Anda di halaman 1dari 10

Variation of Equivalence Ratio and Element Ratios With

Distance From Burner in Premixed One-Dimensional Flames


CHRISTOPHER J. POPE,** RICHARD A. SHANDROSS,*** and
JACK B. HOWARD*
Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139

Differential diffusion effects in a low-pressure premixed, laminar, one-dimensional benzene/oxygen/argon


flame are shown to lead to changes in the elemental mass fractions. The effect is caused by unequal diffusion
of reactants away from the burner and products towards the burner. Analysis of data previously collected from
this flame shows the local equivalence ratio to vary by as much as 25% from its inlet value, with similar
variations found for the atomic C/O ratio (up to 25%) and atomic C/H ratio (up to 10%). Variations of
comparable magnitude are predicted by detailed kinetic modeling of the same flame using a published
mechanism. The variations predicted for a similar flame but at atmospheric pressure are slightly smaller than
those at low-pressure, which is consistent with the constitutive equations. Similar effects would be expected for
any flame in which the mass average flame velocity is comparable to the diffusion velocities of the major flame
species. The results have implications for consideration of molecular-weight growth and soot formation in such
systems, since these processes are strongly dependent upon the C/O and equivalence ratios. © 1998 by The
Combustion Institute

NOMENCLATURE Mk molecular weight of species k


Mk,m weight of element m in species k
A area expansion ratio
p pressure
Am number of moles of atoms of element
R ideal gas constant
m per mole of gas
t time
Bk mass flux fraction of species k due to
T temperature
diffusion
v mass average velocity
Bk,m mass flux fraction of element m in
Vk diffusion velocity of species k
species k due to diffusion
Wm atomic weight of element m
Ball,m mass flux fraction of element m in all
xk mole fraction of species k
species due to diffusion
Yk mass fraction of species k
Dk1,k2 binary diffusion coefficient for species
Yk,m mass fraction of element m in species
k 1 and k 2
k
Dk,mix mixture diffusion coefficient of
Yall,m mass fraction of element m in all
species k
species
Fk molar flux of species k
z distance from burner
G mass flux fraction
Gk mass flux fraction of species k
Greek symbols
Gk,m mass flux fraction of element m in
species k ak,m number of atoms of element m per
Gall,m mass flux fraction of element m in all molecule of species k
species f equivalence ratio
K total number of species r mass density 5 M # p/RT
m inlet mass flux rate 5 r 0 n 0 5 rn A
#
M mean molecular weight of gas Subscripts
0 cold gas
*Corresponding author. E-mail: jbhoward@mit.edu k species index
**Present address: Laboratoire G.R.E., Université de
Haute-Alsace, 68200 Mulhouse, France. m element index
***Present address: Arthur D. Little, Inc., Acorn Park, all sum over all species
Cambridge, MA 02140.

COMBUSTION AND FLAME 116:605– 614 (1999)


© 1998 by The Combustion Institute 0010-2180/99/$–see front matter
Published by Elsevier Science Inc. PII S0010-2180(98)00062-5
606 C. J. POPE ET AL.

INTRODUCTION radial transport, the relevant processes occur-


ring are chemical reaction and axial convection
In diffusion flames, the elemental composition and diffusion. If diffusion did not occur, each
is customarily described in terms of a spatially section of fluid volume could be described as a
varying equivalence ratio [1] or mixture fraction closed system, i.e., as a batch reactor or a
[2], since the flame environment varies between segregated section of a plug flow reactor. Chem-
the extremes of pure fuel and pure oxidant. ical reactions do not transform elements, so
Concentrations of most species in diffusion reactions within a closed fluid volume do not
flames (except products of molecular-weight change the elemental composition. Convection
growth such as aromatics and soot) have been in an ideal one-dimensional system would have
found to correlate directly with the local equiv- no radial gradients in axial velocity and would
alence ratio [2– 6]. Furthermore, in the flame merely translate fluid without affecting its com-
zone of diffusion flames, fuel and oxidant dif- position. Therefore, it is only through diffusion
fuse in opposite directions; if these diffusion that variations in the elemental composition can
rates are not equal, usually due to differing occur.
diffusion coefficients, the element ratios vary The one-dimensional flame equations [17]
across the reaction zone in different ways. The are used below to show how diffusion leads to
effects of this so-called “differential diffusion” changes in the mass fractions of the individual
on the element ratios have been seen by Bilger elements (C, H, O, and Ar). Species diffusion
[2] from analysis of the data of Tsuji and velocities are used in determining which flame
Yamaoka [7], and by Moss and Roberts [8] from species contribute to the variation in equiva-
analysis of the data of Melvin et al. [9]. lence and element ratios, and the extent to
In premixed flames, it has been assumed that which pressure affects the variation of element
the equivalence ratio and elemental composi- ratios is discussed.
tion are spatially invariant [10, 11]. Although
knowledge of the importance of multicompo-
nent diffusion in premixed flames is widespread, EXPERIMENTAL DATA AND
C/H and C/O ratios are invariably given as COMPUTATIONS
single numbers, rather than ranges, in the liter-
ature. Even in soot formation, where elemental The experimental data are from a near-sooting
composition receives much attention because of benzene/oxygen/30% argon laminar flat flame
its critical importance, the C/O ratio has been at a pressure of 20 Torr (2.67 kPa) [13, 14]. The
assumed to have one single value in premixed cold gas velocity and temperature are 50 cm/sec
flames [12]. The present work shows that for and 298 K. The inlet equivalence ratio, defined
one-dimensional premixed laminar flames, con- as the fuel/oxidant ratio in the cold gas divided
siderable variation of elemental composition by the fuel/oxidant ratio required for stoichio-
with respect to distance from the burner occurs metric conversion to CO2 and H2O, is 1.8. The
as the result of axial diffusion of major flame inlet atomic C/H ratio is unity and the inlet C/O
species. Element ratios and the local equiva- ratio is 0.72.
lence ratio are calculated using species concen- The modeling software used was the Sandia
tration profiles from (1) data from a low-pres- flame code [18] with the CHEMKIN-II subrou-
sure laminar premixed benzene-oxygen-argon tine package [19] and transport properties pro-
flame [13, 14] and subsequent analysis of the gram [20]. The detailed kinetic mechanism of
data [15], (2) model results obtained for the Zhang and McKinnon [16], which had been
same flame using a recent detailed kinetic tested on the same flame and found to repro-
mechanism [16], and (3) model results for an duce successfully most of the major species, was
atmospheric-pressure flame similar to the low- used without modification for all the simula-
pressure flame considered. tions performed in the present work. The exper-
While the variation in the local element ratios imental temperature profile [13, 14] and area
and equivalence ratio is calculated from species expansion ratio [13] were used as input for the
concentrations, it is caused by molecular diffu- simulation. Treatment of diffusion in the mod-
sion. For a truly one-dimensional system, i.e., no eling included the full calculation of multicom-
EQUIVALENCE RATIO IN PREMIXED 1-D FLAMES 607

ponent diffusion coefficients [21] and thermal x kM k,m


diffusion factors for H atom and H2. A second Y k,m 5 # , (2)
M
simulation was performed for a 1 atm flame
with the same cold gas composition and velocity where M k,m is the contribution of element m to
as the 20 Torr case, with the following modifi- the molecular weight of species k,
cations in the input parameters: (1) the inlet
M k,m 5 a k,mW m; (3)
mass flux, m, was increased 38-fold to compen-
sate for a 38-fold increase in pressure and (2) then Eq. 1 may be rewritten as
the temperature profile was spatially com-
#
Y all,mM
pressed by a factor of 38 and the experimental Am 5 (4)
profile was extrapolated in the post-flame zone. Wm
Both modifications are consistent with the as- with Y all,m (the total mass fraction of element
sumption that the reaction rates are linearly
m) defined as
proportional to pressure, which is reasonable if
bimolecular reactions predominate.
Diffusion velocities for mass flux fraction
Y all,m 5 OY
k
k,m. (5)

calculations using model results were obtained


Therefore, the element ratios can be written as
via the FBR flat flame analysis program [22],
which uses subroutines from the Sandia A m1 Y all,m1W m2
PREMIX flame code [18]. Dixon-Lewis mul- 5 . (6)
A m2 Y all,m2W m1
ticomponent diffusion coefficients [21] and
thermal diffusivity were incorporated in these Equation 6 is useful for analyzing the source of
calculations as well. deviations from inlet conditions.
Experimental fluxes were calculated [13, 15] For the flames considered in the present
from the measured mole fractions using the work, the only elements present are carbon,
one-dimensional flame equations described be- hydrogen, oxygen, and argon: m 5 {C, H, O,
low. The mole fractions were smoothed and Ar}. Profiles of the atomic C/O ratio (5 A C/
differentiated numerically [23]. Mixture diffu- A O) and C/H ratio (5 A C/A H) (Figs. 1 and 2,
sion coefficients, computed by Wilke mixing respectively) show considerable variation, with
[24], were used. The main difference between the largest perturbations appearing in the reac-
the calculated fluxes used in the present work tion zone of the flame. For the purposes of
and those from the original calculation of Bitt- comparison with the 20 Torr data and modeling,
ner [13] is the inclusion here of all 54 individual the results for the 1 atm modeling are expanded
species profiles measured, as well as profiles for by a factor of 38 in z (distance from burner).
compounds heavier than 200 amu [15]. The local equivalence ratio (f), as defined by
Mitchell et al. [1], can be expressed as

ELEMENT RATIOS AND LOCAL AH Y all,C Y all,H


2*A C 1 2* 1
EQUIVALENCE RATIOS 2 WC 2W H
f5 5 . (7)
AO Y all,O
The set of species mole fraction profiles and the WO
chemical composition of the species, as de-
scribed by ak,m, are needed for the analysis of Calculated values of f are shown in Fig. 3. The
equivalence and element ratios. The number of calculation of f from the data is identical to that
moles of atoms of element m per mole of gas is performed previously [25].
given by The above analysis shows the modeling re-
sults to be qualitatively consistent with the data

Oa
K
analysis, with the model predicting an even
Am 5 k,m x k. (1) greater perturbation in the C/H ratio and the
k51
local equivalence ratio. Results for the 1 atm
If the mass fraction of element m from species k modeling case are similar to those at 20 Torr,
is defined as with the peak heights being diminished slightly
608 C. J. POPE ET AL.

Fig. 1. Calculated atomic C/O ratio vs. distance from burn- Fig. 3. Calculated local equivalence ratio vs. distance from
er: thick line, data; diamonds, 20 Torr modeling; triangles, 1 burner: thick line, data; diamonds, 20 Torr modeling; trian-
atm modeling; dashed line, inlet value. (For this and all gles, 1 atm modeling; dashed line, inlet value.
subsequent figures, the results for the 1 atm modeling are
expanded by a factor of 38 in distance from burner for determine the source of variation in elemental
comparison with the 20 Torr results. Also, the data for all composition. Recall that the molar flux of spe-
figures are the original profiles drawn by Bittner [13, 14],
which have been treated by numerical smoothing [15, 23].)
cies k is defined by
xk p
Fk 5 @v 1 V k#A (8)
and the effect being slightly broadened (when RT
plotted using the expanded scale). Reasons for and the mass flux fraction of species k is
the differences between the 20 Torr and 1 atm
modeling results are discussed below. M kF k
Gk 5 . (9)
r 0v0
SOURCE OF VARIATION: FLUX ANALYSIS Equation 9 is equivalent to

One-Dimensional Flame Equations and


Variation in Yall,m
Gk 5
x kM k x kM k V k
M# 1 M # v F G
. (10)

The terms on the right-hand side of this equa-


The one-dimensional flame equations of Fris- tion are the convective and diffusive contribu-
trom and Westenberg [17] were analyzed to tions, respectively. In terms of the species mass
fractions, the convective and diffusive contribu-
tions to G k are Y k and Y k(V k/v).
Analysis of the equivalence and element ra-
tios also requires study of the mass flux fractions
of the elements. The portion of the total mass
flux fraction of element m arising from species k
is

G k,m 5
x kM k,m x kM k,m V k
M# 1
M# v F G
. (11)

Using Eq. 2 this can be written as

G k,m 5 Y k,m 1 Y k,m F G


Vk
v
. (12)
Fig. 2. Calculated atomic C/H ratio vs. distance from burn-
er: thick line, data; diamonds, 20 Torr modeling; triangles, 1 Conservation of mass requires that the total
atm modeling. (Inlet value 5 1.) mass flux remain constant:
EQUIVALENCE RATIO IN PREMIXED 1-D FLAMES 609

G5 O G 5 O Y 1 O Y vV 5 1.
k
k
k
k
k
k k
(13)
tween the modeling results and the data analysis
to reinforce the assertion that the variations in
mass fractions are real.
Since the individual atoms retain their identi- One might expect that since argon is inert its
ties, element conservation requires that the mass fraction would remain constant, but Fig.
total mass flux fraction of element m (G all,m) 4d shows that assumption to be incorrect. Given
also remain constant: the requirement that G all,Ar is constant, the only

G all,m 5 OG k,m 5 OY k,m


way Y Ar can remain constant is for V Ar/v to
equal 0, which implies that V Ar 5 0. For this to
k k be true, x Ar must be a constant. However, if x Ar
# must also be
O Y vV
and Y Ar are both constant, M
k,m k
1 5 constant. (14) constant, since Y Ar 5 x Ar M Ar /M # . This is un-
k likely, because the mean molecular weight, be-
Since ¥ Y k 5 1, Eq. 13 shows that ¥ (Y kV k/v) is ing the ratio of the total mass to the total
constrained to be zero throughout the flame. In number of moles, can only remain constant if
the elemental balance, the only constraint which the number of moles remains constant. In such
applies to the diffusive contribution to G all,m, a case, the flow would either be non-reacting or
¥ Y k,mV k/v (5 Ball,m), is that it approach zero as one in which the sum of all chemical reactions,
z 3 `, however there is no reason to assume at all points in the flame, produces no net mole
that B all,m will be equal to zero in the reaction change. Such systems might exist, but since
and early post-flame zones (see below). There- flames are driven by free-radical chain branch-
ing reactions, M# can be expected to vary widely.
fore, ¥ Y k,m, which is equal to Y all,m, need not
remain constant to satisfy element conservation. Even when the stoichiometry of the overall
As mentioned above, diffusion is the driving reaction implies no net mole change, for exam-
force of elemental composition variations. Vari- ple, Cx Hy 1 [ x 1 ( y/4)] O2 5 x CO2 1 y/ 2
ations in Y all,m due to B all,m Þ 0 are the source H2O, with y 5 4, there would still be a net
of the changes in the local equivalence ratios change of moles in the reaction zone. There-
and element ratios (cf. Eqs. 6 and 7). fore, even the mass fraction of an inert species
can change in a flame environment, due to
Variation of Measured and Computed diffusion which arises when chemical reactions
Element Mass Fractions cause a change in the total number of moles.

Variations of total element mass fractions, Element Flux Fractions and Species
Y all,m, with distance from the burner are shown Contributions
for carbon, hydrogen, oxygen, and argon in Fig.
4a– d, respectively. The variations are clearly The normalized element flux fraction due to
significant in comparison with the accuracy with diffusion, B all,m/G all,m, is a measure of the mag-
which the species mass balances are being com- nitude and direction of the effect of diffusion on
puted, i.e., four significant figures. Again, the equivalence and element ratio variations. Un-
results from the modeling are comparable to fortunately, experimental G all,m values vary
those obtained from the data, with the excep- from their inlet value by as much as 30% [13],
tion of the Ar mass fraction. Since the experi- primarily because they involve differentiation of
mental argon mole fraction is computed by measured data gathered in regions of high
difference [13], it is affected by the errors in the gradients. Model calculations do not suffer from
concentrations of all other flame species. The this type of error, though there are small devi-
data close to the burner surface ( z , ;0.3 cm) ations in the individual element fluxes due to
are also more prone to experimental uncer- numerical errors. For the 20 torr modeling,
tainty due to probe-burner interactions and flow G all,m deviates by less than ;2% for z , 1 cm,
nonidealities. However, the variations in ele- and by no more than 3.4% for z , 2 cm; for the
ment and equivalence ratios persist well beyond 1 atm modeling, the maximum deviation in
0.3 cm, and there is sufficient agreement be- G all,m is 5%. Therefore, B all,m/G all,m,0 from
610 C. J. POPE ET AL.

Fig. 4. Mass fraction Y all,m vs. distance from burner—thick line, data; diamonds, 20 Torr modeling; triangles, 1 atm modeling;
dashed line, inlet value: (a) carbon, (b) hydrogen, (c) oxygen, and (d) argon.

modeling results are accurate enough to reveal element flux contributions (B k,m/G all,m,0) for
important trends. carbon-containing species. (Acetylene’s behav-
The normalized diffusive flux fractions for 20 ior is explained by the fact that it is an interme-
torr modeling results are shown in Fig. 5. The diate species, reaching its maximum mole frac-
normalized flux fraction is negative for carbon tion at z 5 0.9 cm, so it diffuses both towards
and hydrogen, indicating a net flux due to and away from the burner.)
diffusion towards the burner for these elements, A similar argument holds for hydrogen-con-
while there is a net flux due to diffusion of both taining species: benzene would diffuse much
oxygen and argon away from the burner. An more slowly away from the burner than the
intuitive explanation can be found in the fact major hydrogen-containing products (H2O,
that the fuel benzene is heavier than any of the C2H2, H2, and H) would diffuse upstream.
major carbon-containing products (CO, CO2, Figure 7 shows the normalized element flux
and C2H2) and would diffuse more slowly away fraction contributions from diffusion for hydro-
from the burner than the products would diffuse gen-containing species. From Fig. 5, the magni-
toward the burner, due to the differences in tude of the normalized diffusive oxygen flux is
diffusion coefficients. This would lead to a net considerably less than those of carbon and
diffusive flux toward the burner, which would hydrogen, but it still indicates the net effect of
require a larger convective flux—implying a O2 diffusing away from the burner faster than
larger mass fraction of carbon—to conserve CO, H2O, and CO2 are diffusing toward the
total carbon flux. To quantify the above asser- burner (Fig. 8).
tions, Figure 6 shows the normalized diffusive The species contributing most to the diffusive
EQUIVALENCE RATIO IN PREMIXED 1-D FLAMES 611

Fig. 5. Normalized element flux fraction due to diffusion


Fig. 6. Normalized diffusive element flux fraction contribu-
(B all,m/G all,m,0) vs. distance from burner for the 20 Torr
tions (B k,m/G all,m,0) with m 5 carbon vs. distance from
modeling results: squares, carbon; diamonds, hydrogen;
burner for the 20 Torr modeling results: squares, C6H6;
triangles, oxygen; inverted triangles, argon.
diamonds, CO; triangles, C2H2; inverted triangles, CO2;
thick line, sum of all other species.
flux are those with the largest mole fractions.
All the species for which individual curves are
element ratios, can be seen by a detailed con-
drawn in Figs. 6 – 8 have peak mole fractions
sideration of the terms Y k,m and B k,m. As noted
greater than 0.01, except H atom, which has a
by Fristrom and Westenberg [17], in a flame
peak mole fraction of only 0.007 but has a large
dominated by bimolecular reactions the solu-
diffusion velocity. The reactants (C6H6 and O2)
tion of the flame equations is invariant with
and the major stable products (CO, CO2, H2O,
respect to pressure so long as mass flow rate
H2, and C2H2) contribute most strongly to the
scales linearly with pressure and distance from
variation of element ratios; the only intermedi-
the burner scales inversely with pressure. That
ates playing a significant role are C2H2 (also a
is, for a given cold gas velocity and composition,
product) and H, with the C2H2 diffusion veloc-
nearly identical sets of mole fraction and tem-
ity going from negative to positive within the
perature profiles are seen at all pressures where
reaction zone. Figures 6 – 8 show a concrete
bimolecular reactions predominate, except that
situation in which the diffusive fluxes for a given
the profiles are relatively compressed or ex-
element do not cancel each other: although
panded. Because all the x k would vary similarly
species containing carbon, hydrogen, and/or ox-
with pressure and their magnitudes would re-
ygen are diffusing in both directions, there is a
main unchanged, the mean molecular weight
net diffusive flux of these elements (B all,m Þ 0)
should behave similarly. Therefore, Eqs. 11 and
because of the differences in diffusion coeffi-
14 show that the pressure dependence of the
cients and mole fraction gradients, leading to
magnitude of B all,m/G all,m, which governs varia-
variations in the Y all,m.
tions in element and equivalence ratios, would
depend upon the pressure dependence of the
PRESSURE DEPENDENCE OF ratio V k/v. Since v 5 r0v0/rA, keeping the mass
EQUIVALENCE AND ELEMENT RATIO flow rate (5 r 0 v0) proportional to the pressure
VARIATIONS means that the factor p 0 /p implicit in this equa-
tion becomes p/p, and the velocity profile
The effect of pressure on the magnitude of the should also expand (or contract) with the dis-
diffusive flux, and thus the equivalence and tance-from-burner scale.
612 C. J. POPE ET AL.

Fig. 7. Normalized diffusive element flux fraction contribu- Fig. 8. Normalized diffusive element flux fraction contribu-
tions (B k,m/G all,m,0) with m 5 hydrogen vs. distance from tions (B k,m/G all,m,0) with m 5 oxygen vs. distance from
burner for the 20 Torr modeling results: squares, C6H6; burner for the 20 Torr modeling results: squares, O2;
diamonds, H2O; triangles, C2H2; inverted triangles, H2; diamonds, CO; triangles, H2O; inverted triangles, CO2;
circles, H; thick line, sum of all other species. thick line, sum of all other species.

Next, consider the diffusion velocity V k( z). the extent of reaction is the mean molecular
Whether the diffusion coefficients (D k,mix) are weight M # . From Fig. 9, the slower decrease in M
#
calculated by Wilke mixing [24], as in the data for the 1 atm case indicates a relatively slower
flux analysis, or by the more rigorous method of overall rate of reaction. Examination of the
Dixon-Lewis [21], which was used in the mod- individual species profiles supports this notion.
eling and the FBR flat flame analysis program Figures 10 and 11 show the mass fractions of O2
[22], the diffusion velocity is of the form and CO, respectively, as examples of the slower
consumption of the initial reactants and slower
d@ln x k~ z!#
V k~ z! 5 D k,mix~ z! . (15) production of the major products. A likely rea-
dz son for the slower-than-expected rate of
In both cases D k,mix is calculated from the set of progress is the importance of termolecular rad-
mole fractions and the binary diffusion coeffi- ical recombination reactions, whose depen-
cients. The mole fraction profiles, as argued dence on pressure is one power higher than
above, would retain the same relative shapes, bimolecular reactions. For example, the pre-
but the z-scale and the D k,mix both scale as p 21 , dicted peak mole fraction of H atom at 1 atm is
so the diffusion velocities should also be invari- only 31% of that predicted for 20 Torr. With a
ant. Therefore, the ratio V k/v, and consequently decreased radical pool, all the flame reactions
the variation in equivalence and element ratios, would slow down. Therefore, the proposed scal-
should be unaffected by pressure. ing of chemical composition is not quite correct,
According to the modeling results, the varia- explaining the slightly decreased variation of
tion of element ratios is slightly less at atmo- element ratios for the 1 atm modeling results.
spheric pressure than at 20 Torr. The difference Note that the lack of an experimental tempera-
may be due to the failure of the variation of the ture profile for the 1-atm case introduces some
chemical composition (i.e., the rate of reaction) uncertainty in any comparison of the modeling
to scale as suggested above. A global measure of results for the two cases.
EQUIVALENCE RATIO IN PREMIXED 1-D FLAMES 613

Fig. 9. Mean molecular weight (M # ) vs. distance from


Fig. 11. CO mass fraction vs. distance from burner: thick
burner: thick line, data; diamonds, 20 Torr modeling; trian- line, data; diamonds, 20 Torr modeling; triangles, 1 atm
gles, 1 atm modeling; dashed line, inlet value. modeling.

CONCLUSIONS AND DISCUSSION flames in which the diffusion velocities of the


flame species are of comparable magnitude to
The elemental composition of laminar pre- or greater than the flame velocity, similar effects
mixed flat flames can vary widely (up to ;25%) can be expected. While the variation of element
due to axial diffusion of the initial reactants, the ratios was first noticed [25] by analyzing data
major products, and key intermediates such as from a low-pressure flame, modeling results and
H and C2H2. While diffusion is occurring in scaling arguments show that similar behavior is
both directions in such a way as to conserve expected to occur in atmospheric-pressure flat
both the total mass flux and the total element flames.
fluxes, the net flux fractions due to diffusion for These results are not only of theoretical in-
the individual elements can be nonzero, leading terest but are also pertinent to flame chemistry,
to changes in the elemental mass fractions and for example, in molecular weight growth leading
thus variations in the elemental ratios and the to the formation of polycyclic aromatic hydro-
local equivalence ratio. The key parameter de- carbons (PAH) and soot, since the equivalence
termining the magnitude of the effect of diffu- ratio is a key parameter in characterizing a
sion on the elemental composition is the V k/v flame’s tendency to form soot. In the 20 Torr
factor multiplying the Y k,m. Therefore, for benzene flame considered here, the C/O ratio
and N are noticeably above their inlet values for
z up to ;1.0 cm, which is also the region where
the observed PAH (up to 202 amu) are formed
and destroyed [13, 14]. The increased C/O ratio
in the reaction zone could be contributing to the
formation of PAH or key intermediates in mo-
lecular weight growth chemistry. As an example
of the strong effect of the equivalence ratio on
molecular-weight growth processes, Bittner [13,
14] found that by varying the inlet equivalence
ratio from 1.8 to 2.0 the flame went from
non-sooting to sooting, and the concentration of
compounds heavier than 700 amu increased
roughly 100-fold.
Fig. 10. O2 mass fraction vs. distance from burner: thick
There is no simple way to determine, a priori,
line, data; diamonds, 20 Torr modeling; triangles, 1 atm the direction of the variation of the element
modeling. ratios in a given flame. Conceivably, there are
614 C. J. POPE ET AL.

flames in which the observed trends for one or 4. Mitchell, R. E., Sarofim, A. F., and Clomburg, L. A.,
more of the elements is the opposite of that Combust. Flame 37:227–244 (1980).
5. Smyth, K. C., Miller, J. H., Dorfman, R. C., Mallard,
seen in the present work. While the total flux of W. G., and Santoro, R. J., Combust. Flame 62:157–181
a given element remains constant, the magni- (1985).
tude and direction of the total diffusive flux of 6. Sivathanu, Y. R., and Faeth, G. M., Combust. Flame
an element is a composite of the diffusive fluxes 82:211–230 (1990).
of all the species containing that element. What 7. Tsuji, H., and Yamaoka, I., Thirteenth Symposium
(International) on Combustion, The Combustion Insti-
the present work has demonstrated is that, in tute, Pittsburgh, 1971, pp. 723–731.
the general case, such a diffusive flux is not zero 8. Moss, J. B., and Roberts, P. T., Combust. Flame
throughout the flame. Therefore, it is not safe to 34:335–338 (1979).
assume that the element ratios in a laminar flat 9. Melvin, A., Moss, J. B., and Clarke, J. F., Combust. Sci.
flame remain the same as their inlet values. Technol. 4:17–30 (1971).
10. Gaydon, A. G., and Wolfhard, H. G., Flames: Their
While it has long been known [2, 8] that structure, radiation and temperature (4th Ed.), Chap-
differential diffusion leads to unequal variation man and Hall, London, 1979, p. 39.
of element ratios in diffusion flames, the results 11. Liñán, A., and Williams, F. A., Fundamental Aspects of
of the present work provide additional insights Combustion, Oxford University Press, New York, 1993,
on the interpretation of diffusion flame data. In p. 62.
12. Wagner, H. Gg., in Soot Formation in Combustion,
the analysis of Bilger [2] the carbon mass frac-
Bockhorn, H., Ed., Springer-Verlag, 1994, p. 163.
tion (as expressed by the mixture fraction) 13. Bittner, J. D., Sc.D. thesis, Massachusetts Institute of
exhibited no apparent changes as might be Technology, 1981.
caused by differential diffusion. Given the 14. Bittner, J. D., and Howard, J. B., Eighteenth Sympo-
present finding that the mass fractions of all the sium (International) on Combustion, The Combustion
Institute, Pittsburgh, 1981, pp. 1105–1116.
elements in the flame studied were found to
15. Pope, C. J., M. S. thesis, Massachusetts Institute of
vary, we conjecture that the behavior of the Technology, Cambridge, MA, 1988.
carbon mass fraction in Bilger’s study is specific 16. Zhang, H.-Y., and McKinnon, J. T., Combust. Sci.
to the flame considered [7]. Accordingly, any Tech. 107:261–300 (1995).
generalization that the carbon mass fraction in 17. Fristrom, R. M., and Westenberg, A. A., Flame Struc-
ture, McGraw-Hill, New York, 1965.
diffusion flames is somehow unaffected by dif-
18. Kee, R. J., Grcar, J. F., Smooke, M. D., and Miller,
ferential diffusion may deserve reconsideration J. A., A Fortran Program for Modeling Steady Laminar
in light of the present work. One-Dimensional Premixed Flames, Sandia National
Laboratories Report No. SAND85-8240, 1985.
We are grateful to the Division of Chemical 19. Kee, R. J., Rupley, F. M., and Miller, J. A., Chemkin-
II: A Fortran Chemical Kinetics Package for the Analysis
Sciences, Office of Basic Energy Sciences, Office of Gas-Phase Chemical Kinetics, Sandia National Lab-
of Energy Research, U.S. Department of Energy, oratories Report No. SAND89-8009, 1989.
for financial support under grant number DE- 20. Kee, R. J., Dixon-Lewis, G., Warnatz, J., Coltrin,
FG02-84ER13282; and to the National Energy M. E., and Miller, J. A., A Fortran Computer Code
Research Supercomputer Center for CPU time. Package for the Evaluation of Gas-Phase Multicompo-
nent Transport Properties, Sandia National Laborato-
The authors would also like to thank Prof. J. T. ries Report No. SAND86-8246, 1986.
McKinnon and H.-Y. Zhang of the Colorado 21. Dixon-Lewis, G., Proc. R. Soc. A 307:111–135 (1968).
School of Mines for providing us with copies of 22. Shandross, R. A., Ph.D. thesis, Massachusetts Institute
their kinetic mechanism and associated thermo- of Technology, Cambridge, MA, 1995.
chemical and transport parameters. 23. Savitsky, A., and Golay, M. J. E., Anal. Chem. 36:1627–
1639 (1964). See also corrections by Steinier, J., Ter-
monia, Y., and Deltour, J., Anal. Chem. 44:1906 –1909
REFERENCES (1972).
24. Wilke, C. R., Chem. Eng. Prog. 46:95 (1950).
1. Mitchell, R. E., Sarofim, A. F., and Clomburg, L. A., 25. Pope, C. J., Ph.D. thesis, Massachusetts Institute of
Combust. Flame 37:201–206 (1980). Technology, Cambridge, MA, 1993.
2. Bilger, R. W., Combust. Flame 30:277–284 (1977).
3. Moore, J. G., and Moore, J., Sixteenth Symposium
(International) on Combustion, The Combustion Insti- Received August 18, 1997; revised March 31, 1998; accepted
tute, Pittsburgh, 1976, pp. 1123–1132. April 15, 1998

Anda mungkin juga menyukai