Size-dependent spectroscopy of MoS2 nanoclusters

Article abstract of DOI:10.1021/jp011898x

The synthesis and characterization of 3.5, 4.5, and 8 nm diameter MoS₂ nanoclusters are described. Each size is characterized by TEM images, electron diffraction, and optical spectroscopy. These nanoclusters can be synthesized directly or may be synthesized from samples of polydisperse nanoclusters. Obtaining larger or smaller discrete sizes from polydisperse samples indicates that these nanocluster sizes are thermodynamically favored, that is, these sizes correspond to free-energy minima. Electron diffraction and static spectroscopic results indicate that the particles are three atoms thick, specifically, single S-Mo-S trilayer disks. Static and time-resolved absorption and emission spectroscopy were used to determine the polarizations of the lowest energy allowed transitions in each case. We find that the 8 nm particles have a lowest energy absorption maximum at 473 nm that has mixed x - y and z polarizations, 4.5 nm particles have absorption maxima at 400 and 440 nm that are x - y and z polarized, respectively, and 3.5 nm particles have an absorption maximum at 362 nm that is z polarized. These results indicate that the polarization of the lowest observed transition is size-dependent and that the ordering of the lowest excited states changes as the particle size decreases.

Full text of DOI:10.1021/jp011898x

3794
J. Phys. Chem. B 2002, 106, 3794-3804
Size-Dependent Spectroscopy of MoS2 Nanoclusters
V. Chikan and D. F. Kelley*
Department of Chemistry, Kansas State UniVersity, Manhattan, Kansas 66506-3701
ReceiVed: May 16, 2001; In Final Form: December 17, 2001
The synthesis and characterization of 3.5, 4.5, and 8 nm diameter MoS2 nanoclusters are described. Each size
is characterized by TEM images, electron diffraction, and optical spectroscopy. These nanoclusters can be
synthesized directly or may be synthesized from samples of polydisperse nanoclusters. Obtaining larger or
smaller discrete sizes from polydisperse samples indicates that these nanocluster sizes are thermodynamically
favored, that is, these sizes correspond to free-energy minima. Electron diffraction and static spectroscopic
results indicate that the particles are three atoms thick, specifically, single S-Mo-S trilayer disks. Static and
time-resolved absorption and emission spectroscopy were used to determine the polarizations of the lowest
energy allowed transitions in each case. We find that the 8 nm particles have a lowest energy absorption
maximum at 473 nm that has mixed x - y and z polarizations, 4.5 nm particles have absorption maxima at
400 and 440 nm that are x - y and z polarized, respectively, and 3.5 nm particles have an absorption maximum
at 362 nm that is z polarized. These results indicate that the polarization of the lowest observed transition is
size-dependent and that the ordering of the lowest excited states changes as the particle size decreases.
Introduction
results that show that just such a reversal occurs in the case of
MoS2 nanoclusters.
One of the central motivations for the study of semiconductor
nanoclusters is that many of their properties are size dependent.
When the particle size becomes comparable to or less than that
of the exciton, the optical and electronic properties vary strongly
with the particle size. Properties such as the optical absorption
and emission spectra and the electrochemical redox potentials
change as a result of “quantum confinement” of the conduction
band electrons and valence band holes.1-3 Thus, by varying the
particle size, it often possible to tune the properties of the
semiconductor. Small semiconductor particles generally have
larger band gaps, more negative reduction potentials, and more
positive oxidation potentials than does the bulk material. Several
different treatments of semiconductor nanocluster electronic
structure have been developed and have recently been reviewed.4
The simplest semiquantitative description of how quantum
confinement alters these properties is given by treatments using
the “effective mass” approximation, in which the properties of
the conduction band electrons and valence band holes are
calculated from a simple particle-in-a-box model using the
effective masses of each carrier in the bulk semiconductor.2,3
The effective mass of each carrier depends on the band structure
of the solid and is given by meff ) p2 (∂2E/∂k2)-1. It is important
to note that the effective mass of a carrier depends on the
electronic state of the carrier; different electronic states are
characterized by different effective masses. Highly localized
electron or hole states have large effective masses. In the
effective mass model, the extent of the size-dependent shift of
the band gap is proportional to the inverse of the effective mass.
Thus, quantum confinement causes localized states to shift less
than delocalized states. One could imagine that in some cases
this could cause the ordering of different excited states to reverse
as the particle size is decreased. Such a reversal involving the
lowest excited states would cause the nature of the lowest
excited state to change with particle size. Below, we present
Examination of the size-dependent spectroscopy of any type
of nanocluster requires the synthesis of particles having different
sizes. Nanoclusters are often synthesized as free-standing
particles in solution using colloid chemistry methods. There are
several different methods that can be used to control the size
distributions of semiconductor and metal nanoparticles and
obtain monodisperse samples. These methods fall into three
broad classes, although there is considerable overlap between
them. Monodisperse nanoclusters may be obtained by (1)
controlling the kinetics of nucleation and subsequent crystal
growth,5,6 (2) synthesizing polydisperse samples and selecting
a portion of that distribution,7-9 or (3) adjusting the synthetic
conditions so that a monodisperse size distribution corresponds
to a thermodynamic minimum.10 The crystal growth kinetics
and stability may be strongly influenced by the nanocluster
thermodynamics.5 Thus, thermodynamics may play a key role
in all three methods. The particle surface energy is usually
positive, making larger particles thermodynamically more
favorable. In addition, the surface free energy increases with
the surface curvature, which is the Gibbs-Thomson effect.6 This
variation of surface energy with particle size leads to the well-
known phenomenon of Oswald ripening, in which larger
particles grow at the expense of smaller ones. By controlling
the solution monomer concentration, it is possible to vary the
growth rates of different-sized particles, thereby focusing or
defocusing the size distribution.5 Size-dependent thermodynam-
ics affecting the kinetics of particle growth is a rather general
and often observed phenomenon. However, thermodynamic
control of the static, equilibrium nanocluster size distribution
requires that there be a free-energy minimum corresponding to
a specific nanocluster size and that the particles be sufficiently
labile that this distribution is realized. This situation occurs in
surface-supported nanoclusters11,12 and in solution-phase metal
nanoclusters.10,13-15 The most extensively studied example of
a thermodynamically controlled nanocluster size distribution is
* Corresponding author. E-mail: dfkelley@ksu.edu.
10.1021/jp011898x CCC: $22.00 © 2002 American Chemical Society
Published on Web 03/23/2002

Size-Dependent Spectroscopy of MoS2 Nanoclusters
J. Phys. Chem. B, Vol. 106, No. 15, 2002 3795
that of thiol capped gold particles. When polydisperse nano-
clusters are heated in the presence of thiols in solution
(“digestive ripening”), monodisperse particles are formed, with
their average size determined by the thiol concentration. In this
case, the free-energy minimum results from a combination of
the surface energy, the thiol binding energy, and the Gibbs-
Thomson effect. The basic theory is simple and well understood
and is discussed below. However, thermodynamic control of
solution-phase semiconductor nanoclusters is rarely observed.
Below, we present spectroscopic, dynamical, and imaging results
indicating that under some circumstances the size distribution
of MoS2 nanoclusters is controlled by thermodynamics rather
than by the kinetics of nucleation and crystal growth.
There has recently been considerable study of MoS2 and
related (WS2, MoSe2, and WSe2) nanoclusters16-24 and “inor-
ganic fullerenes”25-35 of these materials. The first solution-phase
MoS2 nanoclusters were synthesized by Wilcoxon and co-
workers.16 Subsequently, we have synthesized and characterized
WS2, MoSe2, and WSe2 nanoclusters.20,21 We have also studied
the electron and hole-trapping dynamics in MoS2 and WS2
nanoclusters20,23 and the trapped electron and hole recombination
dynamics in MoS2 nanoclusters.19,24 In addition, we have studied
the interfacial electron-transfer dynamics between MoS2 nano-
clusters and adsorbed electron acceptors and the electron
injection dynamics between MoS2 nanoclusters and adsorbed
dyes.18,23 The optical properties of MoS2 nanotubes (inorganic
fullerenes) have also been studied and compared to band
structure calculations.27,36 These systems are much larger and
therefore show much smaller quantum confinement effects than
do the colloidal nanoclusters.
Time-Resolved Spectroscopy. Time-resolved emission re-
sults were obtained by time-correlated single photon counting.
In the apparatus used in these studies, the sample is excited
with 7 ps pulses from the second harmonic or the fundamental
of a synch-pumped, cavity-dumped cw dye laser (Spectra-
Physics 3500), which is pumped by a cw-mode-locked YAG
laser (Spectra-Physics 3800). Detection is accomplished with a
Hamamatsu 6m MCP PMT and time-correlated single photon
counting electronics. Wavelength selection is accomplished
using a 0.25 m monochromator with a 150 groove/mm grating.
The emission kinetics reported here are taken at 450 nm, with
a bandwidth of about 10 nm. The excitation wavelength is varied
in the range of 292-315 nm, with pulse energies of about 1.0
nJ focused to a spot size of about 0.5 mm. This combination of
low pulse energies and relatively large spot sizes results in
fluences that are sufficiently low that multiphoton excitations
are completely avoided. Polarized emission detection is ac-
complished using a 1 cm Glan-Taylor excitation polarizer and
a Polaroid emission polarizer in a collinear geometry. Appropri-
ate corrections are made for the polarization-dependent through-
put of the detection system.
The apparatus used to obtain the femtosecond one-color
polarized transient absorption results is similar to that previously
described.23 The brief description of the apparatus is as follows.
The femtosecond light source is based on a Clark-MXR CPA-
2001, which produces 775 nm, 130 fs, 800 µJ pulses at a
repetition rate of 1 kHz. The pulses are used to pump an OPA
(Clark-MXR, Vis-OPA) producing pulses between 450 and 495
nm or are put through a frequency-doubling crystal to produce
387 nm pulses. In the one-color polarized bleach experiments,
about 10% of the pulse intensity is split off and used for the
sample probe. The remainder is used for sample excitation. The
pump beam is reflected through a corner cube on a delay stage
to control the relative arrival time of the pump versus the probe
at the sample. This is followed by a Glan-Taylor polarizer to
ensure that the excitation pulse is completely linearly polarized.
The pump beam is typically focused to a spot size of 0.5-1.0
mm at the sample. The power density can be varied by changing
the position of the sample with respect to the focal point of the
pump beam, thereby changing the focal spot size. In the results
presented here, variation of the power density by a factor of
about 10 has no detectable effect on the observed kinetics. The
low-intensity probe beam is attenuated and then split into
reference and sample probe components. The sample probe
component is then passed through a half-wave plate and a Glan-
Taylor polarizer, which is adjusted to obtain equal amounts of
vertically and horizontally polarized light. The intensity of the
probe pulses is less than 1 µJ at the sample. After the sample,
the probe beam is split into horizontal and vertical polarization
components and imaged onto the detectors. Alternatively, in
the two-color transient absorption experiments, a portion of the
775 nm light is focused onto a sapphire window to generate a
white light continuum. A spectral portion of the continuum is
selected using interference filters and used as the sample probe.
In all cases, the reference and polarized probe beams are imaged
onto UDT Sensors PIN 10D photodiodes, biased at -15 V. The
photodiode outputs are amplified and input into a Stanford
Research Systems gated integrator. The gated integrator output
is measured using a National Instruments 16 bit A/D converter
in the data acquisition computer. The A/D converter and gated
integrator triggering and reset are synchronized with the CPA
2001 Q-switch and controlled by home-built timing electronics.
Data acquisition and movement of the delay stage is controlled
by LabView software running on a Pentium II computer.
Despite the above dynamical and spectroscopic studies, a
systematic study of the size-dependent spectroscopy of MoS2
nanoclusters has not been reported. In this article, we report
the synthesis of several different sizes of MoS2 nanoclusters
and show that these sizes correspond to thermodynamic minima.
We also use static and time-resolved absorption and emission
spectroscopy to determine the polarizations of the lowest
allowed excited states in each case. These results indicate that
the polarization of the lowest observed transition is size
dependent and that the ordering of the lowest excited states
changes as the particle size decreases.
Experimental Section
Materials. Dodecyltrimethylammonium chloride (DTAC)
(99%) and didodecyldimethylammonium bromide (98%) were
purchased from Fluka and recrystallized from ethanol. Trido-
decylmethylammonium iodide (TDAI) was obtained from
Aldrich and repeatedly washed with ether and stored in the
glovebox in the dark. Tridodecylmethylammonium chloride
(TDAC) was synthesized by the reaction of tridodecylamine
(Aldrich) and chloromethane (Aldrich), recrystallized from
hexane, and stored in the glovebox. Octane (reagent grade) was
purchased from Fisher and distilled in a nitrogen atmosphere
from sodium. Hexanol (99%, anhydrous) and ammonium sulfide
(20% solution) were purchased from Aldrich and used without
further purification. DQTCI dye was obtained from Exciton and
used without further purification. Molybdenum(IV) chloride was
synthesized by using standard methods.37,38 Specifically, MoCl5
was refluxed in tetrachloroethylene in an evacuated and sealed
vessel for 3 days. The resulting product was repeatedly washed
with carbon tetrachloride in an inert-atmosphere glovebox to
remove starting material. Acetonitrile was distilled from phos-
phorus pentoxide in a nitrogen atmosphere. Following purifica-
tion, all the reagents except H2S and (NH4)2S were kept in a
glovebox. Typical sample volumes in all syntheses were 10 mL.

3796 J. Phys. Chem. B, Vol. 106, No. 15, 2002
Chikan and Kelley
Figure 1. Absorption and emission spectrum of 3.5 nm MoS2
nanoclusters following extraction into acetonitrile.
Electron Microscopy and Diffraction. The sizes of the
different nanoclusters are determined with transmission electron
microscopy (TEM) using a Philips CM-100 TEM operating at
100 kV. A small fraction of the nanocluster solution (either
acetonitrile or octane phase) is diluted by a factor of 5 , and a
drop of the diluted solution is spread over a copper grid (300
mesh size) supporting a thin film of amorphous carbon. To
reduce the damage from the electron beam, the sample is cooled
to liquid nitrogen temperature during data collection. Electron
diffraction results were obtained with the same instrument and
on the same samples.
Figure 2. TEM image of 3.5 nm MoS2 nanoclusters.
transferred into a gastight vial. The solvent is evaporated, and
the nanoclusters are subsequently redissolved in acetonitrile in
the presence of DQTCI. Repeated evaporation of the solvent
followed by dissolution in acetonitrile results in complete
removal of the hexanol and in the DQTCI adsorbing onto the
particles. The nanoclusters with adsorbed DQTCI are then
redissolved in pentane. DQTCI is almost completely insoluble
in pentane, ensuring that all of the dye in this solution is indeed
adsorbed onto the MoS2 nanoclusters. The DQTCI/nanocluster
solution shows a small spectral shift of the dye absorption, also
indicating that the dye is absorbed on the nanoclusters. The static
emission is of comparable intensity to that of an equal
concentration of DQTCI in ethanol, indicating a lack of
quenching by electron transfer. The rotational diffusion dynam-
ics of the DQTCI absorbed on the nanoclusters are determined
by time-correlated single photon counting, with sample excita-
tion at 580 nm and emission detection at 650 nm. Both parallel
and perpendicular polarizations of the emission are determined.
Typical anisotropy decays are shown in Figure 3. The anisotropy
of DQTCI in acetonitrile (no MoS2 nanoclusters) decays very
rapidly, in about 100 ps. The average nanocluster size is
calculated from these results, on the basis of the Navier-Stokes
equation39 that is solved for an oblate rotor using stick boundary
condition. Two very reasonable assumptions are made to
facilitate these calculations. It is assumed that the thickness of
the nanocluster corresponds to that of a single layer of MoS2
(see below) and that the transition dipole of the absorbed dye
is perpendicular to the principal axis of the disc-shaped MoS2.
A Gaussian constraint on the particle-size distribution is also
assumed. The temperature of the experiment is 298 K, and the
solvent viscosity of n-pentane is 0.224 cP. The probability of
dye absorption is taken to vary with the particle size and to be
proportional to the circumference of the particle. The first 50
ps is not considered in order to neglect the stimulated emission
during the excitation pulse duration. A calculated curve corre-
sponding to 3.4 ( 0.6 nm is also shown in Figure 3. Agreement
between the observed and calculated depolarization kinetics is
quite good. This approximately 3.4 nm average size is also in
good agreement with the TEM results.
Synthesis and Characterization
3.5 nm Nanoclusters. The 3.5 nm nanoclusters are synthe-
sized using the same methods as in previous studies.16,18,19,24
MoCl4 is dissolved in a ternary tridodecylmethylammonium
iodide (TDAI/hexanol/octane, 8/8/84 wt %) inverse micelle
solution at a concentration of 1.0 × 10-3 M. H2S (4.0 molar
equiv) is added via a gastight syringe to the solution. Reaction
occurs within a few minutes. Following reaction, the inverse
micelle solution, along with the MoS2 nanoclusters, are extracted
with an equal volume of acetonitrile. The nanoclusters partition
approximately equally between the octane and acetonitrile
phases.
These nanoclusters are characterized by optical spectroscopy,
TEM, electron diffraction, and time-resolved emission depo-
larization of adsorbed DQTCI dye. The absorption and emission
spectra of these nanoclusters are shown in Figure 1. Previous
studies by a different group have characterized similar nano-
clusters with TEM and optical spectroscopy and have chro-
matographically purified the particles.16,17 The absorption
spectrum has a maximum at 362 nm, in agreement with
previously reported absorption spectra. The absorption spectra
obtained from the octane and acetonitrile phases exhibit the same
absorption maximum. The emission spectra are almost feature-
less and peak at about 420 nm. TEM images of these
nanoclusters are shown in Figure 2 and indicate that the average
particle diameter is about 3.5 nm. The average size may also
be determined by rotational diffusion measurements of an
adsorbed cyanine dye, DQTCI. This dye absorbs and emits at
about 580 and 650 nm, respectively, which is far to the red of
the MoS2 nanocluster absorption and emission. The rate of
emission depolarization is indicative of the rotational diffusion
rate and thus the nanocluster/dye hydrodynamic radius. In these
studies, a small portion of the nanocluster/micelle sample is

Size-Dependent Spectroscopy of MoS2 Nanoclusters
J. Phys. Chem. B, Vol. 106, No. 15, 2002 3797
Figure 3. Emission anisotropy of DQTCI adsorbed on 3.5 nm (MoS2/
TDAI) (lower curve) and polydisperse (MoS2/DDAB) (upper curve)
nanoclusters. Also shown are curves calculated for the rotational
diffusion of 3.4 ( 0.6 nm and 4.5 ( 1 nm MoS2 particles.
Figure 5. Absorption spectra of DDAB/MoS2 nanoclusters before
(dotted line) and after (solid line) addition of TDAI.
The broad size distribution of these samples is also confirmed
by rotational diffusion measurements. DQTCI dyes may be
attached to these particles, and a typical emission depolarization
decay is shown in Figure 3. The decay is multiexponential and
considerably longer than that obtained for the 3.5 nm particles.
This decay can be fit assuming particle sizes of 4.5 ( 1 nm.
Following the above synthesis, 0.5 g of TDAI surfactant is
added to 10 mL of polydisperse nanocluster inverse micelle
solution. The absorption spectrum of this mixture is shown in
Figure 5. The reaction mixture remains sealed under a nitrogen
atmosphere and left in the dark for several days while further
reaction occurs. The resulting spectrum has a well-defined peak
at 362 nm, indicating the presence of 3.5 nm particles. At this
point, DQTCI dye may be attached to the nanoclusters, and the
emission depolarization decay is very similar to that of the 3.5
nm particles (see Figure 3). TEM images also show that the
particle-size distribution has narrowed and become smaller.
Samples were also synthesized using TDAC (tridodecyl-
methylammonium chloride) or mixtures of TDAI and TDAC
as the surfactant. The use of TDAC as the surfactant usually
results in polydisperse samples similar to those obtained using
DDAB as the surfactant. In some cases, a shoulder in the 360-
362 nm region is obtained. Previous studies have shown a
resolved 360-365 nm peak in MoS2 nanocluster samples
synthesized in TDAC/hexanol/octane inverse micelles.16,18 This
is an important observation in considering the origin of the 362
nm peak, discussed below. The reason that our present samples
do not show a resolved peak, while the previous syntheses do,
is not known. However, we note that the previous syntheses
were done with commercial MoCl4, and the syntheses reported
here were done with freshly synthesized, highly purified MoCl4.
(To the best of our knowledge, MoCl4 is no longer commercially
available.) The presence of small amounts of other molybdenum
oxidation states in the commercial MoCl4 may affect the
particle-size distribution.
Figure 4. TEM image of polydisperse MoS2 nanoclusters synthesized
in DDAB/hexanol/octane micelles. Nanoclusters ranging in size from
∼2 to g15 nm can be seen.
Nanoclusters having diameters of 3.5 nm can also be
synthesized from a polydisperse sample simply by the addition
of iodide ions. The polydisperse sample is synthesized using
the same procedure as above except that the ternary inverse
micelle solution consists of DDAB/hexanol/octane (6/10/84 wt
%), where DDAB is didodecyldimethylammonium bromide.
MoCl4 is dissolved in this solution at a concentration of 10-3
M and 4.0 molar equiv of H2S is added via a gastight syringe.
Reaction occurs within a few minutes. The absorption spectrum
of this solution is featureless beyond 300 nm and decreases
smoothly with increasing wavelength, extending out to about
520 nm. This solution is stable for days but will eventually form
a precipitate. A TEM image of these particles is shown in Figure
4. The image shows that the sample is highly polydisperse,
having particle sizes ranging from about 2 nm to greater than
15 nm. Extraction of these polydisperse MoS2 nanoclusters with
acetonitrile results in the nanoclusters going into the octane
phase. The extracted solution is less stable than the micelle
solution, and the particles are usually left in the micelle solution.
Several samples were synthesized using a mixture of TDAI
and TDAC surfactants. The resulting absorption spectra were
superpositions of the polydisperse and 3.5 nm particle spectra.
Increasing the fraction of TDAC increases the polydisperse
component of the spectrum. Subsequent extraction with aceto-
nitrile results in the polydisperse component going into the
octane phase and the component having the absorption maxi-
mum at 362 nm going into the acetonitrile phase. Little or no
shift of this absorption maximum is observed as the TDAI/
TDAC ratio is varied.

3798 J. Phys. Chem. B, Vol. 106, No. 15, 2002
Chikan and Kelley
It is necessary to comment on the possibility of spectral
interferences caused by the formation of I3-. We note that the
absorption spectrum of I3- has a peak at about 366 nm, which
could interfere with the 362 nm peak observed here. Indeed, it
has been suggested that the peaks near 366 nm in the PbI2
nanocluster absorption spectrum may be assigned to ionic
species such as I3 rather than to the nanoclusters,40 although
-
counterarguments have been given.41 Independent of any
controversy in the PbI2 case, the results of blank experiments
rule out such an assignment or the possibility of significant
spectral interference in the present cases. Specifically, perform-
ing the synthesis without the metal chloride results in baseline
absorption in the 330-500 nm range. As long as the sample
remains sealed and in the dark, this blank spectrum is stable.
Subsequent exposure to air and light results in an increase of
absorption in the 366 nm region, presumably because of the
oxidation of I- to I3-. With this in mind, exposure of the samples
used here to air and light was avoided. It is also possible to
Figure 6. Absorption spectrum of a sample produced by the injection
of (NH4)2S into a MoS2/DTAC inverse micelle solution. The solid curve
is obtained immediately after injection and corresponds to a mixture
of 4.5 and 8 nm MoS2 nanoclusters. The dotted curve is obtained a
day later and corresponds to mostly 8 nm MoS2 nanoclusters.
-
imagine that solution-phase I3 could be formed by a redox
reaction with MoCl4, specifically, MoCl4 + 3I- f MoCl2 +
2Cl- + I3-. This possibility is also eliminated by blank
experiments. No peaks in this spectral region are formed in the
TEM images taken from grids prepared at this point are shown
in Figure 7 and show that these particles have an average
diameter of about 4.5 nm. As the samples age, the 400 nm peak
slowly loses intensity and is replaced by a well-defined peak at
473 nm. TEM images taken from grids prepared at this point
indicate the presence of 8 nm nanoclusters (see Figure 7). The
latter TEM image shows that there remains a small but
nonnegligible population of 4.5 nm particles. Consistent with
this observation, there remains a small 400 nm peak in the latter
spectrum in Figure 6. It is important to note that the 400 nm
peak does not gradually shift to 473 nm. Rather, the 473 nm
peak grows at the expense of the 400 nm peak, with an isosbestic
point at 455 nm. This process occurs on the time scale of several
hours and reaches equilibrium in about a day. Following the
reaction, the inverse micelle solution, along with the MoS2
nanoclusters, may be extracted with an equal volume of
acetonitrile. We find that if TDAI is used as the surfactant then
the nanoclusters go into the octane phase. Alternatively, if the
same synthesis is performed using DTAC as the surfactant, the
nanoclusters go into the acetonitrile phase.
-
absence of H2S. The only conceivable way that I3 could be
formed is if the above redox reaction becomes favored in the
-
presence of MoS2 particles (i.e., if I3 is stabilized by being
-
formed directly on the nanocluster surfaces). I3 is known to
have a great affinity for metal atoms on the edges of MoS2-
type surfaces,42-44 which could stabilize I3- formed in this way.
This possibility cannot be eliminated by any simple control
-
experiments, and it is possible that some I3 is formed in this
way, resulting in spectral interference. However, as noted above,
syntheses using TDAC as the surfactant and the commercial
MoCl4 also give a resolved 360-365 nm peak having an
intensity comparable to that observed here.16,18 The commercial
MoCl4 may have impurities corresponding to other molybdenum
oxidation states, but the synthesis of MoCl4 does not involve
any iodine-containing reagents37,38 so it surely does not have
iodide ions. In this case, the 360-365 nm peak can have no
I3- spectral interference. The observation that the peak observed
in the spectrum of nanoclusters synthesized with commercial
MoCl4 has comparable intensity to the 362 nm peak reported
-
here indicates that the majority of this feature is not due to I3
.
Samples of 8 nm nanoclusters may also be synthesized from
samples of polydisperse nanoclusters. Samples of polydisperse
nanoclusters are synthesized as described above. Addition of a
nonaqueous base, such as NaCN or Na2S, to this solution results
in formation of a well-defined peak at 470 nm. TEM images
taken at this point indicate a similar size distribution to that
obtained from the (NH4)2S synthesis, centered at about 8 nm.
Electron Diffraction Results. Electron diffraction patterns
have been obtained for the 3.5 and 8 nm nanoclusters as well
as bulk 2H-MoS2. These results can be interpreted by compari-
son with electron diffraction simulations. MoS2 commonly exists
as one of two different polytypes, 2H and 3R, which differ in
stacking of the S-Mo-S sheets. The more common is the 2H
polytype, in which the unit cell consists of two S-Mo-S
trilayers with the molybdenum atoms of one layer situated
directly above the sulfur atoms of the adjacent layer.45 The
experimental and calculated diffraction patterns show complete
agreement for the bulk 2H-MoS2; all reflections within a
scattering angle of 4° for 100 keV electrons can be assigned.
This result serves as an accurate calibration of the TEM
instrument on which the diffraction patterns are obtained.
Electron diffraction simulations show that the diffraction patterns
for the 2H and 3R crystal structures converge as the size of the
We also note that 4.5 and 8 nm MoS2 nanoclusters may be
synthesized in a somewhat basic TDAI micelle (see below),
without the formation of a 362 nm peak. This synthesis is quite
analogous to the procedure described above and therefore serves
as another control for the formation of I3-. We conclude that
we cannot exclude the possibility of some spectral interference
from adsorbed triiodide formed by an in situ surface redox
reaction. However, we emphasize that there is no reason to
suspect any spectral interference and that the TDAC synthesis
results show that most or all of the resolved 362 nm peak is
-
due to absorption of the nanocluster itself rather than to I3
chemisorbed on the nanocluster edges.
4.5 and 8 nm Nanoclusters. The 4.5 and 8 nm nanoclusters
are synthesized as follows. MoCl4 is dissolved in a ternary
inverse micelle solution of either dodecyltrimethylammonium
chloride (DTAC/hexanol/octane, 3/8/89 wt %) or tridodecylm-
ethylammonium iodide (TDAI/hexanol/octane, 8/8/84 wt %) at
a concentration of 1.0 × 10-3 M. Aqueous (NH4)2S (100 µL of
a 20% solution) is injected into the rapidly stirring solution.
Reaction occurs within about a minute. The absorption spectrum
of this solution is shown in Figure 6. The presence of a well-
defined 400 nm absorption peak indicates nanocluster formation.

Size-Dependent Spectroscopy of MoS2 Nanoclusters
J. Phys. Chem. B, Vol. 106, No. 15, 2002 3799
consist of a single trilayer type of structure. This same
conclusion was reached for WS2, MoSe2, and WSe2 nanoclusters
of comparable size.21,22 This conclusion makes good sense in
terms of the bonding; the only thing holding the trilayer sheets
to each other are comparatively weak van der Waals forces.
When the particles are sufficiently small, this binding is weak,
and the sheets will spontaneously exfoliate. This result is also
consistent with studies of other layered materials, which
exfoliate into single covalently bound sheets.46,47
Thermodynamic Control of Nanocluster Sizes. The above
results may be summarized as follows. Polydisperse samples,
synthesized from DDAB/hexanol/octane micelles, have particle
sizes ranging from ∼2 to g15 nm and featureless absorption
spectra. These samples may be treated with either iodide ions
or with a nonaqueous base. In the former case, TEM images
show that a relatively monodisperse sample of 3.5 nm particles
is formed and that the absorption spectrum develops a resolved
peak at 362 nm. In the latter case, relatively monodisperse 8
nm particles are formed, and the absorption spectrum develops
a resolved peak at 470 nm.
The above observations indicate that the nanocluster size
distributions are controlled by thermodynamics rather than by
the kinetics of nucleation and crystal growth. This is an unusual
situation for semiconductor nanoclusters, but it is well docu-
mented for metal nanoclusters. In most semiconductor syntheses,
narrow size distributions are obtained by separating the time
scales of nucleation and subsequent crystal growth. Conditions
are adjusted such that a burst of nucleation occurs upon mixing
the reactants, followed by much slower growth. Because all
nucleation occurs simultaneously and growth occurs at the same
rate for all of the nascent nanoclusters, a narrow size distribution
is obtained in the kinetically controlled limit. In contrast, under
some circumstances, the size-dependent thermodynamics de-
termines the size distribution of gold nanoclusters. This is not
a “magic number” consideration resulting from the filling of
structural or electronic shells but results from classical consid-
erations of the free energies of bound and solution-phase ligands
and of curved surfaces. The theory of how thermodynamics
controls the size distribution of gold particles in the presence
of alkanethiols has been considered in detail.10 The basic idea
is as follows. Particles establish a size distribution that minimizes
the free energy of the entire system. Thiols bind quite strongly
to the surface of the gold, and the total free energy has
contributions from both the surface energy of the gold particles
and the adsorbed thiols. How this results in a thermodynamic
minimum can be seen from simple scaling arguments. The
surface free energy of the gold particles varies with the number
of atoms per particle for a constant total number of gold atoms.
Specifically,
Figure 7. TEM images of mixtures of 4.5 and 8 nm MoS2/DTAC
nanoclusters. The upper panel corresponds to a sample immediately
following synthesis and is almost exclusively 4.5 nm particles. The
lower panel corresponds to a day after synthesis and is mostly 8 nm
particles.
µsurf ) (number of particles) ×
(number of surface atoms/particle) × (energy/surface atom)
(1)
structures decreases along the crystallographic c axis. These
electron diffraction simulations show that within a scattering
angle of 4° for 100 keV electrons four reflections having indices
of (1,1,0), (3,0,0), (2,2,0), and (4,1,0) remain as the crystal
thickness is decreased to a single unit cell. Other very intense
reflections calculated for bulk 2H-MoS2 (specifically, (0,0,2),
(1,0,0), and (1,0,2)) broaden out and are lost. This result matches
the experimental result for both the 3.5 and 8 nm nanoclusters.
Diffraction rings at radii corresponding to the (1,1,0), (3,0,0),
(2,2,0), and (4,1,0) reflections are observed, and nothing is
observed at radii corresponding to the (0,0,2), (1,0,0), and (1,0,2)
reflections. We therefore conclude that the MoS2 nanoclusters
We may define N ≡ total number of atoms (which is constant)
and n ≡ number of atoms per particle. We have that (number
of particles) ) N/n and that the (number of surface atoms/
particle) scales as the square of the particle radius, or as n2/3
.
Thus, the product of the first two terms in eq 1 is related to the
total surface area and scales as n-1/3. The surface energy per
area (or the energy per surface atom) depends on the surface
radius of curvature through the Gibbs-Thomson effect.6 This
surface energy is taken to be C + D′/R or C + D/n1/3, where C,
D, and D′ are constants and R is the particle radius. Thus, we

3800 J. Phys. Chem. B, Vol. 106, No. 15, 2002
Chikan and Kelley
get µsurf ) A/n1/3 + B/n2/3, where A and B are constants.
Producing surface and decreasing the surface radius of curvature
both cost energy, and A and B are positive constants. The thiol
binding energy is given by
µbind ) (no. of bound thiols) × (binding energy/thiol) (2)
The number of bound thiols is proportional to the total surface
area and hence scales as n-1/3. If the thiol binding energy is
taken to be surface curvature-independent, then µbind ) A′/n1/3
.
Binding of the thiols is energetically favorable, and A′ is a
negative constant. Thus, combining eqs 1 and 2, the total free
energy depends on the particle size and µtot ) µsurf + µbind
)
(A + A′)/n1/3 + B/n2/3. If A + A′ is negative, then µtot has a
minimum at n ) (-2B/(A + A′))3, which corresponds to the
thermodynamically favored nanocluster size. It is important to
note that the free-energy minimum results from the combination
of thiol binding favoring a high surface area and therefore small
particles along with the Gibbs-Thomson effect favoring a flat
surface and therefore larger particles.
Figure 8. Absorption spectra of 3.5 nm (dashed curve), 8 nm (dotted
curve) and a mixture of 4.5 and 8 nm (solid curve) MoS2 nanoclusters.
The mixture curve corresponds to a MoCl4/(NH4)2S synthesis shortly
after injection. A difference curve corresponding to just the 4.5 nm
nanoclusters is shown in the insert. The assigned polarizations of the
absorption transitions are also shown.
The same basic idea may be applied to the 2-D MoS2
nanoclusters studied here. In this case, the edge of the (assumed
circular) nanoclusters plays the role of the surface in the 3-D
gold nanoclusters. The nanocluster edges have exposed molyb-
denum atoms that may bind to ligands in solution. We suggest
that in the neutral pH synthesis using TDAI as the surfactant
the ligand that binds to the surface is I- and in the syntheses
using (NH4)2S or other bases (Na2S or NaCN) the edge-binding
ligand is SH-. Because MoS2 particles are 2-D, the edge area
per particle (i.e., the circumference) scales as n1/2, and the total
edge area at constant N scales as n-1/2. By analogy to the 3-D
case, the Gibbs-Thomson effect gives an energy term that
scales as 1/R. The total free energy may be evaluated for the
2-D case, and we get µtot ) µsurf + µbind ) (A + A′)/n1/2 + B/n.
The presence of a strong binding ligand will result in the quantity
A + A′ being negative, and the free energy will have a minimum
at n ) (-2B/(A + A′))2. A stronger binding ligand will result
in a more negative value for A′ and thus smaller particles. We
suggest that I- lowers the surface free energy of the nanocluster
edges more than does SH-, resulting in smaller particles.
An obvious experiment is to try to vary the nanocluster size
by varying the iodide ion concentration. One could imagine that
this would vary the iodide ion chemical potential and thus vary
the extent to which the nanocluster edges are bound by iodide
ions. The effect of this would be to vary the magnitude of the
A′ constant (above) and thereby change the size corresponding
to the minimum in the total chemical potential. It has been
shown that varying the thiol concentration varies the gold
particle size in the gold/thiol system. However, as described
above, the result of this type of experiment is to obtain a well-
defined peak at 362 nm as well as a broad absorption
characteristic of a polydisperse sample. As indicated above,
acetonitrile extraction of the nanoclusters synthesized in these
mixed surfactant solutions results in a polydisperse sample in
the octane phase and a 3.5 nm sample in the acetonitrile phase.
The relative concentrations of these components depend on the
TDAC/TDAI ratio, with more TDAC favoring a larger poly-
disperse component. This result may be understood in terms of
the inverse micelle solution being a microheterogeneous envi-
ronment, resulting in more complicated phase-dependent chem-
istry. We suggest that adsorption of iodide ions results in
changing the nanocluster charge as well as the edge energy.
Consider the following equilibrium
where [MoS2] refers to an MoS2 particle and “exterior” and
“interior” refer to the nanocluster being in the nonpolar and
polar environments on the exterior and interior of the inverse
micelles, respectively. MoS2 particles on the exterior of the
micelles (in the hydrocarbon phase) are uncharged and have
few iodide ions adsorbed on their edges. We suggest that in
the absence of adsorbed iodide ions there is no thermodynamic
minimum in the size distribution and these nanoclusters are
polydisperse. Decreasing the iodide ion concentration (in the
mixed-surfactant syntheses) shifts this equilibrium to the left,
increasing the fraction of polydisperse nanoclusters. In this
model, nanoclusters on the interior of the micelle are in an
environment that is close to saturated with iodide ions, and these
nanoclusters always have the same minimum in their size
distribution.
The above model considers the surface energy in terms of
iodide ion adsorption. However, the same effects should be seen
for any species that binds tightly to the nanocluster edge, thereby
reducing its chemical potential. The media can be made basic
by using (NH4)2S as the sulfiding agent in the synthesis or by
the addition of Na2S or NaCN following the synthesis of
polydisperse particles. In all of these cases, SH- is formed and
can adsorb onto the nanocluster edges. We suggest that the
extent to which the chemical potential is lowered by HS-
adsorption is somewhat less than that in the case of I-. Thus,
the quantity A + A′ is not as negative in the SH- case, and
larger (8 nm) particles are formed. Why the synthesis with
(NH4)2S initially produces metastable 4.5 nm particles and only
later gives the more stable 8 nm particles is not well understood
at this point. The initial formation of 4.5 nm particles must be
due to both kinetic and thermodynamic factors, but no detailed
mechanism can be determined from the results presented here.
Size-Dependent Polarization Spectroscopy
The absorption spectra of the 3.5 nm MoS2 nanoclusters and
the mixture of 4.5 and 8 nm MoS2 nanoclusters taken shortly
after and a day after the (NH4)2S injection are collected in Figure
8. As discussed above, injection of (NH4)2S into an MoCl4
inverse micelle solution immediately results in the formation
of 4.5 nm particles, which later form almost exclusively 8 nm
particles. Thus, at long times, there are mostly 8 nm particles,
and their spectrum can be obtained with minimal spectral
[MoS2] (exterior) + mI- a [MoS2]Im- (interior)

Size-Dependent Spectroscopy of MoS2 Nanoclusters
J. Phys. Chem. B, Vol. 106, No. 15, 2002 3801
where θ is the angle between the pump and probe oscillators.
A value of 0.4 can be obtained only when both pump and probe
oscillators are linear and θ ) 0°. If both oscillators are in a
plane, then one must integrate over all orientations in the plane,
and an anisotropy of 0.1 is obtained. The other situation relevant
to disklike particles is if one oscillator is in-plane and the other
is perpendicular to the plane. In this case, θ ) 90°, and the
anisotropy is -0.2. Thus, the initial anisotropy of 0.4 indicates
that the absorbing and emitting oscillators are both linear and
aligned. Because the only unique axis is the crystallographic c
axis, we conclude that the absorption at 310 nm and the emission
are both z-polarized. The initial anisotropy is excitation wave-
length-dependent. We find that the anisotropy decreases with
decreasing excitation wavelength in the 290-315 nm range and
approaches 0.0 with 290 nm excitation. This result indicates
the presence of an overlapping x - y polarized transition in the
blue part of this range.
The combination of static absorption, time-resolved emission,
and emission quantum yield measurements suggests that the
emitting state is not the same state as that giving rise to the
362 nm absorption peak. The 362 nm peak has an absorbance
of about 1.0 in a 0.2 cm path-length cell. These nanoclusters
have on the order of 100 Mo atoms, and the original solution
has [Mo] ) 10-3, making the nanocluster concentration about
10-5 M. Thus, these nanoclusters have an extinction coefficient
at 362 nm of roughly 5 × 105 L mol-1 cm-1. In the strong
quantum confinement limit, the excited state (exciton) does not
have translational degeneracy, and we conclude that the observed
transition is allowed. The emission lifetime is several nanosec-
onds, and a high emission quantum yield would be expected
from an allowed transition. However, the emission quantum
yield is observed to be about 10-3. We therefore conclude that
emission is from a low oscillator strength, z-axis-polarized
transition buried under the red edge of the 362 nm peak. The
static polarized excitation spectrum shows small, positive
anisotropy throughout the 330-400 nm excitation wavelength
range, indicating that the emitting state and the 362 nm peak
have the same polarization, that is, they are both z-axis polarized.
The exact magnitude of the splitting between these states cannot
be determined from the results presented here. These assign-
ments are summarized in Figure 8.
Figure 9. Time-resolved total emission and emission anisotropy of
3.5 nm MoS2 nanoclusters excited at 310 nm and detected at 425 nm.
interference from the 4.5 nm particles. The spectrum of the 4.5
nm particles (without spectral interferences of the 8 nm particles)
can be obtained by subtracting the long-time spectrum from
that obtained shortly after injection. Because there is some
concentration of 8 nm particles in the sample shortly after
synthesis, this subtraction is scaled to eliminate the 473 nm peak.
This subtracted spectrum (Figure 8) shows the well-defined peak
at 400 nm and a prominent shoulder at about 440 nm, both of
which may be assigned to the 4.5 nm particles.
It is tempting to assume that the 473, 400, and 362 nm peaks
correspond to the same transition in the 8, 4.5, and 3.5 nm
particles, respectively, and that this transition correlates to the
lowest allowed transition (the A exciton) in bulk MoS2. The
lowest allowed transitions in the bulk material, the A and B
excitons, are polarized perpendicular to the crystallographic c
axis, so polarization studies of the nanocluster absorptions can
determine if this is the case. Both time-resolved polarized
emission and one-color time-resolved polarized absorption
(transient bleach) studies can reveal the polarization of the
absorption oscillator. The 8 nm particles are only very weakly
emissive, making time-resolved emission polarization measure-
ments problematic. However, the polarization of the 473 nm
peak can be determined only by a polarized bleach measurement.
The 3.5 nm particles are emissive, and the polarization of several
of the lowest transitions may be determined from polarized
emission measurements. The 4.5 nm particles are also only
weakly emissive, and we have determined the polarization of
the 400 nm peak and the 440 nm shoulder by polarized bleach
measurements.
4.5 and 8 nm Nanoclusters. The polarizations of the lowest
transitions of the 4.5 and 8 nm particles may be determined
from one-color transient bleach measurements. Similar polariza-
tion considerations apply to these studies as apply to the case
of the time-resolved emission studies, described above. In the
bleach experiments, the sample is excited with polarized light,
and the decrease of the absorbance at the same wavelength is
measured. This may be thought of as a “polarization hole-
burning” experiment. The anisotropy of the absorbance change
will be 0.4 or 0.1 for linear or planar polarized transitions,
respectively. In these experiments, the bleach anisotropy was
determined using femtosecond pulses, with the probe delayed
a few picoseconds from the pump. The results reported here
were obtained in acetonitrile solvent, which relaxes quite
quickly, within a few picoseconds. This delay ensures electronic
and vibrational relaxation as well as relaxation of optical Kerr
effects induced in the solvent. One caveat about these transient
bleach measurements need to be mentioned. The results could
be confused by the presence of an excited-state absorption. To
address this question, transient absorption experiments were
performed with excitation at 475 nm and detection at 550 nm.
This detection wavelength is to the red of the wavelengths at
which a bleach would be observed and provides a measure of
3.5 nm Nanoclusters. Time-resolved total emission and
emission anisotropy decays for the 3.5 nm particles are shown
in Figure 9. These data were obtained following 310 nm
excitation using 450 nm detection. The anisotropy starts out at
close to 0.4 and subsequently decays. The dynamics following
excitation have been discussed in a previous article.23 Photo-
selection theory gives the anisotropy for an incoherent pump/
probe emission experiment as48,49
r ) (Ipar - Iperp)/(Ipar + 2Iperp) ) 2/5P2(cos θ) )
1/5(3 cos2 θ - 1),

3802 J. Phys. Chem. B, Vol. 106, No. 15, 2002
Chikan and Kelley
TABLE 1: Observed Transition and Quantum Confinement
Energies
the transient absorption intensity in this general spectral range.
An extremely small transient absorption is observed, which is
at least a factor of 10 smaller than the magnitude of the observed
bleach signals. We therefore conclude that the polarization of
the bleach transients gives an accurate indication of the
polarization of the transition being excited.
x - y (observed)
z (observed)
∆(x - y)
∆z
nm
/1000 cm-1
nm
/1000 cm-1 /1000 cm-1 /1000 cm-1
bulk
660, 630
15.88a
21.1
25.0
= 540b
473
440
18.5
21.1
22.7
27.6
0
0
8 nm 473
4.5 nm 400
3.5 nm 290
5.3
9.1
18.6
2.6
4.2
9.1
The transient bleach results indicate that excitation of the 4.5
nm nanoclusters at 387.5 nm induces a transient anisotropy very
close to 0.10, which means that that the 400 nm peak is x - y
polarized. Determining the polarization of the 440 nm shoulder
of the 4.5 nm particles is somewhat problematic because of the
spectral overlap with the blue edge of the 473 nm peak (assigned
to the 8 nm particles). However, this interference is minimized
by taking the bleach results immediately after injection of
(NH4)2S, when the sample has almost exclusively 4.5 nm
particles. In this case, excitation at 450 nm induces a transient
anisotropy that is very close to 0.40, indicating that the 440 nm
shoulder is z-axis polarized. As the sample ages, the 450 nm
absorption of the 4.5 nm particles is replaced by a weaker one
assigned to 8 nm particles. As this process occurs, the anisotropy
decreases, and the bleach becomes essentially unpolarized when
the 400 nm peak is completely gone. Consistent with this result,
bleach anisotropy measurements following excitation at 475 and
490 nm also show little or no anisotropy. We conclude that the
473 nm peak, assigned to the 8 nm particles, is close to isotropic,
which could be due to x - y- and z-polarized transitions that
are either overlapping or mixed. It is unlikely that the differently
polarized transitions would be exactly overlapping, and if the
lack of anisotropy were due to overlapping transitions, then some
wavelength dependence of the anisotropy would probably be
seen. The band is observed to be isotropic from 450 to 490
nm, and we suggest that the most likely explanation of the lack
of polarization is that there is mixing of x - y- and z-polarized
transitions at comparable energies. These assignments are also
summarized in Figure 8.
34.5
362
a Taken from the average of the A and B excitons at 660 and 602
nm, respectively. b Estimated from the broad onset in the amorphous
bulk MoS2 spectra.
cut perpendicular to the basal plane and of sufficient optical
quality, is difficult.52,53 However, the approximate energy of
the lowest z-polarized transition may be inferred from the
unpolarized amorphous thin film or colloidal particle spectra54,55
by knowing the energies of the x - y-polarized transitions from
single-crystal spectra. Specifically, while the single-crystal x
- y-polarized spectrum shows no transitions in the 500-600
nm region, the unpolarized amorphous thin film and colloidal
particle spectra show a broad shoulder at wavelengths less than
570 nm. Assigning the energy of this transition is uncertain
because these spectra are quite broad and no peak assignable
to the z-polarized transition is resolved. However, on the basis
of these unresolved spectra, there is probably a z-polarized
transition in the (approximately) 540 nm region.
Quantum confinement is expected to affect the energies of
the x - y- and z-axis-polarized transitions very differently.
Photoexcitation produces a nodal plane that is perpendicular to
the polarization of the light. We expect that the presence of a
nodal plane will increase the spatial extent of the wave function
perpendicular to that nodal plane. Thus, an excited state
corresponding to an x - y-polarized transition will have a larger
x - y spatial extent than will one corresponding to a z-polarized
transition, that is, changes in x - y quantum confinement will
have the largest effects on x - y-polarized transitions. On the
basis of this expectation, an important prediction can be made:
reducing the x - y dimensions of these 2-D particles will cause
a larger blue shift of the transitions that are x - y polarized
than of those that are z polarized. This differential quantum
confinement effect can be seen by comparing the absorption
spectrum of bulk MoS2 with that of the 8, 4.5, and 3.5 nm
particles. These comparisons are summarized in Table 1. The
quantum confinement shifts of the lowest x - y- and z-polarized
transitions are given as ∆(x - y) and ∆z, respectively. These
shifts are calculated by taking the lowest x - y-polarized
transition in bulk MoS2 to halfway between the A and B
excitons. This is appropriate because the nanoclusters are single
trilayers, and this value represents the energy of the zeroth-
order state in the absence of splitting from the interactions of
adjacent trilayers. The lowest z-polarized transition in bulk MoS2
is taken to be at 540 nm, as discussed above. In all cases, the
lowest x - y-polarized transition correlates to the A and B
excitons, and the lowest observed z-polarized transition cor-
relates to the shoulder at 540 nm in bulk MoS2. Table 1 shows
that as the x - y dimensions of the particle are reduced quantum
confinement causes both the x - y- and z-polarized transitions
to move to higher energies. As expected from the above
arguments, quantum confinement in the x - y plane has a larger
effect on the energies of the x - y-polarized transition than on
those of the z-polarized transition. The values in Table 1 indicate
that the effect of changes in x - y quantum confinement is about
twice as large for x - y-polarized transitions. We note that a
reversal of the ordering of the excited states occurs upon
quantum confinement. In particles smaller than 8 nm, the lowest
Spectral Assignments. The polarization results presented
above make it possible to establish the correlations of the
transitions observed in these nanoclusters and in bulk MoS2. In
particular, these results show how quantum confinement affects
the ordering of the excited states. The lowest allowed transitions
in bulk MoS2 are the A and B excitons at about 660 and 602
nm, respectively.50 Both of these absorption features are
polarized perpendicular to the z axis in the x - y plane. Band
structure calculations indicate that these two features correspond
to the same transition, with the approximately 1500 cm-1
splitting caused by the interaction of adjacent trilayer sheets.27,36
Thus, only a single transition is expected for the present case
of single trilayer nanoclusters. The spectroscopy of MoS2 thin
films and inorganic fullerenes shows that z-axis quantum
confinement has only a small effect on this transition energy.
The A and B excitons shift only about 0.1 eV (about 800 cm-1
)
from the bulk energy in the thinnest (a few trilayers thick)
inorganic fullerenes or thin films.27,51 Thus, in the absence of
quantum confinement effects, interlayer interactions split a single
transition at about 15 900 cm-1 (630 nm) into the observed A
and B exciton transitions.
Although these lowest x - y-polarized transitions are
comparatively well understood, assignment of the lowest
z-polarized transition is somewhat problematic. The z-axis-
polarized transitions in bulk MoS2 are at an energy higher than
that of the A and B excitons and have not been studied. Polarized
absorptions in MoS2 and related materials are typically studied
using reflection spectroscopy of single crystals. Related literature
on this subject indicates that producing MoS2 single crystals,

Size-Dependent Spectroscopy of MoS2 Nanoclusters
J. Phys. Chem. B, Vol. 106, No. 15, 2002 3803
allowed transition is z polarized, whereas in bulk MoS2 it is x
- y polarized. We suggest that the crossover occurs in particles
that are about 8 nm in diameter.
somewhat longer wavelengths. The lowest x - y-polarized
transition is at about 290 nm.
(5) The lowest observed transition in bulk MoS2 is x - y
polarized, and we conclude that the ordering of this transition
and the lowest z-polarized transition reverse for particles smaller
than about 8 nm.
We note that the spectra presented here are consistent with
the conclusion made from the electron diffraction results: the
particles are single trilayer disks. This result is seen from the
comparison of the A and B exciton spectrum of bulk MoS2 with
the spectrum of the 8 nm particles. The splitting between the A
(660 nm) and B (602 nm) excitons in bulk MoS2 is about 1500
cm-1. These transitions are of roughly comparable intensity. In
contrast, the spectrum of the 8 nm particles shows a single,
well-resolved peak at 473 nm, with the next transition at about
340 nm. The separation between these transitions is about 8000
cm-1. A peak 1500 cm-1 to the blue of 473 nm would be at
about 442 nm. Figures 6 and 8 show no evidence in the 8 nm
particle absorption spectrum of the sort of splitting that gives
rise to the A and B exciton in bulk MoS2. Because band structure
calculations indicate that the A-B exciton splitting is due to
interlayer interactions,27,36 this result also indicates that these
particles are single trilayers.
Acknowledgment. This work was supported by grants from
the Department of Energy (DE-FG03-96ER14717 and DE-
FG03-15037).
References and Notes
(1) Semiconductor Nanoclusters-Physical, Chemical, and Catalytic
Aspects; Kamat, P. V., Meisel, D., Eds.; Elsevier: New York, 1997.
(2) Brus, L. E. J. Chem. Phys. 1983, 79, 5566.
(3) Brus, L. E. J. Phys. Chem. 1986, 90, 2555.
(4) Efros, A. L.; Rosen, M. Annu. ReV. Mater. Sci. 2000, 30, 475.
(5) Peng, X.; Wickham, J.; Alivisatos, A. P. J. Am. Chem. Soc. 1998,
120, 5343.
(6) Sugimoto, T. AdV. Colloid Interface Sci. 1987, 28, 65.
(7) Chemsiddine, A.; Weller, H. Ber. Bunsen-Ges. Phys. Chem. 1993,
97, 636.
Finally, it is of interest to compare the results presented here
with the morphology-dependent spectroscopy of CdSe. The size-
dependent emission polarization observed here is somewhat
analogous to that observed in CdSe quantum dots and rods.
Emission from CdSe quantum dots is planar polarized, with the
unique axis established by the crystal structure.4 CdSe rods grow
along the unique axis, and when the rods are sufficiently long,
the emission is linearly polarized.56,57 In both the present case
and the case of CdSe, it is the change in the particle aspect
ratio that changes the amount of z-axis versus x - y-plane
quantum confinement and thereby causes the ordering of the
states to reverse.
(8) Micic, O. I.; Sprague, J. R.; Curtis, C. J.; Jones, K. M.; Machol, J.
L.; Nozik, A. J.; Giessen, H.; Fluegel, G.; Mohs, G.; Peyghambarian, N. J.
Phys. Chem. 1995, 99, 7754.
(9) Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc.
1993, 115, 8706.
(10) Leff, D. V.; Ohara, P. C.; Heath, J. R.; Gelbart, W. M. J. Phys.
Chem. 1995, 99, 7036.
(11) Williams, R. S.; Medeiros-Rebeiro, G.; Kamins, T. I.; Ohlberg, D.
A. A. J. Phys. Chem. B 1998, 102, 9605.
(12) Williams, R. S.; Medeiros-Rebeiro, G.; Kamins, T. I.; Ohlberg, D.
A. A. Acc. Chem. Res. 1999, 32, 425.
(13) Lim, X. M.; Sorenson, C. M.; Klabunde, K. J. Chem. Mater. 1999,
11, 198.
(14) Lim, X. M.; Sorenson, C. M.; Klabunde, K. J. J. Nanoparticle Res.
2000, 2, 157.
(15) Lin, X. M.; Wang, G. M.; Sorenson, C. M.; Klabunde, K. J. J.
Phys. Chem. B 1999, 103, 5488.
(16) Wilcoxon, J. P.; Samara, G. A. Phys. ReV. B: Condens. Matter
1995, 51, 7299.
Conclusions
Several conclusions may be drawn from the results presented
here.
(17) Wilcoxon, J. P.; Newcomer, P. P.; Samara, G. A. J. Appl. Phys.
1997, 81, 7934.
(18) Parsapour, F.; Kelley, D. F.; Craft, S.; Wilcoxon, J. P. J. Chem.
Phys. 1996, 104, 4978.
(19) Doolen, R.; Laitinen, R.; Parsapour, F.; Kelley, D. F. J. Phys. Chem.
B 1998, 102, 3906.
(20) Huang, J. M.; Kelley, D. F. J. Chem. Phys. 2000, 113, 793.
(21) Huang, J. M.; Kelley, D. F. Chem. Mater. 2000, 12, 2825.
(22) Huang, J. M.; Laitinen, R.; Kelley, D. F. Phys. ReV. B: Condens.
Matter 2000, 62, 10995.
(1) 3.5, 4.5, and 8 nm MoS2 nanoclusters can be synthesized
using inverse micelle methods. Electron diffraction and optical
spectroscopic results indicate that these particles are a single
trilayer (S-Mo-S) thick. The 3.5 and 8 nm diameter particles
are stable, whereas the 4.5 nm particles are metastable, forming
8 nm particles in a matter of hours.
(2) The 3.5 and 8 nm particles may be synthesized directly
or from previously synthesized polydisperse samples. The
observation that the average size of a polydisperse distribution
can be either increased or decreased by the addition of iodide
ions or an increase of the pH indicates that these sizes are
controlled by thermodynamics and correspond to free-energy
minima.
(23) Chikan, V.; Waterland, M. R.; Huang, J. M.; Kelley, D. F. J. Chem.
Phys. 2000, 113, 5448.
(24) Parsapour, F.; Kelley, D. F. Mater. Res. Soc. Symp. Proc. 2000,
571, 197.
(25) Feldman, Y.; Frey, G. L.; Homyonfer, M.; Lyakhovitskaya, V.;
Margulis, L.; Cohen, H.; Hodes, G.; Hutchison, J. L.; Tenne, R. J. Am.
Chem. Soc. 1996, 118, 5362.
(26) Feldman, Y.; Lyakhovitskaya, V.; Tenne, R. J. Am. Chem. Soc.
1998, 120, 4176.
(3) The free-energy minima corresponding to the 3.5 and 8
nm particles may be understood in terms of the particle-edge
energies, the ligand (I- or SH-) adsorption energies, and the
Gibbs-Thomson effect. This situation is completely analogous
to thermodynamic control of the size distribution of gold
particles with adsorbed thiols.
(4) The combination of transient bleach and emission studies
permits assignment of the polarization of the lowest observed
transitions in each case. We find that the lowest transition in
the 8 nm particles is at 473 nm and is a mixture of x - y and
z polarizations. The 4.5 nm particles have transitions at 440
nm (z polarized) and 400 nm (x - y polarized). The 3.5 nm
particles have z-axis-polarized allowed transitions at about 310
and 362 nm and a z-axis-polarized forbidden transition at
(27) Frey, G. L.; Elani, S.; Homyonfer, M.; Feldman, Y.; Tenne, R.
Phys. ReV. B: Condens. Matter 1998, 57, 6666.
(28) Frey, G. L.; Tenne, R.; Matthews, M. J.; Dresselhaus, M. S.;
Dresselhaus, G. Phys. ReV. B: Condens. Matter 1999, 60, 2883.
(29) Homyonfer, M.; Alperson, B.; Rosenberg, Y.; Sapir, L.; Cohen, S.
R.; Hodes, G.; Tenne, R. J. Am. Chem. Soc. 1997, 119, 2693.
(30) Tenne, R.; Homyonfer, M.; Feldman, Y. Chem. Mater. 1998, 10,
3225.
(31) Zak, A.; Feldman, Y.; Alperovich, V.; Rosentsveig, R.; Tenne, R.
J. Am. Chem. Soc. 2000, 122, 11108.
(32) Rothschild, A.; Popovitz-Biro, R.; Lourie, O.; Tenne, R. J. Phys.
Chem. B 2000, 104, 8976.
(33) Zhu, Y. Q.; Hsu, W. K.; Grobert, N.; Chang, B. H.; Terrones, M.;
Terrones, H.; Kroto, H. W.; Walton, D. R. M.; Wei, B. Q. Chem. Mater.
2000, 12, 1190.
(34) Rothschild, A.; Sloan, J.; Tenne, R. J. Am. Chem. Soc. 2000, 122,
5169.

3804 J. Phys. Chem. B, Vol. 106, No. 15, 2002
Chikan and Kelley
(35) Hsu, W. K.; Chang, B. H.; Zhu, Y. Q.; Han, W. Q.; Terrones, H.;
Terrones, M.; Grobert, N.; Cheetham, A. K.; Kroto, H. W.; Walton, D. R.
M. J. Am. Chem. Soc. 2000, 122, 10155.
(36) Coehoorn, R.; Haas, C.; Dijkstra, J.; Flipse, C. J. F.; deGroot, R.
A.; Wold, A. Phys. ReV. B: Condens. Matter 1987, 35, 6195.
(37) Brown, T. M.; McCann, E. L., I. Inorg. Chem. 1968, 7, 1227.
(38) Larson, M. L.; Moore, F. W. Inorg. Chem. 1964, 3, 285.
(39) Fleming, G. R. Chemical Applications of Ultrafast Spectroscopy;
Oxford: New York, 1986.
(40) Micic, O. I.; Zongguan, L.; Mills, G.; Sulivan, J. C.; Meisel, D. J.
Phys. Chem. 1987, 91, 6221.
(41) Sengupta, A.; Jiang, B.; Mandal, K. C.; Zhang, J. Z. J. Phys. Chem.
B 1999, 103, 3128.
(46) Kaschak, D. M.; Johnson, S. A.; Hooks, D. E.; Kim, H.-N.; Ward,
M. D.; Mallouk, T. E. J. Am. Chem. Soc. 1998, 120, 10887.
(47) Kerimo, J.; Adams, D. M.; Barbara, P. F.; Kaschak, D. M.; Mallouk,
T. E. J. Phys. Chem. B 1998, 102, 9451.
(48) Albrecht, A. J. Mol. Spectrosc. 1961, 6, 84.
(49) Tao, T. Biopolymers 1969, 8, 609.
(50) Evans, B. L.; Young, P. A. Proc. R. Soc. London, Ser. A 1965,
284, 402.
(51) Casadori, F.; Frindt, R. F. Phys. ReV. B: Condens. Matter 1970,
2, 4893.
(52) Beal, A. R.; Hughes, H. P. J. Phys. C 1979, 12, 881.
(53) Liang, W. Y. J. Phys. C: Solid State Phys. 1973, 6, 551.
(54) Frindt, R. F.; Yoffe, A. D. Proc. R. Soc. London, Ser. A 1963,
273, 64.
(55) Gutierres, M.; Henglein, A. Ultrasonics 1989, 27, 259.
(56) Hu, J.; Li, L.; Yang, W.; Manna, L.; Wang, L.-W.; Alivisatos, A.
P. Science 2001, 292, 2060.
(42) Jakubowicz, A.; Mahalu, D.; Wolf, M.; Wold, A.; Tenne, R. Phys.
ReV. B: Condens. Matter 1989, 40, 2992.
(43) Salvador, P.; Chaparro, A. M. J. Phys. Chem. 1996, 100, 760.
(44) Hodes, G.; Watkins, E.; Mantell, D.; Brillson, L. J.; Peisach, M.;
Wold, A. J. Appl. Phys. 1992, 71, 5077.
(45) Evans, B. L. In Physics and Chemistry of Materials with Layered
Structures; Lee, P. A., Ed.; Reidel: Dordrecht, 1976; Vol. 4, p 1.
(57) Li, L. L.; Hu, J.; Yang, W.; Alivisatos, A. P. Nano Lett. 2001, 1,
349.