Investigations of nonstoichiometric tungsten oxide nanoparticles

Journal of Solid State Chemistry 162, 300}314 (2001)

doi:10.1006/jssc.2001.9319, available online at http://www.idealibrary.com on

Investigations of Nonstoichiometric Tungsten Oxide Nanoparticles

G. L. Frey,*ꢃꢀ A. Rothschild,*ꢃꢁ J. Sloan,-ꢃ? R. Rosentsveig,* R. Popovitz-Biro,* and R. Tenne*ꢃꢂ

*Department of Materials and Interfaces, Weizmann Institute of Science, Rehovot 76100, Israel; -Inorganic Chemistry Laboratory, University of Oxford,

South Parks Road, Oxford OX1 3QR, United Kingdom; and ?Department of Materials, University of Oxford, Parks Road,

Oxford OX1 3PH, United Kingdom

Received April 3, 2001; in revised form June 26, 2001; accepted July 16, 2001

IN HONOR OF PROFESSOR PAUL HAGENMULLER ON THE OCCASION OF HIS 80TH BIRTHDAY

are elongated along one axis. In this process, the tungsten

Recently, pure phases of hollow multi-walled WS₂nanotubes suboxide cores, which are encapsulated inside the WS

ꢁ

were prepared from a surface reaction with reduced tungsten

oxide nanowhiskers. During the process, which starts with

WO_3؊xnanoparticles and 5nishes with WS₂nanotubes, particu-

lar care has been devoted to the evolution of the tungsten oxide

cores once the 5rst encapsulating WS₂layer has been formed.

The reduced tungsten oxide phases were previously studied by

a combination of techniques, including high-resolution transmis-

sion electron microscopy and X-ray di4raction. In the present

study, Raman spectroscopy combined with the previous two

techniques is used to give further detail concerning the structure

of the reduced oxide phases. This study sheds some further light

nanotubes, undergo a certain evolution. The microstruc-

tures of most of these encapsulated suboxides could be, at

some point, described in terms of +hkl, crystallographic

0

shear (CS) planes relative to a ReO -type structure. Further-

ꢂ

more, the type of reduced CS phases encountered in the

tungsten suboxide cores appeared to be directly dependent

on the morphology of the sul"dized particles. Quasi-spheri-

cal faceted oxide nanoparticles, which are converted into

nested fullerene-like nanoparticles (onions), were found to

consist of a mixture of complex CS structures. Some oxide

on the reduction process of the tungsten suboxide phases and the encapsulates with +103, CS planes were also found (5).

0

growth mechanism of oxide nanowhiskers and subsequently WS₂

nanotubes from quasi-isotropic tungsten oxide nanoparticles.

( 2001 Elsevier Science

Usually, the encapsulated nanoparticles consisted of more

than one oxide nanograin with clear grain boundaries sep-

arating the reduced oxide phases (1, 5). On the other hand,

some suboxide phases encapsulated inside WS -nanotubes

ꢁ

were found to form ordered +001, CS planes (3, 5). Careful

0

I. INTRODUCTION

examination of the encapsulates using high-resolution elec-

tron microscopy (HRTEM) revealed that, although the CS

Compounds of tungsten oxide with various stoichiomet- were ordered along the growth axis of the encapsulated

ries have been documented. They were studied in the past in oxide nanowhisker, the distances between the CS planes

relation to their importance as, e.g., catalysts and electroch- were not always constant (3b, 5). In other cases, the +001,

0

romic panels. By analogy with the synthesis of WS

CS formed two or more separated domains within the

ꢁ

fullerene-like particles from WO nanoparticles (1, 2), pure nanowhisker with a clearly observable boundary separating

ꢂ

phases of WS nanotubes were grown recently starting from the two phases. Miyano et al. (6) described the +001, CS in

ꢁ

0

needle-like WO

particles as a precursor (3, 4). In this WO

whiskers in terms of a homologous W O

series.

ꢂ\V

L ꢂL\ꢀ

reaction, the W}O bonds in the WO

progressively converted into W}S bonds (in WS ) and the [010] projection of the +001, CS phases of W O and

&&outside in'' reaction leads, after a few hours, to the forma- W₅O₁₄, respectively (5). The CS planes are represented

oxide precursor are Figures 2a and 2b show a schematic representation of the

ꢂ\V

ꢁ

0

ꢄ ꢀꢀ

tion of hollow multi-walled WS nanotubes. Figure 1 shows through the planes of edge sharing WO octahedra. The

ꢁ

ꢅ

a scanning electron micrograph of an assembly of WS

reduction of the oxide entails that n decreases in this series

ꢁ

nanotubes grown from the WO

precursor needles, which and thereby WO

is gradually converted into WO . This

process can be described microscopically by a rapproche-

ꢂ\V

ꢁ

ment of the CS walls separating the ReO domains. A theor-

ꢂ

ꢀ Present address: OE group, Cavendish Laboratory, University of

Cambridge, Madingly Road, Cambridge, CB3 0HE, United Kingdom.

ꢁ Present address: IMEC, SPT Division, Kapeldreef 75, B-3001 Leuven,

Belgium.

etical study of these phases (7) showed that elastic strain

a!ects an attractive interaction between the CS domain

walls, while for n(4 a repulsive interaction predominates.

This explains the great stability of shear structures with low

ꢂ To whom correspondence should be addressed.

300

0022-4596/01 $35.00

( 2001 Elsevier Science

NONSTOICHIOMETRIC TUNGSTEN OXIDE NANOPARTICLES

301

di!raction (XRD) was also performed in order to obtain

more global information on the evolution of the crystalline

structure of the nanoparticles during the reduction process.

Whereas HRTEM and ED can be used to investigate the

microstructure of a single nanowhisker, XRD provides glo-

bal structural information concerning the sample, which

contains many nanoparticles with various reduced phases.

However, because of line broadening e!ects due to the

latter, the interpretation of the XRD data is not unambigu-

ous. Raman spectroscopy is situated somewhat in between

the two extremes, since the optical beam can be focussed to

domains of 1 micro-meter and, if the sample is well disper-

sed, a few nanowhiskers or even a single nanostructure can

be probed, provided the intensity of the signal is su$ciently

large.

Unfortunately, the cross-section (p) for Raman scattering

FIG. 1. Electron micrograph of an assembly of WS nanotubes ob-

tained by the sul"dization of oxide needle-like nanoparticles.

ꢁ

and the absorption coe$cient (a) of the encapsulating WS

ꢁ

layers are much larger than those of the suboxide phases

inside the nanoparticle core. For this reason it was not

possible to detect the Raman signal originating from the

n number against complete reduction. The stability of these encapsulated suboxides using the present setup. To bypass

phases was of key importance in retaining a su$ciently this problem, an alternative strategy was employed. Indeed,

open structure, which permitted slow di!usion of sulphur long oxide (WO ) whiskers (several micro-meters in

ꢂ\V

and oxygen at high temperatures and led to the formation of length) can be produced, starting from precursor needle-like

WS nanotubes (3). Another stable phase in oxide whiskers WO

particles (50 nm in length) by annealing the latter at

ꢁ

ꢂ\V

is the c-W O , which consists of an ordered 2-D lattice of 8403C under a gas mixture of H /N (3, 4). It was often

edge-sharing WO octahedra forming a network of pentag- observed that reduced oxides with ordered superstructures,

onal columns (PC) interspersed with hexagonal channels similar to those oxides formed within the WS nanotubes,

(HC) (8 }10). This phase is represented schematically in occur in the elongated nanowhiskers. Taking advantage of

ꢀꢇ ꢄꢈ

ꢁ

ꢅ

ꢁ

Fig. 2c ([010] projection). No evidence for the existence of this similarity, the evolution of the tungsten suboxides un-

this phase in the encapsulated nanowhiskers was found in der the reductive hydrogen atmosphere was investigated.

the previous study (5). Furthermore, a W O phase, which There appears to exist a relationship between the degree of

ꢆ ꢀꢄ

consists of PC and HC is known as well (11) and is schemati- reduction of the WO₃nanoparticles; their morphology, and

cally represented in Fig. 2d ([010] projection) ꢄ . These two their crystalline structure. Whereas the pure WO₃nanopar-

latter phases are not only highly anisotropic, but they also ticles are essentially isotropic, the shape of the reduced

possess a mechanical strength necessary for the holding nanoparticles (WO ) is generally elongated. This evolu-

ꢂ\V

together of the brittle oxide nanowhiskers during their tion is attributed to the evolution of the lattice structure (see

growth and subsequent conversion into WS nanotubes. below).

ꢁ

Note that although the crystalline structures represented in

Furthermore, the crystalline structure of WO₃is quite

Figs. 2b and 2d are very di!erent, the O/W ratio for these isotropic, but the lattice parameters of the various suboxide

two phases is the same (2.8).

In order to gain further understanding of the WS

phases are highly anisotropic. Moreover, the crystal struc-

ture of WO₂is again quite isotropic, which explains the

ꢁ

nanotube growth mechanism, the evolution of the inter- tendency of this oxide phase to adopt a more symmetric

mediate suboxides was studied during the reduction pro- morphology. Thus, the reduced tungsten-oxide phases go

cess. For this purpose, Raman scattering was used to follow through a cascade of phase transformations during reduc-

the reduction process of the oxide nanowhiskers in addition tion, which a!ects their stoichiometry, and also their mor-

to transmission electron microscopy (TEM), electron dif- phology. Three types of tungsten oxide nanoparticles were

fraction (ED), and high-resolution TEM (HRTEM). X-ray used as precursors in this study. The "rst type, commercially

available, was made of WO₃nanoparticles with a yellow

tint. This phase served as a precursor for the synthesis of

WS₂nanotubes, using the #uidized bed reactor (see below).

ꢄ To distinguish the two forms of W O oxides based on CS planes and

ꢆ

ꢀꢄ

The oxide nanowhiskers were obtained by either interrupt-

ing the reaction after a few minutes, or by having a too little

H₂S #ow in the reactor. The other two types of tungsten

HC columns, we use the notation CS-W O for the crystal shear (i.e.,

ꢀꢄ

Miyano) form of this phase and HC-W O for the form containing

ꢀꢄ

ꢆ

hexagonal channels.

302

FREY ET AL.

FIG. 2. Schematic representation of various reduced tungsten oxide phases with +001, CS (the W O

family of phases): (a) CS-W₄O₁₁lamellar

0

L

ꢂL\ꢀ

phase (6); (b) CS-W₅O₁₄lamellar phase (6); (c) cross section ([010] direction) of a c-W

O

phase containing a network of PC and HC; (d) cross section

ꢀꢇ ꢄꢈ

([010] direction) of HC-W₅O₁₄with a network of PC and HC. The arrows show the growth direction of the nanowhisker.

II. EXPERIMENTAL SECTION

suboxides precursors (types I and II) were prepared in-

house by a documented procedure (3) (see below). Type I

consists of quasi-spherical faceted particles, while type II is

composed predominantly of needle-like particles. Both

II.1. Preparation of the Samples

II.1.a. Synthesis of the WO

particles (type I and type II

ꢂ\V

types of precursor were studied by a combination of TEM, precursors). A few tens of mg of tungsten suboxide par-

HRTEM, and XRD. Analysis with Raman scattering, which ticles (WO ) were produced by heating a tungsten "la-

ꢂ\V

has not been used for characterization of these types of ment (model ME11 from the R. D. Mathis company) in the

materials so far, is being reported in this work.

presence of water vapor inside a vacuum chamber (3). The

NONSTOICHIOMETRIC TUNGSTEN OXIDE NANOPARTICLES

303

following procedure was used: a tungsten "lament was pre- 200B apparatus having a CuKa anode. The samples were

heated for a few minutes in an evacuated (10\ꢄ Torr) bell-jar also analyzed by a TEM (Philips CM-120 (120 kV)) supple-

in order to remove the super"cial oxidized layer. Water mented by a JEM-4000EX HRTEM operated at 400 kV.

vapor was then allowed to di!use into the vacuum chamber The electron di!raction (ED) patterns were obtained on the

through an inlet, until the desired pressure was reached HRTEM instrument and these were calibrated using TlCl

(additional argon was sometimes introduced into the cham- as a reference standard. The accuracy of the d spacings

FIJ

ber in order to increase the total vapor pressure). The was estimated to be$0.005 nm.

"lament was then heated to around 1600$203C, while the

pressure in the chamber was maintained at a constant value

of a few Torr during the evaporation process. After a few II.3. Raman Scattering

minutes of evaporation, a blue powder condensed on the

Raman spectra were recorded by a Renishaw 1000

bell-jar walls. Depending on the water vapor pressure and

Raman microscope at ambient temperatures using a stan-

the total pressure in the chamber, the accrued powder

dard backscattering geometry. An excitation wavelength

consisted of either amorphous material, or faceted and

corresponding to the 623.8 A line of a He/Ne laser, which

isotropic nanoparticles (type I, 25 nm in diameter) or

was capable of supplying 1 mW of power was used. In order

needle-like particles (type II, 50 nm in length and 15 nm in

to prevent damage to the samples, an optical "ber was used

diameter) (3b). These nanoparticles could be converted

to reduce the incident power to roughly 0.6 mW. In addi-

quantitatively into WS₂nanotubes.

tion, a constant #ow of Ar gas over the sample prevented

II.1.b. Reduction of the WO

II). The progressive reduction of both types of WO

particles (isotropic faceted, type I; needle-like, type II) was intensities were measured using a high-resolution CCD

performed in a reactor heated to 8403C under the #ow of camera.

precursors (types I and oxidation of the particles. Several regions from each sample

ꢂ\V

were probed to ensure reproducibility of the data. Scattering

ꢂ\V

a 1% H /99% N gas mixture. The #ow-rate was main-

ꢁ

ꢂ

\ꢀ

tained constant during all the experiments (200 cm min ),

III. RESULTS AND DISCUSSION

while the reaction time was increased progressively from

one batch to another. Starting with type I particles, "ve III.1. Study of the Reduction Ewect on the WO

ꢂ\V

samples corresponding to a reduction period of 1, 5, 10, 20,

and 60 min were prepared. The same procedure was repeat-

ed with the particles of type II (needle-like), and led to

a series of "ve samples corresponding to a reduction time of

2, 5, 10, 15, and 60 min.

Particles by ¹EM/HR¹EM and ED

With respect to the quasi-spherical precursors (i.e., type

I nanoparticles), the reduction was accompanied by rapid

densi"cation of the samples, and the specimens generally

became tinted and more opaque in the TEM as the reduc-

II.1.c. Synthesis of WO

nanowhiskers using the yuidized tion proceeded. Starting with the needle-like particles (i.e.,

ꢂ\V

bed reactor. More recently, a #uidized bed reactor for the type II particles) led also to the same observation. Indeed, as

synthesis of large amounts (20}30 g/day) of nested was evidenced by ED patterns, both sets of samples went

fullerene-like WS nanoparticles from quasi-spherical WO

nanoparticles has been constructed (12). A slightly modi"ed a wide series of reduced tungsten oxide intermediates, such

through an overall reduction process from WO

to W via

ꢁ

ꢂ

ꢂ\V

reactor of this kind was found to produce WO

nanow- as WO (Table 1a). A clear correlation could be observed

ꢂ\V

ꢁ

hiskers and subsequently WS nanotubes in large amounts between the morphology of the particles, as they appeared

ꢁ

in a single process using commercially available WO

in the TEM, and their stoichiometry, which was determined

ꢂ

nanoparticles (METEK Technologies). However, a pure by ED. Indeed, the further the reduction process has pro-

phase of the nanotubes has not so far been obtained (13). ceeded, the less WO

The separation of nanotube bundles from the nanoparticles faceted spheroidal WO or W particles appeared in the

by a system of sieves has been shown to be achievable (13). products (Tables 1a and 1b). The global mechanism could

whiskers were present and more

ꢂ\V

ꢁ

In some cases, the sul"dization of the oxide nanoparticles therefore be summarized as follows. A "rst step, which goes

was not complete and thus, oxide nanowhiskers were found through a wide variety of tungsten suboxides intermediates,

at the end of the reaction.

and ends up with the formation of WO particles. Then,

a second step follows, which leads to the production of

ꢁ

tungsten particles

II.2. X-Ray Powder Diwraction (XRD), Transmission

Electron Microscopy (TEM), and High-Resolution

Transmission Electron Microscopy (HRTEM)

WO #H PWO #(1!x) H O#xH

[1]

[2]

ꢂ\V

ꢁ

XRD of the samples were measured in the transmission

mode (0.3 mm glass capillaries) of a Rigaku Rota#ex RU-

WO #2H PW#2H O

ꢁ ꢁ ꢁ

304

FREY ET AL.

TABLE 1a

meters in length) are formed at the early stages of the

Evolution of the Morphology of the WO_3؊xParticles

with the Reduction Time (ED Analysis)

reduction process (i.e., within the "rst 5 min, or less). How-

ever, the proportion of these whiskers compared to spheri-

cal-faceted particles is clearly dependent on the type of

precursor. While after 5 min reduction, type I particles con-

sisted of spherical-faceted particles as the majority product,

type II particles were predominantly long oxide whiskers

(Table 1b). As the reduction proceeds, the batches were also

distinguished by a time-lag between the appearance of

the new reduced phases. In the case of type II precursor

(Fig. 3a), the long nanowhiskers (N) that have been formed

during the "rst instant of the reduction period ((1 min)

mostly preserved their shape after a 5 min reduction time.

However, their cores were found to be reduced to WO₂(see

also Tables 1a and 1b). For the same reduction period (5

min) the type I precursor are quickly transformed from

Type I precursor

Type II precursor

(spherical-faceted particles)

(needle-like particles)

Reduction

Reduction Stoichiometry (ED)

time (min)

time

(minutes)

Stoichiometry (ED)

2

5

W O

(W

O

)

1

5

W O

$W

(W O )

L

ꢂL\ꢁ

ꢁꢉ ꢆꢇ

O

L

ꢂL\ꢁ

ꢁꢉ ꢆꢇ

WO #$W

#few

O

#W O

ꢁ

ꢁꢉ ꢆꢇ

L

ꢂL\ꢀ

W O

(n"4, 5, 6)#very few

L

ꢂL\ꢀ

WO #(W

O )*

ꢁ

ꢀꢇ ꢄꢈ

10

WO

10

$W

O

#W O

ꢁ

ꢁꢉ ꢆꢇ

L

ꢂL\ꢀ

(n"4, 5, 6)#more WO

ꢁ

15

60

WO

20

60

WO #W

ꢁ

WO #W

W

ꢁ

elongated WO

faceted nanoparticles with fully reduced metallic cores

(Fig. 3b). It is to be noted, that in the case of type I oxide

nanoparticles into WO spherical-

ꢂ\V

ꢁ

Note. The gas mixture going through the reactor is 1%H /99% N and

ꢁ

its #ow-rate 200 cmꢂ min\ꢀ. Sh denotes short whiskers, L denotes long

whiskers, &&S}F'' denotes spherical-faceted particles, * has been found by

HRTEM.

nanoparticles, the presence of H S gas in the reaction atmo-

ꢁ

sphere led to a rapid encasing of the oxide nanoparticles by

a few WS layers, and the quasi-spherical shape of the

ꢁ

ꢂ

The densities of the WO

precursor (ca. 7.16 g cm ) and nanoparticles is therefore preserved (1). Also, typically +102,

ꢂ\V

ꢂ

the reduced tungsten metal (19.35 g cm ) (14), are therefore or +103, CS planes were found in the reduced oxide core of

consistent with the densi"cation phenomenon, which occurs such nanoparticles (5). The rate of reduction of the oxide

during the overall process. Both WO₂(distorted rutile) and was also found to be slowed down considerably in the case

W (fcc) possess more or less an isotropic crystalline habit, of the WS encapsulated WO

nanoparticles. Whereas

ꢁ

ꢂ\V

while the reduced suboxide WO

istropic in their crystalline structure. Not surprisingly there- W metal may take place within a few minutes, for the bare

fore, the chemical transformation of the WO

phases are highly an- complete reduction into WO and subsequently into

ꢂ\V

ꢁ

phases into oxide type I nanoparticles, the WO

core of the WS

ꢂ\V

ꢁ

WO₂and subsequently into W is followed by a change from enfolded nanoparticles is stable for a few hours under sim-

highly anisotropic to a more or less isotropic morphologies ilar reducing conditions.

(see below).

It appears that, whatever the original WO

(i.e., types I or II), very long oxide whiskers (several micro-

precursor

ꢂ\V

TABLE 1b

Evolution of the Morphology of the WO_3؊xParticles

with the Reduction Time (TEM Observations)

Type I precursor

Type II precursor

(spherical-faceted particles)

(needle-like particles)

Reduction

time (min)

Morphology

(TEM)

Reduction time

Morphology

(min)

(TEM)

2

5

10

15

60

Sh#L

&&S-F''ꢀL

&&S-F''

1

5

10

20

60

Sh

Lꢀ&&S-F''

&&S-F''

FIG. 3. TEM micrographs of the evolution of types II and I precursors

after 5 min reduction. (a) Type II precursor: a long oxide nanowhisker,

representative of the batch. While the shell of the nanowhisker consists of

Note. The gas mixture going through the reactor is 1%H /99% N and

ꢁ

its #ow-rate 200 cmꢂ min\ꢀ. Sh denotes short whiskers, L denotes long

whiskers, &&S-F'' denotes spherical-faceted particles, * has been found by

HRTEM.

a WO

phase, the core is further reduced to WO . (b) Type I precursor:

ꢂ\V

ꢁ

mostly spherical-faceted particles of WO and W core are observed.

ꢁ

NONSTOICHIOMETRIC TUNGSTEN OXIDE NANOPARTICLES

305

Figures 4a }4d document a series of reduced intermediates

observed in the oxide needles across the composition range

WO }WO . Figure 4a shows a WO oxide needle in the

ꢂ

ꢁ

ꢂ

early stages of reduction. The microstructure conforms to

that of well-ordered monoclinic WO viewed along [100].

ꢂ

A small amount of disorder is indicated by the di!use

streaking along [001] in the Fourier transform (FFT) image

(inset). Note that in this whisker, the growth axis is along the

[010] direction. Figure 4b shows a typical nonstoichiomet-

ric WO

needle observed within a more reduced prepara-

ꢂ\V

tion. The extent of disorder in this intermediate is clearly

indicated in the inset ED pattern as parallel rows of di!use

streaked re#ections indicating disorder orthogonal to the

needle axis (bold arrow, inset). The disorder can possibly be

attributed also to the formation of oxygen vacancy walls,

which were postulated to precede the formation of the CS

planes or, alternatively, the formation of random CS planes

within the needles (15, 16). In Fig. 4c, a lattice image corre-

sponding to an ordered intermediate is shown which, in this

case, conforms to a [110] projection of W O (c-tungsten

ꢀꢇ ꢄꢈ

oxide). It is important to emphasize here, that this is only

one of a number of possible ordered intermediates and

others were observed. Figure 4d shows an [100] projection

of an ordered fragment corresponding to the WO phase.

The example in Figs. 5a }5c shows a twinned version of an

ꢁ

FIG. 5. TEM image of a twinned nanowhisker of the W

O

phase,

ꢀꢇ ꢀꢈ

which is encapsulated by closed WS layers.

ꢁ

identical projection of the W O phase shown in Fig. 4c.

ꢀꢇ ꢀꢈ

A lower magni"cation micrograph in Fig. 5a shows a prom-

inent twin boundary (arrowed) which is also visible in the

corresponding ED pattern (Fig. 5b) as splitting of the re#ec-

tions in the noncentral rows. The indicated pair of spots in

Fig. 5b correspond to split (113) re#ections (cf. Fig. 4c, inset).

Figure 5c shows an enlargement of the microstructure of the

encapsulate, which is clearly identical to that in the micro-

graph in Fig. 4c. The observed oxide needle is encased

within two layers of the encapsulating sulphide, which is an

early stage in the growth of the WS₂nanotube. An impor-

tant new observation arising from this work is that, with

respect to IF-WS encapsulated tungsten oxides, the encap-

ꢁ

sulation process is apparently independent of the ordering

behavior of the encapsulated oxide. Previously we noted the

formation of ordered tungsten oxides within IF-WS cages.

ꢁ

The encapsulated oxide in Fig. 5 is clearly the same oxide as

the ordered precursor in Fig. 4c. This is consistent with the

observation, also made in the previous work, that dis-

ordered tungsten oxides are also observed within IF-WS

structures. This does not necessarily rule out the possibility

ꢁ

of ordering during encapsulation, which invariably involves

some heat treatment and reduction of the precursors. How-

ever the previously noted ordering phenomenon is an inci-

dental rather than essential feature of the encapsulation

process and the sul"dization reaction is a surface rather

than a bulk phenomenon. A consequence of this observa-

tion is that it ought to be possible to tailor the type of

FIG. 4. TEM images of a series of reduced intermediates observed in

the oxide needles across the composition range WO }WO . (a) WO oxide

needle in the early stages of reduction; (b) typical nonstoichiometric

ꢂ

ꢁ

ꢂ

WO

needle; (c) [110] projection of W (c-tungsten oxide); (d) [100]

O

ꢂ\V

ꢀꢇ ꢄꢈ

projection of an ordered WO crystallite.

ꢁ

306

FREY ET AL.

tungsten oxide encapsulate produced within IF-WS mater- stead of the closed ended nanotubes reported previously (3).

ꢁ

ials, although given the complexity of the WO }WO sys- In these two respects they are more closely related to the

ꢂ

ꢁ

tem, this is a far from trivial exercise.

WS nanotubes reported in (4). This entails that in contrast

ꢁ

Concerning the type of defects encountered in the tung- to the previous report (3), here the oxide nanowhisker grows

sten suboxides core, it was observed that many whiskers are "rst and crystallize (probably in the upper part of the

indeed composed of well distinguishable &&strips'' in the reactor), and are subsequently sul"dized (in the lower part

[010] direction, which is evidence of the presence of +001,

of the reactor). The partially converted nanowhiskers are

0

crystallographic shear (CS) planes along the growth direc- obtained by interrupting the reaction some 15 min after it

tion [010]. The CS defects are found to be randomly distrib- has started. TEM images of the tungsten oxide nanow-

uted inside the core of the whiskers and their density varied hiskers obtained by the #uidized bed reactor are shown in

from one whisker to another. Consequently, the Figs. 6 }8. Figure 6a shows a TEM image of a bunch of

stoichiometry of the whiskers is extremely variable (WO

with x variable), which leads to a wide mixture of non- XRD) as belonging to the HC-W O₁₄phase. Figure 6b

stoichiometric phases in the early stages of the reduction. As shows the TEM image of a tip of single nanowhisker and its

nanowhiskers, which were part of a material identi"ed (by

ꢂ\V

5

the reduction time increases, more and more +001, CS ED in the inset. The long d-spacing of 11.6 A was assigned as

0

planes are created and start to approach each other, ending the (200) re#ection of the HC-W O₁₄phase. Figure 7 shows

5

up with a fully ordered stoichiometric structure of the form: an IF-WS nanowhisker encapsulating a reduced oxide

ꢁ

W O

(4, 5) or W O

(17, 18). Alternatively, stable PC core. Figure 7a shows the image of the whole nanowhisker

L ꢂL\ꢀ

L ꢂL\ꢁ

motifs are formed along the nanowhiskers. Such ordered under a low magni"cation. In Fig. 7b, the magni"ed image

structures inside the whiskers have been routinely observed of the nanowhisker tip is shown. The pencil-shaped tip is

in the samples originated from type II precursor, already open at the end. Most remarkably, while the oxide nanow-

after 5 min annealing (Fig. 5). It is important to notice that hisker is encapsulated by two layers of WS , the number of

ꢁ

no other type of defects were apparently created in the long WS layers increases at the whisker tip to six. The image

ꢁ

tungsten oxide whiskers. This fact underscores the import- also clearly reveals the lattice image of the oxide phase in the

ance of the +001, CS planes and the PC motifs in the

0

formation of WS nanotubes from needle-like tungsten ox-

ꢁ

ide nanoparticles (1).

A second point to emphasize is that the rate of reduction

of the spherical oxide nanoparticles is fast under these

conditions, and consequently, in contrast to the case of

IF-WS nanoparticles with oxide core (5), no de"nitive

ꢁ

answer regarding the type of defects could be obtained. The

small fraction of spherical nanoparticles (type I), which are

converted into nanowhiskers contain +001, CS planes and

0

superstructures with PC, exclusively. This fact tends to

demonstrate that +001, CS planes or an array of PC and

0

HC are the only defects compatible with the growth and

evolution of the nanowhisker morphology. This is not sur-

prising in view of the large anisotropy, typical of such

crystalline structures and consequently their high mechan-

ical strength, which can support these long and slender

nanostructures made of quite a brittle material.

By examining the samples with a HRTEM, the same

general conclusions could be drawn. Unexpectedly however,

another phase could be distinguished that was not observed

by the preceding TEM studies (3, 5). Indeed, the c-W O

ꢀꢇ ꢄꢈ

phase (8), consisting of pentagonal columns (PC) and hexag-

onal channels (HC), was observed in the batches originated

from type II precursors (see Table 1a). This point is parti-

cularly noteworthy and it will be further discussed in what

follows.

Generally, the nanotubes produced by the #uidized bed

reactor are longer (up to 300 lm) than the nanotubes re-

FIG. 6. TEM micrograph of an oxide nanowhisker produced by the

ported earlier (3). Furthermore they grow open ended in- #uidized bed reactor. The HC-W₅O₁₄was identi"ed through ED.

NONSTOICHIOMETRIC TUNGSTEN OXIDE NANOPARTICLES

307

also large amount of WS nanoparticles with nested

ꢁ

fullerene-like structure.

The reduction of WO in the #uidized bed reactor is

ꢂ

accompanied by a unique morphological change, from

round or elongated small WO particles to whiskers

ꢂ

ranging from a few to about 50 lm in length. Two phases of

WO

whiskers have been identi"ed by electron di!rac-

ꢂ\V

tion. HC-W₅O₁₄(WO_2.8) that packs in a tetragonal space

group P-421m a"b"2.33, c"0.38 nm and W₁₈O₄₉

(WO_2.72) that belongs to a monoclinic space group P2/m

a"1.83, b"0.38 and c"1.4 nm. These two phases have

pentagonal columns and pseudo-hexagonal channels run-

ning along the whisker. The growth morphologies of these

whiskers can be related to their crystallographic structures

using the Bravais}Friedel}Donnay}Harker (BFDH) law as

the rule of thumb (19). According to this law the growth

rate of a crystal face is inversely proportional to the inter-

planar spacing of that face. Thus, in both WO

phases

ꢂ\V

observed, the growth direction of the whisker is the short

0.38 axis, while the larger and more developed faces are

those with the largest interplanar spacing. This is in contrast

to the precursor WO₃that in all its polymorphs has three

rather similar cell dimensions, with no preferred growth

direction.

FIG. 7. A pencil-shaped oxide nanowhisker encapsulated by closed

WS layers. The oxide is assigned to the W phase. Note that the tip of

O

ꢁ

ꢀꢇ ꢄꢈ

the nanowhisker is open, i.e., the sul"de layer is not closed at this end. Note

also that the number of encapsulating WS layers is 2 in the bulk of the

ꢁ

nanowhisker and increases to 6 at the tip.

core. The electron di!raction of the oxide core is shown in

Fig. 7c. The long spacing of 12.7 A could be assigned to the

(001) re#ection of the W₁₈O₄₉phase. In other series of

growth experiments however, the spacing of the di!raction

dots of the oxide core (16.5 A) coincided with that of the

CS-W O phase, i.e., the one belonging to the +001, CS

ꢆ ꢀꢄ

0

series, which was discussed in detail in (5). Figure 8a shows

a [010] view of a fragment of a broken W₁₈O₄₉nanowhisker.

This fragment appears to be a pure phase with no obvious

defects. The enlargement in Fig. 8b shows that this fragment

consists of an ordered array of pentagonal column (PC)

units interspersed with domains of ReO₃-type WO₃struc-

ture and hexagonal channels (HC). Figures 8c and 8d show

an image simulation and structure model corresponding to

the same projection of W₁₈O₄₉. The unit cell of this phase is

indicated by the parallelogram in Figs. 8b }8d.

It is thus found that while the unencapsulated nanow-

hiskers can adopt a number of di!erent oxide phases, the

FIG. 8. (a) Low magni"cation HRTEM micrograph showing the regu-

core of the IF-WS nanowhiskers is predominantly made of

lar microstructure of a W

ꢀꢇ ꢄꢈ

in 8a; (c) Image simulation computed for a 0.2 nm thick crystal at !54 nm

O

fragment; (b) Enlargement of boxed region

ꢁ

the anisotropic W O phase, which is typi"ed by a net-

ꢀꢇ ꢄꢈ

work of PC and HC. Notably the present product contains defocus; (d) Corresponding structural model used to calculate (c).

308

FREY ET AL.

W O #W O

(n"4, 5, 6), by the ED analysis. This

III.2. Determination of the Stoichiometry of the

ꢁꢉ ꢆꢇ

L ꢂL\ꢀ

point is particularly relevant since W O was de"nitely

Progressively Reduced WO

Particles by XRD

ꢀꢇ ꢄꢈ

ꢂ\V

detected in the same batches by HRTEM. The proportion of

this speci"c phase compared to the others W O

or

L ꢂL\ꢀ

Irrespective of their morphology, the stoichiometry of the

W O

cannot be readily estimated by HRTEM, due to

ꢀ ꢂL\ꢁ

WO

precursors, could not be easily assigned by XRD

ꢂ\V

the poor statistics of this method. The fact that the XRD

patterns of the reaction product exhibit only the W O

(3b, 5). First, most of the samples were either not su$ciently

crystalline or were of too small particle size for generating

well-de"ned XRD patterns. Moreover, several non-

stoichiometric tungsten oxide phases have been reported in

the literature and all of them exhibit very similar patterns.

Consequently, assigning the stoichiometry of the WO

ꢀꢇ ꢄꢈ

phase and not the W O

related to either a major proportion of the W O phase,

which is unlikely, or to its larger cross section for X-ray

di!raction. Therefore, the evolution of the tungsten subox-

or W O

phases could be

L ꢂL\ꢀ

L ꢂL\ꢁ

ꢀꢇ ꢄꢈ

ꢂ\V ide nanowhiskers along the reduction process does not lend

phases comprehensively from XRD data alone was a rather

itself for an exclusive formation and evolution of +hkl, CS

0

di$cult task. The measurement by electron di!raction was

more informative in this case. Individual crystallites or

groups of crystallites were examined within the TEM. The

absence of such ED pattern in the case of type I precursor

shows that many of the nanoparticles are amorphous (al-

though lattice images indicated that many of these particles

were nanocrystalline), as was clearly evidenced by the XRD

measurements. By contrast, type II nanoparticles di!ract

strongly enough, to the extent that bulk XRD patterns

planes, as was previously assumed (5). (Nevertheless, the

W O phase, which is organized around pentagonal col-

ꢀꢇ ꢄꢈ

umns originates also from the presence of +001, CS planes.

0

This point emphasizes the fact that the tungsten suboxides

phases are systematically related to +001, CS planes, which

develop along the whisker growth axis).

0

Figure 9 shows the XRD patterns of the two oxide

nanowhisker phases obtained by the #uidized bed reactor.

While Fig. 9a shows the XRD pattern of the HC-W₅O₁₄

nanowhiskers, Fig. 9b shows the XRD pattern of the

W O phase encapsulated by the WS closed layers. Note

could be obtained and therefore the values of the d spac-

ings could be calculated. The stoichiometry of individual

FIJ

ꢀꢇ ꢄꢈ

ꢁ

particles of type II precursors can be ascribed to the tetrag-

onal phase W O (WO ) as described recently by Sloan

that the contribution of the encapsulating WS layers to the

ꢁ

ꢁꢉ ꢆꢇ

ꢁꢊꢈ

XRD pattern of this sample is not appreciable. In contrast,

the HC-W₅O₁₄phase, which was not encapsulated by sul"de

layers, exhibited a very pronounced signal of WS . Unfor-

tuantely, the present product was not a pure nanowhisker

et al. (5). This W O (WO ) phase corresponds to the

ꢁꢉ ꢆꢇ

ꢁꢊꢈ

usual stacking of WO with randomly distributed defects

ꢂ

ꢁ

along the whisker axis consisting predominantly of CS-type

or related defects.

phase and contained also substantial amounts of fullerene-

The two sets of XRD measurements originating from

both types of precursors are summarized in Table 2. The

phases are globally consistent with those identi"ed in

the ED patterns, except for the most di$cult part, which is

the assignment of the stoichiometry of the long tungsten

oxide whiskers. Indeed, while the samples consisting of long

like WS nanoparticles, which contributed a component to

ꢁ

the XRD signal.

III.3. Raman Measurements

By recording the Raman spectra at di!erent stages of the

nanotube synthesis, it was possible to monitor the reduction

and sul"dization processes when converting polycrystalline

WO into WS nanotubes, as shown in Fig. 10 (3a). These

oxide whiskers were ascribed by XRD to the phase W O

the same sample was found to consist of a mixture of phases

,

ꢀꢇ ꢄꢈ

ꢂ

ꢁ

spectra show that, in principle, one should be able to follow

the transformation of the remaining tungsten oxide

TABLE 2

nanoparticles in the core of the WS nanotubes by this

technique. Unfortunately, the strong absorption and the

In6uence of the Reduction Time on the Stoichiometry

of the WO_3؊xParticles (types I and II Precursors)

ꢁ

Raman scattering by the WS envelope suppresses the

Raman signal from the oxide core at the wavelength used in

ꢁ

Type I precursor

(spherical-faceted particles)

Type II precursor

(needle-like particles)

these experiments (632.8 nm). The Raman spectrum of WS

ꢁ

nanotubes was compared to that of bulk 2H-WS material

ꢁ

Reduction time

before (20). In the next sections, the Raman spectra of oxide

(min)

XRD

(min)

XRD

nanowhiskers, which were prepared by similar techniques, is

reported.

2

5

10

15

60

W

O

1

5

10

20

60

W

O

ꢀꢇ ꢄꢈ

WO

ꢀꢇ ꢄꢈ

O

#W

ꢁ

ꢀꢇ ꢄꢈ

O

WO

WO #W

III.3.a. Evolution of the Raman spectra of type I isotropic

ꢁ

ꢀꢇ ꢄꢈ

ꢁ

WO #W

W

ꢁ

faceted WO

particles upon reduction. While WO has

ꢂ\V

ꢂ

ꢁ

been extensively studied by this technique (21 }24), no report

NONSTOICHIOMETRIC TUNGSTEN OXIDE NANOPARTICLES

309

FIG. 9. XRD patterns of two nanowhisker phases obtained by the #udizied bed reactor: (a) HC-W₅O₁₄phase which contained also substantial

amounts of fullerene-like WS nanoparticles; (b) Encapsulated W

phase.

O

phase encased by a few closed WS layers. The peak at 14.33 belongs to the WS₂

ꢁ

ꢀꢇ ꢄꢈ

ꢁ

310

FREY ET AL.

FIG. 11. Raman spectra of type I (spherical) nanoparticles under

reduction after di!erent annealing times (in min): the starting precursor is

WO and is designated as 0 (min). With longer annealing times (indicated

ꢁꢊꢈ

for each curve in min) the Raman spectrum of the nanoparticles resemble

more and more that of WO nanoparticles.

ꢁ

(Ref. (3a)). The Raman spectra for the further reduced sam-

ples, 10, 20 min (not shown) and 60 min are similar to that

obtained after 5 min and correspond to WO .

ꢁ

III.3.b. Evolution of the Raman spectra of type II WO

ꢂ\V

nanowhiskers upon reduction. Figure 12 shows the Raman

spectra of the needle-like WO

particles reduced at 8403C

ꢂ\V

under the #ow of a 1% H /99% N gas for 1, 5, 10, 20, and

ꢁ

60 min. After 1 min of reduction the needle-like precursor is

still amorphous resulting in a featureless Raman spectrum.

In the Raman spectrum of the needle-like material reduced

for 5 min several new Raman modes are observed in

FIG. 10. Raman spectra of (a) WO , (b) WO nanoparticles, (c) WO ,

ꢂ

ꢁꢊꢈ

ꢁ

and (d) WS nanotubes.

ꢁ

\ꢀ

the 200}400 and 800}900 cm

regions, above an intense

was apparently published on the Raman of the tungsten

suboxide (i.e., WO ) phases; a fact, which is probably

ꢂ\V

related to the di$culty of isolating pure phases of tungsten

suboxides and therefore in attributing the observed features

to the composite structural components. However, with the

development of commercial Raman microscopes with a res-

olution of below 1 lm, the recording of a Raman spectrum

of a single nanowhisker has become feasible.

Figure 10 shows the Raman spectra of the isotropic

faceted WO

nanoparticles reduced at 8403C under the

ꢂ\V

#ow of a 1% H /99% N for 0, 2, 5, and 60 min. The Raman

ꢁ

spectra of type I (spherical nanoparticles) precursor.

Figure 11, shows an intense background indicating that the

material is amorphous. Upon slight reduction (2 min),

\ꢀ

a broad feature centered at &315 cm is observed above

the background. After 5 min of reduction the Raman spec-

trum is intense and detailed with Raman lines at: 168(w),

189(w), 286(vs), 345(w), 423(w), 479(m), 512(m), 599(m),

FIG. 12. Raman spectra of type II (whisker) nanoparticles under reduc-

tion. The numbers on each curve represent the annealing time in forming-

gas atmosphere. After 60 min annealing time the spectrum resembles that

\ꢀ

617(m) cm , and a mode at 781(s) cm

\ꢀ

which tails to

higher energies (w-weak; m-medium; s-strong; vs-very

strong). This spectrum has been recently attributed to WO

of WO .

ꢁ

NONSTOICHIOMETRIC TUNGSTEN OXIDE NANOPARTICLES

311

background. In the high-frequency region, two broad bands WS nanoparticles, which were in close proximity to the

ꢁ

\ꢀ

appear centered at 825 and 870 cm . In the lower fre- speci"c nanowhisker, they also contributed to the Raman

quency region of the spectrum, a broad band appears be- scattering and their spectrum was superimposed on the

\ꢀ

tween 200 and 400 cm with features at 240, 290, 315, 335, spectrum of the oxide nanowhisker. Unfortunately, the

\ꢀ

375, and 418 cm . Although the XRD patterns obtained Raman spectrum of the W O nanowhisker encapsulated

ꢀꢇ ꢄꢈ

from this needle-like oxide sample indicated that the par- with a few WS layers coincided with that of WS

ticles are amorphous (25), the ED patterns and HRTEM nanotubes (3a). In this case the scattering intensity of the

ꢁ

images revealed that this phase has a highly distorted encapsulating WS phase is rather strong and suppressed

ꢁ

layered structure (26). After 10 min of reduction, the Raman the Raman spectra from the oxide cores.

spectrum does not change relative to that obtained after

5 min, but the features become more clearly pronounced.

III.3.c. Discussion of the Raman spectra of the reduced

Reducing the needle-like precursor for 20 min results in WO

a low background and several new modes in the Raman state Raman scattering is sensitive to both metal}oxygen

spectrum appear, as follows: (i) three bands in the high bond length, coordination number, and symmetry around

particles (type I and type II particles). Since solid

ꢂ\V

\ꢀ

frequency region at 782, 830, and 870 cm , (ii) several a metal center, it is therefore a useful technique in studying

bands in the low frequency region at 186(w), 210(w), 237 (m), phase transformation and evolution within these types of

\ꢀ

283(vs), 318(m), 340(m) cm , and (iii) several new bands in compounds.

the central region of the spectrum at: 425(m), 511(w), 600(w), Recently, the Raman spectra of transition metal oxides

\ꢀ

620(w), and 650(w) cm . The Raman spectrum of the fully has been interpreted using the diatomic approximation

reduced type II precursor (60 min) is similar to that ob- which regards each metal}oxygen bond in the material as

served for the fully reduced type I precursor (Fig. 10, 60 min vibrationally independent of its surroundings (25). Conse-

annealed) and is assigned to WO . Similar features, i.e., quently, in tungsten oxides constructed from WO octahed-

ꢁ

ꢅ

\ꢀ

changes in relative peak intensities, shifts in band positions ral repeat units, all Raman bands around 400 cm

are

and new modes, were observed when following the reduc- assigned to W}O stretching modes and are assumed to

tion process of MoO by in situ Raman spectroscopy (27). directly re#ect the W}O bond lengths. An empirical expres-

ꢂ

Figure 13 shows the Raman spectra of a nanowhisker, sion for relating the W}O bond length to the frequency of

which was produced by the #uidized bed reactor and the Raman modes, which enables a direct determination of

according to XRD belongs to the phase HC-W O

ꢆ ꢀꢄ

Raman peaks are observed at 215, 264, 325, 349, 418, 425, (25). Furthermore, the energy of the highest observed

.

the W}O bond length from the Raman spectra, is derived

\ꢀ

707, and 800 cm . In other nanowhiskers of this kind Raman band associated with the stretching of the shortest

\ꢀ

a series of peaks at 764, 854, and 900 cm were recorded as W}O bond re#ects the regularity of the structure (25, 26);

well. Also, because this phase contained also fullerene-like i.e., for the di!erent octahedrally coordinated WO phases,

V

FIG. 13. Raman spectrum of a CS-W₅O₁₄nanowhisker #anked by WS nested fullerene-like nanoparticles.

ꢁ

312

FREY ET AL.

a higher Raman stretching mode indicates a more distorted the intermediate nonstoichiometric phases do not produce

structure. By comparing the Raman spectra of several tung- distinct Raman spectra indicate that the reduction process

sten oxide compounds with an octahedral coordination, does not advance through a set of well de"ned phase trans-

Horsley et al. (25) showed that the distorted octahedra formations.

\ꢀ

possess Raman bands between 740 and 980 cm

.

The reduction of the needle-like precursor, on the other

hand, is slower and proceeds through nonstoichiometric

The room temperature crystallographic structure of WO

is derived from the ReO structure with the corner-shared phases with detailed Raman spectra (Fig. 12). The Raman

octahedra tilted, and the tungsten atoms distorted from spectrum of the reduced precursor contains features similar

ꢂ

their symmetrically located positions within the oxygen to that of WO , namely, two broad bands in the

ꢂ

\ꢀ

octahedra, to form zig-zag chains with alternating short and 800}900 cm

region and several bands in the

\ꢀ

modes are

long W}O bond distances. The crystalline WO Raman 200}400 cm

region. The 825 and 870 cm

ꢂ

spectrum (Fig. 5a) exhibits major Raman bands at 808, 719, assigned to the W}O}W stretching modes which are shifted

\ꢀ

W}O}W stretching frequencies and the 275 cm

a W}O}W bending mode (25). The shorter W}O}W bonds edge-sharing octahedra. Using the empirical formula de-

\ꢀ \ꢀ

and 275 cm , where the 808 and 719 cm are assigned as to higher energies compared to WO . This shift is the

ꢂ

\ꢀ

as outcome of the shorter W}O}W bonds introduced by the

are responsible for the stretching mode at 808 cm , while scribed in (24), it is possible to associate the 870 cm mode

\ꢀ

longer bonds, are the source of the 719 cm peak. Relative- with the stretching mode of a 1.78 A W}O bond. This value

\ꢀ

ly strong peaks are also observed around 300 cm

\ꢀ

and compares favorably with the calculated distance of 1.77 A

below 150 cm

O}O deformation modes, respectively.

WO has a distorted rutile structure formed by edge region (200}400 cm ), compared to only one in the WO

attributable to W}O deformation and for one of the W}O bond lengths in the CS W O structure

ꢂ ꢇ

(31). The abundance of many bands in the W}O}W bending

\ꢀ

ꢁ

ꢂ

sharing WO units and metal-metal bonds (28, 29). The parent material, indicates that there are several distinct

most intense peak in the Raman spectra of WO (Figs. 10, W}O}W bond lengths. The complex structure of the PC

11d, and Fig. 12e) at 285 cm is assigned to the W}O}W phases, mainly, W O , contain a wide range of W}O}W

ꢅ

ꢁ

\ꢀ

ꢀꢇ ꢄꢈ

\ꢀ

bending mode, which appears at 275 cm for WO (Fig. bond lengths which are expected to result in a very broad,

ꢂ

10). The shift of this band to a higher frequency is attributed featureless band. The CS structures, however, contain sev-

to the constrained W}O}W bending in the more compact eral possible members of the W O

homologous series

L ꢂL\ꢀ

distorted-rutile structure of WO (30). Only one stretching with di!erent CS distance which could result in speci"c

ꢁ

\ꢀ

mode is observed for WO at 782 cm . This mode exhibits, bending modes. Hence, the Raman spectra of the partly

ꢁ

however, a strong asymmetric tailing to higher energies. reduced needle-like particles could be a "nger-print of the

Hence, the WO octahedra forming the rutile structure are CS structures, although, the contribution of the PC phases,

ꢅ

highly distorted and contain a large variety of W}O}W mainly in W O and HC-W O , could not be ruled out.

ꢀꢇ ꢄꢈ

ꢆ ꢀꢄ

bond lengths.

Upon further reduction of the type II precursor, WO is

ꢁ

Upon reduction of WO , oxygen vacancies are formed formed. Hence, the Raman spectrum of the type II precursor

which are compensated by either crystal shear planes (CS) reduced for 20 min coincides with the Raman spectrum of

ꢂ

or pentagonal columns structures (PC). In the CS structure the CS structure and that of WO (Fig. 12).

ꢁ

+102, , +103, crystal-shear planes, along with +001, CS

The results described previously re#ect the fact that along

0

planes of the W O

series were observed [5]. The CS the reduction process, type I and type II precursors undergo

planes are formed by the edge-sharing of neighboring oc- a distinguishable evolution since the Raman spectra di!er

tahedra. Two adjacent CS planes are separated by a com- after a few minutes reduction. In the case of type II precur-

L ꢂL\ꢀ

paratively undistorted WO sublattice. In the PC structures sors, for which the reduction process leads in a "rst step to

ꢂ

(HC-W O , W O , and W O ), pentagonal columns the formation of multiple W O

phases in a single batch,

ꢆ ꢀꢄ

ꢀꢁ ꢂꢄ

ꢀꢇ ꢄꢈ

L ꢂL\ꢀ

couple in pairs by edge-sharing octahedra to form &&tunnel it has been suggested that each of these phases contribute to

compounds.'' Wide, empty pseudo-hexagonal tunnels are new set of Raman peaks, which are related to W}O bond

formed which run through the structure. While WO has lengths di!erent from those in the WO precursor. How-

ꢂ

ꢁꢊꢈ

been extensively studied by Raman spectroscopy, the tung- ever, it is not easy to attribute each of these peaks to

sten suboxides, in general, and more particularly, the homo- a certain W}O bond length. Computer simulation of each of

logous series with the formula W O

with Raman spectroscopy, to the best of our knowledge,

probably because it is di$cult to isolate any pure phase of peaks in the W}O}W stretch region at 707, 800, and

, were not studied the W O

phases might be very useful in this case.

L ꢂL\ꢀ

The Raman spectrum of the HC-W O₁₄phase shows three

5

\ꢀ

900 cm . The position of the "rst two peaks (707 and

the suboxides (5).

\ꢀ

From the evolution of the Raman spectrum of precursor 800 cm ) is similar to the position of the W}O}W stretch

I, upon reduction, it is evident that the isotropic faceted peaks observed for the WO starting material. However, the

ꢂ

WO

particles are rapidly reduced to WO . The fact that peaks are broader and the relative intensity changes. While

ꢂ\V

ꢁ

NONSTOICHIOMETRIC TUNGSTEN OXIDE NANOPARTICLES

313

\ꢀ

the 800 cm peak dominates the WO Raman spectrum, in planes, only. Since PC#HC phases emanate from suboxide

ꢂ

\ꢀ

the Raman spectrum of HC-W O₁₄, the 800 cm peak is phases containing +001, CS planes, these kind of CS planes

5

0

\ꢀ

very broad and has lower intensity than the 707 cm peak. are likely to be present in the beginning of the reduction

\ꢀ

The decrease of the 800 cm peak could be correlated with process of the oxide nanowhiskers, irrespective of their

\ꢀ

the appearance of the new peak at 900 cm . Using the future evolution. However, nothing can be put forward

\ꢀ

empirical formula in (17), the 900 cm mode is associated concerning the type I particles, since they change their

with a 1.76 A W}O bond. As mentioned earlier, a higher morphology into nanowhiskers and subsequently into cry-

Raman stretching mode wavenumber indicates a more dis- stalline nanoparticles in the early stages of the reduction

torted structure. Hence, the reduction of the oxide whiskers process.

e!ected mainly the shorter W}O bonds. The elimination of

Moreover, it has been shown up that di!erent tungsten

oxygen causes the remaining short W}O bonds to shrink suboxide phases belonging to the homologous series

even further, while the longer W}O bonds are only slightly W O

altered. This structural change detected by the Raman spec- stages in the "rst step of the reaction (ED). This has been

trum is associated with the formation and propagation of con"rmed in one speci"c example by Raman spectroscopy,

(with n"3, 4, 5, 6) are formed as intermediate

L ꢂL\ꢀ

the CS planes. The longer W}O bonds re#ect the WO

sublattice situated in between adjacent CS planes, while the a W O structure has been calculated by the method of

shorter W}O bonds are the result of CS planes forming Hardcastle and Wachs (25).

for which a W}O"1.77 A distance, characteristic of

ꢂ

ꢂ ꢇ

upon reduction. In the further reduced phase, CS-W O , the

distance between two adjacent CS is small and causes the reactor tend to be long and crystalline, which explain the

The oxide nanowhiskers which form in the #uidized bed

ꢂ ꢇ

distortion of all W}O bonds. The distorted structure results open tip of the WS nanotubes obtained after sul"dization

ꢁ

in broadening and shifting of the two main W}O stretch of the nanowhiskers. Most of the encapsulated nanow-

\ꢀ

modes to 825 and 870 cm as shown in Fig. 11.

\ꢀ

The absence of Raman bands around 950 cm for all phase with a network of PC and HC features.

measured samples, indicates that, as expected, the heated

hiskers obtained in this reaction contain the c-W O

ꢀꢇ ꢄꢈ

WO

powders do not contain any hydrated phases. The

ꢂ\V

proposed growth mechanism of the needle-like oxides

suggests that presence of water molecules, which are

formed during the reduction process, enhances the growth

of the nanowhiskers (32). Volatile hydrated clusters

ACKNOWLEDGMENTS

We are grateful to Alisa Band and Alex Margolin for their technical help

with the Raman and XRD measurements, respectively. This work was

supported by the ACS-PRF (USA); Krupp Foundation, and France-Israel

R & D Foudation (AFIRST). J.S. is indebted to the Royal Society.

(WO

) nH O) are formed and subsequently condense on

ꢂ\V

ꢁ

the nanowhisker tip. It is evident from the Raman spectra,

however, that the hydrated clusters are reduced upon con-

densation on the tip and the suboxide nanowhiskers are not

hydrated.

REFERENCES

1. (a) Y. Feldman, G. L. Frey, M. Homyonfer, V. Lyakhovitskaya,

L. Margulis, H. Cohen, G. Hodes, J.L. Hutchison, and R. Tenne, J.

Am. Chem. Soc. 118, 5362 (1996); (b) Y. Feldman, V. Lyakhovitskaya,

and R. Tenne, J. Am. Chem. Soc. 120, 4176 (1998).

2. (a) Y. Feldman, E. Wasserman, D. J. Srolovitz, and R. Tenne, Science

267, 222 (1995); (b) A. Zak, Y. Feldman, V. Alperovich, R. Rosentsveig,

and R. Tenne, J. Am. Chem. Soc. 122, 11,108 (2000).

3. (a) A. Rothschild, G. L. Frey, M. Homyonfer, M. Rappaport, and

R. Tenne, Mat Res. Innovat. 3, 145 (1999); (b) A. Rothschild, J. Sloan,

and R. Tenne, J. Am. Chem. Soc. 122, 5169 (2000).

4. Y. Q. Zhu, W. K. Hsu, N. Grobert, B. H. Chang, M. Terrones,

H. Terrones, H. W. Kroto, and D. R. M. Walton, Chem. Mater. 12,

1190 (2000).

5. J. Sloan, J. L. Hutchison, R. Tenne, Y. Feldman, T. Tsirlina, and

M. Homyonfer, J. Solid State Chem. 144, 100 (1999).

6. T. Miyano, M. Iwanishi, C. Kaito, and M. Shiojiri, Jpn. J. Appl.

Phys. 22, 863 (1983) and references therein.

7. E. Iguchi, J. Solid State Chem. 23, 231 (1978).

8. A. Magneli, Ark. Kemi. 1, 223 (1950).

IV. CONCLUSIONS

Studies of three types of WO

nanoparticles (type

ꢂ\V

I*spherical-faceted and type II*needle-like and commer-

cially available WO₃nanoparticles), by a combined ap-

proach (HR)TEM, XRD, and Raman spectroscopy, has

allowed both the morphological evolution of the particles as

well as their stoichiometry during the reduction by hydro-

gen to be followed.

First, it was demonstrated that, irrespective of the starting

precursor, WO

nanowhiskers are produced during the

reduction process. In particular, type I nanoparticles do not

sustain their morphology under hydrogen #ow.

Second, the needle-like tungsten suboxide phases

(type II), and the oxide nanowhiskers produced by the

#uidized bed reactor were also found to include the c-

W₁₈O₄₉and W₅O₁₄phases, which contain an ordered array

of PC and HC features. On the other hand, the phases

ꢂ\V

9. M. Lundberg, Chem. Common. ;niv. Stockholm 12 (1971).

10. W. Sahle, J. Solid State Chem. 45, 324 (1982).

11. I. J. McColm, R. Steadman, and S. J. Wilson, J. Solid State Chem.

23, 33 (1978).

W_nO

and W_nO

, consist of ordered +001,_RCS

ꢂL\ꢀ

ꢂL\ꢁ

314

FREY ET AL.

12. Y. Feldman, A. Zak, R. Popovitz-Biro, and R. Tenne, Solid State

Sci. 2, 663 (2000).

23. S. Hayashi, H. Sugano, H. Arai, and K. Yamamoto, J. Phys. Soc. Jpn.

61, 916 (1992).

13. R. Rosentsveig, A. Margolin, Y. Feldman, R. Popovitz-Biro, and R. 24. M. Scheithauer, R. K. Grasseli, and H. Knozinger, ¸angmuir 14, 3021

Tenne, submitted. (1998).

14. R. C. Weast (Ed.), &&Handbook of Chemistry and Physics,'' 69th ed., 25. F. D. Hardcastle and I. E. Wachs, J. Raman Spectrosc. 26, 397 (1995).

P. B-141. CRC Press, Boca Raton, FL, 1988.

15. P. Gado, Acta Phys. Hung. 18, 111 (1965).

26. J. A. Horsley, I. E. Wachs, J. M. Brown, G. H. Via, and F. D.

Hardcastle, J. Phys. Chem. 91, 4014 (1987).

16. S. Iijima, J. Solid State Chem. 14, 52 (1975).

17. A. Magneli, Acta Chem. Scand. 1, 501 (1948).

27. G. Mestl, P. Ruiz, B. Delmon, and H. Knozinger, J. Phys. Chem. 98,

11,269 (1994).

18. L. A. Bursill and B. G. Hyde, J. Solid State Chem. 4, 430 (1972).

19. (a) J. D. Donnay and D. Harker, Amer. Miner. 22, 446 (1937);

28. A. F. Wells, in &&Structural Inorganic Chemistry,'' 5th ed., p. 541.

Oxford Scienti"c Publications, Clarendon Press, Oxford, 1984.

(b) A. S. Myerson, in &&Molecular Modeling Applications in Crystal- 29. F. A. Cotton and G. Wilkinson, in &&Advanced Inorganic Chemistry,''

lization'' (A. S. Myerson, Ed.). Cambridge Univ. Press, Cambridge,

1999.

20. G. L. Frey, R. Tenne, M. J. Matthews, M. S. Dresselhaus, and G.

Dresselhaus, J. Mater. Res. 13, 2412 (1998).

5th ed. p. 808. Wiley Interscience, New York.

30. F. H. Jones, R. G. Egdell, A. Brown, and F. R. Wondre, Surf. Sci.

374, 80 (1997).

31. The W}O bond lenghts were calculated using the program &&Crystal-

Maker'' (1994}2001 David Palmer).

21. E. Salje, Acta Crystallogr. Sect. A 31, 360 (1975).

22. M. Arai, S. Hayashi, K. Yamamoto, and S. S. Kim, Solid State Comm. 32. (a) T. Millner and J. Neugebauer, Nature 163, 602 (1949); (b) V. K.

75, 613 (1990). Sarin, J. Mater. Sci. 10, 593 (1975).

Article Doi

Investigations of nonstoichiometric tungsten oxide nanoparticles

DOI: 10.1006/jssc.2001.9319

Source and publish data:

Authors:

Article abstract of DOI:10.1006/jssc.2001.9319

Full text of DOI:10.1006/jssc.2001.9319

Products guided by the article

R&D Labs maybe for 7440-33-7

Relevant to this article

Hot Product