Catalytic consequences of borate complexation and pH on the epimerization of l-arabinose to l-ribose in water catalyzed by Sn-Beta zeolite with borate salts

Article abstract of DOI:10.1016/j.molcata.2013.08.021

Sn-Beta zeolite with sodium tetraborate cooperatively catalyzes the epimerization of aldoses via an intramolecular 1,2 carbon-shift mechanism. l-Arabinose is one of the seven common sugars and its epimer, l-ribose, is a valuable rare sugar with applications in antiviral and anticancer agents. Here, a full factorial experimental design is performed to demonstrate the catalytic consequences of varying key reaction parameters such as pH, borate to sugar ratio, and reaction time. Reactivity data revealed that isomerization is favored under acidic pH conditions (pH < 7.0); that epimerization dominates under neutral conditions (pH ranging from 7.0 to 7.8); and that drastic inhibition of both epimerization and isomerization rates occur under basic conditions (pH > 7.8). Using a 5 wt% arabinose feed and 100:1 sugar-metal ratios at 343 K, conversions ranging from 20% to 30% were obtained with selectivities of 75%, 84%, and 91% for boron-sugar ratios of 0.2:1, 0.5:1, and 1:1, respectively. The predominance of epimerization over isomerization products with substoichiometric borate suggests that one borate can influence the reactivity of several sugar molecules and may influence the Sn active site directly. Reaction data obtained under differential conditions revealed that the epimerization reaction follows first order kinetics over a wide temperature range with an apparent activation energy of 98 kJ/mol and pre-exponential factor of 1.9 × 10¹⁴ L mol Sn^-1 s^-1.

Full text of DOI:10.1016/j.molcata.2013.08.021

Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
Contents lists available at ScienceDirect
Journal of Molecular Catalysis A: Chemical
journal homepage: www.elsevier.com/locate/molcata
Catalytic consequences of borate complexation and pH on the
epimerization of l-arabinose to l-ribose in water catalyzed by Sn-Beta
zeolite with borate salts
William R. Gunther, Quynh Duong, Yuriy Román-Leshkov∗
Department of Chemical Engineering, Room 66-456, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 7 July 2013
Received in revised form 19 August 2013
Accepted 20 August 2013
Available online 6 September 2013
Sn-Beta zeolite with sodium tetraborate cooperatively catalyzes the epimerization of aldoses via an
intramolecular 1,2 carbon-shift mechanism. l-Arabinose is one of the seven common sugars and its
epimer, l-ribose, is a valuable rare sugar with applications in antiviral and anticancer agents. Here, a
full factorial experimental design is performed to demonstrate the catalytic consequences of varying key
reaction parameters such as pH, borate to sugar ratio, and reaction time. Reactivity data revealed that iso-
merization is favored under acidic pH conditions (pH < 7.0); that epimerization dominates under neutral
conditions (pH ranging from 7.0 to 7.8); and that drastic inhibition of both epimerization and isomeriza-
tion rates occur under basic conditions (pH > 7.8). Using a 5 wt% arabinose feed and 100:1 sugar–metal
ratios at 343 K, conversions ranging from 20% to 30% were obtained with selectivities of 75%, 84%, and
91% for boron–sugar ratios of 0.2:1, 0.5:1, and 1:1, respectively. The predominance of epimerization over
isomerization products with substoichiometric borate suggests that one borate can influence the reac-
tivity of several sugar molecules and may influence the Sn active site directly. Reaction data obtained
under differential conditions revealed that the epimerization reaction follows first order kinetics over a
wide temperature range with an apparent activation energy of 98 kJ/mol and pre-exponential factor of
Keywords:
Sn-Beta zeolite
Borate
Epimerization
Water tolerant solid Lewis acids
Catalytic asymmetric chemistry
Carbohydrate chemistry
1.9 × 1014 L mol Sn−1 s−1
.
© 2013 Elsevier B.V. All rights reserved.
1. Introduction
effective catalytic pathways to produce rare sugars is of great rele-
vance.
Carbohydrates are involved in important biological processes
and also serve as valuable building blocks in many commer-
cial applications. For instance, the common pentose d-ribose is a
major component in ribonucleic acid (RNA), adenosine triphos-
phate (ATP), and nicotinamide adenine dinucleotide (NADH) in
metabolism. The industrial production of riboflavin also involves
the use of d-ribose [1]. The rare pentose, l-ribose, has found appli-
cations in anti-viral and anti-cancer drugs because its reverse
stereochemistry alters the interaction with enzymes, inhibiting the
nucleoside synthesis-replication process [2]. Additionally, l-RNA
binds tightly to natural RNA and is resistant to degradation by
nucleases, allowing it to silence gene expression. Consequently, l-
ribose is part of several pharmaceuticals presently on the market
or undergoing clinical trials [3]. To meet the demand for such rare
sugars, costly and complex biochemical [4] and synthetic organic
processes are currently used to produce them from common car-
bohydrates [5]. Therefore, the development of simpler and more
The selective epimerization of aldoses is an attractive pathway
for the production of rare sugars. l-Arabinose is a common sugar
found in the hemicellulose fraction of biomass. As depicted in Fig. 1,
arabinose is interconverted into an equilibrium mixture consisting
of ca. 66% arabinose, 11% ribulose and 23% ribose in the pres-
ence of isomerase enzymes or inorganic base catalysts at 333 K [6].
Ribulose is the ketose isomer, while ribose is the aldose epimer, typ-
ically formed by way of a double isomerization process. The direct
epimerization of arabinose into ribose may be accomplished using
epimerase enzymes to generate a distribution of 74% arabinose
and 26% ribose at 333 K. However, most epimerases only func-
tion on substrates modified with phosphate or nucleotide groups.
Direct epimerization yields a product distribution that does not
contain ribulose, and is therefore more amenable for purification.
Epimerization of arabinose to ribose is slightly endothermic, and
consequently, high temperatures lead to higher equilibrium ribose
yields (see Fig. S1). In this respect, inorganic catalysts may be attrac-
tive epimerization candidates given that they are not limited by
temperature or substrate functionalization in the same manner as
biological catalysts [6].
∗
Hydrophobic zeolites with metal atoms substituted in the crys-
talline microporous framework have recently emerged as a new
Corresponding author. Tel.: +1 617 253 7090.
E-mail address: yroman@mit.edu (Y. Román-Leshkov).
1381-1169/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.molcata.2013.08.021

W.R. Gunther et al. / Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
295
on the epimerization of l-arabinose to l-ribose. The conversion of
arabinose to ribose provided a high value added case to investi-
gate the role of sugar–borate complexation. Understanding which
species and reactions are dominant under controlled reaction con-
ditions provides insight into the role of additives in directing the
sugar conversion pathway. Thus, we show a comparison of isomer-
ization activity with the fraction of free sugar under low borate
loadings and use 11B NMR and Raman spectroscopy characteriza-
tion to identify sugar borate dimers and polyborates in solution. A
temperature study is performed to extract kinetic parameters and
optical rotation measurements are used to assess stereospecificity.
2. Experimental
Fig. 1. Schematic representation of the isomerization and epimerization of l-
arabinose. The D form is the mirror image of the L form.
2.1. Catalyst synthesis
Sn-Beta was synthesized as follows: 26.735 g of aqueous
tetraethylammonium hydroxide (Sigma–Aldrich, 35% (w/w)) and
24.069 g of tetraethylorthosilicate (Sigma–Aldrich, 99% (w/w))
were added to a Teflon® (Polytetrafluoroethylene, [PTFE]) dish,
which was magnetically stirred at room temperature for 90 min and
then cooled in an ice bath. Then, 0.261 g of tin (II) chloride dihydrate
(Sigma–Aldrich, 98% (w/w)) dissolved in 15 g of cold deionized
water (DI H2O) was added dropwise. Sn(II) which oxidizes to Sn(IV)
in water was used in place of SnCl4·5H2O and has resulted in Sn-
Beta consistently free of extraframework SnO2. The solution was
left uncovered on a stir plate for 10 h to reach a total mass of
35 g. Next, 2.600 g of aqueous hydrofluoric acid (Sigma–Aldrich,
48% (w/w)) was added dropwise and the mixture was homoge-
nized using a PTFE spatula, resulting in a thick gel. Then, 0.358 g
of previously-made Sn-Beta was seeded into the mixture, which
was allowed to evaporate to 33.776 g, which corresponds to a final
molar composition of SiO2/0.01 SnCl2/0.55 TEAOH/0.54 HF/7.52
H2O. The thick paste was transferred to a 45 ml PTFE-lined stainless
steel autoclave and heated to 413 K for 40 days under static condi-
tions. The solids were recovered by filtration, washed with DI H2O,
dried at 373 K and calcined at 853 K for 10 h with a 1 K min−1 ramp
and 1 h stops at 423 and 623 K, leading to an overall inorganic oxide
yield of 80–90%.
Powder X-ray diffraction (PXRD) patterns were collected using
a Bruker D8 diffractometer with Cu K␣ radiation, which confirmed
that the solid material has the Beta topology (see Supplementary
Fig. S2). Ultraviolet–visible (UV–vis) measurements were recorded
using a Varian Cary 5000 UV–vis-NIR spectrometer with a dif-
fuse reflectance cell after calcination without subsequent drying.
The ultraviolet–visible diffuse reflectance spectrum of the calcined
sample shows the presence of a unique band at ∼200 nm, which has
been associated with Sn tetrahedrally coordinated into the zeolite
framework (see Supplementary Fig. S3); as additional confirmation
of framework incorporation, 119Sn NMR of an analogous sample is
included in Fig. S4. Sn content was measured using a Horiba Jobin
Yvon ACTIVA-S ICP-AES. Catalyst samples were dissolved in a few
drops of 48% HF and diluted in 2% nitric acid after evaporation of
the HF.
class of Lewis acid catalysts capable of activating carbonyl func-
tional groups in the presence of water. The highly dispersed
tetrahedrally-coordinated metal centers can accept electron pairs
through empty d orbitals and expand their coordination shells from
4 to 5 or 6 sites due to their large covalent radii. Tin-containing
zeolites are active in the Meerwein–Ponndorf–Verley reduction
[7,8], Baeyer–Villiger oxidation [9–11], and carbonyl-ene reaction
[12]. In the context of carbohydrate chemistry, the incorporation
of Sn, Ti, and Zr in the framework of Beta zeolite has found appli-
cations in the conversion of hexoses [13,14], pentoses [15–18], and
trioses [19–22] via intramolecular hydride shifts, as well as in the
production of gamma-valerolactone from xylose [23] and alpha-
hydroxyacids [24,25] from C-2 and C-4 oxygenates. Recent work
shows that Sn-Beta catalyzes the chiral Baeyer–Villiger oxidation
of sugar derivatives [26].
Our recent work has shown that Sn-Beta with sodium tetrabo-
rate cooperatively catalyzes the selective epimerization of aldoses
in aqueous media under mild conditions [27]. The reaction mecha-
nism, investigated using 13C-labeled sugars and nuclear magnetic
resonance (NMR), proceeds by way of a 1,2 carbon-shift mecha-
nism whereby C1 forms a new C C bond with C3, while C2 moves
to the C1 position. In the absence of the borate promoter, no carbon
backbone rearrangement occurs and Sn-Beta catalyzes a regular
aldose–ketose isomerization via an intramolecular hydride shift as
reported in previous studies [14]. The epimerization mechanism is
analogous to the Bilik reaction using molybdenum-based catalysts
that form rigid complexes by interacting with the hydroxyl groups
of C1–C4 in the sugar [28,29].
Isomorphic substitution of tin for silicon tetrahedrally-
coordinated in the crystalline zeolite framework generates a closed
site. Hydration with two water molecules produces a Sn site with
octahedral coordination, while hydrolysis of one of the Sn–O–Si
bridges generates an open site. For the isomerization of glucose
to fructose, Davis and co-workers showed that the open-site in Sn-
Beta promotes the ring opening and the hydride transfer steps more
favorably than the closed site [30]. In the presence of borate, the
1,2 hydride-shift isomerization pathway is drastically inhibited in
favor of a 1,2 carbon-shift epimerization pathway. Interestingly, a
similar 1,2 carbon-shift product has been recently reported using
Sn-Beta in methanol, although epimerization rates are lower when
compared to the Sn-Beta/borate system in water [31]. Borates can
interact with the hydroxyl groups found in carbohydrates and with
the silanol groups found near the Sn open site, both with a strong
dependence on pH [32]. We hypothesize that the borate additive
interacts with the sugar and/or the active site to change the reaction
from a hydride-shift isomerization to a carbon-shift epimerization
pathway. Here, we present a full factorial experimental design to
investigate the impact of pH, sugar to borate ratio, and reaction time
2.2. Solutions
Within this study, the sugar content was held constant and the
borate content was varied to obtain the desired ratios. The sodium
to borate ratio was adjusted by mixing boric acid, sodium tetrabo-
rate, sodium metaborate and sodium hydroxide on a mass basis on
a 50 ml solution scale (see Table S1). The pH was measured with a
Mettler Toledo FiveEasy FE20 probe, calibrated with pH 6, 7 and 10
buffers prior to use at 298 K. For the control experiments, the pH-
adjusted solutions without borate were very sensitive and were

296
W.R. Gunther et al. / Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
343 K. 13C NMR was referenced with respect to the C-1 carbon
of d-arabinose at 99.47 ppm as a secondary standard relative to
4,4-dimethyl-4-silapentane-1-sulphonic acid.
therefore measured while mixing. NMR solutions were prepared
analogously in D2O on a 5 ml scale.
Raman spectra were acquired on a Kaiser RamanRXN1 spec-
trometer with a 784.807 nm exciting line from an Invictus diode
laser using a HPG-785 grating and a 20× microscope objective. For
analysis, a drop of sample was placed on a microscope slide coated
with aluminum foil. Post-processing included baseline removal and
normalization of the largest peak to an intensity of 1.
Optical rotation measurements were performed with a Jasco
P-1010 polarimeter using the 589 nm sodium D line at room tem-
perature. After performing separate epimerizations of l-arabinose
and d-arabinose and separation of the products by HPLC, the ribose
fraction was dried using a lyophilizer and redissolved in D2O (see
Fig. S5). The optical rotation was measured and used to calculate
the specific rotation based on the concentrations determined by
HPLC using a power law calibration curve for dilute samples. Solu-
tions were prepared 24 h prior to measurement in order to allow
the sugar conformers to equilibrate.
2.3. Reactions
For the factorial design experiments, Sn-Beta (Si/Sn = 96) was
added at a 100:1 sugar–metal molar ratio to a 5 wt% sugar solution
in a 5 ml thick-walled glass reactor containing a small magnetic
stir bar (typically ∼40 mg of catalyst in 2 ml of sugar solution).
Reactions were performed with d- and l-arabinose interchange-
ably with no difference in kinetics. Initial rate experiments were
done with larger volumes and higher sugar–metal ratios in order
to capture the low conversion data points. In order to determine
the initial rates at higher temperature ranges, an arabinose to Sn
ratio of 1000 or 2000 was used. The thick-walled glass vials were
sealed using a PTFE/silicone septa and metal crimp top, placed in a
temperature-controlled oil bath, and removed periodically to take
samples. The glass reactors were quenched in ice and a small sam-
ple volume was removed. Next, 100 ␮l of the filtered sample and
25 ␮l of a 10 wt% mannitol solution were mixed in a vial. The sam-
ples were analyzed by high-performance liquid chromatography
(HPLC) on an Agilent 1260 system equipped with photodiode array
ultraviolet and evaporative light-scattering detectors. The reaction
products were separated using a Bio-Rad Aminex HPX-87C column
heated to 353 K using deionized water (pH = 7) as the mobile phase
at a flow rate of 0.6 ml min−1. Ribulose concentrations were deter-
mined using an ultraviolet detector at a wavelength 210 nm. At this
wavelength all other sugars had negligible ultraviolet signals. For
all ultraviolet-inactive carbohydrates, the analysis was performed
using the evaporative light scattering detector. Samples were run
in duplicate on the HPLC and the column was regenerated daily
with 0.1 M Ca(NO3)2. Ketoses showed the strongest tendency to
shift retention time in the column. Fractionation and NMR were
used to check peak purity.
119Sn solid-state MAS NMR experiments were measured using
an 8.46-T (1H, 360.336 MHz) Oxford wide bore magnet and a
home-built spectrometer (courtesy of Dr. David Ruben, FBML-MIT).
Experiments were performed using a 4-mm Otsuka Electronics
MAS double-resonance probe, tuned to 119Sn. Powdered sam-
ples were packed into 4 mm outside diameter (o.d.) ZrO2 rotors
equipped with Vespel drive-caps and Kel-F top-caps with O-ring
seals. Dry samples were prepared by heating to 473 K under a
0.1 mbar vacuum and loading the NMR rotor in a glovebox. 119Sn
spectra were acquired using a Hahn-echo experiment, RD = 1.3*T1,
50 kHz vrf, 2–125 second recycle delays, a spinning frequency of
ωr/2ꢀ = 9–10 kHz, and between 70 and 80k co-added transients at
295 K. All 119Sn spectra were referenced with respect to SnO2 at
−604.3 ppm as an external standard relative to trimethyltin. A T1
curve was generated to determine an appropriate relaxation time.
Contour plots were generated by fitting a surface to the exper-
imental data using a linear scattered interpolant, applying five
cycles of 40 × 40 point surface convolution using a 3 × 3 matrix with
a central weight of 0.4, side weights of 0.1 and corner weights of
0.05, and then stretching this smoothed surface to fit the data using
nonlinear parameter fitting. This method avoided data over fitting,
the introduction of anomalous peaks due to cubic splines and the
use of polynomial surface models which have unusual behavior at
the edges of the data. In order to maintain continuity between the
sodium to borate and pH plots the pH contour plots were obtained
by a conformal mapping of the sodium to borate reactivity sur-
face using a separate pH surface response model. This was possible
because reactivity varied smoothly with sodium to borate ratio.
Contour lines were determined using a default Origin Delaunay
triangulation with B-Spline curves.
Conversion, selectivity, yield, and turn over frequency (TOF) are
defined as follows:
Initial moles of reactant − Moles of reactant
Conversion =
Initial moles of reactant
Moles of product − Initial moles of product
Selectivity =
Initial moles of reactant − Moles of reactant
Yield = Conversion ∗ Selectivity
Moles of product − Initial moles of product
=
Initial moles of reactant
Moles of converted
TOF =
Moles of Sn ∗ Time
2.4. Characterization
3. Results and discussion
Solution NMR was performed on a Varian Inova 500 MHz
spectrometer with an Oxford Instruments magnet and 5 mm
double-resonance probe, which was tuned to 1H, 11B, and 13C
before use. Samples were prepared at a sugar concentration of
0.05 g ml−1 in D2O using quartz NMR tubes. No correction was
made for an isotope effect on the ion product of water. The spin-
ning frequency was 20 Hz. For 11B 128 transients were taken and
for 13C 1024 transients were taken. All 11B spectra were referenced
using boric acid (H3BO3, 19.6 ppm) as a secondary standard rela-
tive to boron trifluoride etherate (BF3 EtO2). Boric acid was used
as an external standard for 11B NMR since the chemical shift for
free borate changes with pH. A variable temperature study was
also performed in 10 K increments between room temperature and
Borates have found application in organic synthesis due to their
ability to complex with cis-diols of organic molecules [33]. Thus,
in the presence of borates, sugars in solution can exist as a free
sugar, a one boron one sugar molecular complex (B−L), a one boron
two sugar molecular complex (B−L2) and/or a two boron one sugar
molecular complex (B−)2L (see Fig. S6). The ribose–borate complex
is ca. 41 kJ/mol more stable than the arabinose–borate complex
[34]. Previous studies have also shown that sugar–borate com-
plexes formed with tetrahedral borate are much stronger than
those formed with trigonal borates with complexation timescales
considerably faster than those involved in the epimerization reac-
tions [35]. Borates in solution can exist in an uncharged trigonal
form, a negatively charged tetrahedral form as well as an intricate

W.R. Gunther et al. / Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
297
Fig. 2. Schematic representation of borate species in solution and borate–sugar complexes.
polyanionic mixture (see Fig. 2). The nature of the borate specia-
tion in solution is a strong function of pH. For pH < 5 borate exists
almost exclusively as boric acid H3BO3; for pH > 11 as the tetrab-
orate ion B(OH)4−; and at intermediate pH values, as a complex
mixture of trigonal borate (B), tetrahedral borate (B−), and borate
polyanions [36]. Sodium tetraborate (the borate salt initially used
in the Sn-Beta/borate epimerization system) contains two neutral
B and two negatively charged B− species counterbalanced by two
sodium cations and, due to charge neutrality arguments, the frac-
tion of tetrahedral boron in solution is equal to the ratio of sodium to
boron in the starting material. Thus sodium tetraborate has a B−/B
of 0.5 while, sodium metaborate has a B−/B of 1. Sodium metabo-
rate corresponds to boric acid fully titrated with sodium hydroxide
(NaOH) into tetrahedral borate; therefore, titration of boric acid
Table 1
Full factorial design involving pH, borate to sugar ratio and reaction time as reaction variables for the epimerization of arabinose to ribose.
−
Entry
Borate–
arabinose
B−/B
pH
BL−
BL2
15 min
30 min
45 min
60 min
Ca [%]
Ca [%]
S1b [%]
S2c [%]
Ca [%]
S1b [%]
S2c [%]
Ca [%]
S1b [%]
S2c [%]
S1b [%]
S2c [%]
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
1:1
0.00
0.23
0.41
0.50
0.64
1.00
0.00
0.22
0.45
0.50
0.74
1.00
0.00
0.21
0.45
0.50
0.78
1.00
1.10
0.00
0.21
0.46
0.50
0.80
1.00
1.10
na
3.75
6.50
6.97
7.21
7.50
8.37
3.93
6.51
7.02
7.11
7.49
8.37
3.98
6.5
6.97
7.10
7.50
7.88
8.37
4.10
6.50
6.99
7.09
7.50
7.83
8.38
4.51
0
0
11
31
31
28
18
4
11
73
87
91
65
7
24
5
3
3
1
2
20
6
3
3
1
2
28
8
3
3
3
2
4
29
11
3
4
2
3
4
23
14
39
40
34
27
3
22
71
78
89
46
12
17
61
77
81
60
56
22
60
70
80
73
72
67
26
43
62
69
75
67
71
40
27
4
2
4
1
5
26
8
4
4
1
3
29
11
3
4
3
4
5
30
13
4
6
4
4
5
29
17
44
45
37
26
5
27
64
75
85
59
9
26
5
2
4
1
2
26
9
5
5
2
3
28
12
4
5
3
3
5
27
15
5
7
4
5
6
31
18
48
49
39
29
9
31
60
69
85
52
6
26
5
4
5
1
1
24
8
5
6
2
2
25
13
4
6
4
4
6
24
15
6
8
4
5
6
35
0.05
0.15
0.19
0.24
0.38
0
0.06
0.15
0.16
0.23
0.30
0
0.05
0.15
0.16
0.24
0.30
0.31
0
0.15
0.15
0.16
0.26
0.28
0.32
na
0.12
0.26
0.31
0.38
0.46
0
0.13
0.29
0.33
0.46
0.59
0
0.11
0.30
0.34
0.50
0.59
0.64
0
0.29
0.29
0.33
0.48
0.59
0.64
na
0.5:1
0.4:1
7
5
9
11
33
39
38
41
17
15
30
41
37
37
19
16
17
26
28
31
33
30
27.0
33
22
55
76
80
60
47
27
55
63
75
69
70
62
31
43
57
66
74
69
75
36
14
35
42
41
43
18
19
32
45
39
40
18
17
20
29
30
33
34
34
31
35
25
49
69
75
58
49
30
52
57
73
67
82
56
33
42
56
60
75
64
64
36
24
30
31
30
12
10
22
31
29
27
14
9
64
82
84
65
52
13
66
72
80
70
69
72
15
47
59
70
75
67
71
46
30
36
35
37
14
13
27
37
33
33
16
12
14
24
25
29
30
27
25
28
0.2:1
9
19
20
26
24
22
21
23
0:1
Reactions were performed at a 100:1 sugar–metal molar ratio with a 5 wt% arabinose feed at 343 K in a stirred batch reactor using Sn-Beta with a Si/Sn of 96. B−/B fraction
−
is equal to the Na/B ratio used to make the solutions. The fraction BL− and BL2 represent the normalized 11B NMR areas without correction for different NMR transfer
efficiencies.
a
C refers to arabinose conversion.
S1 refers to ribose selectivity.
S2 refers to ribulose selectivity.
b
c

298
W.R. Gunther et al. / Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
Fig. 3. Factorial design for the conversion of a 5 wt% arabinose solution at 343 K using Sn-Beta and a borate co-catalyst. Reactions were performed at a 100:1 sugar–metal
molar ratio with a 5 wt% arabinose feed at 343 K in a stirred batch reactor, using Sn-Beta with a Si/Sn of 96.
with NaOH provides an effective method to control the B−/B ratio,
and, consequently, reaction pH.
after 15 min. Note that the observed product distribution contains
ribose in excess of the amount predicted from the thermodynamic
equilibrium, indicating that the equilibrium is shifted toward the
epimer by preferential borate–ribose complexation. At high pH val-
ues the reaction rate was drastically inhibited and the low ribose
selectivity values obtained (<10%) are not quantitative since the
conversion measurements are within the error of the analytical
method. When the amount of borate is systematically decreased,
the optimal pH to maximize epimerization activity and selectivity
gradually increases. For instance, using a 0.2:1 borate–arabinose
ratio, a 27% arabinose conversion with 75% selectivity to ribose is
obtained at a pH of 8.38 after 45 min. These data also demonstrate
that epimerization pathways continue to dominate over isomer-
ization pathways even when substoichiometric amounts of borate
are used. The low ribulose yield obtained with substoichiomet-
ric borate suggests that borate inhibits isomerization even in the
presence of free sugar suggesting that the sugar borate complex is
dynamic and could form preferentially in the pores of the zeolite
(see Fig. S7). Control experiments and mass balances accounting
for >98% of the carbon exclude the possibility of low ribulose yields
due to borate–ribulose complexation. These results suggest that
one borate molecule may influence the reactivity of many sugar
molecules in a catalytic cycle or that borate species remain inside
the zeolite pores, potentially influencing the active site directly.
Borate ions are known to catalyze the isomerization of aldoses to
ketoses in alkaline media [37]. Control experiments in the presence
of borates but in the absence of Sn-Beta were performed to quantify
the extent of homogeneous isomerization. Table 2 shows results for
reactions performed with 1:1 borate–arabinose solutions. For all
Table 1 and Fig. 3 show the results of a full 7 × 4 × 4 facto-
rial design involving pH, sugar to borate ratio, and reaction time
as reaction variables for the epimerization of arabinose to ribose.
Experiments were designed to facilitate comparison at constant
B−/B and pH. In this manuscript, low pH refers to values lower than
7, near-neutral pH refers to values in the range of 7–7.8, and high
pH refers to values of higher than 7.8. In order to relate reactivity
to the concentration of borate species, the fraction of B−L and B−L2
was determined with 11B NMR at room temperature (see Supple-
mentary Table S2 and Fig. S10). A separate variable temperature 11
B
NMR study showed similar spectra up to 343 K with a shift toward
monoester at higher temperatures. The 11B data show that the frac-
tion of complexed borate increases steadily with pH and that the
ratio between mono and diester is almost constant across the pH
spectrum.
The experimental data show that pH has a drastic effect on
epimerization selectivity and reactivity. For a solution featuring
a 1:1 borate–arabinose ratio, low pH values hindered epimer-
ization pathways, yielding low ribose selectivities <30%. In the
presence of the borate additive, a conversion of 11% was obtained
at a pH of 3.75 after 15 min at 343 K, whereas in the absence of
borates, a conversion of 23% is achieved under similar reaction
conditions, thus suggesting that borates slightly inhibit reactiv-
ity at low pH values. Conversely, at near-neutral pH conditions
ribose selectivities exceeding 50% were obtained, reaching a max-
imum of 91% at a pH of 7.21. In the pH range of 6.5–7.5, high
arabinose conversions ranging from 18% to 32% were observed

W.R. Gunther et al. / Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
299
Fig. 4. Contour plots as a function of borate to sugar ratio and pH at 60 minutes for: (a) arabinose conversion, (b) ratio of ribose to ribulose selectivities (i.e., instantaneous
selectivity of ribose), (c) ribose selectivity, (d) ribulose selectivity, (e) ribose yield, (f) ribulose yield. The response surface lies the within experimental error of all experimental
points which are labeled with dots on the plot. Reactions were performed at a 100:1 sugar–metal molar ratio with a 5 wt% arabinose feed at 343 K in a stirred batch reactor,
using Sn-Beta with a Si/Sn of 96.
Table 2
the pH values investigated, arabinose conversion never exceeded
Control experiments for arabinose conversion in the presence of borates but in the
absence of Sn-Beta. Reactions were performed with a 5 wt% arabinose feed at 343 K
in a stirred batch reactor.
1% even after 120 min at 343 K. Reactions performed using a com-
bination of Si-Beta and borate were also unreactive.
Fig. 4 shows a series of contour plots mapping the dynamic ara-
Arabinose–Borate
1:1
B−/B
pH
Conversion [%]
60 min
binose conversion and ribose yield as a function of pH ratio and
borate–sugar ratio (see Fig. S8 for ribulose study). Note that a near-
linear relationship exists between the NaOH–HBO3 ratio and pH,
which is characteristic of weak acid titrations (see Fig. S9). Ribose
selectivity and arabinose conversion rates reached their maximum
values in near-neutral pH conditions and rapidly tailed off at either
extremes of the NaOH–HBO3 range investigated, thus leading to
120 min
0.00
0.23
0.41
0.64
1.00
3.75
6.50
6.97
7.50
8.37
<1
<1
<1
<1
<1
<1
<1
<1
<1
<1

300
W.R. Gunther et al. / Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
Fig. 5. Raman spectra for 5 wt% arabinose solutions with borate obtained at 298 K.
−
the appearance of a sharp maximum in ribose yield in the con-
tour plot. In the range where maximum reactivity and selectivity
is observed, the ratio of tetragonal to trigonal boron species (i.e.
B−/B) is ca. ∼0.5. Diverting from this B−/B ratio leads to sharp
decreases in activity and selectivity, possibly suggesting that both
species are required to promote the epimerization pathway. Note
that tetragonal borate readily interacts with internal diols in the
sugar backbone, whereas trigonal borate can react with hydroxyl
groups, such as those present in the Sn open site or in the silanols
present next to the Sn open site, via an sp3d intermediate [32]. The
rate constant for arabinose conversion in the presence of borate
under acidic conditions is approximately half of that calculated in
the absence of borate, thus suggesting that while trigonal borate
does not appreciably bind to arabinose it can inhibit the reaction
presumably by interacting with the Sn site. Additionally, that the
low ribulose selectivity and low unaccounted carbon suggest the
isomerization is virtually shut down at neutral pH conditions. Con-
sequently, borate binding may impact the binding of arabinose to
the active site and alter elementary steps involved in the hydride
shift isomerization mechanism [30]. Studies are currently under-
way to determine if Sn–O–B are formed during the reaction.
The mechanism for the Bilik reaction involves the formation
of a rigid complex between three to four hydroxyl groups of the
sugar and a molybdate dimer [38]. Since polyborates occur in near-
neutral pH conditions it is plausible that an analogous rigid complex
could be formed between a borate dimer and the sugar. Therefore,
Raman spectroscopy was used to probe borate-containing solutions
at different pH values for the presence of polyborates. As shown in
Fig. 5, the resulting spectra clearly show bands at 872 and 745 cm−1
corresponding to B(OH)3 and B(OH)4 species, respectively; how-
ever, no evidence for polyborate species with a sharp maximum
at near-neutral pH conditions was found [36]. Resonances corre-
sponding to borate chains were less apparent likely because sugar
complexation affects their vibrational frequencies or because they
are, in fact, a minor component in the solution. A mechanism that
requires polyborates would be difficult to rationalize in the context
of steric effects inside the pores of the zeolite and also in view of
the carbon-shift epimerization recently observed with Sn-Beta in
methanol.
11B NMR was used to further study the complete inhibition of
isomerization and epimerization reactions observed under basic
conditions (see Fig. S10 and Table S2). At high pH values, the
formation of B−L2 complexes is slightly favored. We hypothe-
size that the B−L2 complexes formed within the pores may cause
pore blocking. Also, the B−L2 complexes also lack open B
O H
bonds, effectively decreasing the number of available protons
that may participate during key reaction steps, including ring-
opening/closing sequences and product desorption from the site
[30,39]. Interestingly, Fig. S10 shows a 11B resonance at 10.9 ppm
in the BL2− region that reaches a maximum as a function of pH for a
given borate–arabinose ratio. The pH value at which this maximum
is observed increases as the borate–arabinose ratio decreases. A
compensation effect is seen wherein the fraction of the BL2− species
increases in such a way that the ratio of BL2− to BL- remains roughly
constant with decreasing borate loadings. Although the presence
or absence of B−L2 complexes cannot fully explain the inhibition or
promotion of reactivity, further studies are required to understand
the role of hydroxyl species on reactions rates.

W.R. Gunther et al. / Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
301
Fig. 6. Madon-Boudart diffusion test at 343 K Experiments were performed with
20 mg of Sn-Beta and a sufficient volume of 5 wt% arabinose to achieve a 2000
sugar–metal ratio in a stirred batch reactor.
Fig. 8. Arrhenius plot for first order epimerization and isomerization reactions with
1000:1 sugar–metal. Epimerization reactions were performed with a 5 wt% arabi-
nose 1:1 borate–arabinose feed at pH 7.21 in a stirred batch reactor using Sn-Beta
with a Si/Sn of 98. Isomerization was performed unbuffered with 5 wt% arabinose.
Kinetic parameters were extracted for arabinose conversion
under epimerization conditions. Prior to acquiring reactivity data,
an appropriate stirring speed was selected to avoid external
intermolecular mass transfer limitations, and a Madon-Boudart
test was performed to rule out internal mass transfer limita-
tions. As shown in Figs. 6 and S11, the initial epimerization TOFs
remained constant for identical catalyst volumes with zeolites
featuring Si/Sn ratios ranging from 98 to 367, thus showing the
absence of intra-pore diffusion artifacts at the reaction conditions
of interest. Next, the reaction order was determined by varying
the feed concentration under differential conditions over a wide
range of concentrations (see Figs. 7 and S12). Turnover rates were
determined by normalizing the moles of arabinose converted by
the total moles of Sn on each catalyst at conversion levels <15%.
As shown in Fig. 7, arabinose epimerization follows first order
kinetics, thus suggesting that sugar coverage at the active site is
low under the reaction conditions investigated. An Arrhenius plot
was used to calculate an apparent activation energy of 98 kJ/mol
and pre-exponential factor of 1.9 × 1014 L mol Sn−1 s−1 for the
epimerization reaction (see Fig. 8, Table S3 and Fig. S13). Ribose
selectivity remained high for all temperatures investigated, indi-
cating that homogeneous side reactions were not dominant. The
amount of ribose formed from ribulose isomerization (i.e., 1,2
hydride-shift pathway) with respect to that formed from arabinose
epimerization (i.e., 1,2 carbon-shift pathway) was estimated from
13C NMR data of reaction products obtained using [13C1]arabinose
feeds. A 5:1 arabinose–borate solution reacted at pH 7.21 for
60 min yielded ca. 29% [13C2]ribose C1, 4% [13C1]ribose, and 2%
[
13C1]ribulose. The ratio of [13C1]ribose to [13C1]ribulose is approx-
imately equal to the isomerization thermodynamic equilibrium
value (see Figs. S14 and S15). The NMR data for reactions performed
with stoichiometric borate show that a similar product distribution
is obtained to that of the 5:1 arabinose:borate case. Further, initial
rate data show that the [13C2]ribose and the [13C1]ribulose selec-
tivity remains constant at low conversions, which is consistent with
a reaction pathway that does not involve ribulose as an intermedi-
ate (see Table S4). These data suggest that the observed arabinose
consumption corresponds predominantly to ribose production via
the 1,2 carbon-shift epimerization pathway.
Analogous kinetic experiments were performed in the absence
of borate. An apparent activation energy of 91 kJ/mol and a pre-
exponential factor of 6.7 × 1012 L mol Sn−1 s−1 were obtained for
arabinose isomerization (see Supplementary Table S3 and Fig. S16).
These values are similar to the energy barriers previously calculated
for glucose [30]. Thus, under optimal conditions, the epimerization
rate constant at 343 K (0.23 L mol Sn−1 s−1) is roughly twice that
of the isomerization without borate (0.093 L mol Sn−1 s−1) and the
similarities between the isomerization and epimerization kinetics
are consistent with a common intermediate such as the open chain
form of the sugar. The differences in rates are mostly associated
with differences in the pre-exponential factors. The higher pre-
exponential factor observed for the epimerization of borates may
result from more favorable open vs. closed site distribution in the
presence of borate or due to a dual Lewis acid activation wherein
borate contributes to spatially arrange the sugar in a more desirable
configuration near the active site.
Optical rotation measurements demonstrated the stereospecific
nature of the epimerization reaction. The specific rotation of a sugar
varies with solvent and borate concentration, and requires time
to equilibrate, making it difficult to determine an enantiomeric
excess [40]. Optical rotation results shown in Table 3, confirm the
Fig. 7. Reaction order plot for initial rates at 343 K and 1000:1 sugar–metal with a
5 wt% arabinose 1:1 borate–arabinose solution at pH 7.21 in a stirred batch reactor.
The catalyst used was Sn-Beta with a Si/Sn of 98.

302
W.R. Gunther et al. / Journal of Molecular Catalysis A: Chemical 379 (2013) 294–302
Table 3
References
Optical rotation measurement used to assess stereospecificity.
[1] R. Wolf, F. Reiff, R. Wittmann, J. Butzke, Process for the preparation of riboflavin,
U.S. Patent US4355158 A, 1982.
Carbohydratea
[˛]D
Literature d-ribose[41]
Literature l-ribose[41]
Ribose from d-Arabinose
Ribose from l-Arabinose
−18.8 to −23.7
18.7–24.0
−14
[2] K. Okano, Tetrahedron 65 (2009) 1937–1949.
[3] N. Wymer, P. Taylor, Production of l-ribose and other rare sugars, US Patent
App. 11/501,997, 2006.
[4] C. Hu, L. Li, Y. Zheng, L. Rui, Appl. Microbiol. Biotechnol. 92 (2011) 449–455.
[5] M. Akagi, D. Omae, Y. Tamura, Chem. Pharm. Bull. 50 (2002) 866–868.
[6] Y. Tewari, R. Goldberg, Biophys. Chem. 22 (1985) 197–204.
[7] A. Corma, M.E. Domine, S. Valencia, J. Catal. 215 (2003) 294–304.
[8] A. Corma, M.E. Domine, L. Nemeth, S. Valencia, J. Am. Chem. Soc. 124 (2002)
3194–3195.
[9] A. Corma, L.T. Nemeth, M. Renz, S. Valencia, Nature 412 (2001) 423–425.
[10] A. Corma, M.T. Navarro, M. Renz, J. Catal. 219 (2003) 242–246.
[11] H.Y. Luo, L. Bui, W.R. Gunther, E. Min, Y. Román-Leshkov, ACS Catal. 2 (2012)
2695–2699.
21
a
Literature values are at 274 K while experimental values are at 298 K.
stereospecific nature of the epimerization reaction using Sn-Beta
in conjunction with borate salts.
4. Conclusion
[12] A. Corma, M. Renz, Arkivoc 2007 (2007) 40–48.
[13] Y. Román-Leshkov, M. Moliner, J.A. Labinger, M.E. Davis, Angew. Chem. Int. Ed.
49 (2010) 8954–8957.
[14] M. Moliner, Y. Román-Leshkov, M.E. Davis, PNAS 107 (2010) 6164–6168.
[15] M.S. Holm, Y.J. Pagán-Torres, S. Saravanamurugan, A. Riisager, J.a. Dumesic, E.
Taarning, Green Chem. 14 (2012) 702.
[16] C.M. Lew, N. Rajabbeigi, M. Tsapatsis, Microporous Mesoporous Mater. 153
(2012) 55–58.
[17] V. Choudhary, A.B. Pinar, S.I. Sandler, D.G. Vlachos, R.F. Lobo, ACS Catal. 1 (2011)
1724–1728.
[18] V. Choudhary, S.I. Sandler, D.G. Vlachos, ACS Catal. 2 (2012) 2022–2028.
[19] M.S. Holm, S. Saravanamurugan, E. Taarning, Science 328 (2010) 602–605.
[20] F. de Clippel, M. Dusselier, R. Van Rompaey, P. Vanelderen, J. Dijkmans, E. Mak-
shina, et al., J. Am. Chem. Soc. 134 (2012) 10089–10101.
[21] L. Li, C. Stroobants, K. Lin, P.A. Jacobs, B.F. Sels, P.P. Pescarmona, Green Chem.
13 (2011) 1175–1181.
[22] P.P. Pescarmona, K.P.F. Janssen, C. Delaet, C. Stroobants, K. Houthoofd, A. Philip-
paerts, et al., Green Chem. 12 (2010) 1083–1089.
[23] L. Bui, H. Luo, W.R. Gunther, Y. Román-Leshkov, Angew. Chem. Int. Ed. 52 (2013)
8022–8025.
A full 7 × 4 × 4 factorial experimental design for NaOH–HBO3
(i.e., pH), sugar to borate ratio, and reaction time facilitated opti-
mization of ribose production and showed that low pH values
favored isomerization pathways, near-neutral pH values favored
epimerization pathways and high pH values inhibited both epimer-
ization and isomerization. Optimal pH values ranging from 7.0 to
7.5 and high borate to sugar ratios were found to promote borate
complexation with the sugar, effectively decreasing the concen-
tration of free arabinose and promoting epimerization via a 1,2
carbon-shift mechanism. Importantly, 1,2 carbon-shift epimeriza-
tion products dominated most product distributions even when
using substoichiometric amounts of borate, thus suggesting a
potential interaction of borate species with the active site that
inhibits isomerization pathways. Reactivity and 11B NMR charac-
terization data showed that although the fraction of complexed
sugar (B−L and B−L2) increases with pH, reactivity declines at high
pH values, thus indicating that the fraction of complexed borate is
not a sufficient descriptor of reactivity. The maximum in reactivity
as a function of pH is consistent with both charged and uncharged
borates partitioning into the zeolite and participating in the reac-
tion. The Sn-Beta/borate system shows high epimerization activity
at low temperature following a first order rate law over a wide
temperature range. Optical rotation measurements confirmed the
stereospecific nature of the epimerization reaction. The isolated
Lewis acid sites inside the hydrophobic zeolite present an inter-
esting opportunity to study the Sn-borate catalytic pair and the
unusual C C bond-forming and bond-breaking events it promotes.
Future studies are geared at understanding the nature of the borate
interaction with the active site.
[24] M. Dusselier, P. Van Wouwe, S. De Smet, R. De Clercq, L. Verbelen, P. Van
Puyvelde, et al., ACS Catal. 3 (2013) 1786–1800.
[25] M. Dusselier, P. Van| Wouwe, F. de| Clippel, J. Dijkmans, D.W. Gammon, B.F.
Sels, ChemCatChem 5 (2013) 569–575.
[26] C. Paris, M. Moliner, A. Corma, Green Chem. 15 (2013) 2101.
[27] W.R. Gunther, Y. Wang, Y. Ji, V.K. Michaelis, S.T. Hunt, R.G. Griffin, Y. Román-
Leshkov, Nat. Commun. 3 (2012) 1109.
[28] L. Petrusˇ, M. Petrusˇová, Z. Hricovíniová, Glycoscience 215 (2001).
[29] A. Köckritz, M. Kant, M. Walter, A. Martin, Appl. Catal.,
112–118.
A 334 (2008)
[30] R. Bermejo-Deval, R.S. Assary, E. Nikolla, M. Moliner, Y. Román-Leshkov, S.-J.
Hwang, et al., Proc. Natl. Acad. Sci. U. S. A. 109 (2012) 9727–9732.
[31] R. Bermejo-Deval, R. Gounder, M.E. Davis, ACS Catal. 2 (2012) 2705–2713.
[32] A. Irwin, J. Holmgren, T. Zerda, J. Jonas, J. Non-Cryst. Solids 89 (1987) 191–205.
[33] C.J. Cowden, I. Paterson (Eds.), Organic Reactions, John Wiley & Sons, Inc., Hobo-
ken, NJ, 2004.
[34] J.E. Sponer, B.G. Sumpter, J. Leszczynski, J. Sponer, M. Fuentes-Cabrera, Chem-
istry 14 (2008) 9990–9998.
[35] R. van den Berg, J.a Peters, H. van Bekkum, Carbohydr. Res. 253 (1994) 1–12.
[36] T. Hirao, M. Kotaka, H. Kakihana, J. Inorg. Nucl. Chem. 41 (1979) 1217–1220.
[37] J. Mendicino, J. Am. Chem. Soc. 402 (1960) 4975–4979.
[38] L. Petrusˇ, M. Petrusˇová, Z. Hricovíniová, in: A. Stütz (Ed.), Topics in Current
Chemistry, Springer Berlin, Heidelberg, 2001, pp. 15–41.
Appendix A. Supplementary data
[39] R.S. Assary, L.A. Curtiss, J. Phys. Chem. A 115 (2011) 8754–8760.
[40] D. Adam, R.P. Jennings, Proc. South African Sugar Technol. Assoc., 1966, pp.
206–213.
Supplementary data associated with this article can be
found, in the online version, at http://dx.doi.org/10.1016/j.molcata.
2013.08.021.
[41] F. Phelps, H. Isbell, W. Pigman, J. Am. Chem. Soc. 3226 (1934) 747–748.