You are on page 1of 8

144 Ind. Eng. Chem. Res.

2001, 40, 144-151

Ammonia Pyrolysis and Oxidation in the Claus Furnace


W. D. Monnery, K. A. Hawboldt, A. E. Pollock, and W. Y. Svrcek*
Department of Chemical and Petroleum Engineering, The University of Calgary,
Calgary, Alberta T2N 1N4, Canada

The modified Claus process is commonly used in oil refining and gas processing to recover sulfur
and destroy contaminants formed in upstream processing. In oil refining, in addition to the
typical modified Claus plant feed of H2S and CO2, NH3 is also often present. NH3 is a process
contaminant and must be destroyed in the front-end furnace of the modified Claus plant,
otherwise, it poses a risk of poisoning the catalyst beds and plugging off downstream equipment
because of the formation of ammonium salts. In this paper, the pyrolysis and oxidation of NH3
was studied under Claus furnace temperatures and residence times. Experimental data was
taken, and new reaction rate expressions were developed for NH3 pyrolysis and oxidation.
The derived rate expressions are outlined as follows: the NH3 pyrolysis rate expression r ) A
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

exp(-Ea/RT)PNH31.25, where A is 0.004 21 mol s-1 atm-1.25 cm-3 and Ea is 16.5 kcal mol-1, and
the NH3 oxidation rate expression r ) A exp(-Ea/RT)PNH3PO20.75, where A is 4430 mol s-1 atm-1.75
cm-3 and Ea is 40.0 kcal mol-1. The rate expression for NH3 pyrolysis matched experimental
Downloaded via CHEVRON USA INC on March 30, 2023 at 09:29:32 (UTC).

data within 13% and matched well with published data. The rate expression for NH3 oxidation
matched experimental data within 10%.

Introduction 5
2NH3 + O2 f 2NO + 3H2O (3)
2
The modified Claus sulfur recovery process is the
most common process used to remove sulfur from acid Incomplete pyrolysis or combustion of NH3 in the
gas streams occurring in oil and gas processes. This furnace results in NH3 and NO carryover into the
process is comprised of a high-temperature furnace catalyst beds. Ammonia can form ammonium salts,
followed by catalytic reactor(s). The reactions occurring which can plug or foul the catalyst beds, other equip-
in the furnace are numerous, and several authors have ment, or piping. Although the formation of SO3 occurs
attempted to delineate the important ones (Clark et al., in the catalyst bed regardless of the presence of NO,
1997). As part of our ongoing study, we have previously the presence of NO in the beds acts as a catalyst for
published kinetic data and reaction rate expressions for the conversion of SO2 to SO3, which in turn causes
H2S cracking/reassociation (Hawboldt et al., 1999a) and catalyst sulfation (Garside and Phillips, 1962). Of the
the Claus reactions (Hawboldt et al., 1999b). primary causes of catalyst activity loss, catalyst sulf-
The reaction furnace has three functions: the conver- ation is regarded as the most significant (Grancher,
sion of one-third of the H2S to SO2 for downstream 1978). It is therefore critical to convert as much of the
catalytic processing, the destruction of any contami- NH3 to N2, H2, and H2O as possible.
nants which may foul downstream equipment, and the One of the obstacles to designing and operating the
production of elemental sulfur, which can account for furnace properly, i.e., determining optimal residence
up to 70% of the inlet sulfur for a “straight-through- times and temperatures, is the absence of kinetic rate
type plant”. The types of contaminants in the sour gas expressions for the key reactions (Monnery et al., 1993).
stream depend on the source of the feed gas. For For ammonia destruction, an empirical rule of thumb
instance, in oil refinery operations, NH3 is formed as a that is generally agreed with in industry is that tem-
byproduct of denitrogenation operations, such as hydro- peratures greater than 1200-1250 °C are required.
cracking and hydrotreating. Subsequent sour water However, there appears to be some debate as to how
stripping results in a sour gas feed stream containing much residence time is required and how ammonia
NH3, which is then directed to the sulfur recovery oxidation competes for oxygen versus hydrocarbon and
facility for destruction. H2S oxidation (Johnson and Rempe, 1997; Goar, 1994).
The desired NH3 destruction reactions in the furnace As such, ammonia pyrolysis and oxidation kinetic data
are the complete pyrolysis and/or oxidation of NH3 are badly needed. In this paper, we will present new
experimental data and new kinetic rate expressions for
2NH3 T N2 + 3H2 (1) the NH3 pyrolysis and NH3 oxidation reactions at Claus
front-end furnace (FEF) conditions and residence times.
3
2NH3 + O2 T N2 + 3H2O (2)
2 Literature Review

Under excess oxygen conditions and incomplete mixing, NH3 Pyrolysis. Davidson et al. (1990) performed a
series of high-temperature pyrolysis experiments at a
the following reaction may also occur (Clark et al., 1997):
temperature range of 2000-3200 K, a pressure range
of 0.8-1.1 atm, and an NH3 concentration range of 0.1-
* To whom correspondence should be addressed. 1.0%. Under these conditions, NH3 is completely con-
10.1021/ie990764r CCC: $20.00 © 2001 American Chemical Society
Published on Web 12/01/2000
Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001 145

Figure 1. Apparatus schematic.

sumed within a 1 ms residence time. A detailed reaction effective in reducing NO at all temperatures and pres-
mechanism was proposed, consisting of 21 free-radical sures of the study. However, when the NH3 concentra-
reactions. These experiments were performed behind tion was comparable to O2, the reduction of NO was
reflected shock waves. The severity of the temperatures inhibited. Miller and Bowman (1989) came to the same
and the mechanism of the experiment (shock tube) conclusions and, in addition, found that the process
prevent direct comparison with data generated in our occurs in the narrow temperature range of 1100-1400
experiments. However, these studies do give an indica- K in the absence of other additives. They also found that
tion of trends and therefore the significance of inter- the presence of H2O had a slight inhibiting effect on NO
mediate species (NH2, NH, N2H2, etc.). reduction. Dou et al. (1992) proposed a simple kinetic
Clark et al. (1998) conducted a study on NH3 pyrolysis model that predicts the rate of NO and NH3 conversion
at FEF conditions to determine the effects of other in the NO-NH3-O2 system. Major conclusions from the
chemical species occurring at furnace temperatures, Dou et al. (1992) study indicate that NH3 conversion
pressures, and residence times. In general, both H2S increases with temperature, whereas NO conversion
and H2O were found to inhibit NH3 pyrolysis, whereas increases with temperature until approximately 1200
SO2 enhances NH3 pyrolysis. Their results also showed K, after which conversion decreases. Furthermore, NO
oxidation being much more rapid than pyrolysis. conversion rapidly levels off within 50 ms, indicating
NH3 Oxidation. Published studies have indicated that longer residence times will not affect NO conver-
that the oxidation of NH3 is a much more rapid process sion. In their experiments, O2 was in large excess.
than pyrolysis (Clark et al., 1998; Lindstedt and Selim, Clark et al. (1998) performed oxidation studies on
1994; Miller et al., 1983; Fujii et al., 1981). In the Fujii NH3 at residence times between 0.34 and 0.55 s and
et al. (1981) study, the reaction was investigated at inlet concentrations of NH3 and O2 of 4 and 3%,
temperatures between 810 and 2100 K, pressures from respectively. The study found that between 700 and
1.1 to 8.4 atm, NH3 concentrations from 0.5 to 75%, and 1100 °C conversion of NH3 is below 10%, whereas at
O2 concentrations from 0.5 to 90%, with the balance of temperatures greater than 1100 °C, the conversion
the gas being argon. increases to between 60 and 100%.
Miller et al. (1983) used a reaction model consisting
of 98 reactions to determine the reaction mechanism for New Experimental Data Generation
NH3 oxidation. The model predictions were compared
with experimental data generated by MacLean and Experimental Apparatus. The experimental ap-
Wagner (1967). The developed kinetic model predictions paratus, in Figure 1, consists of pressurized gas cylin-
were in good agreement for the lean to moderately rich ders delivering reactants through a controlled flow
gases. However, for the rich flames, the agreement was system to a flow reactor housed in a high-temperature
poor, possibly indicating inadequate modeling of am- furnace, a quench system, and a gas chromatograph for
monia pyrolysis. The model and experimental results analyzing reactants and products. The apparatus was
also showed very low concentrations (approximately designed by Fookes (1996) specifically for the kinetic
0.001 mole fraction) of NO in the post-flame gases. study of the gas-phase reactions occurring in the Claus
In 1994, Lindstedt and Selim proposed a reaction FEF.
mechanism consisting of five steps based on experimen- Reactant gases are dropped from cylinder pressures
tal data generated by MacLean and Wagner (1967) and to the reactor pressure across the regulators. Subse-
Vandooren (1992). Both Lindstedt and Selim (1994) and quently, flow rates are controlled by Linde FM4660
Miller et al. (1983) based their reaction mechanisms on mass flow controllers (MFC) connected to a main
flame studies, and therefore temperatures were not console. MFCs were calibrated prior to experimental
stated. work; however, to ensure flow rates were accurately
Other studies on NH3 oxidation have focused on the known, an Alexander Wright wet test meter was also
reduction of NO by NH3 in NH3-NO-O2 systems. Lyon used. Ultimately, flows were consistent to within 1.0%.
and Benn (1978) were the first to conduct a kinetic study Temperatures were measured with thermocouples, also
of the reduction of NO by NH3 in the presence of O2 at accurate to within 1.0%.
temperatures between 872 and 980 °C and pressures Quartz flow reactors were used in these experiments
between 1.07 and 2.14 atm. Provided O2 was present to avoid possible metal or ceramic catalytic effects. Each
in large excess compared to NO, NH3 addition was reactor is 5.0 mm in diameter, and three different
146 Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001

lengths were used to provide residence times from about


0.5 s to about 2.0 s, at temperatures between 800 and
1300 °C, typical of Claus front-end reaction furnace
conditions. The reactors have a preheat length, followed
by a mixing chamber, and then an isothermal hot zone.
The reactors are coiled for two reasons: to have
sufficient length to attain the desired range of residence
times in a compact arrangement and to create secondary
flow, which reduces axial diffusion or Poiseuille flow,
thereby enhancing plug flow (Nauman, 1977; Truesdell
and Alder, 1970). The reactor was designed to satisfy
the conditions Cutler et al. (1988) set out for the design
of a plug-flow reactor. Cutler et al. (1988) showed that,
even in laminar flow, the error in deriving kinetic data
from assuming plug flow is typically a few percent, up
to a maximum of 11%, providing certain dimensional
and heat- and mass-transfer criteria are met. Indeed,
Cutler et al. (1988) showed that in attaining kinetic data
there is more error incurred because of systematic error,
such as thermocouple precision. Furthermore, to verify
that the reactor operated in plug flow, a computational
fluid dynamic (CFD) model of the reactor was developed
by British Oxygen Corp.. The results of this simulation
showed that the flow regime closely approximates plug
Figure 2. Experimental measurements of the deviation of the
flow (Hawboldt, 1998). Last, we verified the apparatus temperature of the reactor from set point temperature along the
by comparing experimental results to accepted model reactor length at a flow rate of 8 SLPM.
results of ethane pyrolysis and matching these to within
5% (Hawboldt et al., 1999b). On the basis of the Cutler and 1200 °C, residence times ranged from 30 to 700 ms,
criteria, ethane pyrolysis results, and the results of the inlet concentrations of NH3 varied between 0.3 and
CFD simulation, the reactor was considered to operate 2.5%, and O2/NH3 ratios varied from 0.85 to 2.55
in plug flow with negligible error. (Hawboldt, 1998). Argon was used as the dilution gas
Product gases from the quartz reactors were quenched for this set of experiments. GC measurements of NH3
rapidly against water in a double pipe exchanger, were made for all experiments, and to check the mate-
typically down to less than 100 °C within 5-10 ms. This rial balance in selected experiments, N2 concentrations
was shown to provide an accurate sample of reactor were also measured. For stoichiometric mixtures of NH3
products (Hawboldt et al., 1999a). Finally, before prod- and N2, the agreement between NH3 consumed and N2
uct gases went to the gas chromatograph, they passed produced was within 10% (Hawboldt, 1998). This excel-
through a separator and filter to remove solid sulfur. lent material balance closure was also found for experi-
A key characteristic of this reactor system is that it ments with O2/NH3 feed ratios greater than the stoichi-
operates nearly isothermally, which simplifies data ometric 0.75. Although some NO may be formed and
analysis because temperature variations within the the GC cannot separate NO from N2, the results of
reacting system do not have to be corrected for. Isother- equilibrium calculations by both ourselves and Miller
mality was ensured with a rapid preheat zone and by et al. (1983) show that the concentration of NO formed
conduction of experiments under dilute conditions. By comprises a maximum of 0.001 mol % of the product.
keeping reactant concentrations lower than 5%, the As such, the excellent material balance that was ob-
effect of any temperature changes due to the heat of served was valid for all experiments.
reaction are minimized, if not eliminated (Hawboldt,
1998). Figure 2 shows the experimental temperature New Experimental Results
profile for the 3.05 m reactor as a plot of the deviation Pyrolysis. Figures 3 and 4 present the conversion of
from the furnace setpoint temperature as a function of NH3 with temperature and residence time for the
length. This figure clearly shows that the temperature ammonia pyrolysis experiments. As the figures show,
rises rapidly to within 30 °C of the setpoint within the between 850 and 1050 °C, the conversion of NH3 was
first 25 cm of length, after which the deviation falls below 25%. At 1150 °C and residence times below 300
rapidly, being within 1.2% at 1.0 m. ms, conversions are less than 20%; however, at longer
Experimental Conditions. Pyrolysis. Our experi- residence times, conversions increase to 55%. At 1200
mental conditions were at much lower temperatures °C, conversions increased substantially and varied from
than those of the Davidson et al. (1990) study. Thirty- 27 to 83%. Material balances were verified by comparing
seven pyrolysis experiments were performed using NH3 conversion measurements with N2 and H2 compo-
temperatures between 850 and 1200 °C, NH3 concentra- sition measurements. These measurements agreed to
tions from 0.5 to 2.0%, and residence time ranges from within 10%, indicating a good material balance. Clark
50 to 800 ms (Hawboldt, 1998). Helium was the gas et al. (1998) showed similar results, in that at residence
chromatograph (GC) carrier gas for the majority of the times from 0.6 to 1.0 s conversions varied from 10% at
experiments; however, to measure the produced N2 and 900 °C to 75% at 1200 °C.
H2 and close the material balance, the carrier gas was Oxidation. Figures 5 and 6 show NH3 conversions
switched to argon. as a function of residence time and temperature for the
Oxidation. Twenty-one NH3 oxidation experiments oxidation experiments. In general, when compared to
were performed. Temperatures were varied between 850 pyrolysis, conversion of NH3 via oxidation is faster and
Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001 147

Figure 3. Experimental and predicted conversion of NH3 in Figure 5. Experimental and predicted conversion of NH3 in
pyrolysis experiments as a function of temperature and residence oxidation experiments as a function of temperature and residence
time. time.

Figure 4. Experimental and predicted conversion of NH3 in Figure 6. Experimental and predicted conversion of NH3 in
pyrolysis experiments as a function of temperature and residence oxidation experiments as a function of temperature and residence
time. time.

occurs to a greater extent for the same temperature. For simulated the reactor. This reactor model contained the
example, at 850 °C, conversion by pyrolysis never proposed kinetic expressions for the reactions, and the
exceeds 20%, whereas at the same temperature, conver- parameters of the kinetic model were regressed using
sion by oxidation is observed at nearly double this value. our experimental data. Details of the modeling and
Direct comparison with the data generated by Lyon regression procedure are provided in Hawboldt (1998).
and Benn (1978) is difficult because their experiments Reactor Model. The reactor simulation numerically
were performed using a large excess of O2. When modeled the experimental apparatus from the inlet
compared to the Clark et al. (1998) data, our NH3 through to the quench, solving a set of differential
conversions were significantly greater at lower temper- equations to determine the reactor product distribution.
atures. Clark et al. (1998) predict conversions of 4-10% The key assumptions in the development of the model
at residence times from 0.34 to 0.55 s in the temperature are plug flow in the reactor, steady-state operation, and
range from 700 to 1100 °C, whereas our conversions in ideal gas behavior. The validity of the plug-flow as-
this same range were 2-60%. The formation of NO in sumption was previously discussed. Because the ap-
these experiments could not be substantiated because paratus operates at high temperatures and close to
of the difficulty in analyzing for this component; how- atmospheric pressure, the assumption of ideal gas
ever, equilibrium calculations suggest negligible quanti- behavior is also valid, given compressibility factors
ties to be present. greater than 0.99.
The set of differential equations is comprised of the
Modeling Methodology plug-flow reactor design (eq 4), the reaction rate equa-
The determination of the new rate expression was tion (eq 5) in terms of reacting species partial pressures,
accomplished by developing a mathematical model that and an equation for the reactor pressure drop. The
148 Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001

Arrhenius law (eq 6) represents the specific rate con-


stant.

dF/dV ) rj (4)

r ) kΠjPjm (5)

k ) Ae-Ea/RT (6)
In addition, an equation of the reactor axial temperature
profile was included, which was established using
experimental data obtained from a temperature calibra-
tion reactor (Hawboldt, 1998). Pure component viscosity
was calculated using Chung’s method, and the mixing
rules of Wilke were used for the prediction of gas
mixture viscosity (Reid et al., 1987). The set of algebraic
and differential equations that describe the physical
system were solved using an adaptive step-size Runge-
Kutta method taken from Press et al. (1992).
Optimization Method. Simulated annealing, de-
tailed in Bohachevsky et al. (1986), was used to mini-
mize the objective function:

OF ) ∑|xp,i - xm,i| (7) Figure 7. Arrhenius plot for NH3 pyrolysis.

where xp,i is the predicted exiting gas conversion and


xm,i is the experimental (measured) gas conversion. The
optimization method uses the reactor model to calculate
the exiting reactor conversions, which are compared to
the experimental conversions. The temperature, pres-
sure, and residence time were set at the beginning of
each experiment, and the independent parameters used
to minimize the objective function were the Arrhenius
constant (A) and activation energy (Ea).
Pyrolysis Modeling Results. Reaction Order.
Previous studies indicated that NH3 conversions under
pyrolysis conditions are small; however, equilibrium
calculations show 100% conversion of NH3 to N2 and
H2, indicating that the pyrolysis of ammonia (eq 1) is
not equilibrium limited at Claus FEF temperatures
(Hawboldt, 1998). Hence, the following rate equation is
proposed for FEF conditions:

r ) kPNH3m (8)

The order and initial estimates of A and Ea were made


prior to parameter optimization. The reaction order (m)
was determined by performing experiments where the
temperature and residence time were kept constant Figure 8. Clark et al. (1998) NH3 conversions compared to model
while the inlet concentration of NH3 was varied from predictions. The Clark et al. data is based on a residence time (J)
0.5 to 2.0%. By plotting inlet versus exit NH3 concentra- range of 0.6-1.0 s.
tions, we calculated the value of m to be 1.25. This value
was verified using the parameter optimization program. between the model predictions and experimental data.
On the basis of the derived form of the rate equation, The average absolute error (AAE) between the model
an Arrhenius plot for NH3 pyrolysis is shown in Figure and data is 13%.
7. The slope of the line is -11.0, yielding an Ea of 21.0 Data generated by Clark et al. (1998) was also
kcal mol-1, which was used as an initial estimate for compared to model predictions, and the results are
the optimization. shown in Figure 8. The experimental conditions of the
Parameter Optimization Results. In the optimiza- Clark et al. (1998) data were an inlet NH3 concentration
tion, parameters A, Ea, and m were optimized to match of 14.5% and a residence time range of 0.6-1.0 s. The
NH3 conversions. The resulting rate equation is shown proposed kinetic model was run at 0.6 and 1.0 s and, as
below as eq 9: shown in Figure 8, the data falls between these two
simulations, further verifying our model.
r ) Ae-Ea/RTPNH31.25 (9) Ammonia Oxidation Modeling. Reaction Order.
Equations 2 and 3 show the two primary oxidation
where A is 0.004 21 mol cm-3 s-1 atm-1.25 and Ea is 16.5 pathways for NH3. In this study, the focus has been on
kcal mol-1. Figures 3 and 4 show the comparison eq 2, and consequently, the experiments were run at
Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001 149

Figure 9. Natural logarithm of the regressed rate constant versus


inverse temperature for the NH3 oxidation reaction. Figure 10. Comparison of model predictions with Clark et al.
(1998) data for NH3 oxidation (residence time ) 0.34-0.55s).
conditions where eq 3 was minimized. The proposed
form of the rate equation is as in eq 10: this more accurately describes the chemistry of the
system. As such, the apparent change in the tempera-
r ) kPNH3mPO2n (10) ture dependence is due to the combined effect of pyroly-
sis and oxidation. Given the above conclusion that the
Because direct measurement of O2 is difficult with our temperature dependence of the rate constant follows the
apparatus, the reaction orders m and n were determined Arrhenius law, the form of the rate equation is as shown
by the method of excess reactants. That is, to determine in eq 12:
m, the O2 was kept in excess of the stoichiometric value
(0.75 O2/NH3). Although this ratio exceeds the quantity r ) kPNH3PO20.75 (12)
of O2 required to form NO, as stated previously,
negligible amounts of NO are formed. In determining Figures 5 and 6 compare the model predictions to the
the reaction order with respect to NH3, we varied the experimental data for the complete range of tempera-
NH3 levels from 0.44 to 1.35% at a temperature of 1050 tures. The AAE for the data set was 10%, indicating an
°C and a residence time of 70 ms. The reaction order of excellent agreement between the predicted and experi-
NH3 was then calculated by plotting the inlet concen- mental values.
tration against the outlet concentration. The model was compared with the data generated by
For the reaction order with respect to O2, n, the NH3 Fujii et al. (1981). In this study, the residence time
was kept constant at 1.37% and the O2 increased from required to consume 5% of NH3 was calculated as a
2.6 to 4.4% at a residence time of 720 ms and 950 °C. function of temperature. According to the study, for a
The order with respect to O2 was calculated via eq 11: temperature range of 909-1250 K, the residence time

( )
to reach 5% was between 10 and 20 ms. Given their ex-
NH3,exit
log ln ) n log O2 + constant (11) perimental conditions as inputs, our model predicts the
NH3,inlet same range of residence times, suggesting a good fit.
Our oxidation data are compared to the data of Clark
The calculated values of m and n were 0.98 and 0.72, et al. (1998) in Figure 10. It is immediately apparent
respectively. These values were set to 1.0 and 0.75 in that our measured conversions are much higher than
the oxidation kinetic rate expression. those of Clark et al. (1998). Our data follow an NH3
Parameter Optimization Results. Previous studies cracking trend until 800 °C, after which oxidation
(Fujii et al., 1981; Clark et al., 1998) have suggested becomes significant. Note that our cracking results are
that the temperature dependence of the rate constant very similar to those of Clark et al. (1998). According
changes at approximately 1050-1100 °C, indicating to Clark (personal communication, 2000), their oxidation
that the Arrhenius constant and activation energy must conversion data are much lower than even cracking
also change. To verify this result, individual rate until oxidation becomes significant at 1100 °C, because
constants were regressed at each temperature and of the inhibition effect of substantial amounts of water
plotted as shown in Figure 9. The linear relationship present. As such, it can be concluded that we did not
between ln(k) and inverse temperature indicates that experience the inhibition effect. It must be noted that,
there is no significant change in the temperature although the experimental feed concentrations and
dependence of the rate constant between 850 °C and temperature were similar, the apparatus of Clark et al.
1200 °C, a result that is in disagreement with previous (1998) is considerably different than ours with a ceramic
work by Fuji et al. (1981). It is worthwhile to note that reactor and a much slower quench. Thus, the reason
in the regression of the rate constants for the oxidation for the large difference in conversions could be at-
reaction, kinetic expressions for both ammonia pyrolysis tributed to the formation of larger amounts of water
and oxidation were included in the simulation, because prior to oxidation in their apparatus. However, this
150 Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001

needs to be investigated in future work, because it is Acknowledgment


beyond the scope of this paper.
The authors thank the following supporters of this
work: Shell Canada, British Oxygen Corp., Stork-
Conclusions Comprimo, HEC Technologies, and NSERC. In particu-
lar, the authors thank the Gas Research Institute for
In this paper, NH3 pyrolysis and oxidation were their continued support of this research project. In
studied at Claus FEF temperatures, pressures, and addition, we thank Alberta Sulfur Research for their
residence times. New experimental data were generated cooperation and contributions to our continuing studies.
in order to develop global kinetic rate expressions for
the pyrolysis and oxidation of NH3 at FEF conditions.
List of Symbols
At conditions similar to those in a Claus FEF,
ammonia oxidation can be much more significant than A ) Arrhenius constant
pyrolysis at temperatures of less than 1150 °C. In fact, D ) diameter (cm)
conversion of NH3 through oxidation varied between 1 Ea ) activation energy (kcal mol-1)
and 50% for temperatures between 850 and 950 °C and k ) reaction rate constant
50-700 ms, whereas pyrolysis conversions at these K ) equilibrium constant
conditions never exceeded 20%. The increase in NH3 m and n ) reaction order
conversion due to oxidation is even more significant at OF ) objective function
P ) pressure (atm)
temperatures between 1050 and 1200 °C, where the
R ) gas constant (kcal mol-1 K-1)
conversion varied from 25 to 90% at the same residence
r ) reaction rate (mol m-3 s-1)
times. T ) temperature (°C or K)
For ammonia pyrolysis, the newly developed rate t ) time (s)
expression is V ) elemental reactor volume (cm3)
xm,I ) experimental fractional conversion
r ) Ae-Ea/RTPNH31.25 xp,I ) predicted fractional conversion
z ) axial position (m)

where A is 0.004 21 mol cm-3 s-1 atm-1.25 and Ea is 16.5 Greek Symbols
kcal mol-1. The AAE between the model and the data τ ) residence time (s)
was 13%. The model was also used to simulate experi- Subscripts
ments performed by Clark et al. (1998), and the model
predictions agreed well with their data. l ) species
For ammonia oxidation, the following rate equation j ) species
was developed: Acronyms
AAE ) average absolute error
r ) kPNH3PO20.75 SLPM ) standard liters per minute

Literature Cited
where A and Ea were 4430 mol cm-3 s-1 atm-1.75 and
40.0 kcal mol-1, respectively. Model predictions and Bohachevsky, I. O.; Johnson, M. E.; Stein, M. L. Generalized
experimental data showed good agreement with an AAE simulated annealing for function optimization. Technometrics
of 10%. 1986, 28 (8), 209-217.
Clark, P. D.; Dowling, N. I.; Huang, M. Chemistry of the Claus
Although oxygen competition needs to be addressed Front-End Reaction Furnace. Hydrocarbon Reactions and the
in a more definitive manner to fully judge industrial Formation and Destruction of CS2. Proceedings of the Brimstone
recommendations for ammonia destruction, simulations Sulphur Recovery Symposium, Vail, CO, Sept 23-26, 1997.
based on our new rate expressions give preliminary Clark, P. D.; Dowling, N. I.; Huang, M. Ammonia Destruction in
indications of what minimum temperatures and resi- the Claus Furnace. Proceedings of the Brimstone 1998 Sulphur
Recovery Symposium, Vail, CO, Sept 15-18, 1998.
dence times are necessary to obtain 60 ppm of ammonia Culter, A. H.; Antal, M. J.; Jones, M. A Critical Evaluation of the
in the furnace effluent, as typically desired for ammonia Plug-Flow Idealization of Tubular-Flow Reactor Data. Ind. Eng.
destruction in the Claus FEF. Specifically, temperatures Chem. Res. 1988, 27 (4), 691-697.
need to be greater than 1200 °C with residence times Davidson, D. F.; Kohse-Hoinghaus, K.; Chang, A. Y. Pyrolysis
greater than 1.0 s or greater than 1250 °C with Mechanism for Ammonia. Int. J. Chem. Kinet. 1990, 22 (5),
residence times greater than 0.5 s. 513-535.
Duo, W.; Dam-Johansen, K.; Østergaard, K. Kinetics of the Gas-
In industry, it is generally accepted that the furnace Phase Reaction between Nitric Oxide, Ammonia and Oxygen.
should be designed with residence times greater than Can. J. Chem. Eng. 1992, 70 (5), 1014-1020.
0.8 s, and flame temperatures of 1300 °C or higher are Fookes, R. B. Development of a High-Temperature Reactor System
desired. As such, guidelines for actual conditions are to Study the Intrinsic Kinetics of Homogeneous Gas-Phase
somewhat more conservative to account for the incom- Sulpur Reactions. M.S. Thesis, Department of Chemical and
Petroleum Engineering, The University of Calgary, Calgary,
plete mixing and oxygen competition that occur in Claus Alberta, Canada, 1996.
plant FEFs. In addition, calculated results based on the Fujii, N.; Miyama, H.; Koshi, M.; Asaba, T. Kinetics of Ammonia
new rate expressions clearly show that ammonia py- Oxidation in Shock Waves. Eighteenth Symposium (Inter-
rolysis accounts for significant destruction. national) on Combustion, Aug 17-22, 1980; PUBL Combust
Finally, it should be noted that there appears to be Inst. 1981; 873-883.
Garside, J. E.; Phillips, R. F. Pure and Applied Chemistry; Pittman
an inhibition of ammonia pyrolysis when significant and Sons Ltd.: London, 1962; pp 751-755.
amounts of water are present and under certain experi- Goar, G. B. Improving Claus Plant Performance by Oxygen
mental conditions. This phenomenon requires further Enrichment. Proceedings of Sulphur 94 International Confer-
study to delineate its occurrence. ence, Tampa, FL, Nov 6-9, 1994; Vol. 1, pp 109-129.
Ind. Eng. Chem. Res., Vol. 40, No. 1, 2001 151

Grancher, P. Advances in Claus Technology. Hydrocarbon Process. Miller, J. A.; Bowman, C. T. Mechanism and Modelling of Nitrogen
1978, 57 (7), 155-160. Chemistry in Combustion. Prog. Energy Combust. Sci. 1989,
Hawboldt, K. A. Kinetic Modeling of Key Reactions in the Modified 15 (4), 287-338.
Claus Plant Front End Furnace. Ph.D. Thesis, Department of Miller, J. A.; Smooke, M. D.; Green, R. M.; Kee, R. J. Kinetic
Chemical and Petroleum Engineering, The University of Cal- Modelling of the Oxidation of Ammonia in Flames. Combust.
gary, Calgary, Alberta, Canada, 1998. Sci. Technol. 1983, 34, 149-176.
Hawboldt, K. A.; Monnery, W. D.; Svrcek, W. Y. A Study on the Monnery, W. D.; Svrcek, W. Y.; Behie, L. A. Modelling the Modified
Effect of Quench Design on the Quality of Experimental Data. Claus Reaction Furnace and the Implications on Plant Design
Ind. Eng. Chem. Res. 1999a, 38 (6), 2260-2263. and Recovery. Can. J. Chem. Eng. 1993, 71 (5), 711-724.
Hawboldt, K. A.; Monnery, W. D.; Svrcek, W. Y. New Experimental Nauman, E. B. The Residence Time Distribution for Laminar Flow
Data and Kinetic Rate Expression for H2S Cracking and Re- in Helically Coiled Tubes. Chem. Eng. Sci. 1977, 32 (3), 287-
Association. Chem. Eng. Sci. 1999b, 55 (5), 957-966. 294.
Johnson, J. E.; Rempe, M. L. Reaction Furnace Design and
Press, W. H.; Vetterling, W. T.; Teukolsy, S. A.; Flannery, B. P.
Operation for Processing Ammonia Bearing GasessA Review
Numerical Recipes in CsThe Art of Scientific Computing, 2nd
of Fundamentals and a Case Study. Proceedings of the Brim-
ed.; Cambridge University Press: Cambridge, U.K., 1992.
stone 1997 Sulphur Recovery Symposium, Vail, CO, Sept 23-
26, 1997. Reid, R. C.; Prausnitz, J. M.; Poling, B. E. The Properties of Gases
Lindstedt, R. P.; Selim, M. A. Reduced Reaction Mechanism for and Liquids, 4th ed.; McGraw-Hill: New York, 1987.
Ammonia Oxidation in Premixed Laminar Flames. Combust. Truesdell, L. C.; Adler, R. J. Numerical Treatment of Fully
Sci. Technol. 1994, 99 (4-6), 277-298. Developed Laminar Flow in Helically Coilded Tubes. AIChE J.
Lindstedt, R. P.; Lockwood, F. C.; Selim, M. A. Detailed Kinetic 1970, 16 (6), 1010-1015.
Modeling of the Chemistry and Temperature Effects on Am- Vandooren, J. Comparison of the Experimental Structure of
monia Oxidation. Combust. Sci. Technol. 1994, 99 (4-6), 253- Ammonia Seeded Rich-Hydrogen-Oxygen-Argon Flame with
276. Calculated Ones Along Several Reaction Mechanisms. Combust.
Lyon, R. K.; Benn, D. Kinetics of the NO-NH3-O2 Reaction. 17th Sci. Technol. 1992, 84 (1-6), 335-344.
Symposium (International) on Combustion, Leeds, England, Aug
20-25, 1978; pp 601-610. Received for review October 22, 1999
Maclean, D. L.; Wagner, H. Gg. The Microstructure of the Reaction Revised manuscript received September 27, 2000
Zones of Ammonia-Oxygen and Hydrazine Decomposition Flames. Accepted September 27, 2000
11th Symposium (International) on Combustion, Berkeley, CA,
Aug 14-20, 1967; pp 871-878. IE990764R

You might also like