Solvent-Activated Hafnium-Containing Zeolites Enable Selective and Continuous Glucose–Fructose Isomerisation

Article abstract of DOI:10.1002/anie.202006718

The isomerisation of glucose to fructose is a critical step towards manufacturing petroleum-free chemicals from lignocellulosic biomass. Herein we show that Hf-containing zeolites are unique catalysts for this reaction, enabling true thermodynamic equilibrium to be achieved in a single step during intensified continuous operation, which no chemical or biological catalyst has yet been able to achieve. Unprecedented single-pass yields of 58 % are observed at a fructose selectivity of 94 %, and continuous operation for over 100 hours is demonstrated. The unexpected performance of the catalyst is realised following a period of activation within the reactor, during which time interaction with the solvent generates a state of activity that is absent in the synthesised catalyst. Mechanistic studies by X-ray absorption spectroscopy, chemisorption FTIR, operando UV/Vis and ¹H–¹³C HSQC NMR spectroscopy indicate that activity arises from isolated Hf^IV atoms with monofunctional acidic properties.

Full text of DOI:10.1002/anie.202006718

A Journal of the Gesellschaft Deutscher Chemiker
Angewandte
International Edition Chemie
www.angewandte.org
Accepted Article
Title: Solvent-activated Hf-containing zeolites enable selective and
continuous glucose-fructose isomerisation
Authors: Luca Botti, Simon A Kondrat, Ricardo Navar, Daniele
Padovan, Juan S Martinez-Espin, Sebastian Meier, and Ceri
Hammond
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Angew. Chem. Int. Ed. 10.1002/anie.202006718
Link to VoR: https://doi.org/10.1002/anie.202006718

Angewandte Chemie International Edition
10.1002/anie.202006718
RESEARCH ARTICLE
Solvent-activated Hf-containing zeolites enable selective and
continuous glucose-fructose isomerisation
[
a][b]
[c]
[b]
[b]
[d]
Luca Botti,
Simon A. Kondrat, Ricardo Navar, Daniele Padovan, Juan S. Martinez-Espin,
[
e]
[a]
Sebastian Meier and Ceri Hammond*
[
[
[
[
[
a]
b]
c]
d]
e]
Department of Chemical Engineering, Imperial College London, London, SW7 2AZ, UK. E-mail: ceri.hammond@imperial.ac.uk
Cardiff Catalysis Institute, Cardiff University, Cardiff, CF10 3AT, UK
Department of Chemistry, Loughborough University, Loughborough, UK
Haldor Topsøe A/S, Haldor Topsøes Allé 1, 2800-Kgs. Lyngby, Denmark
Department of Chemistry, Technical University of Denmark, Kemitorvet Building 207, 2800-Kgs. Lyngby, Denmark
Supporting information for this article is given via a link at the end of the document.
8
Abstract: The isomerisation of glucose to fructose is a critical step
towards manufacturing petroleum-free chemicals from lignocellulosic
biomass. Here we show that Hf-containing zeolites are unique
catalysts for this reaction, enabling true thermodynamic equilibrium
within its lattice. However, although the high activity of Sn-BEA
for GI has received widespread attention, several negative
aspects of its performance are less appreciated, such as its poor
9
stability in the polar solvents required for biomass conversion,
to be achieved in
a
single step during intensified continuous
and its tendency to catalyse a variety of competitive and/or
1
0
operation, which no chemical or biological catalyst has yet been able
to achieve. Unprecedented single pass yields of 58 % are observed
at a fructose selectivity of 94 %, and continuous operation for over
degradation reactions at operational conditions. The latter is
1
b
especially problematic, as it wastes a precious resource and
negatively impacts process economics. However, no better
alternatives to Sn-BEA have yet been found. Consequently,
heterogeneous catalysis is unable to reach the levels of
performance exhibited by enzymes, and renewable chemical
production by scalable chemo-catalytic methods remains limited
by the enduring bottleneck of selective glucose isomerisation.
Here we show that hafnium (Hf)-containing zeolites are
uniquely able to perform GI at unprecedented levels of
selectivity, even at commercially relevant operational conditions.
The optimal catalyst (Hf-BEA, Hf/Si molar ratio = 202, prepared
1
00 hours is demonstrated. The unexpected performance of the
catalyst is realised following a period of activation within the reactor,
during which time interaction with the solvent generates a state of
activity that is absent in the synthesised catalyst. Mechanistic
studies by X-ray Absorption Spectroscopy, chemisorption FTIR,
1
13
operando UV-Vis and H- C HSQC NMR indicate that activity arises
from isolated Hf(IV) atoms with monofunctional acidic properties.
1
1
by hydrothermal synthesis ) achieves single pass fructose
yields of 58 %, at a fructose selectivity of 94 %, even during
continuous operation for over 100 h on stream. Its unique
performance is maintained even at elevated temperature
Introduction
Pressing environmental and societal concerns are driving
researchers to develop new processes with minimised reliance
(
<140 °C), which is an advantage over biological catalysts that
1
on fossil resources. In this context, sustainable chemical
are only active at low temperature, where thermodynamics
favours the reactant side of equilibrium, and reactor productivity
manufacture is a major challenge, since over 95 % of the
organic chemicals currently employed in the chemical industry
7b
is limited. Structure-activity relationships indicate that activity
2
are sourced from coal, oil and natural gas. Amongst renewable
arises from isolated Hf(IV) atoms with monofunctional acidic
properties, which permit high productivity isomerisation to be
achieved without the undesirable side reactions that plague
existing chemo-catalysts.
resources, lignocellulosic biomass is especially promising, given
its abundance, high chemical functionality and geographic
diversity. Yet, its use as a chemical feedstock requires the
development of selective catalytic technologies that are able to
transform this valuable feedstock into useful end products in an
economically and environmentally efficient manner.
Results and discussion
The isomerisation of glucose to fructose (GI) is a key step
towards converting glucose (the most abundant fraction of
lignocellulose) into a variety of commercially relevant chemicals,
Kinetic studies. Our preliminary experiments revealed that
amongst a range of analogous Lewis acidic silicate zeolites -
each of which was prepared by fluoride-media hydrothermal
synthesis (SI Figure S1) - Sn-BEA was the best performing
catalyst for GI during batch operation in methanol (MeOH),
3
including furanics, levulinates and unsaturated hydroxyesters.
Although enzymes catalyse GI to single pass fructose yields of
4
up to 42 %, development of a solid (heterogeneous) catalyst
capable of performing this reaction would be an important step
forward, both by improving operational flexibility (temperature,
which was chosen as solvent as it maximises catalyst stability
9a,b
and facilitates downstream processing (Table 1).
Other Lewis
5
6
pH, feed purity) and facilitating process intensification. To date,
it has become generally appreciated that tin (Sn) containing
zeolites (particularly Sn-BEA) possess the greatest potential for
acidic silicates were either much less active (Ti-BEA, Zr-BEA), or
inactive (Hf-BEA), for GI. At the low levels of substrate
conversion achieved during batch operation (XGlu<20 %), Sn-
BEA was also relatively selective to fructose (SFru<68 %), and
was somewhat more selective than Zr- and Ti-BEA (45 and
7
chemo-catalytic GI. Zeolites are crystalline, microporous
silicates, which in the case of Sn-BEA is a three-dimensional
silicate containing dilute amounts of Lewis acidic Sn(IV) atoms
5
6 %, respectively).
1
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202006718
RESEARCH ARTICLE
converted, from 33.6 % to 66.2 %. Yet, despite operating at
extremely high conversion, the excellent selectivity of Hf-BEA
remained, resulting in an unprecedented single-pass fructose
yield of 57.9 % being achieved. Notably, no loss of carbon
balance was observed in this run, strongly indicating that
competitive and/or consecutive side reactions remained absent
even at high levels of conversion. In fact, the product distribution
achieved at these conditions is representative of the
thermodynamic equilibrium mixture expected at 110 °C,
representing the first time chemo-catalysis has been able to
Table 1. Physical, chemical and reactivity data for various Lewis acidic
silicates explored for glucose isomerisation in batch mode.
Entry
1
Catalyst
Sn-BEA
SSA
V
micro
3
Si/M
198
TOF
S
X
Fru
at
2
-1
-1
(
m g )
(cm g )
for GI
Glu=20%
-
1
(
h )
(%)
398
0.21
95
68
2
3
4
Zr-BEA
Ti-BEA
Hf-BEA
414
390
404
0.21
0.20
0.22
189
204
202
16
15
0
45
56
-
1
4
achieve this feat for GI.
2
Porosity data determined from N isotherms. Si/M molar ratio determined by
ICP-MS. Turnover frequency (TOF) values calculated from batch GI
experiments at t = 2 minutes. Reaction conditions: 1 wt. % glucose in MeOH, 4
g reaction solution, glucose/metal ratio of 50, 110 °C, autogenic pressure.
However, major differences in performance for the Lewis acidic
silicates described in Table 1 were observed when GI was
performed at continuous processing conditions. In continuous
processes, reactants are continuously pumped over
a
suspended solid catalyst bed, and the reaction occurs during the
time the solution resides on the catalyst within the reactor
(
“contact time”). Such reactors are widely employed in the base
chemical industry, due to their increased levels of productivity,
1
2
safety and scalability. Figure 1A-D presents the time on stream
data for two particular Lewis acidic silicates, Sn-BEA and Hf-
BEA, for continuous GI at 110 °C, alongside their conversion vs.
selectivity profiles and product distribution profiles. As can be
seen (Figure 1A), Sn-BEA exhibited an initially high level of
substrate conversion (45%) at the conditions employed (SI
Table S1, Entry 2). However, relatively rapid deactivation
occurred, resulting in more than 40% loss of activity over the
course of 113 h on stream. A single deactivation regime was
determined, characterized by a deactivation rate constant of
Figure 1. Kinetic data for the isomerisation of glucose to fructose over Sn- and
Hf-containing BEA zeolites. (A) Time on stream data for Sn-BEA; (B) Time on
stream data for Hf-BEA; (C) Glucose conversion versus fructose selectivity
profile for Sn- and Hf-BEA; (D) Product distribution achieved with Sn- and Hf-
BEA at various contact times between 4 and 54 seconds. General reaction
conditions: 1 wt. % glucose in methanol, 110 °C. Precise conditions including
WHSV of each experiment are provided in SI Table S1.
-
1
13
0
.00715 X%·h (SI Figure S2). In contrast, Hf-BEA exhibited a
very different kinetic profile (Figure 1B). After an initial induction
period lasting several hours, Hf-BEA reached similar levels of
activity to Sn-BEA. Its performance was maintained over the
following 72 h on stream, with conversion and yield fluctuating
by the typical experimental error of the testing and analytical
protocols (5-8%). After the induction period, a deactivation
We note here that the induction period observed with Hf-
BEA (Figure 1B) could be avoided by activating the catalyst in
MeOH at 110 °C for 20 h prior to introducing glucose into the
feed, which ensured that no valuable substrate was wasted
during operation (SI Figure S4). Furthermore, although Zr-BEA
also exhibited an induction period and reached high levels of
selectivity (SI Figure S5), its absolute activity was 5 times lower
than that of Hf-BEA, and its selectivity was also found to be
approximately 20 % lower, at otherwise identical conditions.
To understand the nature of the overall reaction network,
and hence better understand the origin of the improved
performance of Hf-BEA relative to Sn-BEA, detailed studies of
both product mixtures were undertaken. Firstly, operando UV-
Vis was performed during continuous GI over Sn-BEA and Hf-
-
1
constant of 0.00281 X%.h was determined, demonstrating Hf-
BEA to be more stable than Sn-BEA (SI Figure S3).
In addition, analysis of the product distribution revealed Hf-
BEA to be substantially more selective to fructose than Sn-BEA.
When comparing the fructose selectivity at all levels of
conversion (Figure 1C), Hf-BEA was approximately 80-100 %
more selective than Sn-BEA, converting glucose solely to
fructose and mannose, the two thermodynamic products
expected from GI, at a carbon balance level of 100 % (Figure
15
BEA, according to our recently benchmarked procedure. We
1
D). In contrast, only around half of the glucose converted by
note that the kinetic data obtained with the operando reactor
Sn-BEA became fructose, and a substantial (>20%) loss of
carbon balance was observed (Figure 1D), indicating that
unwanted by-products were formed. Notably, the exceptional
selectivity of Hf-BEA was maintained even at extended contact
times (Figure 1D). As can be seen, increasing the contact time
from 8 seconds to 54 seconds increased the quantity of glucose
matched that obtained in the conventional flow reactor (SI Figure
15
S6), and that both experiments were tailored so that similar
levels of substrate conversion were attained during both
reactions. The resulting spectra of the first 50 h on stream under
the UV probe are shown in Figure 2A-B. Spectra are presented
in difference mode, so that optical features formed during the
2
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202006718
RESEARCH ARTICLE
reaction exhibit positive signals, whereas optical features
decreasing during the reaction exhibit negative signals. Both
experiments exhibited a strong, positive absorption at 300-360
nm, and a minor negative signal at short wavelengths (< 250
nm). Control experiments (SI Figure S7) and our recent
none of which were observed for Hf-BEA. The by-products
observed included methyl fructosides and the branched
aldohexose hamamelose, in addition to various glycolytic end
products including trans-2,5,6-trihydroxy-3-hexenoic acid methyl
1
7
ester (THM) and various 3-deoxy-γ-lactones.
1
5
operando UV-Vis studies
allowed these features to be
assigned to metal-glucose (300-360) and metal-methanol (<250
nm) interactions. As these absorptions were common to both
samples irrespective of performance, they are not discussed
further for brevity. However,
a clear difference at longer
wavelengths was observed. In the case of Sn-BEA (Figure 2A),
a positive absorption was immediately formed between 400-480
nm, the intensity of which slowly decreased over the course of
the experiment. An excellent inverse correlation between the
intensity of this absorption and the selectivity of the catalyst
could be made (SI Figure S8), indicating that this absorption
relates to by-product formation. Notably, this signal was absent
from the operando spectrum of Hf-BEA, suggesting the absence
of various chromophoric by-products when using this catalyst.
1
13
Figure 3. H- C HSQC NMR spectra for NMR analyses of the reaction
effluent from Sn- and Hf-BEA catalysed GI at 110 °C. Spectra were acquired
at 25 ºC on an 800 MHz spectrometer equipped with a cryogenically cooled
probe. The spectra provide an unbiased representation of the vastly reduced
by-product formation when using Hf-BEA instead of Sn-BEA.
Figure 2. Operando UV-Vis measurements for (A) Sn-BEA and (B) Hf-BEA
catalysed GI at 110 °C, permitting the absorption spectra of the system to be
measured continuously alongside the kinetics of the system. General reaction
conditions: 1 wt. % glucose in methanol, 110 °C.
1
13
The diversity of products formed, particularly in the case of Sn-
BEA, indicates that multiple reaction pathways are feasible
during GI at these conditions. With the aim of developing
structure-function relationships to account for these pathways,
and to compare the structure and active site speciation of Hf-
BEA to those previously established for Sn-BEA, spectroscopic
studies of Hf-BEA were performed. Unfortunately, the low
gyromagnetic ratio of Hf, its large quadrupole moment and its
high-energy absorption feature (λmax<200 nm) prohibited its
H- C HSQC NMR spectroscopy was thus used to obtain an
unbiased representation of by-product formation when using
1
13
either catalyst. The H- C HSQC method was employed as it
provided roughly 30-times higher sensitivity than analogous one-
1
3
dimensional C NMR measurements, whilst providing dramatic
improvements in resolution due to its two-dimensional nature.
Moreover, various by-products in the Lewis acid catalysed
1
conversion of glucose have previously been described using H-
1
3
C HSQC NMR assays, thus facilitating the characterization of
1
8
1
6
1
13
characterisation
Accordingly, direct and indirect studies were achieved by means
of Hf L edge X-ray Absorption Spectroscopy (XAS) and
chemisorption coupled with FTIR spectroscopy (cIR).
Hf L edge XAS studies were used to gain insight into the
by
classical
site-selective
methods.
post-reaction solutions. Figure 3A-D shows the H- C HSQC
NMR spectra for Sn- and Hf-BEA catalysed GI at 110 °C across
various spectral regions. The spectra of both reactions were
sampled at different points of time on stream, so that samples of
exactly the same level of conversion (20 %) were measured
3
3
Hf active site within the catalyst, before and after reaction, and
after MeOH activation (SI Figure S4). Samples were not further
treated between operation and measurement, so as to maintain
the relevance of the XAS spectra to kinetic performance. Figure
4A-B shows the X-ray Absorption Near Edge Structure (XANES)
spectra of Hf-BEA, prior to reaction, after methanol activation,
and after 100 h on stream, alongside the spectrum of a
(
10% and 19 % yield of fructose for Sn-BEA and Hf-BEA,
respectively). As such, spectra were normalised to the quantity
of remaining glucose, as verified by NMR standardisation and
HPLC-ELSD. Hence, the volumes of the signals associated to
1
6
unconverted glucose are identical in both spectra (Figure 3A).
Signals attributed to fructose were much more intense for Hf-
BEA than Sn-BEA, verifying the improved yield to fructose and
hence the improved selectivity of Hf-BEA. A similar quantity of
mannose was also observed during both experiments. However,
a number of additional species were observed over Sn-BEA,
2
monoclinic HfO standard. The adsorption edges of all catalyst
samples, as determined by the maxima of the first derivative of
the data, were found to be at 9564.4 eV, compared to 9563.4 eV
3
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202006718
RESEARCH ARTICLE
in monoclinic HfO
2
. The 1 eV shift in adsorption edge of the
was coupled with a significant
2
Table 2. EXAFS fitting data of Hf-BEA catalysts and monoclinic HfO .
catalysts compared to HfO
2
b
2
2
increase in the white line feature labelled 1 (2p → 5d transition)
in the catalyst samples, which was greatest in the fresh sample,
followed by the methanol activated catalyst and the post reaction
material. It is proposed that the formal oxidation state of Hf (IV)
is present in all samples, and that the observed differences
indicate a change in the local environment of Hf within the
Sample
Path
CN
R (Å)
2σ (Å )
E
r
R-
(
eV)
factor
a
b
Mono-
Hf-O (1)
7
2.12(1)
3.42(1)
3.70(3)
4.00(1)
2.00(2)
2.20(4)
2.00(2)
2.20(2)
2.00(2)
2.21(5)
0.011(1)
0.008(1)
0.013(5)
0.005(2)
0.003(2)
7(1)
0.007
HfO
2
a
b
Hf-Hf (1)
7
a
b
Hf-O (2)
Hf-Hf (2)
Hf-O (1)
Hf-O (2)
Hf-O (1)
Hf-O (2)
Hf-O (1)
Hf-O (2)
7
catalyst compared to the monoclinic HfO
2
standard. Feature 2
a
b
was seen in the XANES of HfO , but not in Hf-BEA at any stage
2
4
of reaction, and could be attributed to a Hf-O-Hf multiple
scattering process involving atoms in the second shell. The
absence of this feature in the Hf-BEA catalyst further suggests a
different coordination of oxygen to Hf in Hf-BEA compared to
Hf-BEA
fresh)
4.5(8)
2.1(5)
4.5(8)
2.4(5)
4.8(8)
1.9(5)
7(1)
8(1)
7(1)
0.007
0.005
0.007
(
c
0.004
bulk HfO
material.
2
. Feature 3 was the EXAFS oscillation feature of each
Hf-BEA
used)
0.004(2)
(
c
0.004
Hf-BEA
MeOH
activated)
0.004(2)
(
c
0.00
Fitting parameters: 2.2 < k < 12, 1.1 < R < 2.5 (4 for mono-HfO
2
). Amplitude
a
reduction factor = 0.8 (fitted from Mono-HfO
single path (i.e Hf-O(1) comprises of 7 different path lengths); Coordination
2
); Multiple path lengths fitted as a
b
1
9
c
2
numbers (CN) fixed to known crystallographic values; 2σ defined as 1.2 of
2
2
σ of Hf-O(1)
Fitting of the EXAFS data (Table 2) revealed two clear first shell
Hf-O paths in the catalyst samples, compared to a single Hf-O
path in monoclinic HfO
2
that was fitted to represent the 7 very
2
1
similar Hf-O paths seen in monoclinic HfO
2
.
The first Hf-O path
in all the catalyst samples was found at 2.00 Å, and possessed a
coordination number of 4. These values are in excellent
agreement to those we determined for Sn-BEA during previous
2
0a
XAS studies, and thus this path can be attributed to Hf atoms
present in the BEA framework. A second Hf-O distance, at 2.20
Å with a coordination number of 2, could also be determined in
the catalyst samples. Since this second distance is notably
longer than the average first shell distance of monoclinic or
Figure 4. X-ray Absorption Spectroscopy data for Hf-BEA catalysts. (A)
Hafnium L -edge XANES of Hf-BEA catalyst and monoclinic HfO and (B)
Expansion of the XANES spectra. (Red) fresh catalyst; (grey) methanol
activated catalyst; (blue) used catalyst; (green) HfO . (1) White line feature; (2)
multiple scattering Hf-O-Hf; (3) EXAFS oscillation. (C) k weighted χ data of
Hf-BEA catalysts and monoclinic HfO . (D) Magnitude of Fourier Transform of
the L -edge EXAFS spectra. Samples: (i) monoclinic HfO ; (ii) fresh catalyst;
iii) used catalyst; (iv) methanol activated catalyst.
orthorhombic HfO
interpretation was supported by the poor fit of monoclinic or
orthorhombic HfO models to the second shell paths of all the
2 2
, it is unlikely to arise from bulk HfO . This
3
2
2
2
3
catalyst samples, alongside the absence of splitting in the
oscillation around 9 Å in the χ data of all catalysts (Figure 4C).
As such, the second Hf-O distance could either derive from
water bound to the Hf atoms within the framework, or from a
2
3
2
(
small fraction of the Hf being present as HfO
x
clusters. Whilst the
absence of HfO cluster standards makes it impossible to
x
To complement these findings, Hf L
Absorption Fine Structure (EXAFS) studies were performed. The
3
-edge Extended X-ray
conclusively distinguish between these options, our previous
XAS studies of Sn-BEA materials prior to and following
x
k
weighted data of the samples are shown in Figure 4C,
2
0a
dehydration,
hydrated nature of Lewis acidic heteroatoms within the BEA
alongside previous studies documenting the
alongside the corresponding magnitude of the Fourier transform
2
0
(
FT) in Figure 4D. Based on previous XAS studies of Sn-BEA,
2
2
lattice at ambient conditions, lead us to favour the initial
hypothesis i.e. that the second Hf-O distance relates to the
coordination of two water molecules to the framework Hf species.
Thus, we conclude that at least the major fraction of Hf is within
the BEA framework, and is present as an isomorphously
substituted lattice atom in a hydrated state at ambient conditions.
Although this is the first time such XAS studies have been
performed for Hf-BEA catalysts, our findings are supported by
the first feature in the FT data was assigned to Hf-O scattering
paths, while features between 2.5 and 4 Å could be associated
to Hf-Hf scattering and/or second shell Hf-O paths. Clearly, the
first Hf-O feature in all the catalyst samples (Figure 4D) was at a
2
shorter distance compared to HfO . Moreover, second shell Hf-O
and/or Hf-Hf paths were much less significant in the catalysts
samples.
2
3
previous Pair Distribution Function measurements of Hf-BEA.
4
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202006718
RESEARCH ARTICLE
The lack of changes in the amplitude of features between
compared to those in Hf-BEA. Secondly, whilst vibrations
associated to the protonation of pyridine were clearly generated
2
.5 and 4 Å in the catalyst samples post operation, coupled with
-
1
the constant coordination numbers for first shell Hf-O species
even after 100 h on stream, strongly suggest that the Hf species
present in the catalyst did not change in quantity or structure
during reaction, consistent with the excellent stability of the
catalyst. Moreover, no changes to the properties of Hf-BEA were
evident after methanol activation. This indicates that the
processes occurring during the induction period are too subtle to
be probed by XAS, and evidently do not relate to major
modifications of the catalyst active sites.
by Sn-BEA (particularly at 1637 and 1540 cm ), such vibrations
were totally absent from the spectra of Hf-BEA. Thus, it can be
concluded that Sn-Beta possesses relatively strong Lewis acid
sites alongside Brønsted acidic active sites, whereas Hf-BEA
only contains Lewis acid active sites, which are of a lower
strength to those found in Sn-BEA. We note that the observation
of Brønsted acidity in Sn-BEA is supported by recent NMR
2
3a
studies,
relative to Sn-BEA is supported by N MAS NMR studies of
whilst the weaker Lewis acid strength of Hf-BEA
1
5
2
6
Although it can be concluded that both Sn- and Hf-BEA
possess isomorphously substituted Lewis acidic centres, recent
studies have indicated that the substitution of different Lewis
acids into zeolite lattices can lead to the genesis of materials
various Lewis and Brønsted acidic silicates.
We hypothesise that these differences in acidic nature are
responsible for the excellent performance of Hf-BEA; Whilst the
stronger Lewis and Brønsted acidic sites in Sn-BEA result in a
catalyst that is initially more active than Hf-BEA in terms of
glucose conversion (see TOF in Table 1 and conversion in
Figure 1), increased conversion is achieved due to initiation of
several competing processes over multiple active sites that co-
exist in the catalyst. In contrast, the monofunctional Lewis acidity
in Hf-BEA results in isomerisation being achieved at
unprecedentedly high selectivity (>90%).
2
4
with very different acidic properties. Thus, to gain insight into
the acidic properties of Sn- and Hf-BEA, chemisorption FTIR
spectroscopy (cIR) was performed with pyridine as a probe
molecule sensitive to both Lewis and Brønsted acidic centres.
Figure 5 shows the IR spectra of pyridine adsorbed on (A) Hf-
-
1
and (B) Sn-BEA in the 1650-1400 cm region, following dosing
with pyridine vapour at room temperature, and subsequent
outgassing up to 200 °C to remove gaseous and weakly bound
2
5a
pyridine, according to the method of Harris et al. Spectra were
background corrected by subtraction of the spectrum of the
zeolite catalyst prior to dosing.
Figure 6. Kinetic data for the isomerisation of glucose to fructose over Sn- and
Hf-containing BEA zeolites at 140 °C. (A) Distribution of products attained over
Sn-BEA and Hf-BEA at maximal conversion at 140 °C. (B) Time on stream
analysis of Hf-BEA at 140 °C over a 28 h period. General reaction conditions:
1
wt. % glucose in methanol, 140 °C. Precise conditions including WHSV of
each experiment are provided in SI Table S1.
Figure 5. Chemisorption FTIR studies following the interaction of pyridine over
(
A) Hf-BEA, and (B) Sn-BEA. Spectra are illustrated after outgassing at room
temperature (RT), and following thermal treatment at 100-200 °C.
Given that higher reaction temperatures favour thermodynamic
7
b
equilibrium yields and boost reactor productivity (space-time-
yield), the activity of Hf-BEA for intensified GI at elevated
temperature (140 °C) was explored. As can be seen, the
excellent performance of Hf-BEA was maintained even at
140 °C, with fructose still produced at yields of up to 50% at a
selectivity 94%, but with a three-fold increase in space-time-yield
2
5
Based on previous studies of pyridine adsorption and the
behaviour of the vibrations upon thermal treatment, three main
classes of vibrations could be assigned; i) vibrations related to
physisorbed and/or hydrogen bonded pyridine (1597, 1580,
-
1
1
482, 1440 cm ); ii) vibrations arising from coordination of
-
1
-3 -1
pyridine to Lewis acid centres (1620-1600, 1455-1445 cm ),
and; iii) vibrations related to pyridine protonated by Brønsted
over that achieved at 110 °C (4.12 g (Fru) cm h versus 1.3 g
-
3
-1
(Fru) cm h ) (Figure 6A). In contrast, poor fructose yield,
selectivity and carbon balance were observed with Sn-BEA at
-
1
acid centres (1637, 1540 cm ). Since physisorbed and/or
hydrogen bonded pyridine vibrations do not relate to the active
-
3
-1
140 °C (2.1 g (Fru) cm h ). In addition, excellent stability for Hf-
BEA was observed over a 28 h period of operation (Figure 6B).
Notably, the induction period observed at 110 °C was also
eliminated at higher temperature, resulting in a catalytic process
that was highly productive and selective from the very first
moments of operation without requirement of a preceding
methanol activation step (SI Figure S4). Furthermore, high levels
of productivity were maintained even in the presence of
additional water (SI Figure S9). Notably, separation of the
product mixture by reduced pressure distillation also revealed
that the isolated product formed at 140 °C was a white,
-
1
sites of the catalysts, and vibrations at 1490 cm can be
generated both by Lewis and Brønsed acid sites, these bands
were not followed further.
By interrogation of the cIR profiles (Figure 5A-B), two major
observations could be made. Firstly, although both samples
exhibited vibrations related to interaction of pyridine with Lewis
acidic centres, the stretching frequencies of these vibrations
were present at higher values for Sn-BEA than Hf-BEA (1616 vs.
-
1
-1
1
607 cm , and 1455 vs. 1446 cm , respectively). This indicates
greater activation of pyridine by the Lewis sites of Sn-BEA
5
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202006718
RESEARCH ARTICLE
Natl. Acad. Sci. 2010, 107, 6164–6168. c) Y. Román-Leshkov, M.
Moliner, J. A. Labinger, M. E. Davis, Angew. Chemie - Int. Ed. 2010, 49,
crystalline solid, confirming the absence of Maillard browning
processes even during high temperature operation with Hf-BEA
8
954–8957. d) M. Moliner, Dalt. Trans. 2014, 43, 4197–4208.
(
SI Figure S10).
[8] M. E. Davis, Chem. Mater. 2014, 26, 239–245.
[
9] a) D. Padovan, S. Tolborg, L. Botti, E. Taarning, I. Sádaba, C.
Hammond, React. Chem. Eng. 2018, 3, 155−163. b) D. Padovan, C.
Parsons, M. S. Grasina, C. Hammond, Green Chem. 2016, 18, 5041-
5
049. c) L. Botti, R. Navar, S. Tolborg, J. S. Martinez-Espin, D. Padovan,
Conclusion
E. Taarning, C. Hammond, Top. Catal. 2019, 62, 1178–1191. d) G. M.
Lari, P. Y. Dapsens, D. Scholz, S. Mithell, C. Mondelli, J. Pérez-Ramirez,
Green Chem. 2016, 16, 1249-1260. e) M. J. Cordon, J. N. Hall, J. W.
harris, J. S. Bates, S.-J. Hwang, R. Gounder, Catal. Sci. Tech. 2019, 7,
Hf-containing zeolites are shown to be unique catalysts for
glucose-fructose isomersation, enabling unprecedented single
pass yields of 58% to be achieved at a fructose selectivity of
1
654-1668. f) W. N. P. van der Graaff, C. H. L. Tempelman, F. C.
Hendriks, J. Ruiz-Martinez, S. Bals, B. M. Weckhuysen, E. A. Pidko, E.
J. M. Hensen, Appl. Catal. A: Gen, 2018, 564, 113-122.
9
4%, for over 100 hours on stream. Spectroscopic studies by
[
10] a) I. Tosi, A. Riisager, E. Taarning, P. R. Jensen, S. Meier, Catal. Sci.
Technol. 2018, 8, 2137–2145. b) M. S. Holm, S. Saravanamurugan, E.
Taarning, Science. 2010, 328, 602–605. c) S. Tolborg, I. Sádaba, C. M.
Osmundsen, P. Fristrup, M. S. Holm, E. Taarning, ChemSusChem 2015,
1
13
XAS, chemisorption FTIR, operando UV-Vis and H- C HSQC
NMR suggest the activity of Hf-BEA arises from the presence of
isolated Hf(IV) atoms with monofunctional Lewis acidic
properties. Although catalytic activity is only initially observed
after a period of activation in the reactor, this can be bypassed
by pre-treating Hf-BEA in the reaction solvent, or by operating
the reaction at higher temperature. Optimisation of the system
permits selective glucose-fructose isomerisation to be achieved
even during intensified continuous operation at 140 °C, which no
chemical or biological catalyst has yet been able to achieve. The
reasons behind the induction period observed for Hf-BEA at
lower temperature, and its higher stability against deactivation
compared to other Lewis acidic zeolites, remain the focus of our
on-going investigations.
8
, 613–617.
[11] S. Tolborg, A. Katerinopoulou, D. D. Falcone, I. Sadaba, C. M.
Osmundsen, R. J. Davis, E. Taarning, P. Fristrup, M. S. Holm, J. Mater.
Chem. A, 2014, 2, 20252–20262.
[12] a) D. Padovan, C. Parsons, M. Simplicio Grasina, C. Hammond, Green
Chem. 2016, 18, 5041–5049 b) F. Héroguel, B. Rozmysłowicz, J. S.
Luterbacher, Chimia. 2015, 69, 582–591.
13] O. Levenspiel, Chemical Reaction Engineering, (Eds.: John Wiley &
Sons: New York), 1999; ch. 21, p 473.
14] P. A. Son, S. Nishimura, K. Ebitani, React. Kinet. Mech. Catal. 2014,
[
[
1
11, 183–197.
[15] D. Padovan, L. Botti, C. Hammond, ACS Catal. 2018, 8, 7131–7140.
16] S. G. Elliot, S. Tolborg, I. Sádaba, E. Taarning, S. Meier,
ChemSusChem 2017, 10, 2990–2996.
17] S. Tolborg, S. Meier, I. Sádaba, S. G. Elliot, S. K. Kristensen, S.
Saravanamurugan, A. Riisager, P. Fristrup, T. Skrydstrup, E. Taarning,
Green Chem. 2016, 18, 3360–3369.
[
[
[
[
18] R. K. Harris, E. D. Becker, J. Mag. Reson. 2002, 156, 323-326.
19] K. R. Whittle, G. R. Lumpkin, S. E. Ashbrook, J. Solid State Chem. 2006,
1
79, 512-521.
[
20] a) C. Hammond, D. Padovan, A. Al-Nayili, P. P. Wells, E. K. Gibson, N.
Dimitratos, ChemCatChem 2015, 7, 3322–3331. b) S. R. Bare, S. D.
Kelly, W. Sinkler, J. J. Low, F. S. Modica, S. Valencia, A. Corma, L. T.
Nemeth, J. Am. Chem. Soc. 2005, 127, 12924-12932.
Acknowledgments
CH gratefully appreciates the support of The Royal Society, for
provision of a University Research Fellowship (UF140207,
URF\R\201003) and enhanced research grant funding
[
[
21] S. E. Erenburg, S. V. Trubina, K. O. Kvashnina, V. N. Kruchinin, V. V.
Gritsenko, A. G. Chernikova, A. M. Markeev, J. Exp. Theor. Phys. 2018,
1
26, 815-824.
22] a) P. Wolf, A. J. Rossini, A. Comas-Vives, F. Núñez-Zarur, B. Malaman,
(
RGF/EA/180314). CH and LB are also grateful to Haldor
A. Lesage, L. Emsley, C. Copéret, I. Hermans, Angew. Chem. Int. Ed.
2
014, 53, 10179-10183. b) P. Wolf, M. Valla, F. Núñez-Zarur, A. Comas-
Topsøe A/S for PhD studentship funding. CH and RN appreciate
the support of Consejo Nacional de Ciencia y Tecnologia
Vives, A. J. Rossini, C. Firth, H. Kallas, A. Lesage, L. Emsley, C.
Copéret, I. Hermans, ACS Catal. 2016, 6, 4047-4063. c) V. L.
Sushkevich, P. A. Kots, Y. G. Kolyagin, A. V. Yakimov, A. V. Marikutsa,
I. I. Ivanova, J. Phys. Chem. C. 2019, 123, 5540-5548. d) T. D.
Courtney, C.-C. Chang, R. J. Gorte, R. F. Lobo, W. Fan, V. Nikolakis,
Microp. Mesop. Mat. 2015, 210, 69-76.
(
CONACYT, Fellowship 472256) for PhD studentship funding.
00 MHz NMR spectra were recorded on the spectrometer of
8
the DTU NMR centre supported by the Villum Foundation.
[
23] T. Iida, K. Ohara, Y. Romàn-Leshkov, T. Wakihara, Phys. Chem. Chem.
Phys. 2018, 20, 7194-7919.
Keywords: renewables • biomass • catalysis • zeolites •
[24] a) V. L. Sushkevich, P. A. Kots, Y. G. Kolyagin, A. V Yakimov, A. V
Marikutsa, I. I. Ivanova, J. Phys. Chem. C. 2019, 123, 5540-5548. b) A.
Zheng, S. Bin Liu, F. Deng, Chem. Rev. 2017, 117, 12475–12531.
continuous • glucose • hafnium
[
25] a) J. W. Harris, M. J. Cordon, J. R. Di Iorio, J. C. Vega-Vila, F. H.
Ribeiro, R. Gounder, J. Catal. 2016, 335, 141–154. b) F. Bonino, A.
Damin, S. Bordiga, C. Lamberti, A. Zecchina, Langmuir 2003, 19, 2155–
[
[
1] a) F. Cavani, J. H. Teles, ChemSusChem 2009, 2, 508–534. b) P. N. R.
Vennestrøm, C. M. Osmundsen, C. H. Christensen, E. Taarning, Angew.
Chemie - Int. Ed. 2011, 50, 10502–10509. c) C. O. Tuck, E. Pérez, I.T.
Horváth, R.A. Sheldon, M. Poliakoff, Science 2012, 337, 695–700.
2
161.
[
26] W. R. Gunther, V. K. Michaelis, R. G. Griffin, Y. Romàn-Leshkov. J.
Phys. Chem. C. Nanomater. Interf. 2016, 120, 28533-28544.
2] a) G. W. Huber, S. Iborra, A. Corma, Chem. Rev. 2006, 106, 4044–
4
2
098. b) A. Corma, S. Iborra, A. Velty, Chem. Rev. 2007, 107, 2411–
502. c) C. H. Christensen, J. Rass-Hansen, C. C. Marsden, E.
Taarning, K. Egeblad, ChemSusChem 2008, 1, 283–289.
3] a) X. Zhang, K. Wilson, A. F. Lee, Chem. Rev. 2016, 116, 12328–12368.
b) H. Chang, A. H. Motagamwala, G. W. Huber, J. A. Dumesic, Green
Chem. 2019, 21, 5532-5540.
[
[
[
4] a) H. Li, S. Yang, S. Saravanamurugan, A. Riisager, ACS Catal. 2017, 7,
3
3
010–3029. b) Z. Zhang, A. A. Donaldson, X. Ma, Biotechnol. Adv. 2012,
0, 913–919.
5] a) M. Dusselier, P. Van Wouwe, A. Dewaele, E. Makshina, B. F. Sels,
Energy Environ. Sci. 2013, 6, 1415–1442. b) T. Ennaert, J. Van Aelst, J.
Dijkmans, R. De Clercq, W. Schutyser, M. Dusselier, D. Verboekend, B.
F. Sels, Chem. Soc. Rev. 2016, 45, 584–611.
[
[
6] a) C. Hammond, Green Chem. 2017, 19, 2711–2728. b) I. Sádaba, M. L.
Granados, A. Riisager, E. Taarning, Green Chem. 2015, 17, 4133–4145.
c) J. P. Lange, Angew. Chem. Int. Ed. 2015, 54, 13187-13197. d) L. S.
Scott, ACS Catal. 2018, 8, 8597-8599.
7] a) J. Dijkmans, D. Gabriëls, M. Dusselier, F. De Clippel, P. Vanelderen,
K. Houthoofd, A. Malfliet, Y. Pontikes, B. F. Sels, Green Chem. 2013,
1
5, 2777–2785. b) M. Moliner, Y. Román-Leshkov, M. E. Davis, Proc.
6
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202006718
RESEARCH ARTICLE
Entry for the Table of Contents
Hf-zeolites are shown to catalyse glucose-fructose isomerisation to thermodynamic equilibrium during continuous operation, which no
chemical catalyst has yet been able to achieve. Fructose yields of 58 % are observed at a selectivity of 94 %, and continuous
operation for over 100 hours is demonstrated. Structure-activity relationships are developed with X-ray Absorption Spectroscopy,
chemisorption FTIR, operando UV-Vis and HSQC NMR.
7
This article is protected by copyright. All rights reserved.