Sunteți pe pagina 1din 9

This article was downloaded by: [USC University of Southern California]

On: 09 May 2012, At: 18:48


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Combustion Science and Technology


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/gcst20

III. Coalescence as a Rate-Controlling Process


a

G. D. ULRICH & N. S. SUBRAMANIAN

Department of Chemical Engineering, University of New Hampshire, Durham, New


Hampshire, 03824
Available online: 07 May 2007

To cite this article: G. D. ULRICH & N. S. SUBRAMANIAN (1977): III. Coalescence as a Rate-Controlling Process, Combustion
Science and Technology, 17:3-4, 119-126
To link to this article: http://dx.doi.org/10.1080/00102207708946822

PLEASE SCROLL DOWN FOR ARTICLE


Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.

Gordon and Breach Science Publishers Ltd., 1977

Combustion Science and Technology


1977, Vol. 17, pp, 119-126

Printed in Great Britain

Particle Growth in Flames


III. Coalescence as a Rate-Controlling Process
G. D. ULRICH and N. S. SUBRAMANIAN Department of Chemical Engineering. Universitv ot
New Hampshire. Durham. New Hampshire 03824

Downloaded by [USC University of Southern California] at 18:48 09 May 2012

(Received June 4, 1976; in final form JUlie 17, 1977)


Abstract-Oxide and carbon particles in flames are observed to grow by inter-particle Brownian collision and
fusion. Previous investigators have tacitly assumed that coalescence is instantaneous and that growth is controlled solely by the collision frequency. Growing particles, however, are often found to be flocculated,
suggesting that fusion rather than collision is the rate controlling process. In this study, a growth rate expression
is derived which considers both collision and coalescence phenomena. A new parameter is introduced, N. the
number of primary particles within a floc. Predicted values of Np were compared with those estimated from
electron micrographs of silica aggregates. Comparison of predicted with measured values of N is inconclusive.
Better data, in situ, for aggregates in flames are needed.

aggregates found in soots and fine-particle oxides


were presumed to develop during the quenching
process at the conclusion of the growth period. It
was thought that these particles collided and fused,
but because of increasing viscosity due to tempering
or cooling, did not coalesce. Large particle oxides
or particles in dilute flames, because of reduced
collision frequency, were viewed as passing through
the quench transition without collision and thus
without flocculation. Clusters or floes were observed
by Wersborg et al. (1973) in a soot-forming flame
similar to that used by Bonne. More recently,
rapidly-quenched samples collected in a silicayielding flame were found to be flocculated at all
stages of growth (Milnes, 1975; Subramanian,
1974). Thus, small, growing silica particles apparently are not spherical droplets as was assumed
originally in deriving the collision-coalescence
growth expression. Rather, the colliding particles
appear to be irregular, low-density floes composed
of small primary particles which are in the process
of coalescing. To develop an accurate growth
expression for this model, a term describing the
rate of particle fusion must be included as well as a
modified equation for collision frequency.

BACKGROUND
An earlier study (Ulrich et al., 1976) describes the
collision-coalescence growth model and its application to submicron silica particles in flames.
Predicted growth rates were based on the frequency
of interparticle collisions due to Brownian motion
in the free molecule aerosol regime (Knudsen
numbers greater than ten). Experimentally measured growth rates agreed with theoretical predictions at residence times ranging from 40 to 100
milliseconds if the number of successful collisions
is one in 250. Deviations, however, at shorter and
longer residence times were substantial. Possible
causes of the discrepancy are a variable sticking
coefficient, electrical effects, and slow coalescence.
The effect of slow coalescence is considered in this
discussion.
The assumption of instantaneous fusion has been
tacitly applied by previous investigators who employ
the collision-coalescence model to describe particle
disappearance rates in flames (Fenimore and Jones,
1969; Howard et al., 1973; Ulrich, 1971). Such is
valid if particles have sufficient time to coalesce
between collisions. The assumption is consistent
with the observations of Bonne et al. (1965) and
Homann (1967) who noted single, discrete, spherical
particles in the electron micrographs of small,
rapidly-quenched soot samples. Larger metal
oxide particles from industrial and commercial
combustion sources are frequently observed as
discrete spheres (fly-ash from coal, for example).
In an earlier analysis (Ulrich, 1971), the floes or

THEORY AND DISCUSSION


If N represents the concentration of particle floes
and N p is defined as the number of so-called
primary or individually identifiable units comprising the floc, the total concentration of primary
119

120

G. D. ULRICH AND N. S. SUBRAMANIAN

particles NT is equal to the product


(I)

Assuming particles fuse under the influence of


surface tension and via viscous flow, the time for
complete fusion is given approximately by Frenkel
(1945) as

The rate of disappearance of primary particles is


given by the negative derivative of this expression
with respect to time

Downloaded by [USC University of Southern California] at 18:48 09 May 2012

(2)

The time derivative - N is the rate of disappearance of aggregates because of flocculation caused
by Brownian collisions. The term Np accounts
for the increase in number of primary particles per
aggregate through flocculation as well as a decrease
because of coalescence. This can be expressed as

- Np = (Np/N)N+2Np a/(a+2).,.

(3)

where the first term on the right is the rate of


increase due to flocculation. The second term, the
rate of coalescence, is based on the model of a
single primary particle in contact with several
other discrete primary particles creating a cell within the floc. To develop the fusion rate, each of the
peripheral particles was assumed to coalesce with
the central particle in the cell. The number of
peripheral particles or points of contact is designated by a whereas represents the time required
for fusion. Thus, at the onset of fusion, the cell
consists of a+ I primary particles which, after the
period of time r-, becomes a single primary particle.
Combination of Eqs. (2) and (3) results in cancellation of the collision-dependent terms, yielding
an expression for N7' that is solely a function of the
fusion rate
(4)

This result can be derived intuitively. That is,


assuming floes are large relative to the cells, the
concentration of primary particles would clearly be
dependent only on the rate of fusion and not on
the rate of aggregate collisions. At the other
extreme, if N p is constant and equation to one,
then, from Eq. (2)
(5)

which becomes the collision-coalescence growth


rate developed earlier (Ulrich, 1971).
It is instructive to consider the case where growth
occurs according to Eq. (4), i.e. where N p is large.

(6)

where I-' is the viscosity and a the surface tension of


the medium. Substitution of Eq. (6) into (4) yields

Rp can be related to NT by use of the material


balance equation.
(8)

where A is the Avogadro number, Co is the molecular concentration of condensed matter in the
system, M is its molecular weight and p, its density.
If Eq. (8) is used to eliminate R p from (7) which is
then integrated, the following expression results

NT- 1/ 3 = (2[47TpA/3CoMJI/ 3 a a: 0/3[a+ 2]1-')


+(NTO)-1/3

(9)

where 0 is time measured from the onset of coalescence and NTo is the primary particle concentration at the onset.
Since NT is difficult to determine experimentally,
it is preferable to employ specific surface area. For
a non-porous sphere the surface area is
SA = 3/pR p

(10)

Combination of Eqs. (10), (8), and (9) yields the


desired relationship:

SA-l

(2paaO/9[a+2]1-')
+ (MCOp2/367TA NT)l/3

(II)

Equation (II) is similar in some respects to that


describing growth by collision and coalescence of
single particles (Eq. (10), Ulrich, 1971). For
example, at small values of 0, the second term on
the right dominates and the surface area is constant. At large values of 0, on the other hand, the
first term dominates and the surface area becomes
independent of the initial particle concentration.
However, in this case, the surface area is also
independent of Co at larger values of 0 where it

Downloaded by [USC University of Southern California] at 18:48 09 May 2012

PARTICLE GROWTH IN FLAMES

<> -

10- 3

'0

121

,o~

COALESCENCE TIME, Sec.

FIGURE 1 Theoretical relationship between specific surface area and coalescence time for a silica-synthesis
flame at 2000 "K, The growth rate is as described by Eq. (12) for coalescence as the controlling process.

becomes controlled solely by surface tension,


density, viscosity, and time.
At this point, it is appropriate to examine
Eq. (II) for a specific particle-growth system.
For silica at flame conditions described previously
(Ulrich et al., 1976), it becomes

[The value of a was assumed to be 4 based on the


electron micrographs of our samples. The surface
tension value of 307 dyne/ern and that of viscosity,
106 poise at 2000 OK, were taken from Kingery
(1960). Other property data are those used
previously for silica synthesis flames.]
This relationship is illustrated graphically in
Figure 1 where the surface area is shown as a
function of () for various values of ColNT O, the
number of molecules per initial particle. The
curves, particularly for longer residence times, are
especially interesting for the silica system. First,
growth is independent of flame silica concentration.
Secondly, because of its unusually large viscosity,
the predicted growth of this oxide is controlled
primarily by fusion. Thus, aggregates become
larger and larger through collision while fusion
lags. Finally, at prolonged residence times, fusion
continues and the collision frequency diminishes
to the point where single particles are viable.
Fusion control is consistent with the unrealistically
small sticking coefficient required to fit silica
growth data by the simple Brownian collision

expression (Ulrich, et al. 1976). Titania, on the


other hand, having a viscosity more than a factor
of 105 less than that of silica, is observed to grow
faster, with smaller aggregates, and at a rate consistent with highly efficient single-particle Brownian
collisions (George et al., 1973).
Figure 1 is based on an ideal model of uniform,
spherical particles. The effect of these approximations, however, is undoubtedly much less
significant than two other factors: First, it was
assumed that the surface tension and viscosity of
silica are independent of particle size. Although
this may be true for larger particles, it is certainly
unlikely for the smallest embryos, Andreas (1965).
Secondly, the viscosity is such a strong function of
temperature that a real system, with heat losses, will
coalesce at slower rates than are predicted using
the viscosity at the flame temperature.
It is helpful at this point to compare data from
Ulrich et al. (1976) with the curves of Figure I.
This is illustrated in Figure 2. (For coalescencecontrolled growth, the coalescence time is equal to
the residence time.) The actual growth rate is
clearly slower than that predicted by the coalescence model. However, extrapolation of the
results to shorter residence times indicates early
growth at a rate exceeding the coalescence rate. It
is conceivable, then, that the very small particles
grow by single particle collisions and almost
instantaneous coalescence until they become large
enough to exhibit macroscopic viscosity and surface tension. Hereafter, growth becomes coalesc-

Downloaded by [USC University of Southern California] at 18:48 09 May 2012

122

G. D. ULRICH AND N. S. SUBRAMANIAN

'O",L_:;;.~_~~~L~:;_4,.-'-~~'~''IL~_-=3~~---"
RESIDENCE TIME; Sec.

FIGURE 2 Comparison of experimental surface area results with those predicted by the coalescence
model. Rates for the collision (instantaneous coalescence) model are shown as dotted lines.

ence controlled. The theoretical curve would then


become a composite of the two limiting conditions.
This is illustrated in Figure 3 where the curve for
collision growth (c = 0.5) is combined with that
for coalescence, assuming the transition occurs at
a primary particle size equivalent to 1000 molecules
(CO/N7'o = 1000). (In Figure 3, the residence time
is the sum of residence time in the collisioncontrolled phase plus 8, the coalescence time.)
The composite model is consistent with several
observations, in both soot- and oxide-forming

flames. For example, even though single particles


were not found in our experiments, Homann (1967)
noted their existence early in a soot-producing
flame followed by the formation of aggregates at
longer residence times. Also, as commonly
observed in both oxide and soot-forming systems,
large single particles are produced when growth
times are prolonged. In these flames, coalescence
presumably continues while collision frequency
diminishes. The growth rate then again becomes
collision controlled. This corresponds to the curves

~--.r>TTl--'---'-""T'TIlI T'-~~~,,=-~i i i 11'1


-r-r-rr-r-r-r-rr-r-r-ta

Q8I8
o 0.8 mole % Si02

o 0.9 mole % Si0 2

1_~._~'_""~~_~~__
IOS

10. 4

I_I_l_LLl.l.ll_ _

10- 3

RESIDENCE TIME; Sec.

10- 2

l- .~ ~"': -:,-"

FIGURE 3 The composite theoretical growth curve, controlled by collision rate at short residence
times and coalescence rate at longer residence times. The sticking coefficient is 0.5 and the transition
particle size is 1000 molecules, both selected arbitrarily. Data from Figure 2 are shown for comparison.

Downloaded by [USC University of Southern California] at 18:48 09 May 2012

PARTICLE GROWTH IN FLAMES

obtained by extrapolating the initial straight


portions of the two parallel lines in Figure 3
beyond the points where they intersect the coalescence curve. Finally, anomalous sticking coefficients greater than unity such as those of Wersborg et al. (1973) are conceivable if the collision
rate expression were used to fit data in the regime
of coalescence control. This cannot fully explain
their results, however, since clusters were apparently counted rather than single particles.
The theoretical curve of Figure 3 can be modified
to fit the data by adjusting the value of the viscosity
as it presumably changed because of a drop in
temperature of the post-combustion gases. Though
tenuous, this approach permits the comparison of
theoretical and experimental results for one other
parameter, N p , the number of particles per aggregate. Assuming CO/NTo = 1000, viscosity values
were computed from Eq. (II) for residence times
from 10 to 140 milliseconds. These are listed in
Table I. The temperature of bulk silica corresponding to the viscosity is also shown. Unfortunately,
the true temperature profile in the flame is not
available, but the data in Table 1 are plausible.
The population of primary particles per aggregate, N p , is calculated from both the rates of
collision and of fusion. This is expressed through a
rearrangement of Eq. (2) as follows:

123

mechanism proposed above, during early growth,


N p is unity, NT is equal to Nand Np is zero. Nor
NT is accordingly the conventional coagulation
rate. If aggregates exist, the third term of (13) can
be expressed by Eq. (7). Assuming uniform floes
and free-molecule Brownian behavior, the rate of
inter-floc collisions, N, is as expressed by Eq. (4)
in Ulrich (1971):

N = -(c/2) (167TkT/m)1/ 2(2R)2 N2

(14)

but the values of m and R must be modified somewhat because the clusters are non-spherical and
porous.
Based on the analysis of electron micrographs
of numerous different soot samples and additional
computer-simulated floes, Medalia (1967) and
Medalia and Heckman (1969) define a "radiusequivalent elipse" which has an area approximating
the collision cross-section of the floc. This area is
given by
(15)

where B is a bulkiness factor to account for loose


packing of particles in the aggregate. If the ellipse
is characterized as a circle of the same area,
A = 7TR2 and R can be expressed as:

(13)

(16)

The first term on the right accounts for the increase


in N p resulting from interaggregate collisions
whereas the second term corresponds to the
decrease in N p through coalescence. For the

The aggregate mass, based on the mass of a primary


particle is

and Eq. (14) becomes

TABLE I
Calculated values of viscosity determined from
applying Eq. (II) experimental surface area data from
a silica-forming flame.
Flame
residence
time (10- 3 sec)

10
20
40
60
80
100
120
140

Specific
surface
area (m 2 /g )

380
340
285
250
220
195
175
160

(17)

Calculated
viscosity
(10 poise)

Temperatures
(OK)

9.9
15.1
21.2
25.2
27.3
28.5
29.3
30.0

1805
1785
1770
1760
1758
1755
1754
1753

Data from Kingery (1960) for bulk silica.

N=

4c(3kT/P)1/2 R p l/2 N po.37 B N2

(18)

Substitution of (18) and (7) into (13) yields

Np

= 4c(3kT/p)1/2 R p l/2 N p1.37 BN - 2Np

xaa/(a+2)/-,R p

(19)

The value of N can be expressed in terms of R p


and s, through Eqs. (I) and (8) so that Eq. (19)
becomes

Np = (3cBCoM/7TpA)(3kT/p)1/2 Rp-S/2
X

Np O.37-2Np aa/(a+2)/-'

s,

(20)

G. D. ULRICH AND N. S. SUBRAMANIAN

124

"\'(

r tY'

'j
,

Downloaded by [USC University of Southern California] at 18:48 09 May 2012

I ~l

'lit'
~

7-

?..~

.~;i; . .

t,:~";
I

i
1R"'NSMISSIO~

0""

l00.ootl J(
lllC'flOH MICFlOQR.. ,"

$URfAC( .orA'

'o.s~ I

100.001) J(

UlANSMlSS10N ELEClRO'" MICROQA"'W

WUAC .utA H, ...1IG

jr,3 t12 f C

(a.)

l b)

(
I ' ', ..' , \
100,001I )C

I'

O.5f/!

, I

1/lO,DOOX
TA.4NUlllIS10'" HEClROH IIlICROCIU,,"

TPlANSMISlIOH IELECTAON MICROGRA'H

su'r"cr A~[A

(e)

SURI/.C[ APU al ,.,1,(;

20l Mt,tC

(d)

FIGURE 4 Electron micrographs of silica samples collected at various residence times: (a) 8, (b) 13,
(c) B6, and (d) 137 milliseconds.

Downloaded by [USC University of Southern California] at 18:48 09 May 2012

PARTICLE GROWTH IN FLAMES

To integrate Eq. (20), R p was evaluated from the


experimental surface area data at each value of t.
Also, it was necessary to select a boundary value
for N p .
Though limited in accuracy, N can be measured
from electron micrographs. Subramanian (1974)
employed the techniques described by Medalia and
Heckman (1969) to prepare electron micrographs
of four silica samples having residence times of
8, 13, 86, and 137 milliseconds and produced at a
silica mole fraction of 0.009. The micrographs are
reproduced in Figure 4. Size distributions, aggregate populations, EM surface areas and N p were
evaluated using a Zeiss counter. Primary particle
size distributions were similar to those observed by
Bonne et al. (1965) and Wersborg et al. (1973).
Values of EM surface area and N p are shown in
Table 2. As expected, the less dependable EM
surface areas are lower than nitrogen adsorption
values. Values of N p , from EM photographs, on
the other hand, though somewhat uncertain for
larger aggregates, are considered more reliable.
Equation (20) was integrated numerically to
provide theoretical values of N p Bulkiness factors
were found by Medalia (1967) and Medalia and
Heckman (1969) to range from 1.0 to 1.4 for a large
number of actual and simulated carbon black floes,
Seventy-five percent of the values, however, were
between 1.1 and 1.3. The value of B used here was
1.2. Other constants were those employed previously. Values of viscosity from Table 1 and R p
from Eq. (10) were based on experimental specific
surface area data. The boundary values of N p and
t were selected as N p = 25 at t = 10 milliseconds
based on the first two values in Table 2. The
predicted relationship between N p and residence
time is shown by solid curves in Figure 5. Each
curve corresponds to a different value of the
sticking coefficient as shown. The circles represent

125

RESIDENCE TIME; Sec.

FlGURE 5 Comparison of predicted values for N p (the


solid lines) and experimental values taken from electron
micrographs.

data taken from electron micrographs. The predicted N p increases gradually until the fusion rate
becomes greater than the collision rate, causing it
to decrease toward unity.
Agreement with the experimental values is poor.
This may be caused by a defect in the model or the
limited accuracy of N p determined from electron
micrographs of a dispersed bulk sample. In any
event, N p must ultimately become unity in a fusing
cluster. Thus, the decrease in N p at longer fusion
times is a physical reality even though not evident
from the data available for these silica samples.
More precise values of N p obtained over a broader
time span are necessary to provide a rigorous test
of the theory.
CONCLUSIONS

TABLE 2
Specific surface areas and aggregate populations
determined from electron micrographs of four silica
samples (silica concentration: 0.9 mole %)
Flame

Specific surface area (m 2Jg)

residence

time (10- 3 sec)

Nitrogen adsorption

E.M.

Np

8
13
86
137

358
349
203
151

207
165
130
108

33
27
44
77

Until now, theoretical coagulation rates for particles in flames have been based on the collision
frequency. That aggregates exist in a growing
system suggests coalescence is a controlling factor
as well. Thus, the population of primary particles
in an aggregate, N p , is a function of both the collision rate and the fusion rate. The theoretical
relationship presented above predicts an increase
in N p when the Brownian collision frequency is
large followed by a decrease as particles become
larger. Quantitative agreement between measured

126

G. D. ULRICH AND N. S. SUBRAMANIAN

values of N p for silica and those predicted is poor.


More extensive experiments are in progress
employing a broader range of concentrations, in situ
methods of measuring N p , and oxides having different physical properties to provide a more rigorous
test of the theory.

Downloaded by [USC University of Southern California] at 18:48 09 May 2012

ACKNOWLEDGEMENTS
Financial support of this work by the National Science
Foundation (Grant number G K-37465), Cabot Corporation, The Public Service Company of New Hampshire,
and New England Electric System is gratefully acknowledged. Advice and analytical assistance provided by
scientists at the Cabot Coroporation Technical Center are
also appreciated. N. S. Subramanian is presently at
Dacron Research Laboratory, E. I. du Pont de Nemours,
Kinston, N.C.

REFERENCES
Andres, R. P. (1965). Homogeneous nucleation from the
vapor phase, Ind. Eng. Chem., 57, 17.
Bonne, U., Homann, K. H., and Wagner, G.Gg. (1965).
Carbon formation in premixed flames. Tenth Symposium (lnternattonat) on Combustion, The Combustion
Institute, pp. 503-512.
Fenimore, C. P., and Jones, G. W. (1969). Coagulation of
soot to smoke in hydrocarbon flames. Combustion
anti Flame, 13, 303.
Frenkel, J. (1945). Viscous flow of crystalline bodies
under the action of surface tension, J. Physics (Moscow), 9, 385-391.

George, R. P., Murley, R. D., and Place, E. R. (1973).


Formation of Ti02 aerosol from the combustion
supported reaction of TiCI. and 02. Faraday Symposia of the Chemical Society, No.7, 63-71.
Homann, K. H. (1967). Carbon formation in premixed
flames. Combustion and Flame, ll, 265-286.
Howard, J. B., Wersborg, B. L., and Williams, G. C. (1973).
Coagulation of carbon particles in premixed flames.
Faraday Symposia of the Chemical Society, No.7,
109-119.
Kingery, W. D. (1960). Introduction to Ceramics, Wiley,
p.574.
Medalia, A. I. (1967). Morphology of aggregates-I.
Calculation on shape and bulkiness factors; application to computer-simulated random floes. J.
Colloid and Interfacial Science, 24, 393-404.
Medalia, A. I., and Heckman, F. A. (1969). Morphology
of aggregates-II. Size and shape factors of carbon
black aggregates from electron microscopy. Carbon,
7. 567-582.
Milnes, B. A. (1975). An experimental study of particle
growth in oxide-forming flames.
M.S. Thesis,
Department of Chemical Engineering, University of
New Hampshire.
Subramanian, N. S. (1974). Growth of oxide particles in
flames: A theoretical and experimental study of
silica-forming flames. Ph.D. Thesis, Department of
Chemical Engineering, University of New Hampshire.
Ulrich, G. D. (1971). Theory of particle formation and
growth in oxide synthesis flames. Combustion Science
and Technology, 4, 47-57.
Ulrich, G. D., Milnes, B. A. and Subramanian, N. S.
(1976). Particle growth in flames-II. Experimental
results for silica particles. Combustion Science and
Technology, 14, 243.
Wersborg, B. L., Howard, J. B., and Williams, G. C. (1973).
Physical mechanisms in carbon formation in flames.
Fourteenth Symposium (International) on Combustion,
The Combustion Institute, pp, 929-940.

S-ar putea să vă placă și