Kinetics of nested inorganic fullerene-like nanoparticle formation

Article abstract of DOI:10.1021/ja973205p

Recently, a model for the growth mechanism of inorganic fullerene-like (IF) nanoparticles of MS₂ (M = Mo, W) from the respective oxides was presented (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). According to this model, sulfidization of oxide particles of < 150 nm leads to the formation of a sulfide encapsulate with oxide core, which is progressively converted into a hollow IF nanoparticle. Using transmission electron microscopy, the sulfidization of a group of oxide nanoparticles is demonstrated step by step. This study provides direct evidence for the quasi spiral growth of the sulfide layers into the oxide nanoparticle core. However, the mechanism for the formation of the first closed sulfide layer which engulfs the oxide encapsulate remained a puzzle thus far. The analysis of the kinetics of simultaneous reduction and sulfidization of WO₃ powders suggests the occurrence of a unique driving force for the fast growth of the first curved sulfide layer (001) around an oxide nanoparticle. According to the present model, a synergy between the reduction and sulfidization processes which occurs in a very narrow window of parameters leads to the formation of the first one or two closed sulfide layers. The present study is not limited to the sulfides. The formation of IF-WSe₂ (Tsirlina, T.; Feldman, Y.; Homyonfer, M.; Sloan, J.; Hutchison, J. L.; Tenne, R. Fullerene Sci. Technol. 1998, 6, 157) is found to be consistent with the same kind of model.

Full text of DOI:10.1021/ja973205p

4176
J. Am. Chem. Soc. 1998, 120, 4176-4183
Kinetics of Nested Inorganic Fullerene-like Nanoparticle Formation
Y. Feldman, V. Lyakhovitskaya, and R. Tenne*
Contribution from the Department of Materials and Interfaces, Weizmann Institute of Science,
RehoVot 76100, Israel
ReceiVed September 11, 1997
Abstract: Recently, a model for the growth mechanism of inorganic fullerene-like (IF) nanoparticles of MS2
(M ) Mo, W) from the respective oxides was presented (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). According to this model, sulfidization of oxide particles of <150 nm leads to the formation of a
sulfide encapsulate with oxide core, which is progressively converted into a hollow IF nanoparticle. Using
transmission electron microscopy, the sulfidization of a group of oxide nanoparticles is demonstrated step by
step. This study provides direct evidence for the quasi spiral growth of the sulfide layers into the oxide
nanoparticle core. However, the mechanism for the formation of the first closed sulfide layer which engulfs
the oxide encapsulate remained a puzzle thus far. The analysis of the kinetics of simultaneous reduction and
sulfidization of WO3 powders suggests the occurrence of a unique driVing force for the fast growth of the first
curved sulfide layer (001) around an oxide nanoparticle. According to the present model, a synergy between
the reduction and sulfidization processes which occurs in a very narrow window of parameters leads to the
formation of the first one or two closed sulfide layers. The present study is not limited to the sulfides. The
formation of IF-WSe2 (Tsirlina, T.; Feldman, Y.; Homyonfer, M.; Sloan, J.; Hutchison, J. L.; Tenne, R.
Fullerene Sci. Technol. 1998, 6, 157) is found to be consistent with the same kind of model.
1. Introduction
A general mechanism for fullerenes’ formation has been
proposed14,15 in which the small graphitic sheets bend in an
attempt to eliminate the large number of highly energetic
dangling bonds present at the edges of the growing nanostruc-
ture. The experimental techniques for the production of carbon
fullerene molecules include laser vaporization,14,16 arc dis-
charge,17-20 and electron irradiation.21-26
Real crystals have many random point defects that lead to a
curvature of the crystallographic planes in their close proximity.
However, in some materials, like mineral structures, such as
clino-asbestos, cylindrite, and halloysite, long-range order of
curved atomic layers exists.3,4 Bursill developed the idea of a
spiral lattice as a new type of crystallography, based on the
Archimedian spiral.5
Boron nitride9 and MoS2 onion-like nanoparticles were
27
obtained under intense irradiation in transmission electron
microscope (TEM). MoS2 onion-like nanoparticles were ob-
tained also from amorphous MoS3 by short electrical pulses from
the tip of a scanning tunneling microscope.28 Most likely, the
driving force here involves energy minimization of the dangling
The most outstanding example of sphere-like curved atomic
structure is found in the case of the fullerenes.6 Nested
fullerenes (onion-like structures)7 and nanotubes8 further extend
the concept of spherical symmetry in molecules up to the size
of nanoparticles (>3 nm) and beyond. The discovery of
fullerenes stimulated the search for similar structures in other
layered compounds. Hexagonal boron nitride, having a layer
structure similar to that of graphite, shows a tendency toward
curling and forming concentric shells with onion-like struc-
tures.9,10 Earlier, layered compounds with more complicated
structure, such as MS2 (M ) Mo, W), were shown to form
onion-like nanoparticles (inorganic fullerene-like, IF) and
nanotubes.11-13
(11) Tenne, R.; Margulis, L.; Genut, M.; Hodes, G. Nature 1992, 360,
444.
(12) Margulis, L.; Salitra, G.; Tenne, R.; Talianker, M. Nature 1993,
365, 113.
(13) Feldman, Y.; Wasserman, E.; Srolovitz, D. J.; Tenne, R. Science
1995, 267, 222.
(14) Zhang Q. L.; et al. J. Phys. Chem. 1986, 90, 525.
(15) Goroff, Nancy S. Acc. Chem. Res. 1996, 29, 77.
(16) Kroto, H. W. Nature 1987, 329, 529.
(17) Kra¨tschmer, W. A.; Lamb, L. D.; Fostiropoulos, K.; Huffman, D.
R. Nature 1990, 347, 354.
(18) Guo, T.; Jin, C.; Smalley, R. E. J. Phys. Chem. 1991, 95, 4948.
(19) Saito, Y.; Yoshikawa, T.; Inagaki, M.; Tomita, M.; Hayashi, T.
Chem. Phys. Lett. 1993, 204, 277.
(20) Jiao, J.; Zhou, D.; Seraphin, S. ICEM 13, Paris, July 17-22, 1994.
(21) Ugarte, D. Nature 1992, 359, 707; Eur. Phys. Lett. 1993, 22, 45;
Chem. Phys. Lett. 1993, 207, 473; MRS Bull. November 1994.
(22) Zhang, Y.; Wei, Y.; Jian, L.; Zhan, K.; Lin, S. Phys. ReV. Lett.
1995, 74 (14), 2717.
(23) Zwanger M. S.; Banhart, F. Philos. Mag. B 1995, 72 (1), 149.
(24) Burden, A. P.; Hutchison, J. L. J. Cryst. Growth 1996, 158, 185.
(25) Zwanger, M. S.; Banhart, F.; Seeger, A. J. Cryst. Growth 1996,
163, 445.
(1) 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.
(2) Tsirlina, T.; Feldman, Y.; Homyonfer, M.; Sloan, J.; Hutchison, J.
L.; Tenne, R. Fullerene Sci. Technol. 1998, 6, 157.
(3) Whittaker, E. J. W. Acta Crystallogr. 1955, 571.
(4) Yada, K. Acta Crystallogr. A 1971, 27, 659.
(5) Bursill, L. A.Int. J. Modern Phys. B 1990, 4, 2197.
(6) Kroto, H. W.; Heath, J. R.; O’Brien, S. C.; Curl, R. F.; Smalley, R.
E. Nature 1985, 318, 162.
(7) Iijima, S. J. Cryst. Growth 1980, 50, 675.
(8) Iijima, S. Nature 1991, 354, 56.
(9) Banhart, F.; Zwanger, M.; Muhr, H.-J. Chem. Phys. Lett. 1994, 231,
98.
(26) Fuller, Th.; Banhart, F. Chem. Phys. Lett.
(10) Boulager, L.; Andriot, B.; Cauchetier, M.; Willaime, F. Chem. Phys.
Lett. 1995, 234, 227.
(27) Jose-Yacaman, M.; Lorez, H.; Santiago, P.; Galvan, D. H.; Garzon,
I. L.; Reyes, A. Appl. Phys. Lett. 1996, 69 (8), 1065.
S0002-7863(97)03205-8 CCC: $15.00 © 1998 American Chemical Society
Published on Web 04/18/1998

Inorganic Fullerene-like Nanoparticles
J. Am. Chem. Soc., Vol. 120, No. 17, 1998 4177
For the synthesis of IF-MoS2,13 a portion of 30 mg of MoO3 powder
(>99% pure) is heated (>800 °C) and sublimes and effuses out of a
nozzle where it crosses a stream of H2S gas mixed with a forming gas
(95% N2/5% H2). It takes some 3-5 min for the entire load of MoO3
to sublime. The reaction products are collected on a quartz substrate,
which is positioned 3 cm away from the crossing point of the two gas
streams and is maintained at the same temperature (>800 °C).
Since WO3 is not volatile at these temperatures, the solid (WO3)-
gas (H2S + H2) reaction was preferred, in this case.1 The starting
material for the synthesis of IF-WS2 is a WO3 powder (>99% pure),
with particle sizes smaller than ca. 150 nm. To avoid agglomeration
and fusion of the heated nanoparticles, the powder was carefully
dispersed on the entire floor of the reactor boat, resulting in a complete
exposure of the nanoparticle surface to the reacting gas. 2H-WS2
platelets were predominantly obtained under the following experimental
conditions: packing of the powder was too compact; oxide particles
with sizes above 0.2 µm were used; reaction temperature exceeded 900
°C.
Samples were cooled to room temperature before removing them
from the reactor. They were immediately transferred to a vacuum
desiccator and kept there until shortly before the analysis. A sample
from the reaction product was collected on a Cu grid covered with
amorphous carbon film and examined by CM12 (Philips) TEM. For
the XRD experiments, a Rigaku Rotaflex RU-200B powder diffracto-
meter with a Cu anode was used.
For the HREM analysis, a JEOL 3010 microscope equipped with
300-keV beam was used. A sample of molybdenum oxide was first
sulfidized in the GPR for 5 min.13 A group of the partially sulfidized
nanoparticles was dispersed on a gold grid with SiO support which
can withstand the severe conditions in the reactor chamber was used.
Moreover, this grid has ABC marks which help to locate the same
specimen again and again. After the first HREM examination, the same
grid was inserted back into the reactor and the sulfidization reaction
was repeated for another 10 min. Subsequently, the grid was removed
from the reactor and reexamined by the HREM.
bonds which are formed during crystallization of the amorphous
nuclei.
Curved layers can be formed also by a chemical reaction on
a curved template. Bursill shows high-resolution TEM (HREM)
images of nickel sulfide core structures surrounded by tungsten
sulfide sheets.5 Sanders showed MoS2 spherical nanoparticles,29
which were observed in hydrodesulfurization catalysts of some
oxides. HREM was used for investigating the influence of
curvature on electron diffraction of the nanoparticles. An almost
pure nested fullerene-like (IF) phase (hollow onion-like nano-
particles) and nanotubes of MoS2 were synthesized by a gas-
phase reaction (GPR).13 Another setup for the synthesis of about
1 g/day of IF-WS2 by solid-gas reaction (SGR) was realized.1
In both reactions, oxide particles serve as precursors for the IF
phase formation.
Recently, a model for IF-MS2 formation from oxide nano-
particles in H2S atmosphere was suggested.1 According to this
model, under certain conditions, one or two layers of molyb-
denum/tungsten disulfide encage the surface of the correspond-
ing oxide nanoparticles in a very fast process (<5 s). Subse-
quently, the reduced oxide core of the nanoparticle is progressively
converted into sulfide nanoparticle with an empty core. This
process may take 15-120 min depending on the experimental
parameters.
To unravel the mechanism of the IF formation, the kinetics
of the oxide reduction/sulfidization was studied. A detailed
description of the synthesis of IF has been reported previously1,13
and will be repeated only briefly here. Both GPR and SGR
are used for IF synthesis by sulfidization of oxide powder in
reducing atmosphere (mixture of H2S and forming gas, typically
(95% N2/5% H2)) at elevated temperatures (∼800 °C). Since
molybdenum oxide is volatile above 700 °C, the GPR was
adopted for synthesis of IF-MoS2. Consequently, the size of
IF-MoS2 is determined primarily by the reaction parameters.
In this case the shape and size of the precursor (molybdenum
oxide) particles can be controlled by varying the reaction
conditions. Tungsten oxide is not volatile at least until 1000
°C, and hence, SGR was opted for the synthesis of IF-WS2.
Therefore the size of IF-WS2 depends only on the size of the
incipient oxide nanoparticles which must be less than 200 nm.
For this reason, the tungsten oxide into IF-WS2 conversion
was preferred for the present kinetic study.
3. Results
3a. Direct Evidence for the IF Growth Mechanism. In
previous works,1,13 IF formation from oxide nanoparticles was
followed by retracting a portion of the powder from the reactor
at different times and analyzing the powder using various
techniques. To get a more firm picture of the IF growth model,
it would be desirable to follow the (sulfidization) reaction on a
given group of oxide nanoparticles.
Toward this goal, the time evolution of a group of nanopar-
ticles was examined by sequential HREM viewing/reaction
procedure. Figure 1A shows an assortment of molybdenum
oxide particles which were sufidized in the GPR for 5 min.13
The nanoparticles are made of a MoO2 core covered by five or
six closed layers of MoS2. Subsequently, the grid was placed
again in the reaction chamber and withdrawn after an additional
10-min reaction. Figure 1B shows the HREM image of the
same group of nanoparticles. Clearly, an additional number of
internal sulfide layers were formed at the expense of the
encapsulated oxide and a hollow core is left in the center of
each nanoparticle. These results give direct evidence in support
of the growth mechanism of the inorganic fullerene-like
nanoparticles from an oxide precursor as proposed before.1
3b. Influence of Hydrogen on IF Formation (Hydrogen-
Free Reduction). According to thermodynamic calculations,2,30
a reducing atmosphere is not required for the sulfidization of
molybdenum or tungsten oxides. To test this point, WS2
synthesis in hydrogen-free atmosphere was carried out. WO3
powder with an average particle size of 150 nm was placed in
the reactor (820 °C). Sulfur pressure (1 Torr) was provided by
While previous studies helped to elucidate the growth
mechanism of the IF particle after the first few layers have been
formed, the great puzzle which remains to be understood is the
formation of the first one or two closed sulfide layers on top of
the oxide nanoparticle. This closed sulfide layer serves as a
template for the furthering of IF growth. Its formation is
counterintuitive to thermodynamic arguments, since the elastic
energy of bending must be accounted for. To address this
central issue, the kinetics of the IF synthesis from the respective
oxide nanoparticles and the chalcogen gas are further studied.
A kinetic model which takes into account the synergy between
the formation of shear planes within the oxide nanoparticle and
the surface sulfidization reaction is described. The predictive
power of this model is tested by using a number of test cases
which are described in the text.
2. Experimental Section
The procedures for the synthesis of the IF particles are similar to
the ones described in earlier publications.1,13
(28) Homyonfer, M.; Mastai, Y.; Hershfinkel, M.; Volterra, V.; Hutchi-
son, J. L.; and Tenne, R. J. Am. Chem. Soc. 1996, 118, 7804.
(29) Sanders, J. V. J. Electron Microsc. Tech. 1986, 3, 67.
(30) Knacke, O.; Kubaschewski, O. Thermochemical Properties of
Inorganic Substances; Springer-Verlag, Berlin, 1991.

4178 J. Am. Chem. Soc., Vol. 120, No. 17, 1998
Feldman et al.
Table 1. Kinetics of the Reduction of WO3 Nanoparticles into
Suboxides
average
size of
starting
gas
atmosphere, oxide
product of
reaction
(XRD)
average
particle size
(TEM) (µm)
flow rate
particles temp reaction
(µm)
(cm3/min)
(°C) time (h)
H2 (1%)/N2,
100
∼0.1
630
1.5
WO3
650
1.5
W20O58
∼0.02 500
1.5
1.5
WO3
W20O58
530
not changed
560
570
2
1.5
WO3
W20O58
600
1
W20O58
W18O49
W
+
0.06-0.1
0.03-0.1
+
H2 (5%)/N2,
100
∼0.1
650
770
1.5
WO2 + W
Figure 1. TEM images of a group of molybdenum oxide/molybdenum
sulfide nanoparticles: (A) after 5 min anneal in H2S/forming gas
atmosphere; (B) after additional 10 min anneal in the same atmosphere.
0.2
0.03
0.3
W18O49
0.1-1
0.03-0.3
1-2
820
970
W18O49
WO2 + W
0.3
1-6
630
650
1
1.5
WO3
WO3 +
WO2
not changed
0.1-3
3-5
750
970
0.2
0.3
W20O58
WO2
1-6
The experiments were performed in the open (flow) system,
which was previously used for the SGR synthesis of IF.1,13 The
composition of the forming gas used for most experiments was
(5% H2/95% N2). This composition was found to be the most
suitable for IF synthesis. For the sake of comparison, another
composition of the forming gas (1% H2/99% N2) was used in
a few experiments. To cover the entire existence zone for IF
formation, the temperature was varied over a wide interval
(300-1000 °C). The reduction process of WO3 goes through
several intermediate stages. Initially a few suboxides phases
are formed, followed by dioxide. Eventually the pure metal is
obtained. The different intermediate phases can be observed
by careful variation of the experimental parameters. With
everything kept constant, all the intermediate phases of the
reduction process could be followed by simply varying the
annealing time of the experiment. The various crystalline phases
were deduced from XRD measurements. The average particle
size was estimated from the TEM images. Furthermore, TEM
observations clearly demonstrated that each oxide particle
consisted of smaller crystallites which were joined through their
grain boundaries. Representative samples were taken for TEM
examination during the reduction process. The results of these
experiments are summarized in Table 1.
On the basis of the results presented in Table 1, the following
inferences can be made:
1. The lower the concentration of hydrogen in the forming
gas mixture, the slower was the reduction process and the higher
was the temperature necessary for the reaction.
2. The size of the oxide particles plays an important role in
the reduction process. The smaller are the oxide particles, the
lower the temperature required to begin the reduction process.
3. The integrity of the oxide particles is preserved so that
their initial shape is unchanged until some temperature is reached
(∼600 °C). This temperature depends on the size of the starting
oxide particles and the hydrogen concentration. Using TEM it
was found that the grain boundaries between the crystallites
Figure 2. X-ray diffractograms of tungsten oxide converted into
tungsten disulfide in the absence (curve A) and presence (curve B)
of H2.
preheating sulfur powder at 180 °C. The gaseous sulfur was
transported to the reactor by the carrier gas (N2). An TEM/
XRD study shows that 2H-WS2 platelets were obtained almost
exclusively under these conditions. Figure 2A shows XRD of
the reaction product. The position of the (0002) peak is in
accordance with the conclusions of the TEM observation.
However, when forming gas (5% H2/95% N2) was used instead
of pure N2, IF-WS2 (Figure 2B) was obtained under the same
experimental conditions. Both the shift of the (0002) peak of
the XRD spectra (Figure 2B) and the TEM observations support
this conclusion.13 Thus, the IF phase cannot be obtained from
the oxide nanoparticles without hydrogen. To further understand
the influence of the reduction process on IF formation, the
kinetics of the oxide reduction by hydrogen were investigated.
3c. Kinetics of WO3 Reduction by Hydrogen. It is well-
known that the reduction kinetics of oxides depend on the
specific surface area of the powder.31 Therefore, three kinds
of WO3 powders (>99% pure) with different average particle
sizes were used. The first two oxide powders with average
particle sizes of 3-5 µm and 150 nm were commercially
available products. The third precursor oxide with particle sizes
<30 nm was obtained by ball milling the second powder for 2
days. The oxide particles were multicrystalline with clear grain
boundaries separating the different crystallites.
(31) Sloczynski, J.; Bobinski, W. J. Solid State Chem. 1991, 92, 436.

Inorganic Fullerene-like Nanoparticles
J. Am. Chem. Soc., Vol. 120, No. 17, 1998 4179
Table 2. Kinetics of the Reduction/Sulfidization of WO3 Nanoparticles into Suboxides and Its Conversion into IF-WS2
average size of
starting oxide
particles (µm)
gas atmosphere,
reaction
time (h)
product of
reaction (XRD)
average particle
size (TEM) (µm)
flow rate (cm3/min)
temp (°C)
S2 + N2 subl., 100
∼0.1
840
1.5
2H-WS2
<0.5
S2 + H2 (5%)/N2 subl., 100
IF-WS2
not changed
650
750
0.5
0.5
IF-WS2/W20O58
2H-WS2 + W18O49
∼0.02
<0.5
150
300
400
500
700
820
820
18
9
4
18
0.3
0.3
2
WO3
WS2/WO3
WS2/W20O58
a
PN-WS2
not changed
IF-WS2/W20O58
IF-WS2/W18O49
IF-WS2
∼0.1
H2S
4.5
870
970
2
1
IF-WS2 + 2H-WS2 + W18O49
2H-WS2 + W18O49 + WO2
H2 (5%)/N2, 100
2-10
500
3
WO3
not changed
560
620
820
3
1
1
WS2 + W20O58
WS2 + W20O58
3-5
WS2 + W20O58 + W18O49
0.1-3
2-10
870
970
2
1
2H-WS2 + W18O49
2H-WS2 + WO2 + W18O49
a PN ) polycrystal nanoparticle.
disjoin and therefore the oxide particles tend to disintegrate into
smaller particles. This process is believed to be facilitated by
the crystallographic shear process (vide infra) occurring within
the crystallites. The size of the shattered particles decreases
with increasing temperature. XRD measurements did not
discern changes in the (average) crystallites’ size during the
reduction process.
4. The two most abundant intermediate phases are W20O58
(WO2.9) and W18O49 (WO2.72). The first one is stable at lower
temperatures, while the second phase is more typical of higher
temperatures.
5. At elevated temperatures (above 700 °C), a secondary
process of recrystallization of the oxide nanoparticles takes
place. The recrystallized particles (typically >0.5 µm) are
almost monocrystalline platelets even though the starting
material consisted of polycrystalline oxide particles. In contrast
to the shattered particles, the size of the recrystallized oxide
particles increases with temperature.
6. Thus, one can distinguish between three temperature
regions in which a different reduction behavior of trioxide
particles is observed. Up to 600 °C the oxide particles do not
change their initial shape during reduction; they break into
smaller nanoparticles (less than 0.1 µm) at intermediate tem-
peratures (600-700 °C) and grow to large platelets (more than
1 µm) at high temperatures.
3d. Kinetics of WO3 Reduction/Sulfidization Combined.
All the experiments of this series were performed in the same
apparatus and using the same WO3 powder as in the previous
section. The addition of H2S in the reaction chamber was the
sole change. The data of this study are summarized in Table
2. As was reported in the previous work (ref 1), the sulfidization
process of the oxide nanoparticles can be divided into two time
domains. In the first stage, which lasts only a few seconds,
the surface of the oxide nanoparticle is converted into a sulfide
shell consisting of one or two monolayers of metal disulfide.
In the second stage lasts 30-120 min or more.
2. At low temperatures, the rate of oxide sulfidization is
higher than the reduction rate. For example, tungsten sulfide
is obtained on the surface of the oxide particles (0.1 µm) after
9 h of annealing in forming gas + H2S atmosphere already at
300 °C, while no reduction occurs under similar conditions for
the same oxide particles.
3. The rate of reduction of the oxide particles increases with
temperature much faster than the rate of sulfidization of the
crystallites.
4. At elevated temperatures (>900 °C), large 2H-WS2
crystallites are obtained by the reduction/sulfidization reaction
from oxide particles of any size (0.1 µm to >3 µm).
5. Most surprisingly, while in the absence of H2S gas (Table
1) both small and large particles breakdown into smaller
particles, at temperatures between 600 and 850 °C, only the
large particles disintegrate in H2S atmosphere while small
particles maintain their integrity.
6. There is a wide temperature interval (650-850 °C) in
which the sulfide layer wraps an oxide nanoparticle (<200 nm),
i.e., IF nanoparticle with oxide core is obtained. In this range
of experimental parameters, the oxide core is progressively
consumed and is converted into a hollow IF after 60-120 min.
7. Typically, large IF nanoparticles (100-150 nm) are
produced from the corresponding oxide nanoparticles at 780-
830 °C, while appreciably lower temperatures (600-700 °C)
are required to obtain IF from the smaller oxide nanoparticles
(10-30 nm).
8. During the conversion of oxide nanoparticles into IF-
WS2, the W20O58 (WO2.9) phase is observed at 700 °C, while
W18O49 (WO2.72) is found to be the stable phase at 820 °C. As
was noted above, MoO3 is transformed into MoO2 during IF-
MoS2 formation at 800 °C.
In view of the air sensitivity of sulfide particles, a series of
experiments was conducted to check the time invariance of the
sulfide/oxide composite nanoparticles. It was found that the
XRD of a few samples which were analyzed in various time
laps after they have been prepared did not show any time
variation. This result is not surprising in view of the fact that
the sulfide shell of nanoparticles is essentially seamless and the
The following inferences can be drawn on the basis of the
results presented in Table 2:
1. Comparison between the results of Tables 1 and 2 shows
that H2S serves as an auxiliary reducer of the oxide.

4180 J. Am. Chem. Soc., Vol. 120, No. 17, 1998
Feldman et al.
II in Figure 3) initially. Eventually, when the grains of reduced
oxide coalesce in the oxide particle, the rate of reduction
decreases with time (termination stage III). This explains the
sigmoidal shape of the reduction curve.
Initially, the reduction of molybdenum trioxide to MoO2 was
considered to be a one-step reaction. Burch37 was the first to
recognize that Mo4O11 is an intermediate product of the reaction.
This hypothesis was confirmed experimentally by Ueno et al.,39
who measured the rate of formation and disappearance of
Mo4O11 in the course of reduction of MoO3 in a high-
temperature X-ray chamber. Sloczynski,42 analyzing the data
of Ueno et al., has concluded that changes in the content of the
intermediate and final product of the MoO3 reduction can be
ascribed to an autocatalytic effect. Note also that the intermedi-
ate MoO3-x phases are substantially less stable than the
intermediate phases of tungsten, i.e., W20O58 and W18O49. In
the range of temperatures used in the present study, MoO2 is
the prevalent phase obtained upon reduction of MoO3.
Figure 3. Schematic drawing describing the time evolution of the
fraction of reduced oxide (R) for the nucleation model.
oxide core is protected from the ambient atmosphere by the
sulfide shell.
According to the nucleation model, surface oxygen atoms
are removed from the lattice by reduction, leaving behind an
anion vacancy.43 When the concentration of vacancies reaches
a critical value, the vacancies are annihilated by rearrangement
of the lattice with the eventual formation of small grains of
lower-valency oxide. Such rearrangement can be readily
accomplished if the oxide particle tolerates the presence of
crystallographic shear (CS) planes. The incipient oxide grains
(on the surface of the oxide particle) grow inward by rearrange-
ments of the reduced metal ions and diffusion of the oxygen
ions outward. The shear process is more facile in the case of
molybdenum oxide layered structure than for tungsten oxide
(distorted ReO3 structure), which explains the difference in the
reduction kinetics of both phases (vide supra).
According to the calculations of Cormack et al.,44 the
concentration of vacancies in equilibrium with shear planes
decreases with temperature. Therefore, lower critical concentra-
tion (threshold value) of vacancies is required for the shear
process to occur at higher temperatures. On the other hand,
the rate of vacancy formation increases with temperature.
Therefore, the induction period for generation of threshold
concentration of vacancies on the crystallite surface decreases
with temperature. Indeed, experimental data show that at
elevated temperatures (above 600 °C) the trioxide particles
disintegrate and are broken into smaller suboxide crystallites
within a few minutes. Apparently, the fast shear process causes
disintegration of the oxide particles. Thus, the first step of
reduction (WO3 f WO3-x) at high temperatures cannot be
described by the nucleation model and the shear process
becomes a rate-limiting step.
4. Discussion
4a. MO3 (M ) Mo, W) Reduction Kinetics. In the
majority of papers34-40 dedicated to the reduction of MoO3 by
hydrogen or by hydrocarbons at temperatures below 600 °C,
the degree of reduction, i.e., the ratio R of the reduced phase/
total oxide, with time (t) was shown to exhibit a sigmoidal curve,
which indicates that the rate of reaction versus t - R′ ) dR/dt
goes through a maximum (Figure 3). A similar characteristics
for the reduction process of WO3 has been documented.41
Figure 3 shows a schematic representation of the fraction of
reduced oxide with respect to time R(t) for the nucleation
model.40 According to this model, the first stage of the reduction
process is typified by an induction period during which oxygen
vacancies accumulate at the oxide surface until a critical value
is reached. If the critical value is high and the induction period
is long, the first stage is a rate-limiting step for the entire process.
Presumably, an autocatalysis of the reduction occurs in the
second stage (Figure 3). Since dissociative adsorption of
hydrogen might be enhanced by the presence of reduced metal
oxide, which has a higher density of coordinately unsaturated
metal cations of lower oxidation states, the reduction process
could be autocatalytic (propagation stage II). Therefore, the
nuclei of reduced oxide, which are formed on the oxide particle
surface, self-activate the reduction process. Moreover, as the
grain size of the reduced oxide increases, the interfacial area
between the reduced oxide grain and the oxide core expands
and the supply of hydrogen increases, thereby increasing the
rate of reduction as the process proceeds. Thus the fraction of
reduced oxide as a function of time is concave upward (stage
The shear process (stage II) is equivalent to the hopping of
an atom to neighboring vacancy in the lattice, however it is
believed that the octahedron hops as a whole at the instant of
the event (Figure 4C). The diffusion of a vacancy can be
described as a sequence of jumps which are described by the
Frenkel model.45 According to this model, the hoping fre-
quency, p, which is in fact the probability that sometime during
a unit time the atom will have sufficient thermal energy to pass
over the barrier between neighboring atoms, is
(32) Muijsers, J. C.; Weber, Th.; van Hardeveld, R. M.; Zandbergen, H.
W.; Niemantsverdriet, J. W. J. Catal. 1995, 157, 698.
(33) Spevack, P. A.; McIntyre, N. S. J. Phys. Chem. 1993, 97, 11031.
(34) Batist, Ph. A.; Karteijns, C. J.; Lippens, B. C.; Shuit, G. C. A. J.
Catal. 1967, 7.
(35) Chijikov, D. M.; Ratner, Yu, E.; Tsvetkov, Yu, V. IzV. Akad. Nauk
SSR Metal. 1970, 70, 8.
(36) Ratnasamy, P.; Ramaswamy, A. V.; Banerjee, K.; Sharma, D. K.;
Ray, N. J. Catal. 1975, 38, 19.
(37) Burch, R. J. Chem. Soc., Faraday Trans. 1978, 74 (12), 2982.
(38) Gajardo, P.; Grange, P.; Delmon, B. J. Chem. Soc., Faraday Trans.
1980, 76, 929.
(39) Ueno, A.; Kotera, Y.; Okuda, S.; Bennet, C. D. Chem. Uses
Molybdenum Proc. Int. Conf. 5th 1982, 250.
(42) Sloczynski, J. React. Solids 1989, 7, 83.
(43) Miyano, T.; Iwanishi, M.; Kaito, C.; Shiojiri, M. Jpn. J. Appl. Phys.
1983, 22 (5), 863.
(40) Transition metal oxides: Surface Chemistry and Catalysis, Kung,
H. H., Ed.; Amsterdam, Oxford, New York, Tokyo, 1989.
(41) Guo, J.-D.; Whittingham, M. S. Int. J. Modern Phys. B 1993, 7,
4145.
(44) Cormack, A. N.; Jones, R. M.; Tasker, P. W.; Catlow, C. R. A. J.
Solid State Chem. 1982, 44, 174.
(45) Introduction to solid-state physics, 6th ed.; Kittel, Ch. Ed.; Brisbane,
Toronto, Singapore, 1986; p 520.

Inorganic Fullerene-like Nanoparticles
J. Am. Chem. Soc., Vol. 120, No. 17, 1998 4181
Figure 5. Calculated rate of reduction of small (A) and large (B) oxide
nanoparticles and sulfidization (C) and selenization (D) as a function
of temperature.
dissociative adsorption of hydrogen. Therefore, the addition
of sulfur accelerates the rate of the reduction process.
4b. MO3 Reduction/Sulfidization Kinetics. The mecha-
nism of the reaction MO3 f MS2 is still under debate.32,33,46,47
Although MO3 sulfidization occurs more readily in H2/H2S
mixtures than in pure, dry sulfur, the role of hydrogen in the
sulfidization process can be different depending on the reaction
temperature.
Temperature-programmed sulfidization was used32,47 for
studying the sulfidization of MoO3 thin films on SiO2 support.
Analysis of the Mo 3d and S 2p lines of the XPS spectra
suggested that two temperature regimes could be recounted for
this reaction. At low temperatures and up to 600 °C, the first
step is an oxygen/sulfur exchange at the surface of the oxide
precursor without reduction of the metal atoms. This exchange,
which takes place already at room temperature, results in S2-
ligands bonded to MoVI centers (Figure 4B). In the next stage
of the sulfidization process, reductive elimination of S atom
Figure 4. Schematic drawing of the various processes which occur
during the sulfidization of the metal oxide particles. The oxide precursor
is represented as two octahedra connected via a corner oxygen atom.
(A) Preliminary step: oxygen atom leaves the octahedron, and oxygen
vacancy is thus formed. The three possible subsequent steps are shown
below: (B) Sulfidization of the oxide at low temperatures. The rates
of the processes are represented schematically through the size of the
arrows. Thus, the rate of the shear process is slow compared with sulfur
trapping by the vacancy and consequently sulfur trapping predominates.
(C) Sulfidization of the oxide at high temperatures. The rates of both
the shear process and sulfur trapping increase with temperature; the
former one predominates however. Therefore, shear (reduction of the
oxide) prevails at these temperatures. (D) At intermediate temperatures
(600-850 °C), the rates of both processes are on the same order of
magnitude and the sulfidization and shear appear at the same time,
leading to the formation of the superficial sulfide layer engulfing the
oxide nanoparticle. Representation of the situation (a) prior to the
process and (b) after the process has been completed.
2-
from the bridging S2 ligands accompanied by the formation
of Mo-S-Mo bonds and the release of H2S takes place.32 At
high temperatures reduction of the metal atoms prevails (vide
infra).
Spevack and McIntyre33 also established that O-S exchange
occurred on MoVI atoms at low temperatures. They further
speculated that much higher energies would be required for
sulfur to break the multiple Mo-Mo bonds in MoO2.33 Indeed,
Grange46 and the present study have both indicated that the
formation of MoO2 hinders (slows down) the sulfidization even
at high temperatures.
p ) V exp(-Eb/kBT)
(1)
where Eb is the energy barrier height between neighboring
atoms, kB is the Boltzmann constant, and T is the absolute
temperature. Since the octahedra form a kind of an extended
network, the vibration frequency (V) of the surface octahedra
grows up as the size of the oxide particle decreases (Figure 5A).
The relaxation time of vacancy annihilation by the shear
mechanism is equal to 1/p and decreases exponentially with
temperature.
As noted above, the sulfidization rate prevails over the
reduction step at low temperatures and the reactivity of hydrogen
as reductant of MO3 increases with temperature. Since the rate
of reduction at elevated temperatures (up to 1000 °C) is very
fast, the sulfidization of the oxides takes place after the reduction
of MoO3 to MoO2 has been completed.47 Sulfidizing in this
range of temperatures can be therefore described mainly as O-S
exchange bound to MoIV atoms.47 The present study shows that
at high temperatures (>600 °C) WO3 reduces very quickly to
suboxide phase WO3-x and subsequently it is slowly transformed
into the respective sulfide (see Table 2).
In a number of papers,33,46 including the present one, the
presence of sulfur in the reaction chamber was found to catalyze
the reduction process. Apparently, oxygen-sulfur exchange
influences the autocatalytic reduction mechanism. Indeed, when
sulfur replaces oxygen in the corner of an octahedron, the degree
of unsaturation of the metal cation increases, since sulfur is less
electronegative compared with oxygen. Furthermore, the higher
degree of unsaturation of the metal cation increases the
Therefore, the relationship between reduction and sulfidization
rates determines the kinetics of reduction/sulfidization of the
oxide. At low temperatures, the induction period of the
(47) Arnoldi, P.; van den Heijkant, J. A. M.; de Bok, G. D.; Moulijn, J.
(46) Grange, P. Catal. ReV.-Sci. Eng. 1980, 21 (1), 135.
A. J. Catal. 1985, 92, 35.

4182 J. Am. Chem. Soc., Vol. 120, No. 17, 1998
Feldman et al.
reduction is rather long and the sulfidization is accomplished
before the reduction of the nanoparticles has started. At high
temperatures, the first step of reduction from trioxide to suboxide
is very quick and the sulfidization takes place only after
reduction of the oxide nanoparticles. This situation further slows
the rate of sulfur/oxide exchange, since much higher energies
would be required for sulfur to break the stronger oxygen-
metal bond of sub- or dioxide as compared with this bond in
trioxide. The transition zone between the two regimes is
discussed separately below.
The creation of oxygen vacancies on the surface of the oxide
particles, by hydrogen, is most probably the first step in the
entire (reduction-sulfidization) process (Figure 4A). Thence
two competing events follow: the first one is vacancy annihila-
tion by the shear process; the second event is sulfur occupation
of the oxygen vacancy site. As noted above, sulfur occupation
of the vacancy sites is more rapid than vacancy annihilation by
a shear process at low temperatures (Figure 4B), while the
reduction process prevails at high temperatures (Figure 4C).
The exponential dependence of the relaxation time of the
shear process on temperature was discussed above. The
relaxation time of the oxygen vacancy on the crystallite surface
through sulfur occupation can be estimated from the kinetic
theory of gases. The sulfur concentration defines an average
distance between a vacancy on the particle surface and the
nearest sulfur atom in the gas atmosphere. The kinetic energy
of a particle in the gas atmosphere is equal to kBT, and hence,
the average velocity (V) of a sulfur atom is
to a chemical reaction or substitution at this instant of time.
When the probability of oxygen leaving (reduction of oxide)
equals the probability of the sulfur occupation of the oxygen
vacancy site, the kinetic rate of the oxygen substitution by sulfur
is the largest (Figure 4D). On the other hand, as noted above,
the sulfur-oxygen exchange accelerates the autocatalytic reduc-
tion effect. Thus, a synergy between the autocatalytic reduction
and oxygen-sulfur exchange prevails when the rates of the two
processes are equivalent.
From the microscopic viewpoint, there is a significant
difference between the low (high)-temperature processes and
those at intermediate temperatures. In the low and high-
temperature processes (Figure 4B,C), no oxygen atom is
abstracted. In contrast to that, one oxygen atom is released when
the synergy between the two processes (shear/sulfidization)
occurs (Figure 4D). Moreover, this situation leads to the
bonding of the captured sulfur atom to two neighboring MV
atoms, which is opposite to the situation at low temperatures.
Subsequent abstraction of oxygen atoms and shear and con-
comitant sulfur atom capturing from the gas phase promotes
the growth of the MS2 surface layer encasing the oxide
nanoparticle.
As shown above (section 3b), in the absence of hydrogen,
sulfidization of the oxide leads to the formation of 2H-WS2.
It is believed that the synergy between the reduction and
sulfidization processes proVides the conditions for IF formation
in a certain range of experimental parameters. Following the
formation of sulfide nuclei on the surface of an oxide nano-
particle, these two processes proceed simultaneously. The
reduction process extends quickly inward by the fast rearrange-
ments of the lattice and the formation of shear planes. This
process leads to the “recrystallization” of the suboxide nanoc-
rystal. On the other hand, oxygen/sulfur exchange cannot extend
inward (into the particle core) due to steric hindrance of sulfur
diffusion into the oxide core. The fast-growing sulfide layer
replicates the morphology of the surface of the oxide nanopar-
ticle. The anisotropy of the sulfide growth rate provides no
time for a 3-D growth pattern of the nanocrystal and, thus,
further promotes the formation of the curved sulfide 2-D (001)
layer. It should be noted that the growth rate of the sulfide
layer on the surface of the large oxide particles is not sufficient
to wrap the entire surface, while reduction of the oxide takes
place and they all break up into smaller crystallites (Table 2),
which precludes IF formation from large oxide particles.
V ) (2kBT/m)1/2
(2)
where m is the mass of a sulfur atom. Thus, for a certain sulfur
concentration, the relaxation time of oxygen vacancy by sulfur
occupation is inversely proportional to the square root of the
temperature.
4c. Synergistic Model. Figure 5 illustrates the schematic
rates of reduction (A and B), sulfidization (C), and selenization
(D) as a function of temperature. To plot this graphs, the
functional dependence of the shear (eq 1) and sulfidization (eq
2) on temperature were calculated using arbitrary parameters,
so that a single cross point between the two rates was obtained.
Therefore the plots have qualitative significance only and they
do not intend to represent a detailed calculation of the kinetic
process. The sulfidization (selenization) reaction goes faster
at low temperatures. Conversely, the reduction rate is much
larger at high temperatures. The cross point of the reduction
and sulfidization (selenization) curves shows the temperature
range where the reduction rate is equal to the sulfidization
(selenization) rate. It is suggested that instead of a competition
between the two processes, a synergy of the two processes exists
at this range of temperatures.
Figure 4 shows a schematic drawing of the shear (reduction)
and sulfidization processes at different temperatures. The first
step (step I) of vacancy formation by abstraction of one oxygen
atom by a hydrogen atom is common to all temperatures. In
the next step (II), any of three processes can take place; their
relative rates vary with the temperature. At low temperatures,
sulfur trapping is relatively faster than shear (Figure 4B).
Conversely, at high temperatures, the shear process is faster
(Figure 4C). At intermediate temperatures (600-850 °C), the
rates of the two processes are equivalent (for a given experi-
mental parameters this range is much narrower). During the
shear of an octahedron, the oxygen atom, which hops from one
corner to a neighboring vacancy, has to break its chemical bond
to its nearest metal neighbor, and consequently, it is more prone
The formation of a thin sulfide skin controls the size of the
fullerene-like nanoparticle. Once the size and shape are fixed,
further transformation of the oxide into sulfide takes place
internally. The growth of the next inner sulfide layers does
not change the morphology of the outer layers (Figure 1).
Therefore, sulfur and oxygen diffuse through available faults
in the sulfide shell.1
The quasi-spiral inward growth process of the curved sulfide
layers predetermines the hollow core of the fullerene-like
nanoparticle. Every inner sulfide layer grows under a neighbor-
ing outer one. The radius of curvature of the innermost layers
decreases as the process proceeds. The smallest radius of
curvature of the sulfide layer, which corresponds to the
minimum hollow core size, was found to be 1.4 nm.13 However,
most IFs have larger hollow cores.13 The oxide core engulfed
by the first closed sulfide shell is the precursor for the
subsequent sulfide growth. The density of the crystalline metal
oxide is higher than that of the metal sulfide. The presence of
a hollow core in the IF particle suggests that the density of the
primary oxide nanoparticles is much less than the formal density

Inorganic Fullerene-like Nanoparticles
J. Am. Chem. Soc., Vol. 120, No. 17, 1998 4183
2
Synthesis of IF-WSe2 shows also a good agreement with
the present model. The cross point between the reduction and
selenization processes (Figure 5) is observed at lower temper-
atures (typically 760 °C) compared with 840 °C for IF sulfide
formation. Higher reaction temperatures yielded platelets (2H)
of the respective selenides. Hence, the synergism between
selenization and reduction, which is a prerequisite for the IF
formation, occurs at lower temperatures than that of the
respective sulfide. This tendency is confirmed by the available
experimental data2 and is further supported by preliminary data
for IF of metal telluride’s (630 °C). However, in this last case,
it is unlikely that dislocation-free IF will be obtained at such
low temperatures.
Figure 6. TEM images of polycrystalline (A) and onion-like (B)
tungsten sulfide nanoparticles.
5. Conclusions
of the (monocrystal) oxide. Moreover, it is sometimes observed
that the volume of the first shell of the sulfide layer (skin) can
be appreciably larger than the volume of the starting oxide
nanoparticle, leaving thus a free void between the sulfide skin
and the oxide. Thus, the size of the hollow core of the IF is
defined by the experimental conditions of this concrete nano-
particle.
Direct evidence for the formation mechanism of the first
fullerene-like layer of MX2 (M ) W, Mo; X ) S, Se, Te) on
the surface of oxide nanoparticle is demonstrated. This process
is followed by a chalcogen/oxide exchange for which a direct
proof is demonstrated on the same group of nanoparticles which
were reacted step by step. It is shown that hydrogen is a
necessary precursor for IF formation from metal oxide nano-
particles.
The synergy between the reaction of the chalcogen molecules
coming from the gas phase and the reduction process of the
oxide for IF-WX2 (X ) S, Se) synthesis is consistent with the
available experimental data. Further work is necessary to
provide a direct proof of this unique process.
As noted above,1 the size of the precursor oxide particle
defines the final dimensions of the IF particle. According to
the synergy model (Figure 5), smaller oxide nanoparticles are
wrapped by a sulfide layer at lower temperatures than bigger
ones, which is confirmed by the experimental data (see Table
2). It is noticed that the size and shape of the IF particles follow
closely those of the oxide precursor (see Figure 2 of ref 1).
Figure 6 shows TEM images of a polycrystalline (A) and
fullerene-like (B) nanoparticles of WS2. The polycrystalline
nanoparticles were grown at 500 °C. This temperature is below
the temperature corresponding to the synergy effect. The
polycrystalline sulfide nanoparticle, like its oxide precursor
nanoparticle, consists of very small randomly oriented nanoc-
rystallites. Oppositely, the IF particle, which is formed via the
synergy effect at much higher temperatures (800 °C), consists
of a single domain and consequently has the long-range order
of the curved atomic (metal disulfide) layers.
Acknowledgment. We are grateful to Dr. J. Sloan of the
University of Oxford for his useful suggestions regarding the
SC process. This research was funded by the following grants:
UK-Israel Science and Technology Foundation; NEDO (Japan)
International research program; Israeli Ministry of Science
(strategic research program on nanomaterials); Petroleum
Research Foundation of the American Chemical Society;
Minerva Foundation (Munich).
JA973205P