Size-distribution and emission spectroscopy of W nanoparticles generated by laser-assisted CVD for different WF6/H2/Ar mixtures

Article abstract of DOI:10.1021/jp0343077

Tungsten nanoparticles were produced by ArF excimer laser-assisted chemical vapor deposition from WF₆/H₂/Ar gas mixtures. Subsequent pulses excited the gas-phase particles, allowing optical emission spectroscopy to monitor the intensity of the emitted radiation and temperature of the laser-heated particles. A systematic study on size-distributions of the deposited particles, determined by electron microscopy, in connection with emission spectroscopy and rate of deposition measurements is presented, with respect to different partial pressures of the reactants. The rates of deposition of W nanoparticle films were determined by X-ray fluorescence spectroscopy. In addition, the intensity of the scattered 193 nm laser line was also monitored as the partial pressure of H₂ was varied.

Full text of DOI:10.1021/jp0343077

J. Phys. Chem. B 2003, 107, 11615-11621
11615
Size-Distribution and Emission Spectroscopy of W Nanoparticles Generated by
Laser-Assisted CVD for Different WF6/H2/Ar Mixtures
†,
‡
§,#
Lars Landstr o1 m, * Jun Lu, and Peter Heszler
The Ångstr o¨ m Laboratory, Department of Materials Chemistry, Uppsala UniVersity, Box 538,
S-751 21 Uppsala, Sweden, The Ångstr o¨ m Laboratory, Department of Analytical Materials Physics,
Uppsala UniVersity, Box 534, S-751 21 Uppsala, Sweden, and The Ångstr o¨ m Laboratory,
Department of Solid State Physics, Uppsala UniVersity, Box 534, S-751 21 Uppsala, Sweden
ReceiVed: February 5, 2003; In Final Form: August 11, 2003
Tungsten nanoparticles were produced by ArF excimer laser-assisted chemical vapor deposition from
WF /H /Ar gas mixtures. Subsequent pulses excited the gas-phase particles, allowing optical emission
6
2
spectroscopy to monitor the intensity of the emitted radiation and temperature of the laser-heated particles.
A systematic study on size-distributions of the deposited particles, determined by electron microscopy, in
connection with emission spectroscopy and rate of deposition measurements is presented, with respect to
different partial pressures of the reactants. The rates of deposition of W nanoparticle films were determined
by X-ray fluorescence spectroscopy. In addition, the intensity of the scattered 193 nm laser line was also
2
monitored as the partial pressure of H was varied.
Introduction
troscopy (OES) measurements on the laser induced light
emission, which was identified as thermal (blackbody-like)
radiation.
Nanoparticles and nanocrystalline materials often exhibit
unique properties well differing from bulk, e.g., size-dependent
melting point and optical characteristics, e.g., absorption and
scattering cross-sections for single particles. Furthermore,
collective properties such as increased hardness, super-plasticity,
and improved magnetic properties are also observed in nano-
1
0
The unique properties of nanomaterials have a significant
sometimes critical) size-dependence on the nm sized building
(
elements. Thus, the understanding of the formation processes
is of great importance in order to control the particle size and
achieve narrow size-distributions. In our earlier papers on W
nanoparticle formation during LCVD using WF6 precursor, the
optical spectroscopy studies were not coupled to size-distribution
1
,2,3
materials.
Numerous methods exist for nanoparticle generation, e.g.,
4
5
6
mechanical (ball-milling), template based, sol-gel, or gas-
phase processes (e.g., gas condensation, chemical vapor dep-
osition (CVD), etc.), where the particles are condensed from
9
,10,11
and rate of deposition measurements.
Therefore, in this
7
report, results on size-distribution and deposition rate measure-
ments are presented and related to optical spectroscopy of the
light emission originating from the nanoparticles for different
mixtures of the WF6/H2/Ar precursors, i.e., varying both the
WF6 and H2 partial pressures. It is shown that by applying
simple OES measurements the optimal precursor gas composi-
tion for achieving narrow particle size-distribution can easily
be monitored.
an oversaturated vapor. The use of a laser to assist chemical
vapor deposition (LCVD) during nanoparticle production in-
troduces an extra well-controlled parameter in terms of energy
and timing of the energy deposition if pulsed laser-sources are
used.
LCVD of tungsten thin films from WF6/H2 gas mixtures is
an established metalization process, where the photolysis step
either lowers the deposition temperature or increases the
In addition, inelastic quenching effects between the nano-
particles and ambient gas were also examined based on OES.
8
deposition rate. However, at certain pressures and gas composi-
tions, tungsten nanoparticles are formed, and it is worth noting
9
Experimental Section
that no particles could be detected without H2, stressing the
importance of CVD processes for nanoparticle formation. Par-
ticle formation during film deposition is an undesirable effect,
but on the other hand it can be utilized for nanoparticle and/or
nanomaterial fabrication. Because the particles are formed by
homogeneous nucleation in the gas-phase, subsequent laser
pulses heat up the particles that allows optical emission spec-
The experimental setup consisted of a stainless steel vacuum
chamber with quartz windows at the front (laser beam in)
and on top of the reactor, respectively. The flow-through re-
2
actor had a cross-section of ∼12 cm , and the laser induced
light emission was observed perpendicularly to the laser beam
through the top window of the chamber. The laser beam was
slightly focused by a 38 cm focal length cylindrical lens. The
place of optical observation was situated 9 cm downstream with
respect to the WF6/H2 inlet. An optical detection system
consisting of a Czerny-Turner type grating spectrograph, a
charge coupled device (CCD) detector, and an optical multi-
channel analyzer (OMA III of EG&G) was used. A 150/mm
grating allowed 1.8 nm spectral resolution and (0.2 nm
wavelength accuracy.
*
To whom correspondence should be addressed. E-mail: Lars.
Landstrom@mkem.uu.se. Tel: +46 (0)18 471 3769. Fax: +46 (0)18 513
5
48.
†
Department of Materials Chemistry.
Department of Analytical Materials Physics.
Department of Solid State Physics.
‡
§
#
Also at Research Group on Laser Physics of the Hungarian Academy
of Sciences, Zzeged, Box 406, H-6721 Hungary.
1
0.1021/jp0343077 CCC: $25.00 © 2003 American Chemical Society
Published on Web 09/30/2003

11616 J. Phys. Chem. B, Vol. 107, No. 42, 2003
Landstr o¨ m et al.
Figure 1. Typical corrected OES spectra from the laser-heated W
nanoparticles. Delay 500 ns and gate pulse 1 µs. Smooth solid line
1
0
represents best fit by corresponding Planck curve.
The gateable CCD detector was set to a 500 ns delay with
respect to the laser pulse, and a 1 µs gate pulse with overall
exposure time of 1 s was used during OES measurements.
Typical corrected OES spectra, with the best fit Planck curves
and corresponding temperatures, are shown in Figure 1. The
2
Figure 2. TEM BF picture of deposited particles at 120 mJ/cm laser
2 6
fluence, 165 Pa H and 115 Pa WF partial pressures, respectively,
with electron diffraction pattern as inset.
5
00 ns delay was used to avoid uncertainties for the fitting
Results
procedure on the thermal radiation, because at shorter delays
the high temperature of the nanoparticles resulted in evaporation
of excited W atoms, and the de-excitation of these could be
observed as superimposition of W elemental lines on the thermal
radiation.12
The relative transfer function of the OMA was determined
by using a tungsten-strip calibration lamp, that of the emissivity
variation over the visible region was less than 0.5%. Measured
spectra were corrected by the relative transfer function and, to
determine the temperature of the nanoparticles, fitted by
corresponding Planck curve, where the emissivity of tungsten
Tungsten nanoparticles were formed by photolytic LCVD of
different WF6/H2/Ar gas mixtures. A TEM bright field (BF)
picture of the particles is shown in Figure 2. As can be seen,
the particles have a regular round shape and are well distin-
guishable even though some agglomeration has occurred. The
electron diffraction pattern originating from the particles (inset
in Figure 2) corresponds well to â-W, a metastable phase that
is assumed to be impurity stabilized.
13,14
Energy dispersive X-ray
spectroscopy showed low concentrations of contaminants such
as fluorine and oxygen, which supports the assumption of
impurity stabilization. (X-ray photoelectron spectroscopy on
deposited nanoparticle films also revealed O and F contamina-
tion, but the amount incorporated in the nanoparticles could not
be determined with sufficient accuracy because of, e.g., post-
oxidation of deposited particles and adsorbed tungsten subfluo-
-
1.43
nanoparticles was found proportional to λ
in the observed
spectral region (400-700 nm),10 where λ is the wavelength of
the emitted radiation. Because a gate pulse of 1 µs was used,
an average temperature during this interval was observed. The
integrated intensity was obtained by summation of the thermal
emission in the monitored wavelength-window.
15
rides. )
The measurements of intensity of the scattered 193 nm laser
line were performed at 90°, relative to the excitation, with an
acquisition time of 1 s.
In parts a and b of Figure 3, the intensity of thermal emission
and particle temperature dependence on the H2 partial pressure
(
pH ) are shown, respectively. In Figure 3a, it can be seen that
2
A constant total pressure of 2000 Pa was used while varying
the Ar, H2, and WF6 partial pressures (Ar was introduced as
purge-gas on the front window to prevent deposition on it). The
H2 partial pressure was varied between 0 and 1000 Pa with a
constant WF6 partial pressure of 30 Pa. WF6 was varied between
the intensity of the emission increases rapidly until pH2 ∼ 50
Pa. Then the intensity is almost constant for some tens of Pa
after which the intensity decreases with increasing p . Normal-
H2
ized intensity, calculated using Stefan-Boltzmann’s radiation
law and temperatures from Figure 3b, is also shown in Figure
0
and 170 Pa with constant H2 partial pressure at 165 Pa. The
3a (×).
linear gas velocity was held constant at 2 cm/s during all
experiments; that is, the H2 + Ar partial pressure was constant
when H2 was varied, and WF6 + Ar was constant as WF6 varied.
Laser parameters were as follows; wavelength 193 nm
The temperature is approximately constant (∼3400 K at 500
ns delay and 1 µs gate pulse) above the threshold partial pressure
(where a sufficient number of particles are formed to allow
measurements with reasonable statistics) up to ∼100 Pa, see
(Lambda Physik ArF excimer laser), nominal pulse duration
Figure 3b. Above this p the temperature decreases linearly
H2
1
5 ns (fwhm), and repetition rate 50 Hz. The laser fluence at
(∼ -0.5 K/Pa) with increasing p .
H2
the place of optical observation and deposition was set to 120
mJ/cm during all experiments.
Size-distribution histograms for five different p ’s are shown
H2
2
in Figure 4. It can be seen that the geometric mean diameter is
Rates of deposition were measured for a constant deposition
time (10 min) by X-ray fluorescence spectroscopy (XRF) by
∼12 nm for all examined partial pressures, and only a slight
increase in mean diameter of the particles is observed, see also
integrating the WL peak for different partial pressures of the
Figure 5a. log-probability plots for the first two H partial
R
2
precursors WF6 and H2. In addition, the intensity of the scattered
laser line was also monitored as the H2 partial pressure was
varied.
pressures (Figure 6, top panels) show a deviation from a log-
normal size-distribution, and as the H2 partial pressure is
increased, this deviation remains (see Figure 4). The dependence
For determination of the size distributions, particles were
deposited on carbon covered Cu-grids for transmission electron
microscopy (TEM) analysis.
of the rate of deposition of tungsten on pH is shown in Figure
5b. A steep increase of the XRF signal can be observed up to
∼60 Pa, after which the rate of deposition is saturated and
2

W Nanoparticles Generated by Laser-Assisted CVD
J. Phys. Chem. B, Vol. 107, No. 42, 2003 11617
negative y axis. Again, this negative intercept suggests that a
threshold value in pWF6 has to be reached for nanoparticle
formation.
Discussion
Relevant Processes. To explain the experimental results,
several chemical and physical processes have to be considered,
and the most relevant are listed and discussed below.
The net chemical reaction for tungsten formation from WF6
and H2 precursors is
WF + 3H f W + 6HF
(1)
6
2
The dominant reaction pathway for particle formation during
LCVD of W contains several steps, and most likely (i) the
photolytic generation of sub-fluorides (mostly WF3 and WF4),
(
(
ii) formation of W-W bonds between the sub-fluorides, and
iii) further reduction by H2 are the most important.17 After the
particles are formed, they travel downstream by the gas-flow
and are heated by following laser pulses (because the particles
are still within the reaction zone). The temperature of these laser-
heated particles, shortly after the laser pulse, will be independent
of nanoparticle size in the first approximation, since the specific
3
heat of one particle is proportional to r (volume) and the
Figure 3. Integrated intensity of the thermal radiation (a) and
temperature (b) of the particles (at 500 ns after the laser pulse and
using 1 µs gate pulse) vs H partial pressure. WF partial pressure 30
2 6
Pa with a total pressure of 2000 Pa. Data marked by × in Figure 3a
were calculated by using Stefan-Boltzmann’s radiation law and the
measured temperatures in Figure 3b.
3
absorption cross-section of a particle also varies with r for
spherical particles much smaller than the incident wavelength.18
Considering that certain cooling mechanisms, e.g., evaporation,
radiative and heat transfer by ambient, exhibit size-dependent
3
rates of cooling which are not proportional to r , the rate of
decrease of temperature is somewhat size-dependent. However,
this effect is not taken into consideration in this work.
Thermal CVD (TCVD) processes on the surface of the
nanoparticles may also be significant for the particle growth in
approximately constant up to 500 Pa (it is noted that for LCVD
of thin films with comparable experimental parameters, the
1
6
deposition rate of tungsten was found to be ∼ 10 nm/min ).
Above this value a slight increase of the integrated intensity
10
a certain temperature range. (The deposition rate expression
(
WL ) can be observed. The intensity of the scattered laser line,
1/2
0
WF6
19
R
for TCVD is proportional to pH2 and p .) However, the
see Figure 5c, follows approximately the same behavior as the
rate of deposition depicted in Figure 5b.
particles reach high temperatures (see Figures 3b and 7b), and
taking into account the delay time and temperature decreasing
Intensity of thermal emission and dependence of the W
12
rate, one can deduce that the initial temperature of the particles
nanoparticle temperature on WF6 partial pressure (pWF ) are
6
is very close to the melting point of tungsten (∼3695 K) shortly
depicted in Figure 7, parts a and b, respectively. A rapid increase
of the intensity up to a partial pressure of ∼30 Pa can be
observed, after which the intensity decreases slowly to a local
minimum at ∼75 Pa (see Figure 7a). Then the intensity increases
again to a maximum point at ∼110 Pa only to decrease again
after each laser pulse. This gives rise to high evaporation rates
12
of tungsten atoms; furthermore, the rates of desorption for the
reactants in eq 1 at high temperatures are largely unknown.
Therefore, an estimation of the net effect from these two
competing processes becomes very uncertain.
Intensity of the emitted thermal radiation (I) from the laser-
heated particles follows Stefan-Boltzmann’s radiation law, i.e.
on further increase of pWF .
6
As for the nanoparticle temperature dependence on pWF (see
6
Figure 7b), it can be seen that the temperature decreases
approximately linearly (∼ -1.5 K/Pa) until a partial pressure
of ∼120 Pa WF6, after which the temperature drops rapidly
∞
4
I )
∫
n (r)A (r)ꢀ(r)T dr
(2)
c
s
0
(
∼ -10 K/Pa) with increasing pWF .
6
where n is the size-distribution function of the particles in the
c
Size-distributions of the generated W particles for different
observed volume, ꢀ is the emissivity, A is the surface area, r is
s
WF6 partial pressures are shown in Figure 8. Only at the lowest
the radius of the particle, and T denotes the absolute temperature.
pWF (15 Pa), does the distribution follow the log-normal type
The total LCVD produced tungsten mass is considered to be
6
(
see also Figure 6). At higher pWF (>30 Pa), this deviation
proportional to pWF6 (within most of the investigated partial
6
3
(from a log-normal distribution) becomes more significant, and
pressure region, see Figure 9a), so it follows that ∫n r ∝ p
c
WF6
2
it can be seen that the mean size of the particles increases and
that the distributions are broadened.
and, because ꢀ ∝ r (ref 18) and As ∝ r , eq 2 can be rewritten
as
The rate of deposition of tungsten increases continuously as
4
I ∝ pWF6
T
(3)
pWF is elevated, slightly deviating from a linear behavior, see
6
Figure 9a, and a threshold value for W nanoparticle generation
seems to be present. The geometric mean volume of the particles
follows an approximate linear dependence on pWF (for pWF g
However, eqs 2 and 3 are only valid until agglomeration of
particles occurs. For agglomerated particles (e.g., fractal-like
structures), interference effects of the light emission from the
particles within the aggregates has to be taken into consideration.
6
6
1
5 Pa), as can be seen in Figure 9b, but it should be noted that
the extrapolated intercept (for pWF ) 0) is situated on the
6

11618 J. Phys. Chem. B, Vol. 107, No. 42, 2003
Landstr o¨ m et al.
Figure 4. Normalized size-distribution histograms (sample size ∼400 from TEM BF micrographs) for five different H
2 6
partial pressures. WF
partial pressure 30 Pa with a total pressure of 2000 Pa.
Figure 6. log-probability plots for the two first histograms in Figures
and 8, respectively. Only at 15 Pa WF partial pressure can a log-
normal size-distribution be found.
4
6
also remove energy from the particles, and the energy dissipation
2
0
per unit time (P) can be described as
1
8
p〈V〉
Ta
P ) ê(l + 1)
A (T - T )
(5)
s
a
where l is the degrees of freedom for colliding molecule (l )
and 36 for H2 and WF6, respectively). The parameter ê is the
7
accommodation coefficient (ê e 1) and denotes the degree of
inelasticity of the collisions.
Figure 5. (a) Geometric mean diameter (d
rate, in measures of the intensity of the W
and (c) intensity of scattered 193 nm laser line dependence on H
pressure, respectively. Total pressure 2000 Pa with a constant WF
partial pressure of 30 Pa.
g
), (b) tungsten deposition
peak from XRF analysis,
L
R
2
partial
As for the scattering from separated (noninterfering) gas-phase
nanoparticles (r , λ), the scattering cross-section follows the
6
6
4
Rayleigh law; that is, the intensity is proportional to r /λ .
Furthermore, the formation of gas-phase aggregates will also
affect the intensity of the scattered light in a similar fashion as
for the thermal radiation.
OES measurements were performed with a 1 µs gate pulse,
and during this time, the ambient gas collides with the par-
ticles, and some of these these collisions quench the photon
H2 Partial Pressure Dependence. There is no significant
9
emission. The amount of collisions (Q) on a particle can be
change in the measured size-distributions as pH is varied (see
2
written as
Figure 4). The observed distributions deviate from the log-
normal type (see Figure 6, top panels), and this is explained by
that agglomeration and further coalescence of the particles occur
1
4
p〈V〉
Q )
As
(4)
k T
B
a
at ∼30 Pa pWF . This assumption is verified in the following
6
subsection.
where kB is the Boltzmann constant, p is the pressure, 〈V〉 is the
mean velocity, and Ta is the temperature of the ambient gas
Because of the chemistry of the net reaction (see eq 1), no
particles can be formed if no H2 is present, which can be seen
in Figure 3a (no thermal radiation was observed at zero H2
(considered as room temperature, i.e., 295 K). Inelastic collisions

W Nanoparticles Generated by Laser-Assisted CVD
J. Phys. Chem. B, Vol. 107, No. 42, 2003 11619
in Figure 3a), based on measured temperatures in Figure 3b
and the observed intensity (O in Figure 3a). The measured
Stern-Volmer plot also gives evidence for the quenching of
9
light emission. The microscopic explanation of the increased
heat conductivity of the gas mixture can be related to the
increased number of inelastic collisions between H2 molecules
and W particles, resulting in a cooling of the particles as
observed in Figure 3b. The measured data allow for the
estimation of the degree of inelasticity (ê) of the H2 collisions
by using eq 5. The collision rate of H2 molecules with a 12 nm
in diameter particle (see Figure 5a) increases with H2 partial
pressure as ∼50 collisions/(Pa µs) according to eq 4. Thus,
taking the gate pulse length (1 µs) and the slope of Figure 3b
into consideration, ê was calculated and found to be ∼0.15.
(This value should be interpreted as that 15% of the maximum
allowed energy, Emax, is transferred as an average for every H2
molecule - W nanoparticle collision. Emax ) CH (T - Ta) where
2
CH is the heat capacity of a H2 molecule.)
2
The nearly constant mean size of the particles and deposition
rate (Figure 5, parts a and b) as pH was varied is in line with
2
earlier results on photolytic LCVD of tungsten from WF6
precursor, where the rate of deposition was found being
2
2
independent of pH above a threshold value. The slight mean
2
size increase at high pH is most likely due to TCVD on the
2
1
0
Figure 7. Intensity of the thermal radiation (a) and temperature of the
surface of the hot particles.
generated W particles (b) vs WF
delay time 500 ns, H partial pressure 165 Pa with a total pressure of
000 Pa.
6
partial pressure. Gate pulse 1 µs,
WF6 Partial Pressure Dependence. The WF6 partial pressure
dependence of intensity of thermal radiation and of the tem-
perature are depicted in parts a and b of Figure 7, respectively.
2
2
(
Also recall that the total LCVD produced tungsten volume is
partial pressure). When a certain threshold value is reached (∼10
Pa), the number of generated particles allow sufficient sta-
tistics for the optical measurements. Above this threshold value,
increasing over the whole examined pressure range [see Fig-
ure 9a].)
If a similar size-distribution is assumed during the first,
the intensity increases rapidly with pH , which indicates that
2
rapidly increasing part, the intensity increase (at low pWF ) can
6
the particle concentration (nc) also increases rapidly according
to eq 2, considering that the temperature and mean size of the
particles are approximately constant (see Figures 3b and 5a).
be attributed to an increasing number of particles (see eq 2,
considering a constant temperature [see Figure 7b]). At a certain
critical concentration, collisions between the particles will be
significant resulting in agglomeration in the gas phase. As a
consequence, the coalescence in the agglomerates results in a
distortion of the initial size distribution. This distortion can easily
be seen in Figures 6 and 8, where a deviation from log-normal
distribution begins at ∼30 Pa. Agglomerated particles can also
This is interpreted as that in the ∼10-50 Pa pH interval access
2
of H2 within the reaction-zone is in deficit with respect to the
3
0 Pa WF6 partial pressure, in analogy with eq 1. The maxi-
mum intensity is observed at ∼60 Pa H2, and this suggests that
a saturation of the total volume of generated tungsten nano-
particles is reached (for the experimental parameters and
setup used). This is confirmed by the rate of deposition and
scattering measurements (see Figure 5). It is noted that,
according to the stoichiometry of the net reaction (see eq 1),
the maximum emitted intensity, in accordance with the partial
pressure of WF6 (30 Pa), should be found at 90 Pa H2 partial
pressure. This discrepancy is explained by the change in
be seen in Figure 2 (pWF ) 115 Pa), which confirms the
6
aggregation of the nanoparticles.
In the 30-120 Pa partial pressure region, the behavior of
the emitted light intensity is connected to the formation of gas-
phase aggregates. Further and detailed investigations of these
phenomena, concerning emission from aggregates, are in
progress in connection with light scattering measurements from
stoichiometry of the net reaction due to the photolytic contribu-
2
1
tion during LCVD.17 (Less H2 is needed for the tungsten
aggregated W nanoparticles.
formation since the laser photons strip away fluorine atoms.)
Because neither significant decrease nor increase of the scattered
At ∼120 Pa WF6, the intensity of the emission starts to de-
crease, and this change can be explained by attenuation of the
laser beam due to increased absorption of the generated particles
and the precursor gas at elevated pWF6 (e.g., ∼35% of th)e
incident laser light is absorbed by WF6 molecules at pWF6
120 Pa).
The linear decrease of temperature (see Figure 7b) up to ∼120
Pa is explained by inelastic collisions between the nanoparticles
and WF6 molecules, in combination with an attenuation of the
laser beam. The collision rate of WF6 molecules with a 12 nm
in diameter particle increases with WF6 partial pressure as ∼4
collisions/(Pa µs), but the high number of degrees of freedom
for WF6 results in an effective quenching, see eq 5. Above ∼120
Pa the temperature drops rapidly (see Figure 7b), which confirms
the laser attenuation.
intensity could be observed (at pH > 60 Pa) during the scattering
2
experiments (see Figure 5c), a change in size of gas-phase
aggregates is ruled out (at 30 Pa WF6 an increasing number of
aggregates was observed, see section C, and a change in size
of these aggregates would otherwise alter the scattering char-
acteristics).
The decrease of intensity of the thermal radiation after the
observed maximum can be attributed to (i) a decrease of the
temperature of the particles at elevated H2 pressures (see Figure
3
a), due to the increased heat conductivity of the gas mixture,
and (ii) to quenching of the emission by collisions with excess
H2 molecules. The quenching is evident if one considers the
intensity difference between calculated intensity (marked by ×

11620 J. Phys. Chem. B, Vol. 107, No. 42, 2003
Landstr o¨ m et al.
Figure 8. Normalized size-distribution histograms (sample size ∼400 from TEM BF micrographs) for five different WF
6 2
partial pressures. H
partial pressure 165 Pa with a total pressure of 2000 Pa.
can be applied to explain the resulting log-normal distribution.24
In the RTA model, the particle growth is assumed to be
governed purely by vapor condensation and that no aggregation
or coagulation of particles occur. Furthermore, a log-normal
distribution can also be preserved above the coagulation limit,
if the total mass is conserved and a sufficient number of col-
lisions takes place, because of the self-preserving behavior of
2
5
the coagulation process and initial distribution.
Log-normal size distributions were found at low WF6 partial
pressures (<30 Pa) in accordance with the RTA model (see
Figure 6). At high oversaturation ratios (i.e, above the coagula-
tion limit at high WF6 partial pressures), distorted distributions
were found instead of preserved initial (log-normal) distribu-
tions. Because an open system was used, the total mass of the
generated particles was not conserved, and in addition, the
residence time in the reaction-zone was small (∼4.5 s). These
facts explain the distorted distributions (relative to log-normal
and shifted toward larger diameters) for high partial pressures
of WF6 (>30 Pa).
To obtain narrow size distribution, one has to avoid the
agglomeration/coagulation, and for that, a low WF6 partial
pressure is needed. (If higher pWF is used, the critical point in
6
particle concentration for aggregation/coagulation can also be
shifted by decreasing the residence time.) The H2 partial pressure
is almost irrelevant, with the exception that it has to exceed
Figure 9. (a) Rate of deposition of tungsten from XRF measurements
and (b) cubic geometric mean radius, from Figure 8, dependence on
WF partial pressure. H partial pressure 165 Pa at a total pressure of
6 2
about twice of the value of pWF . In addition, OES is capable
6
of monitoring the agglomeration of the particles during LCVD,
and precise evaluation of this phenomenon is under way.21
Modeling of the particle growth within the aggregates is in
progress to understand the linear particle volume on WF6 partial
pressure dependence when agglomeration of particles occurs.
2
000 Pa. The dashed line in part b represents the best linear fit.
The rate of deposition of W increases close to linearly up to
∼
80-90 Pa WF6 (see Figure 9) which is consistent with earlier
22
studies. Extrapolation of the measured XRF signal to pWF )
6
0
indicates that a threshold value of generated sub-fluorides has
Conclusions
to be reached for nanoparticle formation. At pWF > 80-90 Pa,
6
a slower increase of deposition rate, compared to lower partial
pressures, is observed and is explained by two effects: (i) the
H2:WF6 ratio is less than 2, resulting in less tungsten formation
according to the net reaction (eq 1 and taking into account the
photolytic contribution) and (ii) the attenuation of the incoming
laser light due to absorption and scattering from precursor
molecules and generated W particles.
Size-distribution, temperature and intensity of emitted thermal
radiation dependence of precursor composition was examined
for LCVD generated tungsten nanoparticles from WF6/H2/Ar
gas mixtures. The precursor gas composition was systematically
varied and spectra of the emitted thermal (blackbody-like)
radiation, originating from the laser-heated nanoparticles, was
monitored by means of OES. Scattering measurements were
also performed while varying the partial pressure of H2.
Variation of the H2 concentration between 1 and 50% (at 1.5%
WF6 concentration) did not result in significant change of the
size-distributions, and only a slight increase in mean diameter
was observed. A fast increase of the intensity of the thermal
radiation was observed up to 3% H2, indicating that an in-
creasing number of particles were formed. However, above 3%
H2 concentration, a decrease of both the emitted intensity and
temperature of the particles could be observed because of (i)
the quenching of excited states and (ii) an increased heat transfer
Recall that the intercept (in particle mean volume dependence,
see Figure 9b) is situated on the negative y axis. This confirms
the above statement that a threshold value for pWF exists for
6
generation of tungsten nanoparticles.
On the Observed Size Distributions. If the particles are
formed during an equilibrium condensation from an oversatu-
rated vapor, with constant oversaturation ratio, the expected
result is a log-normal distribution of the particles.23 Below the
so-called coagulation limit (the point where significant coales-
cence takes place), the residence time approach (RTA) model

W Nanoparticles Generated by Laser-Assisted CVD
J. Phys. Chem. B, Vol. 107, No. 42, 2003 11621
by the increasing number of H2 collisions, respectively. The
accommodation factor ê (degree of inelasticity of the H2-W
nanoparticle collisions) was found to be ∼0.15. The rate of
deposition of tungsten and the intensity of the scattered laser
line dependence on H2 partial pressure revealed that a saturation
in W deposition was reached at ∼3% H2. This saturation was
found at a WF6:H2 ratio of 1:2 (instead of 1:3 ratio according
to stoichiometry) and was explained by the photolytic contribu-
tion of the UV laser energy to the W deposition rate.
acknowledged. One of the authors (P.H.) is also thankful for
partial supports from the OTKA foundation with Grant Nos.
T34381 and TS040759.
References and Notes
(
1) Hayashi, C.; Uyeda, R.; Tasaki, A. Ultra-Fine Particles; Noyes
Publications: New Jersey, 1997.
2) Nalwa, H. S. Handbook of Nanostructured Materials and Nano-
technology; Academic Press: San Diego, 2000.
(
(
3) Gleiter, H. Acta Mater. 2000, 48, 1.
The rate of deposition of W dependence of WF6 concentration
increased almost linearly in the observed partial pressure range
(4) Eckert, J.; Holzer, J. C.; Krill, C. E., III.; Johnson, W. L. J. Mater.
Res. 1992, 7, 1751.
(
(
5) Huczko, A. Appl. Phys. A 2000, 70, 365.
6) Brinker, C. J.; Scherer, G. W. Sol-Gel Science; Academic Press:
(0-8.5%), and the mean volume of the generated tungsten
nanoparticles depended linearly on the WF6 concentration. A
rapid increase of the emitted thermal radiation could be observed
up to 1.5% WF6 (at 8% H2 partial pressure). In the partial
pressure range of 1.5-6%, the behavior of the intensity of
emitted light can be explained by the formation of gas-phase
aggregates. Importantly, size-distribution data verified this
concept, because a broadening and shift of the distributions (due
to agglomeration and further coalescence) was found above
Boston, 1990.
(7) Hahn, H. Nanostruct. Mater. 1997, 9, 3.
(8) Shintani, A.; Tsuzuku, S.; Nishitani, E.; Nakatani, M. J. Appl. Phys.
987, 61, 2365.
1
(
9) Heszler, P.; Mogyor o´ si, P.; Carlsson, J.-O. J. Appl. Phys. 1995,
7
8, 5277.
(10) Heszler, P.; Landstr o¨ m, L.; Lindstam, M.; Carlsson, J.-O. J. Appl.
Phys. 2001, 89, 3967.
(
11) Heszler, P.; Carlsson, J.-O. Vacuum 1995, 46, 955.
(12) Landstr o¨ m, L.; Heszler, P. submitted.
1
.5% WF6 concentrations. Above 6% WF6, both the emitted
(13) H a¨ gg, G.; Sch o¨ nberg, N. Acta Crystallogr. 1954, 7, 351.
(14) Tang, C. C.; Hess, D. W. Appl. Phys. Lett. 1984, 45, 633.
thermal intensity and temperature of the W particles decreased
rapidly because of increased absorption of the incident laser
light by the WF6 precursor molecules and the generated
nanoparticles.
Log-normal and narrow size distributions were found for gas
mixtures consisting of low WF6 concentrations, where no
significant agglomeration of the particles leading to coalescence
occurred. Thus, to obtain narrow distributions, a low concentra-
tion of WF6 should be used (independently of the H2 concentra-
(
15) Landstr o¨ m, L.; Kokavecz, J.; Lu, J.; Heszler, P. submitted.
(16) Mogyor o´ si, P.; Carlsson, J.-O. J. Vac. Sci. Technol. A 1992, 10,
3
131.
(17) Heszler, P.; Carlsson, J.-O.; Mogyor o´ si, P. J. Vac. Sci. Technol. A
993, 11, 2924.
18) Bohren, C. F.; Huffman, D. R. Absorption and scattering of light
1
(
by small particles; John Wiley & Sons: New York, 1983.
(19) Hitchman, M. L.; Dobson, A. J. Appl. Surf. Sci. 1989, 38, 312.
(20) B a¨ uerle, D. Laser Processing and Chemistry, 3rd ed.; Spriger
Verlag: Berlin, 2000.
(21) Landstr o¨ m, L.; Heszler, P. to be published.
tion if pH > 2pWF ). Alternatively, the residence time within
2
6
(22) Matsuhashi, H.; Nishikawa, S.; Ohno, S. Mater. Res. Soc. Symp.
Proc. 1989, 129, 63.
the reaction zone should be decreased if higher concentrations
of WF6 are to be used.
(
(
23) Scheibel, H. G.; Porstendorfer, J. J. Aerosol Sci. 1983, 14, 113.
24) S o¨ derlund, J.; Kiss, L. B.; Niklasson, G. A.; Granqvist, C. G. Phys.
ReV. Lett. 1998, 80, 2386.
25) Lai, F. S.; Friedlander, S. K.; Pich, J.; Hidy, G. M. J. Colloid
Interface Sci. 1972, 39, 395.
Acknowledgment. Financial support from the Swedish
Research Council and G o¨ ran Gustafsson foundation is gratefully
(