Acid monolayer functionalized iron oxide nanoparticles as catalysts for carbohydrate hydrolysis

Article abstract of DOI:10.1039/c3gc41420e

Superparamagnetic iron oxide nanoparticles were functionalized with a quasi-monolayer of 11-sulfoundecanoic acid and 10-phosphono-1-decanesulfonic acid ligands to create separable solid acid catalysts. The ligands are bound through carboxylate or phosphonate bonds to the magnetite core. The ligand-core bonding surface is separated by a hydrocarbon linker from an outer surface with exposed sulfonic acid groups. The more tightly packed monolayer of the phosphonate ligand corresponded to a higher sulfonic acid loading by weight, a reduced agglomeration of particles, a greater tendency to remain suspended in solution in the presence of an external magnetic field, and a higher catalytic activity per sulfonic acid group. The particles were characterized by thermogravimetric analysis (TGA), transmission electron microscopy (TEM), potentiometric titration, diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), inductively coupled plasma optical emission spectrometry (ICP-OES), and dynamic light scattering (DLS). In sucrose catalysis reactions, the phosphonic-sulfonic nanoparticles (PSNPs) were seen to be incompletely recovered by an external magnetic field, while the carboxylic-sulfonic nanoparticles (CSNPs) showed a trend of increasing activity over the first four recycle runs. The activity of the acid-functionalized nanoparticles was compared to the traditional solid acid catalyst Amberlyst-15 for the hydrolysis of starch in aqueous solution. Catalytic activity for starch hydrolysis was in the order PSNPs > CSNPs > Amberlyst-15. Monolayer acid functionalization of iron oxides presents a novel strategy for the development of recyclable solid acid catalysts.

Full text of DOI:10.1039/c3gc41420e

Green Chemistry
View Article Online
View Journal | View Issue
PAPER
Acid monolayer functionalized iron oxide
nanoparticles as catalysts for carbohydrate
hydrolysis†
Cite this: Green Chem., 2014, 16,
836
Myles Ikenberry,a Leidy Peña,b Daming Wei,a Hongwang Wang,c
Stefan H. Bossmann,c Trenton Wilke,a Donghai Wang,b Venugopal R. Komreddy,d
D. Paul Rillemad and Keith L. Hohn*a
Superparamagnetic iron oxide nanoparticles were functionalized with a quasi-monolayer of 11-sulfo-
undecanoic acid and 10-phosphono-1-decanesulfonic acid ligands to create separable solid acid cata-
lysts. The ligands are bound through carboxylate or phosphonate bonds to the magnetite core. The
ligand-core bonding surface is separated by a hydrocarbon linker from an outer surface with exposed sul-
fonic acid groups. The more tightly packed monolayer of the phosphonate ligand corresponded to a
higher sulfonic acid loading by weight, a reduced agglomeration of particles, a greater tendency to
remain suspended in solution in the presence of an external magnetic field, and a higher catalytic activity
per sulfonic acid group. The particles were characterized by thermogravimetric analysis (TGA), trans-
mission electron microscopy (TEM), potentiometric titration, diffuse reflectance infrared Fourier transform
spectroscopy (DRIFTS), inductively coupled plasma optical emission spectrometry (ICP-OES), and
dynamic light scattering (DLS). In sucrose catalysis reactions, the phosphonic–sulfonic nanoparticles
(PSNPs) were seen to be incompletely recovered by an external magnetic field, while the carboxylic–
sulfonic nanoparticles (CSNPs) showed a trend of increasing activity over the first four recycle runs. The
activity of the acid-functionalized nanoparticles was compared to the traditional solid acid catalyst
Amberlyst-15 for the hydrolysis of starch in aqueous solution. Catalytic activity for starch hydrolysis was in
the order PSNPs > CSNPs > Amberlyst-15. Monolayer acid functionalization of iron oxides presents a
novel strategy for the development of recyclable solid acid catalysts.
Received 18th July 2013,
Accepted 8th November 2013
DOI: 10.1039/c3gc41420e
www.rsc.org/greenchem
development of manufacturing processes which inexpensively
achieve the necessary material outputs without degrading the
Introduction
Oil resources are non-renewable, and the expanding econo- environment. Current industrial processes, mainly involving
mies of the world will require renewable material sources and petrochemicals, consume more than 15 million tons of sulfu-
processing methods if economic development is to be sustain- ric acid every year as a non-recyclable catalyst.2 Replacing sul-
able. Fortunately, the world’s annual production of biomass is furic acid with reusable acid catalysts would be a major
greater than 170 billion metric tons, and three fourths of that achievement towards the goal of sustainable, environmentally-
material is carbohydrates, which are thought to be a potential friendly chemical production.
feedstock for a sustainable world chemical industry in the
Biofuel production often requires the same acid and base
future.1 One key to this sustainable chemical market is the catalysts that are used in other industrial processes, signifi-
cantly increasing the production costs and environmental
impacts of such commodities.3 For example, the hydrolysis of
cellulose into glucose is efficiently catalyzed by sulfuric acid,
aDepartment of Chemical Engineering, Kansas State University, 1016 Durland Hall,
Manhattan, KS 66506-5102, USA. E-mail: hohn@ksu.edu; Fax: +1-785-532-7372;
but the cost associated with waste separation and disposal has
Tel: +1-785-532-4315
prevented an economically efficient large-scale saccharification
of cellulosic materials.4 A heterogeneous catalyst capable of
efficiently catalyzing cellulose hydrolysis at mild conditions
could reduce the costs associated with waste disposal and
enable large-scale utilization of cellulosic biofuels. Typical
solid acid catalysts include sulfonated silicas, zeolites, hetero-
polyacids, and ion-exchange resins.2,5 Stability in water at
bDepartment of Biological and Agricultural Engineering, Kansas State University,
Manhattan, KS, USA
cDepartment of Chemistry, Kansas State University, Manhattan, KS, USA
dDepartment of Chemistry, Wichita State University, Wichita, KS 67260-0051206,
USA
†Electronic supplementary information (ESI) available: Complete reaction
details for 10-phosphono-1-sulfonodecanic acid, and starch hydrolysis table. See
DOI: 10.1039/c3gc41420e
836 | Green Chem., 2014, 16, 836–843
This journal is © The Royal Society of Chemistry 2014

View Article Online
Green Chemistry
Paper
elevated temperatures remains an important challenge. A
recent addition to the family of solid acid catalysts is the
material composed of superparamagnetic iron oxide cores,
typically surrounded with an organic polymer or silica
shell.6–12 Superparamagnetism provides a facile method of
separation which offers these solid acids an advantage over
those which require centrifugation or filtration.12 However, at
temperatures and aqueous conditions relevant to biomass con-
version, the magnetite–silica–sulfonic acid composites require
regeneration with sulfuric acid,11,12 or lose acid groups.9 Mag-
netite–polymer–sulfonic acid composites also lose acid groups,
and require a high degree of crosslinking to protect the mag-
netic nanoparticle cores.10 This paper details a new approach
for preparing an acid functionalized superparamagnetic solid
acid catalyst: attaching molecules containing acid groups
directly to the iron oxide surface in a monolayer. The mono-
layer approach to acid functionalization of nanoparticles may
have the potential to surpass polymerization and silica function-
alization by producing more active, stable, and tunable acid
catalysts for application in green chemical processes.
One way to attach acidic ligands is to choose ligands that
have two acids groups: one that can bind to a metal oxide,
such as magnetite, and one that will eventually serve as the cata-
lytic site. Portet et al.13 attached di-sulfonic acids to magnetite
leaving one sulfonic acid group exposed to solution.
However, using a sulfonic acid to bind to magnetite is not
optimal due to the sulfonic acid group’s weak binding affinity
toward the magnetite surface.13–15 The binding strength of
typical functional groups to a magnetite surface is given by
the order: phosphonic > carboxylic > hydroxyl > sulfonic.15,16
While phosphonic acids form the strongest bonds to magnet-
ite, they have been associated with substantial decreases in
magnetic properties,17 although there is some disagreement.18
Fig. 1 A diagram of the CSNP material composed of a magnetite core,
carboxylate–magnetite bonds, an alkane layer, and an outer surface with
exposed sulfonic acid groups.
functionalized nanoparticles is demonstrated by following
their activity over multiple uses.
Experimental
Synthesis of 11-sulfoundecanoic acid and 10-phosphono-
1-decanesulfonic acid
This paper demonstrates that a single ligand can contain a 11-Sulfoundecanoic acid was synthesized from 11-bromoun-
functional group for binding to the nanoparticle surface and decanoic acid and sodium sulfite according to a previously
an additional catalytically active sulfonic acid group facing published procedure.19 The 11-bromoundecanoic acid was
outward into solution. This approach takes advantage of differ- combined with sodium sulfite in a 10 : 1 ratio and the reagents
ences in the binding affinity of magnetite for different func- were heated at 50 °C for 1 hour, 70 °C for 2 hours, 80 °C for
tional groups. In this research, carboxylic/phosphonic acids 6 hours, then 90 °C for 3 hours. The product was separated
attach to the magnetic particle and a sulfonic acid group from solution by first lowering the pH to 5 with HCl and filter-
extends away from the particle toward the surrounding solu- ing under vacuum at 50 °C, then placing the filtrate in an oven
tion. The sulfonic acid group is, therefore, available for for one week at 80 °C to fully oxidize the sodium sulfite impur-
catalysis. The hydrocarbon chains separating the carboxylic/ ity salt into sodium sulfate. The baked precipitate was then
phosphonic groups from the sulfonic acid group have hydro- placed in concentrated hydrochloric acid solution and centri-
phobic ligand–ligand interactions which are thought to reduce fuged to precipitate the purified ligand and leave the remain-
core–solvent interactions and provide some shielding of the ing sodium sulfate and sodium bromide in the acid solution.
carboxylic/phosphonic acid–magnetite bonds from chemical Full oxidation of the sodium sulfite is necessary to ensure that
attack and subsequent ligand dissociation. The structure of hydrochloric acid addition does not produce dangerous sulfur
the functionalized particles is indicated in Fig. 1.
trioxide gas. The centrifugation was repeated three times.
The activity of these acid-functionalized nanoparticles is Then, the precipitated ligand was vacuum dried overnight at
demonstrated through their use in two acid-catalyzed reac- room temperature to remove excess HCl.
tions: hydrolysis of sucrose to glucose and fructose, and hydro-
10-Phosphono-1-sulfonodecanic acid was synthesized as
lysis of starch to glucose.5 The activity of magnetite illustrated in Fig. 2.† Briefly, the intermediate compound 1,10-
nanoparticles bound to carboxylic and phosphonic acid dibromodecane was synthesized from 1,10-butanediol by
ligands are compared with that of a common solid acid cata- heating in hydrobromic acid at 100 °C for 48 hours. After the
lyst, Amberlyst-15. Furthermore, the reusability of the acid- reaction, it was extracted with chloroform, washed with brine,
This journal is © The Royal Society of Chemistry 2014
Green Chem., 2014, 16, 836–843 | 837

View Article Online
Paper
Green Chemistry
agglomerates. Increasing the concentration of re-suspended
particles did not increase the measured size of particle
agglomerates, possibly indicating that individual particles
were present in solution, instead of agglomerates. The com-
bined supernatant–particle mixture was then subjected to cen-
trifugation (10 000 rpm for 1 hour at 4 °C). The supernatant
was discarded, and the centrifuged solid was then re-
suspended in DI water and vigorously agitated for 5 minutes.
The centrifugation–washing was repeated twice more.
Washing the CSNP particles five times gave a suspension
which was not fully separable by the supermagnet. After a few
days of sitting at room temperature, the particles had agglo-
merated to the stir bar magnet, leaving a clear supernatant.
They were washed another 5 times by supermagnetic particle
separation to further remove unbound ligand. The PSNPs and
CSNPs were not dried after washing. Instead, a washed catalyst
solution of known catalyst dry weight concentration was added
to DI water to obtain reaction solutions.
Fig. 2 Sequence of steps20–22 used to prepare 10-phosphono-
1-decanesulfonic acid (4).
stirred over sodium bicarbonate, then filtered and passed
through Na2SO4. The obtained compound was purified
by column chromatography on silica gel, yielding 1,10-dibro-
modecane as white crystals. The 1,10-dibromodecane was
heated to 155 °C in an oil bath under argon, then triethylpho-
sphite was added dropwise over 8 h. Cooling to room tempera-
ture produced a white precipitate, which was separated by
filtration and purified by flash column chromatography.
Diethyl(10-bromodecyl)phosphonate was then heated in a satu-
rated solution of Na2SO3 at reflux for one hour then at 120 °C
until the solvent evaporated. The residue was then slurred with
CH3OH under reflux and the hot solution was filtered. The fil-
trated was then evaporated and the residue acidified with
hydrobromic acid and refluxed for 48 hours. Evaporation of the
acid under reduced pressure produced a pale-black powder.
Catalyst characterization
To calculate the acid loading of all catalysts, 0.05 grams (dry
wt) of wet catalyst solution were titrated with 0.01 M NaOH
solution. For TGA, the samples were analyzed in a PerkinElmer
Pyris1 TGA (Norwalk, CT). Approximately 5 mg of each sample
was heated under a nitrogen atmosphere from 50 °C to 600 °C
at a rate of 10 °C min−1. Transmission electron microscope
(TEM) images were used to estimate the size distribution of
the nanoparticles. A model CM100 TEM (FEI Company, Hills-
boro, OR), equipped with an AMT digital image capturing
system, was operated at 100 kV. The images were taken of dis-
Nanoparticle synthesis and functionalization
Iron oxide nanoparticle cores were prepared using a modifi- persed particles deposited from aqueous solutions by contact-
cation of Massart’s precipitation method.23,24 Iron(II) chloride ing a drop of liquid with Formvar/carbon-coated, 200-mesh
and iron(III) chloride in a 1 : 2 molar ratio were precipitated copper grids (Electron Microscopy Sciences, Fort Washington,
with ammonium hydroxide at 80 °C, heated for 30 minutes, PA) for 30 seconds at room temperature. Nanoparticle mean
and then sonicated for 20 minutes. 1 g of the synthesized di- diameters were calculated from the TEM images using the soft-
acid ligand was added to 1 g of the sonicated particles, and ware ImageJ, available from the National Institute of Health.
the mixture was sonicated for 10 minutes, then heated to Particle agglomerate data was taken on a ZetaPALS Zeta Poten-
95 °C for 30 minutes. The 95 °C particles were sonicated for an tial Analyzer (Brookhaven Instruments Corporation) with
additional 10 minutes, then returned to the oil bath. After hydrodynamic light scattering.
24 hours of stirring at 95 °C, the solution was allowed to
XRD analysis was conducted using a Rigaku MiniFlex II
return to room temperature under constant stirring for desktop X-ray diffractometer, using a fixed time scanning
12 hours, then reheated to 95 °C and stirred for another method with a 0.01° step size and 6 seconds counting time at
24 hours, and then again allowed to return to room each step. The phase composition of the nanoparticles was
temperature.
estimated using the relative intensity ratio method within
Washing at room temperature was achieved by repeatedly Rigaku’s PDXL software. Crystallite sizes were calculated using
mixing the nanoparticles with water and magnetically extract- the Scherrer formula. The crystallite size is used as an approxi-
ing the particles from solution. The washwater was initially mation of the magnetite particle size, as described by other
clear and then on subsequent washes became darker as the investigators.23 Diffuse reflectance infrared Fourier transform
particles became increasingly difficult to separate from solu- spectroscopy (DRIFTS) was conducted on a Thermo Nicolet
tion. As the concentration of unbound ligand in the super- NEXUS 870 spectrometer with a ZnSe window, operated by the
natant decreased during the washing process, the supernatant software program OMNIC. Samples were dried at 50 °C for
did not become clear in the presence of the neodymium super- 48 hours. KBr was used for a background scan, then immedi-
magnet (with field strength of 3661 Gauss).
ately added to a spec of dried sample, ground up by mortar
After 5 washes, the PSNP supernatant, which was not clear, and pestle, and returned to the cell for analysis. ICP-OES was
was measured by DLS as 50 nm agglomerates. The separable conducted with a Varian 720-ES ICP Optical Emission Spectro-
particles, when re-suspended, also measured as 50 nm meter. Standards were prepared from stock solutions of Fe, S,
838 | Green Chem., 2014, 16, 836–843
This journal is © The Royal Society of Chemistry 2014

View Article Online
Green Chemistry
Paper
and P, and used for calibration. The lines selected for each
element were Fe 259.94 nm, P 178.29 nm, S 181.97 nm.
Catalytic activity measurement
Sucrose hydrolysis was conducted by 80 °C for 6 hours in a
sealed vial with magnetic stirring, using 0.05 g of catalyst and
0.1 g of sucrose in 5 g of water. Reaction temperature was con-
trolled using an oil bath held at constant temperature, with
reactor temperature periodically checked with a temperature
probe. After the completion of each run, the catalyst was separ-
ated either by supermagnet or by centrifugation. Fresh sucrose
and 5 g of water were placed in the reaction vessel, which was
returned to the oil bath for the next round of reaction. To test
the reproducibility of the sucrose experiments, three runs were
conducted on three different CSNP samples. Control runs with
no catalyst present showed no conversion. Catalytic activity is
measured in terms of sucrose conversion to the two hydrolysis
products glucose and fructose. The mass balance indicated
that fructose formed by-products which are less than 3% of
the total mass.
For starch hydrolysis, 0.002 g (dry wt) of PSNPs, 0.017 g (dry
wt) of CSNPs, or 0.0006 g Amberlyst-15 were added with 0.09 g
starch and 5 g of water so as to yield an acid site : starch :
water mol ratio of 1 : 200 : 100 000 for Amberlyst-15 and
1 : 400 : 200 000 for the CSNPs and PSNPs. In a sealed glass
tube with magnetic stirring at 300 rpm, the mixtures were
placed in an oil bath at 130 °C for 24 hours. CSNPs and PSNPs
were both separated by centrifugation at 10 000 rpm for 1 hour
at 4 °C. The centrifuged catalyst was combined with 5 g of
water and 0.09 g of starch and returned to the reaction vessel
for the next run. To estimate the variability of the experimental
setup, 3 runs with fresh Amberlyst-15 were run, and a standard
deviation of 3.7% was found. Control runs with no catalyst
present showed no conversion. Catalytic activity is measured
in terms of glucose turnover frequency per acid site. Oligomers
were not measured. A table including maltose (dimer) pro-
duction is included in the ESI.†
Fig. 3 DRIFTS spectra of the (a) CSNPs, (b) 11-sulfoundecanoic acid, (c)
NaOH neutralized PSNPs and (d) 10-phosphono-1-decanesulfonic acid.
1556 cm−1 which may be due to sodium sulfonate. The
sodium in the CSNPs may originate from sodium impurity in
the synthesized 11-sulfoundecanoic acid, while sodium in the
PSNPs comes from the addition of NaOH to a limited supply
of catalyst. For CSNPs, an overlapping peak at 1544 cm−1 is
due to asymmetric COO− stretching. The CSNP peak at
1626 cm−1 corresponds to COO− symmetric stretch. The dis-
tance between the asymmetric and symmetric COO− peaks
being less than 110 cm−1 is an indication of chelating biden-
tate attachment25 or chelating mononuclear attachment.26
When carboxylic acid COOH groups are uncoordinated, a peak
around 1710 cm−1 is present.27,28 The peak for free carboxylic
acid is observed in the 11-sulfoundecanoic acid. The lack of
this peak in the spectrum of the CSNPs indicates a lack of free
COOH. The alkane chain vibrations appear at 2852 and
2921 cm−1 for PSNPs, with CSNPs showing corresponding
peaks at 2850 and 2918 cm−1. The NP alkane vibrations are at
lower wavenumbers than the corresponding free acids, indicat-
ing confinement and ordered packing on the particle sur-
faces.29 The ligands and NPs all had a broad OH vibration
band between 3200 and 3600 cm−1 (not shown).
Sugar analysis was conducted using an HPLC RCM-Ca+2
monosaccharide column (300
× 7.8 mm; Phenomenex,
Torrance, CA) with a refractive index detector. The HPLC was
calibrated to external standards for sucrose, glucose, fructose,
and maltose, using 80 °C deionized water at a flow rate of
0.6 mL per minute. During HPLC analysis, samples were alter-
nated with known standards to ensure accuracy of the
measurements.
TGA results for both PSNPs and CSNPs are shown in Fig. 4.
By 200 °C, the CSNPs had lost about 1% of its weight, and the
Results and discussion
Catalyst characterization
DRIFTS spectra of the ligands, CSNPs, and NaOH neutralized
−
−
PSNPs are shown in Fig. 3. RSO3 and RPO3 groups appeared
around 1050 and 1180 cm−1 due to asymmetric and symmetric
vibrations, respectively.17 Both NPs have a peak at 1467 cm−1
corresponding to the CH2 scissoring bend.17 CSNPs, PSNPs,
Fig. 4 TGA graph of catalyst weight % vs. the temperature (°C). The
and the 11-sulfoundecanoic acid show
a
peak around total weight loss is (a) 5.6% for the CSNPs and (b) 18.7% for the PSNPs.
This journal is © The Royal Society of Chemistry 2014
Green Chem., 2014, 16, 836–843 | 839

View Article Online
Paper
Green Chemistry
PSNPs had lost about 2%. Some of the weight loss may be due
to water. For CSNPs, the total weight loss was 5.6% by weight.
If this entire weight loss is ascribed to ligand desorption and
decomposition, it corresponds to a sulfur content of 0.69%, or
0.2 mmol g−1 sulfur in the CSNPs. That is higher than the acid
loading of 0.08 mmol H+ per gram as determined by titration
of the CSNPs. The discrepancy between titration and TGA may
indicate that the measured acid loading is reduced by cationic
impurities, or that agglomeration hinders solution access to
some acid sites. As seen in the CSNP sucrose reaction data, the
acid sites become more active during initial reaction cycles.
For the PSNPs, the weight loss during TGA was 18.7%, and the
titrated acid loading was 0.65 mmol H+ per gram. An acid
loading of 0.65 mmol H+ per gram corresponds to a ligand
and sulfur loading of 19.7% and 2.1% by weight, respectively,
so the TGA measurement of 18.7% is roughly in agreement
with the titration. Impurities and agglomeration might also
affect the measurement of PSNP acid sites, but be counter-
acted to some extent by incomplete removal of the ligand
layer. During TGA of functionalized iron oxide nanoparticles,
it is thought that some carbon remains on the particle surface
due to incomplete decomposition of the organic ligands, and
the formation of graphitic species.26 Incomplete decompo-
sition may be more significant for the higher ligand loading of
the PSNPs.
We can further understand how the acid-containing ligands
pack on the surface by using particle surface area to estimate
the packing density. The packing density of the ligand layer on
the particle surface is important for many reasons. A more
tightly packed full monolayer will obviously result in a higher
acid site density on the outer surface and a higher ligand
loading by weight. Additionally, a tightly packed ligand layer
may be more likely to protect the magnetite layer from oxi-
dation and a corresponding reduction in magnetic properties.
Prevention of magnetite degradation, along with the loss of
acid groups, is a primary concern in the design of acid function-
alized iron oxide nanoparticles.
Fig. 5 XRD for (a) PSNPs after 18 hours of starch hydrolysis at 130 °C,
(b) PSNPs as synthesized, (c) CSNPs after 18 h starch hydrolysis at
130 °C, (d) CSNPs as synthesized and (e) bare particles. Lines denote
goethite peaks.
coverage is significantly less than monolayer for the CSNPs,
with TGA putting the ligand loading at 5.7% and the titration
indicating only 0.08 mmol H+ per gram. If the lower than
expected titration measurement is due to restriction of acid
sites within agglomerates, then the TGA value may be a better
indicator of the ligand packing density. According to this
analysis, the CSNPs have approximately 1/3 of a monolayer.
However, this packing density is likely an average over signifi-
cantly different packing densities for different particle sizes, as
indicated by other work with oleic acid and magnetite nano-
particles with monodisperse diameters.25
X-ray diffraction results (Fig. 5) indicate significant differ-
ences in the iron oxide phase compositions of CSNPs and
PSNPs. Magnetite and maghemite are present in both CSNPs
and PSNPs, with PSNPs showing substantially higher mag-
hemite content. Additionally, the goethite phase of iron oxide
is present in the PSNPs but not in the CSNPs or bare particles,
as demonstrated by the XRD peaks at 21.18°, 33.20°, 39.93°,
41.16°, 50.65°, 53.16°, 58.98°, and 61.29°, which are marked
in Fig. 5. Carboxylic acids and other ligands are thought
to rearrange the surfaces of metal oxide nanoparticles.31,32
This does not necessarily mean a degradation of magnetic
properties.18,33
The Scherrer equation was used to calculate an average
PSNP diameter of 10.9 nm from the XRD data, which corre-
sponds to a specific surface area of 107 m2 g−1 for a constant
magnetite density of 5.18 g cm−3. Some researchers believe
that phosphonates bind in a tridentate fashion30 but others
suggest only bidentate binding.17 For a ligand footprint of
0.24 nm2, there is an 18% weight loading of 10-phosphono-
1-decanesulfonic acid, and 0.60 mmol H+ per gram. The
measured TGA ligand loading of 18.7% and titration loading
of 0.65 mmol H+ per gram are close to what is expected,
Catalysis of starch and sucrose hydrolysis
suggesting a complete monolayer with a ligand footprint of The conversion of sucrose for three different CSNP samples
0.24 nm2.
over multiple consecutive runs is shown in Fig. 6. The CSNP
For CSNPs, the XRD indicates an average particle size of catalytic activity increases through the first 4 runs, using cen-
11.5 nm, which corresponds to specific surface area trifugation for separation of particles between runs. After the
a
of 100 m2 g−1. The slightly smaller carboxylate footprint of 4th cycle, the catalyst was refrigerated and stored for one week
0.21 nm2 and molecular weight of 266.4 g mol−1 for 11-sul- before being used again in recycle reactions 5–9, with mag-
foundecanoic acid correspond to a 17% ligand weight loading netic particle separation. Sucrose conversion for trials 5–9 is
for monolayer coverage, with a sulfonic acid loading of relatively constant at ∼60%. As seen in Fig. 6, the increasing
0.66 mmol H+ per gram. TGA and titration indicate that the activity over the first four recycles was noted for three different
840 | Green Chem., 2014, 16, 836–843
This journal is © The Royal Society of Chemistry 2014

View Article Online
Green Chemistry
Paper
Fig. 6 Sucrose hydrolysis at 80 °C using 0.05 g catalyst, 0.1 g sucrose,
and 5 g water.
Fig. 8 Sucrose hydrolysis at 80 °C using 0.05 g catalyst, 0.1 g sucrose,
and 5 g water.
CSNP samples, providing strong evidence that this was a real
trend. The much higher acid loading of Amberlyst-15 results in
a higher conversion for sucrose hydrolysis for equal catalyst
weight. However, the CSNPs and PSNPs have a higher turnover
frequency per acid site, possibly due to the location of acid
groups on the outer surface of the material. The turnover fre-
quency for CSNPs (9th run), PSNPs (17th run) and Amberlyst-15
are 7.1, 0.64, and 0.20 mol sucrose converted per mol H+ per
hour, respectively, based on initial acid loadings. The higher
turnover frequency of acid sites is apparent in the starch
hydrolysis, which compares equal numbers of acid sites
instead of equal catalyst weights.
An increase in catalytic activity during recycle runs is
unusual. One possible explanation for this trend is that some
ammonium may have remained coordinated to the sulfonate
groups after synthesis and become detached during the cata-
lytic runs. Another possibility is that changes in agglomerate
size may impact the catalytic activity. In aqueous solutions of
functionalized iron oxide nanoparticles, agglomeration often
decreases at higher temperatures.34
Fig. 9 6 Hour sucrose reaction supernatants with PSNPs remaining in
suspension in the presence of a powerful magnetic field.
primary particles leaves only a single non-agglomerated par-
ticle to be seen in the TEM image. In the image on the right,
6 hours of sucrose hydrolysis at 80 °C has caused the agglo-
meration to decrease significantly.
The CSNP de-agglomeration during sucrose reactions was
also measured with DLS, by adding drops of hot reaction solu-
tion to a heated cell containing 70 °C sucrose solution. These
conditions were chosen to minimize agglomerate changes
during the DLS measurement of the reaction mixture. DLS
measurements of the CSNPs taken at the beginning and end
of a 6 hour sucrose reaction indicated that the initial effective
agglomerate size of 3000–4000 nm was reduced to less
than 1000 nm by the end of a reaction cycle. The necessary
dilution for DLS measurement may have caused some
de-agglomeration.
The de-agglomeration of CSNPs was observed in TEM
images, as shown in Fig. 7. On the left, the agglomeration of
The PSNP catalytic activity for sucrose hydrolysis is shown
in Fig. 8 for 17 consecutive 6-hour runs. Catalyst separation
was achieved by using an external magnetic field in runs 1–13
and by centrifugation in runs 14–17. As seen in Fig. 8, conver-
sion is between 40 and 50% for the first four runs, but then
shows a decrease to less than 20% by the 13th run. This
decrease is due to the loss of magnetic particles in the re-
cycling process. PSNPs are quite stable in solution, so a certain
amount is not able to be extracted even by the high powered
magnet. Fig. 9 is a photograph of the supernatants following
reaction and removal of PSNPs with the magnet. The darker
colors observed are an indication that some PSNPs remain in
the supernatant.
To understand whether the loss in activity noted over 13
runs was entirely due to the loss of PSNP particles or whether
some other deactivation was occurring, the supernatants col-
lected in the first 13 runs were centrifuged to recover more of
the suspended catalyst for reuse. The recovered catalyst was
combined with the catalyst that had been magnetically recov-
ered successfully. The recombined catalyst was then used in
Fig. 7 TEM images of CSNPs deposited (a) at room temperature in pure
water and (b) from an aqueous 80 °C sucrose solution after a 6 hour
reaction.
This journal is © The Royal Society of Chemistry 2014
Green Chem., 2014, 16, 836–843 | 841

View Article Online
Paper
Green Chemistry
further reactions. As seen in Fig. 8, PSNP conversion returned
to over 40% for runs 15–17, indicating that the catalyst had
not been deactivated.
At this time, it is not possible to eliminate the detachment
of ligands as partially responsible for the lower conversion
noted for recycle runs 5–13. The phosphonate–magnetite bond
is expected to be quite strong, although the ICP data may indi-
cate ligand loss from the monolayer of some particles. For
runs 15–17, centrifugation was used to recycle the PSNPs.
A conversion above 40% for all recycle runs indicates that cen-
trifugation is superior to magnetic separation when recycling
the PSNPs. This might be an indication that the magnetic
character of the magnetite core is damaged by the attachment
of the phosphonate ligands. This is supported by the XRD data
that shows the presence of a new iron phase goethite
(α-FeOOH) in addition to substantially increased maghemite
(Fe2O3), after phosphonate ligand attachment.
The PSNPs are not easily separated from solution. Centrifu-
gation is more effective, but neither centrifugation nor
magnetic field is able to achieve complete separation of par-
ticles from supernatants. This may be due to a combination of
factors such as increased hydrophilic character from the
increased number of sulfonic acid groups, increased hydro-
philic character due to the more highly packed and less
exposed lipid layer, decreased agglomerate sizes resulting in
smaller effective magnetic force per particle, and possibly
decreased particle magnetic strength. PSNPs titrated to a pH of
7 gave distinct TEM images: individual particles averaged
5.2 nm, and agglomerates of larger particles averaged 12.9 nm.
ICP was used to gain some information about the atomic
composition of the supernatants after catalyst removal. These
data (shown in Table 1) indicate a ratio of iron to phosphorus
that is around 2 : 1 for the combined supernatants of runs
1–13, and around 4 : 1 for the individual supernatants of runs
14–16. For a monolayer of 10-phosphono-1-decanesulfonic
acid calculated using a ligand head area of 0.24 nm2, the Fe : P
ratio varies from 15 to 58, for 5 nm and 20 nm particles,
respectively.
Fig. 10 TOF [mol glucose (mol H+)−1 h−1] for three runs of 24 hour
starch hydrolysis at 130 °C.
seen for the CSNPs may be present in the PSNPs as well, but
counteracted by the loss of excess ligand. The issue of incom-
plete catalyst recovery in the PSNP sucrose runs may compli-
cate an accurate assessment of the catalytic activity.
To compare the activity of CSNPs vs. PSNPs, the starch
hydrolysis data may be more useful since centrifugation is
used to maximize catalyst recovery after each reaction. Acid
site turnover frequency (moles of glucose formed per mole of
acid per hour) for PSNPs, CSNPs, and Amberlyst-15 is shown
in Fig. 10. The PSNPs show the highest activity for starch
hydrolysis, surpassing the CSNPs and Amberlyst-15. The
increased ligand loading achieved using the phosphonic acid
attachment apparently also results in an outer surface whose
acid sites are individually more active. Unfortunately, there is a
trade-off between the benefits and downsides of a more highly
hydrophilic outer surface. Even in centrifugation, the PSNPs
leave the supernatant visibly darkened with the presence of
particles. The presence of particles was again confirmed with
ICP data showing the presence of iron, sulfur, and phosphorus
in the post-centrifugation supernatants. The downside of
decreased catalytic recovery is a significant one. If larger par-
ticles are more easily recovered, the lower acid loading by
weight may be a reasonable trade-off. Further research should
investigate monodisperse particles for activity and recycle
potential.
For both the CSNPs and PSNPs, the activity seen is an
average over particles of significantly different radii and
surface curvature. Beyond the immediate geometrical trends,
the smaller particles are likely to behave differently from larger
ones. For example, smaller iron oxide nanoparticles have been
seen to form more tightly packed monolayers than larger par-
ticles.25 This may compound the effect of increased surface
area and further increase the weight loading of ligand on
smaller particles. However, since phosphonic acids bind
more strongly than carboxylic acids to magnetite, the
monolayer density as a function of particle size may be less
pronounced in PSNPs than in CSNPs. Another interesting
effect of the nanoparticle curvature is the tendency to push
acid groups closer together at larger particle sizes.35 It
is unclear whether this effect increases acid site activity36 or
decreases it.35
The incomplete separation of particles from solution, com-
bined with the correlation between ligand loading and particle
size, makes the interpretation of ICP data difficult. A shift in
Fe : P ratio can reflect not only a change in the relative
amounts of free ligand and particles, but also in the diameter
distribution of particles which remain suspended. The ICP
data seems to indicate that some excess ligand is lost,
although this does not result in a decrease in catalytic activity
after the initial 4 runs. The trend of catalytic activity increase
Table 1 ICP data for the inseparable catalyst remaining in solution after
6 hour sucrose runs
Run
Fe (ppm)
P (ppm)
S (ppm)
1 to 13
14
15
20
19
47
26
9
5
10
6
9
5
12
8
16
842 | Green Chem., 2014, 16, 836–843
This journal is © The Royal Society of Chemistry 2014

View Article Online
Green Chemistry
Paper
15 R. Cornell and U. Schwertmann, The iron oxides: Structure,
properties, reactions, occurrences, and uses, Wiley-VCH,
2nd edn, 2003.
Conclusions
Monolayer acid functionalized iron oxide nanoparticles have
the potential for facile recyclability, high acid loading, high 16 Y. Sahoo, H. Pizem, T. Fried, D. Golodnitsky, L. Burstein,
acid site catalytic activity, and stability in aqueous solutions at
elevated temperatures. CSNPs have been shown to possess an
C. N. Sukenik and G. Markovich, Langmuir, 2001, 17, 7907–
7911.
interesting trend of increasing catalytic activity through initial 17 C. Yee, G. Kataby, A. Ulman, T. Prozorov, H. White, A. King,
catalytic reactions, likely because of deagglomeration of par-
ticles at the high temperatures of reaction. PSNPs have higher
M. Rafailovich, J. Sokolov and A. Gedanken, Langmuir,
1999, 15, 7111–7115.
acid loading and acid site activity compared to CSNPs, but 18 T. J. Daou, J. M. Grenèche, G. Pourroy, S. Buathong,
lower recyclability. The analysis of PSNP data is complicated by
incomplete catalyst recovery and variations in ligand loading
a. Derory, C. Ulhaq-Bouillet, B. Donnio, D. Guillon and
S. Begin-Colin, Chem. Mater., 2008, 20, 5869–5875.
between various particle sizes. Future work should explore 19 J. Cui, D. Cao and C. Xu, Chin. J. Synth. Chem., 2006, 14,
functionalizations of monodisperse iron oxide particles to 529–531.
evaluate the optimal particle size which results from the 20 S. De, V. K. Aswal and S. Ramakrishnan, Langmuir, 2010,
tradeoff between recyclability and acid site loading/activity.
26, 17882–9.
21 X. Wang and M. Lieberman, Langmuir, 2003, 19, 7346–
7353.
22 D. Taffa, M. Kathiresan and L. Walder, Langmuir, 2009, 25,
5371–5379.
23 Y. Sahoo, A. Goodarzi, M. T. Swihart, T. Y. Ohulchanskyy,
N. Kaur, E. P. Furlani and P. N. Prasad, J. Phys. Chem. B,
2005, 109, 3879–3885.
Notes and references
1 A. Corma, S. Iborra and A. Velty, Chem. Rev., 2007, 107,
2411–2502.
2 M. Hara, T. Yoshida, A. Takagaki, T. Takata, J. N. Kondo,
S. Hayashi and K. Domen, Angew. Chem., Int. Ed., 2004, 43,
2955–2958.
24 R. Massart, IEEE Trans. Magn., 1981, 17, 1247–1248.
3 F. Carvalheiro, L. C. Duarte and F. M. Gírio, J. Sci. Ind. Res., 25 L. Zhang, R. He and H.-C. Gu, Appl. Surf. Sci., 2006, 253,
2008, 67, 849–864. 2611–2617.
4 S. Van de Vyver, L. Peng, J. Geboers, H. Schepers, F. de 26 P. Roonasi and A. Holmgren, Appl. Surf. Sci., 2009, 255,
Clippel, C. J. Gommes, B. Goderis, P. a. Jacobs and
B. F. Sels, Green Chem., 2010, 12, 1560.
5 P. L. Dhepe, M. Ohashi, S. Inagaki, M. Ichikawa and
A. Fukuoka, Catal. Lett., 2005, 102, 163–169.
6 N. T. S. Phan and C. W. Jones, J. Mol. Catal. A: Chem., 2006,
253, 123–131.
7 C. Gill, B. Price and C. Jones, J. Catal., 2007, 251, 145–152.
8 J. Lee, Y. Lee, J. K. Youn, H. Bin Na, T. Yu, H. Kim,
5891–5895.
27 K. V. P. M. Shafi, A. Ulman, X. Yan, N. Yang and
C. Estourne, Langmuir, 2001, 17, 5093–5097.
28 K. Yang, H. Peng, Y. Wen and N. Li, Appl. Surf. Sci., 2010,
256, 3093–3097.
29 K. Nakamoto, Infrared and Raman Spectra of Inorganic and
Coordination Compounds, John Wiley & Sons, New York,
1997.
S.-M. Lee, Y.-M. Koo, J. H. Kwak, H. G. Park, H. N. Chang, 30 S. Mohapatra and P. Pramanik, Colloids Surf., A, 2009, 339,
M. Hwang, J.-G. Park, J. Kim and T. Hyeon, Small, 2008, 4,
143–152.
9 L. Peña, M. Ikenberry, B. Ware, K. L. Hohn, D. Boyle,
35–42.
31 T. Rajh, L. X. Chen, K. Lukas, T. Liu, M. C. Thurnauer and
D. M. Tiede, J. Phys. Chem. B, 2002, 106, 10543–10552.
X. S. Sun and D. Wang, Biotechnol. Bioprocess Eng., 2011, 32 S. Laurent, D. Forge, M. Port, A. Roch, C. Robic, L. Vander
16, 1214–1222.
Elst and R. N. Muller, Chem. Rev., 2008, 108, 2064–
2110.
10 M. Feyen, C. Weidenthaler, F. Schüth and A.-H. Lu, Chem.
Mater., 2010, 22, 2955–2961.
11 D. Lai, L. Deng, J. Li, B. Liao, Q. Guo and Y. Fu, Chem-
SusChem, 2011, 4, 55–58.
12 D. Lai, L. Deng, Q. Guo and Y. Fu, Energy Environ. Sci.,
2011, 4, 3552.
33 H. Duan, M. Kuang, X. Wang, Y. A. Wang, H. Mao and
S. Nie, J. Phys. Chem. C, 2008, 112, 8127–8131.
34 H. Kikura, J. Matsushita, M. Matsuzaki, Y. Kobayashi
and M. Aritomi, Sci. Technol. Adv. Mater., 2004, 5, 703–
707.
13 D. Portet, B. Denizot, E. Rump, J.-J. Lejeune and P. Jallet, 35 D. Wang, R. J. Nap, I. Lagzi, B. Kowalczyk, S. Han,
J. Colloid Interface Sci., 2001, 238, 37–42.
14 L. Vekasa, D. Bica and M. Avdeev, China Particuol., 2007, 5,
43–49.
B. A. Grzybowski and I. Szleifer, J. Am. Chem. Soc., 2011,
133, 2192–2197.
36 I. Mbaraka and B. Shanks, J. Catal., 2006, 244, 78–85.
This journal is © The Royal Society of Chemistry 2014
Green Chem., 2014, 16, 836–843 | 843