Available online at www.sciencedirect.

com

Colloids and Surfaces A: Physicochem. Eng. Aspects 316 (2008) 210216

Preparation and characterization of water-soluble monodisperse magnetic

iron oxide nanoparticles via surface double-exchange with DMSA
Z.P. Chen a,b,c , Y. Zhang a,b , S. Zhang a,b , J.G. Xia a,b , J.W. Liu a,b,c , K. Xu a,b,c , N. Gu a,b,
a

State Key Laboratory of Bioelectronics, Southeast University, Nanjing 210096, PR China

b Jiangsu Laboratory for Biomaterials and Devices, Nanjing 210096, PR China
c School of Chemistry and Chemical Engineering, Southeast University, Nanjing 210096, PR China
Received 20 April 2007; received in revised form 17 August 2007; accepted 1 September 2007
Available online 12 September 2007

Abstract
A simple, but efficient method for preparation of water-soluble iron oxide nanoparticles has been developed. Monodisperse Fe3 O4 nanoparticles
were synthesized by thermal decomposition of iron-oleate. Surface double-exchange of oleic acid capped monodisperse Fe3 O4 nanoparticles
with a familiar 2,3-dimercaptosuccinnic acid (HOOCCH(SH)CH(SH)COOH, DMSA) was first performed in chloroform in the presence of
triethylamine, and then this process was repeated in ethanol under the same conditions. The resulting Fe3 O4 nanoparticles could be transferred
into water to form stable magnetic fluid without post-treatment processes such as filtration and re-concentration. TEM images show that watersoluble Fe3 O4 nanoparticles remain monodisperse and even form a monolayer of ordered assembly, and the results of TGA, VSM show that Fe3 O4
nanoparticles via surface double-exchange possess more DMSA molecules through intermolecular disulfide cross-linking between DMSA, as
confirmed by Raman spectra. Zeta potential measurements show that nanoparticles after surface double-exchange are negatively charged in the
range of pH = 114, and stability assays exhibit their excellent stability in water and other physiological environments.
2007 Elsevier B.V. All rights reserved.
Keywords: Monodisperse; Iron oxide nanoparticle; DMSA; Surface double-exchange; Water-soluble

1. Introduction
Magnetic nanoparticles have been widely studied because
of their many biological applications such as magnetic separation [1,2], DNA detection [3], magnetic resonance imaging
[4,5], target-drug delivery [6], and magnetic hyperthermia [7,8].
Among magnetic nanoparticles, iron oxide nanoparticles are of
particular interests for applications because of their unique magnetic properties and biocompatibility. For future highly sensitive
magnetic nanodevices and biological applications, iron oxide
nanoparticles with controlled-shape, -size, and a narrow size distribution are urgent. Very recently, several groups have reported
that such high-quality iron oxide nanoparticles could be synthesized by thermal decomposition of different types of iron
precursors such as iron acetylacetonate [9], iron pentacarbonyl
Corresponding author at: State Key Laboratory of Bioelectronics, Southeast
University, Nanjing 210096, PR China. Tel.: +86 25 83794960;
fax: +86 25 83792576.
E-mail address: [email protected] (N. Gu).

0927-7757/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.colsurfa.2007.09.017

[10], and iron-oleate [11,12]. However, the direct products of

the above-mentioned approaches are organic-soluble, which to
some extent limits their applications in biological fields.
To take advantage of their high-quality properties in
biological applications, it is necessary to transfer iron oxide
nanoparticles from organic to aqueous solution. Some groups
have reported surface modification techniques including
surface exchange with cyclodextrin [13], copolypeptides [14],
anionic octa(tetramethylammonium)-polyhedral oligomeric
silsesquioxane (TMA-POSS) [15], and intercalation of surfactants [16]. Especially, some reports have given prominences to
surface exchange with DMSA by which iron oxide nanoparticles
modified are fairly stable in water over wide ranges of pH and
salt concentrations [4,5,1720], which makes them preferable
in biological applications. From the point of surface chemistry,
the success of any surface exchange process relies on a careful
balance of the intermolecular forces driving the interaction
between the molecules to be exchanged and the outermost layer
of the substrate surface. Bruce et al. reported an inexpensive
method for the introduction of more amino groups onto the

Z.P. Chen et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 316 (2008) 210216

surface of silica-coated iron oxide nanoparticles under various

reaction conditions such as solvent systems and temperatures
[21]. Here, we developed surface double-exchange of oleic acid
capped monodisperse Fe3 O4 nanoparticles with DMSA in the
presence of triethylamine. The resulting Fe3 O4 nanoparticles
could be transferred into water to form stable magnetic fluid
without post-treatment processes and kept stable for several
months, even at high concentration.
2. Experimental
2.1. Materials
1-octadecene was purchased from Alfa Aesar. 4Morpholineethanesulfonic acid (MES) was purchased from
Pierce and RPMI-1640 was purchased from Gibco. Note that
RPMI-1640 used in our experiments contains 10% (v/v) fetal
calf serum. The other chemicals were analytical reagents and
purchased from Shanghai Chemical Reagent Corporation,
China. All chemicals were used as received. Distilled water
was used for all the experiments.
2.2. Synthesis of monodisperse Fe3 O4 nanoparticles
Known method was followed to synthesize monodisperse
Fe3 O4 nanoparticles [11]. Monodisperse Fe3 O4 nanoparticles
were synthesized in two steps: first, to prepare an iron-oleate
precursor and second, to decompose the precursor. In a typical experiment, 2.7 g of FeCl3 6H2 O was dissolved in 50 mL
of methanol followed by addition of 8.5 mL oleic acid. And
then, a solution with 1.2 g of NaOH in 100 mL of methanol was
dropped into above solution under magnetic stirring conditions.
The observed brown precipitate was washed with methanol 45
times and dried under vacuum overnight to remove all solvents.
The obtained waxy iron-oleate was dissolved in 1-octadecanol
at 70 C and reserved as a stable stock solution at room temperature.
One millilitre of the above stock solution (0.39 mol/mL) was
mixed with 4 mL 1-octadecanol and 0.5 mL oleic acid. The reaction mixture was heated to 320 C at a constant heating rate
of 3.3 C/min under a nitrogen atmosphere, and then kept at
that temperature for 30 min. The resulting solution was cooled
and precipitated by addition of excess ethanol and centrifugation. And then, the precipitate containing Fe3 O4 nanoparticles
was washed 45 times with ethanol. This sample was labeled as
MNP-1.
2.3. Preparation of water-soluble Fe3 O4 via surface
double-exchange
One hundred milligrams of MNP-1 was dissolved in 10 mL
chloroform followed by addition of 50 L triethylamine. Fifty
milligrams of DMSA was dispersed in 10 mL dimethyl sulfoxide (DMSO). And the solution was added into the above solution
containing MNP-1. The resulting solution was vortexed at 60 C
for 12 h. The solution became turbid, and the black precipitate
was observed. The final sample labeled as MNP-2 was obtained

211

by centrifugation, followed by washing with ethanol carefully,

and dissolved in 10 mL water for characterization and comparison. This process was repeated, and obtained MNP-2 was
dissolved in 100 mL ethanol for next reaction.
We found that MNP-2 was more easily dissolved in ethanol
than water, so surface double-exchange was repeated in ethanol
to introduce more DMSA molecules onto the surface of Fe3 O4
nanoparticles. Especially, 50 L triethylamine was added into
above ethanol solution containing MNP-2, followed by addition
of a solution with 50 mg DMSA in 10 mL DMSO. The solution
was vortexed at 60 C for 12 h. The final sample labeled as MNP3 was centrifuged and washed with ethanol 45 times carefully.
MNP-3 was collected using a permanent magnet and transferred
into 10 mL water.
2.4. Characterization
The size and morphology of the particles were determined
by a JEM-2000EX TEM operating at 120.0 kV. Samples were
dropped, either from chloroform or water, onto a carbon-coated
copper grid and dried under room temperature. IR spectra were
recorded on a Nicolet Nexus 870 FT-IR spectrometer and powder samples were dried at 100 C under vacuum for 24 h prior
to fabrication of the KBr pellet. Spectra were recorded with
a resolution of 2 cm1 . Raman spectra were collected on a
JY HR800 spectrometer with a resolution of 1 cm1 . Thermogravimetric analysis (TGA) was performed for powder samples
(5 mg) with a heating rate of 20 C/min using a Perkin-Elmer
TGA 7 Thermogravimetric Analyzer in synthetic N2 atmosphere
up to 700 C. Magnetic measurements were carried out with a
Lakeshore 7470 vibrating sample magnetometer (VSM).
Photon correlation spectroscopy (PCS) was used to determine
hydrodynamic size distribution using a Beckman Coulter N4
Plus Submicron Particle Analyzer, and surface charge measurements were performed with a Beckman Coulter Delsa 440SX
Zeta Potential Analyzer. Furthermore, to investigate stability
of MNPs dispersed in different media, 1 mL MNP-2 solution
was added to four 50 mL beakers containing 50 mL water,
RPMI-1640, PBS and MES buffer solution, respectively. The
suspension was settled at room temperature and the upper solution was taken to record UVvisible absorbance spectra every
day for 5 days [15], using a U-4100 spectrophotometer. Stability
assays for MNP-3 were performed as above-mentioned process.
3. Results and discussion
3.1. TEM
Monodisperse Fe3 O4 nanoparticles (MNP-1) were synthesized by thermal decomposition of iron-oleate in the presence of
oleic acid. TEM image (Fig. 1a) shows that MNP-1 is monodisperse with about 10 nm average diameter and well-dispersed in
chloroform. According to the ED pattern (Fig. 1b), the d-spacing
can be calculated in the following equation:
L = dR
where L is the distance between the test sample and the film
(L = 137 cm), the wavelength of electron beam ( = 0.0251 A )

212

Z.P. Chen et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 316 (2008) 210216

Fig. 1. (a) TEM image of MNP-1 dispersed in chloroform; (b) ED image of MNP-1; (c) TEM image of MNP-3 dispersed in water; (d) enlarged image of region
marked with the arrow in (c).

and R is the radius of the diffraction ring. The calculation

results are shown in Table 1, which are well accorded with
theory values of the inverse cubic spinel structure of Fe3 O4 .
Water-soluble Fe3 O4 nanoparticles (MNP-3) were prepared in
chloroform followed by repeated performance in ethanol via
surface double-exchange with DMSA in the presence of triethylamine. TEM images (Fig. 1c and d) show that MNP-3
is well dispersed in water (the TEM image of MNP-2 is not
provided because there is no obvious differences between MNP2 and MNP-3). In addition, these images indicate that no
degradation takes place upon transfer from organic to aqueous
solution.

3.2. IR and Raman

IR spectra of MNP-1, MNP-2, MNP-3 and pure DMSA are
shown in Fig. 2. IR spectra for all samples exhibit one interesting region located between 1800 and 1500 cm1 . In spectrum
a, the sharp peak at 1715 cm1 assigned to carbonyl indicates
that there is free oleic acid [22] and the two peaks at 1569 and
1519 cm1 due to the symmetric and asymmetric carboxylate
(COO ) stretch, respectively, indicate that oleic acid is bound
to the surface of Fe3 O4 nanoparticles through covalent bond
between carboxylate (COO ) and Fe atom [2325]. The appearance of a broad peak at 1580 cm1 in spectrum b and two broad

Table 1
ED experimental data of MNP-1 in comparison with theory values

R (cm)

ED results, d (A)

Theory values, d (A)

Crystalline lane (h l k)

0.70
4.89
4.85
(1 1 1)

1.16
2.95
2.97
(2 2 0)

1.36
2.52
2.53
(3 1 1)

1.64
2.09
2.10
(4 0 0)

2.01
1.70
1.71
(4 2 2)

2.12
1.62
1.62
(5 1 1)

2.32
1.48
1.48
(4 4 0)

Z.P. Chen et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 316 (2008) 210216

213

Fig. 2. IR spectra of (a) MNP-1, (b) MNP-2, (c) MNP-3 and (d) DMSA.

Fig. 4. TGA curves of MNP-1, MNP-2 and MNP-3.

peaks at 1670 and 1580 cm1 in spectrum c derived from the

symmetric and asymmetric stretch of carboxylate (COO ) of
DMSA indicates that DMSA is bound to the surface of Fe3 O4
nanoparticles. In spectrum d, the weak peak at 2562 cm1 is
attributed to thiol group and the sharp peak at 1700 cm1 is
assigned to the carbonyl stretch of DMSA as a diacid. Therefore,
the appearance of the low frequency of the peak at 1711 cm1
both in spectra b and c, compared to the frequency for free oleic
acid (1715 cm1 ), is attributed to free carbonyl stretch of DMSA
on the surface of Fe3 O4 nanoparticles. Additionally, the presence
of the sharp methylene peaks at 2922 and 2854 cm1 , in spectra
ac, indicates that some of initial oleic acid remains on the surface of Fe3 O4 nanoparticles after modification. The reason of the
lack of evidence of thiol group and disulfide group, in spectra b
and c, is that most of thiol groups have been oxidated into disulfide groups which are lower absorption in IR analysis. So Raman
spectra are used to confirm disulfide groups, as shown in Fig. 3.
No peak is observed for MNP-1 between 300 and 800 cm1 ,
whereas a sharp peak for MNP-3 appears at about 503 cm1 ,

due to strong absorption of disulfide groups. This result suggests that intermolecular disulfide cross-linking between DMSA
molecules can introduce more DMSA molecules onto the surface of nanoparticles.
3.3. TGA
TGA has been performed to confirm the coating formation
and estimate the binding efficiency on the surface of Fe3 O4
nanoparticles. Fig. 4 shows the weight loss for MNP-1, MNP-2
and MNP-3. A slight weight loss is observed up to 250 C in
all curves, probably due to adsorbed water, while a significant
weight loss takes place between 250 and 500 C. The weight loss
for MNP-1, attributed to decomposition of oleic acid, is about
25%, corresponding to a monolayer of oleic acid on the surface
[26]. The weight loss for MNP-2 and MNP-3 is increased to
34% and 65%, respectively, mainly due to the decomposition of
DMSA. Interestingly, the molecular weights of oleic acid and
DMSA are 282 and 182, respectively. Assuming one DMSA
molecule exchanged with one oleic acid molecule, the weight
loss for both of MNP-2 and MNP-3 should be 13.6% in contrast
with MNP-1, while if one DMSA molecule exchanged with two
oleic acid molecules, the weight loss for them should be 7.8%.
Therefore, the increased weight loss for MNP-2 and MNP-3
suggests that a multilayer of DMSA exists on their surfaces.
Especially, the more significant weight loss for MNP-3 than
MNP-2 (69%:39%) shows that more DMSA molecules exist on
the surface of MNP-3.
3.4. Magnetic measurements

Fig. 3. Raman spectra of MNP-1 and MNP-3.

Magnetic measurements indicate superparamagnetic behavior at room temperature for all samples, with no hysteresis and
perfect Langevin behavior (Fig. 5). The saturation magnetization
value (Ms ) for MNP-1 is 50.2 emu/g, while the Ms for MNP-2
and MNP-3 are decreased to 33.4 and 16.6 emu/g, respectively.
There are several approaches that can explain the reduction of
the Ms for coated magnetic nanoparticles [2729]. In this case,

214

Z.P. Chen et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 316 (2008) 210216

Table 2
The MS for Fe3 O4 cores based on TGA and VSM results
Sample

Percentage of Fe3 O4 core (% determined by TGA)

Total MS (emu/g; determined by VSM)

The MS of Fe3 O4 core (emu/g)

MNP-1
MNP-2
MNP-3

72
61
31

50.2
33.4
16.6

69.72
54.75
53.55

Fig. 5. Hysteresis loops at room temperature for MNP-1, MNP-2 and MNP-3.

the presence of nonmagnetic DMSA molecules on the surface

of MNP-2 and MNP-3 leads to decrease of the Ms , and more
decrease of the Ms for MNP-3 than MNP-2 indicates that there
are more nonmagnetic DMSA molecules on the surface of MNP3, which is consistent with TGA results. Furthermore, since the
weight percentage for Fe3 O4 core is about 72%, 61%, 31%,
determined by TGA, the Ms for Fe3 O4 cores of MNP-1, MNP2 and MNP-3 are 69.72, 54.75, and 53.55 emu/g, respectively
(Table 2).

Fig. 6. Hydrodynamic size distribution of MNP-2 and MNP-3 dispersed in

water.

respectively (Fig. 8b). These results indicate that MNP-3 is more

stable than MNP-2 in water and physiological environments.
Interestingly, UVvisible absorbance of MNP-2 dispersed in
RPMI-1640 changes to 0 after 5 days, indicating that there
are almost no Fe3 O4 nanoparticles in the upper solution. However, UVvisible absorbance of MNP-3 dispersed in RPMI-1640
changes from 0.328 to 0.251, indicating that it is the most stable
for all the samples.

3.5. Stability assays

3.6. Structure and water-solubility test

Fig. 6 shows hydrodynamic size distribution of MNP-2 and

MNP-3. The average hydrodynamic sizes of MNP-2 and MNP3 are about 250 and 320 nm, respectively, indicating that both
of MNP-2 and MNP-3 exist as aggregates in water. However,
smaller hydrodynamic size of MNP-3 than MNP-2 suggests that
MNPs-3 be of better stability. Zeta potential results (Fig. 7) show
that MNP-3 is always negatively charged in the range of pH
114, whereas MNP-2 has an isoelectric point (IEP) at pH 1.62,
indicating that total charge of DMSA molecules on the surface
keeps MNP-3 sterically and electrically stable in water.
Many biological applications require magnetic nanoparticles
to exhibit stability in physiological environments, so we investigate stability of MNP-2 and MNP-3 in water, RPMI-1640 (a
complex medium for cell culture) and two kinds of buffer solutions (MES and PBS) based on UVvisible absorbance (Fig. 8).
After settling for 5 days, for MNP-2 dispersed in water, MES,
PBS, UVvisible absorbance changes from 0.294, 0.303 and
0.31 to 0.089, 0 and 0, respectively (Fig. 8a). While for MNP-3,
it changes from 0.289, 0.303 and 0.306 to 0.168, 0.101 and 0.12,

In our surface double-exchange reactions, DMSA is used to

exchange with long-chain oleic acid on the surface of Fe3 O4

Fig. 7. Zeta potentials of MNP-2 and MNP-3.

Z.P. Chen et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 316 (2008) 210216

215

Fig. 9. (a) Schematic illustrations for the structure of MNP-2 and MNP-3 and
(b) water-solubility test of MNP-2 and MNP-3.

indicates that the surface of MNP-3 possesses more DMSA

molecules.
Fig. 8. Stability assays of MNP-2 (a) and MNP-3 and (b) dispersed in water,
RPMI-1640 (pH 7.0), PBS (pH 7.4) and MES buffer solution (pH 4.9), respectively.

nanoparticles. Furthermore, it is noticeable that alkalescent triethylamine plays an important role as catalyst. It can deprive
of hydrogen ions of carboxyl of DMSA to form carboxylate
which is preferable to exchange with long-chain oleic acid. The
surface double-exchange reaction without triethylamine was
carried out for comparison, and we found that the resulting
Fe3 O4 nanoparticles were not stable in water at high concentration (data not provided). DMSA is exchanged onto the surface
of nanoparticles by at least one carboxyl. DMSA first forms
a stable coating through its carboxylic chelating bonding and
further stability is obtained through intermolecular disulfide
cross-linking between DMSA molecules. The remaining carboxylates ensure surface charges and can be used for conjugating
with biological molecules and further applications. Fig. 9a is
schematic illustration for the structure of Fe3 O4 nanoparticles
modified by DMSA. We consider that the difference between
MNP-2 and MNP-3 is the amount of DMSA on the surface of
Fe3 O4 nanoparticles, as confirmed by TGA, VSM and watersolubility test (Fig. 9b). Although no precipitate is observed
when both of MNP-2 and MNP-3 are dispersed in water at
the high concentration of 10 mg/mL, MNP-2 is precipitated
thoroughly and MNP-3 remains stable after 10 days, which

4. Conclusions
Oleic acid capped monodisperse Fe3 O4 nanoparticles have
been successfully transferred into aqueous solution via surface
double-exchange with DMSA in the presence of triethylamine.
The characterization results show that surface double-exchange
is an efficient method that can introduce more DMSA molecules
onto the surface of nanoparticles through intermolecular disulfide cross-linking between DMSA. By this means, we obtain
stable nanoparticles with negative charges in the range of
pH = 114. Moreover, the resulting products can be dispersed
in water and other physiological solutions. Because DMSA is a
nontoxic and biocompatible product used in many applications,
high-quality magnetic nanoparticles modified by DMSA by this
means may be promising in biologically relevant applications.
Acknowledgments
This work has been carried out under financial support of the
National Natural Science Foundation of China (Nos. 60571031,
60501009 and 90406023) and National Basic Research Program of China (Nos. 2006CB933200 and 2006CB705600). The
author would like to thank Mr. A.Q. Xu, from the Analysis and Testing Centre of Southeast University for technique
assistances.

216

Z.P. Chen et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 316 (2008) 210216

References
[1] J.P. Dailey, J.P. Phillips, C. Li, et al., J. Magn. Magn. Mater. 194 (1999)
140.
[2] A. Jordan, R. Scholz, P. Wust, et al., J. Magn. Magn. Mater. 194 (1999)
185.
[3] L. Josephson, J.M. Perez, R. Weissleder, Angew. Chem. Int. Ed. 40 (2001)
3204.
[4] Y.-M. Huh, Y.-W. Jun, H.-T. Song, et al., J. Am. Chem. Soc. 127 (2005)
12387.
[5] Y.-W. Jun, Y.-M. Huh, J.-S. Choi, et al., J. Am. Chem. Soc. 127 (2005)
5732.
[6] T.K. Jain, M.A. Morales, S.K. Sahoo, et al., Mol. Pharm. 2 (2005) 194.
[7] A. Jordan, R. Scholz, P. Wust, et al., J. Magn. Magn. Mater. 201 (1999)
413.
[8] R.Z. Xu, Y. Zhang, M. Ma, et al., IEEE Trans. Magn. 43 (2007) 1078.
[9] S. Sun, H. Zeng, J. Am. Chem. Soc. 124 (2002) 8204.
[10] T. Hyeon, S.S. Lee, J. Park, et al., J. Am. Chem. Soc. 123 (2001) 12798.
[11] N.R. Jana, Y. Chen, X. Peng, Chem. Mater. 16 (2004) 3931.
[12] J. Park, K. An, Y. Hwang, et al., Nat. Mater. 3 (2004) 891.

[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]

Y. Wang, J.F. Wong, X. Teng, et al., Nano Lett. 3 (2003) 1555.

L.E. Euliss, S.G. Grancharov, S. OBrien, et al., Nano Lett. 3 (2003) 1489.
B.L. Frankamp, N.O. Fischer, R. Hong, et al., Chem. Mater. 18 (2006) 956.
T. Pellegrino, L. Manna, S. Kudera, et al., Nano Lett. 4 (2004) 703.
N. Fauconnier, J.N. Pons, J. Roger, et al., J. Colloid Interface Sci. 194
(1997) 427.
M. Auffan, L. Decome, J. Rose, et al., Environ. Sci. Technol. 40 (2006)
436.
F. Bertorelle, C. Wilhelm, J. Roger, et al., Langmuir 22 (2006) 5385.
M.P. Garcia, R. Miranda Parca, S. Braun Chaves, et al., J. Magn. Magn.
Mater. 293 (2005) 277.
I.J. Bruce, T. Sen, Langmuir 21 (2005) 7029.
A.L. Willis, N.J. Turro, S. OBrien, Chem. Mater. 17 (2005) 5970.
L. Zhang, R. He, H. Gu, Appl. Surf. Sci. 253 (2006) 2611.
Y. Ren, K.-I. Iimura, T. Kato, Langmuir 17 (2001) 2688.
Z. Li, L. Wei, M.Y. Gao, et al., Adv. Mater. 17 (2005) 1001.
A.G. Roca, M.P. Morales, K. Grady, et al., Nanotechnology 17 (2006) 2783.
T. Sato, T. Iijima, M. Seki, et al., J. Magn. Magn. Mater. 65 (1987) 252.
J. Popplewell, L. Sakhnini, J. Magn. Magn. Mater. 149 (1995) 72.
M. Mikhaylova, D.K. Kim, C.C. Berry, et al., Chem. Mater. 16 (2004) 2344.