Structure-performance correlations of cross-linked boronic acid polymers as adsorbents for recovery of fructose from glucose-fructose mixtures

Article abstract of DOI:10.1039/c9gc03151k

Recovery of bio-oxygenates from reaction mixtures is one of the major challenges for future bio-refineries. Isolation of fructose produced by isomerization of glucose presents a typical example at a very early stage of the value chain. We propose to recov

Full text of DOI:10.1039/c9gc03151k

Green Chemistry
View Article Online
View Journal
PAPER
Structure-performance correlations of cross-
linked boronic acid polymers as adsorbents for
recovery of fructose from glucose–fructose
mixtures†
Cite this: DOI: 10.1039/c9gc03151k
b
Guido Schroer,a Jeff Deischter,a Tobias Zensen,a Jan Kraus,b Ann-Christin Pöppler,
a
Long Qi, c Susannah Scott d,e and Irina Delidovich
*
Recovery of bio-oxygenates from reaction mixtures is one of the major challenges for future bio-
refineries. Isolation of fructose produced by isomerization of glucose presents a typical example at a very
early stage of the value chain. We propose to recover fructose from a solution containing a mixture of
glucose and fructose by adsorption on polymers bearing phenylboronate moieties. p-Vinylphenylboronic
acid was polymerized with various cross-linkers, namely polar aliphatic, low-polarity aliphatic, and aro-
matic. The cross-linker content was in the range 5–40 mol%. The polymers exhibit high capacities for
fructose, with a maximum loading of up to 1 molFru molB−1. Fructose loading depends significantly on the
length and content of cross-linker, as well as pre-treatment of the polymer. In general, the maximum
fructose capacity correlates with the swelling ability of the polymers, since phenylboronate moieties
become available for adsorption upon swelling. In contrast, maximum glucose loadings are much lower,
in the range 0.1–0.3 molGlu molB−1, and depend only slightly on the type of cross-linker. The structures of
the glucose and fructose complexes and the kinetics of their uptake were studied by in situ MAS NMR.
Efficient desorption of fructose was observed in acidic medium, and more importantly, using CO2. The
structures of the polymers after repeated adsorption and desorption remain unchanged, as confirmed by
solid-state NMR. Adsorption-assisted isomerization of glucose catalyzed by soluble carbonates was also
studied. A 56% yield of fructose was achieved after 8 successive cycles of reaction and adsorption.
Received 6th September 2019,
Accepted 13th December 2019
DOI: 10.1039/c9gc03151k
rsc.li/greenchem
Keq ≈1 at 50 °C.6,7 Fructose is considered a more valuable
product, as it can be selectively converted to promising bio-
Introduction
Carbohydrates are major compounds obtained from biomass based platform chemicals, such as 5-HMF8 and levulinic
valorisation. They are frequently discussed as key intermedi- acid.9,10 Aside from its relevance for bio-refineries, the isomeri-
ates for the production of renewable fuels as well as bulk and zation reaction is of further industrial importance. Glucose-
fine chemicals in future bio-refineries.1,2 Glucose, as a primary fructose syrups are widely used as sweeteners. Due to the
product, is readily available from cellulose via hydrolysis.3 higher sweetening power of fructose compared to glucose,
Fructose is obtained from glucose by isomerization.4,5 high fructose contents are desired. Glucose-fructose syrups are
Importantly, the isomerization is an equilibrium reaction with currently manufactured by enzymatic isomerization, followed
by a separation process. Fructose concentrations of 42 mol%
after isomerization can be increased to 90 mol% by chromato-
aChair of Heterogeneous Catalysis and Chemical Technology, RWTH Aachen
graphic procedures such as simulated moving bed chromato-
graphy.11 Developing a more facile method for fructose syn-
University, Worringerweg 2, 52074 Aachen, Germany.
E-mail: delidovich@itmc.rwth-aachen.de
thesis would promote its industrial use in value chains for pro-
duction of fuels or commodities. In this regard, chemo-cata-
lytic processes, which are typically less expensive than enzy-
matic processes, associated with a suitable technique for
separation of the product, are highly desirable. Significant
efforts have been made to design suitable chemo-catalysts for
the glucose–fructose isomerization.4,12 At the same time,
developing facile and energy-efficient techniques for fructose
bInstitute of Organic Chemistry, University of Würzburg, Am Hubland, 97074
Würzburg, Germany
cU.S. DOE Ames Laboratory, Iowa State University, Ames, Iowa 50011, USA
dDepartment of Chemistry & Biochemistry, University of California, Santa Barbara,
California 93106, USA
eDepartment of Chemical Engineering, University of California, Santa Barbara,
California 93106, USA
†Electronic supplementary information (ESI) available. See DOI: 10.1039/
c9gc03151k
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
isolation represents a challenge. Conventional separation tech-
niques, such as distillation, cannot be used to isolate fructose
since the compound is highly polar and non-volatile.13
One method to separate glucose and fructose is based on
their reversible esterification with boronates. The latter form
cyclic esters with molecules bearing a 1,2-diol motif, such as
saccharides.14 The affinity of a saccharide for the boronate
depends on the saccharide structure.15–17 Fructose typically
exhibits much higher complexation constants with boronates
than glucose. With phenylboronates the association constants
are K = 110 for glucose and K = 4370 for fructose, respect-
ively.18 This phenomenon is called “molecular recognition”:
boronates “recognize” fructose in the presence of glucose and
selectively form a complex with the former.19
Fig. 1 Schematic representation of the proposed separation of fructose
using adsorption on boron-containing polymers. Structures of cross-
linkers are provided in Table 1.
Experimental
Chemicals
Unless specified, all chemicals were used without further puri-
fication. p-Vinylphenylboronic acid (p-VPBA, 97%) was sup-
plied by ChemPUR. D-(−)-Fructose (Ph. Eur.), 1,4-phenylene
dimethacrylate (PDMA; >99%) and 2,2′-azobis(2-methyl-propio-
nitrile) (AIBN, >98%) were purchased from Sigma-Aldrich. D-
(+)-glucose (Reag. Ph. Eur.) and sodium dihydrogen phosphate
monohydrate (Reag. Ph. Eur.) were provided by Merck. Sodium
hydrogen carbonate (>99%) and chloroform (>99%) from Roth
and conc. hydrochloric acid (37 wt%) from Honeywell were
used. Toluene (>99.7%), sulfuric acid (98%), ethanol (99.9%),
anhydrous benzyl alcohol (>99.8%) and sodium hydroxide
Several analytical methods have been designed based on
molecular recognition, including boronate affinity chromato-
graphy20 and boronate affinity saccharide electrophoresis.21
Recently, a number of preparative methods were proposed
based on molecular recognition with boronates for recovery of
saccharides, diols, and polyols produced from biomass.22–34
A
few studies were aimed at fructose isolation. Typically, these
approaches are based on the extraction of a fructose-boronate
anionic complex into an organic phase, while glucose remains in
the aqueous phase.29,35–38 However, some challenges connected
with the extraction process remain unsolved. These include the
necessity of using an organic solvent as well as the need for envir-
onmentally unfriendly quaternary amines as counter-cations for
the fructose-boronate anionic complex in the organic phase.
Leaching of boron into the aqueous phase presents another chal-
lenge. These disadvantages of liquid–liquid anionic extraction
may be overcome by implementing an adsorption-based separ-
ation process using polymers bearing boronate functionalities.
Barker et al. proposed the idea of adsorption-assisted
glucose isomerization to fructose for the first time in the
microgranulate
(≥98.8%)
were
obtained
from
GeyerChemSolute. Divinyl sulfone stabilized with 0.05% hydro-
quinone (DVS, 97%), ethylene glycol dimethacrylate stabilized
with 100 ppm mehq (EDMA, 98%), triethylene glycol dimeth-
acrylate stabilized with 200–300 ppm mehq (TDMA; 95%), 1,7-
octadiene (OD; 98%), 1,9-decadiene (DD, 98%) and 1,13-tetra-
decadiene (TD, 90+%) were purchased from Th. Geyer.
Diethanolamine (DEA, 99%) and divinylbenzene (DVB, 80% di-
vinylbenzene: 20% ethylbenzene mixture, 99%) were provided
by Alfa Aesar. Distilled water was used to prepare solutions.
1970s.39 They used
a p-vinylphenylbronic acid (p-VPBA)
polymer cross-linked with ca. 5% divinylbenzene (DVB) for
in situ adsorption of fructose upon isomerization of glucose in
the presence of NaOH. This enabled isolation of adsorbed
fructose and an increase in fructose yield up to 57% was
reported. The process was emphasized as promising and
future studies proclaimed, but research ceased without further
reasoning.
In this study, we revisit the previously proposed approach,
aiming to elucidate relationships between the molecular struc-
ture of the polymers and their adsorption capacities. Fig. 1
schematically represents the steps undertaken. First, the poly-
mers were synthesized with various types and contents of
cross-linkers. Adsorption of fructose and glucose on the poly-
mers, as well as their subsequent desorption, were studied. To
obtain molecular-level insight on the structure-performance
relationships, significant effort was expended in material
characterization including classical solid-state NMR and in situ
Esterification of p-VPBA
The esterification method described in the literature was opti-
mized.40 In a typical procedure, stoichiometric amounts of
p-VPBA (16.00 g, 108 mmol) and DEA (11.36 g, 10.36 mL,
108 mmol) were added to 640 mL toluene used as an azeotro-
pic solvent. The reaction mixture was heated to 115 °C and
water was successively removed with a water separator. After
2 h, the reaction mixture was cooled in an ice bath and the
cold solution was filtered over a glass frit. The recovered imi-
nodiethyl p-vinylphenylboronate was extensively washed with
toluene, dried under high vacuum, and stored in a freezer. The
product was obtained as a white powder in yields ranging from
80 to 95%. The structure was confirmed by H- and 13C NMR.
Spectra are provided in ESI (Fig. 1S and 2S†).
1
Synthesis of 2,6-divinylnaphthalene
monitoring of adsorption by MAS NMR. Finally, adsorption- 2,6-Divinylnaphthalene is the only cross-linker that was not
assisted isomerization of glucose to fructose was performed in commercially available. Its synthesis is based on a literature
the presence of carbonate buffer (Na2CO3 + NaHCO3, pH 10) procedure.41 2,6-Dibromonaphthalene (2.63 g, 9.20 mmol),
as catalyst.
potassium vinyltrifluoroborate (3.7 g, 27.62 mmol), PdCl2
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
(0.10 g, 0.55 mmol), Cs2CO3 (18.0 g, 55 mmol) and PPh3 Desorption and recycling
(0.43 g, 1.64 mmol) were dissolved in 50 mL of a 9 : 1 v : v
After adsorption, each polymer was filtered over a glass frit
and dried under high vacuum. Different acid solutions were
screened for desorption (Table 5). In a typical procedure, the
desorption solution (2.63 mL) was added to 0.100 g polymer
containing adsorbed fructose, and the solution was allowed to
stir for 4 h at 500 rpm. The solutions were processed and ana-
lysed by HPLC as described above.
For polymer recycling, each polymer was filtered over a
glass frit, dried under high vacuum, and used in a second
adsorption–desorption cycle.
mixture of degassed THF and water. The solution was refluxed
at 75 °C under a nitrogen atmosphere for 16 h. After cooling to
RT, the solvent was removed under reduced pressure. Water
(75 mL) was added to the residue and the aqueous phase was
extracted with ethyl acetate (3 × 20 mL). The combined organic
extracts were then successively washed with 50 mL water and
25 mL brine, dried over MgSO4, and the solvent removed
under reduced pressure. After column chromatography with
neat n-heptane and solvent removal under reduced pressure,
the product was obtained as a white solid (1.18 g, 71%). The
structure was confirmed by NMR (Fig. 3S†).
Adsorption-assisted isomerization
A 50 mL two-necked round-bottom flask was charged with 1 g
glucose and 9 mL carbonate buffer (resulting in a 10 wt%
glucose solution). Isomerization was allowed to proceed for
1 h at 63 °C. These reaction conditions were chosen based on
previous screening studies (Fig. 32S–35S†). After the reaction,
the solution was cooled quickly using an ice bath and once RT
was reached, 6.57 mL solution was added to 0.250 g polymer.
The resulting mixture was stirred at 500 rpm for 3 h at RT. The
solution was filtered with a polyamide syringe filter (Chromafil
PA 20/25) to remove the polymer. To restore the initial sugar
concentration, 4 mL of the filtrate was combined with 6 mL of
a fresh 10.5 wt% glucose solution for use in the next isomeri-
zation step. Eight consecutive adsorption-separation cycles
were performed. 1 mL aliquots of the solution prior to isomeri-
zation, after isomerization and after adsorption were analysed
by HPLC.
Synthesis of cross-linked p-VPBA polymers
Polymerization was based on a previously reported pro-
cedure.39 Iminodiethyl p-vinylphenylboronate (5.00 g,
23.03 mmol) and AIBN (14 mg, 0.4 mol%) were dissolved in
14 mL dried and degassed benzyl alcohol. Various amounts
and types of cross-linker were added (Table 1S, ESI†). The
mixture was degassed in a 50 mL Schlenk-flask for 30 min. The
reaction was allowed to proceed under an Ar atmosphere for
20 h at 80 °C. The solid product was ground, washed with
chloroform, dried over a glass frit, and stored in a vacuum dessi-
cator (60 °C). After drying, the product was finely ground with a
mortar and pestle, washed three times successively with 70 mL
aqueous 1 M hydrochloric acid solution, water, and an 0.5 M
aqueous phosphate buffer solution (pH 7.5, prepared by drop-
wise addition of 4 M NaOH to 0.5 M aqueous NaH2PO4 until
pH 7.5 was reached). The product was obtained as a white
powder after drying in a vacuum dessicator at 60 °C. Polymer
yields ranged from 62% (not cross-linked) to 99% (Table 1S†).
Solid-state NMR
All solid-state NMR experiments with unlabelled material
were performed on a Bruker Avance III NMR Spectrometer at
14.1 T with either 20 or 24 kHz MAS. A 3.2 mm HX probe
Buffer solutions
Two types of buffer solutions were used in this study, namely,
a phosphate and a carbonate buffer. The carbonate buffer
solution was obtained by mixing of 0.5 M aqueous carbonate
buffer (Na2CO3 + NaHCO3, pH 10, prepared by dropwise
addition of 4 M NaOH to 0.5 M aqueous NaHCO3 until pH 10
was reached) with ethanol in a ratio 80 : 20 v : v. The phosphate
buffer solution was obtained by mixing of 0.5 M aqueous phos-
phate buffer (Na2HPO4 + NaH2PO4, pH 7.5, prepared by drop-
wise addition of 4 M NaOH to 0.5 M NaH2PO4 until pH 7.5 was
reached) with ethanol in a ratio 80 : 20 v : v.
1
was used. In all cases, the H 90° pulse duration was 2.5 µs.
For 1H–13C cross-polarization (CP) experiments, a 2 ms contact
1
time with a ramp on the H channel was used. During acqui-
sition, SPINAL64 heteronuclear decoupling was employed.
1H–13C CP FSLG-HETCOR experiments were recorded using
the same setup, but with shorter and varying contact times.
Chemical shifts were referenced using adamantane. 11B NMR
spectra were acquired using the Hahn-echo sequence to reduce
the background signal from the probe. All experiments were
nominally performed at RT. However, frictional heating due to
MAS results in a higher actual sample temperature. Spectral
processing was done using the Bruker TopSpin 3.5 software. In
particular, FIDs in 11B Hahn-echo experiments were left-
shifted to the echo maxima.
Adsorption isotherms
Stock solutions for recording adsorption isotherms were pre-
pared by dissolving 0.063–1.250 g of either glucose or fructose
in 18.75 mL carbonate buffer solution. Each stock solution
(2.63 mL) was added to 0.100 g polymer and the resulting solu-
tion was stirred at 500 rpm for 3 h at room temperature (RT).
In situ 13C MAS NMR
The solutions were filtered with polyamide syringe filters The kinetics of carbohydrate sorption were measured on an
(Chromafil PA 20/25) and analysed by HPLC.
Avance IPSO 500 MHz WB NMR Spectrometer (11.7 T)
For competitive sorption studies, stock solutions were pre- equipped with an MAS double resonance broadband 1H/X
pared in the same manner, using a 70 : 30 mol% : mol% probe. The 4 mm NMR rotor was loaded with 4 mg of either
mixture of glucose and fructose.
1-13C-glucose or 2-13C-fructose (22 µmol), 6 mg of 20 mol%
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
DVB-cross-linked polymer and 75 µL of an 80 : 20 mixture (v : v) molecule was initially located in the centroid of the cylinder.
of 0.5 M NaHCO3 buffer in D2O (pH adjusted to 10 with 4 M The optimization function rotates the molecule in space until
NaOH) and EtOH. Time-dependent 13C direct polarization the maximum squared radius is minimized. Since the cylinder
1
spectra without H decoupling were recorded at a MAS rate of fit is rotationally symmetric, the function takes only the x and
5 kHz. Each spectrum acquisition required 20 min, for a total y coordinates into account to reduce the number of variables.
experiment time of 4 h.
Thus, the optimization is reduced to a 2D solver problem. The
function was solved using the Matlab® globalSearch function
from the Global optimization toolbox.43 An additional optimiz-
ation was performed to translate the molecule into cylindrical
form to further reduce the cylinder radius.
Elemental analysis
Elemental analysis was performed using an ICP-OES
Spectroblue Instrument from Spectro Analytical Instruments.
20–30 mg sample, 1 g KOH and 0.12 g KNO3 were fused in a
China jar. The melt cake was added to 50 mL aqueous 10 wt%
hydrochloric acid solution and analyzed. Carbon, hydrogen
and nitrogen contents were measured with a 2400 Series II
CHNS/O Elemental Analyzer from PerkinElmer. Blank samples
and acetanilide as standard were used for instrument cali-
bration. Selected samples of polymers were analyzed in
Mikroanalytisches Laboratorium Kolbe in Mülheim.
HPLC
The concentrations of fructose and glucose were determined
by high performance liquid chromatography (HPLC) using a
Shimadzu Prominence LC-20 system. Samples containing only
fructose or glucose were injected into two successively con-
nected organic acid resin columns (CS-Chromatography, 100 ×
8.0 mm and 300 × 8.0 mm) connected via a 6-port switching
valve equipped with two pumps (A + B). The columns were
heated to 40 °C and the eluent (154 μL CF3COOH in 1 L of
water) was supplied at a flow rate of 1 mL min−1. During the
first 7 min after injection, the eluent was supplied with pump
A and pumped through the connected columns as well as a
refractive index (RI) detector. Then, the valve position was
switched to allow elution through each column separately for
the next 18 min. Using pump A, the eluent was pumped
through the first column (100 × 8.0 mm), and a photodiode
array (PDA) detector with a detection wavelength of 270 nm, to
analyse possible furanic by-products. Simultaneously, the
eluent was pumped through a second column (300 mm ×
8.0 mm) and the RI detector using pump B, to analyse mono-
saccharides and acidic by-products.
Nitrogen and water vapor physisorption
Nitrogen adsorption/desorption isotherms were recorded on a
Quadrasorb
SI
instrument
from
Quantachrome/3P
Instruments. Prior to the measurement, the sample was
degassed at 60 °C for 16 h. Total pore volume was monitored
at relative pressures of 0.900 p/p0. Water vapour adsorption
was measured at 20 °C on an Autosorb iQ instrument from
Quantachrome/3P Instruments. Before the measurement, the
sample was degassed at 60 °C until it passed the degassing
test (21 mTorr min−1). Pore volume was calculated at relative
pressures of 0.900 p/p0 after an equilibration time of 5 min.
Swelling degree
Samples containing both glucose and fructose were diluted
by factor of 10 prior to analysis. 10 mL of each diluted sample
was treated successively with 200 mg of Amberlyst® 15 for
30 min, and 500 mg ion-exchange resin Amberlyte® IRA-96
Polymer swelling was assessed gravimetrically, according to a
literature procedure.42 0.100 g polymer was placed in a 5 mL
centrifuge tube inlay. The weights of the empty (m1) and filled
(m2) inlays were noted. The filled inlay was inserted into a
15 mL centrifuge tube and the tube was filled with the specified
soaking solution. Swelling was monitored in the carbonate
buffer. For the determination of the swelling degrees in pres-
ence of glucose and fructose, 0.500 g of the respective sugar
were added to the solution and samples ultra-sonically mixed
for 3 h. HPLC samples were taken prior and after the procedure
to determine and subtract the adsorbed amounts of glucose
and fructose from m3. After standing for 24 h, the supernatant
liquid was removed, and the tube was centrifuged for 15 min at
4000 rpm. The inlay was recovered and its weight (m3) noted.
The swelling degree SD was determined using eqn (1).
free base for
1 h each. Samples were injected into a
Phenomenex 5 μL Luna-NH2 column (250 × 4.6 mm) and
eluted at 35 °C with an eluent (acetonitrile and water in a ratio
80 : 20 v : v) supplied at a rate of 1 mL min−1. The system was
equipped with an RI detector. An example chromatogram and
the corresponding calibration curves are provided in the ESI
(Fig. 4S†). An HPLC error of 2% was estimated for determi-
nation of glucose and fructose concentrations. Selected adsorp-
tion experiments were performed in duplicate or triplicate.
Results and discussion
Polymer synthesis and characterization
ðm3 ꢀ m1Þ ꢀ ðm2 ꢀ m1Þ
SD ¼
ꢁ 100%
ð1Þ
m2 ꢀ m1
Polymers with different cross-linkers were synthesized as
shown in Fig. 1. p-Vinylphenylboronic acid was protected with
DEA and co-polymerized with the respective cross-linker. A
Determination of the cross-linker length
Molecular structures were geometry optimized using the high boron content is a prerequisite for high sorption capacity,
MM2 method. First, the rigid molecule was placed in a cylin- since adsorption takes place by stoichiometric esterification.
der with a minimum diameter. The center of gravity of the Therefore, the cross-linker content was kept low, in the range
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
Table 1 Composition and properties of synthesized polymers, SD stands for swelling degree
Cross-linker
Boronic acid content
Pore volume
Length
[Å]
Amount
[mol%]
Entry
Name
Structure
Calc.
Found
N2
H2O
0.23
SD [%]
1
2
Not cross-linked
Divinylsulfone (DVS)
—
9.6
0
20
6.733
5.617
4.672 (69%)
4.117 (73%)
107
67
0.004
3
4
5
6
Ethylene glycol
dimethacrylate (EDMA)
14.9
22.0
5
6.275
5.212
4.252
4.725
4.080 (65%)
3.663 (70%)
2.895 (68%)
3.441 (73%)
74
60
45
56
20
40
20
0.004
0.002
0.20
0.24
Triethylene glycol
dimethacrylate (TDMA)
1,7-Octadiene (OD)
1,9-Decadiene (DD)
1,13-Tetradecadiene (TD)
Divinylbenzene (DVB)
7
8
9
10
11
12
13
13.0
15.5
20.6
11.9
20
20
20
5
20
40
20
5.668
5.493
5.062
6.353
5.432
4.552
5.252
4.330 (76%)
4.791 (87%)
4.293 (85%)
3.996 (63%)
3.506 (65%)
3.219 (71%)
3.717 (71%)
0.005
0.006
0.006
0.26
0.25
0.26
77
80
85
75
68
46
70
0.005
0.22
2,6-Divinyl naphthalene (DVN)
13.6
15.1
0.002
0.006
0.21
0.23
14
1,4-Phenylene dimethacrylate (PDMA)
20
4.911
3.219 (66%)
71
of 5 to 40 mol%. The types of used cross-linkers are listed in in interaction with components of the solvent (i.e., water and
Table 1. They include polar aliphatic (DVS, EDMA, and TDMA), ethanol). Thus, highly polar cross-linkers are most probably
low-polarity aliphatic (OD, DD, and TD), and aromatic (DVB, solvated by water, whereas low-polarity cross-linkers are sol-
DVN, and PDMA) molecules. The cross-linkers differ in length, vated by ethanol. Apparently, these solvation differences reflect
as estimated by the cylinder fitting procedure. The polymers an inverse dependence of the swelling degree on the length for
were obtained in high yields of 62 to 99%. The boronic acid highly polar cross-linkers. The polymers were further charac-
contents range from 60 to 90% of the theoretical amounts terized by 1H, 13C and 11B MAS NMR. For the cross-linked poly-
(Table 1). Loss of boron moieties can be explained by partial mers, linker-specific resonances were present in addition to
hydrolysis during synthesis. In addition, IR spectra of poly- resonances observed for the not cross-linked polymer. To
mers containing EDMA, TDMA and PDMA exhibit a character- facilitate assignments, the solution-state spectrum of the not
istic carbonyl stretching band at ca. 1700 cm−1. As expected, cross-linked polymer was recorded in a mixture of THF-d8 and
this band is absent in the spectra of other polymers (Fig. 5S– D2O (9 : 1 v : v). The resonances for the dissolved polymer gen-
7S†). The dry polymers are non-porous, as shown by nitrogen erally appeared at the same chemical shifts as for the solid
physisorption (Table 1). Estimated specific surface areas of the sample (Fig. 8S†), but the solution-state resonances were much
polymers were ca. 1 m2 g−1. However, pore volumes of all narrower than the solid-state resonances. Using this infor-
materials are significantly higher as determined in physisorp- mation in combination with literature data on polystyrene, all
tion with water than nitrogen, indicating that the polymers polymer signals were assigned.44 An atactic structure was
swell in aqueous solution. Hence, the swelling was studied by deduced from the distribution of 13C chemical shifts around
gravimetric analysis in the carbonate buffer, particularly 40 to 50 ppm corresponding to the CH2 group of the backbone
because the same buffer was later used for the adsorption iso- in both the solid state and in solution. In addition to the
therm experiments as well as during the adsorption-assisted characteristic polymer signals, resonances were observed at 52
isomerization of glucose into fructose. Swelling degrees deter- and 62 ppm. They are assigned to DEA, used as the protecting
mined for all synthesized polymers are listed in Table 1. The agent, although the nitrogen content is below the detection
content of cross-linker was varied for EDMA and DVB. In both limit of the CHN analysis. Moreover, the polymer
cases, the swelling degree decreases upon increasing the cross- contains residual benzyl alcohol used as the solvent for
linker amount. The swelling degree can be further correlated polymerization, based on the presence of a 13C resonance at
with the length of the cross-linker: for low-polarity aliphatic 64 ppm as well as a signal at 4.5 ppm in the 1H NMR
(entries 7–9, Table 1) and aromatic (entries 11, 12 and 14, spectrum. 1H-13C HETCOR MAS NMR experiments (Fig. 9S†)
Table 1) cross-linkers, longer cross-linkers result in higher confirm these assignments.
swelling degrees. Unexpectedly, the swelling degree of poly-
In addition to the general 1H and 13C NMR characterization
mers containing highly polar cross-linkers decreases as the of the polymers (Fig. 10S†), the nature of the boron functional-
length of the cross-linker increases (entries 2, 4, and 6, ity is important for their desired application as fructose recov-
Table 1). This effect is probably associated with the higher ery materials. Two characteristic broad signals at 15 to 30 ppm
oxygen content of the longer cross-linkers, leading to a change and 0 to 10 ppm corresponding to sp2- and sp3-hybridized
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
boron, respectively, were observed in the 11B MAS NMR spectra
(Fig. 11S†).45
Adsorption isotherms
After characterization, the polymers were tested for fructose
and glucose sorption. First, the sorption conditions were
screened and optimized using a polymer cross-linked with
20 mol% DVB. Addition of an organic solvent to the aqueous
buffer solution was necessary to ensure polymer dispersion
since the polymers are not wetted in pure aqueous solution.
Different organic co-solvents were tested to improve the wett-
ability of the polymers (Table 4S†). The best results were
achieved by adding 20 vol% ethanol to the aqueous sugar-
buffer solution. Sorption experiments were performed in both
a carbonate buffer solution (pH 7.5 and pH 10) and a phos-
phate buffer solution (pH 7.5 and pH 10). Significantly lower
Fig. 2 (a) 11B MAS NMR spectra of polymers cross-linked with 20 mol%
DVB recorded with the Hahn-echo-sequence after various pre-treat-
ments, (b) fructose sorption isotherms of the variously pre-treated and
dialysed polymers. Conditions: 0.100 g polymer, 2.63 mL carbonate
buffer solution, 0.063–1.250 g fructose, 3 h, RT.
sugar adsorption was noted at pH 7.5, which can be explained bonate. Such interactions have been reported, although the
by its lower pH. A greater fraction of boron is sp3-hybridized at contributions of such complexes are expected to be small due
pH 10 than at pH 7.5.46 In addition, a significant decrease of to competition with OH− anions under basic conditions
the pH is noted when performing the sorption at pH 10 in the (Fig. 14S†).48 The ratio of available sp3 to sp2 boron environ-
phosphate buffer, as the pH is above the buffer regime of the ments decreases in the following order: polymer treated with
buffer system.47 Sorption isotherms were therefore recorded in carbonate buffer, polymer dialysed after treatment with car-
the carbonate buffer solution (pH 10) to which 20 vol% EtOH bonate buffer, polymer treated with phosphate buffer, and
was added (Table 5S†). The sorption equilibrium was polymer dialysed after treatment with phosphate buffer
reached after approximately 3 h for fructose, and 90 min for (Fig. 2a). Interestingly, the polymer pre-treated with the phos-
glucose (Fig. 12S and 13S†). The Langmuir model was used to phate buffer exhibits a significantly higher fructose sorption
analyse the measured adsorption isotherms. Importantly, the capacity than the material pre-treated with carbonate buffer,
glucose adsorption isotherms are saturation curves, whereas even though the latter contains a higher fraction of sp3-hybri-
saturation was not typically obtained for fructose sorption. dized boron atoms (Fig. 2b). Pre-treatment with the phosphate
Thus, the maximum fructose capacity was deduced using the buffer results in a more pronounced swelling of the polymers.
best fit of the Langmuir isotherm equation. In addition, 4-bro- The swelling degrees after pre-treatment with phosphate or
mopolystyrene cross-linked with 20% DVB was tested for sugar carbonate were 68 and 61%, respectively (Table 6S†). The influ-
sorption under optimized conditions. No sorption was found, ence of the counter-anions on the swelling can be attributed to
confirming the necessity of boronate moieties for sugar the polyelectrolyte swelling effect.49 During pre-treatment,
sorption.
Na2CO3 and NaHCO3 as well as Na2HPO4 and NaH2PO4 are
After optimization, the influence of the polymer pre-treat- occluded by the polymer. During swelling, i.e. at pH 10, car-
−
2−
ment on the sorption behaviour was studied. The synthetic bonate ions are present as a mixture of HCO3 and CO3
,
2−
procedure includes the polymerization of p-vinylphenylboronic whereas all phosphate species are present as HPO4 anions.
acid protected with DEA, followed by washing with HCl to This leads to a higher repulsion between the doubly-charged
remove the protecting group (Fig. 1). The resulting polymers hydrophosphate anions and the negatively charged boronate
contain mainly boronic acid groups in sp2 hybridization, moieties, facilitating swelling. Dialysis of the polymers against
whereas a sp3 hybridization of the boronic acid (“boronate”) is water results in partial removal of the occluded phosphate and
more favourable for complexation.33,46 For anionic extraction carbonate anions, causing a drop in swelling degree and
of saccharides, the boron-containing species were treated with adsorption capacity (Fig. 2b). In general, a positive effect of
a basic solution to convert the boronic acid sites into boro- phosphates on the adsorption properties is observed.
nates to facilitate sorption.35 In this study, we washed the poly- Therefore, all polymers were subsequently pre-treated with the
mers with either a phosphate buffer (pH 7.5) or a carbonate phosphate buffer prior to adsorption experiments.
buffer (pH 10.0) and dried them afterwards. Subsequently, the
The pre-treated polymers contained predominantly sp2
washed polymers were purified by dialysis against water. hybridized boron moieties along with a small share of the
Fig. 2a compares the 11B MAS NMR spectra of polymers after negatively charged sp3 hybridized boronates (Fig. 2a). The
the washing pre-treatments and dialysis. A resonance corres- portion of the negatively charged sp3-hybridized boron
ponding to sp3-hybridized boron is present at a lower fre- moieties increases significantly upon fructose adsorption
quency (4 ppm) after pre-treatment with the phosphate buffer according to the solid-state 11B NMR (Fig. 11S†). We also tried
compared to pre-treatment with the carbonate buffer (5 ppm). to measure the zeta potential of the polymers before and
The different shifts might be caused by molecular complexa- after adsorption in order to detect negatively charged surface
tion of the phenylboronic acid with either phosphate or car- moieties. Unfortunately, the measurement failed due to
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
Table 2 Maximum fructose and glucose sorption capacities for the polymers with various content of DVB and EDMA cross-linkers, based on
Langmuir isotherm fits
Fructose
Glucose
mmolFru g−1
molGlu molB
mmolGlu g−1
−1
−1
Entry
Cross-linker
molFru molB
1
2
3
4
5
6
DVB, 5 mol%
1.478a
0.854
0.499
1.108a
0.618
0.344
5.906a
2.993
1.607
4.519a
2.261
0.997
0.266
0.201
0.139
0.206
0.171
0.111
1.064
0.704
0.477
0.840
0.627
0.322
DVB, 20 mol%
DVB, 40 mol%
EDMA, 5 mol%
EDMA, 20 mol%
EDMA, 40 mol%
Conditions: 0.100 g polymer, 2.63 mL carbonate buffer solution, 0.063–1.250 g fructose or glucose, 3 h, RT. a Fit based on adsorption at low con-
centrations, since the highly viscous solution became impossible to filter at high concentrations.
insufficient dispersion of the particles in solution (for further measured at low sugar concentration. This lowers the accuracy
details see p. 14, ESI†).
of the fit, such that the maximum sorption capacity exceeds
−1
After determining the effect of the polymer pre-treatment 1 molFru molB (since complexation is based on a stoichio-
on the sorption behaviour, the influence of the cross-linker metric reaction between the boronate and the sugar, the theore-
type and amount was evaluated. Polymers cross-linked with 5, tical maximum sorption capacity cannot exceed 1 molFru
20, or 40 mol% DVB and EDMA were tested for fructose and molB−1). A concentration of 20 mol% cross-linker was con-
glucose sorption. Maximum sorption capacities are listed in sidered as a good compromise in maintaining sufficient
Table 2 (sorption isotherms, Fig. 15S–18S†). The maximum polymer durability, while preserving a high sorption capacity.
sorption capacity decreases as the amount of cross-linker
To examine the influence of the nature of the cross-linker
increases. This is a combined consequence of the accessibility in more detail, polymers containing 20 mol% of different
of sorption-sites and the degree of polymer swelling. As cross-linkers were tested for glucose and fructose sorption.
expected, the higher the content of cross-linker, the lower the Results are summarized in Table 3 (sorption isotherms,
swelling degree of the polymer (Table 1).50 As noted for the Fig. 20S–25S†). Three series of cross-linkers as listed in Table 1
influence of the pre-treatment conditions, more readily-swelling were tested, namely, polar aliphatic (DVS, EDMA, TDMA), low-
polymers exhibit higher adsorption capacities. Additionally, the polarity aliphatic (OD, DD, TD), and aromatic (DVB, PDMA,
durability of the polymer, i.e. the capability of the polymer gran- DVN) cross-linkers. In general, the fructose capacity increases
ules to maintain the shape under adsorption conditions, plays a with increasing length of the cross-linker, while the glucose
crucial role for application. For example, the not cross-linked capacity does not vary significantly for polymers with different
polymer dissolves completely in the fructose solution cross-linkers (Fig. 3). Based on their commercial availability
(Fig. 19S†). Polymers with 5 mol% cross-linking give highly and a good compromise between high sorption capacity and
viscous liquids at high fructose concentrations, making it sufficient durability, the cross-linked polymers containing
impossible to separate the polymer from the solution. In these 20 mol% DVB or EDMA were identified as most promising can-
cases, the Langmuir isotherm is based only on sorption values didates and were therefore used in subsequent studies.
Table 3 Maximum fructose and glucose sorption capacities based on a Langmuir isotherm fit for polymers cross-linked with 20 mol% of different
cross-linkers
Fructose
Glucose
mmolFru g−1
molGlu molB
mmolGlu g−1
−1
−1
Entry
Cross-linker
molFru molB
1
2
3
4
5
6
7
8
9
DVS, 20 mol%
EDMA, 20 mol%
TDMA, 20 mol%
OD, 20 mol%
DD, 20 mol%
TD, 20 mol%
DVB, 20 mol%
DVN, 20 mol%
PDMA, 20 mol%
0.294a
0.618
0.819
0.472
0.550
0.722
0.854
0.947
0.879
1.652a
2.261
2.818
2.044
2.634
3.099
2.993
3.523
2.831
0.262
0.171
0.165
0.255
0.253
0.278
0.201
0.228
0.187
1.079
0.627
0.566
1.102
1.211
1.194
0.704
0.848
0.602
Conditions: 0.100 g 20 mol% cross-linked polymer, 2.63 mL carbonate buffer solution, 0.063–1.250 g fructose or glucose, 3 h, RT. a Fit based on
adsorption at low concentrations, since the highly viscous solution became impossible to filter at high concentrations.
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
adsorption is a self-sustaining process, due to repulsion
between negatively charged complexes. This description
explains why the maximum fructose adsorption capacity
depends significantly on parameters that influence swelling,
such as the pre-treatment method and the content and nature
of the cross-linker.
The dependence of the fructose sorption capacity on the
length of cross-linker is shown in Fig. 3a. Interestingly, the
trends are the same for the polar and low-polarity aliphatic
cross-linkers. Polymers containing aromatic cross-linkers
Fig. 3 Influence of the length of cross-linker on: (a) the maximum
fructose sorption capacity and (b) the maximum glucose sorption exhibit higher sorption capacities for the same length, likely
capacity.
due to the more rigid structure compared to the more flexible
aliphatic cross-linkers.51 This trend is most obvious when com-
paring the maximum sorption capacities of the aromatic cross-
Adsorption: a general discussion
linkers DVB, DVN, and PDMA. The maximum fructose capacity
increases with the length of the aromatic cross-linker from
DVB to DVN (Fig. 3a). However, a decrease in the maximum
fructose capacity occurs for PDMA, although the cross-linker is
longer. This decrease may be caused by the presence of polar
groups in PDMA (see Table 1 for structures). This explanation
is supported by the measurement of the swelling degree in
presence of fructose: the PDMA-containing polymer swells sig-
nificantly less compared to the polymers cross-linked with
DVB and DVN (Table 15S†).
The results for the adsorption of glucose and fructose can be
interpreted using the scheme in Fig. 4.
Boron-containing polymers are non-porous. After pre-treat-
ment with a basic solution, phenylboronic acid sites are partly
converted to phenylboronates, i.e., they become negatively
charged (a). Upon dispersion in carbonate buffer at pH 10, the
polymers swell. Pre-treatment with a phosphate buffer facili-
tates swelling, owing to electrostatic repulsion between the
phenylboronate sites and occluded HPO4
2−
anions.
Adsorption of glucose is also depicted schematically in
Fig. 4(f). The association constant for complexation with
phenyl-boronate (K = 110) is significantly lower than for fruc-
tose.18 As a result, the density of anionic complexes after
adsorption is much lower. Thus, the polymers do not swell
further, or swell insignificantly, upon interaction with glucose.
The swelling degree in the presence of glucose and fructose
was determined for the polymers cross-linked with 20 mol%
DVB or 20 mol% EDMA (Table 4). The amount of adsorbed
fructose and glucose is subtracted, thus the swelling degree is
representative of the pure polymers.
These results confirm the insignificant swelling in the pres-
ence of glucose compared to pure buffer, whereas swelling is
significantly enhanced in presence of fructose. Consequently,
most of the phenylboronate moieties remain inaccessible to
glucose. Since swelling does not take place during glucose
adsorption, the sorption capacity for glucose is largely inde-
pendent of the pre-treatment method used, the cross-linker
content, and the nature of the cross-linker. The maximum
sorption capacities in dependence of the length of cross-linker
are illustrated in Fig. 3b.
Complexation of phosphate by phenylboronate moieties
(Fig. 14S†) likely contributes to the improved swelling as well
(b). Depending on the nature of the adsorbate, two scenarios
are now possible. Fructose exhibits a high association constant
with phenylboronates (K = 4370).18 Therefore, a high density of
the fructose-phenylboronate complexes is expected (c). Since
these complexes are negatively charged, further swelling takes
place due to electrostatic repulsion between the polymer
chains (d). During swelling, the polymer chains unfold and
more phenylboronate moieties become available for adsorp-
tion. Eventually, the polymer is completely unfolded and all
phenylboronate sites are accessible for adsorption (e). Such
polymers exhibit their maximum capacity, corresponding to an
equimolar fructose to boron loading. However, if the swelling
is constrained, for example, by short cross-linkers or a high
density of cross-linkers, some fraction of the phenylboronate
sites remains inaccessible, i.e., case (d). Consequently, fructose
Table 4 Swelling degree in the presence and absence of glucose and
fructose for polymers cross-linked with 20 mol% DVB or 20 mol%
EDMA, measured in the carbonate buffer solution
Swelling degree [%]
Pure
buffer
In presence
of fructose
In presence
of glucose
Entry
Cross-linker
Fig. 4 Schematic representation of the adsorption of fructose and
glucose by phenylboronic acid-containing polymers.
1
2
20 mol% DVB
20 mol% EDMA
68
60
212
178
68
76
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Fig.
Paper
4
also rationalizes the kinetics of adsorption. three β-Frup monoboronates with signals at 114.7, 114.6, and
Homogeneous reactions between boronates and saccharides 113.4 ppm. The first two are the 2,3-exo and endo isomers
are typically complete within minutes.52 However, it takes (Fig. 5, 1a and 1b), respectively, whereas the last is a 2,3,6-tri-
roughly 3 h and 90 min, respectively, for the adsorption of dentate complex (Fig. 5, 2). The three monoboronates form pre-
fructose and glucose to reach the equilibrium. In the first few ferentially when the fructose-to-boron ratio is close to 1. Two
minutes, rapid uptake of the sugar is observed, followed by additional signals, at 103.4 and 103.0 ppm, are most likely
slower adsorption (Fig. 12S and 13S†). Apparently, the rate of diboronate ester isomers.55 Nicholls and Paul reported that
adsorption is not controlled solely by the complexation reac- diboronate esters can be formed at the 2,3- and 4,5-positions of
tion but also by the swelling of the polymer. Equilibration is the boat/twist conformers of β-Frup.56 Two phenyl rings can
faster for glucose than for fructose, since the polymers swell have either endo or exo orientations, resulting in a total of four
much less during adsorption of glucose.
diastereomers (Fig. 5, 3). The structures of the four possible
It is worth noting that the model proposed in Fig. 4 does diboronate esters are similar, and thus the corresponding reso-
not consider changes in the electronic structure of boron nances are not well-resolved. Both mono- and di-boronate esters
caused by the varying nature of the cross-linker. Such changes have been reported in the presence of organic diboronic acids.57
cannot be completely excluded, since differences are present Analysis of the different complexes formed with glucose, as well
in the 11B MAS NMR spectra of the materials (Fig. 11S†).
as time-resolved spectra showing the changes in concentration
of free and sorbed fructose and glucose can be found in the ESI
(Fig. 26S–29S†). According to the in situ NMR studies, the time
required to establish equilibria between dissolved and adsorbed
monosaccharides are 2 h for fructose and 1 h for glucose.
Structure of sugar-boronate complexes
For an in depth analysis, the composition of the sugar complexes
formed upon adsorption was studied by in situ 13C MAS NMR.53
To enhance the sensitivity and simplify the spectra, fructose
and glucose were selectively 13C-labelled at the anomeric
carbons, which coordinate to the boronate in all complexation
modes. For fructose, 2-13C-labelled fructose was used. Time-
resolved spectra and assignments of signals to free fructose
tautomers and fructose-boronate esters are shown in Fig. 5.
Three fructose tautomers are identifiable in the spectra:
β-fructopyranose (β-Frup), β-fructofuranose (β-Fruf ), and
α-fructofuranose (α-Fruf ), at 97.8, 101.2, and 104.2 ppm,
respectively. Concentrations of fructoketose (Fruk) and
α-fructopyranose (α-Frup), are below the detection limit. Over
time, the β-Frup signal declines in intensity while the concen-
trations of β-Fruf and α-Fruf are almost unchanged. In addition
to the fructose tautomers, five other resonances are present in
the spectra. They are assigned to fructose-boronate esters,
based on multiple literature reports.54,55 The signals of the
sorbed species are not significantly broadened compared to
those of the free fructose tautomers, suggesting that the fruc-
tose esters remain mobile. The most abundant species are
Desorption
In addition to sugar adsorption, the subsequent desorption is
important in the development of
a separation process.
Desorption is possible in acidic media, since the sugar complexes
are unstable at low pH. High desorption efficiency was observed
with 0.5 M aqueous H2SO4 acid, which we previously used for
extraction-assisted processes.35,58–60 Analogous to the adsorption
procedure, 20 vol% EtOH was generally added to the desorption
solutions to ensure polymer dispersion. Typical desorption
experiments were performed at RT for 4 h. There is no significant
influence of temperature (RT, 50, or 70 °C), and sugar concen-
tration reaches a stable value after 4 h (Tables 11S and 12S†).
Desorption efficiencies for various acidic solutions are
high, in the range 79 to 93% (Table 5). The nature of the acidic
Table 5 Fructose desorption efficiency in various acidic solutions, from
polymers cross-linked with 20 mol% EDMA. Repetitive desorption
(entries 9a–9c) was performed with the 20 mol% DVB cross-linked
polymer
Entry
Desorption solution
DE [%]
Overall DE [%]
1
2
3
4
5
6
7
8
0.5 M H2SO4 + 20 vol% EtOHa
0.5 M H2SO4 + 20 vol% EtOH
0.5 M H2SO4 + 20 vol% EtOHb
1 M formic acid
89
88
91
90
93
88
79
92
88
9
1 M acetic acid
1 M acetic acid + 20 vol% EtOH
1 M acetic acid/acetate buffer
CO2 desorption, 30 bar, 50 °C
0.5 M H2SO4 + 20 vol% EtOH
0.5 M H2SO4 + 20 vol% EtOHc
0.5 M H2SO4 + 20 vol% EtOHd
9a
9b
9c
97
99
2
Fig. 5 In situ array of direct polarization 13C MAS NMR spectra of fruc-
tose sorption on a 20 mol% DVB cross-linked phenylboronic acid
polymer, showing three fructose tautomers and four fructose-boronate
esters. Conditions: 80 : 20 v : v mixture of 0.5 M NaHCO3-buffer in D2O
(pH adjusted to 10 with 4 M NaOH) and EtOH, RT.
Conditions: 100 mg polymer, 2.63 mL desorption solution, 4 h, RT.
a Measured using half the amount of desorption solution (1.32 mL).
b Measured using double the amount of desorption solution (5.25 mL)
c The polymer used in 9a was reused for a 2nd desorption. d The
polymer used in 9b was reused for a 3rd desorption.
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
solution has no notable influence on the desorption efficiency. although resonances of Frup complexes are still visible (Fig. 6,
An advantage of the organic acids (formic and acetic acid) is inset 2). After three consecutive desorption runs, no reso-
that both can be used without EtOH addition, while maintain- nances corresponding to boronate-saccharide complexes are
ing sufficient polymer dispersion (Table 5, entries 4 and 5). present in the 13C CP/MAS NMR spectrum.
Fructose can be further desorbed into acidic solutions of
To examine the possibility of polymer recycling, consecutive
smaller volume (Table 5, entries 1–3). This finding suggests a adsorption–desorption cycles were performed under optimized
way to increase the final fructose concentration. Desorption conditions. Five repetitive adsorption–desorption cycles using
into an acetate buffer at pH 4.75 was also efficient, indicating 0.5 M H2SO4 and three recycling runs using pressurized CO2
that even a slightly acidic pH is sufficient to induce boronate for desorption were performed (Fig. 30S and 31S†). In each
ester cleavage (Table 5, entry 7). Most importantly, efficient de- cycle, the polymer exhibited around 90% of the sorption
sorption was observed into
a 80 : 20 v : v water : ethanol capacity of the previous one. This result indicates a rather high
mixture in an autoclave pressurized with CO2 (Table 5, durability of the polymer materials under adsorption and de-
entry 8). The pH after desorption was 6.8. Utilization of CO2 sorption conditions. Our current work focuses on investigation
presents an opportunity to improve the sustainability of the of long–term stability of the polymers in a continuous setup.
desorption process, yielding streams that are not contami- This will enable further insight into fouling mechanisms of
nated with molecular acids. Desorption of glucose and the polymers and required regeneration procedures. It is par-
saccharides from glucose–fructose mixtures with CO2 were ticularly noteworthy that no boron leaching was detected in
also examined. These experiments showed feasibility of the liquid phase upon recycling.
glucose desorption with CO2, though a moderate desorption
efficiency of 51% was observed (Table 14S†).
Adsorption-assisted isomerization
Table 5 shows that three consecutive desorption runs using
H2SO4 are necessary to completely desorb fructose (entries 9a–
c). The progress of desorption was studied by 13C CP/MAS
NMR. In addition to the resonances previously assigned to the
polymer network, signals in the range 65 to 120 ppm are
present after fructose adsorption. They are assigned to
adsorbed fructopyranose (α/β-Frup) and fructofuranose (α/
β-Fruf, Fig. 6, inset 1). The lines are broad since multiple com-
plexes with various configurations are present, as shown by
in situ 13C MAS NMR (see section Structure of sugar-boronate
complexes). Nevertheless, the ranges corresponding to reso-
nances of furanose and pyranose complexes are recognizable
(Fig. 6, insets 1 + 2). The specific assignments shown in both
insets are supported by the work from Norrild and Eggert, who
Finally, we explored the separation of glucose–fructose mix-
tures via polymer adsorption and the possibility of enhancing
the fructose yield by combining separation with glucose–fruc-
tose isomerization. Competitive sorption from a model solu-
tion containing
a 70 : 30 mol% : mol% glucose : fructose
mixture demonstrates the feasibility for selective fructose
recovery (Fig. 7). Adjusting the sugar-to-boron ratio has a
major influence on selective recovery. At a low sugar-to-
polymer ratio, free sorption sites remain available to bind
glucose, whereas a high sugar-to-polymer ratio leads to satur-
ation of the sorption sites with fructose.
The base-catalyzed isomerization of glucose to fructose
was studied in the carbonate buffer, which was also used as a
solvent for measuring sorption isotherms. The reaction is rela-
tively fast at moderate temperatures, and a high selectivity
towards fructose was observed. For example, a fructose yield of
24% was achieved at 25% glucose conversion after 1 h at 63 °C
(Fig. 32S and 33S†). Barker et al. proposed that in situ adsorp-
tion of fructose on boron-containing polymers improves the
fructose yield and facilitates product separation during
glucose isomerization.39 Using NaOH as the catalyst and a
1
employed JCC coupling constants to investigate p-tolylboronic
acid-fructose complexes.55 Interestingly, no signals corres-
ponding to Fruf were observed after the first desorption,
Fig. 6 13C CP/MAS NMR (24 kHz MAS, contact time 2.0 ms, 3072
scans) of the DVB-cross-linked polymer after adsorption and desorption.
The recovered polymer was filtered and dried under high vacuum prior
to analysis.
Fig. 7 Competitive adsorption from a 70 : 30 mol% : mol% glucose :
fructose model substrate mixture at varying total sugar concentrations.
Conditions: 100 mg 20 mol% cross-linked EDMA polymer, 2.63 mL car-
bonate buffer solution, 3 h, RT.
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
p-vinylphenylboronic acid polymer containing 5 mol% DVB, the polymer material as well as the process design to improve
they reported a maximum fructose yield of 57% in the pres- the isolated fructose yield, i.e., optimizing the glucose isomeri-
ence of the polymer. Therefore, we added a 20 mol% DVB zation conditions.
cross-linked polymer to a reaction mixture containing glucose
and carbonate buffer as catalyst. However, the isomerization
rate decreased significantly in the presence of the polymer, Conclusions
such that a maximum fructose yield of 27% at a glucose con-
This study described a method for fructose production via
adsorption-assisted isomerization of glucose. Cross-linked
version of 58% was achieved only after 12 h. The reduction in
reaction rate is consistent with glucose adsorption by the
poly(p-vinylphenylboronate) adsorbs fructose selectively from
polymer, which lowers the effective glucose concentration in
reaction mixtures after glucose isomerization, making use of
solution. Similar decreases in fructose yield and selectivity
the preferential complexation of fructose by boronate moieties.
were observed during attempts to reproduce the experiments
In order to reach high fructose loading, high boron content is
of Barker et al., and the reported yield of 57% was not
achieved. In the presence of NaOH or carbonate buffer, the
required. Our results show, however, that a compromise
between boron content and durability is needed. High load-
maximum yields of fructose were limited to 36% (at a fructose
−1
ings of ca. 1 molFru molB can be achieved for polymers with
selectivity of 50%) and 27% (at a fructose selectivity of 46%),
low cross-linker content, but these materials become viscous
respectively. Furthermore, the polymers became colored,
substances under adsorption conditions. Polymers containing
implying the adsorption of sugar degradation products
(Fig. 40S†).
20 mol% DVB or 20 mol% EDMA represent an attractive com-
promise, maintaining polymer durability and high sorption
Hence, we studied the adsorption-assisted isomerization of
capacities of 0.854 and 0.618 molFru molB−1, corresponding to
glucose in consecutive isomerization and adsorption experi-
540 and 407 mgFru gpolymer−1, respectively. Moreover, these
ments. Glucose was first isomerized in the carbonate buffer,
then the polymer was added to separate the fructose. The
residual solution was used in a second isomerization cycle, after
cross-linkers are commercially available. Ex situ and in situ
MAS NMR provided molecular insight into the kinetics of
sugar sorption as well as the nature of complex formation
adding more glucose to restore the initial total sugar concen-
between sugar and boronate. These insights will help to
tration. Eight consecutive isomerization-separation cycles were
further develop a highly efficient material for the separation of
performed. The results summarized in Fig. 8 demonstrate the
potential of the adsorption-assisted process. The maximum fruc-
fructose from glucose. Desorption of fructose was possible
using pressurized CO2, which represents a promising sustain-
tose yield almost doubled, from 29% in absence of the polymer
able alternative to the use of non-volatile mineral acids.
to 56% for the adsorption-assisted reaction (compare Fig. 8a
Overall, the adsorption-assisted approach is an attractive
and b), comparable to the yield reported by Barker et al.39
process with high potential for integration into value chains
Comparing the results to the previously reported extraction-
for the conversion of renewable feedstocks.
assisted process,35 a slight increase of fructose yield of 56%
compared to 51% is obtained during the adsorption-assisted
approach. Further advantages are the avoidance of an organic
co-extraction solvent as well as the necessity of adding quatern-
Conflicts of interest
ary amines as counter-cations for the stabilization of the mole-
cular boronate complex in the organic phase. The process is,
therefore, highly promising. We are now working on improving
There are no conflicts to declare.
Acknowledgements
We thank Simon Petring, Noah Avraham, Jens Heller, Heike
Bergstein and Ines Bachmann-Remy for HPLC, CHN, ICP-OES
and solution-state NMR analysis. We thank Christian Gierlich
for computing the lengths of the cross-linkers. We thank DFG
(Project 391305926) for financial support. G. S. acknowledges
the German Federal Environmental Foundation (DBU) and the
German Academic Exchange Service (DAAD) through its
Thematic Network “ACalNet” funded by the German Federal
Ministry of Education and Research (BMBF) for fellowship
support. As part of the ACalNet project, G. S. and S. L. S. thank
Fig. 8 (a) Adsorption-assisted isomerization: fructose yield after each
of eight consecutive isomerization-separation cycles, (b) dependence of
the fructose yield on the glucose conversion in glucose isomerization in
the absence of polymer. Conditions: for glucose isomerization: 10 mL of
the U.S. National Science Foundation for support under award
CBET-1512228. L. Q. is supported by the U.S. Department of
Energy, Office of Basic Energy Sciences, Division of Chemical
Sciences, Geosciences, and Biosciences through the Ames
Laboratory. The Ames Laboratory is operated for the U.S.
10 wt% glucose solution in carbonate buffer, 1 h at 63 °C; for adsorption:
6.57 mL isomerization solution added to 0.250 g polymer, 3 h at RT.
This journal is © The Royal Society of Chemistry 2019
Green Chem.

View Article Online
Paper
Green Chemistry
Department of Energy by Iowa State University under Contract 25 G. Griffin, Sep. Sci. Technol., 2005, 40, 2337–2351.
No. DE-AC02-07CH11358.
26 B. R. Caes, R. E. Teixeira, K. G. Knapp and R. T. Raines,
ACS Sustainable Chem. Eng., 2015, 3, 2591–2605.
27 B. Li, S. Varanasi and P. Relue, Green Chem., 2013, 15,
2149–2157.
28 G. J. Griffin and L. Shu, J. Chem. Technol. Biotechnol., 2004,
79, 505–511.
29 B. Li, P. Relue and S. Varanasi, Green Chem., 2012, 14,
2436–2444.
30 H. A. Aziz, A. Kamaruddin and M. A. Bakar, Sep. Purif.
Technol., 2008, 60, 190–197.
Notes and references
1 S. E. Jacobsen and C. E. Wyman, Twenty-First Symposium on
Biotechnology for Fuels and Chemicals, 2000, pp. 81–96.
2 I. Delidovich, K. Leonhard and R. Palkovits, Energy Environ.
Sci., 2014, 7, 2803–2830.
3 Y.-B. Huang and Y. Fu, Green Chem., 2013, 15, 1095–1111.
4 I. Delidovich and R. Palkovits, ChemSusChem, 2016, 9, 547– 31 S. Alipour, Green Chem., 2016, 18, 4990–4998.
561.
32 S. Alipour and H. Omidvarborna, RSC Adv., 2016, 6,
111616–111621.
33 R. R. Broekhuis, S. Lynn and C. J. King, Ind. Eng. Chem.
Res., 1996, 35, 1206–1214.
34 S. S. Gori, M. V. R. Raju, D. A. Fonseca, J. Satyavolu,
C. T. Burns and M. H. Nantz, ACS Sustainable Chem. Eng.,
2015, 3, 2452–2457.
35 I. Delidovich and R. Palkovits, Green Chem., 2016, 18, 5822–
5830.
5 L. M. Hanover and J. S. White, Am. J. Clin. Nutr., 1993, 58,
724S–732S.
6 J. M. Carraher, C. N. Fleitman and J.-P. Tessonnier, ACS
Catal., 2015, 5, 3162–3173.
7 Y. B. Tewari, Biotechnol. Appl. Biochem., 1990, 23, 187–203.
8 A. A. Rosatella, S. P. Simeonov, R. F. Frade and
C. A. Afonso, Green Chem., 2011, 13, 754–793.
9 J. J. Bozell and G. R. Petersen, Green Chem., 2010, 12, 539–
554.
36 P. J. Duggan, Aust. J. Chem., 2004, 57, 291–299.
10 D. W. Rackemann and W. O. Doherty, Biofuels, Bioprod. 37 S. J. Gardiner, B. D. Smith, P. J. Duggan, M. J. Karpa and
Biorefin., 2011, 5, 198–214. G. J. Griffin, Tetrahedron, 1999, 55, 2857–2864.
11 K.-U. Klatt, F. Hanisch and G. Dünnebier, J. Process Control, 38 M. Di Luccio, B. Smith, T. Kida, C. Borges and T. Alves,
2002, 12, 203–219. J. Membr. Sci., 2000, 174, 217–224.
12 H. Li, S. Yang, S. Saravanamurugan and A. Riisager, ACS 39 S. Barker, B. Hatt, P. Somers and R. Woodbury, Carbohydr.
Catal., 2017, 7, 3010–3029. Res., 1973, 26, 55–64.
13 S. Ramaswamy, H.-J. Huang and B. V. Ramarao, Separation 40 A. K. Hoffmann and W. Thomas, J. Am. Chem. Soc., 1959,
and purification technologies in biorefineries, John Wiley &
81, 580–582.
Sons, 2013.
41 R. Boulos and J. Feutrill, U.S. Patent Application, 15/
748,514, 2018.
42 V. Davankov, M. Tsyurupa and S. Rogozhin, Angew.
Makromol. Chem., 1976, 53, 19–27.
43 Matlab 2018a and Global Optimization Toolbox, The
MathWorks, Inc., Natick Massachusetts, United States,
2018.
14 J. A. Vente, H. Bosch, A. B. de Haan and P. J. Bussmann,
Adsorpt. Sci. Technol., 2006, 24, 771–780.
15 G. Springsteen and B. Wang, Chem. Commun., 2001, 1608–
1609.
16 G. Springsteen and B. Wang, Tetrahedron, 2002, 58, 5291–
5300.
17 J. Yan, G. Springsteen, S. Deeter and B. Wang, Tetrahedron, 44 J. W. Wackerly and J. F. Dunne, J. Chem. Educ., 2017, 94,
2004, 60, 11205–11209. 1790–1793.
18 J. P. Lorand and J. O. Edwards, J. Org. Chem., 1959, 24, 769– 45 H. Nöth and B. Wrackmeyer, Nuclear magnetic resonance
774.
spectroscopy of boron compounds, Springer Science
Business Media, 2012.
46 J. A. Peters, Coord. Chem. Rev., 2014, 268, 1–22.
&
19 S. Shinkai, K. Tsukagoshi, Y. Ishikawa and T. Kunitake,
J. Chem. Soc., Chem. Commun., 1991, 1039–1041.
20 D. C. Klenk, G. T. Hermanson, R. I. Krohn, E. K. Fujimoto, 47 R. M. C. Dawson, D. C. Elliott, W. H. Elliott and
A. K. Mallia, P. Smith, J. England, H. Wiedmeyer, R. Little
K. M. Jones, Data for biochemical research, Clarendon Press,
and D. Goldstein, Clin. Chem., 1982, 28, 2088–2094.
Oxford, UK, 1986.
21 T. R. Jackson, J. S. Springall, D. Rogalle, N. Masumoto, 48 L. Bosch, T. Fyles and T. D. James, Tetrahedron, 2004, 60,
H. C. Li, F. D’Hooge, S. P. Perera, T. A. Jenkins, T. D. James
and J. S. Fossey, Electrophoresis, 2008, 29, 4185–4191.
22 Y.-H. Zhao and D. F. Shantz, Langmuir, 2011, 27, 14554–
14562.
23 T. C. Brennan, S. Datta, H. W. Blanch, B. A. Simmons and
B. M. Holmes, Bioenergy Res., 2010, 3, 123–133.
24 J. Shi, J. M. Gladden, N. Sathitsuksanoh, P. Kambam,
11175–11190.
49 S. T. Dubas and J. B. Schlenoff, Langmuir, 2001, 17, 7725–
7727.
50 G. Wulff, J. Vietmeier and H. G. Poll, Angew. Makromol.
Chem., 1987, 188, 731–740.
51 M. Takayanagi, T. Ogata, M. Morikawa and T. Kai,
J. Macromol. Sci., Part B: Phys., 1980, 17, 591–615.
L. Sandoval, D. Mitra, S. Zhang, A. George, S. W. Singer and 52 W. A. Marinaro, R. Prankerd, K. Kinnari and V. J. Stella,
B. A. Simmons, Green Chem., 2013, 15, 2579–2589. J. Pharm. Sci., 2015, 104, 1399–1408.
Green Chem.
This journal is © The Royal Society of Chemistry 2019

View Article Online
Green Chemistry
Paper
53 L. Qi, R. Alamillo, W. A. Elliott, A. Andersen, 57 S. P. Draffin, P. J. Duggan, S. A. Duggan and J. C. Norrild,
D. W. Hoyt, E. D. Walter, K. S. Han, N. M. Washton, Tetrahedron, 2003, 59, 9075–9082.
R. M. Rioux and J. A. Dumesic, ACS Catal., 2017, 7, 3489– 58 P. Drabo, T. Tiso, B. Heyman, E. Sarikaya, P. Gaspar,
3500.
J. Förster, J. Büchs, L. M. Blank and I. Delidovich,
ChemSusChem, 2017, 10, 3252–3259.
59 I. Delidovich, M. S. Gyngazova, N. Sánchez-Bastardo,
J. P. Wohland, C. Hoppe and P. Drabo, Green Chem., 2018,
20, 724–734.
54 R. van den Berg, J. A. Peters and H. van Bekkum,
Carbohydr. Res., 1994, 253, 1–12.
55 J. C. Norrild and H. Eggert, J. Chem. Soc., Perkin Trans. 2,
1996, 2583–2588.
56 M. P. Nicholls and P. K. Paul, Org. Biomol. Chem., 2004, 2, 60 N. Sánchez-Bastardo, I. Delidovich and E. Alonso, ACS
1434–1441.
Sustainable Chem. Eng., 2018, 6, 11930–11938.
This journal is © The Royal Society of Chemistry 2019
Green Chem.