Characterization of titanium dioxide nanoparticles dispersed in organic ligand solutions by using a diffusion-ordered spectroscopy-based strategy

Article abstract of DOI:10.1002/chem.200601722

Diffusion-ordered NMR spectroscopy (DOSY NMR) is presented as a tool for the determination of the diffusion coefficients of organic ligands in suspensions of titanium dioxide nanoparticles. The nanoparticles were prepared by a sol-gel process by hydrolysis and condensation reactions of titanium tetra-n-butoxide in the presence of pentane-2,4-dione (acacH: acetylacetone), as well as para-toluenesulfonic acid (pTsA) and n-butanol (nBuOH). NMR spectroscopic studies were performed in various deuterated solvents, on both dispersed xerosols and diluted sols. The bipolar-pulsed field gradient longitudinal eddy-current delay (LED) pulse sequence was used for data acquisition. The data were processed by inverse Laplace transformation (ILT), by using a maximum entropy algorithm, to afford 2D DOSY spectra. Different diffusion regimes for organic ligands in the bound and unbound states were successfully discriminated, more particularly in [D₃]acetonitrile, thus allowing assessment of their interactions with the nanoparticles.

Full text of DOI:10.1002/chem.200601722

FULL PAPER
DOI: 10.1002/chem.200601722
Characterization of Titanium Dioxide Nanoparticles Dispersed in Organic
Ligand Solutions by Using a Diffusion-Ordered Spectroscopy-Based Strategy
[b]
Luk Van Lokeren,[a] GØraldine Maheut,[a] FranÅoisRibot, Virginie Escax,[a, b]
Ingrid Verbruggen,[a] ClØment Sanchez,[b] JosØ C. Martins,[c] Monique Biesemans,*[a] and
Rudolph Willem[a]
Abstract: Diffusion-ordered NMR
spectroscopy (DOSY NMR) is present-
ed as a tool for the determination of
the diffusion coefficients of organic li-
gands in suspensions of titanium diox-
ide nanoparticles. The nanoparticles
were prepared by a sol-gel process by
hydrolysis and condensation reactions
of titanium tetra-n-butoxide in the
presence of pentane-2,4-dione (acacH:
acetylacetone), as well as para-toluene-
sulfonic acid (pTsA) and n-butanol
(nBuOH). NMR spectroscopic studies
were performed in various deuterated
solvents, on both dispersed xerosols
and diluted sols. The bipolar-pulsed
field gradient longitudinal eddy-current
delay (LED) pulse sequence was used
for data acquisition. The data were
processed by inverse Laplace transfor-
mation (ILT), by using a maximum en-
tropy algorithm, to afford 2D DOSY
spectra. Different diffusion regimes for
organic ligands in the bound and un-
bound states were successfully discrimi-
nated,
more
particularly
in
[D3]acetonitrile, thus allowing assess-
ment of their interactions with the
nanoparticles.
Keywords: diffusion
· ligand ef-
fects · nanostructures · NMR spec-
troscopy · titanium dioxide
Introduction
ligands remaining on the isolated solid. Both approaches re-
quire tedious separation protocols (centrifugation, ultrafil-
tration, etc.), which can be invalidated if the exchange rate
of the ligands is faster than the timescale of the separation
process. This hampers the ability to gain insight into the
actual nature of the interaction phenomena occurring at the
surface of the particle. Various standard methods, for exam-
ple, dynamic light scattering (DLS), enable one to deter-
mine the size distribution of particles suspended in solution,
but cannot address the issue of the structural nature of the
ligand–nanoparticle interactions. The development of new
tools to characterize dispersed nanoparticles and to under-
stand these interactions remains a major issue. Pulsed field
gradient NMR spectroscopy is an in situ methodology which
enables both the resolution and quantification of complex
mixtures of different species in solution on the sole basis of
their differentiated translational mobility.[1,2] The use of two-
dimensional (2D) NMR experiments, which combine a
chemical shift scale and a diffusion coefficient scale in so-
called 2D diffusion-ordered (DOSY) NMR spectra, pro-
vides a straightforward assignment of the diffusion correla-
tion peaks associated with the various molecular species
present in the mixture. Therefore, this analytical tool is com-
plementary to conventional separation techniques,[3] for in-
stance, size exclusion chromatography (SEC), in which the
Measuring in situ the affinity between an organic ligand and
an inorganic metal-oxide nanoparticle remains a difficult
task. Generally, this affinity is measured from adsorption
isotherms, quantifying either the organic ligands remaining
in the solution freed from the nanoparticles or the organic
[a] L. Van Lokeren, Dr. G. Maheut, Dr. V. Escax, I. Verbruggen,
Prof. Dr. M. Biesemans, Prof. Dr. R. Willem
High Resolution NMR Centre (HNMR)
Vrije Universiteit Brussel
Pleinlaan 2, 1050 Brussel (Belgium)
Fax : (+32)2-629-3291
E-mail: mbiesema@vub.ac.be
[b] Dr. F. Ribot, Dr. V. Escax, Dr. C. Sanchez
Laboratoire de Chimie de la Mati›re CondensØe de Paris
UniversitØ Pierre et Marie Curie - Paris6, CNRS UMR 7574
Tour 54, 5e Øtage, 4 Place Jussieu
75252 Paris Cedex 05 (France)
[c] Prof. Dr. J. C. Martins
NMR and Structural Analysis Unit
Vakgroep Organische Chemie, Universiteit Gent
Krijgslaan 281 S4, 9000 Gent (Belgium)
Supporting information for this article is available on the WWW
under http://www.chemeurj.org/ or from the author.
Chem. Eur. J. 2007, 13, 6957 – 6966
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
6957

D ¼ kBT=fT
ð1Þ
components of the mixture are separated chemically rather
than spectroscopically. It is likewise of major interest for
mixtures of soluble or dispersible compounds (surfactants),[4]
and products with reduced translational mobility (dendrim-
ers,[5] polymers,[6] and macromolecules[7]) whenever species
of different sizes or molar masses are to be distinguished.[8]
Another advantage of this methodology is its potential to in-
vestigate interactions between various species in solution
and to gain insight into their mutual clustering (for example,
organic tin-oxide based clusters) as well as their affinity to
the solvent.[9,10] More recently, DOSY NMR was explored
as a tool for investigating to what extent measuring diffusion
coefficients enables ligands grafted to the surface of parti-
cles to be distinguished from unbound ones in solution
(CeO2, InP).[11,12] In the case of InP, kinetic aspects of the
ligand exchanges involved in association/dissociation equili-
bria were likewise investigated.[12]
in which kB is the Boltzmann constant, T the temperature in
Kelvin and fT the friction factor. Assuming the diffusing
structural unit has a spherical shape, Equation (1) can be re-
written as the well-known Stokes–Einstein equation
[Eq. (2)][16]
:
D ¼ kBT=cphrH
ð2Þ
in which h is the viscosity of the solvent, rH is the hydrody-
namic radius of the structural unit and c is a parameter that
goes asymptotically to six as the hydrodynamic radius in-
creases, which is generally achieved for structures with a di-
ameter of 1 nm or more.[17]
Equation (2) is convenient to estimate the molecular radii
of species in solution from diffusion coefficient measure-
ments. Given that structural units generally do not have a
spherical shape, the hydrodynamic radius rH deduced from a
diffusion coefficient measurement by using Equation (2)
should simply be considered as the radius of a fictitious,
spherical structural unit, which diffuses with the same trans-
lational rate as the structural unit under investigation.
In a first stage, the DOSY measurements were performed
on model solutions prepared from tetramethylsilane (TMS)
and only a single component involved in the sol or xerosol,
namely, acacH, nBuOH, or pTsA in a given deuterated sol-
vent. Subsequently, DOSY experiments were conducted for
the three types of samples of interest—blank, sol S3, and
Sanchez and co-workers described the synthesis of solu-
ble,[13] size-tailored nanoparticles of titanium dioxide (ana-
tase)[14] by hydrolysis–condensation reactions of titanium
tetra-n-butoxide in the presence of pentane-2,4-dionate
(acac: acetylacetonate) and p-toluenesulfonic acid (pTsA) in
wet n-butanol (nBuOH). It is well known that, in solution,
acetylacetone (acacH) exists as two distinct species, a
ketone and an enol in mutual tautomeric equilibrium. In the
case of TiO2, only the deprotonated enolic tautomer can act
as a bidentate complexing agent for the nanoparticle. This
system appeared to be a particularly relevant and experi-
mentally suitable model for translational mobility investiga-
tions on nanoparticles by DOSY NMR, given that TiO2
enjoys a wide range of applications and its properties are
well described and easily controllable as far as the particle
size is concerned. We focused more particularly on the so-
called S3 sol featuring nanoparticles with diameters ranging
from approximately 2 to 6 nm, the size of which depends on
the sample preparation conditions, as revealed previously by
DLS.[13] The xerogel was obtained by drying this sol, and
was thoroughly investigated by DOSY NMR. We believed
A
sion coefficients of the species in these model solutions with
those in the blank solutions, in which all the components are
present apart from the nanoparticles under identical condi-
tions of concentration, provides valuable information as to
their possible mutual interactions, independent of their pos-
sible interactions with the nanoparticles.
It is convenient to circumvent changes of viscosity from
one sample to another by defining a reduced diffusion coef-
ficient Dj(S) as the ratio of the diffusion coefficient of the
species j in sample S of interest to that of TMS in the same
sample [Eq. (3)]. TMS is a nonpolar and nonpolarizable spe-
cies, and is therefore presumed to have diffusive properties,
which are sensitive only to the viscosity of the sample and
not to any other interaction (j: acacH, nBuOH, pTsA;
S: models, blank, S3, X3):[18]
1
that 2D DOSY H NMR spectra would enable us to address
ligand-exchange phenomena taking place at the surface of
the titanium dioxide nanoparticles and possible interactions
between the reagents present in the saturated “sol”
medium, as well as in the re-dispersed xerosol.[15] In order
not to overlook any possible interaction between reagents
and/or solvent molecules, DOSY NMR experiments were
also performed on so-called blanks, prepared under exactly
the same conditions as the above-mentioned sols except that
titanium tetra-n-butoxide was substituted for pure nBuOH.
Herein, we present the successful achievement of these
goals.
DjðSÞ ¼ DjðSÞ=DTMSðSÞ
ð3Þ
The quantity Dj(S) thus provides information on the sensitivi-
ty of the diffusion of species j in sample S to the polarity of
the medium, as well as on other possible interactions in
which the species j might be involved.
Results and Discussion
The timescale to separate fast from slow exchange is dif-
ferent in DOSY NMR compared with regular 1D NMR
spectroscopy. In DOSY the timescale to be considered is the
diffusion time used in between the coding and decoding gra-
Methodological considerations: In solution, only translation-
al motions enable one to determine the diffusion coefficient
by using the Debye–Einstein equation [Eq. (1)]:
6958
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 6957 – 6966

Characterization of Nanoparticles
FULL PAPER
Table 1. Diffusion coefficients[a] [10À10 m2 sÀ1] and reduced diffusion coefficients, in [D3]acetonitrile, for the
pure compounds (model solutions), blank sample, and redispersed xerosol X3. The relative fractions are also
given for species in different states (for example, free and bound).
dient pulses (that is, about
100–200 ms in the experiments
reported here), while for
1D NMR spectroscopy the
timescale corresponds to the
inverse of the frequency differ-
ence in between the two chem-
ical shifts of the species con-
Compounds
Model solutions
Blank
Dj(blank)
X3
Dj(X3)
Dj(model)
Dj(model)
Dj(blank)
Dj(X3)
%
free
hindered/free
bound
free acacH
bound acac
free acacH
free
15.9
0.55
0.97
10.2
0.43
pTsA
9.7
4.3
29.0
3.4
27.5
30.4
36.0
0.27
0.12
0.81
0.09
0.76
0.84
1.00
50
50
85
29.1
24.1
1.02
enolic 60%
15
cerned (that is, for
0.5 ppm difference
a
d=
ketonic 40%
nBuOH
TMS
27.2
28.7
29.8
0.91
0.94
1.00
21.7
20.6
23.6
0.92
0.87
1.00
100
100
100
at
250 MHz, the characteristic
time for the coalescence is
about 8 ms).
free
[a] The relative accuracy of the diffusion coefficients is Æ1% for models, Æ1% for the blank, and Æ10% for
X3, and the relative accuracy of the fractions determined from DOSY NMR is Æ10% (see the Experimental
Section).
Evidence for chemical reac-
tionsin the blank : Apart from
the expected diffusion correlation peaks from the species of
interest (pTsA, acacH/acac, and nBuOH), the 2D DOSY
1H NMR spectrum of the blank solution shows some addi-
tional peaks. In the singlet methyl area of the blank solution
in CD3CN, an additional unexpected diffusion correlation
peak is observed with a chemical shift (d=2.10 ppm) be-
tween those of the enolic (d=2.01 ppm) and ketonic (d=
2.15 ppm) tautomers of acacH, which arises from a species
not observed in any of the spectra of the pure components.
This additional compound appears to be acetone generated
as a degradation product upon mixing the different compo-
nents of the blank solutions. This byproduct is obtained
most likely by decomposition of acacH by a retro-Claisen
reaction. This reaction also produces acetic acid, which
reacts further with free nBuOH to yield n-butyl acetate, of
which only the a-CH2 butyl resonances are well resolved at
d=4.02 ppm. The resonance of the methyl moiety of the
acetyl group and the other n-butyl resonances overlap at
250 MHz with the resonances of the residual solvent (d=
the change of diffusion behavior of each species when
switching from model to blank is mostly related to the vis-
cosity change. The enolic acacH tautomer (Dblank =24.1”
10À10 m2 sÀ1) diffuses slightly faster than the ketonic one
(Dblank =21.7”10À10 m2 sÀ1) in the blank, which appears al-
ready to be the case in model solutions. As compared to the
others, the pTsA species diffuses markedly slower than TMS
in the model solution. The size of acacH and pTsA being
similar,[19] the slower diffusion of pTsA is likely to be related
to aggregation phenomena through dipole–dipole interac-
tions, p stacking, and/or hydrogen bonds. The presence of
acacH and nBuOH in the blank slows down further, albeit
slightly, the translational diffusion of pTsA, which is not sur-
prising in view of the hydrogen bridge donor capacities of
acacH and nBuOH. That the diffusion coefficients of the
latter two substances remain roughly unaffected by the pres-
ence of pTsA appears acceptable, as both acacH and
nBuOH are present in large molar excess with respect to
pTsA (acacH/pTsA 15:1, nBuOH/pTsA 50:1), which, ac-
cordingly, has to be much more sensitive to intermolecular
interactions from other species.
1
1.94 ppm) and the nBuOH. However, a H NMR spectrum
acquired at 700 MHz clearly shows the resonances of the
acetyl group at d=1.97 ppm and of the b-CH2 at d=
1.58 ppm. The g-CH2 and the methyl resonances of n-butyl
acetate are also visible in the spectrum, but only partially as
shoulders of the corresponding nBuOH resonances. Finally,
it also appears that acetone is partially reduced further into
isopropanol, as revealed by the two characteristic multiplet
resonances at d=1.09 and 3.86 ppm.
Some additional unidentified, variably broad resonances,
which appear with poorly reproducible chemical shifts in the
spectrum, coalesce away upon addition of [D4]methanol, in-
dicating that these resonances originate from chemically ex-
changeable protons, mostly hydroxylic ones.
The 1D 1H NMR spectra of the redispersed xerosol X3
and of the sol S3 in [D3]acetonitrile display novel resonances
not observed in the blank (Figure 1). The observed AA’XX’
line pattern for the aromatic protons of pTsA, appearing as
two doublets with reasonably sharp resonances in the blank,
is somewhat broadened in X3 and S3. It is accompanied by
additional very broad resonances at the low (X3) and the
high (S3) frequency sides. These additional broad bands, ex-
tending well over d=0.2 ppm, are assigned to the pTsA li-
gands interacting directly with TiO2 nanoparticles, as evi-
denced from their slower translational diffusion measured
by 2D DOSY (Figure 2). The resulting slower rotational re-
orientation rate of the ligands interacting directly (or indi-
rectly) with the nanoparticles explains this dramatic line
broadening. The relative displacement of the broad bands
with respect to the narrower doublets, from X3 to S3, is most
likely related to the presence of nBuOH in large excess in
S3.
DOSY NMR investigations in [D3]acetonitrile: The diffu-
sion coefficients of the different species, as obtained from
the model solutions of the pure organic compounds, the
blank and the redispersed xerosol X3 in CD3CN, are collect-
ed in Table 1. Here, the data indicate that, apart from pTsA,
all other species display rather similar reduced diffusion co-
efficients in the model solutions and in the blank. Therefore,
The overlapping and broadened resonances in the aromat-
ic, as well as in the methyl, spectral range of pTsA in both
Chem. Eur. J. 2007, 13, 6957 – 6966
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
6959

M. Biesemans et al.
Figure 3. Aromatic and methyl proton resonance ranges of the pTsA
ligand in the 2D DOSY spectrum of xerosol X3 redispersed in
[D3]acetonitrile, with projections along the chemical shift and diffusion
axes. Contour levels and projections are not identical in the left (10”)
and right (1”) panels, for the sake of optimal representation.
Figure 4. Vinylic spectral range of the enolic tautomer (left panel) and
methylene spectral range of the ketonic tautomer (right panel) of
acac(H) in the 2D DOSY spectrum of the xerosol X3 redispersed in
[D3]acetonitrile, with projections along the chemical shift and diffusion
axes.
Figure 1. 1D 1H NMR spectra of model solutions; a) pTsA, b) acacH,
c) nBuOH, d) blank, e) S3, and f) X3 in [D3]acetonitrile.
Figure 5. Spectral expansions in the 2D DOSY spectrum of the xerosol
X3 redispersed in [D3]acetonitrile, with projections along the chemical
shift and diffusion axes, in the resonance ranges of nBuOH (a), ketonic
acacH (b), and n-butyl acetate (c), a byproduct generated in situ. The n-
butyl acetate is clearly visible with low-intensity diffusion correlation
multiplets around d=1.59 (b-CH2 (n-butyl acetate), right panel) and
4.02 ppm (a-CH2 (n-butyl acetate), left panel); the correlation peaks in
the range d=1.25–1.55 ppm are assigned to b-CH2 (nBuOH) and g-CH2
(nBuOH, n-butyl acetate) moieties (right panel). The diffusion correla-
tion multiplets around d=3.49 (left panel) and 0.9 ppm (right panel) cor-
respond to the a-CH2 (nBuOH) and CH3 (nBuOH, n-butyl acetate) moi-
eties, respectively. Contour levels and projections are not identical in the
left (2”) and right (1”) panels, for the sake of optimal representation.
Figure 2. Complete 2D DOSY spectrum of xerosol X3 redispersed in
[D3]acetonitrile, with projection along the chemical shift axis.
X3 and S3 can clearly only originate from the presence of
the TiO2 nanoparticles, as such broadened resonances are
completely absent in the blank solution (Figure 1). The dif-
fusion coefficient analysis resulting from the 2D DOSY
spectrum of the xerosol X3 is presented in Table 1, and some
of the corresponding spectral ranges of interest are dis-
played in Figures 3, 4, and 5.
Integration of the resonances of the enolic and ketonic
tautomers of acacH, in the standard 1D spectrum, enables
6960
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 6957 – 6966

Characterization of Nanoparticles
FULL PAPER
one to deduce that approximately 60% of the acacH in the
mixture appears in enolic form, and thus only approximately
40% is in the ketonic form.
exchange in which the ligand remains bound to the surface
for an average time that is longer than the diffusion time
used in the bipolar pulse pair–longitudinal eddy-current
delay (BPP-LED) sequence, that is, 100–200 ms. Stated in
other terms, the exchange kinetics are slow on the diffusion-
al timescale.
Figures 3 and 4 (left panel) clearly reveal the presence of
2D DOSY diffusion correlation peaks lying in the reasona-
bly narrow range of ÀlogD values between 9.3 and 9.6, even
though they are considerably smeared out along the hori-
zontal chemical shift axis, as a consequence of the line
broadening resulting from slowed down rotational molecular
dynamics. In particular, the pTsA ligand gives rise to a
broad ÀlogD range centered between 9.3 and 9.5 for both
aromatic proton resonances (d=7.0–7.6 ppm) as well as for
the methyl singlet (d=2.2–2.4 ppm; Figure 3). Analogously,
the enolic acac ligand gives rise to two rather well-defined
diffusion correlation peaks in the ÀlogD range between 9.4
and 9.6, as found at the cross section of the 2D DOSY spec-
trum for both the vinylic (at d=5.62 ppm; Figure 4, left
panel) and methylic (at d=2.02 ppm; not shown) resonances
of the enolic acac(H) ligand. By contrast, the ketonic acacH
(Figure 4, right panel) and nBuOH resonances (Figure 5) do
not show any DOSY diffusion correlation peak at all in the
range 9.3<ÀlogD<9.6, which clearly demonstrates that
none of them plays a specific role in the functionalization of
the titanium dioxide nanoparticles. As further shown in
Figure 5, the diffusion correlation multiplets around d=1.59
(right panel) and 4.03 ppm (left panel) arise only from
n-butyl acetate, while the correlation multiplets around d=
3.49 (right panel) and 0.9 ppm (left panel) only arise from
nBuOH. All other resonances arise jointly from both
n-butyl acetate and nBuOH, resulting in the observable
peak dispersion along the diffusion axis. Although only a
minor fraction, n-butyl acetate as a byproduct displays two
observable diffusion correlation multiplets that are suffi-
ciently separated to reveal a diffusion coefficient slightly
lower than that for nBuOH.
Slowly diffusing species provide diffusion cross-peaks in
the range 9.3<ÀlogD<9.6, which corresponds to a range of
5”10À10 to 2.5”10À10 m2 sÀ1 for the diffusion coefficients and
to a range of 1.5–3.0 nm for the hydrodynamic radii (h=
0.29 cP). Such a hydrodynamic radius range is in excellent
agreement with the radii of TiO2 nanoparticles determined
previously for the present system by DLS, and confirms
their reasonably narrow size distribution.[13]
The assignment of the enolic and ketonic acac(H) reso-
nances is straightforward. Both forms exhibit diffusion cor-
relation peaks for all their resonances in the ÀlogD range
around 8.5, with diffusion coefficients of 2.90”10À9 and
2.75”10À9 m2 sÀ1 for enolic and ketonic acacH, respectively,
characteristic of unbound, free-moving ligands. Only the
enolic acac tautomer additionally exhibits diffusion correla-
tion peaks around ÀlogD=9.5 (3.4”10À10 m2 sÀ1), character-
istic of the bound ligand. This unambiguously confirms that
the enolic acac ligand is basically the only form of acac that
chelates the Ti atoms at the nanoparticle surface. The two
separated diffusion correlation peaks observed for enolic
acac, on the other hand, enable one to conclude that its
bound and free forms undergo an association–dissociation
The reduced diffusion coefficient for nBuOH in X3 is very
similar to that observed in the blank (Table 1), indicating
that interactions with the nanoparticles are negligible. In the
case of the ketonic acacH, its reduced diffusion coefficient
in X3 is decreased by approximately 20% as compared to
the blank (Table 1). This result suggests that ketonic acacH
exists essentially in free form, but might undergo some slow-
ing down in its diffusion due to the presence of the nanopar-
ticles. According to the very small volume fraction occupied
by the particles (estimated at approximately 1% maximum
for 10 mg of X3 in 750 mL of solvent), obstruction effects
should be negligible.[7] Alternatively, a small fraction of the
ketonic acacH might be bound to the nanoparticles under a
fast exchange regime (for example, with an interaction time
shorter than 100–200 ms) with the corresponding bulk spe-
cies. The same argument holds for the free enolic ligand, the
reduced diffusion coefficient of which is also decreased by
approximately 20% when comparing X3 to the blank
(Table 1). Globally, this apparent slowing down of ligands
not strongly bound to nanoparticles remains a minor,
though significant, effect (see further discussion below).
Deconvolution of the volume of the diffusion correlation
peaks of acac/acacH as a function of the diffusion coefficient
allows the estimation that roughly 5 to 25% of the enolic
form is bound as acac to the surface of the nanoparticles. By
contrast, as for the blank, the behavior of the pTsA ligand
again deserves special focus. The result of a twice as large
relative hydrodynamic radius for pTsA when compared to
nBuOH and both tautomers of acacH (Table 1) is even am-
plified for the xerosol X3, the reduced diffusion coefficient
for the “faster” pTsA fraction (labeled “hindered/free” in
Table 1) going down to 0.27. This particular pTsA fraction is
different from the tightly nanoparticle-bound fraction,
which exhibits an even smaller reduced diffusion coefficient
(0.12) and a well-separated diffusion correlation peak on
both diffusion and chemical shift scales in the aromatic
region. This “hindered/free” pTsA is ascribed to an ex-
change between the free aggregate complex, as observed in
the blank, and an alternative hindered form which is loosely
but significantly experiencing the influence of the slower dif-
fusion of the nanoparticles, this exchange now being fast on
both the chemical shift and diffusion timescales. The validity
of this interpretation is evidenced by the fact that all the res-
onances in both the aromatic and methylic regions exhibit
diffusion correlation peaks in the same ÀlogD range and
also appear in the 1D spectrum as resonances displaying sig-
nificant broadening with respect to the blank solution
(Figure 1). In particular, this broadening reflects a fast ex-
change between the pTsA ligand with a diffusive pattern
typical to the blank, and another slowed down species,
which, however, cannot be pTsA strongly bound to the
Chem. Eur. J. 2007, 13, 6957 – 6966
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
6961

M. Biesemans et al.
nanoparticles. The interactions between the nanoparticles
and these hindered species may be of electrostatic and/or di-
polar nature, and are sufficiently strong to influence their
diffusion rates, but are too weak to prevent fast exchange
with the pTsA ligands in the bulk solution.
Deconvolution of the correlation peaks summed along the
diffusion axis in the chemical shift ranges 7.80–7.00 or 2.45–
2.25 ppm reveals that between 40 and 60% of pTsA ligands
are tightly bound to the surface of the nanoparticles (9.3<
ÀlogD<9.5), while the remaining fraction (8.8<ÀlogD<
9.1) is in fast exchange between the pool of loosely bound
pTsA ligands (hindered) and the free ones.
The slow diffusion regimes of both enolic acac and pTsA
(9.3<ÀlogD<9.6) clearly correspond to a strongly nanopar-
ticle-bound state. Thus, it seems unrealistic that an averaged
diffusion coefficient from a hypothetical fast diffusion-ex-
change regime would result in virtually identical values for
ligands as different as pTsA and enolic acac, certainly in
view of their very different concentrations. Furthermore,
both pTsA and enolic acac ligands do exhibit other
2D DOSY diffusion correlation peaks in other, faster but
different diffusion regimes, as discussed below. Last but not
least, as stated above, the resulting estimated hydrodynamic
radii appear to match very well the numerical values from
DLS.[13]
ficient of the bound enolic acac (3.4”10À10 m2 sÀ1) while
Df(X3) is approximated by the diffusion coefficient of the free
ketonic acacH in the blank, corrected from the viscosity
change (0.92”3.60”10À9 m2 sÀ1). This estimation results in
about 20% of the ketonic acacH interacting weakly with the
nanoparticles and 80% as free as in the blank. For the
enolic acac ligand, one obtains xh(X3) ꢀ0.25 which, together
with the approximately 15% of tightly bound enolic acac
ligand, represents more than a third of the enolic acac expe-
riencing translational slowing down through strong, moder-
ately strong, or weak interactions with the titanium oxide
particles. In contrast, for nBuOH, on the basis of a-CH2 and
the CH3 resonances, which are well isolated on the chemical
shift scale, and therefore the most reliable resonances to be
used, xh(X3) <0.05 confirms unambiguously that nBuOH
plays a marginal role, if any, in the stabilization and tailoring
of the nanoparticles.
For the sol S3, only a limited number of data could be ex-
ploited from the spectra, because the large excess of
nBuOH in this kind of sample causes dynamic range prob-
lems. The aromatic part of the 1D spectrum of the sol S3
also exhibits resonances broader than those for the blank
and X3. The DOSY analysis of this system led to only one
diffusion coefficient for every chemical species: DTMS
25.4”10À10 nBuOH =21.6”10À10 ketonic acacH =22.3”10À10
enolic acacH =25”10À10, and DpTsA =7.6”10À10 m2 sÀ1. For all
=
,
D
,
D
,
For the “faster” fraction of pTsA ligands not tightly
bound to the nanoparticles, an averaged diffusion constant
can be defined as [Eq. (4)]:
D
species but the pTsA, the reduced diffusion coefficients
(DnBuOH =0.88, Dketonic acacH =0.91, and Denolic acacH =1.02) are
remarkably similar to the ones in the blank (Table 1). Only
the pTsA fraction, with a reduced diffusion coefficient of
0.31, is similar to the hindered/free pTsA fraction in X3. It is
thus the only ligand for which the interaction with the nano-
particles could be clearly revealed in S3. The larger amount
of acacH in S3, as compared to X3, does not allow the inter-
action with the nanoparticles to be visible any more.
< D > X3 ¼ xfðX3ÞDfðX3Þ þ xhðX3ÞDhðX3Þ
ð4Þ
in which Df(X3) and xf(X3) refer, respectively, to the diffusion
coefficient and the molar fraction of the free ligand fraction,
while Dh(X3) and xh(X3) refer to those of the hindered ligand
fraction of pTsA. In Equation (4), the quantity <D>X3 is
the experimentally measured parameter, and when consider-
ing only the pool of pTsA ligands not tightly bound to the
nanoparticles, xf(X3) +xh(X3) =1. Dh(X3) is not directly accessi-
ble but the diffusion coefficient measured for the pTsA
tightly bound to the nanoparticles (4.3”10À10 m2 sÀ1) repre-
sents a quite reasonable estimate because, even though their
interaction is weaker, the hindered pTsA moves along with
the nanoparticles when they are bound to them. Df(X3) is
also not directly accessible but, as it corresponds to the free
pTsA, the diffusion coefficient measured for the pTsA in
the blank, corrected for the change of viscosity (0.43”36”
10À10 m2 sÀ1), can be taken as a reasonable estimate. From
these values xf(X3) and xh(X3) both turn out to be roughly 0.5.
To summarize, the pTsA in X3 can be split between three
different types: the tightly bound (ꢀ50%), the weakly
bound (ꢀ25%), and the free one (ꢀ25%), the latter two
being in fast exchange on both the diffusion and chemical
shift timescales.
DOSY NMR investigations in [D4]methanol: With respect
to [D3]acetonitrile, changing the solvent to [D4]methanol re-
sults in a significant loss of spectral resolution. nBuOH also
exhibits a lower D(model) than the other ligands, which can be
logically ascribed to hydrogen bonding with the solvent mol-
ecules being stronger than for the other ligands (Table 2).
Finally, the comparison of the reduced diffusion coefficients,
between the blank and the models, indicates that mixing the
ligands in [D4]methanol slows down their diffusion with re-
spect to the corresponding model solutions. In
[D3]acetonitrile, this effect was only observed for pTsA.
This result is quite normal, as the total concentration of
solute molecules is higher in the blank than in each of the
individual model solutions, and therefore all molecules are
prone to more attractive intermolecular forces. This is also
in line with nBuOH being a better hydrogen-bond donor
than methanol.
A similar estimation can be performed for the ketonic
acacH ligand, which does not display any tight bonding to
the nanoparticles. The measured <D>X3 equals 2.75”
10À9 m2 sÀ1, and Dh(X3) is approximated by the diffusion coef-
For the xerosol X3 in [D4]methanol, several striking fea-
tures can be outlined in the 1D 1H NMR spectrum when
compared to data in [D3]acetonitrile. First, only relatively
sharp doublets and a methyl singlet are observed for pTsA,
6962
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 6957 – 6966

Characterization of Nanoparticles
FULL PAPER
Table 2. Diffusion coefficients[a] [10À10 m2 sÀ1] and reduced diffusion coefficients, in [D4]methanol, for the pure
compounds (model solutions), blank sample, and redispersed xerosol X3. The relative fractions are also given
for species in different states (for example, free and bound).
For the xerosol X3 (Table 3),
the enolic acac tautomer shows
again two well-separated diffu-
sion correlation peaks, indicat-
ing a free and a bound form.
Compounds
Model solutions
Blank
Dj(blank)
X3
Dj(X3)
Dj(model)
Dj(model)
Dj(blank)
Dj(X3)
%
free
hindered/free
free acacH
bound acac
free acacH
free
9.3
0.49
1.06
6.4
0.40
Differently
from
pTsA
7.9
15.7
2.4
15.8
14.7
16.2
0.49
0.97
0.15
0.98
0.91
1.00
100
85
15
100
100
100
[D3]acetonitrile, but similarly
to [D4]methanol, only a single
diffusion regime is observed
for pTsA in [D6]DMSO, which
is very similar to that of the
blank. This is assigned to
DMSO molecules which are
22.1
14.6
0.92
enolic 80%
ketonic 20%
nBuOH
TMS
20.8
14.6
20.3
1.00
0.75
1.00
13.2
10.4
15.9
0.83
0.65
1.00
free
[a] The relative accuracy of the diffusion coefficients is Æ1% for models, Æ1% for the blank, and Æ10% for
X3, and the relative accuracy of the fractions determined from DOSY NMR is Æ10% (see the Experimental
Section).
present in
a
much larger
without any additional broad
Table 3. Diffusion coefficients[a] [10À10 m2 sÀ1] and reduced diffusion coefficients, in [D6]DMSO, for the pure
compounds (model solutions), blank sample, and redispersed xerosol X3. The relative fractions are also given
for species in different states (for example, free and bound).
resonances. The 2D DOSY
spectrum confirms that no
nanoparticle-bound pTsA is
observed any more. No reso-
nance related to the enolic
form of acac is observed in the
olefinic resonance range. This
finding can readily be ex-
plained by the fact that the
central vinylic proton of the
enolic acac ligand can be sub-
stituted by the hydroxylic deu-
teron of [D4]methanol. How-
ever, in the methyl range, the
enolic form exhibits a single
Compounds
Model solutions
Blank
Dj(blank)
X3
Dj(X3)
Dj(model)
Dj(model)
Dj(blank)
Dj(X3)
%
free
hindered/free
free acacH
bound acac
free acacH
free
3.2
0.55
1.08
3.8
0.55
pTsA
7.0
7.0
2.4
7.0
6.3
7.0
0.51
1.00
0.35
0.99
0.89
1.00
100
85
15
100
100
100
9.4
7.9
1.14
enolic 70%
ketonic 30%
nBuOH
TMS
8.5
7.0
7.0
0.98
0.88
1.00
6.9
6.2
7.0
1.00
0.88
1.00
free
[a] The relative accuracy of the diffusion coefficients is Æ1% for models, Æ1% for the blank, and Æ10% for
X3, and the relative accuracy of the fractions determined from DOSY NMR is Æ10% (see the Experimental
Section).
methyl resonance displaying two diffusion regimes, a fast
and a slow one. According to the viscosity of the medium
(h=0.75 cP), the slower diffusion regime corresponds to a
hydrodynamic radius of approximately 1.2 nm and is related
to the enolic acac bound to the surface of the nanoparticles.
As for [D3]acetonitrile, the exchange between the bound
and free enolic acac in [D4]methanol is slow on the diffusion
timescale. For [D3]acetonitrile dispersions, only a single dif-
fusion regime is observed for the ketonic tautomer of acacH
and nBuOH (Table 2). All compounds show a faster diffu-
sive behavior in the xerosol X3 than in the blank, which is
probably due to the smaller relative amount of nBuOH
present in the former than in the latter. The main difference
with respect to the xerosol X3 in [D3]acetonitrile is that, in
[D4]methanol, the bound pTsA ligands are now driven away
from the nanoparticle surface. This feature is probably relat-
amount than pTsA moieties, therefore overruling them in
coordinating titanium at the nanoparticle surface, the strong
complexing properties of DMSO towards metal atoms being
well known. However, the two diffusion regimes of enolic
acac(H) indicate that this strong coordination of DMSO
does not drive the enolic acac ligand away from the nano-
particle surface. The diffusion behavior of nBuOH and ke-
tonic acacH is not influenced by the presence of nanoparti-
cles in [D6]DMSO, which has nearly the same polarity as
[D3]acetonitrile.
DOSY NMR investigations in [D]chloroform: In CDCl3,
phase separation occurs in the pTsA model solution, blank
and S3 system. Accordingly, only the xerosol X3 in CDCl3
was investigated (Table 4).
The molecules in the X3 xerosol in CDCl3 have a nearly
identical diffusive behavior to that in CD3CN. Three differ-
ent diffusion regimes are clearly distinguished: a free diffu-
sion regime for both acacH tautomers and nBuOH, a bound
diffusion regime for enolic acac and pTsA and a hindered/
free diffusion regime for pTsA due to interactions, as al-
ready discussed for [D3]acetonitrile. The similarity of these
findings with those in the [D3]acetonitrile dispersion is
really striking, even for the thermodynamic fractions. Also,
the hydrodynamic radius (ꢀ2.5 nm) confirms the size of the
nanoparticles already mentioned above for other solvents.
À
ed to the esterification of the surface Ti OH moieties by
CD3OD.
DOSY NMR investigations in [D6]dimethyl sulfoxide:
DOSY data on the blank and model solutions in [D6]DMSO
are gathered in Table 3. With respect to [D3]acetonitrile
(Table 1) and [D4]methanol (Table 2), the mixing effect to-
tally disappears upon changing the solvent to [D6]DMSO, as
all the reduced diffusion coefficients remain unchanged
upon mixing.
Chem. Eur. J. 2007, 13, 6957 – 6966
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
6963

M. Biesemans et al.
Table 4. Diffusion coefficients[a] [10À10 m2 sÀ1] and reduced diffusion coef-
ficients, in CDCl3, for the redispersed xerosol X3. The relative fractions
are also given for species in different states (for example, free and
bound).
values of the diffusion coefficients, but does not affect fun-
damentally the interaction behavior of the ligands at the
nanoparticle surface. The only significantly different behav-
ior is found for methanol, as a consequence of its specific
hydrogen-bond donor capacities and ability to esterify the
Compounds
X3
Dj(X3)
Dj(X3)
%
À
surface Ti OH, with the result that the slow association/dis-
hindered/free
bound
free acacH
bound acac
free acacH
free
7.2
2.2
18.9
1.8
22.5
21.9
22.3
0.32
0.10
0.85
0.08
1.01
0.98
1.00
50
50
85
pTsA
sociation equilibrium of the pTsA ligand is speeded up.
Generally, all compounds show for all solvents, in model
solutions as well as in the blank, diffusion characteristics
which are similar to that of TMS; pTsA, the diffusion of
which is hindered most likely by intermolecular interaction
phenomena, constitutes a noticeable exception. In nanopar-
ticle dispersions, the same free fractions and the hindered/
free pTsA fraction are found, within experimental scatter.
Furthermore, a nanoparticle-related diffusion regime is
found for enolic acac and pTsA.
enolic 60%
15
ketonic 40%
nBuOH
TMS
100
100
100
free
[a] The relative accuracy of the diffusion coefficients is Æ10% and of the
fractions determined from DOSY NMR is Æ10% (see the Experimental
Section).
This clearly illustrates that changing the polarity of the sol-
vent, while keeping the viscosity almost constant, does not
significantly alter either the global diffusive behavior of the
organic compounds or the size characteristics of the nano-
particles in the system.
Every mixture component and diffusion regime was ana-
lyzed with different H magnetic probes. In this way, all data
1
are self-consistently cross-checked, thus confirming the reli-
ability of the method, including data on the nanoparticle hy-
drodynamic radius of approximately 2 nm confirmed by
DLS measurements. A global overview of the interaction
and chemical-exchange model at the nanoparticle surface
and in the bulk solution is given in Scheme 1.
Conclusion
It can be concluded that DOSY NMR is a powerful meth-
odology enabling one to analyze qualitatively and semiquan-
titatively the thermodynamics and kinetics of interaction
and exchange processes of organic ligands in all nanoparti-
cle dispersions in different solvents. Undoubtedly, the meth-
odology proposed goes far beyond the restricted frame of
TiO2 nanoparticles addressed here.
Diffusion-ordered NMR spectroscopy appears to be an ele-
gant methodology to measure diffusion coefficients of or-
ganic structures interacting with nanoparticles. Coupling this
technique to diffusion measurements on model solutions of
these organic species provides valuable information on how
their intrinsic translational mobility is altered in more com-
plex nanoparticle dispersions. Thus, a systematic investiga-
tion going from model solutions to nanoparticle dispersions
enables one to discriminate between solvation and interac-
tion phenomena also present in model and blank solutions,
and association/dissociation phenomena only present in the
nanoparticle dispersion.
The analysis of 2D DOSY spectra, in both the chemical
shift and diffusion dimensions, by using appropriate mathe-
matical software tools for integration and deconvolution of
the spectra or Gaussian decays, results in the quantification
of thermodynamic fractions of the tautomeric equilibrium of
acetylacetone and the dissociation/association equilibrium of
the ligands at the surface of the nanoparticles. However, be-
cause of the ill-conditioned nature of inverse Laplace trans-
formation (ILT), the fractions of the bound and free ligands
should be considered with appropriate caution if the distinc-
tion of the species is only possible in the diffusion dimension
of the 2D DOSY spectra as a consequence of identical
chemical shifts.
Although a priori expected, the polarity does not influ-
ence significantly the behavior of the compounds in the
nanoparticle dispersion, as clearly demonstrated by the very
similar diffusion behaviors observed in [D]chloroform and
[D3]acetonitrile, which exhibit a dipolar moment/dielectric
constant couple of 1.04 D/4.8, and 3.93 D/36.6, respective-
ly.[20] The viscosity of the solvents determines the absolute
Scheme 1. Summarizing scheme of all possible equilibria and interactions
between species in solution and nanoparticles in the dispersion mixture,
as deduced from DOSY NMR spectroscopic data acquired from
[D3]acetonitrile dispersions.
Experimental Section
Synthesis: The sol S3 and xerosol X3 were prepared according to pub-
lished procedures.[13a] For the sol S3, Ti
ACHTREUNG
6964
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 6957 – 6966

Characterization of Nanoparticles
FULL PAPER
(3 equiv, 44.1 mmol). After stirring for 15 min, pTsA (0.2 equiv) in pure
water (10 equiv) was added and the mixture was stirred for 2 h at room
temperature, then heated at 658C for 24 h. For the corresponding xerosol
X3, the mixture was dried under vacuum for 4 h at approximately
0.01 mm Hg at room temperature, and subsequently overnight at 608C
under atmospheric pressure, which provided a mass reduction of slightly
above 90%. For the blanks, a solution containing nBuOH (10 equiv,
147 mmol) and acacH (3 equiv, 44.1 mmol) was prepared. After stirring
for 15 min, pTsA (0.2 equiv) in pure water (10 equiv) was added and the
mixture was stirred for 2 h at room temperature, then heated at 658C for
24 h.
Instruments for NMR spectroscopy: Most 1H NMR spectroscopic data
were acquired on
a Bruker Avance 250 spectrometer operating at
250.13 MHz. The spectral width was 3000 Hz, with an acquisition time of
1.36 s (digital resolution=0.367 Hz per point), a recycling delay of 6.0 s
and a 908 pulse width of 10.0 ms. The data were processed with an expo-
nential broadening of 0.30 Hz. Some of the 1H NMR spectroscopic data
were acquired on a Bruker Avance II 700 spectrometer operating at
700.13 MHz. The spectral width was 9800 Hz, with an acquisition time of
3.34 s (digital resolution=0.150 Hz per point), a recycling delay of 6.0 s,
and a 908 pulse width of 7.0 ms. The data were processed with an expo-
nential broadening of 0.60 Hz.
DOSY NMR measurements: Samples were prepared for NMR acquisi-
tions by dissolving appropriate amounts in deuterated solvents (0.75 mL)
containing TMS (3 mL) as a reference. The diffusion experiments were
performed by pulsed field gradient (PFG) 1H NMR spectroscopy by
means of a BPP-LED pulse sequence,[21] with shaped gradient pulses of
Figure 6. Stacked plot of the aromatic spectral range of the 2D DOSY
spectrum of the xerosol X3 redispersed in [D3]acetonitrile. The plot clear-
ly shows the less intense, broader (along the chemical shift scale) peaks
of pTsA ligands bound to nanoparticles. The deconvolution shown in the
inset was performed on the projection resulting from summing up all
cross sections parallel to the diffusion axis of the 2D DOSY spectrum
within the chemical shift range d=7.0–7.8 ppm.
amplitude G and duration d modulated as the G
N
N
periments were carried out at 300 K, without sample spinning, on a
Bruker Avance 250 spectrometer equipped with a Bruker BGU z-gradi-
ent unit providing a maximum gradient of 51 GcmÀ1. For each solution,
the 1D spectra were acquired by using a BPP-LED pulse sequence to op-
timize the gradient pulse duration (d) and the diffusion delay (D) in
order to detect properly the full signal decay. The amplitude of the gradi-
ents was varied from 2 to 95% of their maximum value, while the gradi-
ent recovery delay (t) and the eddy current delay (te) were fixed at 1 and
5 ms, respectively. The diffusion delay (D) varied from 50 to 340 ms and
the gradient pulse duration (d) from 0.5 to 3.5 ms.
Viscosity measurements: For the three types of samples analysed in each
solvent, the blank, the sol (S3), and the re-dispersed xerosol (X3), the
actual dynamic viscosity of the medium was determined by using TMS as
an internal reference.[26] TMS appeared particularly appropriate for this
purpose, as it is inert, nonpolar, and nonpolarizable, and therefore hardly
interacts with any of the substrates of interest, including the solvent. This
leads to the reasonable assumption that the product of its hydrodynamic
radius (rH) and the parameter c of the Stokes–Einstein relation is con-
stant in a given solvent. The viscosities of the pure deuterated solvents
were taken from Holz et al.[27] A complete list of viscosities is given in
the Supporting Information.
DOSY NMR data processing: The DOSY NMR data were processed by
ILT,[22] by using a maximum entropy algorithm, MaxEnt,[23] as implement-
ed in the Gifa Processing software interfaced with TOPSPIN software.
The ILT–MaxEnt 2D DOSY spectra obtained in this way are represented
on the vertical axis in f1 as the negative logarithmic scale of the diffusion
coefficient, and on the horizontal axis in f2 as the chemical shift (in ppm),
for each measurable moiety of the corresponding diffusing species. The
resulting 3D spectrum was (numerically) integrated over a given reso-
nance range along the chemical shift axis, which provides an intensity
function depending only on the ÀlogD variable. A sum of Gaussian func-
tions was fitted to these data (see Figure 6) with different independent
software packages (Excel, TableCurve, Perch), which increased the
number of independent Gaussian distributions until (at least) two Gaussi-
ans equal within statistical error (clearly indicating that they represent
the same diffusion regime) or the fraction of (at least) one Gaussian was
reduced to zero (clearly indicating there are fewer diffusion regimes than
represented by Gaussian distributions). The peak maxima thus obtained
were used as the representative diffusion coefficients. The accuracy of
the diffusion coefficients depends on the system. For spectra with no
overlapping resonances, that is, models and blank, the ILT–MaxEnt 2D
DOSY gave the same results as a simple monoexponential fitting, the ac-
curacy of which is below 1%.[24] For strongly overlapping systems, the ac-
curacy of the diffusion coefficients determined by ILT–MaxEnt
2D DOSY can be estimated at Æ10% by using a Monte Carlo ap-
proach.[25] However, the spread of the diffusion coefficients measured for
the different resonances of a unique chemical species (for example,
pTsA) was always narrower. With the same Monte Carlo approach, the
accuracy of the fractions of the different components is also estimated as
Æ10%.[25]
1H NMR spectroscopic data (250 MHz): For all the studied samples,
chemical shifts, integrals, homonuclear proton scalar coupling constants,
and assignments are given in the Supporting Information.
Acknowledgements
The authors wish to thank Professor M. A. Delsuc (UniversitØ de Mont-
pellier I) for helpful discussions concerning the MaxEnt implementation
of the ILT. Financial support by the fund of Scientific Research Flanders
(Belgium) (FWO; Grants G.0016.02 and G.0469.06) and the Research
Council (Onderzoeksraad) of the Vrije Universiteit Brussel (Concerted
Research Action, Grant GOA31) to R.W., L.V.L., G.M., and M.B. is
gratefully acknowledged. V.E. is indebted to the fund of Scientific Re-
search Flanders (Belgium) (FWO) for a postdoctoral foreign fellowship.
R.W. and J.C.M. acknowledge the financial support by the fund of Scien-
tific Research Flanders (Belgium) (FWO; Grant G.0064.07). The FWO
Scientific Research Network on Magnetic Resonance is gratefully ac-
knowledged for financially facilitating exchange of researchers between
the different labs involved in this work (Grant WO.01407). The “Minis-
t›re des affaires Etrang›res” is acknowledged for its support through the
PHC-Tournesol No. 11537UF.
Chem. Eur. J. 2007, 13, 6957 – 6966
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
6965

M. Biesemans et al.
[13] a) E. Scolan, C. Sanchez, Chem. Mater. 1998, 10, 3217–3223; b) E.
Scolan, C. Magenet, D. Massiot, C. Sanchez, J. Mater. Chem. 1999, 9,
2467–2474.
[14] D. T. Cromer, K. Herrington, J. Am. Chem. Soc. 1955, 77, 4708–
4709.
[1] a) B. Antalek, Concepts Magn. Reson. 2002, 14, 225–258; b) S.
Abrahmsen-Alami, P. Stilbs, J. Colloid Interface Sci. 1997, 189, 137–
143; c) C. S. Johnson Jr, Prog. Nucl. Magn. Reson. Spectrosc. 1999,
34, 203–256.
[2] a) L. H. Lucas, W. H. Otto, C. K. Larive, J. Magn. Reson. 2002, 156,
138–145; b) W. S. Price, H. Ide, Y. Arata, J. Phys. Chem. A 2003,
107, 4784–4789.
[15] a) E. J. Cabrita, S. Berger, P. Brꢁuer, J. Kꢁrger, J. Magn. Reson.
2002, 157, 124–131; b) E. J. Cabrita, S. Berger, Magn. Reson. Chem.
2002, 40, S122–S127.
[16] D. Canet, M. Decorps in Dynamics of Solutions and Fluid Mixtures
by NMR (Ed.: J. J. Delpuech), Wiley, Chichester, 1995, pp. 336–340.
[17] D. Zuccaccia, A. Macchioni, Organometallics 2005, 24, 3476–3486.
[18] E. J. Cabrita, S. Berger, Magn. Reson. Chem. 2001, 39, S142–S148.
[19] From the molecular structure of acacH and pTsA, the average
atomic distance to their geometric centre can be calculated at 2.5
and 2.2 ꢂ for pTsA and acacH, respectively. The maximum distance
to their geometric centre can be similarly calculated at 3.8 and 3.2 ꢂ
for pTsA and acacH, respectively. In conclusion, both molecules ex-
hibit a similar size.
[20] CRC Handbook of Chemistry and Physics, 82nd ed., CRC Press,
Boca Raton, 2001.
[21] D. Wu, A. Chen, C. S. Johnson Jr. , J. Magn. Reson. A 1995, 115,
260–264.
[22] a) R. Huo, R. Wehrens, J. van Duynhoven, L. M. C. Buydens, Anal.
Chim. Acta 2003, 490, 231–251; b) B. Antalek, J. M. Hewitt, W.
Winding, P. D. Yacobucci, T. Mourey, K. Le, Magn. Reson. Chem.
2002, 40, S60–S71.
[23] F. Mariette, J. P. Guillement, C. Tellier, P. Marchal in Signal Treat-
ment and Signal Analyses in NMR (Ed.: D. N. Rutledge), Elsevier,
Amsterdam, 1996, pp. 218–234.
[3] P. Groves, M. O. Rasmussen, M. D. Molero, E. Samain, F. J. Canada,
H. Driguez, J. Jimenez-Barbero, Glycobiology 2004, 14, 451–456.
[4] O. Sçderman, P. Stilbes, Prog. Nucl. Magn. Reson. Spectrosc. 1994,
26, 445–482.
[5] M. Valentini, P. S. Pregosin, H. Ruegger, Organometallics 2000, 19,
2551–2555.
[6] C. Roberts, T. Cosgrove, R. G. Schmidt, G. V. Gordon, Macromole-
cules 2001, 34, 538–543.
[7] P. Bernado, J. Garcia de la Torre, M. Pons, J. Mol. Recognit. 2004,
17, 397–407.
[8] a) M. J. Stchedroff, A. M. Kenwright, G. A. Morris, M. Nilsson,
R. K. Harris, Phys. Chem. Chem. Phys. 2004, 6, 3221–3227; b) M.
Nilsson, I. F. Duarte, C. Almeida, I. Delgadillo, B. J. Goodfellow,
A. M. Gil, G. A. Morris, J. Agric. Food Chem. 2004, 52, 3736–3743;
c) A. R. Bilia, M. C. Bergonzi, F. F. Vincieri, P. Lo Nostro, G. A.
Morris, J. Pharm. Sci. 2002, 91, 2265–2270; d) R. K. Harris, K. A.
Kinnear, G. A. Morris, M. J. Stchedroff, A. Samadi-Maybodi, N.
Azizi, Chem. Commun. 2001, 23, 2422–2423; e) P. Hodge, P. Monvi-
sade, G. A. Morris, I. Preece, Chem. Commun. 2001, 3, 239–240.
[9] F. Ribot, V. Escax, J. C. Martins, M. Biesemans, L. Ghys, I. Verbrug-
gen, R. Willem, Chem. Eur. J. 2004, 10, 1747–1751.
[24] M. Nilsson, M. A. Connell, A. L. Davis, G. A. Morris, Anal. Chem.
2006, 78, 3040–3045.
[25] M. A. Delsuc, T. E. Malliavin, Anal. Chem. 1998, 70, 2146–2148.
[26] E. J. Cabrita, S. Berger, Magn. Reson. Chem. 2001, 39, S142–S148.
[27] M. Holz, X. Mao, D. Seiferling, A. Sacco, J. Chem. Phys. 1996, 104,
669–679.
[10] J. M. Santos, R. Ribeiro da Silva, A. L. Formiga, L. W. Tinoco, J. D.
Figueroa-Villar, Chem. Phys. 2004, 306, 143–151.
[11] F. Ribot, V. Escax, C. Roiland, C. Sanchez, J. C. Martins, M. Biese-
mans, I. Verbruggen, R. Willem, Chem. Commun. 2005, 1019–1021.
[12] a) Z. Hens, I. Moreels, J. C. Martins, ChemPhysChem 2005, 6, 2578–
2584; b) I. Moreels, J. C. Martins, Z. Hens, ChemPhysChem 2006, 7,
1028–1031.
Received: December 1, 2006
Published online: June 12, 2007
6966
www.chemeurj.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2007, 13, 6957 – 6966