem-
perature-programmed desorption of isopropylamine (IPA-TPD)

126
R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
Temperatures in the reactor were measured by means of a thermo-
couple in contact with the solution. All reaction solutions were
mixed at a maximum constant rate of 600 rpm using an internal
stirrer. The temperature and stirring were controlled by a 4848
Controller provided by Parr. The reaction vessel was initially pres-
surized to 800 psi with industrial grade helium (Airgas). Samples
were taken periodically through a sampling port. The samples
were immediately quenched in an ice water bath and filtered with
Ammonia TPD revealed a higher concentration of total acid sites
for the zirconium phosphates compared to the tin phosphate cata-
lysts. The three zirconium phosphate samples showed similar con-
centrations of total acid sites as measured by NH3-TPD (Fig. S2 in
the Supplementary information). Sample ZrP2 showed the highest
total acid concentration despite the equivalent P/Zr molar ratios to
sample ZrP3. This higher total acid concentration for ZrP2 could be
due to its higher surface area. SnP1 had a comparable acid concen-
tration to that reported in the literature [47]. In contrast, an in-
creased amount of acid sites was observed for SnP2 despite the
similar amounts of phosphorus incorporated in the framework.
The two tin phosphate samples showed similar acid strength
according to the NH3-TPD profiles.
a 0.2-lm syringe filter prior to analysis. The reactor was repressur-
ized with helium after each sampling. The dip tube was covered
with a stainless steel woven wire cloth, mesh size 400 Â 400 pro-
vided by McMaster-Carr. This was done to prevent clogging and
loss of catalyst during sampling.
Brønsted acid sites were quantified according to isopropyl-
amine TPD [68–70]. The concentrations of Brønsted acid sites ob-
tained for the zirconium phosphate samples were at least 3
times higher than the concentration of Brønsted acid sites for the
tin phosphates (Fig. S3 in the Supplementary information). ZrP2
showed the highest Brønsted acid concentration among all of the
catalysts tested, containing more than 26 times the amount of
Brønsted sites observed with SnP1. In this respect, extensive stud-
ies have been reported on the use of 1H MAS NMR spectroscopy to
determine the strength of Brønsted acid sites. Acid sites of increas-
ing acid strength are characterized by higher 1H NMR chemical
shifts [73]. Fraissard and co-workers have measured the acid
strength of several types of acid catalysts by two 1H NMR tech-
niques in the presence of water [74–79]. As part of this study,
the zirconium phosphate samples were characterized by solid-
state 1H MAS NMR spectroscopy. Results reveal that all three sam-
ples showed a broad resonance at 7.3 ppm. The placement is
downfield of the shift for water (4.8 ppm) and represents a
weighted average of the acid proton chemical shifts and water
[77]. The chemical shift of all three materials is the same implying
that their acid strengths are comparable. This corresponds with re-
sults obtained from NH3-TPD. It is also notable to mention that the
fraction of Brønsted acid sites for ZrP2 in this study was somewhat
lower than that reported by us in our previous study [40]. This
could be due to a number of factors including differences in the
catalyst washing routine during the synthesis procedure or differ-
ent characterization techniques. NH3 FT-IR spectroscopy was used
to quantify the ratio of Brønsted to Lewis sites in our previous
study. Isopropylamine TPD was used in this study. This could also
lead to inconsistencies in the quantification of the Brønsted acid
sites by different techniques, as different adsorption techniques
quantify sites differently.
2.4. Analysis
Reaction product samples were analyzed by high-pressure liquid
chromatography (HPLC) with a ShimadzuÓ LC-20AT. Carbohydrates
were detected with a RI detector (RID-10A), and reaction products
were detected with a UV–Vis detector (SPD-20AV) at wavelengths
of 210 and 254 nm. The column used was a BioradÓ Aminex HPX-
87H sugar column. The mobile phase was 0.005 M H2SO4 flowing
at a rate of 0.6 mL/min. The column oven was set to 30 °C.
3. Results
3.1. Characterization of solid acid catalysts by adsorption and ICP
The characterizationfor the solidacidcatalysts by adsorption and
ICP appear in Table 1. Zirconia was used as a reference material due
to its high Lewis acidity [71]. All of the phosphate catalysts in this
study were confirmed to be amorphous by XRD (Fig. S1 in the Sup-
plementary information). Previous studies have shown that the
presence of phosphate stabilizes the amorphous zirconia phase
[72]. It has also been reported that amorphous zirconium phosphate
is thermally stable even after calcination at 800 °C [66]. The phos-
phorustometal(IV)molarratiosobtainedfromICPanalysisweredif-
ferent than the phosphorus to metal(IV) molar ratios used in the
precipitation solutions with the exception of ZrP2. The sample
ZrP3 had the same phosphorus content as ZrP2 despite the addi-
tional loading of the phosphorus precursor during synthesis. The
molar ratios for the two tin phosphate samples were comparable
to each other, but lower than the phosphorus to tin ratio used in
the prescription solution.
Zirconium phosphate samples ZrP1 and ZrP3 had similar BET
surface areas to that reported by Kamiya et al. (130 m2 gÀ1) [51].
The highest surface area was obtained for ZrP2 (276 m2 gÀ1). In
contrast, tin phosphate SnP1 had the lowest surface area
(11 m2 gÀ1), which was appreciably lower than that reported by
Patel et al. following the same preparation procedure
(141 m2 gÀ1) [47].
3.2. Solid-state 31P MAS NMR spectroscopy
The zirconium phosphate samples were characterized by solid-
state 31P MAS NMR spectroscopy. This characterization method
serves as a useful technique to identify the coordination states of
the phosphorus atoms in the bulk phase. Generally, a transition
Table 1
Characterization of solid acid catalysts by adsorption and ICP studies.
Catalyst BET surface areaa
Metal(IV)b
(molar%)
Phosphorusb
(molar%)
P/metal(IV) molar Total acid sitesc
Brønsted acid sitesd
Fraction of Brønsted
acid sites
(m2 gÀ1
)
ratio
(mmol gÀ1
)
(mmol gÀ1
)
ZrP1
ZrP2
ZrP3
SnP1
SnP2
ZrO2
173
276
123
11
142
143
9.45
8.29
8.07
9.20
11.03
–
12.07
16.60
15.95
6.64
7.20
N/A
1.28
2.00
1.98
0.72
0.65
N/A
1.942
2.146
1.834
0.463
1.260
0.905
0.240
0.818
0.388
0.031
0.068
0.087
0.12
0.38
0.21
0.07
0.05
0.10
a
b
c
BET ‘‘C’’ constants for the catalysts in descending order are 88, 89, 73, 44, 185, and 80.
Determined from elemental analysis (Galbraith Laboratories, Inc.).
Determined from ammonia TPD.
d
Determined from isopropylamine TPD.

R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
127
assigned to a tetrahedral phosphate connected with three zirconia
groups (Zr–O)3–PO. The high intensity peak at À27 ppm implies
relatively high amounts of polyphosphates, P–O–P, due to conden-
sation of phosphate species [54,66]. The resonance peaks associ-
ated with zirconium phosphate ZrP1 were broader and less
discrete than those for the other two samples.
-27
(a)
-21
-14
The NMR spectra were deconvoluted using Origin software. The
four resonance bands were fit with Gaussian line shapes. The band
areas were normalized according to the sample weight, assuming
equal density for the samples. Table 2 shows the relative amounts
of the different phosphate species for each catalyst. The spectrum
of ZrP1 has just three inflections and gave unreasonable fitting re-
sults until (Zr–O)–PO(OH)2 was constrained to a small value. Over-
all, from these results, it can be concluded that ZrP2 and ZrP3 are
quite similar materials with respect to their bulk properties. In
contrast, ZrP1 showed a higher amount of (Zr–O)2–PO(OH) species
at À14 ppm and less polyphosphates at À27 ppm compared to
ZrP2 and ZrP3. Panda et al. reasoned that the Brønsted acid sites
in their study of mesoporous zirconium phosphate came primarily
from geminal P(OH) groups [66]. This corresponds to the coordina-
tion state detected at À7 ppm. Others have postulated that the
Brønsted acid sites come from the phosphate species on the surface
bonded to two and three zirconia groups (chemical shifts À14 ppm
and À21 ppm, respectively) [81,83]. It is unlikely that these former
conclusions are valid for this study, as the samples with the highest
concentration of Brønsted acid sites (ZrP2 and ZrP3) are predomi-
nantly comprised of polyphosphate species. Consequently, even
though the formation of polyphosphates species is attributed with
deprotonation, their increased amount in the bulk phase could be a
primary cause of the higher Brønsted acidity observed for ZrP2 and
ZrP3 in relation to ZrP1. According to Segawa et al., P–O–P species
in the bulk phase can withdraw electrons from the residual phos-
phate groups on the surface, thus enhancing the acidic properties
of P–OH groups on the surface [84,85].
-7
(b)
(c)
20
10
0
-10
-20
δ / ppm
-30
-40
-50
-60
Fig. 2. Solid-state 31P MAS NMR spectra of the zirconium phosphate catalysts: (a)
ZrP3, (b) ZrP2, and (c) ZrP1.
of the chemical shift (dP) to more negative values is indicative of
both an increase in the number of P–O–Zr bonds [80,81], as well
as an increase in the chain length of the phosphorus atoms [67].
The latter corresponds to progressive deprotonation and subse-
quent condensation of phosphate species (P–O–P bonds), typically
during calcination [80]. Fig. 2 shows the NMR spectra for the three
zirconium phosphate samples. All of the samples showed relatively
broad peaks due to their amorphous nature [66,82]. Samples ZrP2
and ZrP3 showed very similar spectra which were characterized by
four different resonance peaks. The resonance peak at À7 ppm is
attributed to tetrahedral phosphates bonded to one zirconia group
and two hydroxyl groups (Zr–O)–PO(OH)2. This coordination gives
way to geminal P-(OH) groups. The signal at À14 ppm signifies the
presence of two zirconia and one hydroxyl group bonded to the
phosphate (Zr–O)2–PO(OH). Similarly, a peak at À21 ppm can be
3.3. XPS analysis
The zirconium phosphate catalysts were characterized by XPS
in order to study their surface composition and oxidation states.
Fig. 3 shows the high resolution XPS spectra of zirconium, phos-
phorus, and oxygen for the three zirconium phosphate catalysts.
The Zr3d line is composed of two peaks assigned to Zr3d5/2 at
183 eV and Zr3d3/2 at roughly 185 eV. This is characteristic of tet-
ravalent zirconium Zr4+ [86]. All samples had typical P2p binding
energies of 133 eV, characteristic of pentavalent tetracoordinated
phosphorus (P5+) [86,87]. The O1s line shows a broad peak which
could correspond to oxygen bonded to both zirconium and phos-
phorus [86]. The O1s component characteristic of the P–O bond
Table 2
Relative amounts of the different phosphate species for the zirconium phosphate catalysts from solid-state 31P MAS NMR spectroscopy.
Chemical shift (ppm)
Coordination statea
À7
À14
À21
À27
Catalyst
Relative amount (%)
ZrP1
ZrP2
ZrP3
2.3b
3.6
5.0
30.9
12.9
11.3
15.7
11.8
13.5
51.1
71.7
70.2
a
Adapted from Ref. [66].
This value was constrained. Any small value is acceptable.
b

128
R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
100
90
80
70
60
50
40
30
20
10
0
Zr3d5/2
(a)
Zr3d3/2
P2s
194
192
190
188
186
184
182
180
178
Binding energy (eV)
(b)
(c)
P2p
0
20
40
60
80
100 120 140 160
time (min.)
140
138
136
134
132
130
128
Fig. 4. Aqueous phase glucose dehydration at 160 °C in a stirred batch reactor with
different solid acid catalysts. Feed was 10 wt.% glucose aqueous solution. The solid
catalyst loading was constant at 5 wt.%. Catalysts: ZrP1 (j), ZrP2 (d), ZrP3 (N), SnP1
(h), SnP2 (s). Blank run (dashed line) (H).
Binding energy (eV)
O1s
lower than the theoretical ratio predetermined in the preparation
procedure (P/Zr = 3).
Electron binding energy shifts observed for Zr, P, and O can re-
flect variations in the polarity of the bonds on the surface of zirco-
nium phosphate catalysts. A slight downward shift in the Zr3d
binding energies was observed for sample ZrP1 compared to ZrP2
and ZrP3. This indicates a reduced polarization of the Zr–O bonds
in the ZrP1 sample, which could be due to fewer hydroxyl groups
on the surface relative to the other samples [88]. As for the P2p
538
536
534
532
530
528
526
Binding energy (eV)
Fig. 3. High-resolution XPS spectra of the (a) Zr3d, (b) P2p, and (c) O1s regions for
the catalysts: ZrP1 ( ); ZrP2 ( ); ZrP3 (—).
binding
energies,
the
shifts
increased
as
follows:
ZrP1 < ZrP2 < ZrP3. Higher P2p binding energies are a result of in-
creased polarity of the P–O bonds on the surface which can be en-
hanced by the hydroxyl groups bonded to the phosphorus atoms
[88]. This suggests that higher P2p binding energies are indicative
of increased amounts of acid sites on the surface, as shown with
ZrP2 and ZrP3. The O1s binding energies increase according to:
ZrP1 < ZrP3 < ZrP2. Higher O1s binding energy suggests an in-
creased amount of surface hydroxyl groups [88].
is dominant for all samples with an average binding energy of
531 eV.
The surface atomic compositions of the zirconium phosphate
catalysts appear in Table 3. These values were determined by the
XPS line areas of Zr3d, P2p, and O1s, and the phosphorus to zirco-
nium atomic ratios on the surface were calculated accordingly.
There is a gradual increase in phosphorus on the surface in agree-
ment with the increase in phosphorus loading in the preparation
procedure. The P/Zr molar ratio on the surface increased as follows:
ZrP1 < ZrP2 < ZrP3. The atomic compositions on the surfaces
showed only slight differences to the bulk compositions (see Tables
1 and 3). Catalyst ZrP1 showed practically identical P/Zr surface
and bulk molar ratios, as determined by XPS and ICP, respectively.
However, higher zirconium and phosphorus contents were ob-
served on the surface than in the bulk (Table 1). The relatively
higher amount of zirconium on the surface of ZrP1 could be a plau-
sible cause for its relatively higher concentration of Lewis acid sites
compared to ZrP2 and ZrP3. For sample ZrP2, the phosphorus con-
tent on the surface was slightly lower than in the bulk. The oppo-
site trend was found for zirconium. The highest P/Zr molar ratio on
the surface was observed for ZrP3. Nevertheless, this value was
3.4. Aqueous phase glucose dehydration studies with solid metal(IV)
phosphate catalysts
Glucose dehydration studies were carried out at 160 °C with
the different solid acid catalysts in the aqueous phase. The con-
centration of the catalysts was kept constant at 5 wt.%. Fig. 4 de-
picts glucose conversion as a function of reaction time for the
various catalysts. The tin phosphate catalysts showed a higher
activity compared to the zirconium phosphate catalysts. As part
of this study, a run without any catalyst was also carried out
and showed that glucose underwent hydrothermal decomposi-
tion. Similar observations were noted in our previous study
where we studied the kinetics of levulinic acid production from
Table 3
Electron binding energies and atomic surface composition determined by XPS for the zirconium phosphate catalysts.
Catalyst
Binding energy (eV)
Surface atomic compositiona (%)
P/Zr molar ratio
Zr3d3/2
Zr3d5/2
P2p
O1s
Zr
P
O
ZrP1
ZrP2
ZrP3
185.2
185.5
185.5
183.0
183.1
183.1
133.3
133.6
133.7
531.1
531.4
531.3
11.1
8.8
8.1
13.6
16.3
16.6
75.3
75.0
75.2
1.26
1.85
2.05
a
Based on XPS line areas of Zr3d, P2p and O1s.

R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
129
60
55
50
45
40
35
30
25
20
15
10
5
36
30
24
18
12
6
(a)
(b)
(c)
0
42
35
28
21
14
0
SnP1
SnP2
ZrP1
ZrP2
ZrP3
7
Fig. 5. Aqueous phase glucose dehydration at 160 °C in a stirred batch reactor.
Initial turn-over frequency (according to total acid sites concentration determined
from NH3-TPD) based on glucose disappearance for the different solid acid catalysts.
Feed was 10 wt.% glucose aqueous solution. The solid catalyst loading was constant
at 5 wt.%. The homogeneous glucose reactions were subtracted out for calculation
purposes.
0
25
20
15
10
5
glucose [26]. There we found that nearly full glucose disappear-
ance was attained at 180 °C after 150 min. without an acid cata-
lyst, and 75% of the overall carbon content went to form water
soluble and insoluble humic species. In this study, the catalyst
activity obtained from sample ZrP3 was lower than that from
the hydrothermal reactions with glucose at longer reaction times.
This suggests that zirconium phosphate suppresses some of the
undesired reactions encountered under hydrothermal conditions.
Fig. 5 shows the initial turnover frequency (TOF) of glucose disap-
pearance for the different catalysts in this study. The TOF calcula-
tions are based on the total concentration of acid sites as
determined by NH3-TPD. On a per site basis, the catalytic activity
increased as follows: ZrP3 < ZrP2 < ZrP1 < SnP2 < SnP1. As men-
tioned previously, results from the NH3-TPD and 1H MAS NMR
studies showed that the acid strengths for each series of metal(IV)
phosphates were similar. Therefore, it is reasonable to assume
that acid strength does not play a significant role in determining
the catalyst activity. For calculation purposes, the homogeneous
(blank) glucose reactions were subtracted out of the TOF
estimations.
Fig. 6 shows the carbon selectivities of the major products as a
function of glucose conversion. The tin phosphate catalysts had a
higher fructose selectivity compared to the zirconium phos-
phates, as shown in Fig. 6a. This is in agreement with the work
of Davis and co-workers [31–33] and Dumesic and co-workers
[34,35,89], where they find that Lewis acid sites catalyze the
isomerization reaction of glucose to fructose. The fructose selec-
tivity increased as follows: blank < ZrP3 < ZrP2 < ZrP1 < SnP1 ꢀ
SnP2. In all cases, the fructose selectivity decreased with in-
creased glucose conversion. It is notable to mention that lactic
acid was also detected as a by-product of this reaction. The
amounts were not quantified, but were observed for all of the so-
lid catalysts tested here due to the presence of Lewis acid sites.
Recently, Chambon et al. published a study on the production
of lactic acid from cellulose with solid Lewis acid catalysts [90].
They found that Lewis sites improve the extent of cellulose depo-
lymerization compared to Brønsted acid sties. Holm et al. also re-
cently discovered that solid Lewis acid catalysts, such as Sn-Beta,
are able to convert pentose and hexose sugars to methyl lactate
in methanol [89,91].
0
0
10 20 30 40 50 60 70 80 90 100
Glucose conversion (%)
Fig. 6. Carbon selectivity of major products as a function of glucose conversion for
different metal phosphate catalysts at 160 °C: (a) fructose, (b) HMF, and (c) levulinic
acid. Feed was 10 wt.% glucose aqueous solution. The solid catalyst loading was
constant at 5 wt.%. Catalysts: ZrP1 (j), ZrP2 (d), ZrP3 (N), SnP1 (h), SnP2 (s). Blank
run (dashed line) (H).
being produced from glucose and fructose. The HMF selectivity
then decreased as it reacted to produce levulinic acid, formic acid,
and decomposition products. Higher HMF selectivities were ob-
served for ZrP2 and ZrP3 at relatively low glucose conversions.
This coincides with the trend that was observed for catalysts with
relatively high Brønsted to Lewis ratios [40]. The HMF selectivity
reached a maximum of 38% with ZrP2 at a glucose conversion of
38%. Similar selectivities were achieved with the tin phosphate
catalysts at a glucose conversion of 80%. The lowest selectivity
was observed for non-catalyzed glucose decomposition (blank
run).
A relatively high levulinic acid selectivity was observed for the
zirconium phosphate catalysts ZrP2 and ZrP3 compared to the tin
phosphates, as shown in Fig. 6c. For the tin phosphates, levulinic
acid was detected only at glucose conversions above 80%. The cat-
alyst ZrP2 showed the highest levulinic acid selectivity of 22% at a
glucose conversion of 64%. In contrast, catalyst ZrP1 showed no
production of levulinic acid. No levulinic acid was observed when
only water was used. The levulinic acid selectivity increased as fol-
lows: blank ꢀ ZrP1 < SnP2 < SnP1 < ZrP3 < ZrP2. In all cases, the
levulinic acid selectivity increased with increasing glucose
conversion.
In addition to the major products mentioned earlier, quantifi-
able amounts of cellobiose, levoglucosan, and furfural were also
detected as by-products for all of the catalysts tested in this study.
Cellobiose was detected primarily for catalysts with relatively high
Brønsted to Lewis ratios, such as ZrP2 and ZrP3. Carbon selectivity
to cellobiose decreased with glucose conversion. The maximum
carbon selectivity was 36% at 8% glucose conversion for ZrP3.
Fig. 6b depicts the HMF selectivity for the different metal(IV)
phosphates. Initially, the HMF selectivity increased as it was

130
R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
3.5. Aqueous phase dehydration of glucose with homogeneous acid
catalysts
100
90
80
70
60
50
40
30
20
10
0
Aqueous phase glucose dehydration studies were carried out
with water-soluble Lewis and Brønsted acids. Ytterbium(III) tri-
fluoromethanesulfonate hydrate, Yb(OTf)3, and hydrochloric acid
were used, respectively. The former is considered a stable water-
soluble Lewis acid [92,93]. Reactions were performed at varying
Brønsted to Lewis ratios by combining the two acids accordingly,
while holding the total acid concentration constant at 0.1 M.
Fig. 7 shows the results of glucose conversion for the different ra-
tios of the homogeneous catalysts. The activity for glucose disap-
pearance increased systematically as the relative amount of
Lewis sites increased. Pure Yb(OTf)3 showed the highest glucose
conversion, whereas pure HCl had the lowest catalytic activity.
These results indicate that the Lewis acid sites have a higher cata-
lytic activity toward glucose disappearance compared to the
Brønsted acid sites, which is consistent with the results from the
heterogeneous catalysts.
Fig. 8 depicts the carbon selectivities of the major compounds
as a function of glucose conversion for the various Brønsted to Le-
wis ratios in the homogeneous regime. The selectivity for fructose
production increased with the concentration of Lewis sites, as
shown in Fig. 8a, reaching a maximum of 34% at 45% conversion
with pure Yb(OTf)3 as a catalyst. Fig. 8b shows the HMF selectivity
for the different Brønsted to Lewis ratios in the homogeneous re-
gime. As with the solid acid catalysts, a maximum selectivity was
observed for all ratios. This maximum value shifts toward higher
glucose conversions as the relative amount of Lewis acid sites in-
creases. At lower glucose conversions, HMF was selectively pro-
duced by catalysts with more Brønsted acid sites. Lewis acid sites
showed a higher selectivity for HMF production at higher glucose
conversions. The selectivity toward levulinic acid production from
glucose systematically increases with the concentration of
Brønsted sites, as depicted in Fig. 8c. Pure HCl showed a signifi-
cantly higher levulinic acid selectivity compared to Yb(OTf)3, with
a maximum selectivity of 46% at 60% glucose conversion. These re-
sults are consistent with those reported for the heterogeneous
catalysts.
0
20
40
60
80
100 120 140 160
time (min.)
Fig. 7. Aqueous phase glucose dehydration in a stirred batch reactor. Effect of the
Brønsted to Lewis acid ratio on glucose conversion with homogeneous acid
catalysts at 160 °C. Feed was 10 wt.% glucose aqueous solution. Total acid
concentration was kept constant at 0.1 M. Brønsted to Lewis acid ratio = 1:0 (j),
3:1 (d), 1:1 (N), 1:3 ( ), and 0:1 (r). Blank run (dashed line) (H).
36
(a)
30
24
18
12
6
0
21
(b)
18
15
12
9
6
3
4. Discussion
0
The reaction chemistry for levulinic acid production from glu-
cose consists of three key steps. Glucose undergoes isomerization
to produce fructose and mannose; hexoses undergo dehydration
to form HMF, and HMF undergoes a rehydration reaction to pro-
duce levulinic acid and formic acid. Additional undesirable decom-
position reactions can also come about from glucose, fructose, and
HMF to ultimately form humins. For this study, humins were con-
sidered as all unidentified water soluble and insoluble compounds.
Further characterization of humic species was not carried out in
this study, as this goes beyond the scope of this paper. Table 4
shows the product distribution of all identified products for the
metal(IV) phosphates. The values were interpolated to correspond
with a glucose conversion of 40%. The results show that the forma-
tion of humins was predominant for the reaction with ZrP1. The
catalyst ZrP2 showed the lowest selectivity toward these unde-
sired compounds, which is consistent with its high selectivity to-
ward HMF and levulinic acid. The tin phosphate catalysts showed
similar trends to each other for humins formation. Following the
reactions in this study, it was observed that solid humins were
deposited on all of the solid acid catalysts. This was also noticed
in our recent study with ZrP for cellulose decomposition [43].
The amounts of carbon deposited on the catalysts were not deter-
mined, and regeneration was not carried out in this publication.
50
(c)
40
30
20
10
0
0
10 20 30 40 50 60 70 80 90 100
Glucose conversion (%)
Fig. 8. Effect of the Brønsted to Lewis acid ratio on the carbon selectivity of major
products from glucose dehydration with homogeneous acid catalysts at 160 °C: (a)
fructose, (b) HMF, and (c) levulinic acid. Feed was 10 wt.% glucose aqueous solution.
Total acid concentration was kept constant at 0.1 M. Brønsted to Lewis acid
ratio = 1:0 (j), 3:1 (d), 1:1 (N), 1:3 ( ), and 0:1 (r).
Similar trends were observed for levoglucosan with a maximum
carbon selectivity of 13% at 8% at glucose conversion for ZrP3.
Furfural was also detected, but at carbon selectivities no higher
than 4%. Higher selectivities to furfural were observed for catalysts
with higher amounts of Lewis acid sites.

R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
131
Table 4
Product distribution at 40% glucose conversion for the metal(IV) phosphate catalysts at 160 °C. Feed was 10 wt.% glucose aqueous solution. The solid catalyst loading was constant
at 5 wt.%.
Catalyst
Carbon selectivity (%)
Fructose
Cellobiose
Levoglucosan
HMF
Furfural
Levulinic acid
Formic acid
Huminsa
ZrP1
ZrP2
ZrP3
SnP1
SnP2
12.6
4.6
2.3
21.6
18.7
1.8
6.2
6.3
3.7
3.3
0.6
3.8
4.0
1.7
1.4
15.6
37.5
30.4
23.9
23.2
2.3
1.5
1.0
2.4
2.2
0.0
17.3
10.5
0.0
4.1
5.4
3.9
1.2
1.8
63.0
23.7
41.6
45.5
49.4
0.0
a
Considered all unidentified water soluble and insoluble products.
70
60
50
40
30
20
10
0
100
90
80
70
60
50
40
30
20
10
0
0
0.25
0.5
0.75
1
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Fraction of Brønsted acid sites
ZrP2
ZrP1
ZrP3
Fig. 10. Carbon selectivity of water-soluble products as a function of the fraction of
Brønsted acid sites for the homogeneous acid catalysts at 50% glucose conversion
and 160 °C. Total acid concentration was kept constant at 0.1 M. Feed was 10 wt.%
Fraction of Brønsted acid sites
Fig. 9. Carbon selectivity of major products at 40% glucose conversion as a function
of the concentration of Brønsted acid sites (determined by isopropylamine TPD) for
the zirconium phosphate catalysts at 160 °C. Feed was 10 wt.% glucose aqueous
solution. The solid catalyst loading was constant at 5 wt.%. Major products: fructose
(j), HMF (d), levulinic acid (N), humins (H).
glucose aqueous solution.
fructose;
cellobiose;
levoglucosan;
HMF;
furfural; levulinic acid;
formic acid;
humins.
suggesting this type of acid site predominantly catalyzes the rehy-
dration of HMF to form levulinic acid and formic acid. In contrast,
Lewis acid sites favor the isomerization reaction of glucose to fruc-
tose. Similar conclusions have been reported in the past by Davis
and co-workers in their studies of tin-containing zeolites as highly
active solid Lewis acid catalysts [31–33] and by Dumesic and
Shanks [34,35]. As with the dehydration step, it is apparent from
the results that both Brønsted and Lewis acid sites catalyze the
production of HMF from glucose and fructose, reaching a maxi-
mum carbon selectivity of 19% at 50% glucose conversion with a
Brønsted acid site fraction of 0.25. The selectivity to humins does
not vary much as a function of the Brønsted to Lewis acid site ratio;
however, a slight decrease is observed at higher fractions of
Brønsted sites.
Two classes of metal(IV) phosphates have been prepared based
on tin and zirconium precursors. Despite the same phosphorus to
metal(IV) molar loadings in the preparation procedure, the tin
phosphates have lower phosphorus concentrations than the zirco-
nium phosphates according to elemental analysis. The phosphorus
to metal(IV) ratios obtained for the zirconium phosphates are more
than two fold the amount obtained for the tin-based analogs. Con-
sequently, the zirconium phosphates all have higher total acid and
Brønsted acid concentrations according to ammonia and isopropyl-
amine TPD, respectively. Prior studies which have compared differ-
ent types of metal(IV) phosphates have reported conflicting results.
Patel et al. [47,50] compared tin, zirconium, and titanium phos-
phates for cyclodehydration of 1,n-diols. They found tin phosphate
to be the most active and selective catalyst, which they attributed
to its relatively high surface area and strong acid sites. Similar
However, we have shown in our previous studies that we can ob-
tain a satisfactory carbon balance for these aqueous dehydration
reactions based on various characterization techniques [26,43].
This indicates that the method used in this study to calculate the
product selectivities is valid.
Fig. 9 shows the carbon selectivity of the major products as a
function of the fraction of Brønsted acid sites for the three zirco-
nium phosphate catalysts. The values were interpolated to corre-
spond with
a glucose conversion of 40%. There is a direct
correlation between the product distribution and the fraction of
Brønsted sites as determined by TPD measurements with ammonia
and isopropylamine. HMF and levulinic acid production increase
with the amount of Brønsted acid sites. Isomerization to produce
fructose from glucose is predominant at higher Lewis acid concen-
trations, as shown for sample ZrP1. The formation of humic species
is greatest at lower Brønsted to Lewis ratios. Consequently, a high-
er rate of glucose disappearance and lower selectivities to HMF and
levulinic acid are observed for catalysts with increased Lewis sites.
This claim is further supported by the studies carried out with
homogeneous Brønsted and Lewis acid catalysts. Fig. 10 shows
the carbon selectivities of all the identified products as a function
of the fraction of Brønsted acid sites for the homogeneous acid cat-
alysts. These calculated values were interpolated to correspond
with a glucose conversion of 50%. Levulinic acid and formic acid
selectivities increase as the fraction of Brønsted sites increases. A
minimal HMF selectivity is observed with pure Brønsted acid sites

132
R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
conclusions were deduced by Gu et al. [59] in their study of sorbi-
tol dehydration to produce isosorbide. It is notable to mention that
the catalysts tested in these previous studies were not calcined.
The discrepancy with our study could be attributed to the differ-
ences in catalyst preparation procedures. Similarly, Tran et al. have
shown a direct correlation between the acidic and catalytic proper-
ties of sulfated zirconia and its calcination temperature [94]. It has
been reported in the literature that the concentration of Brønsted
acid sites for crystalline zirconium phosphate is maximized at a
calcination temperature of 400 °C [65].
The bulk and surface properties of the zirconium phosphate cat-
alysts differ according to their phosphorus loadings. Results from
elemental analysis and 31P solid-state NMR spectroscopy show
that catalysts ZrP2 and ZrP3 are quite similar in the bulk phase. De-
spite the different phosphorus loadings in the preparation proce-
dure, both catalysts contain nearly identical phosphorus amounts
in the bulk phase, as well as on the surface. The highest attainable
phosphorus to zirconium molar ratio is 2:1. This is similar to the
maximum P/Zr ratio of 2.21 reported by Sinhamahapatra et al.
who studied the effect of phosphate concentration and calcination
temperature on the catalytic properties of mesoporous zirconium
phosphate [66].
Characterization techniques carried out in this study show
noticeable differences in surface properties among the three zirco-
nium phosphate catalysts. As shown with ZrP1 and ZrP2, increased
phosphorus loading results in higher phosphorus content in the
bulk and on the surface. This in turn increases the total acidity
and Brønsted acidity [66]. Conversely, a higher zirconium loading
increases the Lewis acidity due to increased amounts of tetravalent
zirconium (Zr4+), as shown for ZrP1. Further increasing the phos-
phorus loading in the preparation step from a P/Zr molar ratio of
2 to 3 does not increase the phosphorus content in the bulk or
on the surface, as shown for ZrP2 and ZrP3. Among the three zirco-
nium phosphate catalysts, ZrP2 has the highest surface area, as
well as the highest overall acid and Brønsted acid concentrations.
Consequently, ZrP2 was found to contain the highest amount of
hydroxyl groups on its surface, as confirmed with XPS analyses.
This could explain its high concentration of Brønsted acid sites
[44]. Kellum and Hahn found a direct correlation between the con-
centration of surface hydroxyl groups and the surface area for a
series of trimethylsiloxy-treated ammonium silicates [95]. There-
fore, the increased amount of surface hydroxyl groups on ZrP2
could also be a cause for its relatively high surface area. On the
other hand, other studies have claimed that the preparation condi-
tions have a strong effect on the structure and surface properties of
zirconia catalysts. Specifically, the nature of the zirconium precur-
sor, the pH of the solution during precipitation, temperature and
time of digestion, and calcination temperature all play key roles
that influence these properties [96–99]. In this study, the zirco-
nium phosphate catalysts differ by the P/Zr molar ratios in the
preparation step. This was achieved by varying the relative
amounts of phosphorus and zirconium precursors (ammonium
phosphate monobasic and zirconium oxychloride octahydrate,
respectively). This in turn alters the pH of the solution during pre-
cipitation, which could be a cause for varying surface areas ob-
served for the different samples.
conclusions are entirely valid for this study, as the relative
amounts of the polyphosphate species for ZrP2 and ZrP3 are nearly
identical (Table 2), whereas ZrP2 shows a higher concentration of
acid sites compared to ZrP3. However, it has been reported in the
literature that the length of the polyphosphate chain is a function
of the metal oxide to phosphate ratio [100]. An increase in the
polyphosphate chain length is observed with increased amounts
of P2O5. Determining the composition and chain length of the poly-
phosphate species was not a focus of this study; however, differ-
ences in these parameters between ZrP2 and ZrP3 could be a
reason for the discrepancies in their catalytic properties. Determin-
ing the optimal phosphate species for this reaction is beyond the
scope of this paper and will be the focus of future studies. Another
analogous study was reported by Mishra and Parida who examined
the effect of sulfate loading on sulfated zirconia catalysts [71].
According to their results, increasing the sulfate loading from
10 wt.% to 15 wt.% resulted in decreased sulfur content as deter-
mined by elemental analysis. The increase in sulfate loading also
resulted in inferior catalytic properties including lower surface
area, lower overall acidity, and decreased surface hydrophilicity.
Another study was carried out by Ahmed et al. on the effect of sul-
fate loading with sulfated zirconia [101]. They incorporated SO24À in
zirconia ranging from 5 to 30 wt.% and found the catalyst with a
loading of 15 wt.% sulfate to possess the highest surface area, as
well as highest total acidity and Brønsted to Lewis ratio. Conse-
quently, this catalyst was found to exhibit optimal catalytic activ-
ity for ethanol dehydration.
5. Conclusions
We have prepared and characterized a series of metal(IV) phos-
phate catalysts and tested them for aqueous phase dehydration of
glucose to levulinic acid. Adsorption studies with ammonia and
isopropylamine as probe molecules reveal a higher overall concen-
tration of acid sites for the zirconium phosphates compared to the
tin phosphate catalysts. Sample ZrP2 shows the highest amount of
total acid sites, as well as the highest concentration of Brønsted
sites among all of the catalysts tested. XPS analysis corroborates
these findings by revealing a high concentration of surface hydro-
xyl groups for the zirconium phosphate catalysts, specifically ZrP2.
Four phosphorus coordination states have been identified by solid-
state 31P MAS NMR spectroscopy, among which the polyphosphate
species has the highest relative amount. The higher amounts of
polyphosphate species detected in ZrP2 and ZrP3 could be a reason
for its enhanced acidity compared to ZrP1. Likewise, the length of
the polyphosphate chain could also play a key role in determining
the concentration of the acid sites.
We have demonstrated here that both heterogeneous and
homogeneous Lewis and Brønsted sites share different functions
as related to the proposed reaction scheme for glucose dehydra-
tion. The catalytic activity and selectivity for all of the metal(IV)
phosphates tested in this study vary according to the Brønsted to
Lewis ratio. Catalysts with high Lewis acidity, such as the tin phos-
phates, show the highest activity on a per site basis, whereas the
zirconium phosphates with relatively high Brønsted acidity (i.e.
ZrP2 and ZrP3) demonstrate the lowest activity. Fructose selectiv-
ity increases with an increase in the Lewis acid concentration of
the catalyst. This is due to induced isomerization reaction cata-
lyzed by Lewis acid sites. Both types of acid sites catalyze the dehy-
dration reaction to produce HMF from glucose. However, the HMF
selectivity increases with increased concentration of Brønsted acid
sites, particularly at lower glucose conversions for the heteroge-
neous catalyst. The levulinic acid selectivity is also a function of
the relative concentration of Brønsted to Lewis sites. The levulinic
acid selectivity increases with an increase in the Brønsted to Lewis
Various groups have studied the effect of phosphate and sulfate
loading on different acid catalysts. Sinhamahapatra et al. discov-
ered that a phosphate to zirconium ratio of 2 yields the highest
surface area as well as the highest concentration of total acid sites
and Brønsted acid sites as determined by NH3-TPD and DRIFTS
spectra for pyridine adsorption, respectively [66]. A further in-
crease in phosphate loading (P/Zr ratio 3) resulted in a decrease
in total acidity and Brønsted acidity. They reasoned the decrease
in Brønsted acidity was due to the formation of polyphosphate,
which in turn diminishes the P–OH groups. It is unlikely that these

R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
133
[17] D.C. Elliott, J.G. Frye, Hydrogenated 5-Carbon Compound and Method of
Making, US, 1999.
[18] S. Bayan, E. Beati, La Chim. L’Ind. 23 (1941) 432–434.
[19] H. Heeres, R. Handana, D. Chunai, C. Borromeus Rasrendra, B. Girisuta, H. Jan
Heeres, Green Chem. 11 (2009) 1247–1255.
ratio for both the heterogeneous and homogeneous acid catalysts.
The formation of humic species increases with increased Lewis
acidity for the heterogeneous catalysts.
Increase in the P/metal(IV) molar ratio from 1 to 2 in the precur-
sor solution results in a higher surface area, as well as increased
overall concentration of acid sites and Brønsted acid sites. How-
ever, for the tin phosphates, the fraction of Brønsted sites remains
the same for both catalysts in the series (SnP1 and SnP2). Conse-
quently, both catalysts show nearly identical selectivities for the
major reaction products. For the zirconium phosphates, a further
increase in phosphorus loading to a P/Zr molar ratio of 3 does
not alter the bulk phase of the catalyst, rather only the surface
properties. Moreover, all three zirconium phosphates are predom-
inantly comprised of phosphate species in the polyphosphate form.
Among the zirconium phosphate catalysts, the catalyst with a P/
Zr ratio of 2 (ZrP2) exhibits the highest surface area, as well as the
highest amount of total acid sites and fraction of Brønsted sites.
This catalyst in turn exhibits the highest selectivity to HMF and
levulinic acid production. A relatively high concentration of surface
hydroxyl groups is most likely the source of the relatively high
amounts of Brønsted acid sites, as well as the high surface area.
Differences in preparation conditions, such as pH, could also lead
to variations in the surface properties, as well as deviations in
the composition of the phosphate species in the bulk. This study
lays the grounds for further design of improved hydrothermally
stable solid acid catalysts displaying high activity and selectivity
to HMF and levulinic acid from carbohydrates.
[20] L. Deng, Y. Zhao, J. Li, Y. Fu, B. Liao, Q.-X. Guo, ChemSusChem 3 (2010) 1172–
1175.
[21] D.J. Braden, C.A. Henao, J. Heltzel, C.C. Maravelias, J.A. Dumesic, Green Chem.
13 (2011) 1755–1765.
[22] I.T. Horvath, H. Mehdi, V. Fabos, L. Boda, L.T. Mika, Green Chem. 10 (2008)
238–242.
[23] J.-P. Lange, R. Price, P.M. Ayoub, J. Louis, L. Petrus, L. Clarke, H. Gosselink,
Angew. Chem. Int. Ed. 49 (2010) 4479–4483.
[24] J.Q. Bond, D.M. Alonso, D. Wang, R.M. West, J.A. Dumesic, Science 327 (2010)
1110–1114.
[25] D.W. Rackemann, W.O.S. Doherty, Biofuels Bioprod. Biorefin. 5 (2011) 198–
214.
[26] R. Weingarten, J. Cho, R. Xing, W.C. Conner Jr., G.W. Huber, ChemSusChem 5
(2012) 1280–1290.
[27] K.R. Heimlich, A.N. Martin, J. Am. Pharm. Assoc. Sci. Ed. 49 (1960) 592–597.
[28] P.C. Smith, H.E. Grethlein, A.O. Converse, Sol. Energy 28 (1982) 41–48.
[29] H.E.v. Dam, A.P.G. Kieboom, H.v. Bekkum, Starch 38 (1986) 95–101.
[30] Q. Xiang, Y. Lee, R. Torget, Appl. Biochem. Biotechnol. 115 (2004) 1127–
1138.
[31] M. Moliner, Y. Román-Leshkov, M.E. Davis, Proc. Natl. Acad. Sci. 107 (2010)
6164–6168.
[32] Y. Román-Leshkov, M. Moliner, J.A. Labinger, M.E. Davis, Angew. Chem. Int.
Ed. 49 (2010) 8954–8957.
[33] E. Nikolla, Y. Román-Leshkov, M. Moliner, M.E. Davis, ACS Catal. 1 (2011)
408–410.
[34] Y.J. Pagán-Torres, T. Wang, J.M.R. Gallo, B.H. Shanks, J.A. Dumesic, ACS Catal. 2
(2012) 930–934.
[35] T. Wang, Y. Pagán-Torres, E. Combs, J. Dumesic, B. Shanks, Top. Catal. (2012)
1–6.
[36] K. Kato, T. Doihara, F. Sakai, N. Takahashi, Kenkyu Hokoku – Nippon Senbai
Kosha Chuo Kenkyusho 108 (1966) 361–364.
[37] A. Ohnishi, K. Kato, E. Takagi, Polym. J. 7 (1975) 431–437.
[38] R. Krishna, M.R. Kallury, C. Ambidge, T.T. Tidwell, D.G.B. Boocock, F.A.
Agblevor, D.J. Stewart, Carbohydr. Res. 158 (1986) 253–261.
[39] G.C.A. Luijkx, F.v. Rantwijk, H.v. Bekkum, Carbohydr. Res. 242 (1993) 131–
139.
[40] R. Weingarten, G.A. Tompsett, W.C. Conner Jr., G.W. Huber, J. Catal. 279
(2011) 174–182.
[41] A.P. Dunlop, Ind. Eng. Chem. 40 (1948) 204–209.
[42] D.L. Williams, A.P. Dunlop, Ind. Eng. Chem. 40 (1948) 239–241.
[43] R. Weingarten, W.C. Conner Jr., G.W. Huber, Energy Environ. Sci. 5 (2012)
7559–7574.
[44] A. Clearfield, D.S. Thakur, J. Catal. 65 (1980) 185–194.
[45] A.L. Ginestra, P. Patrono, M.L. Berardelli, P. Galli, C. Ferragini, M.A. Massucci, J.
Catal. 103 (1987) 346–356.
[46] R.A.W. Johnstone, J.-Y. Liu, D. Whittaker, J. Chem. Soc. Perkin Trans. 2 (6)
(1998) 1287–1288.
[47] S.M. Patel, U.V. Chudasama, P.A. Ganeshpure, Green Chem. 3 (2001) 143–145.
[48] A. Clearfield, D.S. Thakur, Appl. Catal. 26 (1986) 1–26.
[49] A.L. Ginestra, P. Patrono, Mater. Chem. Phys. 17 (1987) 161–179.
[50] S.M. Patel, U.V. Chudasama, P.A. Ganeshpure, React. Kinet. Catal. Lett. 76
(2002) 317–325.
Acknowledgments
This material is based upon work supported as part of the Insti-
tute for Atom-efficient Chemical Transformations (IACT), an Energy
Frontier Research Center funded by the US Department of Energy,
Office of Science, Office of Basic Energy Sciences. The authors
would also like to gratefully acknowledge financial support from
NSF-CBET (Grant # 0756663) and NSF MRI (Grant # 0722802).
We also sincerely thank Jacob Hirsch from the Department of Poly-
mer Science and Engineering at the University of Massachusetts–
Amherst for performing the XPS measurements and the data
analysis.
Appendix A. Supplementary material
[51] Y. Kamiya, S. Sakata, Y. Yoshinaga, R. Ohnishi, T. Okuhara, Catal. Lett. 94
(2004) 45–47.
[52] J. Jimenez-Jimenez, J. Merida-Robles, E. Rodriguez-Castellon, A. Jimenez-
Lopez, M. Lopez Granados, S. del Val, I. Melian Cabrera, J.L.G. Fierro, Catal.
Today 99 (2005) 179–186.
[53] C. Srilakshmi, K. Ramesh, P. Nagaraju, N. Lingaiah, P.S.S. Prasad, Catal. Lett.
106 (2006) 115–122.
[54] K.N. Rao, A. Sridhar, A.F. Lee, S.J. Tavener, N.A. Young, K. Wilson, Green Chem.
8 (2006) 790–797.
[55] C. Carlini, M. Giuttari, A. Maria Raspolli Galletti, G. Sbrana, T. Armaroli, G.
Busca, Appl. Catal. A 183 (1999) 295–302.
[56] T. Armaroli, G. Busca, C. Carlini, M. Giuttari, A.M. Raspolli Galletti, G. Sbrana, J.
Mol. Catal. A: Chem. 151 (2000) 233–243.
[57] F. Benvenuti, C. Carlini, P. Patrono, A.M. Raspolli Galletti, G. Sbrana, M.A.
Massucci, P. Galli, Appl. Catal. A 193 (2000) 147–153.
[58] C. Carlini, P. Patrono, A.M.R. Galletti, G. Sbrana, Appl. Catal. A 275 (2004) 111–
118.
[59] M. Gu, D. Yu, H. Zhang, P. Sun, H. Huang, Catal. Lett. 133 (2009) 214–220.
[60] N. Li, G.A. Tompsett, G.W. Huber, ChemSusChem 3 (2010) 1154–1157.
[61] N. Li, G.A. Tompsett, T. Zhang, J. Shi, C.E. Wyman, G.W. Huber, Green Chem. 13
(2011) 91–101.
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2013.03.023.
References
[1] G.W. Huber, J.A. Dumesic, Catal. Today 111 (2006) 119–132.
[2] L.R. Lynd, J.H. Cushman, R.J. Nichols, C.E. Wyman, Science 251 (1991) 1318–
1323.
[3] R.R. Davda, J.W. Shabaker, G.W. Huber, R.D. Cortright, J.A. Dumesic, Appl.
Catal. B 56 (2005) 171–186.
[4] J.N. Chheda, J.A. Dumesic, Catal. Today 123 (2007) 59–70.
[5] J.N. Chheda, G.W. Huber, J.A. Dumesic, Angew. Chem. Int. Ed. 46 (2007) 7164–
7183.
[6] D.A. Simonetti, J.A. Dumesic, ChemSusChem 1 (2008) 725–733.
[7] D.A. Simonetti, J.A. Dumesic, Catal. Rev. 51 (2009) 441–484.
[8] D.M. Alonso, J.Q. Bond, J.A. Dumesic, Green Chem. 12 (2010) 1493–1513.
[9] J.C. Serrano-Ruiz, R.M. West, J.A. Dumesic, Annu. Rev. Chem. Biomol. Eng. 1
(2010) 79–100.
[10] J.C. Serrano-Ruiz, J.A. Dumesic, Energy Environ. Sci. 4 (2011) 83–99.
[11] G.W. Huber, S. Iborra, A. Corma, Chem. Rev. 106 (2006) 4044–4098.
[12] L. Petrus, M.A. Noordermeer, Green Chem. 8 (2006) 861–867.
[13] A. Corma, S. Iborra, A. Velty, Chem. Rev. 107 (2007) 2411–2502.
[14] J.J. Bozell, L. Moens, D.C. Elliott, Y. Wang, G.G. Neuenscwander, S.W.
Fitzpatrick, R.J. Bilski, J.L. Jarnefeld, Resour. Conserv. Recycl. 28 (2000) 227–
239.
[15] P. Gullón, A. Romaní, C. Vila, G. Garrote, J.C. Parajó, Biofuels Bioprod. Biorefin.
6 (2012) 219–232.
[16] S.W. Fitzpatrick, ACS Symp. Ser. 921 (2006) 271–287.
[62] F.S. Asghari, H. Yoshida, Carbohydr. Res. 341 (2006) 2379–2387.
[63] R.M. West, D.J. Braden, J.A. Dumesic, J. Catal. 262 (2009) 134–143.
[64] Y.J. Pagán-Torres, J.M.R. Gallo, D. Wang, H.N. Pham, J.A. Libera, C.L. Marshall,
J.W. Elam, A.K. Datye, J.A. Dumesic, ACS Catal. 1 (2011) 1234–1245.
[65] T. Hattori, A. Ishiguro, Y. Murakami, J. Inorg. Nucl. Chem. 40 (1978) 1107–
1111.
[66] A. Sinhamahapatra, N. Sutradhar, B. Roy, A. Tarafdar, H.C. Bajaj, A.B. Panda,
Appl. Catal. A 385 (2010) 22–30.

134
R. Weingarten et al. / Journal of Catalysis 304 (2013) 123–134
[67] D. Spielbauer, G.A.H. Mekhemer, T. Riemer, M.I. Zaki, H. Knözinger, J. Phys.
Chem. B 101 (1997) 4681–4688.
[84] K.-I. Segawa, Y. Kurusu, Y. Nakajima, M. Kinoshita, J. Catal. 94 (1985) 491–
500.
[68] O. Kresnawahjuesa, R.J. Gorte, D.d. Olivera, L.Y. Lau, Catal. Lett. 82 (2002)
155–160.
[85] K.-I. Segawa, Y. Nakajima, S.-I. Nakata, S. Asaoka, H. Takahashi, J. Catal. 101
(1986) 81–89.
[69] J.G. Tittensor, R.J. Gorte, D.M. Chapman, J. Catal. 138 (1992) 714–720.
[70] R.J. Gorte, Catal. Lett. 62 (1999) 1–13.
[86] Z.-Y. Yuan, T.-Z. Ren, A. Azioune, J.-J. Pireaux, B.-L. Su, Catal. Today 105 (2005)
647–654.
[71] H.K. Mishra, K.M. Parida, Appl. Catal. A 224 (2002) 179–189.
[72] U. Ciesla, G. Stucky, F. Schiith, F.B.C.D.S.G., L. Bonneviot, S. Kaliaguine, Studies
in Surface Science and Catalysis, Elsevier, pp. 527–534.
[73] R.R. Pinto, P. Borges, M.A.N.D.A. Lemos, F. Lemos, J.C. Védrine, E.G. Derouane,
F.R. Ribeiro, Appl. Catal. A 284 (2005) 39–46.
[74] P. Batamack, C. Dorémieux-Morin, R. Vincent, J. Fraissard, Chem. Phys. Lett.
180 (1991) 545–550.
[75] P. Batamack, C. Dorémieux-Morin, R. Vincent, J. Fraissard, J. Phys. Chem. 97
(1993) 9779–9783.
[76] L. Heeribout, V. Semmer, P. Batamack, C. Dorémieux-Morin, R. Vincent, J.
Fraissard, Studies in Surface Science and Catalysis, Elsevier, pp. 831–840.
[77] V. Semmer, P. Batamack, C. Dorémieux-Morin, R. Vincent, J. Fraissard, J. Catal.
161 (1996) 186–193.
[78] L. Heeribout, V. Semmer, P. Batamack, C. Dorémieux-Morin, J. Fraissard,
Microporous Mesoporous Mater. 21 (1998) 565–570.
[79] V. Semmer-Herlédan, L. Heeribout, P. Batamack, C. Dorémieux-Morin, J.
Fraissard, A. Gola, E. Benazzi, Microporous Mesoporous Mater. 34 (2000)
157–169.
[87] A.M. Puziy, O.I. Poddubnaya, A.M. Ziatdinov, Appl. Surf. Sci. 252 (2006) 8036–
8038.
[88] J.L. Colón, D.S. Thakur, C.-Y. Yang, A. Clearfield, C.R. Martin, J. Catal. 124 (1990)
148–159.
[89] M.S. Holm, Y.J. Pagan-Torres, S. Saravanamurugan, A. Riisager, J.A. Dumesic, E.
Taarning, Green Chem. 14 (2012) 702–706.
[90] F. Chambon, F. Rataboul, C. Pinel, A. Cabiac, E. Guillon, N. Essayem, Appl. Catal.
B 105 (2011) 171–181.
[91] M.S. Holm, S. Saravanamurugan, E. Taarning, Science 328 (2010) 602–
605.
[92] T. Okuhara, Chem. Rev. 102 (2002) 3641–3666.
[93] A. Corma, H. Garcia, Chem. Rev. 103 (2003) 4307–4365.
[94] M.T. Tran, N.S. Gnep, G. Szabo, M. Guisnet, Appl. Catal. A 171 (1998) 207–
217.
[95] G.E. Kellum, J.R. Hahn, Anal. Chem. 40 (1968) 952–956.
[96] D. Tichit, D. El Alami, F. Figueras, Appl. Catal. A 145 (1996) 195–210.
[97] G.K. Chuah, S. Jaenicke, S.A. Cheong, K.S. Chan, Appl. Catal. A 145 (1996) 267–
284.
[80] Y. Sun, P. Afanasiev, M. Vrinat, G. Coudurier, J. Mater. Chem. 10 (2000) 2320–
2324.
[81] W. Liu, Z. Song, T. Ikegawa, H. Nishiguchi, T. Ishihara, Y. Takita, Mater. Lett. 58
(2004) 3328–3331.
[98] G.D. Yadav, J.J. Nair, Microporous Mesoporous Mater. 33 (1999) 1–48.
[99] K.T. Jung, A.T. Bell, J. Mol. Catal. A: Chem. 163 (2000) 27–42.
[100] J. Griffith Edward, Phosphorus Chemistry, American Chemical Society, pp.
86–101.
[82] G.L. Zhao, Z.Y. Yuan, T.H. Chen, Mater. Res. Bull. 40 (2005) 1922–1928.
[83] J. Jiménez-Jiménez, P. Maireles-Torres, P. Olivera-Pastor, E. Rodríguez-
Castellón, A. Jiménez-López, D.J. Jones, J. Rozière, Adv. Mater. 10 (1998)
812–815.
[101] A.I. Ahmed, S.A. El-Hakam, S.E. Samra, A.A. El-Khouly, A.S. Khder, Colloids
Surf., A 317 (2008) 62–70.