Gold nanoparticles stabilized by tripod thioether oligomers: Synthesis and molecular dynamics studies

Article abstract of DOI:10.1246/bcsj.78.1773

Gold nanoparticles (1.5-1.7 nm) were prepared by use of three tripod thioether oligomers as stabilizing molecules. The particle sizes were independent of the sizes of the stabilizers, and were smaller than those stabilized with simpler thioethers. Molecul

Full text of DOI:10.1246/bcsj.78.1773

ꢀ 2005 The Chemical Society of Japan
Bull. Chem. Soc. Jpn., 78, 1773–1782 (2005) 1773
Gold Nanoparticles Stabilized by Tripod Thioether Oligomers:
Synthesis and Molecular Dynamics Studies
ꢀ;1;2
Youichi Hosokawa,1 Suguru Maki,1 and Toshi Nagata
1Institute for Molecular Science, 5-1 Higashiyama, Myodaiji, Okazaki 444-8787
2Department of Structural Molecular Science, The Graduate University for Advanced Studies,
5-1 Higashiyama, Myodaiji, Okazaki 444-8787
Received June 2, 2005; E-mail: toshi-n@ims.ac.jp
Gold nanoparticles (1.5–1.7 nm) were prepared by use of three tripod thioether oligomers as stabilizing molecules.
The particle sizes were independent of the sizes of the stabilizers, and were smaller than those stabilized with simpler
thioethers. Molecular dynamics simulations were performed to examine how the stabilizers adsorb on the nanoparticles.
Parameters were developed for modeling the interactions between gold and organic molecules. Simulations in explicit
chloroform solvent at 300 K revealed multiple-point adsorption of the stabilizers on the nanoparticles.
Recently, metal nanoparticles have received much attention
as potentially useful materials1 showing novel electronic prop-
erties derived from the quantum size effects,2 as well as inter-
esting catalytic reactivities.3 The pioneering work of Brust4
opened the field of monolayer protected gold nanoparticles,5
which exhibit excellent stability and tractability in organic sol-
vents. These materials are so stable that even chemical transfor-
mations of the pendant functional groups are possible through
usual techniques of organic syntheses.6 Thus, these materials
can be regarded as ‘‘molecules,’’ if somewhat exotic ones.
On the other hand, there is another class of metal nanopar-
ticles in solutions, which are stabilized by polymeric materials.
These nanoparticles are classical in concept, but still extremely
useful, particularly as catalysts.7 The most characteristic aspect
of this type of nanoparticles, compared with the monolayer
protected ones, is that the protecting molecules adsorb on the
surface of the nanoparticles through weak, non-covalent inter-
actions (van der Waals and/or electrostatic). The stability of
the nanoparticles is then ensured by both the multiple-point
interactions and the steric bulkiness of the polymers. The
absence of strong covalent bonding between the nanoparticle
surfaces and the protecting molecules is beneficial for these
particles to exhibit catalytic activities. There have been efforts
toward obtaining ‘‘molecularly dispersed’’ nanoparticles of this
class by use of large, well-defined molecules such as dendri-
mers.8 These approaches have been successful in presenting
new nanoparticles that exhibit interesting chemical and physi-
cal properties, whereas elucidation of their structures and dy-
namics at molecular levels is yet to be done.
simulations. Computational chemistry will thus provide a rea-
sonable basis for design and development of new organic–
nanoparticle composite materials.
Results
Synthesis of the Tripod Thioether Oligomers (TTO).
Scheme 1 shows the synthesis of the tripod thioether oligomers
1a–c. The synthetic scheme is straightforward; the key reac-
tion is the formation of dibenzyl sulfide function via a two-step
procedure: (i) benzyl bromide or mesylate was allowed to react
with thioacetic acid in the presence of DBU (1,8-diazabi-
cyclo[5.4.0]undec-7-ene, an organic base), (ii) the resulting
benzyl thioacetate was treated with hydroxide in the presence
of another benzyl halide. Similar procedures were utilized by
Ono et al. for synthesis of oligo(dibenzyl sulfide)s as precur-
sors of oligo(phenylene vinylene)s,10 and more recently by
Chow et al. for oligo(dibenzyl sulfide) dendrimers.11
Synthesis and Characterization of Gold Nanoparticles
Stabilized with TTO. The gold nanoparticles stabilized by
1a–c were prepared by the two-phase method4 with slight
modifications. The resulting deep brown solution was succes-
sively passed through a 0.2-mm syringe filter and an alumina
column, followed by three repeated centrifugations of a turbid
solution in CH2Cl2/EtOH at 10000 rpm for 5 min. The filtra-
tion and alumina chromatography removed the large or associ-
ated particles (monitored by disappearance of the plasmon
resonance at 520 nm in the UV–vis absorption spectra) and
unchanged tripod molecules (monitored by TLC); the centrifu-
gation removed the phase-transfer reagent Oct4NBr (moni-
1
In this article, we wish to report studies on gold nanoparti-
cles stabilized with tripod thioether oligomers 1a–c (hereafter
abbreviated as ‘‘TTO’’), and applications of molecular-dynam-
ic (MD) simulations for these materials. Since thioethers are
expected to adsorb on gold nanoparticles via non-covalent
interactions,9 these are classified as weakly protected metal
nanoparticles. The dynamic behavior of TTO on the nanopar-
ticles was examined by analyzing the trajectory of the MD
tored by H NMR). The purified materials, expressed as 1a–
Au, 1b–Au, and 1c–Au hereafter, were deep brown waxy oils
that were soluble in CHCl3, THF, and toluene, and insoluble in
MeOH, EtOH, and water. Both the chloroform solutions and
the evaporated waxy materials can be kept in a freezer under
Ar for several weeks without change. When the above proce-
dure was carried out without addition of 1a–c, only black pre-
cipitates and colorless supernatants formed. This suggests that
Published on the web October 4, 2005; DOI 10.1246/bcsj.78.1773

1774 Bull. Chem. Soc. Jpn., 78, No. 10 (2005)
Gold Nanoparticles Stabilized by Tripod Thioether Oligomers
O
O
O
O
O
O
O
O
i, ii
PBr3, Py
PBr3, Py
OH
CH2Cl2
R1
R2
HO
OH
Br
S
S
n
Br
HO
S
S
n
CH2Cl2
THF, 0 °C
1
2: R1 = R2 = Br, 28%
3: R1 = Br, R2 = OH, 36%
5a: n = 1, 74% from 4a
5b: n = 3, 66% from 4b
4a: n = 1, 79% from 3
i, ii
4b: n = 3, 91% from 5a
4c: n = 5, 60% from 5b
O
O
O
O
O
O
O
O
O
DHP, MSA
+
HO
S
S
n
OTHP
THPO
S
S
n
OTHP
HO
S
S
n
OH
THF, rt, overnight
7a: n = 1, 48%
7b: n = 3, 60%
7c: n = 5, 41%
6a: n = 1, 36%
6b: n = 3, 27%
6c: n = 5, 15%
4a: n = 1
4b: n = 3
4c: n = 5
O
O
OTHP
O
Br
S
1. MsCl/Et3N
CH2Cl2, 0 °C
S
Br
O
O
O
S
S
n
Br
7a-c
AcS
S
S
n
OTHP
O
2. i
NaOH/MeOH-H2O
THPO
O
S
O
O
THF, 0 °C
8a: n = 1, 58%
8b: n = 3, 57%
8c: n = 5, 87%
S
S
S
O
S
OTHP
O
1a: n = 1 (68%)
1b: n = 3 (52%)
1c: n = 5 (48%)
Conditions: i, AcSH, DBU, 2.5 h. ii, 3 (2 equiv.), NaOH/MeOH-H2O
Scheme 1.
standard deviations. These results suggest that the particle sizes
are mainly controlled by the thermodynamics and/or kinetics
of formation of the particles, rather than by the size and shape
of the stabilizing molecules. Interestingly, the sizes of these
nanoparticles were smaller than the sizes of other gold nanopar-
ticles stabilized by thioethers10 or other weakly binding stabi-
lizers (1.8–6.5 nm),13 and were close to the sizes of gold nano-
particles protected by straight-chain alkanethiols.4 We attribute
this preferential formation of small particles to the multiple-
point interaction between 1a–c and the nanoparticles.
Abs
1a-Au
1b-Au
1c-Au
1
0.5
300
500
700
wavelength / nm
The ICP-AES analyses of 1a–Au, 1b–Au, and 1c–Au gave
relative abundances of Au/S ¼ 0:77, 0.54, and 0.63 (in num-
ber of atoms). On the other hand, the mean number of gold
atoms in nanoparticles were 175 ꢁ 130, 187 ꢁ 170, and 126 ꢁ
82 for 1a–Au, 1b–Au, and 1c–Au, as estimated from the TEM
images. From these values, the number of protecting molecules
per nanoparticles were estimated as 25 ꢁ 19, 23 ꢁ 21, and
10 ꢁ 6. Although molecular models suggest that 1a–c are of
comparable sizes with the sizes of nanoparticles observed by
TEM, the observed number of protecting molecules per nano-
particle was much greater than unity. We will discuss later
about a plausible model to account for these results.
Molecular Dynamics. In order to get insights about the
structures of these composite materials, we carried out molec-
ular-dynamic simulations. Our study consisted of three stages:
(i) development of the parameters that describe the interaction
between the gold nanoparticles and TTO, (ii) preliminary
simulations of TTO alone, and (iii) simulations of the nanopar-
ticle–TTO composite systems. For the sake of simplicity, we
will hereafter limit our discussion to the smallest TTO mole-
cule (1a).
Fig. 1. UV–vis absorption spectra of 1a–Au, 1b–Au, and
1c–Au in CHCl3.
1a–c indeed worked as a stabilizer of gold nanoparticles. The
ligands 1a–c were slowly replaced with stronger ligands such
as alkanethiols. When 1c–Au (0.60 mg) was treated with 1-
dodecanethiol (0.43 mg) in CH2Cl2 (1 mL) for 7 d, free 1c
and dodecanethiol-protected gold nanoparticle were detected
1
by H NMR, UV–vis, and TLC (silica gel, CH2Cl2/EtOAc),
together with some unidentified decomposition products.
The UV–vis absorption spectra of the purified 1a–Au, 1b–
Au, and 1c–Au in CHCl3 are shown in Fig. 1, revealing the
absorbance of aromatic rings (280 nm) and gold nanoparticles
(400–800 nm). The absence of characteristic plasmon reso-
nances (520 nm) suggests that the particle sizes were less than
2 nm.12 Figure 2 shows the TEM images of gold nanoparticles
and the size-distribution histograms made by counting the
particles, together with the plot of the size-distribution curves
obtained from the SAXS analyses. The mean particle sizes
estimated from the TEM images were 1.5–1.7 nm, with no par-
ticular trends with the sizes of the protecting molecules. The
SAXS analyses gave somewhat smaller values, but the differ-
ences with the TEM results were well within the estimated
Development of Gold–TTO Parameters. There are many
studies on molecular-dynamic simulations of chemisorbed
thiols on a gold surface,14 and some studies on simulations

Y. Hosokawa et al.
Bull. Chem. Soc. Jpn., 78, No. 10 (2005) 1775
Table 1. The Pair-Specific van der Waals Parameters for the
Au(surface)–X Pairs
60
50
40
30
20
10
1a-Au
ꢀ
r0(X)/A
X
"(X)/(kcal/mol)
CMe
CCH
0.81
0.32
0.48
1.75
1.26
3.84
3.84
3.73
3.56
3.56
2
2
Csp
SSH
SSR
0
1
2
3
Diameter / nm
TEM 1.69 0.41 nm
SAXS 1.36 0.26 nm
where i covers the indices of non-hydrogen atoms in the
adsorbent molecule and ViðriÞ is the atom–surface potential
of the adsorbent atom i. Our next problem is to express
ViðriÞ as a realistic and computationally tractable function.
Beardmore et al.,18 in their MD simulations of alkanethiol
monolayers on Au(111), fitted the potential as a Morse function
of the distance between the adsorbent atom and the gold
surface. This is a reasonable formulation, but in our case it is
not applicable because the distance from the ‘‘surface of the
nanoparticle’’ can not be defined unequivocally. Instead, we
use a pair-specific van der Waals potential function between
the adsorbent atom and the surface gold atom, and assume that
ViðriÞ can be expressed as the sum of the pair potentials over all
the surface gold atoms. Then Eq. 1 becomes:
50
40
1b-Au
30
20
10
0
1
2
3
Diameter / nm
TEM 1.72 0.64 nm
SAXS 1.18 0.32 nm
60
50
40
30
20
1c-Au
(
)
ꢀ
ꢁ
12ꢂ2
ꢀ
ꢁ
6
X X
where i and j are the indices of the adsorbent (non-hydrogen)
r0ðXiÞ
rij
r0ðXiÞ
rij
Vads
¼
"ðXiÞ
;
ð2Þ
i
j
10
atoms and the surface gold atoms, respectively, Xi is the atom
2
0
1
2
3
Diameter / nm
type of the adsorbent atom i (which is one of CMe, CCH , Csp
,
2
TEM 1.51 0.40 nm
SAXS 1.42 0.15 nm
SSH, and SSR, corresponding to the five adsorbent atom types
described above), and "(X) and r0(X) are the pair-specific
van der Waals parameters which depend on the adsorbent atom
types. The values of "(X) and r0(X) were optimized so that the
calculated adsorption energies of selected molecules matched
the experimental enthalpies (see: Computational Methods for
details). The obtained values are shown in Table 1.
Simulations of TTO. Prior to the simulations of the nano-
particle–TTO systems, we examined the molecular-dynamic
simulations of TTO alone. The purpose of this preliminary
study was to examine the effect of the solvent on the confor-
mation of TTO. Thus, we compared two independent simula-
tions of 1a, one without solvent and the other with explicit
CHCl3 solvent, starting from the same, arbitrary initial confor-
mation. Figure 3 shows the changes of the distances between
the center of the central benzene ring and the three acetal car-
bons of the THP groups. These distances conveniently indicate
the conformations of the three legs of the tripod molecule.
Without solvent (part (a)), all the legs quickly collapsed to
Fig. 2. TEM images and histograms with SAXS plots (in
THF) of 1a–Au, 1b–Au, and 1c–Au.
of naked gold nanoparticles,15 but there are very few reports on
simulations of composites of gold nanoparticles and organic
molecules.16 In particular, there are no reports on simulations
of physisorbed organic molecules on gold nanoparticles.
Therefore, we first need to develop reasonable parameters that
can describe the interactions between the gold nanoparticles
and physisorbed organic molecules.
Because the physisorption energies of organic molecules
on gold nanoparticles have not been measured, we used the
physisorption energies on the Au(111) surface as the basis of
our parameter development. Lavrich and co-workers17 mea-
sured the physisorption enthalpies of various sulfur-containing
molecules on the Au(111) surface. Their results revealed that
the physisorption enthalpies could be predicted as the sum of
the following values (in kJ/mol) for each group which directly
contacts with the surface: 15.5 (CH3–), 6.1 (–CH2–), 9.1 (sp2
–CH=), 33.5 (HS–), and 24.1 (–S–). This empirical formula
implies that the adsorption energy is additive in regard to the
number and kind of the surface-contacting non-hydrogen
atoms. Thus we can reasonably assume that the adsorption
potential energy is the sum of the atom–surface interaction
ꢀ
folded conformations (distances less than <15 A) and never
got unfolded again, because of strong intramolecular van der
Waals attractive interactions. On the other hand, with solvent
(part (b)) each leg went back and forth between the folded
and extended conformations during 4 ꢃ 106 MD steps (¼8
ns). Collapse was avoided because intramolecular van der
Waals interaction was compensated by the solvation energy.
These results suggest that the solvent must be included in
the MD simulation of these systems. We carried out all the
following simulations with explicit solvents.
potentials:
X
Vads
¼
ViðriÞ;
ð1Þ
i

1776 Bull. Chem. Soc. Jpn., 78, No. 10 (2005)
Gold Nanoparticles Stabilized by Tripod Thioether Oligomers
(a)
(a)
a
b
c
140
120
100
80
25
TTO 5
TTO 3
30
20
15
25
10
5
0
20
0
100 200 300 400 500
15
10
5
60
40
TTO 8
Other TTOs
20
0
0
0
500 1000 1500 2000 2500
MD steps/1000
0
1000
2000
3000
4000
MD steps/1000
(b)
(b)
140
120
100
80
30
25
20
15
10
5
60
40
20
0
0
0
500 1000 1500 2000 2500
MD steps/1000
0
1000
2000
MD steps/1000
3000
4000
Fig. 4. The contact numbers of (a) 1a molecules and (b) the
gold nanoparticle in the simulation of 16 1a and Au147
with explicit CHCl3 solvent. The vertical lines noted a,
b, and c in part (a) correspond to the snapshots shown in
Fig. 5.
Fig. 3. Distances between the center of the central benzene
ring and the acetal carbons of the THP groups in the MD
trajectory at 300 K. (a) 1a in a vacuum, (b) 1a with explic-
it CHCl3 solvent. The three traces correspond to the three
‘‘legs’’ of 1a.
Simulations of Gold Nanoparticle–TTO Systems. Simu-
lation of gold nanoparticle–TTO systems were carried out by
placing TTO molecules around a naked gold nanoparticle
and subjecting the whole system to MD runs with explicit
CHCl3 solvent at 300 K. Although this does not reproduce
the real process under the experimental conditions (which
includes simultaneous growth of gold nanoparticles and
adsorption of TTOs), it will give us useful information about
the possible structures of the resulting composites of gold
nanoparticles and TTOs. A cuboctahedron Au147 (diameter
1.7 nm) was used as a model of the gold nanoparticle.
adsorbed to the gold nanoparticle. As more atoms in the
TTO molecules get adsorbed, the contact number of the gold
nanoparticle (part (b)) monotonically increases to 88–92,
which means that more than 95% of the nanoparticle surface
is covered with TTOs (the number of surface atoms in Au147
is 92). We performed two more runs under similar conditions
with different starting locations and orientations of the TTO
molecules; the numbers of adsorbed TTO molecules were 3
and 4, and the final contact numbers of gold nanoparticles were
85–87 and 90–92. These results indicate that 3–4 molecules of
TTO (1a) are sufficient to cover the entire surface of an Au147
nanoparticle.
Figure 5 illustrates how the TTO molecules are adsorbed on
the surface of a gold nanoparticle. Three representative snap-
shots were selected so that one, two, and three TTO molecules
were adsorbed (later we will discuss the 1:1 complexes in
more detail). It is easily seen that the surface is completely
covered with three TTO molecules in this case. Figure 5 also
illustrates the distribution of contacting atoms in each adsorb-
ing TTO molecule, indicating the multiple-point interaction
between the gold nanoparticle and TTOs. It should also be
noted that not only the sulfur atoms but also the carbon atoms
(both aromatic and benzylic) participate in adsorption. This is
reasonable because the pair-specific van der Waals parameters
Figure 4 shows the results from a single run, in which 16
TTO molecules (1a) and an Au147 nanoparticle were placed
ꢀ 3
in a box of CHCl3 with dimensions 112:2 ꢃ 96:3 ꢃ 97:7 A .
Part (a) shows the plots of ‘‘contact numbers’’ of the individual
TTO molecules versus the MD steps, where the contact num-
ber of a TTO molecule is defined by the number of atoms of
ꢀ
the TTO molecule that are within 3.84 A of any surface atom
ꢀ
in the gold nanoparticle (the choice of 3.84 A as the cutoff
value is based on the pair-specific van der Waals parameters
in Table 1). Part (b) shows the contact number of the gold
nanoparticle, which is defined by the number of the surface
ꢀ
gold atoms that are within 3.84 A of any atom in any TTO
molecule. From part (a), we see that 3 TTO molecules are

Y. Hosokawa et al.
Bull. Chem. Soc. Jpn., 78, No. 10 (2005) 1777
(a)
TTO5
TTO5
TTO3
(b)
TTO3
TTO5
TTO5
(c)
TTO8
TTO3
TTO3
TTO8
TTO5
TTO5
Fig. 5. Three selected snapshot structures from the 1a16/Au147 simulation shown in Fig. 4. Left column: the space-filling views
from two opposite directions (green: gold, cyan: carbon, white: hydrogen, red: oxygen, and yellow: sulfur). Middle column:
the space-filling views with different colors for different molecular fragments; the gold nanoparticle is shown in white, and the
TTO (1a) molecules are shown in blue, yellow, and gray. Right column: the contacting atoms in each 1a molecule (black circles).
Parts (a), (b), and (c) correspond to the vertical lines shown in Fig. 4a.
in Table 1 indicate that both aliphatic and aromatic carbon
atoms have significant interaction energies (25–65% of that
of thioether sulfur atom).
the van der Waals interaction energies (not including the sol-
vents) at each MD step. As expected, the van der Waals ener-
gies decrease with increasing contact numbers due to the
attractive force between 1a and Au147. Figure 7 illustrates
how each single 1a covers the surface of Au147, by showing
the conformations of the Au147–1a composites and the contact-
ing atoms in the 1a molecule at the timesteps which gave the
largest (i.e. the most negative) van der Waals interaction ener-
gies for each run. Table 2 compiles the timesteps, interaction
energies, and the contact numbers of 1a and Au147 for the
snapshots shown in Fig. 7. The conformations from Runs 1
and 3 show that all three legs of the tripod molecule 1a bind
to the Au147 surface via multiple-point interactions, whereas
The next question we would like to answer is this: how
many surface gold atoms can a single TTO molecule cover?
The results in the above paragraph suggest an average number
of 92=3 ꢄ 31, but this is the number when multiple TTO mole-
cules are bound simultaneously. To examine the coverage by a
single TTO molecule, we performed simulations of a 1:1 com-
plex of Au147 and 1a. Figure 6 shows the results from three
independent runs with different starting orientations. The final
contact numbers of Au147 range from 40 to 65. This corre-
sponds to 45–70% coverage. Part (c) of Fig. 6 also shows

1778 Bull. Chem. Soc. Jpn., 78, No. 10 (2005)
Gold Nanoparticles Stabilized by Tripod Thioether Oligomers
Run 1
(a)
Run 1
160
140
120
100
80
Run 3
Run 2
1200
60
40
20
0
0
400
800
MD steps/1000
Run 2
(b)
60
50
40
30
20
10
0
Run 3
0
400
800
1200
MD steps/1000
(c)
50
0
-50
-100
Fig. 7. The structures of lowest interaction energies in the
three independent simulations of 1:1 1a and Au147. The
left column shows the space-filling views from two oppo-
site directions (green: gold, cyan: carbon, white: hydro-
gen, red: oxygen, and yellow: sulfur) and the right column
shows the contacting atoms in 1a (black circles).
0
400
800
1200
MD steps/1000
Fig. 6. The contact numbers of (a) 1a and (b) Au147, and (c)
the interaction energy between 1a and Au147, in the three
independent simulations of 1:1 1a and Au147 with explicit
CHCl3 solvent.
Table 2. The Energies and Contact Numbers at the Maxi-
mum Interaction StepsaÞ in 1a/Au147
the conformation from Run 2 shows that one of the legs
remains almost unbound. It is likely that Run 2 would have
given a similar result with Runs 1 and 3 if the simulation
was performed for a longer period; actually, Figures 6a and
6b indicate that Run 1 was temporary trapped in a similar con-
formation as the final state of Run 2 during 4{6 ꢃ 105 MD
steps, and then ‘‘escaped’’ to the more stable conformations.
This type of ‘‘trapping’’ phenomena, which is caused by the
presence of many local minima in the potential energy surface,
is a well-known problem in simulations of large molecules.19
MD step Interaction energiesbÞ
Contact numbers
Run
/1000
/kcal molꢂ1
1a
Au147
1
2
3
1457
971
1386
ꢂ124:2
ꢂ82:4
149
96
129
64
46
54
ꢂ123:8
a) The ‘‘maximum interaction step’’ is defined as the MD step
in which the van der Waals interaction energy of the 1a/Au147
complex is the lowest. b) The van der Waals interaction ener-
gies of the 1a/Au147 complex (not including the solvent).

Y. Hosokawa et al.
Bull. Chem. Soc. Jpn., 78, No. 10 (2005) 1779
1H and 13C NMR spectra were recorded on JEOL Lambda 500.
Melting point were determined by Yanaco MP-500V (uncorrected
values). UV–vis spectra were measured by Shimadzu UV-
2500PC. Transmission electron microscopy (TEM) images were
recorded by the JEOL JEM-2000EX. Medium-pressure column
chromatography was carried out by use of Wakogel C-200 or
neutral alumina Merck Art. 1097 Aluminiumoxid 90. Analytical
thin layer chromatography (TLC) was performed on Merck TLC
aluminum sheets (silica gel RP-18 F254 and aluminum oxide 60
There are various techniques to address this problem, and
applications of these techniques to the present system are
currently under way.
Discussion
The MD simulations showed that 3–4 molecules of 1a were
sufficient to cover the surface of a 1.7-nm gold nanoparticle,
whereas the experiments showed as many as 25 molecules of
1a were present per one gold nanoparticle. Although it is not
impossible to accommodate more than 20 1a’s around the sur-
face of a gold nanoparticle, the more appropriate model would
be coexistence of free, unbound 1a molecules and gold nano-
particles covered with a small number of 1a. The free and
bound 1a molecules should be in dynamic equilibrium in
CHCl3 solutions, as indicated by the following observations:
(i) the 1H NMR spectra only showed one set of slightly broad-
ened signals which were characteristic of nanoparticle-bound
molecules and (ii) the gel permeation chromatography of
1a–Au showed a complex feature with many peaks, and when
a fraction corresponding to one peak was separated and inject-
ed again to the same GPC apparatus, a similar complex chro-
matogram was obtained again.
The sizes of TTOs are comparable to those of gold nanopar-
ticles, if a little too small in the case of 1a. Hence, in principle
it is possible to design the TTO to form 1:1 complexes with
gold nanoparticles, like in the beautiful work by Inomata and
Konishi.20 However, in the present study, the molecular design
of TTOs was too crude to realize such well-defined compo-
sites. De Cola et al.21 reported gold nanoparticles by use of a
series of oligo(thioether) dendritic compounds with a similar
aim, but even their extensive study did not lead to conclusive
evidence for 1:1 complexes. On the other hand, works of
Watanabe22 and Aida23 introduced a concept of well-defined
composites of nanoparticles and proteins with predetermined
hollows. The use of proteins is beneficial in that they have
well-defined structures and are even controllable by use of
biological functions, but such use has the drawback that their
designs are determined by nature and modification of their
functions is limited. Admittedly, the apparent problem in
designing the artificial molecules from scratch to form well-
defined complexes with nanoparticles is the great conforma-
tional flexibility that makes prediction of the structure too dif-
ficult. The lack of experimental methods to examine the solu-
tion structure is also problematic.
F
254). Preparative gel permeation chromatography (GPC) was car-
ried out on LC-908 (Japan Analytical Industry Co.) equipped with
the polystyrene-based columns JAIGEL-2.5H and 2H (CHCl3,
3.5 mL/min). Small-angle X-ray scattering was measured on a
Rigaku ATX diffractometer. The average diameter and variance
were obtained by assuming the ꢀ-distribution function for the size
distribution of the nanoparticles and fitting the observed scattering
intensities to the analytical expression.24
Synthetic Procedures. Only general procedures are described
in the following. The amounts of reagents for the individual steps
and the compound data for characterization can be found in the
Supporting Information.
Bromination with PBr3. A solution of the diol in pyridine/
CH2Cl2 was placed in a round-bottomed flask with a magnetic
stirring bar and a three-way stopcock. Phosphorus tribromide
was slowly added at 0 ꢅC and the reaction mixture was stirred
under argon at rt. After 24 h, the reaction mixture was quenched
with water and extracted with CH2Cl2. The organic phase was
dried over anhydrous MgSO4. The solvent was evaporated, and
the residue was purified by medium-pressure column chromatog-
raphy (CH2Cl2).
Thioacetylation (Step (i)). A mixture of the bromide (or
mesylate), thioacetic acid, and THF was placed in a two-necked
flask with a magnetic stirring bar, a three-way stopcock, and a
pressure-equalizing dropping funnel protected with a rubber
ꢅ
septum. A THF solution of DBU was dropped slowly at 0 C un-
der argon, and the reaction mixture was stirred at rt. After 5 h, the
reaction mixture was quenched with water and extracted with
hexane. The organic phase was dried over anhydrous MgSO4.
The solvent was evaporated to give the desired thioacetate as a
yellow oil, which was used in the next reaction without further
purification.
Preparation of Dibenzyl Sulfides (Step (ii)). A solution of
the thioacetate and the bromide in THF was placed in a round-
bottomed flask with a magnetic stirring bar and a pressure-equal-
izing dropping funnel equipped with a three-way stopcock. Argon
gas was bubbled through the solution for 30 min. A solution of
NaOH in MeOH/H2O (v/v ¼ 4=1) was added dropwise under
argon at 0 ꢅC, and the solution was stirred overnight at rt. The
reaction mixture was quenched with water and extracted with
CH2Cl2. The organic phase was dried over anhydrous MgSO4.
The solvent was evaporated, and the residue was purified by
medium-pressure column chromatography (hexane/EtOAc).
Monoprotection of Diols by THP. A solution of the diol and
3,4-dihydro-2H-pyran in THF was placed in a round-bottomed
flask with a magnetic stirring bar and a three-way stopcock with
a rubber septum. Methanesulfonic acid was added at rt under
argon, and the reaction mixture was stirred overnight at rt. The
reaction mixture was quenched with water and extracted with
CH2Cl2. The organic phase was dried over anhydrous MgSO4.
The solvent was evaporated, and the diol, the mono-THP ether,
and the di-THP ether were separated by medium-pressure col-
umn chromatography (hexane/EtOAc). The di-THP ether was
Molecular dynamic simulations can be a powerful tool to
overcome these difficulties. Although the MD methods also
have their specific technical problems (among them is the
‘‘trapping in the local minima’’ problem mentioned earlier),
they have already been used extensively for studying dynamics
of large molecules in solutions, including proteins. The present
article demonstrates the applicability of MD to the gold nano-
particles stabilized with thioether organic molecules, and our
ongoing studies will unveil more power of these methods in
development of new soft nanomaterials.
Experimental
General. Tetrahydrofuran (THF) and diethyl ether were dis-
tilled from sodium diphenylketyl. Dichloromethane was distilled
from calcium hydride. Other chemicals were used as received.

1780 Bull. Chem. Soc. Jpn., 78, No. 10 (2005)
Gold Nanoparticles Stabilized by Tripod Thioether Oligomers
converted back to the diol for recycle by treatment with p-tolu-
enesulfonic acid (0.25 equiv per THP group) in CH2Cl2/MeOH
(v/v ¼ 2=1, rt, overnight).
enthalpies were obtained in the following way: (i) a planar
hexagonal array of 169 gold atoms with the interatomic dis-
ꢀ
tance of 2.88 A was prepared, (ii) the target molecule (one
Preparation of Mesylates. A mixture of the alcohol, meth-
anesulfonyl chloride, and CH2Cl2 was placed in a round-bottomed
flask with a magnetic stirring bar and a three-way stopcock. Tri-
ethylamine was added at 0 ꢅC under argon atmosphere, and the
reaction mixture was stirred for 2–4 h. The reaction mixture
was quenched with water and extracted with CH2Cl2. The organic
phase was dried over anhydrous MgSO4. The solvent was evapo-
rated to give the corresponding mesylate as colorless oil, which
was used in the next reaction without further purification.
Preparation of Gold Nanoparticles Stabilized with TTO.
An aqueous solution of HAuCl4 (2.25 mM, 1 mL) was combined
with a CHCl3 solution (1 mL) of Oct4NBr (2.25 mmol) and the
tripod molecules 1a–c (1a: 0.55 mg, 0.25 mmol; 1b: 0.50 mg,
0.15 mmol; 1c: 0.48 mg, 0.11 mmol). The two-phase mixture
was vigorously stirred (ca. 1500 rpm) for 10 min at rt until the
aqueous layer turned to colorless and the organic layer turned to
yellow. To this mixture, a freshly prepared aqueous solution of
NaBH4 (75 mM, 30 mL) was added with vigorous stirring. After
10 min, the color of organic layer changed to deep brown. The
organic layer was filtered through a 0.2-mm syringe filter and then
purified by alumina column chromatography eluting with 4:1
dichloromethane–methanol, followed by three repeated centrifu-
gations of a turbid solution in CH2Cl2/EtOH at 10000 rpm
for 5 min.
ꢀ
of the above nine) was placed along the plane 4.0 A above
the gold plane, (iii) the total energy of the system (E0) was cal-
culated with the surface interaction potential (Eq. 2) switched
off, (iv) the surface interaction potential was switched on and
the system was subjected to molecular dynamic simulation at
300 K for 2 fs ꢃ 10000 steps, then to the energy minimization
(E1), and (v) the calculated physisorption enthalpy ꢁHcalc was
obtained as E1 ꢂ E0.
Optimization of "(X) and r0(X) was carried out as follows.
Since only 9 experimental values were available to determine
10 parameters, all parameters cannot be optimized independ-
ently. Instead, four coefficients (K1, K2, K3, and K4) were
introduced and "(X) and r0(X) were expressed as follows:
"ðXÞ ¼ K1 AðXÞ;
ð3Þ
ꢆ
r0ðXÞ ¼ Kn ðrvdWðXÞ þ rvdWðAuÞÞ
ꢆ
ðn ¼ 2 for X ¼ CMe and CCH , n ¼ 3
2
2
for X ¼ Csp , n ¼ 4 for X ¼ SSH and SSRÞ;
ð4Þ
where A(X) is the empirical adsorption energy per group
consisting of an atom X and attached hydrogens (Ref. 17;
the values are also quoted in the text), and rvdW(X) is the
van der Waals radius of atom X. The following iterative algo-
rithm was employed to find the optimized set of coefficients
K1–K4:
Computational Methods
General. The initial geometries of the organic molecules
were obtained by building the molecular model with the
Chem3D program package (CambridgeSoft). The structure
and coordinate files of appropriate format were generated by
a homemade program suite named PSFLIB. The atom types
and force field parameters were taken from the parm99 force
field distributed with AMBER 7.25 The RESP (restrained elec-
trostatic potential fit) atomic charge26 were obtained by quan-
(1) Initial guess values of K1 ¼ 0:22, K2 ¼ K3 ¼ K4 ¼ 1:0
and a ‘‘scaling factor’’ k ¼ 0:01 are set.
(2) The ꢁHcalc of the nine model molecules and the squared
2
sum of residuals R ¼ ꢂðꢁHcalc ꢂ ꢁHexptÞ are calculated. The
initial value of R is kept as Rini.
(3) A trial move is attempted by the amount ꢁKi ¼ ðRanf ꢂ
0:5Þ kR=Rini, where Ranf is a uniform random number in [0, 1]
ꢆ
(different Ranf for different i).
(4) For the new set of Ki0 ¼ Ki þ ꢁKi, ꢁHcalc and R are
calculated.
ꢀ
tum mechanical calculation (HF/6-31G ) with GAMESS,27
followed by treatment with the RESP module of AMBER with
the aid of the R.E.D. script.28 The molecular dynamics calcu-
lation was performed with NAMD package.29 A timestep of 2
fs was used. The temperature was maintained by reassigning
the velocities every 500 steps. The RATTLE constraint30
(5) If R is smaller than the previous value, then this trial is
accepted and Ki valus are updated. Return to (4) with the same
trial move ꢁKi.
(6) If R is larger than the previous value, then this trial is
rejected and Ki are unchanged. Return to (3) to try different
direction.
(7) If the probability of rejection becomes too high (>0:8),
then k is reduced (multiplied by 0.25). If k is sufficiently small,
then the local minimum is achieved and the calculation ends.
The final R was 3.10 kcal2/mol2, and the maximum/
minimum of residuals were þ0:97 and ꢂ1:28 (in kcal/mol),
respectively.
ꢀ
was applied to all hydrogen atoms. A 10-A cutoff was applied
to non-bonding interactions. Simulation with explicit solvents
was performed under periodic boundary conditions with a con-
stant volume, and the long-range electrostatic forces were
evaluated by the particle mesh Ewald method.31 The trajectory
was analyzed with VMD32 and homemade programs. Calcula-
tions were performed on a Silicon Graphics SGI2800 computer
of Research Center of Computational Science (Okazaki Re-
search Facilities, National Institutes of Natural Sciences) and
a local cluster of Apple PowerMac G5 machines.
Development of Gold–TTO Parameters. The follow-
ing molecules were used to determine the gold–TTO parame-
ters: CH3SH, CH3CH2SH, CH3(CH2)3SH, CH3(CH2)5SH,
CH3(CH2)7SH, CH3(CH2)8SH, (C2H5)2S, (C4H9)2S, and thio-
phene. The experimental values of physisorption enthalpies
(ꢁHexpt) of these molecules on Au(111) were taken from
Ref. 17. On the other hand, the ‘‘calculated’’ physisorption
Simulations of TTO.
To calculate the RESP charges
efficiently, we divided the target molecule 1a into ten frag-
ments so that each fragment contained one benzene ring, and
the charges were calculated for each fragment plus appropriate
capping groups. The structure of each fragment and the result-
ing RESP charges can be found in the Supporting Information.
After the charges were assigned to all atoms in 1a, a rectan-
ꢀ
gular box of CHCl3 was added providing at least 10 A of
CHCl3 around any solute atom. The atomic coordinates and

Y. Hosokawa et al.
Bull. Chem. Soc. Jpn., 78, No. 10 (2005) 1781
parameters for CHCl3 were taken from Caldwell’s work.33 The
system was subjected to energy minimization for 1000 steps,
gradual heating from 0 to 300 K over 15000 steps, preequili-
bration at 300 K for 250000 steps, and production run at 300
K for 4000000 steps. For the vacuum simulation, a snapshot
was taken from the solvated simulation after the preequilibra-
tion, all the solvent molecules were removed, and the resulting
naked 1a was subjected to production run at 300 K (without
periodic boundary conditions) for 4000000 steps.
Simulations of Gold/TTO Systems. A set of atomic
coordinates for an ideal cuboctahedral Au147 was created with
ꢀ
the Au–Au distances of 2.88 A (= the interatomic distance in
metallic gold). In the present study, the conformational change
of the gold nanoparticle is not taken into account; hence, Au–
References
a) D. L. Feldheim, C. A. Foss, Jr., and M. Dekker, ‘‘Metal
Nanoparticles,’’ New York (2002). b) M.-C. Daniel and D. Astruc,
Chem. Rev., 104, 293 (2004).
2 J. J. Pietron, J. F. Hicks, and R. W. Murray, J. Am. Chem.
Soc., 121, 5565 (1999).
1
3 P. Claus, A. Brueckner, C. Mohr, and H. Hofmeister,
J. Am. Chem. Soc., 122, 11430 (2000).
4
M. Brust, M. Walker, D. Bethell, D. J. Schiffrin, and R.
Whyman, Chem. Commun., 1994, 801.
A. C. Templeton, W. P. Wuelfing, and R. W. Murray, Acc.
Chem. Res., 33, 27 (2000).
5
6 J. G. Worden, A. W. Schaffer, and Q. Huo, Chem.
Commun., 2004, 518.
ꢀ
Au bonds with equilibrium length 2.88 A and force constant
ꢀ
200 kcal/mol/A were assumed between all neighboring gold
7
a) A. B. R. Mayer and J. E. Mark, ACS Symp. Ser., 622,
137 (1996). b) N. Toshima, NATO ASI Ser., Ser. 3, 12, 371 (1996).
atoms. Bond angles and dihedrals were not included. Different
atom types were assigned for the surface and internal gold
atoms, so that the pair-specific van der Waals potentials (see:
text for details) between the gold atoms and the TTO atoms
should work only for the surface gold atoms. Ninety-two gold
atoms out of 147 were assigned as the surface atoms.
The starting structures of the simulation of 16 TTOs and
Au147 were generated in the following protocol: (i) the Au147
was placed at (0, 0, 0), (ii) 16 points were generated by use
of the Halton sequence of quasi-random points34 within a cube
8
R. M. Crooks, M. Zhao, L. Sun, V. Chechik, and L. K.
Young, Acc. Chem. Res., 34, 181 (2001).
a) X.-M. Li, M. R. de Jong, K. Inoue, S. Shinkai, J.
9
Huskens, and D. N. Reinhoudt, J. Mater. Chem., 11, 1919
(2001). b) E. J. Shelley, D. Ryan, S. R. Johnson, M. Couillard,
D. Fitzmaurice, P. D. Nellist, Y. Chen, R. E. Palmer, and J. A.
Preece, Langmuir, 18, 1791 (2002).
10 N. Ono, H. Tomita, and K. Maruyama, J. Chem. Soc.,
Perkin Trans. 1, 1992, 2453.
11 H.-F. Chow, M.-K. Ng, C.-W. Leung, and G.-X. Wang,
J. Am. Chem. Soc., 126, 12907 (2004).
ꢀ
of 78-A edges centered at (0, 0, 0), (iii) for each of the 16
12 a) M. M. Alvarez, J. T. Khoury, T. G. Schaaff, M. N.
Shafigullin, I. Vezmar, and R. L. Whetten, J. Phys. Chem. B,
101, 3706 (1997). b) T. G. Schaaff, M. N. Shafigullin, J. T.
Khoury, I. Vezmar, R. L. Whetten, W. G. Cullen, P. N. First, C.
points, a random conformation was selected from the trajecto-
ry of the simulation of solvated 1a and placed at the point, and
the placed molecule was rotated by random angle until no
atom collisions were found, and (iv) if the molecule could
not placed in any orientation, this point was discarded and
another point was generated. The mean intermolecular dis-
´
´
Gutierrez-Wing, J. Ascensio, and M. J. Jose-Yacaman, J. Phys.
Chem. B, 101, 7885 (1997).
13 a) S. Chen and R. W. Murray, Langmuir, 15, 682 (1999). b)
R. Wang, J. Yang, Z. Zheng, M. D. Carducci, J. Jiao, and S.
Seraphin, Angew. Chem., Int. Ed., 40, 549 (2001). c) M.-K.
Kim, Y.-M. Jeon, W. S. Jeon, H.-J. Kim, S. G. Hong, C. G. Park,
and K. Kim, Chem. Commun., 2001, 667. d) Y. Zhou, H. Itoh,
T. Uemura, K. Naka, and Y. Chujo, Langmuir, 18, 277 (2002).
14 a) B. Park, M. Chandross, M. J. Stevens, and G. S. Grest,
Langmuir, 19, 9239 (2003). b) Y. Morikawa, T. Hayashi, C. C.
Liew, and H. Nozoye, Hyomen Kagaku, 23, 423 (2002). c) M.
Tarek, K. Tu, M. L. Klein, and D. J. Tobias, Biophys. J., 77,
964 (1999). d) H. H. Jung, Y. D. Won, S. Shin, and K. Kim,
Langmuir, 15, 1147 (1999). e) T.-W. Li, I. Chao, and Y.-T.
Tao, J. Phys. Chem. B, 102, 2935 (1998). f) T. Bonner and A.
Baratoff, Surf. Sci., 377–379, 1082 (1997). g) Z. Zhang, T. L.
Beck, J. T. Young, and F. J. Boerio, Langmuir, 12, 1227 (1996).
h) J. Hautman and M. L. Klein, J. Chem. Phys., 91, 4994 (1989).
15 a) T. X. Li, S. M. Lee, S. J. Han, and G. H. Wang, Phys.
Lett. A, 300, 86 (2002). b) H.-S. Nam, N. M. Hwang, B. D. Yu,
and J.-K. Yoon, Phys. Rev. Lett., 89, 275502 (2002).
16 a) K. L. Yin, Q. Xia, H. T. Xi, D. J. Xu, X. Q. Sun, and
C. L. Chen, THEOCHEM, 674, 159 (2004). b) B. Arezki, A.
Delcorte, A. C. Chami, B. J. Garrison, and P. Bertrand, Nucl.
Instrum. Methods Phys. Res., Sect. B, 212, 369 (2003).
17 D. J. Lavrich, S. M. Wetterer, S. L. Bernasek, and G.
Scoles, J. Phys. Chem. B, 102, 3456 (1998).
18 K. M. Beardmore, J. D. Kress, N. Grønbech-Jensen, and
A. R. Bishop, Chem. Phys. Lett., 286, 40 (1998).
19 a) U. H. E. Hansmann and Y. Okamoto, J. Comput. Chem.,
ꢀ
tance was about 40 A. To the resulting set of molecules, a rec-
ꢀ
tangular box of CHCl3 was added, providing at least 4 A of
CHCl3 around any solute atom, and the system was subjected
to the following sequence of simulations: (i) energy minimiza-
tion for 1000 steps, (ii) gradually heating from 0 to 300 K over
15000 steps, and (iii) production run at 300 K for 3000000
steps. The simulation of 1:1 complexes of TTO and Au147
was carried out in a similar manner, except that only one
TTO was placed around the gold nanoparticle.
We thank Dr. Osamu Oishi (IMS) for TEM measurement,
Mr. Akito Sasaki (Rigaku Co.) for SAXS measurement and
analysis, and the Research Center for Computational Science
of Okazaki Research Facilities, National Institutes of Natural
Sciences (NINS) for the use of the SGI2800 computer. This
work was supported by a Grant-in-Aid for Scientific Research
in Priority Area, ‘‘Reaction Control of Dynamic Complexes’’
(No. 16033262) and ‘‘Nanotechnology Support Project’’ of
the Ministry of Education, Culture, Sports, Science and Tech-
nology (MEXT), Japan.
Supporting Information
Details of synthetic procedures and compound data (6 pages),
and the details of RESP charge calculations (1 page). This mate-
rial is available free of charge on the web at http://www.csj.jp/
journals/bcsj/.

1782 Bull. Chem. Soc. Jpn., 78, No. 10 (2005)
Gold Nanoparticles Stabilized by Tripod Thioether Oligomers
18, 920 (1997). b) A. F. Voter, F. Montalent, and T. C. Germann,
Annu. Rev. Mater. Res., 32, 321 (2002).
20 T. Inomata and K. Konishi, Chem. Commun., 2003, 1282.
21 A. D’Aleo, R. M. Williams, F. Osswald, P. Edamana, U.
Hahn, J. van Heyst, F. D. Tichelaar, F. Vogtle, and L. De Cola,
¨
Adv. Funct. Mater., 14, 1167 (2004).
22 T. Ueno, M. Suzuki, T. Goto, T. Matsumoto, K.
Nagayama, and Y. Watanabe, Angew. Chem., Int. Ed., 43, 2527
(2004).
23 D. Ishii, K. Kinbara, Y. Ishida, N. Ishii, M. Okochi, M.
Yohda, and T. Aida, Nature (London), 423, 628 (2003).
24 a) K. Omote, Y. Ito, and S. Kawamura, Appl. Phys. Lett.,
82, 544 (2003). b) O. Nagao, G. Harada, T. Sugawara, A. Sasaki,
and Y. Ito, J. Appl. Phys., 43, 7742 (2004).
25 D. A. Case, D. A. Pearlman, J. W. Caldwell, T. E.
Cheatham, III, J. Wang, W. S. Ross, C. L. Simmerling, T. A.
Darden, K. M. Merz, R. V. Stanton, A. L. Cheng, J. J. Vincent,
M. Crowley, V. Tsui, H. Gohlke, R. J. Radmer, Y. Duan, J. Pitera,
I. Massova, G. L. Seibel, U. C. Singh, P. K. Weiner, and P. A.
Kollman, Amber 7, University of California, San Francisco
(2002).
M. S. Gordon, J. H. Jensen, S. Koseki, N. Matsunaga, K. A.
Nguyen, S. J. Su, T. L. Windus, M. Dupuis, and J. A.
Montgomery, J. Comput. Chem., 14, 1347 (1993).
28 A. Pigache, P. Cieplak, and F.-Y. Dupradeau, Automatic
and highly reproducible RESP and ESP charge derivation: Appli-
cation to the development of programs RED and X RED, submit-
ted to the 227th ACS National Meeting, Anaheim, CA, March 28,
April 1, 2004. http://www.u-picardie.fr/labo/lbpd/RED/.
29 NAMD was developed by the Theoretical and Computa-
tional Biophysics Group in the Beckman Institute for Advanced
Science and Technology at the University of Illinois at Urbana-
Champaign: L. Kal, R. Skeel, M. Bhandarkar, R. Brunner, A.
Gursoy, N. Krawetz, J. Phillips, A. Shinozaki, K. Varadarajan,
and K. Schulten, J. Comput. Phys., 151, 283 (1999). http://www.
ks.uiuc.edu/Research/namd/.
30 H. C. Andersen, J. Comput. Phys., 52, 24 (1983).
31 T. Darden, D. York, and L. Pedersen, J. Phys. Chem., 98,
10089 (1993).
32 W. Humphrey, A. Dalke, and K. Schulten, J. Mol. Graph-
ics, 14, 33 (1996). http://www.ks.uiuc.edu/Research/vmd/.
33 P. Cieplak, J. W. Caldwell, and P. A. Kollman, J. Comput.
Chem., 22, 1048 (2001). The electronic version is available at the
AMBER web site: http://amber.scripps.edu/.
26 C. I. Bayly, P. Cieplak, W. D. Cornell, and P. A. Kollman,
J. Phys. Chem., 97, 10269 (1993).
27 M. W. Schmidt, K. K. Baldridge, J. A. Boatz, S. T. Elbert,
34 J. H. Halton, Numer. Math., 2, 84 (1960).