You are on page 1of 10

Chemical Engineering Science 57 (2002) 1753 – 1762

Design of metal nanoparticle synthesis by vapor $ow condensation

Karsten Wegner, Bernhard Walker, Stavros Tsantilis, Sotiris E. Pratsinis ∗
Institute of Process Engineering, Swiss Federal Institute of Technology, Sonneggstrasse 3, ETH Zentrum ML F26,
CH–8092 Zurich, Switzerland
Received 16 January 2001; received in revised form 26 November 2001; accepted 12 December 2001


A systematic engineering study on continuous synthesis of bismuth metal nanoparticles by vapor condensation in tube $ows is presented.
Simulations of aerosol nucleation, condensation and coagulation are cast in a design diagram format to guide experimental studies on
the e4ect of process parameters on product nanoparticle characteristics. Flow visualization, tracer gas analysis and computational $uid
dynamics are used to unravel the e4ect of particle residence time distribution on product morphology during evaluation of alternate
quenching designs for the metal vapor. Bismuth nanoparticles of average diameter 12–37 nm, as determined by nitrogen adsorption and
X-ray di4raction, were made by controlling the quenching gas $ow rate, carrier gas $ow rate and process pressure. ? 2002 Elsevier
Science Ltd. All rights reserved.

Keywords: Design; Nanostructure; Particle formation; Fluid mechanics; Bismuth; Inert gas condensation

1. Introduction aluminum particles by melt evaporation in helium and re-

ported the formation of bigger particles for increased inert
A variety of metal nanoparticles are made today for a gas pressure, evaporation temperature and residence time
spectrum of niche applications. Condensation of the corre- in the reactor. Granqvist and Buhrman (1976) and ThKolLen
sponding metal vapors is widely used for the manufacture of (1979) made similar observations for other metal nanopar-
these nanoparticles (Gleiter, 1989; Siegel, 1991). Very little ticles when they evaporated the metals from a crucible into
systematic design and operation of these processes has been the reduced atmosphere of an inert gas.
performed since metal nanoparticles are still sold at rather Sattler, MKuhlbach, and Recknagel (1980) and MKuhlbach,
high prices. As this technology matures and product prices Recknagel, and Sattler (1981) generated small clusters of
become competitive, the need for optimal process design bismuth, antimony and lead by vapor expansion in a free
and engineering will rise. Here, such an e4ort is made for jet from an oven source in a liquid-nitrogen-cooled conden-
continuous synthesis of bismuth nanoparticles that are used sation cell. They reported the formation of single bismuth
in dry phototools (Eickmans, Leenders, Lamotte, Dierksen, atoms to 270 atom-clusters as determined by time of $ight
& Jacobsen, 1996). (TOF) mass spectroscopy when the cold He atmosphere of
Pfund (1930) and Burger and van Cittert (1930) synthe- 10 Torr containing the Bi clusters was expanded into the
sized bismuth nanoparticles by evaporating bismuth from high-vacuum TOF chamber (≈ 10−7 Torr). However, only
a tungsten wire at low pressure. Particles formed in the small quantities of clusters can be produced by this free
colder region of the unit were transported by free con- jet expansion technique while the common inert gas con-
vection to a cold surface (“cold Cnger”) where they were densation (IGC) method has the potential to produce gram
collected by thermophoresis. Since these early experi- quantities of nanoparticles in a single run (Samy El-Shall &
ments, direct evaporation of metals from heated crucibles Edelstein, 1996).
into a gas has been widely studied (Gleiter, 1989; Uyeda, Andres et al. (1989) and Siegel (1991) proposed the uti-
1991). Yatsuya, Kasukabe, and Uyeda (1973) synthesized lization of forced gas $ows in the IGC process for production
of larger quantities of nanoparticles. The method was ap-
∗ Corresponding author. Tel.: +41-1-632-2510; plied by Haas, Gleiter, and Birringer (1993) for the synthe-
fax: +41-1-632-1595. sis of Cu and Pd nanoparticles. They modiCed the classical
E-mail address: (S. E. Pratsinis). IGC process by introducing a continuous inert gas jet above

0009-2509/02/$ - see front matter ? 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 0 2 ) 0 0 0 6 4 - 7
1754 K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762

the heated crucible. The particle transport by forced convec- 2:2 mm i.d. Inconel tube. This inlet tube is supported by a
tion resulted in a reduction of number concentration and res- stainless steel $ange which also closes the feed side of the
idence time of the nuclei so that smaller nanosize powders pipe and contains connectors for pressure and temperature
could be produced. Konrad, Haubold, Birringer, and Gleiter sensors.
(1996) applied continuous gas $ow also for the synthesis of On the other side of the Inconel pipe a stainless steel lance
nanosize copper–bismuth alloys by co-evaporation. Copper consisting of a center tube and three annular tubes forms a
crystallites with up to 10% bismuth could be synthesized by diluter, heat exchanger and condenser outlet (Fig. 1). The
directing the gas $ux over Crst the copper and then the bis- 16:5 mm i.d. center tube of this lance (diluter-tube) carries
muth melt. The jet aerosol $ow condenser was investigated the product aerosol to a glass Cber Clter (Whatman GF=A) in
further by Haas, Birrringer, Gleiter, and Pratsinis (1997) for a stainless steel Clter holder with the help of a vacuum pump
the synthesis of palladium nanoparticles with primary par- (Vacuubrand RZ 16). Cooling water of 5◦ C $ows through
ticle median diameters of 2–11 nm. Compared to IGC pro- the Crst and third annulus of the lance at a $ow rate of 60 l=h.
cesses relying on free convection, an 80% reduction in the An argon stream (Pan Gas 99.999%) is introduced into the
primary particle median diameter and a 17% reduction in condenser through the second annulus between the cooling
the geometric standard deviation of the primary particle size water shell as quenching gas. This quenching gas discharges
distribution could be achieved with this aerosol $ow con- in counter$ow to the aerosol jet through 12 equally sized
denser. Tsantilis, Pratsinis, and Haas (1999) combined com- nozzles with an inner diameter of 2:5 mm which are uni-
putational $uid and monodisperse particle dynamics to de- formly arranged at the front side of the lance as shown in
scribe the $ow pattern and particle formation in this aerosol the photograph in Fig. 1 (cap-free diluter). The schematic
$ow condenser. in Fig. 1 shows a stainless steel gas return cap which can be
Here, the synthesis of metal nanoparticles by vapor con- mounted on the diluter to guide the quenching gas perpen-
densation is systematically investigated. Aerosol process de- dicularly to the Bi-laden carrier gas stream, thus forming a
sign is used, for the Crst time to our knowledge, to guide T-mixer. This cap can separate the evaporation zone from
experimental studies on the e4ect of process parameters (va- the condensation (quenching and particle formation) zone
por quenching rate, temperature and pressure) on product of the unit (cap-diluter). Here, the performance of the two
metal nanoparticle characteristics. Hence, synthesis of bis- diluter (or quenching gas) conCgurations in gas cooling and
muth nanoparticles is investigated as these particles are used mixing is investigated for their e4ect on particle formation.
for dry phototools in the production of printed circuit boards An additional water-cooled heat exchanger is connected
(Eickmans et al., 1996). The 200 nm-thin bismuth imaging with a gate valve to the outlet of the diluter and cools the
layer of these phototools must be uniform and requires parti- aerosol to a temperature of 30◦ C at the Clter. This temper-
cles to have average diameters of 50 nm or less, a low degree ature is recorded at the center of the Clter with a type K
of agglomeration and a reasonably narrow primary particle thermocouple and controls the cooling water $ow rate. A
size distribution (Eickmans et al., 1996). Here, bismuth va- precision needle valve $ow controller (Bronkhorst EL-Flow
por is produced in an externally heated tube $ow condenser 201 A) connected to the gas-side inlet of the heat exchanger
containing a crucible with the Bi melt. Transport of the va- (10 cm after the diluter and 30 cm before the Clter) allows
por in an argon carrier gas jet is followed by rapid quench- the addition of trace amounts of oxygen (Pan Gas, 99.999%)
ing with a second argon stream (in a diluter) and by particle to the aerosol stream. All tube connections are made with
collection on a Clter. This experimental design is selected vacuum $anges (Balzers KF-series) to allow gas-tight op-
as it separates the evaporation from the particle formation eration at subatmospheric pressure. The carrier and quench-
zone (during quenching) allowing to systematically study ing gas streams are controlled by calibrated mass $ow
the in$uence of cooling rate and quenching conCguration controllers (Bronkhorst EL-Flow 201 A and AC). The
on product particle characteristics using nitrogen adsorption, pressure in the condenser is measured by a capacitative
transmission electron microscopy and X-ray di4raction. pressure sensor (MKS Baratron) and is controlled by a
butter$y valve (MKS 653B) at the inlet side of the vacuum
2. Experimental

2.1. Apparatus 2.2. Procedure

Fig. 1 shows a schematic of the jet aerosol $ow con- Prior to heating the unit, the ceramic crucible was loaded
denser unit. The middle section of an Inconel pipe (length with about 5 g of bismuth granules (Fluka, 99.9995%) and
1000 mm, inner diameter 52:5 mm) is heated to the desired placed at the appropriate position in the $ow condenser.
process temperature by a tubular furnace (Carbolite CFM The condenser was closed air-tight, purged with argon to
14=1) to evaporate bismuth from an alumina crucible (Alsint a pressure of 1500 mbar and evacuated to 0:1 mbar after-
99.7). From the pipe inlet (feed side) argon carrier gas (Pan wards. This procedure was repeated three times to remove
Gas, 99.999%) is introduced above the crucible through a air from the system. The condenser was then closed with
K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762 1755

Fig. 1. Schematic of the aerosol $ow condenser with its Inconel pipe (1), the crucible containing the bismuth melt (2), the furnace (3), the evaporation
zone (4), and the cap-free diluter (5). An inset schematic shows the diluter conCguration with gas return cap (6) used to introduce the quenching gas
perpendicular to the Bi-laden carrier gas jet.

the gate valve, Clled with argon to atmospheric pressure (Micromeritics Gemini 2306) employing the BET isotherm.
and heated at a rate of 25◦ C=min to the operating temper- Assuming all particles to be monodisperse spheres,
ature of 1000◦ C. After the set temperature was reached, the BET-equivalent particle diameter was calculated as
the gate valve was opened slowly and the argon $ow rates dp = 6=(Bi SSA), with the density of bismuth at room
and system pressure were adjusted to the desired operating temperature being Bi = 9790 kg=m3 (Lide, 1997). Trans-
conditions. The oxygen concentration in the particle col- mission electron microscopy (Hitachi H600, operated at
lection region was set to 1000 ppm with the needle valve 100 kV) was employed to determine the morphology of the
$ow controller in order to form a passivating oxide layer bismuth nanoparticles. Therefore, 0:2 mg of powder was
at the surface of the freshly produced bismuth particles suspended in 1 ml of toluene (Fluka, ¿ 99:5%) and 0:1 ml
(Russell, 1967; Uchikoshi, Sakka, Yoshitake, & Yoshihara, of the surfactant octyltrimethoxysilane (Fluka, ¿ 97%) be-
1994). Without oxygen, bismuth nanoparticles with diame- fore it was applied to a TEM grid. X-ray di4raction analysis
ters smaller than 18 nm rapidly oxidized after exposure to (XRD, Siemens D 5000, Cu K radiation) was performed
ambient air. In a standard experiment the furnace temper- over an angle of 20◦ ¡ 2 ¡ 60◦ in steps of 0:02◦ . Care
ature was 1000◦ C, the pressure 1000 mbar, while the oxy- was taken that all characterization methods were carried out
gen, carrier and quenching gas $ow rates were 11:5 × 10−3 , right after synthesis since bismuth nanoparticles slowly oxi-
1.5 and 10 l=min, respectively. Experiments were carried dize in air and can completely transform into bismuth oxide
out from 200 to 1000 mbar, carrier gas $ow rates of 0.75 after a few weeks, as was observed in the course of this
–3:0 l=min and quenching gas $ow rates of 5 –20 l=min. All work.
volumetric $ow rates are reported at 25◦ C and 1013 mbar. Centerline temperatures in the $ow condenser were
After each experimental run the gate valve was closed and recorded at 1000 mbar and a furnace set temperature of
the Clter was changed in ambient air. If no further exper- 1000◦ C for quenching gas $ow rates of 5, 10, 15 and
iments were performed, the condenser was slowly cooled 20 l=min and a carrier gas $ow rate of 1:5 l=min. Therefore,
down to room temperature. Each experiment was performed a 750 mm long S-type thermocouple with a 0:5 mm bead
twice and found to be reproducible within 5% of the product supported by an alumina tube (o:d: = 4 mm) was introduced
speciCc surface area (SSA). from the unit outlet. The thermocouple was mounted on an
xy-positioning table allowing precise positioning inside the
2.3. Measurements condenser. Temperatures were recorded every 2 cm inside
the evaporation zone and every 1 cm in the quenching zone.
The average speciCc surface area (SSA) of the pow- Measurements were performed for both the cap-free and
ders was determined by nitrogen adsorption at 77 K the cap-diluter.
1756 K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762

2.4. Flow characterization jet at an outlet temperature of 1000◦ C is about 400 for an
argon $ow rate of 1:5 l=min, indicating that the $ow in the
Cold $ow visualization experiments were carried out in a condenser is in the transition regime (Bejan, 1984). There-
transparent (quartz glass) unit of similar dimensions at room fore, the simulations were performed using the RNG k–
temperature and pressure in order to visually study the in$u- model (Yakhot & Orszag, 1986) provided by FLUENTTM
ence of the diluter cap on the $ow pattern of particles inside for $ows in the transient regime (Fluent Inc., 1998) ac-
the condenser. Therefore, the crucible was Clled with am- counting for low Reynolds number e4ects. The tempera-
monia solution (NH3; aq , Aldrich 24%) while the carrier gas tures at the condenser walls were measured and set as con-
was saturated with hydrochloric acid (HClaq , Aldrich 25%) stant boundary conditions for the simulations. At a furnace
in a bubbler prior to introduction into the unit, leading to the set point of 1000◦ C, the plate supporting the carrier gas
formation of ammonium chloride particles above the cru- nozzle, the nozzle itself and the carrier gas had a temper-
cible (Dahlin, Su, & Peters, 1981). A vertical plane through ature of about 980◦ C. A temperature of 680◦ C was mea-
the center axis of the condenser was illuminated with a laser sured for the walls of the cap, while the temperature of the
sheet created by an argon laser (Lexel 95, operated at 2:5 W) diluter-tube and the quenching gas was 5◦ C. The outer walls
and a plano-concave cylindrical lens (Newport CKV513-C). of the evaporation chamber exhibited a temperature proCle
The laser light was scattered by the ammonium chloride par- which could by approximated with the parabolic function
ticles (corresponding to the bismuth particles of the present T (◦ C) = −2:58x2 − 1:50x + 980, where x (cm) is the dis-
experiment) and made the $ow pattern in the unit visible. tance from the condenser inlet. The wall temperatures did
Photographs of $ow patterns were taken with a digital cam- not vary for di4erent operating conditions.
era (Minolta RD 175) for carrier and quenching gas $ow A monodisperse particle dynamics model accounting
rates of 1.5 and 10 l=min, respectively. for simultaneous nucleation, condensation and coagulation
The residence time distribution of the carrier gas in the (Panda & Pratsinis, 1995) was employed to derive a de-
$ow condenser was investigated by introducing CO2 (Pan sign diagram for synthesis of bismuth nanoparticles in a jet
Gas 99.995%) as a tracer into the carrier gas stream at the aerosol $ow condenser (Tsantilis, Pratsinis, & Haas, 1999).
unit inlet and detecting the CO2 concentration at the unit out- SpeciCcally, the e4ects of the evaporation temperature Tv
let with a mass spectrometer (Balzers GSD 300 O). There- (K) and the inert gas $ow rate Qr (l=min) on bismuth aver-
fore, a sampling loop was loaded with 3 ml CO2 and in- age particle diameters were investigated. The Bi properties
stantly discharged into the argon carrier gas stream with the of importance were surface tension (Frurip & Bauer, 1977):
help of a GC valve (C. Erba S.A. 40541), thus creating a
Dirac peak. The time between the discharge of the sampling  (J=m2 ) = (418 − 0:07 × T )=1000 (1)
loop and the drop of the CO2 concentration in the outlet gas
to its initial value was recorded along with the CO2 concen- and vapor pressure (Hultgren, Desai, Hawkins, Gleiser, &
tration. Kelley, 1973)

ps (Pa) = 1:013 × 105 exp(12:398 − 22; 570=T ): (2)

3. Simulation of temperature, velocity and particle growth

Temperature and velocity proCles inside the $ow con- Furthermore, the nucleation rate was given by Girshick and
denser were simulated using the commercially available Chiu (1990), while the collision frequency coeTcient for
computational $uid dynamics (CFD) software FLUENTTM Brownian coagulation was based on the Fuchs interpolation
5.2. The geometry of the condenser can be taken as axisym- function from the free molecular to the continuum regime
metric if the in$uence of the crucible on the $ow is neg- (Seinfeld, 1986). The design calculations were carried out
ligible and the 12 outlet nozzles of the quenching gas are for a tubular aerosol $ow condenser with dimensions con-
modeled as an annular slit providing the same average gas sistent with those of the diluter (quenching zone) of the ac-
outlet velocity. The in$uence of these simpliCcations on the tual experimental setup (i:d: = 16:5 mm). In addition, it is
$ow characteristics inside the unit was evaluated by carry- assumed that the argon carrier gas is saturated with bismuth
ing out a three-dimensional (3-D) simulation for the actual vapor and enters the condenser at evaporation temperature
geometry at a system temperature of 1000◦ C, a pressure of TV (ranging from 1000 to 1150 K) and $ow rate Qr (quench-
1000 mbar and a carrier gas $ow of 1:5 l=min and compar- ing and carrier gas, ranging from 5 to 20 l=min) at 1 atm. For
ing it to the two-dimensional (2-D) axisymmetric conCgura- every set gas $ow rate, the temperature was assumed to drop
tion without crucible. The time-consuming 3-D simulation linearly with a constant gradient of (TV − TL )=L = 50 K=cm
revealed only a minor de$ection of the carrier gas jet by the over the investigated length of L = 10 cm. Therefore, the
crucible presence which was negligible for the study of the cooling rate at the exit of the condenser was calculated as
overall $ow pattern in the condenser. Thus, all CFD simu-  
TV − T L 4Qr TL
lations here were made for the 2-D axisymmetric geometry C (K=s) = (3)
described above. The Reynolds number of the carrier gas L D2 Tr
K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762 1757

where Tr is the reference temperature (296 K) and TL the

gas temperature at x = L. For TV and Qr the product av-
erage particle diameter was calculated following Tsantilis,
Pratsinis, and Haas (1999).

4. Results and discussion

4.1. Design diagram

Fig. 2 shows the product bismuth particle diameter pre-

dicted by the model as a function of the cooling rate at
four evaporation temperatures, TV . The largest particles of
about 95 nm diameter are produced at TV = 1150 K and a
cooling rate of about 4000 K=s, corresponding to a gas $ow
rate of 5 l=min. Decreasing TV and increasing the cooling
rate leads to the formation of smaller particles, the smallest
of which have diameters of about 5 nm. Lower evaporation
Fig. 2. Design diagram for the synthesis of bismuth nanoparticles in
temperatures result in lower metal vapor pressures leading the jet aerosol $ow condenser showing the particle size predictions of a
to the formation of smaller particles in agreement with Yat- monodisperse model for Bi-vapor nucleation, condensation and coagula-
suya, Kasukabe, and Uyeda (1973), Granqvist and Buhrman tion. Lowering the evaporation temperature TV and increasing the cooling
(1976) and Panda and Pratsinis (1995). Higher cooling rates rate decreases the product particle size.
result in a higher supersaturation of the vapor and the for-
mation of a large number of small bismuth particles. Since visualization, CO2 tracer gas analysis and computational
high cooling rates are achieved by larger gas $ow rates, the $uid dynamics.
residence time in the $ow condenser is also shorter, decreas-
ing the particle size further. The design diagram indicates 4.2.1. Flow pattern
that bismuth nanoparticles with the envisioned size of about Visualization of the interplay between $ow and particle
50 nm (Eickmans et al., 1996) can be produced with the jet formation can be accomplished following the formation of
aerosol $ow condenser when it is operated at TV = 1150 K NH4 Cl particles at the standard experimental conditions but
with cooling rates of 10; 000 K=s and higher. at room temperature. Fig. 4a shows how quenching gas from
one of the uppermost nozzles of the cap-free diluter dis-
4.2. Synthesis of bismuth nanoparticles charges into the evaporation zone. The carrier gas stream
is diluted directly behind the crucible, creating a zone of
Fig. 3a shows a typical TEM picture of bismuth nanopar- almost uniform particle concentration, as indicated by the
ticles synthesized at standard operating conditions and rather uniform light scattering intensity there. When the cap
200 mbar with counter$ow introduction of the quenching is placed on the diluter-tube (Fig. 4b), the quenching gas
gas (cap-free diluter). The TEM picture shows a number does not enter the evaporation zone. This allows the core
of large particles of 50 –100 nm in diameter along with $uid of the carrier gas jet to $ow nearly undisturbed through
small particles with diameters of ¡ 20 nm. The broad par- the condenser and to enter the diluter through the cap open-
ticle size distribution indicates that particles have di4erent ing. Gas at the outer boundary of the jet de$ects at the cap
temperature and residence time histories, possibly due to and back$ows.
ineTcient mixing with quenching gas. As T-mixers are Velocity and temperature proCles were calculated by com-
known to be most eTcient in rapid $uid mixing (Chilton & putational $uid dynamics for a furnace set temperature of
Genereaux, 1930; Forney, NaCa, & Vo, 1996), the cap was 1000◦ C, a pressure of 1000 mbar, a carrier gas $ow rate
placed on the diluter to form such a T-mixing conCguration of 1:5 l=min and a quenching gas $ow rate of 10 l=min for
and experiments were repeated. A TEM picture of product the conCgurations with and without cap. Fig. 5 shows the
particles synthesized with this cap-diluter conCguration is CFD-calculated velocity vectors of unit length. The color
shown in Fig. 3b. The primary particle size distribution is of the velocity vectors represents gas temperature. In both
much narrower than in the conCguration with the cap-free diluter (quenching gas) conCgurations, the 1000◦ C hot car-
diluter (Fig. 3a), speciCcally, no large primary particles rier gas jet discharges into the evaporation chamber from
with diameters above 50 nm are found. This strong in$u- the left at a velocity of about 30 m=s and entrains $uid at its
ence of mixing (diluter or quenching design) on the product outer boundaries. Fluid from the downstream jet is recircu-
particle size distribution motivated us to systematically in- lated to the region near the nozzle exit, setting up a recir-
vestigate the residence time distribution in the unit by $ow culating vortex. The core $ow of Bi-vapor-laden carrier gas
1758 K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762

Fig. 4. Visualization of the $ow in the condenser using ammonium

chloride particles freshly formed from NH3 (in the crucible) and HCl
($owing gas) at ambient temperature and pressure for (a) the cap-free
diluter and (b) the cap-diluter. Carrier and quenching gas $ow rates were
1.5 and 10 l=min, respectively.

Fig. 3. TEM pictures of bismuth nanoparticles synthesized at standard

experimental conditions and 200 mbar with the cap-free diluter (a) and the
cap-diluter (b). The primary particles have a broader size distribution for
the cap-free diluter design, indicating di4erent temperature and residence
time histories.

enters the diluter-tube. In the cap-free diluter conCguration

(Fig. 5a), quenching gas at 5◦ C is introduced in counter-
$ow and mixes with the Bi-vapor-laden carrier gas jet be-
fore the diluter. A fraction of the quenching gas is entrained
by the carrier gas jet and directly enters the diluter-tube at
a temperature of 200 to 400◦ C. The rest of the quenching
gas and part of the carrier gas back$ows along the outer
pipe walls toward the entry of the condenser and entrains
into the fresh carrier gas jet. A small pocket of recirculating Fig. 5. Velocity vectors calculated by FLUENTTM for the conCgurations
$uid is formed in front of the diluter. With the cap-diluter without cap (a) and with cap (b) on the diluter. The carrier and quenching
(Fig. 5b) the quenching gas directly enters the diluter-tube gas $ow rates were 1.5 and 10 l=min, the system pressure was 1000 mbar
in co-$ow with the Bi-vapor-laden carrier gas. Thus, only and the furnace temperature was 1000◦ C.
Bi vapor and carrier gas are present in the evaporation zone,
as was shown in the $ow visualization (Fig. 4b). Gas at the
outer boundaries of the carrier gas jet, however, is de$ected 10 cm axial extension and 2:5 cm width in front of the cap.
at the face of the cap and streams back towards the carrier A substantial amount of Bi-vapor-laden carrier gas is thus
gas inlet, setting up a toroidal recirculating vortex of about trapped in these dead volumes.
K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762 1759

Fig. 7. Centerline temperatures (symbols) measured under standard ex-

perimental conditions and saturation ratio S = p=ps (lines) for the two
diluter conCgurations. Gas cooling in the case of the cap-free diluter leads
to high supersaturation and particle formation in the evaporation zone.
Placing a cap on the diluter results in steeper supersaturation and particle
formation later in the process.

is reentrained. Most of this gas now exits the evaporation

chamber, resulting in a rather narrow residence time distri-
bution. For the cap-diluter, the strong recirculating vortex
of pure carrier gas in front of the cap releases the tracer gas
over a long period of time, resulting in the broad residence
time distribution observed in Fig. 6b. The broader RTD
for the cap-diluter is in contradiction with the narrower
Fig. 6. Residence time distribution (RTD) of carrier gas $owing through particle size distribution observed for this conCguration
the aerosol $ow condenser at a rate of 1:5 l=min and standard experimental
conditions. The exit age distribution function E(t) is determined by
(Fig. 3). Therefore, temperature measurements are neces-
introducing a pulse signal of CO2 tracer gas. Placing the cap on the sary to evaluate the e4ect of gas recirculation on particle
diluter (b) broadens the RTD. formation and growth.

The formation of dead volumes shown in the $ow simu- 4.2.2. Temperature
lation was veriCed by measuring the residence time distri- Fig. 7 shows the condenser temperature along the center-
bution (RTD) of carrier gas, following the CO2 tracer gas line with the corresponding saturation ratio S =p=ps for both
concentration. Fig. 6 shows the response signal to a Dirac the cap-free and the cap-diluter, where p is the partial pres-
peak of CO2 that is inserted at t = 0 into the carrier gas sure of bismuth calculated with the evaporation rate from the
stream for each diluter conCguration. For the cap-free di- crucible and ps is the saturation vapor pressure of bismuth
luter (Fig. 6a) most of the tracer leaves the unit after about (Eq. (2)). The temperature in the unit with cap-free diluter
3:5 s, creating a sharp peak with a small shoulder toward decreases almost linearly with the distance from the carrier
longer residence times, which indicates that only a small gas nozzle (925◦ C) and reaches 570◦ C at the entrance of the
fraction of the gas is trapped in dead volumes, as was pre- diluter-tube, 11 cm from the nozzle exit. This rather steep
dicted by the $ow simulation (Fig. 5a). After 10 s, all CO2 temperature decrease is due to the dilution of the Bi-laden
has left the unit. Placing the cap on the diluter drastically carrier gas with cold quenching gas, as is shown in Figs. 4
alters the residence time distribution, as can be seen by com- and 5. The temperature gradient remains unchanged inside
paring Fig. 6a with Fig. 6b. The maximum of the response the diluter-tube and a temperature of 380◦ C is reached after
still exits the unit after about 3:5 s, the height of the peak 19 cm from the nozzle exit. Temperatures are much higher
representing a normalized CO2 concentration, however, is in the evaporation zone with the cap-diluter since quenching
only about half the value of the one detected before. The gas does not enter this region (Figs. 4 and 5). A tempera-
RTD exhibits a distinct shoulder for t ¿ 4 s, indicating that ture of 780◦ C is obtained at the cap face, 10 cm from the
a considerable fraction of the tracer gas is trapped in dead carrier gas nozzle. With the cap-diluter a steep temperature
volume and is slowly released over a period of about 20 s. gradient results inside the diluter-tube, giving a temperature
This di4erence in RTD can be explained by the visualization of 260◦ C after 19 cm from the inlet nozzle. This is about
and simulation of the $ow patterns shown in Figs. 4 and 5. 120◦ C lower than for the cap-free diluter and shows the su-
In the cap-free diluter conCguration, carrier gas which does perior performance of this conCguration in gas quenching
not directly exit the condenser is transported back to the in agreement with literature (Chilton & Genereaux, 1930;
upstream region by the counter-$owing quenching gas and Forney, NaCa, & Vo, 1996).
1760 K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762

Fig. 9. Axial centerline temperatures for the cap-diluter and four quenching
gas $ow rates (carrier gas $ow rate: 1:5 l=min, pressure: 1000 mbar, fur-
nace set temperature: 1000◦ C). Steeper temperature gradients are achieved
at higher quenching gas $ow rates.

nitrogen-cooled TEM grid. The sampling location was

Fig. 8. TEM picture of spherical bismuth particles collected by ther- downstream of the diluter at the center axis of the $ow,
mophoretic sampling from the gas phase, 10 cm in front of the Clter. approximately 10 cm ahead of the Clter. Particles directly
Particle synthesis was carried out in the condenser with cap at standard deposited from the gas phase are similarly sized spheres.
conditions at 200 mbar.
It is worth noting that these particles are essentially
non-agglomerated as Au generated by vapor expansion (Fig.
The saturation ratio S in Fig. 7 indicates whether particle 2a in Andres et al., 1996). The state of agglomeration seen
formation can already occur in the evaporation zone extend- in Fig. 3b is attributed to particle necking while on the Clter.
ing 10 cm from the carrier gas nozzle. For the cap-free di-
luter S ¿ 1 is already obtained after 3:5 cm from the nozzle 4.3. E7ect of process variables on product powder
exit at a temperature of 805◦ C. Thus, particle formation is characteristics
likely to occur before the gas enters the diluter-tube. Flow
simulation (Fig. 5) and RTD measurement (Fig. 6) for the In the production of thin Clms, small particles with a nar-
cap-free diluter conCguration have shown that a fraction of row size distribution are preferred (Eickmans et al., 1996).
the gas recirculates in the evaporation zone. Particles en- This is best obtained with the cap-diluter (T-mixer) which
trapped in these $uid parcels grow larger and may lead to allows controlled particle formation and guaranties the
the observed brand particle size distribution (Fig. 3a). In the steep cooling rates necessary for synthesis of small particles
cap-diluter conCguration, a saturation ratio larger than unity (Fig. 2). Therefore, the e4ect of operating conditions on
is reached only after about 8:5 cm from the carrier gas noz- product particle characteristics was investigated using the
zle (1:5 cm from the cap face). Particles formed in this re- cap-diluter (gas quenching) design.
gion are likely to directly enter the diluter-tube as is apparent Fig. 9 shows the in$uence of the quenching gas $ow rate
from the $ow visualization and simulation (Figs. 4 and 5). on the centerline temperature for the cap-diluter. Flow rates
The strong vortex in front of the cap and the large dead vol- of 5, 10, 15 and 20 l=min are investigated for a carrier gas
ume detected with the RTD measurement (Fig. 6) for this $ow rate of 1:5 l=min, a furnace set temperature of 1000◦ C
conCguration are unlikely to in$uence the particle forma- and 1000 mbar. Increasing the quenching gas $ow rate leads
tion process since only vapor exists in this high-temperature to steeper temperature gradients in the dilutertube but does
recirculation zone. Rapid quenching with the T-mixer (Fig. not a4ect the temperature in the evaporation zone. 19 cm
7), however, leads to high supersaturation of the gas in the away from the carrier gas nozzle exit, the lowest tempera-
diluter and formation of a large number of similarly sized ture of 195◦ C is reached for a quenching gas $ow rate of
particles. These observations underline the importance of 20 l=min whereas a $ow rate of 5 l=min results in a center-
process design for synthesis of metal nanoparticles by con- line temperature of 365◦ C.
densation. Fig. 10 shows the BET-equivalent diameter of product
The narrow particle size distribution obtained with bismuth nanoparticles made at a furnace set temperature
the cap-diluter (T-mixer) is also conCrmed from Fig. 8 of 1000◦ C and a carrier gas $ow rate of 1:5 l=min as
showing a TEM micrograph of nanoparticles collected in a function of the quenching gas $ow rate at 200, 600
the gas phase by thermophoretic deposition on a liquid- and 1000 mbar. Decreasing the pressure produces smaller
K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762 1761

Fig. 11. BET-equivalent particle diameter (left ordinate) and speciCc

surface area (right ordinate) of bismuth nanoparticles as a function of
Fig. 10. BET-equivalent diameter of product bismuth nanoparticles as the carrier gas $ow rate for the cap-diluter. Furnace set temperature,
a function of the quenching gas $ow rate for the cap-diluter at three pressure and quenching gas $ow rate are 1000◦ C, 200 mbar and 10 l=min,
process pressures. The carrier gas $ow rate and furnace set temperature are respectively. Higher carrier gas $ow rates produce smaller particles.
1:5 l=min and 1000◦ C. Smaller particles are produced at higher quenching
gas $ow rate and lower pressure.

particles: at 10 l=min quenching gas $ow rate, particles with

a BET-equivalent diameter of 31 nm (speciCc surface area:
20 m2 =g) are produced at 1000 mbar while at 200 mbar
18 nm particles (speciCc surface area: 35 m2 =g) are made.
As the gas density increases with pressure, the gas velocity
in the condenser decreases. As a result, the aerosol experi-
ences longer residence times and bigger primary particles
are formed. The decrease in particle size with decreasing
pressure is consistent with the predictions of Panda and
Pratsinis (1995) and with the measurements of Yatsuya,
Kasukabe, and Uyeda (1973) and Granqvist and Buhrman
Increasing the quenching gas $ow rate from 5 to 20 l=min
leads to an almost linear decrease in particle diameter at all
pressures. At 1000 mbar, the BET-equivalent particle diam-
eter is 37 nm at a quenching gas $ow rate of 5 l=min and
19 nm at 20 l=min. The corresponding speciCc surface ar-
eas are 16 and 33 m2 =g, respectively. The smallest parti-
cles of 12 nm diameter (speciCc surface area: 50 m2 =g) are
produced at a pressure of 200 mbar and a quenching gas
$ow rate of 20 l=min. This is consistent with the design di-
agram of Fig. 2. The prediction of larger particle sizes by
the model may be explained by the assumption of a fully
saturated vapor while only 10% saturation of the carrier gas
was achieved in the experiment. This was determined by
gravimetric analysis which gave production rates of 0:1 g=h Fig. 12. XRD pattern of product bismuth particles synthesized at stan-
(1000 mbar) –0:8 g=h (200 mbar) and is in agreement with dard experimental conditions, 200 mbar and 1000 ppmO2 in Ar showing
the characteristic peaks of Bi (a) and XRD pattern of powder oxidized
Panda and Pratsinis (1995) for the present unit.
in ambient air after synthesis in oxygen-free atmosphere (b). The char-
Fig. 11 shows the in$uence of the carrier gas $ow rate on acteristic peaks of Bi2 O3 are labeled with “O”.
the product bismuth particle size for a furnace set temper-
ature of 1000◦ C and a quenching gas $ow rate of 10 l=min
at 200 mbar. Increasing the carrier gas $ow rate from 0.75 data of Haas et al. (1997) and the model of Tsantilis, Pratsi-
to 3:0 l=min decreases the average primary particle diame- nis, and Haas (1999).
ter from 19 to 12 nm. As the carrier gas $ow rate increases, Fig. 12a shows a typical XRD pattern of product bismuth
the initial monomer concentration decreases, leading to the particles synthesized under the standard experimental con-
formation of smaller particles consistent with experimental ditions and 200 mbar with introduction of trace amounts of
1762 K. Wegner et al. / Chemical Engineering Science 57 (2002) 1753–1762

oxygen (1000 ppm). Only the distinct peaks of bismuth are Fluent Inc. (1998). Fluent 5: User’s Guide, Vol. 2. Fluent Incorporated,
visible indicating that the passivating oxide layer at the bis- Lebanon, NH 03766 (Chapter 9).
muth surface is either too thin to be detected by XRD or is Forney, L. J., NaCa, N., & Vo, H. X. (1996). Optimum jet mixing in a
tubular reactor. A.I.Ch.E. Journal, 42, 3113–3122.
X-ray amorphous. Particles synthesized in an oxygen-free Frurip, D. J., & Bauer, S. H. (1977). Homogeneous nucleation in metal
condenser atmosphere are rapidly oxidized in ambient air as vapors. 3. Temperature dependence of the critical supersaturation ratio
is apparent from the bismuth oxide re$ections highlighted for Iron, Lead, and Bismuth. Journal of Physical Chemistry, 81,
in the corresponding XRD pattern (Fig. 12b). 1001–1006.
Girshick, S. L., & Chiu, C. (1990). Kinetic nucleation theory: A new
expression for the rate of homogeneous nucleation from an ideal
supersaturated vapor Journal of Chemical Physics, 93, 1273–1277.
5. Conclusions Gleiter, H. (1989). Nanocrystalline materials. Progress in Materials
Science, 33, 223–315.
Synthesis of bismuth nanoparticles by condensation of Granqvist, C. G., & Buhrman, R. A. (1976). UltraCne metal particles.
metal vapor was systematically investigated following de- Journal of Applied Physics, 47, 2200–2219.
sign diagrams accounting for aerosol nucleation, condensa- Haas, V., Gleiter, H., & Birringer, R. (1993). Synthesis of nanostructured
materials by the use of a thermophoretic forced $ux system. Scripta
tion and coagulation. Cold $ow visualization, CO2 tracer Metallurgica et Materialia, 28, 721–724.
gas analysis and computational $uid dynamics showed that Haas, V., Birrringer, R., Gleiter, H., & Pratsinis, S. E. (1997). Synthesis
rapid quenching of the Bi-vapor $ow in a T-mixing conCg- of nanostructured powders in an aerosol $ow condenser. Journal of
uration produces particles with narrow size distribution as Aerosol Science, 28, 1443–1453.
observed by TEM. In contrast, counter$ow quenching re- Hultgren, R., Desai, P. D., Hawkins, D. T., Gleiser, M., & Kelley, K. K.
(1973). Selected values of thermodynamic properties of the elements.
sults in early particle formation and broad particle residence Metal Park, Ohio: American Society for Metals.
time distributions producing particles with a wide size dis- Konrad, H., Haubold, T., Birringer, R., & Gleiter, H. (1996).
tribution. Bismuth nanoparticles of average size from 12 to Nanostructured Cu-Bi alloys prepared by co-evaporation in a
37 nm were made, as determined by N2 adsorption, trans- continuous gas $ow. Nanostructured Materials, 7, 605–610.
mission electron microscopy and X-ray di4raction. Increas- Lide, D. R. (1997). CRC handbook of chemistry and physics (78th ed.).
Boca Raton: CRC Press.
ing the carrier or quenching gas $ow rates and decreasing MKuhlbach, J., Recknagel, E., & Sattler, K. (1981). Inert gas condensation
the pressure decreases the product particle size consistent of Sb, Bi and Pb clusters. Surface Science, 106, 188–194.
with particle formation design diagrams. Panda, S., & Pratsinis, S. E. (1995). Modeling the synthesis of aluminum
particles by evaporation-condensation in an aerosol $ow reactor.
Nanostructured Materials, 5, 755–767.
Acknowledgements Pfund, A. H. (1930). Bismuth-black and its applications. Physical Review,
2, 1434.
Russell, L. E. (1967). The sintering of pyrophoric powders. Powder
Financial support by ETH-Gesuch No. 33=99-2 is ac- Metallurgy, 10, 239–263.
knowledged as well as stimulating discussions with Niko- Samy El-Shall, M., & Edelstein, A. S. (1996). Nanomaterials: synthesis,
laas de Jaeger (AGFA-Geavert NV, Belgium). properties and applications. In A. S. Edelstein & R. C. Cammarata
(Eds.), Bristol and Philadelphia: Institute of Physics Publishing, pp.
References Sattler, K., MKuhlbach, J., & Recknagel, E. (1980). Generation of metal
clusters containing from 2 to 500 atoms. Physical Review Letters, 45,
Andres, R. P., Averback, R. S., Brown, W. L., Brus, L. E., Goddard,
Seinfeld, J. H. (1986). Atmospheric chemistry and physics of air pollution
W. A., Kaldor, A., Louie, S. G., Mascovits, M., Peercy, P. S., Riley,
(pp. 334 –337). New York: Wiley.
S. J., Siegel, R. W., Spaepen, F., & Wang, Y. (1989). Research
Siegel, R. W. (1991). Cluster-assembled nanophase materials. Annual
opportunities on clusters and cluster-assembled materials. Journal of
Review of Materials Science, 21, 559–578.
Materials Research, 4, 704–736.
ThKolLen, A. R. (1979). On the formation and interaction of small metal
Andres, R. P., Bielefeld, J. D., Henderson, J. I., Janes, D. B., Kolagunta,
particles. Acta Metallurgica, 27, 1765–1778.
V. R., Kubiak, C. P., Mahoney, W. J., & Osifchin, R. G. (1996).
Tsantilis, S., Pratsinis, S. E., & Haas, V. (1999). Simulation of synthesis
Self-assembly of a two-dimensional superlattice of molecularly linked
of palladium nanoparticles in a jet aerosol $ow condenser. Journal of
metal clusters. Science, 273, 1690–1693.
Aerosol Science, 30, 785–803.
Bejan, A. (1984). Convection heat transfer (p. 203). New York: Wiley.
Uchikoshi, T., Sakka, Y., Yoshitake, M., & Yoshihara, K. (1994). A study
Burger, H. C., & van Cittert, P. H. (1930). Die Herstellung
of the passivating oxide layer on Cne nickel particles. Nanostructured
von Wismut-Antimon-Vakuumthermoelementen durch Verdampfung.
Materials, 4, 199–206.
Zeitschrift Fuer Physik, 66, 210–217.
Uyeda, R. (1991). Studies of UltraCne Particles in Japan: Crystallography.
Chilton, T. H., & Genereaux, R. P. (1930). The mixing of gases for
methods of preparation and technological applications Progress in
reaction. A.I.Ch.E. Transaction, 25, 102–122.
Materials Science, 35, 1–96.
Dahlin, R. S., Su, J., & Peters, L. K. (1981). Aerosol formation in
Yakhot, V., & Orszag, S. A. (1986). Renormalization group analysis of
reacting gases: Theory and application to the anhydrous NH3 -HCl
turbulence. I. Basic theory. Journal of Scienti@c Computing, 1, 3–51.
system A.I.Ch.E. Journal, 27, 404–417.
Yatsuya, S., Kasukabe, S., & Uyeda, R. (1973). Formation of ultraCne
Eickmans, J., Leenders, L., Lamotte, J., Dierksen, K., & Jacobsen, W.
metal particles by gas evaporation technique. Japanese Journal of
(1996). Mastertool: A new dry phototool in the production of printed
Applied Physics, 12, 1675–1684.
circuit boards Circuit World, 22, 26–33.