The E#ect of Metal Salts on Premixed

Hydrocarbon-Air Flames
W. A. ROSSER, Jr, S. H. INAMI and H. WISE
Stanford Research Institute, Menlo Park, Calif.

(Received November 1961)

A study has been made of the inhibition of premixed flames by finely divided metal salts,
some of which weye found to be effective inhibitors of hydrocarbon combustion. The mechanism
of inhibition by powders probably involves this succession of steps: (a) heating of the powder
particles by the hot game gases, (b) evaporation of the powder particles, (c) decomposition
OY reaction of the evaporated material to provide metal atoms, and (d) inhibition of the
combustion process by the metal atoms.

Introduction acetone in a small laboratory ball mill. After

IT IS known that finely divided salts such as milling, the powder was first oven-dried, then
sodium monohydrogen carbonate (NaHCO,) pulverized by brief ball milling and finally
can be used to extinguish fires or to prevent sieved through a 250 or a 325 mesh stainless
ignition of combustible gaseous mixtures. Var- steel screen. Two methods of further size separa-
ious studies1 of combustion inhibition have tion were tried, sieving and gas elutriation.
revealed that powdered materials vary widely Neither technique was satisfactory, consequently
in their ability to inhibit combustion, that most of the test powders were used as obtained
alkali metal salts are generally effective inhibi- from the ball mill.
tors, and that the effectiveness of at least some
powders is proportional to the specific surface The dispersal of powder in a combzlstible gas
area of the materials. The available data do not A satisfactory dispersing device was assembled
permit identification of the significant processes for the introduction of powders into combust-
which result in inhibition. The present study of ible gas mixtures. The powder generator (Figure
combustion inhibition by finely divided powders 1) consisted of a Pyrex cylinder (6 cm i.d.)
was undertaken in order to elucidate the mech- fitted at each end with large rubber stoppers.
anism of flame inhibition by powders. The generator was provided with five jets
through which gas could enter the system, a
Experimental Details and Results four-bladed propeller to agitate the powder bed,
Materials and an exit tube for the gas-powder mixture.
Various gases were used as obtained from com- The propeller was driven by a 1500 rev/min
mercial sources. These included C.P. methane, electric motor. Four of the gas jets were made
propane, ammonia, argon, industrial grade from glass tubing and had an opening of about
oxygen, and commercial dry nitrogen. Com- 1 mm in diameter. The fifth jet, used only to
pressed laboratory air was dried prior to use by obtain high powder concentrations, was made
passage through a bed of granular calcium from brass tubing with a hemispherical cap
chloride. through which five holes 0.015 in. in diameter
All the powders used were C.P.-grade anhy- had been drilled. One of these holes was on the
drous materials. Some were of sufficient fineness axis of the tube, the other four were arranged
to be used as received; most, however, required symmetrically around it.
some size reduction. These were wet milled with In order to introduce the powder into the gas
107
10s W. A. Rosser, Jr, S. H. Inami and H. Wise Vol. 7

was inserted a short distance into the generator.

p--Powder
-----__ -collector
_--__ _____~
The gas-powder mixture from the generator
flowed vertically upward through the tube and
was vented to the atmosphere at the top. After
a steady flow of gas-powder mixture had been
established, the powder output was measured
by collecting on a glass filter paper* the powder
content of the gases. A collection time of 30 or
60 seconds was usually sufficient to give a
weighable amount of powder on the filter. The
,--Electrodes
weight concentration of powder in the gas was
calculated from the collection time, the weight
of collected powder, and the flow rate of the
gas stream. Immediately after a powder sample
had been collected, the tlow of gas was stopped,
the damper valve closed, the side vent opened
to the atmosphere, the top of tube capped with
a rubber stopper, and the mixture ignited by
spark. The time required for the flame to travel
Lt- 1500 revlmin motor between two fixed points on the tube was meas-
ured by phototubes. The phototube signals were
Figure 1. Schematic diagram of pouder generator
and flame tube displayed on an oscilloscope face and photo-
graphed by Polaroid camera. The average pro-
mixture a steady flow of premixed fuel and pagation velocity of the flame was calculated
oxidizer of known composition from a conven- from the known distance between the photo-
tional gas-handling system was split by means tubes and the measured time of transit.
of needle valves into two streams; one passed In the absence of powder the flames were
directly to the vertical output tube of the pow- approximately hemispherical in shape: after the
der generator, and the other entered the powder first few inches of travel, they moved through
generator through the glass jets. The concen- the tube at an apparently uniform rate. Slight
tration of powder in the final mixture could be tilting of the flame was sometimes observed. The
changed by varying the relative amounts of the presence of powder resulted at times in irregu-
two gas streams up to the point when all of lar flame shapes. Because the propagation velo-
the gaseous mixture passed through the genera- city v is related to flame speed S by v = SA,I A,
tor. If even higher powder concentrations were (where A, is the area of the flame and A,, the
desired, it was again necessary to split the gas cross-sectional area of the tube), these perturba-
flow, this time into a fraction which entered the tions of flame shape are a source of error. In
generator through the four glass jets and the some cases the flames were very long and highly
remainder through the metal jet. In this manner tilted, probably as a result of non-uniform pow-
it was possible to obtain powder concentrations der distribution in the tube.
up to about ten per cent by weight of the gas
flow. Measurement of particle size and of specific
surface area
Measurement of the effect of powder on Specific surface area of the experimental pow-
combustion ders was measured by an optical method3.
The effectiveness of the powder was determined About 20 mg of powder was thoroughly dis-
from the change in flame propagation velocity persed in about 300 ml of acetone contained in
as measured in the apparatus shown in Figure an optical cell and the optical density of the
1. A long glass cylinder about 2 cm in diametezr *Mine Safety Appliance Co.
June 1963 The effect of metal salts on premixed hydrocarbon-air flames 109

mixture measured. The total specific area of the The data in Table 1 refer to samples collected
powder, A*, was calculated from the approxi- at the exit of the flame tube. Comparison of
mate relation those data with surface data for the powder
initially present in the generator revealed little
A*= -4ln&/2cL . . . . [l]
or no difference for fine powders (A* zz 10000
where & is the initial optical transmission, c is cm2 / g). For coarse powders the material collec-
the powder concentration in g/cm3, and L is ted at the burner tube was finer than the original
the optical path length (5 cm in this case). powder charge. It appears that the generator
also functioned to some extent as an elutriation
When size analysis was desired, the optical
device. To avoid error from such elutriation
density was recorded for a period of time suffi-
only a few per cent of the initial powder charge
cient for settling of the powder suspension. The
in the generator was used in a related series of
distribution of particle sizes can be derived from
measurements.
the chart record of -In e versus time3v4.
The specific surface data listed ‘in Table 1
were obtained in the manner described. The The effect of powders on premixed Fames
Several of the powders tested were not effective
average particle size, 2, is defined by equa-
inhibitors of premixed methane-air combustion.
tion 2
About 5 wt% of calcium fluoride, talc, calcium
d=61p~* .... carbonate or calcium hydroxide was required to
[21
reduce by ten per cent the speed at which a
where p is the density of the powder material. stoichiometric methane-air flame propagated
One gramme of powder containing only particles through a tube. Heat and momentum loss by
of diameter 2 would have the same surface area the flame are sufficient to account for that
as one gramme of the actual powder. Size degree of reduction. On the other hand, the re-
analysis revealed that 2 is only a crude measure mainder of the powders tested, sodium fluoride,
of powder fineness. All powders listed in Table sodium chloride, sodium bromide, cuprous
1, even the coarse ones, contained a substantial chloride, potassium sulphate, sodium mono-
number of particles with diameters of only 2 or hydrogen carbonate, potassium monohydrogen
3 microns. For example, the weight distribution carbonate and sodium carbonate were effective
of sodium fluoride particles passed through a inhibitors of methane-air combustion to a
degree which indicates chemical interference
maximum at about 3.5~ although l=6.5~.
with the combustion process.
Table I. Specific area of powders used

Powder A* (~m~/g) 2 c/J’)

Talc 12 600 2.4
E;L432 9 000 2.9
2 000 11.0
NaBr3 2 000 9.0
CaF2 3 500
NaF 3 300 ;:;
CuCl 3 000
NaCl 4 500 i:;
&SO4 5 200
NaHC03 7 760 ;:;
NaHC03 11 900 2.3
KHC03 12 400
Na?COz 4 940 f:i
Na&Os 10 800 2.2

c7=6/,,Ak where N+03, mg/cm3

a denotes average particle diameter.
p is the density of material, and Figure 2. Stoichiometric methane-air: effect of sodium
Aj, is the specific surface area. carbonate on propagation velocity
110 W. A. Rosser, Jr, S. H. Inami and H. Wise Vol. 7

As shown in Figure 2, relatively little sodium However, a size effect is clearly evident. In
carbonate is required to produce a significant view of the fact that sodium monohydrogen
reduction in the propagation velocity of a stoi- carbonate decomposes to sodium carbonate at
chiometric methane-air flame. These following relatively low temperature, it is probable that
features should be noted: (a) the weight con- the observed effectiveness of sodium mono-
centration of powder when expressed in mg/cm3 hydrogen carbonate merely reflects the effect-
is very nearly the same as the weight fraction; iveness of sodium carbonate. The data shown
(6) the scatter in the data is primarily due to in Figures 2 and 3 are consistent with this inter-
perturbations of flame shape; (c) the ratio of pretation. The effect of potassium monohydro-
flame area to tube cross section, apart from gen carbonate, A* = 12 400 cm21 g, on propaga-
erratic perturbations, increases from about two tion velocity is shown in Figure 4. Comparison
at zero inhibitor concentration to about four at
concentrations corresponding to points on the
nearly horizontal portion of the curve; (d)
the powder concentration corresponding to the
intersection of the dotted straight lines may be
used as a measure of inhibitor effectiveness. The
lower this critical concentration, the more effec-
tive the inhibitor. For fine sodium carbonate
this intercept concentration is about 0.6 wt%
whereas for the coarser sodium carbonate the
intercept concentration is about 1.1 per cent.
Comparison of these values with the specific
surface area of the two powders indicates that
0 0.01 0.02 0 3
powder effectiveness as defined above is approx-
KHC03, mglcm3
imately proportional to specific surface area.
Figure 4. Stoichiometric methane-air: effect of potas-
sium monohydrogen carbonate on propagation
“70 ” velocity
Z
-Z with the data for sodium monohydrogen car-
”
0 A211600 cm*/g bonate, A*= 11 900 cm2/g, reveals that potas-
d=2,3tJ sium monohydrogen carbonate powder is twice
.Z
9 ??A*=7760 cm*/g as effective as the sodium powder. It is note-
s c/ ~3.5~ worthy that this potassium powder has been
i 30 found to be twice as effective as sodium mono-
G50ii hydrogen carbonate as a fire extinguishers.
s \
\ Because both sodium and potassium mono-
8 I \
hydrogen carbonates decompose readily to the
a- 10 :\--“I’\-r, I .
004 006 respective carbonates, the effective inhibitor in
0 0.02
Na HCOs , mglc m3 both cases is probably the alkali carbonate.
The effectiveness of sodium monohydrogen car-
Figure3. Stoichiometric methane-air: effect of sodium
monohydrogen carbonate on propagation velocity bonate for non-stoichiometric methane-air
mixtures (methane from 8 to 12 per cent by
As shown in Figure 3, similar results were volume) was comparable with or slightly greater
obtained using sodium monohydrogen carbon- than that for the stoichiometric flame.
ate. Because of flame perturbations for low con- The effect of sodium bromide, sodium chlor-
centrations of sodium monohydrogen carbonate, ide, sodium fluoride, cuprous chloride and
the relation between powder effectiveness and potassium sulphate on stoichiometric methane-
A* could not be determined with precision. air flames was also measured. Precise results
June 1963 The effect of metal salts on premixed hydrocarbon-air flames 111

were obtained for sodium chloride and cuprous phate (see Figure 8). Separate experiments
chloride (Figures 5 and 6). Qualitatively sodium revealed that Cab-0-Sil itself is not an effective
bromide, sodium fluoride and potassium sul- inhibitor of hydrocarbon combustion.
Flame inhibition by dispersed powder is affec-
ted by oxidizer concentration. A stoichiometric
methane-oxygen flame diluted with nitrogen so
that 0, / (0, + NJ =0*25 required 1.7 wt% of

70 / I

-n-- With 0.66% CH3Cl

- Without CH3Cl

I I
0.01 0.02
NaCl , mg/cm3
Figure 5. Stoichiometric methane-air: efiect of sodium
chloride OTZ
propagation velocity I I
0.01 0.02 0.03
phate were also effective inhibitors, but because
Na2C03,mg/cm3
of flame instability precise data could not be
obtained. This instability was apparently asso- Figure 7. Stoichiometric methane-air: effect of
ciated with poor dispersibility of the powder. methyl chloride on effectiveness of sodium
carbonate
The dispersibility of pure potassium sulphate
powder could be greatly improved by the addi- fine sodium carbonate to reduce the linear speed
from 142 cm/set to the intercept point which in
I I I I I I
this case occurs at about 20 cm/ sec. The in-
crease in powder requirement reflects primarily
A*= 3000 cm219 the higher initial propagation velocity. Support
d = 5.7 p for this view is provided by another experiment

CuCl,mg/cm3
Figure 6. Stoichiometric methane-air: effect of
cuprous chloride on propagation velocity

tion of a few tenths of a per cent by weight of

Cab-0-Sil*, an extremely fine silica powder 10. I I
0 0.01 0.02 0.03
available commercially. Precise results were
K,SO, , mg/cm3
obtained with the dispersible potassium sul-
Figure 8. Stoichiometric methane-air: effect of
*Godfrey L. Cabot, Inc.. Boston, Mass. potassium sulphate on propagation velocity
112 W. A. Rosser, Jr, S. H. Inami and H. Wise Vol. 7

using a methane-oxygen-argon mixture with an zones: a preheat zone in which the temperature
initial propagation velocity very nearly the same of the unburnt gas rises exponentially with
as that for a stoichiometric methane-air flame, distance, a reaction zone in which the tempera-
65 cm/set. The effectiveness of potassium mono- ture rises linearly with distance, and a post-
hydrogen carbonate in the two cases was much combustion zone of uniform temperature. The
the same, about 0.6 wt% powder. In contrast, assumed temperature profile has been termin-
the presence of about one per cent by volume of ated at a definite position x=S, because where
methyl chloride in a stoichiometric methane-air
mixture considerably reduced the effectiveness T,:2200°K
of sodium carbonate although the methyl chlor- Ti= 1n1t1.31 temperature of
unburnt mixture
ide itself has very little effect on the initial pro-
To= an ‘ignition’
pagation velocity of the mixture (see Figure 7). temperature
The effectiveness of sodium carbonate (or T,,:adiabatic flame
temperature
sodium monohydrogen carbonate) was found to
vary with the nature of the fuel. About 2.5 wt%
of sodium carbonate, A* = 10 800 cm2/g, was
required to reduce the propagation velocity of a
stoichiometric propane-air flame from 80 cm/ _-
set to 20 cm jsec, compared to about 1 wt% for Negative X- 0- Positive X
a stoichiometric methane-oxygen-nitrogen mix- Figure 9. Assumed temperature profile for a
ture of the same initial propagation velocity. stoichiometric methane-air flame
The non-carbonaceous system ammonia-oxy-
gen-nitrogen was found to be only slightly strong chemical inhibition occurs it must occur
affected by powdered sodium monohydrogen where the reaction rate is high, that is, in or
carbonate. For a near-stoichiometric composition near the visible flame. The reactions which take
with an oxidizer concentration 0, / (0, + N,) = place in the post-flame gases do so gradually
0.4,2 wt% of sodium monohydrogen carbonate, over a distance large compared with the thick-
A* = 11980 cm”/g, only reduced propagation ness of the visible flame. Particles in the region
velocity from 62 cm/set to 52 cm/ sec. x > 6 may affect these processes and will event-
ually reach thermodynamic equilibrium with
The temperature history of small particles in a the hot combustion gases. In this analysis, how-
premixed flame ever, the region x > 6 will not be considered.
In addition, because the observed sensitivity of
During passage through a flame the individual
flame speed to the presence of some powders is
particles of a dispersed powder will be heated
far greater than can be accounted for by heat
and may partially evaporate. The heat trans-
loss alone, the effect of the powder on T,, the
ferred from the flame gases to the particles may
adiabatic flame temperature, has been neglected.
not represent a significant thermal drain on the
In the preheat zone the reaction rate is taken
flame, but the evaporated material may affect
to be zero. Consequently, the temperature pro-
the combustion reactions. Two problems must
be considered: the degree of heating of small file has the form6
particles by the flame, and the extent of evap-
oration resulting from this heating. The two . . . . [3]
problems are coupled and in their entirety are
extremely difficult to solve. However, with some where pi denotes the initial gas density, T, is
simplifying assumptions it is possible to obtain the gas temperature at a point x, Ti is the initial
solutions which are adequate for our purposes. gas temperature, T, is an ignition temperature,
The actual temperature profile through a pre- c is the average heat capacity of the gas, h is
mixed flame is replaced by the profile shown in the average thermal conductivity of the gas,
FigzGre 9, which has been divided into three and S is the flame speed.
 June 1963 The effect of metal salts on premixed hydrocarbon-air flames 113

The preheat zone ends at the point x=0, and the relatively low gas velocities used. In
where T, = T, . For mathematical convenience these semi-quantitative calculations the effect of
T, is taken as the average of Tj and the adia- evaporation on heat transfer is not taken into
batic flame temperature T,; the quantities c and account.
h were evaluated at T,. Equation 5 may be converted to a form in-
The gas temperature in the reaction zone in- volving distance rather than time by means of
creases linearly from T, to T, and can be repre- the relation
sented by the expression
dT/dt=??dT/dx . . . . [6]
(T,-T,)/(T,-Ti)=x/6 . +. m[4]
where 1, is the average velocity of the gas-
where 6, the thickness of the reaction zone, Powder mixture. The integral of equation 6 for
equals A/ CpiS* values of x < 0 is given by
In both the preheat .and the reaction zones
T-Ti @Sex/s
the time rate of temperature increase of non- =p . . . . [7]
evaporating spherical particles has been repre- T,-7-t 1+@8
where
sented by equation 5.
Q = 1211 z,c,p,# . . . [S]
psc,V8 d$ =Zndh (T, - T) . . * * ]51 Based on the conservation of mass equation the
average flow velocity of the gas-powder mix-
where I’, denotes the volume of the particle, d is ture, YE;;is approximately equal to 4s for near-
the diameter of the particle, T, is the gas tem- stoichiometric methane-air flames. For values
perature, T is the particle temperature, c, is the of x > 0, the integral of equation 6 is given by
specific heat of the particle material, and ps is
exp[-@3(x/6)]+@8(x/8)-1
the density of the particle. T=T,+(T,-Ti)
The most important assumptions implied by a,6

1
equation 5 are: _exp[-~~(W).l_ **__[9]
(1) The thermal conductivity of the particle l+@S
material is very large, consequently the
temperature within the particle is uniform and at x=6
and equal to the surface temperature.
(2) The course of particle heating may be
regarded as a succession of incremental
steady states. The particle temperature within the reaction
(3) The Nusselt number for heat transfer zone is seen to depend on the product para-
equals two. meter QS and on the dimensionless distance-
The velocity of a particle being swept along ratio x/S. The particle temperature at the end
by a moving gas will always differ from that of of the reaction zone is a function of @8 alone.
the gas itself. The importance of this lag can The temperature T, has been calculated as a
be estimated by comparing the relaxation time function of particle size for these specific values:
associated with velocity lag to the transit time cg, = 3 and 1, V= 4S, X = 2 x 10m4 cal cm-l set-’
of the particle through the flame. When this is deg-I, pi= 1 x 1O-3 g cm-3, c =0.25 cal g-l
done it is found that velocity lag is not a serious deg-I, and S=40, 25, 15 and 10 cm set-I. The
Perturbation for small particles (r N 1 or 2~) cited values for cg, are applicable to sodium
but will occur to some extent for larger particles chloride and sodium fluoride respectively. It is
(r N 5~). The effect of velocity lag will be to clear from the variation of T, with particle size
lengthen the time of exposure of the particle to (Figure 10) that the degree of particle heating
the flame. The rate of heat transfer will not be can be great and that particle temperature can
significantly affected because of the low Rey- rise high enough to result in evaporation of
nolds number associated with small particles materials not usually considered volatile.
114 W. A. Rosser, Jr, S. H. Inami and H. Wise Vol. 7

transfer alone, it is assumed that steady state

expressions can be used to describe the rate at
which mass and heat are transferred.
Evaporation (an endothermic process) is
limited by, and can proceed only to, the extent
that heat is provided to maintain the process.
The rate of conductive heat transfer, q, from
the flame gases to the particles is given by

q = 2~0% (T, - T) . . ..[14]

Initially, q >>&,, and both the particle tempera-
ture and n, increase. Eventually there may
arrive a time when q = &, and evaporation may
be regarded as having begun at that time.
Further addition of heat to a particle will result
0 0 16 24 32 primarily in evaporation and only secondarily
Particle diameter, TV
in an increase in particle temperature. This
Figure 10. Temperature reached by small man- temperature increase is not thermally signifi-
evaporating particles during passage through a cant but is important in that n, is a strongly
sto:chiom.etric m,ethane-air flame
varying function of particle temperature. The
In an isothermal system the steady state temperature at which evaporation may be con-
molar rate of flow, &, of material M diffusing sidered to have begun can be obtained by com-
away from a sphere of radius r may be repre- bining equations 13 and 14.
sented by h (T, - T) = n,LD . . . . [15]
&= -4nDrlzln (1 -n,/n) . . . . [ll]
The quantity n, is related to particle tempera-
lvhere n denotes total molar concentration, n, ture by
is the molar concentration of ‘M’ at the surface n, =Ae-LIRT . . . . [I61
of the sphere, and D is the diffusion constant
of ‘M’. Equation 16 when substituted in equation 15
If the ratio n, / n < 1, equation 11 may be results in
approximated by
T,-T=r4e-L~RTDL,lh . . . . [17]
#X=47c Dm, . . . . [12]
The initial evaporation temperature is that tem-
Relative to un-evaporated material the enthalpy perature T, consistent with equations 4, 9 and
flow +, associated with the molar flow, &, is 17 (for x> 0).
given by By graphical methods T, was determined as
a function of particle size for sodium chloride
& = 47~Dm,L . . . . [13] and sodium fluoride. The necessary thermo-
where L is the molar heat of evaporation. In a dynamic data were obtained from ref. 7 and
non-isothermal system, such as a flame, the used to derive expressions of the type of equa-
material diffusing away from a particle will be tion 16 for n,. For sodium chloride these data
further heated as it diffuses. Compared with L were used: b.pt= 1738”K, L at b.pt=41 kcal/
which is about 40 to 50 kcal/mole for the mole, D=4 cm2/sec. The quantity n, was
materials we will consider this enthalpy increase represented by
is small and will be neglected. The diffusion
n,=0.924 x 10-*900iTmole/cm3 . . [IS]
and heat transfer processes are then coupled
only at the surface of the particle. As for heat For sodium fluoride these data were used:
June 1963 The effect of metal salts on premixed hydrocarbon-air flames 115

b.pt=l977”K, L=48 kcal/mole, D=4 cm2/ be transformed into equation 23, a form suit-
set, and n, was represented by able for determining the degree of evaporation.

n,=l.lZx 10-104001Tmole/cm3 . . [19]

In both cases the diffusion constant D is an
estimated value.
The variation of T, (initial) with particle size The parameter (G?), is the value of @a appro-
was determined graphically. Because S and r priate to d, and to an assumed value of the
enter into the parameter @ in the same way, flame speed. The quantity 1 -xe/6 can be ob-
once T, is shown as a function of particle size tained from a graph of (T, - Te)i as a function
for a given flame speed, T, can readily be cal- of (@a),, the value of (GG),, and equation 4.
culated as a function of size for other flame The quantity (T, - T) may be taken without
speeds. For either sodium chloride or sodium significant error as the average of (T, - T& at
fluoride the initial evaporation temperature in- x=x, and (T,-T,), at x=6. For sodium
creases with increasing particle size up to a chloride (T, - T,), = 720°K and for sodium
maximum value, Tecmax.,. The temperature of fluoride (T,-T,),=550”K.
an evaporating particle cannot exceed (although Proceeding as indicated, we calculated the
it will reach) T,c,,,, at x =6 unless the particle fractional degree of evaporation, f = 1 - d3,/ d:,
evaporates completely during passage through for sodium chloride and sodium fluoride. The
the flame. For sodium chloride, T,(,,.,= results are shown in Figure 11. The product
1480”K, and for sodium fluoride, T,(,,., = Sd, is constant for a given degree of evapora-
1650°K. tion. Consequently, the variation of f with
Equation 12 implies that a particle evaporat- particle size for any flame speed can be easily
ing under steady state conditions will decrease obtained once the variation is known for a single
in size at a rate described by the expression flame speed.
The calculation of degree of evaporation car-
d==d;-8MDnltlp, . . . . [20] ried out for x =6 can be done for other selected
where M is the molecular weight of the material. values of x. For the purpose of demonstrating
A similar steady state expression can be derived that small particles of materials such as sodium
for a particle evaporating in a flame chloride and sodium fluoride will partially or
f completely evaporate during passage through a
flame, the results shown in Figure 10 for x =F
dz=d:-=~(Tl-T)dt . . [21]
PSL are sufficient.
0 The mathematical analysis just described
In equation 21, n, has been eliminated by means can only be successfully applied to materials for
of equation 15. Time is measured from the on-
set of evaporation as previously defined. Let x,
be the point in the flame corresponding to this
time. In the cases we will consider, xe > 0. This -NaF
distance will be traversed in a time 7 = (6 - xe)/V- ---NaC 1

$=r-~(T’T . . . . [22]
0 s
where d, denotes diameter of the particle at
x =6 and (T, - T) is the average temperature
difference between the gas and the particle in
the interval S-x,. Initial particle diameter, t_h
Further, V=4S, and after substitution for T
Figure I I. Fractional evaporation of sodium chloride
and V and some manipulation, equation 22 can and of sodium fluoride on leaving the reaction zone
116 W. A. Rosser, Jr, S. H. Inami and H. Wise Vol. 7

which there is available a relation between the which were not effective inhibitors, talc, calcium
vapour pressure and the temperature of the carbonate, calcium hydroxide and calcium
material. Substances such as sodium carbonate fluoride are not volatile at the flame tempera-
which decompose during evaporation cannot be tures encountered in our system and would not
treated. Therefore, the relative volatility of be expected to evaporate to a significant extent
several such materials was determined in the during passage through a methane-air flame. It
following way. A small quantity of material was appears that evaporation and inhibition effect-
fused by flame on to the end of a platinum wire iveness are companions and that the observed
until a small bead about 1 mm in diameter had inhibition is due to evaporated materials.
been formed. The bead and wire were suspen-
ded in a Meker burner flame about 1 cm above Discussion
the primary cones. By stopwatch, the time was The theoretical analysis presented in the pre-
measured from the appearance of colour in the ceding section revealed that a significant amount
flame gases to the disappearance of that colour. of evaporation will occur for those powders
The results of such measurements are shown in which were found to be effective inhibitors. The
Table 2. extent of evaporation at any given point in the
flame will depend on flame speed, the volatility
Table 2. Relative volatility of various solids of the material, and the particle size distribu-
tion. Consequently the shape of curves rela-
Substance i (set) B.pt (“K)
ting propagation velocity or flame speed to
NaI -1 1 577 powder concentration will vary from powder to
KC1 1 680
cuci 5
powder. Depending on circumstances powder
1 639
NaBr -2 1 GGG effectiveness may or may not depend on specific
NaCl 3 1 738 surface area.
KzC0.j --
KpSOi lo’ -- In the limit of a volatile fine powder, evapora-
Li&O.i 11 -- tion will be complete early in the flame and one
NaF 13 1 977
Na&,, 16 -- would expect to obtain a dependence of propa-
-. gation velocity on powder concentration similar
7 is the lif&me of 1 mm bead of material when exposed to a
in form to that produced by a gaseous inhibitor,
Meker flame.
for example phosphorus trichloride. The data
Consideration of the data in Table 2 reveals shown in Figures 5 and 6 for sodium chloride
a number of interesting facts. Sodium fluoride and cuprous chloride are of this type and sug-
and sodium carbonate have roughly the same gest that there is little or no dependence on
volatility. Consequently, the degree of evapora- particle size for particles with initial diameters
tion derived for sodium fluoride can be expected less than about 5~. This conclusion is consistent
to apply roughly to sodium carbonate as well. with the analysis presented in the preceding
Next, potassium carbonate is seen to be about section. On the other hand, the data shown in
twice as volatile as sodium carbonate. This is Figure 8 for potassium sulphate do imply a
also about the same as the ratio of inhibition dependence on particle size. The mean particle
effectiveness for these two materials. Lastly, all size for potassium sulphate is comparable to
the test substances except sodium carbonate are that for cuprous chloride but the volatility is
more volatile than sodium fluoride and small much less (see Table 2). Consequently, the
particles of these substances would be expected potassium sulphate data may be interpreted in
to evaporate during passage through a flame the following way. The initial decrease in pro-
to an even greater extent than sodium fluoride pagation velocity is attributed to material which
particles of the same size. Powders of all but has evaporated from smaller than average par-
two of the materials listed in Table 2 were ticles, and to heat loss. As the propagation
tested experimentally and found to be effective velocity is further reduced by higher concentra-
inhibitors of methane-air combustion. Materials tions of powder, larger and larger particles
June 1963 The effect of metal salts on premixed hydrocarbon-air flames 117

begin to contribute substantially to the overall relatively inactive. One must conclude that in-
degree of evaporation because of their longer hibition is associated with the sodium atom,
residence time under these conditions. The and that the overall effectiveness of a sodium
process becomes self-accelerating to a degree compound depends not only on volatility but
which depends on the particle size distribution, also on the availability of the atom after
and the propagation velocity drops abruptly at evaporation. By analogy other alkali metal
a concentration of about 6 x 10e3 mg/cm3. The compounds would behave similarly. The ability
abrupt change in slope, therefore, reflects an to inhibit combustion is probably not limited to
abrupt change in overall fractional evaporation the alkali metal atoms in view of the similar
from a value much less than one to a value of results obtained with sodium chloride and cup-
about one. rous chloride. The two are of comparable
The data shown in other figures can be inter- volatility and both would yield some metal
preted in a similar way. Precision of measure- atoms. Further, both tetraethyl lead and iron
ment was not as good with other powders as it pentacarbonyl strongly inhibit hexane-air
was with cuprous chloride and potassium sul- flameslO. The same mechanism may be opera-
phate (plus Cab-0-Sil), both of which were tive in those cases, although in the flame both
easily dispersible materials. iron and lead may be partially converted to
Two essential tasks remain: to identify the stable oxides or hydroxides.
species responsible for inhibition and the reac- The presence of metal atoms in the flame may
tions involved. To that end, consider first the result in an increase in the rate of recombina-
relatively simple case of sodium chloride. The tion of radicals associated with flame propaga-
condensed material will evaporate at high tem- tion-H, 0 and OH. Recombination reactions
peratures primarily as NaCl molecules. The limit the concentration of radicals in the flame
sodium chloride in diffusing through the hot and ensure their removal in the post-flame
reaction gases may be converted to sodium by gases. Consequently, an increase in recombina-
reactions such as tion rate in the reaction zone can be expected
to reduce flame speed. A metal 2 (e.g. sodium
H+NaCl+Na+HCl . . . . [24] or potassium) can increase the rate of recom-
Experimental observation of flames in which bination by a two-step process.
sodium and chlorine were both present indicates
that the conversion will not be completes. Con- A+Z+M+ (A-Z)+M . . . . [25]
sequently, both sodium and sodium chloride (A-Z)+B--+AB+Z . _. . [26]
must be accepted as major species involving
sodium. Traces of sodium hydroxide can also in which A and B may be H, or 0, or OH, and
be present, as well as minute concentrations of M is a third body. The collision-stabilized com-
sodium hydride or sodium oxides. However, the plex (A-Z) need not be a major species but it
oxides have never been observed spectroscopic- must have a sufficient lifetime to react with B.
ally in flames which also contained sodiuma. If the intermediate (A-Z) does survive this
Consider next the more complicated case of length of time and if reaction 26 is sufficiently
sodium carbonate which evaporates with de- exothermic, the rate of reactions 25 and 26
composition. The initial products in the course relative to the direct recombination
of diffusing away from the particle will decom- A+B+M+AB+M . . . . [27]
pose or be converted by reaction to the Na atom.
The presence of chlorine species in the gas will will be of the order of the ratio Z/B. In order
convert a portion of that sodium into sodium to evaluate the feasibility of this mechanism,
chloride. The observed decrease in powder the cited ratio in the flame zone must accord-
effectiveness as observed in the presence of ingly be known.
sodium carbonate powder and methyl chloride The equilibrium mole fractions Y of H, OH,
(see Figure 7) indicates that sodium chloride is 0 in the post-flame gases of a one-atmosphere
118 W. A. Rosser, Jr, S. H. Inami and H. \&se Vol. 7

stoichiometric methane-air flame with an adia- tration of metal atoms in the gas required to
batic flame temperature of 2214°K have been reduce flame speed by half is, therefore, taken
reportedI : YH =3.57 x 10-4, YOH= 27.8 x 10-4, as equal to the experimental value for potas-
and Y,=2.11 x lo-“. In the flame zone itself sium monohydrogen carbonate, X+ = 1 x 10-3.
the mole fractions of these three species are This concentration of metal atoms is about the
probably higheP, consequently the cited values same as the estimated value for H and 0, but
represent lower limits. For present purposes the much less than that for OH. In the reaction
mole fractions in the flame zone will be taken zone the ratios Z/H and Z 10 are accordingly
as 1 x low3 for H and 0, and 1 x lo-’ for OH. of the order unity while the ratio Z/OH is much
less than one. A recombination mechanism in
Table 3. Relative effectiveness of various powders which B=O or H could therefore contribute
Substance 2 (11) substantially to the rate of recombination of A
Xf
and B. Consider now a specific reaction mech-
KzSOa (+Cab-0-M) 1.8 X 10-S 4.3 anism such as
NazC03 (fine) 1.6 X 10-Z 2.2
NaCl 3.6 X 10-S 6.0 OH+Na$ (Na-OH) . . . . [28]
CuCl 2.5 x 10-S 5.7
NaHC03 (fine) 2.2 x 10-S 2.3
H+ (Na-OH) + Na+ H,O . . . [29]
KHC03 (fine) 1.0x 10-3 2.2
.rf denotes the mole fraction of metal in the unburnt gas required Both reactions 28 and 29 are exothermic. The
to reduce the propagation velocity of a stoichiometric methaneair intermediate Na-OH is sufficiently stable to
flame from an initial value of about 70 cm/w to 35 cmlsec.
persist for the required length of time, and the
The concentration of metal atoms required to ratio Na/H is of the order one. Such a reaction
produce a significant reduction in the propaga- pair would then result in a rate of recombina-
tion velocity of a stoichiometric methane-air tion comparable to that of the uncatalysed
flame can be estimated from the data shown in reaction
Table 3. For each test powder we calculated for
the unburnt gas, the overall mole fraction of H+OH%H,O . . . . [30]
metal atoms x+ which would reduce propagation
and the total rate of recombination would be
velocity from an initial value of about 70 cm/
twice that of the uncatalysed reaction. The
set to half that value, 35 cm/set. The range of
consequent reduction in H atom concentration
variation is not great. The listed values of X+
will further enhance the importance of the re-
are also the maximum mole fractions of metal
actions 28 and 29 compared with reaction 30
in the reaction zone. Either incomplete evapora-
because of the associated increase in the ratio
tion or failure to liberate the metal atom after
Na/H.
evaporation will result in a true mole fraction
Other specific reactions exemplifying the
less than the listed value. The values for sodium
general recombination scheme, equations 25 and
chloride and cuprous chloride are high because
26, can be devised. For all these, the important
some metal is tied up as halide molecules. The
criteria are : (I) Z/B> 1 in order to obtain a
value for potassium sulphate is high in view of
significant increase in recombination rate; (2)
the interpretation of Figure 8. The best values
the bond Z-A must be strong enough to pre-
are those for tine sodium carbonate (2=2.2~),
vent decomposition prior to reaction with B;
fine sodium monohydrogen carbonate (d=2.3,~), and (3) reaction 26 should be exothermic by
and for fine potassium monohydrogen carbon- several tens of kilocalories if the product mole-
ate (d= 2.2~~). The average particle size is much cule A-B is a major species (e.g. H,O). If
the same for all three substances but potassium criterion (3) is not satisfied, the inverse of re-
monohydrogen carbonate, or rather its decom- action 26 may nullify reaction 26 early in the
position product potassium carbonate, is the combustion process.
more volatile and would, therefore, be expected Conceivably, inhibition could also result from
to evaporate the most completely. The concen- the deactivation of energetically excited species
June 1963 The effect of metal salts on premixed hydrocarbon-air flames 119

such as O,*, C02*, C,*, CH* and OH*. An References

inhibition mechanism involving the removal of 1 FRIEDMAN, R. and LEVY, J. B. ‘Survey of funda-
excitation energy by metal atoms could provide mental knowledge of mechanisms of action of
a qualitative explanation for inhibition by pow- flame-extinguishing agents’. W.A.D.C. tech. Rep.
No. 56-548, January 1957
ders. Such a mechanism would require that one 2 DOLAN, J. E. and DEMPSTER, P. B. I. aPPl.
or more excited species be important inter- Chem., Lond. 1955, 5, 510
mediates in the combustion of hydrocarbons. 3 ROSE, H. E. J. appl. Chem., Lond. 1952, 2, 80
However, there is little evidence that energetic- 4 ROSE, H. E. ‘The measurement of particle size in
very fine powders’, Lecture III. Chemical Publish-
ally excited species are anything other than
ing-Co. : fiew York, 1954
subsidiary performers dependent on but not 5 Fire Research Abstracts and Reviews, 1959, 1,61,
affecting the rate of flame propagation. abstract of N.R.L. Rep. No. 5183, August 1958
Chemical inhibition can reduce but not elim- 6 GAYDON, A. G. and WOLFHARD, H. G. Flames,
inate the supply of reactive radicals present in Their Structure, Radiation and Temperature, 2nd
ed. rev., p 91. Chapman and Hall : London, 1960
the flame zone. If the radical population in the ’ KUBASCHEWSKI, 0. and EVANS, E. L. Metallurgical
flame is reduced below that which prevails at Thermochemistry. Wiley: New York, 1956
equilibrium in the post-flame gases, the result- 8 PHILLIPS, L. F. and SUGDEN, T. M. ‘Some observa-
ing diffusive flux of radicals from the product tions on the radiative combination of atomic hydro-
gen with atomic halogens in burner flames’, paper
zone into the flame zone will provide a new presented at a Symposium on Some Fundamental
source of reactivity which can only be elimin- Aspects of Atomic Reactions, held at McGill
University, Montreal, 6 and 7 September 1960
ated by a gross temperature reduction. Neces-
9 GAYDON, A. G. The Spectroscopy of Flames, p 224.
sarily then, heat loss to the powder particles Chapman and Hall: London, 1957
and to the container walls must administer the 1” LASK, G. and WAGNER, H. G. ‘Influence of addi-
co@ de g&e. This view is consistent with the tives on the propagation velocity of laminar
flames’, p 55 of abstracts of papers presented at the
marked decrease in powder effectiveness which Eighth Symposium (International) on Combus-
appears when propagation velocity is reduced tion, California Institute of Technology, August
to about 15 cm /set and with recent theoretical 1960
studies of quenching13s14. 11 CLINGMAN, W. H., BROKAW, R. S. and PEASE,
R. N. Fourth Symposium (International) on
The major portion of this research was car- Combustion, p 310. Williams and Wilkins: Balti-
more, 1953
ried out under Contract No. AF 33(616)-6583 12 KASKAN, W. E. Combustion & Flame, 1959, 3, 49
with the Aeronautical Systems Division, United 1s MAYER, E. Combustion & Flame, 1957, 1,438
States Air Force. 14 SPALDING, D. Proc. Roy. Sot. A, 1957, 240, 83