sevier.com/locate/powtec

Powder Technology 162

Liquid spray vs. gaseous precursor injection — Its influence on the

performance of particle coating by CVD in the fluidized bed

Gregor S. Czok, Joachim Werther *

Hamburg University of Technology, Particle Technology, D-21073 Hamburg, Germany

Received 20 April 2005; received in revised form 13 September 2005

Available online 31 January 2006

Abstract

Aluminum coatings were created onto glass beads by chemical vapor deposition in a fluidized bed reactor at different temperatures. Two

different routes were examined. First, tri-isobutyl-aluminum (TIBA) vapor was enriched in nitrogen and thermally decomposed inside the

fluidized bed to deposit elemental aluminum. On the other hand, liquid injection of TIBA via a two-fluid nozzle directly into the fluidized bed was

tested. To ensure homogeneous coating on the bed material, the fluidizing conditions necessary to avoid agglomeration were investigated for a

broad range of temperatures. Also, different glass types and pretreatments of the substrate surface were investigated to elucidate the influence of

the surface chemistry on the growth and morphology of the layers deposited.

D 2005 Elsevier B.V. All rights reserved.

Keywords: CVD; Precursor; TIBA; Liquid spray injection; Fluidized bed

1. Introduction

For various applications in modern powder technology

coatings help to produce desired solids properties. Most

pigments for metallic color paints consist of a core particle

surrounded by a well-defined layer of a certain refraction

index to give the impression of a certain color, depth and

shine [1]. While wet processing of pigments requires solid–

liquid separation, drying, destruction of agglomerates and

often subsequent tempering of the fine solids, the combina-

tion of chemical vapor deposition (CVD) and fluidized bed

technology facilitates coating without additional process

steps.

The fluidized bed provides incentives for powder coating

due to intense solids mixing, excellent heat and mass transfer

and homogeneous temperatures. Although there are already

industrial applications of chemical vapor deposition in the

fluidized bed reactor (CVD-FBR) for pigment coating [2],

investigations in the open literature are scarce and the

governing mechanisms in the fluidized bed are not yet fully

understood. The technology is further mentioned for the

0032-5910/$ - see front matter D 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.powtec.2005.12.011

* Corresponding author.

E-mail address: Werther@tuhh.de (J. Werther).

production of noble metal catalysts [3,4], where metal organic

precursors are either decomposed thermally or by micro wave

or radio frequency plasma. In the patent literature CVD-FBR

can be found for the production of layered luminescent

pigments [5,6]. Here, one or more very thin but closed

protective coating layers of alumina or SiO2 are applied to

prevent moisture contact and the hereby caused decay of

brightness.

A first pioneering work was carried out by Liu et al. [7],

who investigated the thermal decomposition of tri-isobutyl-

aluminum (TIBA) for the coating of glass beads with high

refractive index. The application – although not named

explicitly – seems to be the production of color pigments

with angle-dependant change in color. Although the tempera-

ture range investigated was almost the same than in the present

work, the focus was on the surface chemistry of the reaction

pathways, the carbon incorporation into the surface layer and

the layer quality for different temperatures and operation times

from a chemist’s point of view. The fluidizing conditions, the

agglomeration tendency and the homogeneity of the coated

material were not further discussed. In the present paper the

fluidizing conditions and solids mixing rates were systemati-

cally studied. Special focus was set on the mode of precursor

injection. Two routes have been investigated; one being the

injection of vaporized precursor in a carrier gas stream, the

(2006) 100 – 110

www.el

Fig. 1. Injection mechanisms of A) gaseous and B) liquid precursor injection.

Fig. 2. Scheme of liquid precursor forming an agglomerate with particles.

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110 101

other being liquid spray injection of TIBA directly into the

fluidized bed.

2. Theory

The transfer of the coating process from classical semicon-

ductor production at clean vacuum conditions to the coating of

powders with a high volume specific surface in the fluidized

bed requires a higher throughput of precursor, since, for

example, 1 g of 60 Am glass beads has a total surface area of

already 3.8 I10�2 m2. Therefore, a high feed rate of precursor is

required to shorten operation times and thus to keep attrition

effects as low as possible. Following classical CVD processes,

the precursor can be introduced into the reaction zone by a

carrier gas. The precursor feed rate is limited by the saturation

capacity of the carrier gas in this case. The enriched carrier gas

is usually kept well below the reaction or decomposition

temperature of the precursor to prevent gas phase reactions.

Inside the fluidized bed, the precursor gets in contact with hot

particles that are entrained into the cooler gas jet (Fig. 1A). The

decomposition reaction will preferably take place at the hot

particle surfaces, as long as the reaction is fast compared to the

mass transport. Thus, gas phase nucleation can be suppressed

largely.

To increase the precursor dosage, liquid spray injection can

be considered. The mechanisms here are displayed in Fig. 1B

for the application with a two-fluid nozzle. Those hot bed

particles, which are entrained in the jet, impinge onto liquid

precursor droplets or ligaments. Although the vaporization of

the liquid precursor is very fast (Leclere et al. [8] found a

vaporization time of 16 ms for a 300 Am FCC feedstock

droplet), according to Bruhns and Werther [9] some neighbor-

ing particles will be sticking to the droplet to form an

agglomerate (Fig. 2). Heat supply by the hot particles of the

agglomerate will cause the precursor to vaporize and since this

occurs locally very close to the particle surface, the gaseous

precursor decomposes on the neighboring particle surface.

Since each agglomerate acts as local source of gaseous

precursor with locally high vapor concentration near the liquid

surface, it can be expected that this mode of precursor supply

will cause a layer structure which differs from the morphology

caused by gaseous precursor injection into the fluidized bed.

Bruhns and Werther [9] have shown, that the agglomerates

formed in the spray jet easily fell apart after evaporization of

the liquid. In the present application it could be necessary to

agitate the bed to such an extent, that the formation of

permanent agglomerates is suppressed. This question is also

subject of the present experimental investigation.

3. Experimental

3.1. Experimental setup

The fluidized bed reactor for the injection of gaseous precursor had a cross-section area of 8�8 mm2, a height of 200 mm and

an enlarged freeboard section of 45 mm inner diameter for gas solids separation (Fig. 3). The fluidized bed reactor was built with

these small dimensions in order to be operated with a solids inventory of 4 g only. This made it possible to simulate the deposition

process in the fluidized bed reactor based on a discrete particle simulation, which is subject of another publication [10].

High-purity nitrogen (>99.99990%) was used as a carrier gas to be saturated with TIBA vapor. The gas flow rate was adjusted

by a mass flow controller (FIC) that was connected to a pressure controller to prevent over pressure (>150 kPa) at the precursor

flask. The precursor bottle was placed into a heated oil bath and used as a bubbler by inserting an immersible pipe. The oil bath was

kept at 82 -C to gain a TIBA temperature of (80T0.5)-C, measured by a Pt100 temperature probe (TI-1). The saturated nitrogen

was transported in pipes, which were heated to a temperature of 85 -C, to the fluidized bed reactor and entered the reactor through afour-hole nozzle. The diameter of the four holes was 0.8 mm each. The geometry of this four-hole nozzle is given in detail in Fig. 4.

Fig. 3. CVD fluidized bed reactor setup for injection of gaseous precursor.

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110102

Several attempts to use a porous plate of diverse materials as a gas distributor failed, because the pores of the distributor were

blocked immediately by deposits.

For the injection of liquid TIBA precursor two different setups were used. In a first set of experiments for bottom spray mode a

reactor with 25 mm inner diameter and a height of 450 mm was equipped with an externally mixing two-fluid nozzle at the bottom

(Fig. 5). The nozzle used was a commercially available one manufactured by Spraying Systems Germany, model number VSS110-

D01. It was mounted in upward oriented position so that the gas flow was used for both, bed fluidization and transporting the liquid

precursor away from the nozzle. The nozzle had an open jet angle of 25- with water in air (Fig. 6). Three different inserts with

orifices of 100, 200 and 300 Am diameter, respectively, were used to modify the precursor flow rate.

In the second part of the investigation an additional setup with side spray mode was used, where the nozzle was mounted in

horizontal orientation on the side wall at a height of 25 mm above the porous plate gas distributor. This latter reactor had an inner

diameter of 54 mm and a height of 450 mm. The fluidizing gas was heated to the desired fluid bed temperature by an electrical gas

pre-heater in case of side spray to minimize temperature gradients within the bed.

For both spray setups the liquid precursor was transported hydraulically from the storage bottle to the nozzle by the pressure of

an adjustable liquid head in a bubble column filled with petroleum ether (max. 20 kPa (gauge)). The temperature of the fluidized

bed was measured at a height of 15 mm (TI-1), 100 mm (TI-2) and 150 mm (TI-3), respectively, above the bottom. The reactor

walls of both reactors were electrically heated to the desired bed temperatures controlled by PID controllers.

3.2. Bed materials

Three different types of glass beads were investigated as substrates for the coating:

& Type A: SwarcoForce (Swarco GmbH, Amstetten, Austria), surface mean diameter dp=64 Am, sodium glass used for

blasting

Fig. 4. Details of four-hole nozzle gas distributor.

Fig. 5. CVD fluidized bed reactor setup for the injection of liquid precursor.

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110 103

& Type B: Spheriglas (Potters Europe, Kirchheimbolanden, Germany), dp=46 Am, high reflex glass containing Ba and Ti, used for

traffic signs and as a filler for cosmetics

& Type C: Plus9beads (Swarco GmbH, Amstetten, Austria), dp=66 Am, high reflex glass used for street marks.

The particle size distributions of these materials are given in Fig. 7 and the minimum fluidization velocities calculated by the

Ergun equation [11] are given in Fig. 8. All glass beads were washed in sodium hydroxide or hydrochloric acid liquor for 30 min

and rinsed with de-ionized water until pH=7 to ensure cleanliness of the substrate surfaces.

3.3. Precursor

As a simple example, the thermal decomposition of tri-isobutyl-aluminum (TIBA) was investigated to produce aluminum

coatings on glass beads. TIBA is a clear, pyrophoric liquid, which decomposes at temperatures above 200 -C:

Above approximately 50 -C TIBA loses one isobutyl ligand to form diisobutyl-aluminumhydride (DIBAH), which is a

hydrogen bridged trimer with a substantially lower vapor pressure of only 1.33 Pa at 40 -C compared to the vapor pressure of TIBA

of 13.3 Pa at 20 -C [12].

TIBA VT>50-C

p>3barDiisobutyl� aluminumhydride DIBAHð Þ þ isobutene

The saturation of the carrier gas with precursor inside the bubbler requires a moderate heating to increase vaporization of TIBA

and heating of all subsequent carrier gas feed lines to prevent condensation of liquid at cold spots. The dissociation reaction in turn

limits the temperature of bubbler and feed lines to avoid the formation of a larger amount of DIBAH, which would condense

immediately. Furthermore, it is thus very difficult to measure TIBA concentrations in the gas phase. The vapor pressure and the

dissociation rate found in Ref. [13] are given in Fig. 9.

3.4. Experimental procedure

In both cases – liquid spray and gaseous precursor injection – the glass beads were filled into the reactor from the top under

constant gas flow through the nozzle gas distributor or the two fluid nozzle, respectively, to prevent blocking of the nozzles by glass

Fig. 6. Open jet angle of the two-fluid nozzle with water in air.

0 20 40 60 80 100 120 140 1600.0

0.2

0.4

0.6

0.8

1.0

Type A (Swarcoforce) Type B (Spheriglas) Type C (Plus9beads)

cum

. mas

s fr

actio

n, -

particle size dp, µm

Fig. 7. Particle size distributions of the glass beads used in the present experiments.

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110104

beads. After closing the reactor, the gas flow rates were adjusted to the operating conditions and the reactor walls were heated to

adjust the desired fluid bed temperature. In case of the gaseous precursor injection the storage bottle was placed into an oil bath

with 82 -C to heat the precursor to 80 -C. The storage bottle was closed and the immersible pipe was not inserted so that the gas

bypassed the bubbler during this phase.

In case of the liquid spray, the immersible pipe was also not inserted, so that a constant nitrogen flow purged the liquid

channel of the two-fluid nozzle preventing glass beads from clogging it. After the fluidized bed had reached the operation

temperature, the storage bottle valve was opened and the immersible pipe was introduced into the precursor liquid. In case of the

gaseous precursor injection, the nitrogen was enriched with TIBA while bubbling through the precursor liquid. The enriched gas

was transported through heated pipes into the reactor. In case of the liquid spray, the pressure of the liquid head of the bubble

column hydraulically transported the liquid precursor towards the two-fluid nozzle, where it was atomized by a nitrogen stream

(see Fig. 5). The gas stream for the spray nozzle was kept constant at a flow rate of 7 l/min (STP) throughout the experiments,

while the gas stream through the porous plate gas distributor was modified to study different intensities of fluidization and thus

solids mixing. During the experiments the temperatures inside the fluidized bed were recorded every 2 s to document the state of

operation. A typical temperature plot for an experiment with liquid spray injection is given in Fig. 10. The sharp decay of the

bed temperature at 14 min is the result of colder bed material which has settled down on the conical walls of the freeboard

sliding back into the bed upon knocking at the reactor wall. To stop the precursor injection after a desired period of time the

immersible pipe was pulled out and the storage bottle valve was closed. The heating was switched off and all nitrogen flows

were kept constant until the fluid bed was cooled down below 60 -C. Then the reactor was opened and the bed material could

be examined visually for flowability and the color on a greyscale.

The layer morphologies produced by both routes on the different glass types, with base or acid pre-treatment and at

temperatures between 225 and 500 -C were investigated by Scanning Electron Microscopy (SEM) analysis. Therefore, the

particles displayed in this article were chosen to be as representative as possible for the sample material of the according

experiment. The deposited mass of aluminum was determined by dissolution in 30% H2O2/HCl at 50 -C and analyzed by optical

200 250 300 350 400 450 500

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

Type C (Plus9beads)

Type B (Spheriglas)

Type A (SwarcoForce)

u mf, 1

0-3m

/s

temperature, °C

Fig. 8. Minimum fluidization velocities at a pressure of 1 bar in nitrogen, calculated with Ergun’s equation [11].

60 80 100 120 1400

2

4

6

8

10

12

0

1

2

3

4

5

6

7

8

vapo

r pr

essu

re, k

Pa

temperature, °C

diss

ocia

tion

rate

, %/m

in

Fig. 9. Vapor pressure and dissociation rate of tri-isobutyl-aluminum (TIBA)

[13].

0 5 10 15 20 25 30300

350

400

450

500

10 minutesliquid spray

injection

knocking onreactor wall

TI-1 TI-2

tem

pera

ture

, °C

time, minutes

Fig. 10. Typical temperature plot of a liquid spray injection experiment

(experimental setup of Fig. 5, TI-1 and TI-2 are two thermo-couples located

in the fluidized bed).

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110 105

emission spectroscopy within an inductively coupled plasma (OES/ICP) in the Central Chemical Laboratory of the Technical

University Hamburg–Harburg.

4. Results and discussion

4.1. Injection of the gaseous precursor

Whereas in classical CVD processes like for very large

scale integration (VLSI) applications the furnace temperatures

are usually 200–300 -C [12] under reduced pressure, we have

investigated the injection of gaseous precursor in the

temperature range of 225–500 -C at ambient pressure for

different fluidizing conditions and pre-treatments. The oper-

ation time for each experiment was 30 min.

4.1.1. Influence of glass type and pre-treatment

The different chemical composition of the three glass types

used was expected to exhibit different layer morphologies of

the aluminum coating. All glass beads shown in Fig. 11 were

washed in 10-N sodium hydroxide liquor for 30 min and then

rinsed with de-ionized water until pH=7 to ensure equal

conditions. It can be seen, that the glass types A and C show

very voluminous deposits in comparison to the type B glass in

the temperature range of 235–250 -C. The Spheriglas (B)

exhibits only small islands of aluminum (white spots in Fig.

11B). The vast majority of the particles do not receive enough

aluminum to cover the surface.

To elucidate the influence of the base pre-treatment, a

further sample of type A glass was washed in 10-N

hydrochloric acid before the coating. The coated particles

are shown in Fig. 12B. No significant difference can be made

out between the two samples, indicating that the pretreatment

with acid or base had much less influence than the chemical

composition of the glass surface itself. These differences in the

morphology and affinity of the deposited aluminum are less

dominant at higher reaction temperatures, as can be seen in

Figs. 13 and 14. This can be explained by the increased

reaction rate at elevated temperatures. Aluminum atoms in the

glass matrix as well as edges and the surface roughness act as

nucleation spots. Due to the autocatalytic character of the

decomposition reaction, aluminum deposits preferably on

spots with such low nucleation energy [12] or onto already

deposited aluminum, leading to voluminous ‘‘cauliflower’’

structures (Fig. 12A and B). Also, there was no visible

.

Fig. 11. Deposits formed at 235–250 -C on the different glass types, A) SwarcoForce, u =8.2 cm/s; B) Spheriglas, u =8.1 cm/s; C) Plus9beads, u =8.3 cm/s; all

NaOH-washed, the white bar indicates the length of 10 Am.

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110106

difference for the acid treated samples of the other two glass

materials.

4.1.2. Influence of reaction temperature

For lower temperatures in the range of 225–250 -C, nocomplete aluminum coating could be achieved. Instead, the

deposited aluminum formed lumps of agglomerates, partially

on the glass surface, but more often in between (Figs. 11 and

12). This can be explained by the autocatalytic character of the

TIBA decomposition reaction. The tendency towards agglom-

eration often led to clogging of the reactor and the superficial

gas velocity needed to be at least 4 cm/s to ensure

homogeneous deposition throughout the bed material for this

inlet concentration of the TIBA precursor in this special

geometry. With gas velocities lower than 3.5 cm/s, there is still

deposition of aluminum at the gas inlet and in the upper part of

the reactor, but in the lower section there is not enough bed

movement to avoid clogging as illustrated in Fig. 15.

For temperatures in the range of 275–360 -C the aluminum

forms a coarse coating that does not fully cover the glass beads

(Fig. 13). Again, the propensity towards agglomeration

requires superficial fluidizing velocities of at least 4 cm/s, i.e.

well above the minimum fluidization velocity (Fig. 8) to ensure

good mixing throughout the bed.

At further increased temperatures in the range of 400–

500 -C the reaction takes place all over the surface resulting in

a complete and uniform coating of small grain size. There was

almost no difference observed between the three types of glass

(Fig. 14). Type C (Plus9beads) shows the most even layer

morphology; the grain size is too small to be seen at this

magnification. On Fig. 14 B there is a scratch on the aluminum

Fig. 12. Deposits formed at 235 -C (u =8.1 cm/s) on glass type A (Spheriglas) (A) N

film providing a slight insight into the film thickness. The

surface diffusivity allows reacting species to migrate to a

growth site along the surface and on the already deposited

aluminum [12]. At higher temperatures of the surface this

diffusivity increases as well as the reaction rate does. Once an

energy threshold is reached, these species decompose on the

spot to deposit elemental aluminum, leading to more flat and

even layers.

4.1.3. Layer thickness

By dissolving the coating with 30% HCl/H2O2 and OES/

ICP analysis the aluminum mass on the glass was determined.

Assuming a plane surface the layer thickness can be calculated

using the known Sauter diameter. The results are shown in

Table 1.

To compare the calculated layer thickness with experimental

results, one sample was cooled down in liquid nitrogen and

crushed into fragments to achieve a brittle fracture of the

aluminum (Fig. 16). It can be seen that the deposit layer is not

uniform and shows the typical equiaxed structure of crystal

growth occurring at conditions T /Tm>0.5 [14], where Tm

denotes the melting temperature. The assumption of a porosity

of the coating therefore had to be considered. The calculated

thickness of 217 nm for a porosity of the coating of 40% is in

good agreement with the observed order of magnitude.

4.1.4. Deposition rate

To achieve a constant dosage of precursor the carrier gas

flow rate through the bubbler operated at 80T0.5 -C was kept

constant at 0.0035 g/s. The deposited mass of aluminum (also

aOH-washed and B) HCl-washed, the white bar indicates the length of 10 Am).

Fig. 14. Deposits formed at 400–500 -C on different glass types, A) SwarcoForce, u =10.8 cm/s; B) Spheriglas, u =12.1 cm/s; C) Plus9beads, u =11.4 cm/s; all

NaOH-washed, the white bar indicates the length of 10 Am.

Fig. 13. Deposits formed at 275–360 -C on different glass types, (A) SwarcoForce, u =10.1 cm/s; B) Spheriglas, u =9.6 cm/s; C) Plus9beads, u =8.8 cm/s; the white

bar indicates the length of 10 Am).

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110 107

Table 1) on the different glass substrates for this constant

nitrogen flow rate through the bubbler within 30 min is

compared with the data of Liu et al. [7] to derive the deposition

kinetics. Different from these latter authors we related the

deposited mass of aluminum to the glass surface area available

for the deposition, assuming the following expression for the

deposition rate:

DmAl

Dt¼ kSlocalCTIBA;local ð1Þ

The calculated overall deposition rates are plotted in the

manner of an Arrhenius plot for the three glass types and the

Fig. 15. A) Agglomerates formed due to excessive growth, B

recalculated data from Liu et al. [7] (Fig. 17), showing one

range with temperature dependence and one range, that is

independent from temperature. The latter one is dominated by

mass transport limitation while the temperature dependence is

caused by the deposition rate depending on the reaction

temperature. The apparent activation energy EA can be

obtained from the slope of the temperature dependent section

of the Arrhenius plot. The glass type B (Spheriglas) shows

deposition rates very close to the data obtained by Liu et al. [7]

for a high reflex glass. But the slope of the temperature

dependent part and the resulting values for EA is a little

different. The other two glass types show a much smaller value

) solid neck C) agglomerated glass beads at the nozzle.

Table 1

Experimental results of gaseous precursor injection obtained with a constant mass flow rate of the carrier gas of 0.0035 g/s and an operation time of 30 min

Glass type Temperature Deposited mass Al per mass glass mg Al/m2 glass Calculated layer thickness Deposition rate r

T [-C] [mg/kg] [nm] [mg/(m2 s)]

Type A SwarcoForce 235 4400 92 33 0.051

250 269 106 2 0.060

275 4800 24 36 0.014

280 652 130 5 0.073

292 5430 138 41 0.078

297 6325 201 48 0.114

316 5400 189 41 0.107

350 10030 317 76 0.179

362 3300 186 25 0.105

392 6500 196 49 0.111

400 19500 399 147 0.226

402 11400 220 86 0.124

450 5900 121 45 0.085

500 17200 352 130 0.200

Type B Spheriglas 253 3236 96 67 0.068

259 3701 110 110 0.079

262 6050 180 159 0.070

268 8721 259 165 0.180

293 9040 269 170 0.171

318 9300 276 209 0.154

350 11450 340 140 0.133

375 7681 228 223 0.156

400 12250 364 140 0.202

Type C Plus9beads 247 751 35 13 0.023

250 2110 99 38 0.024

271 2190 102 39 0.049

275 1890 88 34 0.022

286 2839 133 51 0.083

306 8739 409 157 0.257

358 882 41 16 0.353

380 8094 379 145 0.241

410 16250 760 292 0.423

500 450 400 350 300 250

0.12 0.11 0.10 0.09

E =163.4 kJ/mol

T, °C

u, m/s

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110108

for EA, that could be caused by surface roughness or higher

aluminum content within the glass matrix.

4.2. Injection of the liquid precursor

It has to be stated here, that the experiments for spray

injection of liquid TIBA precursor needed quite some testing

to find out a working setup and feasible operating conditions.

Special care has to be taken to avoid blocking of the nozzle

orifice by solids formed inside the system by TIBA reacting

Fig. 16. Fracture of type A glass coated by injection of gaseous precursor at

500 -C bed temperature (u =12 cm/s).

with air or moisture. A sieve screen of 40 Am mesh size was

placed directly in the precursor feed line in front of the

nozzle block to overcome this problem. Another challenge

was to avoid defluidization of the bed due to the formation of

0.0014 0.0016 0.0018 0.0020

-6

-5

-4

-3

-2

-1

EA=159.3 kJ/mol

EA=87.2 kJ/mol

A

EA=102.1 kJ/mol

SwarcoForce Spheriglas Plus9beads Liu et. al. (2000) [7]

ln r

, 10-6

kg A

l / m

2 s

1/T, K-1

Fig. 17. Deposition rates at different temperatures for constant carrier-gas mass

flow rate of 0.0035 g/s; the resulting superficial velocities in the fluidized bed

are plotted on the additional scale on top of the Fig. For comparison, the results

obtained by Liu et al. [7] have also been plotted in this graph (the velocity scale

is not valid in this latter case).

Fig. 18. A) Liquid spray of 12.5 g, TIBA onto 75 g glass within 15 min (bottom spray, u =18 cm/s, 1.11 I10�2 gTIBA/gglass min) and B) gaseous injection of 0.7 g

TIBA onto 4.8 g glass within 30 min (4.86 I10 �3 gTIBA/gglass min, u =10.1 cm/s). Both type A glass at 330 -C bed temperature.

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110 109

agglomerates; this occurred preferably at the thermocouples

and near the nozzle outlet. In case of the bottom spray the

fluidizing gas was not preheated to prevent clogging of the

nozzle; this led to some temperature gradients within the bed.

Due to the cooling effect of evaporating TIBA the experi-

ments were conducted at bed temperatures above 300 -C to

prevent larger amounts of unreacted TIBA vapor from

condensing in the filter or vent line. Extensive testing

showed that the fluidization velocity created by the nozzle

gas had to be kept above 15 cm/s in order to prevent the

formation of agglomerates with subsequent defluidization of

the bed.

First, the results achieved with the 100 Am orifice insert in

the nozzle (cf. Fig. 5) and a fluidization velocity of 18 cm/s in

bottom spray mode are compared with the morphologies

obtained by injection of gaseous precursor (Fig. 18).

Both deposits were generated at the same fluidized bed

temperature of 330 -C and no significant difference could be

seen between both samples at first look. The surfaces are

completely covered by small grains and larger piles of grains

are visible. These piles might be the effect of either solids

formed in the gas phase or of attrited fines. This fine material

may adhere to the surface by van der Waals forces.

At closer observation of a sample coated by liquid spray

injection in the side spray arrangement, some beads could be

found with a flake-like structure (Fig. 19). Less than about 5%

of the total bed material exhibited these structures, which might

be traces of the formation of agglomerates with liquid TIBA

bridges leading to locally high deposition rates on the particles

in the agglomerate.

Fig. 19. Deposits formed at 320 -C by liquid injection of 7 g TIBA onto

The side spray setup with preheating of the fluidizing gas

was used. The gas flow rate for the two-fluid nozzle was kept

constant at 7.5 l/min (STP) to ensure homogeneous atomization

of the liquid. At fluidization velocities larger than 10 cm/s the

spray rate of 1.8 gTIBA/min did not lead to defluidization or the

formation of agglomerates visible with the eye for an operation

time of 10 min. No significant differences could be observed

between the samples produced with side spray mode and

bottom spray mode with respect to the quality of the coating.

For fluidization velocities lower than 5 cm/s slight agglomer-

ation was observed after 10 min of operation. The agglomer-

ates were few and very fragile, breaking up to powder at the

lightest contact.

Fig. 20 shows a brittle fracture of a very thick deposit layer

generated in side spray mode by spraying 80 g of liquid TIBA

onto 115 g glass beads within 20 min at 340 -C fluidized bed

temperature, using the 300 Am orifice nozzle insert at a

fluidization velocity of 8.6 cm/s (calculated without nozzle

gas). This experiment led to the formation of a thick layer of

agglomerates on the reactor walls and the temperature probes,

while approximately 70% of the bed material was not

agglomerated. The sample used in Fig. 20 was part of the

non-agglomerated bed material. The calculated layer thickness

for this experiment is 583 nm, assuming a planar surface and a

density of 2600 kg/m3 for the aluminum film. This calculation

does not include losses due to attrition of the deposits. Because

of the porosity of the coarse, voluminous deposits formed at

this temperature, the observed thickness in Fig. 20 is of the

order of magnitude of 1.5 Am. One can spot almost identical

structures of the aluminum in both pictures (Figs. 16 and 20),

75 g type A glass beads within app. 5 min (side spray, u =15 cm/s).

Fig. 20. Fracture of type A glass coated by liquid precursor injection of 80 g

TIBA onto 115 g glass at 340 -C (side spray, u =8.6 cm/s).

G.S. Czok, J. Werther / Powder Technology 162 (2006) 100–110110

indicating similar growth conditions at the surfaces in both

experiments.

5. Conclusions

CVD coating of particles in the fluidized bed is feasible and

provides a good means for the treatment of larger quantities of

powders in a cost-efficient way. For the thermal decomposition

of TIBA onto different glass types it could be shown that a

minimum temperature of about 400 -C is required to achieve a

compact and relatively even aluminum layer. At lower

temperatures, the reaction preferably takes place on already

deposited aluminum leading to voluminous deposits.

The injection of gaseous precursor in this geometry requires

a superficial gas velocity of at least 4 cm/s and reaction

temperatures above 400 -C to achieve even layers without

agglomerates.

The results further demonstrate that liquid injection of TIBA

by a two-fluid nozzle is feasible to produce aluminum coatings,

even for glass beads with a Sauter-diameter of only 46 Am. The

tendency towards agglomeration and clogging of the fluidized

bed is much higher for the injection of liquid than for gaseous

precursor. While for the introduction of the gaseous precursor,

the feed rate is limited by the saturation capacity of the carrier

gas, liquid injection allows higher precursor fluxes resulting in

shorter operation times. The reduction of the operating time to

roughly one third reduces the extent of attrition of the deposited

layers while the risk of defluidization due to agglomeration is

increased, which has to be overcome by suitable geometrical

setup and operating conditions.

Nomenclature

cTIBA,local local TIBA concentration, kg/kg

dp particle diameter, m

dp surface mean diameter, AmEA apparent activation energy, J

k reaction rate constant, kg/m2 s

Q3 cumulative mass fraction,�Slocal local particle surface area, m2

T temperature, K

Tm melting temperature, K

u superficial gas velocity, m/s

umf minimum fluidization velocity, m/s

DmAl differential change in aluminum mass, kg

Dt differential change in time, s

Acknowledgements

This work was funded by the German Research Foundation

(Deutsche Forschungsgemeinschaft, DFG) under grant no. WE

935 / 8-1.

References

[1] Merck: Industrial Pigments at: http://www.merck-pigmente.de/servlet/

PB/menu/1015750/index.html, downloaded 21.3.2005.

[2] BASF: Production of Variochrom \ effect pigments with CVD-FBR at:

http://www.corporate.basf.com/en/innovationen/labors/wirk_effektstoffe/

wirk_effekt//variocrom.htm?id=60A2K5zt9bcp0xn#2, downloaded 1.12.

2004.

[3] J.-C. Hierso, R. Feurer, J. Poujardieu, Y. Kihn, P. Kalck, Metal-organic

chemical vapor deposition in a fluidized bed as a versatile method to

prepare layered bimetallic nanoparticles, Journal of Molecular Catalysis.

A, Chemical 135 (1998) 321.

[4] C. Jouanny-Tresy, M. Vardavoulias, M. Jeandin, C.V.D. coating of

alumina reinforcing particles to improve the wear resistance of H.S.S.

based composites, Materials and Manufacturing Processes 9 (1991) 901.

[5] Kenton, Minnesota Mining and Manufacturing Company: Encapsulated

Electroluminescent Phosphor and Method for Making same; European

Patent No. EP 0 455 401 A2 (1991).

[6] Klinedinst et al., GTE Lab. Inc.: Method for Coating Phosphor Using

Aluminum Isopropoxide Precursors and an Isothermal Fluidized Bed;

United States Patent 4,999,219 (1991).

[7] Liu Shuqing, Shi Daxin, Cui Yihong, Cao Zhuyou, Microstructure and

surface topography of aluminum films deposited on glass microspheres in

a fluidized bed CVD reactor, Chemical Journal of Chinese Universities 21

(3) (2000) 339.

[8] K. Leclere, C. Briens, T. Gauthier, J. Bayle, P. Guigon, M. Bergougnou,

Experimental measurement of droplet vaporization kinetics in a fluidized

bed, Chemical Engineering and Processing 43 (2004) 693.

[9] S. Bruhns, J. Werther, An investigation of the mechanism of liquid

injection into fluidized beds, AIChE Journal 51 (2005) 3, 766.

[10] G. Czok, J. Werther, M. Ye, J.A.M. Kuipers, Modelling of Chemical

Vapor Deposition in a Fluidized Bed Reactor by Discrete Particle

Simulation, International Journal of Chemical Reactor Engineering 3

(2006) A57, (http://www.bepress.com/ijcre/vol3/A57).

[11] S. Ergun, Fluid flow through packed columns, Chemical Engineering

Progress 48 (1952) 2–89.

[12] T. Kodas, M. Hampden-Smith (Eds.), The Chemistry of Metal CVD, VCH

Weinheim, ISBN: 3-527-29071-0, 1994, p. 60.

[13] Data taken from MSDS for Triisobutylaluminum at http://www.albermarle.

com, downloaded 1.5.2004.

[14] D.G. Teer, in: E. Lang, et al., (Eds.), Coatings for High Temperature

Applications, Applied Science Publishers, London, 1981, pp. 79–120.