88

Ind. Eng. Chem. Res. 1992, 31, 88-94

Particle Formation by NH
3
-S0
2
Reactions at Trace Water Conditions
Hsunling Bai, Pratim Biswas,* and Tim C. Keener
Department of Civil and Environmental Engineering, University of Cincinnati, Cincinnati, Ohio 45221-0071
formation from NH
3
-S0
2
reactions at trace water vapor concentrations has been studied
expenmentally and possible solid products-NH
3
S0
2
, (NH
3
hS0
2
, and
(NH4)2S205-were considered, and calculatiOns were carried out to predict equilibrium concentrations
of A gas to con.version model for the NH
3
-S0
2
system was developed,
simultaneous chem1cal reactiOn, nucleation, condensation, and coagulation. Potential
apphcat10ns of this system of reactions are also discussed.
Introduction
The gas to particle conversion process associated with
the reaction of sulfur dioxide and ammonia has attracted
the attention of many researchers. It has application in
the areas of ammonium sulfate production and for en-
hanced capture of sulfur dioxide in flue gas desulfurization
processes (Hartley and Matteson, 1975; Scargill, 1971).
The direct reaction of these two gases may also be prev-
alent in the lower stratosphere where an aerosol layer is
expected to form (Scott et al., 1969).
Various reaction products have been proposed in the
literature depending on the reactant concentrations, re-
action time, moisture content, and reaction temperature.
Different products of the anhydrous reaction between NH
3
and so2 have been reported in studies dating to the
nineteenth century (Divers and Ogawa, 1900; Scott et al.,
To whom correspondence should be addressed.
1969, historical review). Badar-ud-Din and Aslam (1953)
were the first to conclusively report that the products of
the reaction were amidosulfurous acid (NH
3
S0
2
) and am-
monium amidosulfite ( (NH
3
hS0
2
) below a temperature
of 10 oc in the absence of water vapor. Hartley and
Matteson (1975) conducted room temperature experiments
of NH
3
and S0
2
and found by X-ray diffraction that the
most likely products were NH
3
S0
2
and (NH
3
)
2
S0
2
at water
vapor concentrations approaching those of NH
3
and 80
2

Ammonium sulfate was found to be the product under
excess moisture conditions. Vance and Peters (1976a)
studied the anhydrous reactions in a concentric flow re-
actor and stated that (NH
3
)
2
S0
2
being less volatile is the
favored solid reaction product, except under excess 80
2
concentrations. They also measured size distributions
using an impactor and a diffusion battery, with particle
sizes in the range of 0.01-7.5 Size distributions were
also reported by Carabine et al. (1971) with a mean size
of around 0.12
0888-5885/92/2631-0088$03.00/0 1992 American Chemical Society
A
J!!J Mass flow controller
e Bypass either outlet stream
r, Pressure gauge/regulator
ea Filter
I Rotameter
Reactor

Ind. Eng. Chern. Res., Vol. 31, No.1, 1992 89
so2
'tank
NH3
Tank

::7
Computer
B (a). Setup 1
to
hood
....
Air
Air Filter
(b). Setup 2 (c). Setup 3
Figure 1. (A) Schematic of the experimental setup to study the NH
3
-S0
2
reaction. (B) Schematic diagram of three different setups to check
for mixing conditions.
Studies in the literature thus indicate that the products
of the anhydrous reaction are primarily NH
3
S0
2
and
(NH
3
hS0
2
However, Hartley and Matteson (1975) and
Vance and Peters (1976a) also reported that water vapor
acts as a catalyst during the initial particle formation
process. Trace quantities of water vapor ( <400 ppm) can
also act as a reactant (St. Clair, 1937; Scargill, 1971), and
the most likely reaction product is ammonium pyrosulfite
(NH
4
hS
2
0
5
In this paper, we present an analysis of the
NH
3
-S0
2
reaction under trace water conditions by con-
sidering three reaction products: NH
3
S0
2
, (NH
3
hS0
2
, and
(NH
4
)
2
S
2
0
5
The equilibrium calculations performed at
these conditions are compared to experimental data. The
evolution of the aerosol size distribution is measured in
a laminar flow reactor and compared to the predictions
of a lognormal model accounting for simultaneous nu-
cleation, condensation, and coagulation.
Experimental Setup
The schematic of the experimental system to study the
NH
3
-S0
2
reaction is shown in Figure 1A. The reactor
consists of either a 40-mm-i.d. quartz tube or a 6.4-mm-i.d.
stainless steel tube. The quartz tube is approximately 80
em long. 1.02% S0
2
gas in N
2
is fed through a 1.5-mm-o.d.
tube along the reactor centerline. The length of this feed
tube can be adjusted to obtain desired reactor residence
times (2-32 s). A mixture of 4.5% NH
3
gas (in N
2
gas) and
clean air are introduced through a side port of the quartz
tube reactor. The stainless steel tube reactor was used for
shorter residence times ( <1 s). Sulfur dioxide and am-
monia gases were mixed in a union tee junction and then
reacted in the stainless steel tube reactor. Concentrations
of the gases were controlled by using MKS mass flow
controllers. Compressed air was passed through a Balston
filter to remove C0
2
, hydrocarbons, and oil and then dried
and cleaned by passing through a silica gel air dryer and
a Gelman air filter (AcroFlow II). The water vapor
content in both S0
2
and NH
3
tanks was provided by the
manufacturer to be less than 5 ppm. The relative humidity
(RH) of the dried, clean air was close to 0%, but water
vapor content was estimated from the data of Bower (1934)
and Trusell and Diehl (1963) to be between 30 (42 ppm,
0.15% RH) and 70 (94 ppm, 0.3% RH). The PMS
model LAS-X Optical Particle Counter was used as the
particle measurement instrument, and the Horiba PIR-
2000 Infrared Gas Analyzer was used for measuring the
S0
2
gas concentration. An attempt was made to use the
condensation nuclei counter (CNC) to measure the total
particle number density of the reaction products. No
particles were detected by the CNC, and it is suspected
that the anhydrous reaction products evaporate in the
CNC's saturation zone where the temperature is about 35
C. Similar observations were made by Vance and Peters
(1976a).
Preliminary experiments were conducted using three
different setups (Figure 1B) to ensure that proper mixing
was obtained in the reactor. The first setup was as de-
scribed above for the quartz tube reactor, with so2 being
injected through the center and NH
3
through the side port
of the reactor. The second setup was same as the first, but
the two gas inlet ports were exchanged; that is, NH
3
en-
tered through the center tube and so2 through the side
port. The third setup used was the stainless steel tube
reactor with the union tee mixing head, and provided a
90 Ind. Eng. Chern. Res., Vol. 31, No. 1, 1992
Table I. Comparison of the Three Different Reactor Setups
(S0
2
,1n = 3000 ppm, NHa,in = 6000 ppm, H201n = 50 ppm)"
80
2
removal at outlet, o/o
reaction time, s setup 1 setup 2 setup 3
0.4
0.6
1.2
2.2
4.5
32
28
30
30
16
12
15
27
27
28
a The comparison cannot be made at the same reaction residence
time due to the geometrical limitations of the reactors. However,
it is seen that the 80
2
concentration reached quasi-equilibrium
state in less than 0.4 s. The standard error of the experiment is
around ::1:4 o/o of S0
2
removal.
rapid mixing of the reactant gases. The results (Table I)
show that the first and the third setup produce the same
S0
2
removal after a quasi-equilibrium state is attained, and
the removal of S0
2
is higher than that from the second
setup. This is because the diffusivity of NH
3
in air is about
twice the diffusivity of S0
2
in air. Thus on injecting NH
3
from the side port, better mixing is obtained amongst the
gases as compared to injecting S0
2
from the side port.
Similar observations have been made by Kodas et al.
(1987) in their aerosol reactor studies.
Theory
Gas-Phase Equilibria. Past studies (Hartley and
Matteson, 1975; Vance and Peters, 1976a) have assumed
NH
3
S0
2
and (NH
3
)
2
S0
2
to be the only products under
anhydrous conditions. However, (NH
4
hS
2
0
5
could also
be formed at low water vapor conditions (St. Clair, 1937;
Scargill, 1971). In this work, we examine the following
three reactions to be simultaneously occurring:
NH
3
(g) + S0
2
(g) ~ NH
3
S0
2
(s) (1)
2NHa(g) + S0
2
(g) ~ (NH
3
hS0
2
(s) (2)
2NHa(g) + 2S0
2
(g) + H
2
0 ;:::t (NH
4
)
2
S20
5
(s) (3)
The equilibrium constants of (1) and (2) have been mea-
sured by Scott et al. (1969, 1970) and Landreth et al. (1974,
1985) and are listed in Table II with the appropriate tem-
perature regimes. The computed values of the equilibrium
constant differ from each other by several orders of mag-
nitude (Kp
1
= 1.9 X IQ-4 and 4.5 X 10- atm
2
from data of
Landreth et al. (1985) and Scott and Lamb (1970), re-
spectively, at 23 C). The equilibrium constant for reaction
3 has been computed by St. Clair (1937) and Scargill (1971)
and is listed in Table II. Reasonable agreement is obtained
in the values of the equilibrium constant computed by
these two expressions.
Assuming that the three reaction products form an ideal
solution with x mole fraction of NH
3
S0
2
and y mole
fraction of (NH
3
hS0
2
, the following expressions can be
written (Scott and Lamb, 1970)
Kp! = PN,.?s,e/ X
Kp2 = PN/Ps,e/Y
Kp3 = PN,e2Ps,e2PH,e/(1- X- y)
(4)
(5)
(6)
where the subscripts N, S, and H refer to ammonia, sulfur
dioxide, and water vapor, respectively, and the subscript
e denotes the equilibrium state. Combining (1)-(3), the
overall reaction is
(2 - x)NH
3
+ (2 - x - y)S0
2
+ (1 - x - y)H
2
0 ~
xNH
3
S0
2
+ y(NH
3
)
2
S0
2
+ (1 - x - y)(NH
4
hS
2
0
5
(7)
Table II. Equilibrium Constants for Three NH
8
-80
2
Reactions
equilib temp
possible reactions expressions range, C
NH
3
S0
2
(s) = NH
3
+ 80
2
Kp
1
= exp(22.7 5 to 45
- 9259/T)
Kp
1
= exp(42.7 -10 to -70
- 16205/T)
(NHal2S02(s) = 2NH
3
+ KP
2
= exp(43.8 5 to 45
802 - 16606/T)
Kp2 = exp(88.0 -10 to -70
- 31306/T)
(NH4)2S
2
0
5
(s) = 2NH
3
+ Kpa = exp(94.6 0 to 23
280
2
+ H
2
0 - 39144/T)
Kpa = exp(96.5 60 to 110
- 40767/T)
source
2
1
2
3
4
a (1) Landreth et al., 1985. (2) Scott and Lamb, 1970. (3) Scar-
gill, 1971. (4) St. Clair, 1937.
Using (4)-(7), the equilibrium vapor pressure of each gas
is
y Kp2
PN = - - = Pw - m(2 - x) (8)
,e X Kp! ,tn
x2 Kp12
Ps = - - = P
8
- m(2 - x - y) (9)
,e Y Kp2 ,m
1- X- y Kp3
PH,e = 2 K 2 = PH,in - m(1 - X - y) (10)
X pl
where m times the stoichiometric constant of the gas
reactant in (7) gives the vapor pressure consumption of
that gas when equilibrium is reached. Equations 8-10 are
a set of three nonlinear equations in x, y, and m and are
solved using a nonlinear equation solver, DNEQNJ (IMSL,
1987).
Particle Formation Mechanisms. The general dy-
namic equation for the gas to particle conversion process
is (Friedlander, 1977)
an a(Gn)
- + --- l'(v*) o (v-v*) =
at au
1 fv
2 Jo {3(v-iJ,iJ) n(v-iJ,t) n(iJ,t) diJ-
n(v,t) ~ {3(v,v) n(v,t) diJ (11)
The first term on the left-hand side (lhs) is the rate of
change of the particle distribution function in the particle
volume interval v to v + dv, the second term on the lhs
accounts for the effect of condensation at rate G, and the
third term on the lhs describes the formation of new
particles of critical volume v* at rate I'. The terms on the
right-hand side account for the effect of Brownian coag-
ulation. A monomer balance for the aerosol growth can
be written in terms of the system saturation ratio S
(Friedlander, 1983):
dS Rr' l'K* ( k8 T )
1
1
2
- = - - - - (S - 1) - A
dt n. n. 211"m
1
(12)
where the first term on the right-hand side is the monomer
formation rate by chemical reaction, the second and third
terms account for the nucleation and condensation effects,
respectively, n. is the monomer concentration at saturation,
K* is the number of monomers in the critical size nucleus,
ks is the Boltzmann constant, m
1
is the monomer mass,
and A is the total surface area of the particles. The mo-
Table III. Computed Mole Fraction of the Three Solid
Products for Different Water Vapor Concentrations
H
2
0, NH
3
S0
2
, (NH
3
)
2
S02, (NH4)2S20s,
ppm X y 1-x-y
80
2
= 3000 ppm; 0
NH
3
= 3000 ppm 50 0.045 0.005 0.950
100 0.042 0.005 0.953
200 0.036 0.004 0.960
400 0.024 0.002 0.974
80
2
= 3000 ppm; 0
NH
3
= 6000 ppm 50 0.086 0.020 0.894
100 0.081 0.019 0.900
200 0.077 0.018 0.905
400 0.057 0.011 0.932
80
2
= 3000 ppm; 0
NH
3
= 9000 ppm 50 0.130 0.046 0.824
100 0.123 0.043 0.834
200 0.117 0.041 0.842
400 0.088 0.029 0.883
The computation is based on thermodynamic data of Landreth
et al. (1985) and Scargill (1971).
ment formulation is employed to obtain the integral
properties of the size distribution using:
Mk(t) = foro vkn(v,t) dv (13)
Multiplying (11) by v on both sides, and integrating over
all particle sizes and using the lognormal distribution
function as an approximation of the entire size spectrum,
the following equations are obtained for a laminar flow
reactor (Pratsinis, 1988):
2(1 - r
2
) dS! dO = R. - IK* - 71(S - 1)M
0
(14)
2(1 - r
2
) dM
0
jdO = I- rM
0
2
(15)
2(1- r
2
) dMifdO = IK* + 71(S- 1)M
0
(16)
2(1- r
2
) dM2fdO = IK*
2
+ 2f(S- 1)M
1
+ 21/;M
1
2
(17)
where M
0
, M
1
, and M
2
are the total particle number con-
centration, total particle volume, and second volume mo-
ment, respectively, r is the dimensionless radial distance
(r = r*jR), 0 is the dimensionless time (=t/T, Tis the
characteristic time for particle growth); 11 and fare coef-
ficients for the condensation rate, and ~ and 1/; are coef-
ficients for the coagulation rate.
Equations 14-17 form a set of four coupled ordinary
differential equations which were solved using a differential
equation solver, DIVPAG (IMSL, 1987). The initial con-
ditions for the above equations are S = Sin and M
0
= M
1
= M
2
= 0 for all radial positions (Bai and Biswas, 1990).
The predominant solid product is (NH
4
)
2
S
2
0
5
(as can be
seen in Table III), and this is expected as it is less volatile
than the other two products. The saturation ratio for
forming the (NH
4
)
2
S
2
0
5
product can be written as
(18)
if chemical equilibrium is attained before particle growth
begins, and Sin is equal to 1 if chemical reaction and
particle growth occur simultaneously.
Results and Discussion
The measured S0
2
gas-phase concentration at the outlet
of the reactor as a function of the inlet ammonia concen-
tration is plotted in Figure 2. The residence time was
varied to ensure that the reaction attained a quasi-equi-
librium state. Calculations were also done at water vapor
Ind. Eng. Chern. Res., Vol. 31, No. 1, 1992 91
4000
3000
E
a.
0.2000
~ : : : : :
!
0
(f)
1000
0
3000
:::::
"'
"'
H
2
0=0 ppm
50 ppm
T
200 ppm
!
400 ppm
"'
......
.........
~ ~ H2o .. o
"' ...... PPrn
so 'PP;n ...... _
4500 6000 7500 9000
NH3,in ppm
Figure 2. Measured and calculated 80
2
equilibrium vapor pressure
as a function of inlet NH
3
concentration. Symbols are experimental
data and lines are theoretical results of our model using different
equilibrium constants for reactions 1 and 2: (-, Landreth et al.,
1985; -- -, Scott and Lamb, 1970).
concentrations using the equilibrium expressions listed in
Table II, and the results are shown in Figure 2. The zero
water vapor concentration results are the Landreth et al.
(1985) and Scott and Lamb (1970) model of two reactions.
Significant difference is obtained between the two pre-
dictions, and this may be due to the different temperature
regions in which the equilibrium constants were deter-
mined. The calculations were then done using the three-
reaction model proposed in this work and using equilib-
rium constants for reactions 1 and 2 proposed by both
Scott and Lamb (1970) and Landreth et al. (1985). When
the Scott and Lamb (1970) equilibrium constants are used,
not much change in the predicted equilibrium concentra-
tion is obtained at 50 ppm water vapor content. Also, as
seen in Figure 2, the curve tends to move further away
from the experimental data. However, there tends to be
better agreement with the measured concentrations when
the Landreth et al. (1985) expressions are used for a water
vapor content between 200 and 400 ppm. This corresponds
to a RH between 0.75 and 1.5%, which is slightly higher
than the estimated water content of silica gel dried air
streams ("'0.2% RH) (Trusell and Diehl, 1963). The
better agreement of predicted and measured concentra-
tions when the Landreth et al. (1985) equilibrium constants
are used as compared to the Scott and Lamb (1970) ex-
pressions may be due to the temperature range of the
Landreth et al. (1985) experiment being closer to that used
in this work. The equilibrium mole fractions of the dif-
ferent solid products at different water vapor contents are
summarized in Table III.
Using the Scargill (1971) computed equilibrium con-
stant, Kpa, the initial saturation ratio (18) is of the order
of 10
4
for inlet concentrations of NH
3
, S0
2
, and H
2
0 of
6000, 3000, and 50 ppm, respectively. This is much higher
than the saturation ratios of the other two compounds.
Therefore we can safely assume that (NH
4
)
2
S
2
0
5
is the
primary nucleation product, with subsequent condensation
of NH
3
S0
2
, (NH
3
)
2
S0
2
, and (NH
4
)
2
S
2
0
5
monomers onto
those nuclei. Our experimental data indicate particle
number density of around 10
5
/ cm
3
and geometric mean
sizes of around 0.1-o.2 ~ m which compared well with
results of Carabine et al. (1971), who also utilized a
light-scattering device to detect particles. The size dis-
tribution of the aerosol is measured by the OPC and
plotted in Figure 3.
The integral properties of the size distribution are
computed and compared to the predictions of the log-
92 Ind. Eng. Chern. Res., Vol. 31, No. 1, 1992
.........
a.
"0
.........
0'1
.2
<J
...........
c
0
:;::;
u
a
....
I.J...
6.0
4.0
2.0
0.0
0.08
r-
r-
r-
h
0.1 0.2
Particle size, I'm
0.3
Figure 3. Size distribution of NH
3
/S0
2
reaction products at room
temperature with inlet 80
2
concentration of 3000 ppm and inlet NH
3
concentration of 6000 ppm at 0.6-s residence time.
normal model described earlier, and the results are shown
in Figure 4. Since

is the predominant product,
we assumed it is the only product in the particle formation
model. Both particle density and the surface tension are
unknown for carrying out the gas to particle conversion
calculation. Particle density was estimated from data of
(NH
4
)
2
S0
4
(pP = 1.77 gjcm
3
) and NH
4
HS0
3
(pp = 2.03
gfcm
3
) and was varied in the range of 1.8-2.0 gfcm
3
in our
model. The value for surface tension is not available and
was adjusted to fit the experimental results (symbols in
Figure 4). Chemical reaction and particle growth may be
occurring simultaneously. As indicated by (14), the satu-
ration ratio increases due to chemical reaction (till the
reaction reaches equilibrium) and decreases due to nu-
cleation and condensation of the vapors. Limited infor-
mation is available in the literature regarding the chemical
reaction order or the reaction rate for the formation of
(NH
4
)
2
S
2
0
5
Hence an average value of R. was estimated
from our experimental data. The time required to reach
equilibrium, t., is between 0 and 0.4 sin our experiments
(Bai et al., 1991). The average value of Rr' was thus cal-
culated by {Ps,in- P
8
,.)N
8
v/82t.T, where Navis Avogadro's
number. This approach is better than assigning arbitrary
values to reaction rates as done by Kodas et al. (1986).
Three sets of model predictions are plotted in Figure 4.
All three sets are plotted by appropriate choice of param-
eters so that there are best fits with the number concen-
tration, geometric mean particle diameter, and geometric
standard deviation. The solid lines are plotted with the
results of model predictions accounting for simultaneous
chemical reaction and particle growth; the dashed lines are
plots for very fast reaction, or assuming that reaction
equilibrium is attained before particle growth begins. On
assuming very fast chemical reaction and on varying the
surface tension, reasonable agreement of experimental data
(number concentration, geometric mean particle size, and
standard deviation) with model predictions was obtained
for a surface tension of 70 dynfcm, this being close to the
value of H
2
S0
4
(Liu and Levi, 1980). The other dashed
line was plotted by using a higher surface tension of 93
dyn/ em, this being less than the value of NH
4
N0
3
used
in the literature [u = 113-119 dynjcm, Raes et al. (1990)].
The nucleation rate had to be scaled up by a factor of 10
9
for agreement between model predictions and experimental
data. The solid line is plotted using a t. of 0.04 s, and
reasonable agreement is obtained with the experimental
10 "!
10 '1
10
1
'*'
g 10'
:.:>
E
c
10 I I
10.
U
a
10
....
.,
.c
10'
E
::l
z
10.!
c
0
10
f-
1
0.0
A
__ a70 ergs/.em', g/.em:
0
----- aQ93 ergs;tcm!. p02.0 g;tcm,. \,=10 ,
_ a93 ergs/em . p
0
2.0 g/cm, K,6 X 10
---------------

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Average Residence Time (sec)
10 -...------------------,
E'
$
.... 10 _,

E
B
__ a=70 ergs/.cm:. p,t.B g/cm!
1
_____ a93 ergs/.em,. g/./em,. 1,10 ,
_ a-=93 ergs/em, p,2.0 g em. R, ... s X 10
.!2 10 ... ------------------------------
0
" . !
/ i
t:: 10 - I !!
cE '
c:
10 -
:IE
0
;:
., 10 _,
E
0
"
(!)

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Average Residence Time (sec)
2.0 ..,.--------------------,
1.9
1.8
1.7
"
0
'E 1.6
0
-g 1.5
.B
Ul 1.4
0
1.3
E
g1.2
(!)
1.1
c
__ a=70 ergs/.em', p
0
=1.8 g/.em:
0
_____ a93 ergs/em', p,2.0 g;tem,. 1,=10
0
_ a93 ergs/em', p
0
2.0 g/cm, R,=6 X 10
1 .0 +---,----.---,:---r-r--.--r----r:--::T"-:-1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Average Residence Time (sec)
Figure 4. (A) Measured and simulated results of cup mixing average
total particle concentration as a function of average residence time.
The inlet 80
2
and NH
3
concentrations are 3000 and 6000 ppm, re-
spectively. (B) Measured and simulated results of cup mixing av-
erage geometric mean particle diameter concentration as a function
of average residence time. The inlet 80
2
and NH
3
concentrations
are 3000 and 6000 ppm, respectively. (C) Measured and simulated
results of cup mixing average geometric standard deviation as a
function of average residence time. The inlet 80
2
and NH
3
con-
centrations are 3000 and 6000 ppm, respectively.
data. Increasing t. (slower reaction) does not yield good
agreement with the experiments, and hence the results are
not shown. The two dashed curves indicated that there
is a burst of nucleation and formation of particles, which
then primarily grow by condensation. This is also con-
flrmed by the fl.nal geometric standard deviation tending
to 1. In the case of simultaneous reaction and particle
growth, nucleation is delayed by a fraction of a second
(short delay as reaction is still rapid). The fmal geometric
standard deviation is higher, indicating growth occurs by
both condensation and coagulation. Also, the computed
characteristic time for particle growth (T) is about 0.12 s,
much larger than t, (=0.04 s). On comparing the experi-
mental data to the three sets of model predictions, it can
be assumed that chemical reaction is extremely rapid and
completed before particle growth begins.
Scott et al. (1969) and Vance and Peters (1976b) have
considered the formation of two adducts, NH
3
S0
2
and

under dry conditions in the stratosphere. Such

gas-phase reactions have often been neglected in atmos-
pheric chemistry. However, in the stratosphere where the
temperature is very low, gas-phase reactions could be an-
other mechanism of forming ammonium sulfate aerosols
in addition to photochemical reactions. The reaction of
NH
3
and S0
2
in the presence of water vapor, as described
in this work, to form (NH
4
)zS
2
0
5
is another possible
mechanism that has not been considered in atmospheric
chemistry. The reaction is fast and requires very low
partial pressures of the reactant gases. For example, the
equilibrium constant at -70 C is 4.9 X 10-
16
ppm
5
for the
reaction that forms (NH
4
)zS
2
0
5
(extrapolation from
Scargill (1971)). This implies that at a 10% relative hu-
midity, if the vapor pressure product of ammonia and
sulfur dioxide exceeds 4 X 10-s ppm
2
, (NH
4
)
2
S
2
0
5
particles
will possibly form. (NH
4
)zS
2
0
5
is highly soluble and forms
ammonium bisulfite (NH
4
HS0
3
), which is then easily ox-
idized to form a stable product-ammonium sulfate (St.
Clair, 1937).
Reactions involving ammonia and sulfur dioxide gases
could also be applied for industrial sulfur dioxide emission
control. Removal of sulfur dioxide is possible when in-
jecting ammonia gas into flue gases at lower temperatures
(4o--60 C). For example, assuming (NH
4
)zS
2
0
5
as the
major product, and with initial S0
2
and NH
3
concentra-
tions of 3000 ppm at 50 C at 8% (by volume) H
2
0 vapor
concentration, the 80
2
removal could be over 90% based
on the equilibrium calculation. If this equilibrium is truly
achievable, then removal of S0
2
from a high sulfur coal
combustion process by ammonia gas injection provides
another route for industrial gas cleaning. A bench-scale
experiment is currently under way to evaluate the possi-
bility of S0
2
removal by ammonia gas injection.
Conclusions
The gas-phase reactions between sulfur dioxide and
ammonia at trace water vapor concentrations were studied.
Quasi-equilibrium gas concentrations and particle prop-
erties were measured experimentally and then compared
to the predictions of a theoretical model. A three-reaction
mechanism leading to NH
3
S0
2
(s), (NH
3
)zS0
2
(s), and
(NH
4
)
2
S
2
0
5
(s) as products was used to carry out the
equilibrium calculations. A theoretical model of the gas
to particle conversion process accounting for chemical
reaction, nucleation, condensation, and coagulation was
developed. Since (NH
4
)
2
S
2
0
5
(s) is the predominant
product of the NH
3
-S0
2
reaction, it was considered as the
source of primary nuclei and further growth of particles
by condensation and coagulation. The measured particle
size appears to be lognormally distributed with a geometric
mean diameter of 0.1-o.2 geometric standard deviation
of 1.1-1.2, and total particle number concentration of
around 10
5
/ cm
3

Acknowledgment
This work was funded by Ohio Coal Development Or-
ganization Grant RF3-88-014.
Ind. Eng. Chem. Res., Vol. 31, No. 1, 1992 93
Nomenclature
A = total particle surface area
G = particle growth rate by the condensation effect
I = particle formation rate (nucleation rate)
I. = nucleation rate scale
K* = number of monomers in the critical size nucleus
ka = Boltzmann constant
J!.I' = equilibrium constant of the NH
3
-S0
2
reaction
Mk = kth aerosol volume moment (k = 0, 1, 2)
m
1
= monomer mass
n, n(v,t) = particle density function
Nav = Avogadro's number (=6.02 X 10
23
molecules/mole)
n. = monomer concentration at saturation, molecules/cm
3
P = partial pressure of the gaseous reactant, atm
R = pipe radius
R, = monomer formation rate from chemical reaction
(=R/T/n
8
)
S = saturation ratio
T = temperature
t = dimensional time
t, = time required to reach chemical equilibrium
v = monomer volume
v* = monomer critical volume
r = radial distance
x = mole fraction of NH
3
S0
2
y = mole fraction of (NH
3
hS0
2
Greek Symbols
fJ = collision frequency function
o = delta function
E = dimensionless condensation coefficient
= dimensionless coagulation coefficient
71 = dimensionless condensation coefficient
8 = dimensionless time
Pp = particle density
u = surface tension
1/1 = dimensionless coagulation coefficient
T = characteristic time for particle growth [=[n.s
1
(k
8
T/
211'mr)l/2j-1]
Subscripts
N =ammonia
S = sulfur dioxide
H = water vapor
e = equilibrium state
in = inlet condition
Superscript
' = dimensional property
Registry No. NH
3
, 7664-41-7; 80
2
, 7446-09-5; H
2
0, 7732-18-5;
NH
3
S0
2
, 25310-87-6; (NH
3
)S0
2
, 29307-29-7; (NH
4
)S
2
0
5
, 32736-
64-4.
Literature Cited
Badar-ud-Din; Aslam, M. Ammonia-Sulfur Dioxide Reaction. Pak.
J. Sci. Res. 1953, 5, 6-10.
Bai, H.; Biswas, P. Reaction of Sulfur Dioxide with Ammonia to
Form Aerosols: Applications in Flue Gas Desulfurization. Ab-
stracts of Papers, Annual Meeting of the AAAR, Philadelphia;
AAAR: Philadelphia, 1990; P1F-2, p 265.
Bai, H.; Biswas, P.; Keener, T. C. Sulfur Dioxide Removal by Re-
action with Ammonia Gas to Form Solid Products. Presented at
the 84th Annual Meeting of Air and Waste Management Associ-
ation, 1991; paper 91-103.24.
Bower, J. H. Comparative Efficiencies of Various Dehydrating
Agents Used for Drying Gases. Bur. Stand. J. Res. 1934, 12,
241-248.
Carabine, M.D.; Maddock, J. E. L.; Moore, A. P. Particle Size Dis-
tributions in Aerosols Form from Gaseous Reactants. Nature
Phys. Sci. 1971, 231, 18.
Divers, E.; Ogawa, M. Ammonium Amidosulfite. J. Chem. Soc.
Trans. 1900, 77, 327-335.
94
Ind. Eng. Chem. Res. 1992, 31, 94-100
Friedlander, S. K. The General Dynamic Equation for the Continu-
ous Distribution Function. In Smoke, Dust and Haze; Wiley:
New York, 1977; pp 263-265.
Friedlander, S. K. Dynamics of Aerosol Formation by Chemical
Reaction. Ann. N.Y. Acad. Sci. 1983, 404, 354-364.
Hartley, E. M.; Matteson, M. J. Sulfur Dioxide Reactions with Am-
monia in Humid Air. Ind. Eng. Chem. Fundam. 1975, 14, 67-72.
IMSL Contents Document, Vol. 2, Version 1.0, International Math-
ematical and Statistical Libraries, Houston, 1987.
Kodas, T. T.; Pratsinis, S. E.; Friedlander, S. K. Aerosol Formation
and Growth in a Laminar Core Reactor. J. Colloid Interface Sci.
1986, 111, 102-111.
Kodas, T. T.; Friedlander, S. K.; Pratsinis, S. E. Effect or Reactant
Mixing on Fine Particle Production in a Tubular Flow Reactor.
Ind. Eng. Chem. Res. 1987, 26, 1999-2007.
Landreth, R.; de Pena, R. G.; Heicklen, J. Thermodynamics of the
Reactions (NH
3
)
0
S0
2
(s) = nNH
3
(g) + S0
2
J. Phys. Chem. 1974,
78, 1378-1380.
Landreth, R.; de Pena, R. G.; Heicklen, J. Redetermination of the
Thermodynamics of the Reactions (NH
3
)
0
S0
2
(s) = nNH
3
(g) +
S0
2
J. Phys. Chem. 1985, 89, 1690-1692.
Liu, B. Y. H.; Levi, J. Generation of Submicron Sulfuric Acid Aerosol
by Vaporization and Condensation. In Generation of Aerosols;
Willeke, K., Ed.; Ann Arbor Science: Ann Arbor, Ml, 1980;
Chapter 15.
Pratsinis, S. E. Simultaneous Aerosol Nucleation, Condensation and
Coagulation in Aerosol Reactors. J. Colloid Interface Sci. 1988,
124, 416-427.
Raes, F.; Kodas, T. T.; Friedlander, S. K. Aerosol Formation by a
Reversible Chemical Reaction: Laser-induced NH
4
N0
3
Aerosol
Formation in a Tubular Reactor. Aerosol Sci. Techno/. 1990, 12,
856-868.
Scargill, D. Dissociation Constants of Anhydrous Ammonium Sul-
phite and Ammonium Pyrosulphite Prepared by Gas-phase Re-
actions. J. Chem. Soc. (A): Inorg. Phys. Theor. 1971,2461-2466.
Scott, W. D.; Lamb, D. Two Solid Compounds which Decompose
into a Common Vapor. Anhydrous Reactions of Ammonia and
Sulfur Dioxide. J. Am. Chem. Soc. 1970, 92, 3943-3946.
Scott, W. D.; Lamb, D.; Duffy, D. The Stratospheric Aerosol Layer
and Anhydrous Reactions Between Ammonia and Sulfur Dioxide.
J. Atmos. Sci. 1969, 26, 727-733.
St. Clair, H. W. Vapor Pressure and Thermodynamic Properties of
Ammonium Sulphites. U.S. Bur. Mines Rep. 1937, Inv. No. 3339,
19-29.
Trusell, F.; Diehl, H. Efficiency of Chemical Desiccants. Anal.
Chem. 1963, 35, 674-677.
Vance, J. L.; Peters, L. K. Aerosol Formation Resulting from the
Reaction of Ammonia and Sulfur Dioxide. Ind. Eng. Chem.
Fundam. 1976a, 15, 202-206.
Vance, J. L.; Peters, L. K. Comments on the Direct Aerosol Forma-
tion by Reaction of Ammonia and Sulfur Dioxide. J. Atmos. Sci.
197Gb, 33, 1824-1825.
Received for review May 16, 1991
Revised manuscript received August 22, 1991
Accepted September 5, 1991
Productive and Parasitic Pathways in Dilute Acid-Catalyzed Hydrolysis
of Cellulose
William Shu-Lai Mok and Michael Jerry Antal, Jr.*
Department of Mechanical Engineering and the Hawaii Natural Energy Institute, University of Hawaii at
Manoa, Honolulu, Hawaii 96822
Gabor Varhegyi
Research Laboratory for Inorganic Chemistry, Hungarian Academy of Sciences, Budapest, Hungary
Cellulose hydrolysis experiments were conducted in a percolating reactor at 34.5 MPa. A glucose
yield of 71% of the theoretical maximum was obtained at 215 C with 0.05% by weight of sulfuric
acid in the percolating solution. The classical model of glucose formation from cellulose followed
by secondary sugar degradation did not describe the reaction chemistry under these conditions. A
parasitic pathway which leads to the formation of nonhydrolyzable oligomer was discovered in the
absence of acid. In the presence of acid, kinetic modeling of the measured, temperature-dependent
rates of glucose evolution indicates that an acid-catalyzed parasitic pathway operates in competition
with the glucose production pathway. No chemical changes were detected in the solid phase during
the course of reaction.
Introduction
Although the acid-catalyzed hydrolysis of cellulosic
materials was industrialized almost a century ago (Harris,
1949), the underlying chemistry is still a focus of research
interest today. Three approaches characterize the state-
of-the-art technology. Dilute acid hydrolysis processes
employ flow reactors which only accept a finely ground
feedstock. Typical glucose yields of about 55% can be
achieved with acid concentrations below 3% and a resi-
dence time of a few seconds (Thompson and Grethlein,
1979; Church and Wooldridge, 1981; Brenner and Rugg,
1985). Somewhat higher yields are obtained with perco-
lator reactors at the cost of reduced reaction rates, batch
operation, and the dilution of sugar products (Faith, 1945;
Harris and Beglinger, 1946; Gilbert et al., 1952). Finally,
low-temperature, concentrated acid processes are some-
times employed (Moore and Barrier, 1987; Sharples, 1957,
1958; Dunning and Lathrop, 1945). In most cases, the yield
of glucose is considerably less than 100%. The goal of this
research was to elucidate the underlying chemistry which
limits the yield of glucose from cellulose.
The classical explanation for the low yield of glucose was
first enunciated by Saeman (1945) and employed more
recently by McParland et al. (1982) and Conner et al.
(1985). It posits the role of secondary degradation reac-
tions in reducing the yield of glucose as the sole primary
product of cellulose hydrolysis: cellulose - glucose -
degradation products. Contradicting this classical picture,
important recent work by Abatzoglou et al. (1986) has
shown that the initial products of cellulose hydrolysis are
primarily soluble oligosaccharides. Moreover, in the
presence of 40 mM H
2
S0
4
at 190 C, Bouchard et al. (1989)
detected very significant alterations in the chemical
structure of the unconverted "cellulose" after 30 min at
reaction conditions. However, Bouchard et al. (1989) did
not actually measure the rate of glucose formation from
0888-5885/92/2631-0094$03.00/0 1992 American Chemical Society