Article

pubs.acs.org/IECR

Ammonia Synthesis at Reduced Pressure via Reactive Separation

Mahdi Malmali,† Yongming Wei,‡ Alon McCormick,† and Edward L. Cussler*,†
†
Department of Chemical Engineering and Materials Science, University of Minnesota, 421 Washington Ave. SE #151, Minneapolis,
Minnesota 55455, United States
‡
State Key Laboratory of Chemical Engineering, Membrane Science and Engineering R&D Lab, Chemical Engineering Research
Center, East China University of Science and Technology, 130 Meilong Rd., Shanghai 200237, China

ABSTRACT: Ammonia is normally made at high temperature and pressure

using a promoted iron catalyst. High temperatures are needed to get fast
kinetics; the high pressure is used to ensure high conversion. Alternatively,
ammonia can be made at high temperature but lower pressure if the product
ammonia is rapidly separated. Here, we have systematically studied the effect of
temperature and pressure on the rates of reaction. We then have qualitatively
investigated the absorptive separation of ammonia using calcium chloride in a
reaction−separation process. Rapid separation reduces the constraint of
reversible reaction and enables us to obtain appropriate reaction rates at
relatively lower pressure. The effect of different operating conditionsreaction
temperature, pressure, absorption temperature, and gas transporton
production rates is carefully measured, and this elucidates the potential and
the limits of this type of low-pressure ammonia synthesis.

■ INTRODUCTION
Those in the global chemical enterprise now agree that the
ammonia-based fertilizer is needed. Thus, by working on
ammonia, we are investigating a possible chemical synthesis
future of the chemical industry is dependent on becoming more valuable both as a carbon-neutral liquid fuel4,5 and as a
sustainable.1 Such a goal includes developing energy sources fertilizer.6
that are not based on nonrenewable fossil fuels, and that do not We have begun to investigate making ammonia from wind
release large quantities of carbon dioxide. This goal implies using a small ammonia plant operating at the West Central
improving the efficiency of energy collected from the sun and Research and Outreach Center (WCROC) in Morris, MN.6
the wind. Solar energy is continuing to get less expensive, but is This plant lets us explore what parts of a conventional process
still more expensive than fossil fuels. Wind energy is more work well, and what parts are potential problems. The plant
immediately attractive and is being rapidly developed in areas of uses a reactor, a condenser, and a compressor, all of which are
high population density, especially in northern Europe.2 scaled down parallel to a conventional large-scale plant. The
However, both wind and solar energy are periodic and, reaction rates observed in the plant are consistent with parallel
hence, may not be available at times of highest demand. No one laboratory experiments with the same catalyst, and with values
wants light only in the middle of a sunny day; similarly, no one from the literature.7−10 Analysis of this process gives three
wants power only when the wind is blowing. Moreover, wind characteristic times: one for the reaction, one for the
and solar productions are distributed geographically, with condensation, and one for the pump recycling unreacted
practical limitations in supplying the electrical grid in remote gases. As the process is currently operated, the reaction time is
locations. Thus, both of these resources must be coupled to largest, so that the small process is currently controlled by the
methods of energy storage, especially as liquid fuels. Methanol chemical kinetics. However, the analysis also shows how
and ammonia are two such fuels. Methanol is attractive because process changes could increase productivity until one of the
it meshes well with the existing infrastructure for distributing other characteristic times became largest, and hence the limit of
fuel. Ammonia is interesting because it can be made from production. Thus, this plant supplies a test for process changes
totally renewable resources: nitrogen produced from pressure that may make sense for a small-scale plant, even when they are
swing adsorption of air, hydrogen produced from electrolysis of known to have limited effectiveness in a large fossil-fuel-based
water, and electricity obtained from stranded solar arrays or synthesis.
wind turbines. A major advantage of ammonia over methanol as In this paper, we explore the potential of process changes to
a fuel is that ammonia does not require a carbon source for the reduce the pressure needed for production. In seeking these
synthesis.
We are especially attracted by ammonia because much of the Received: May 16, 2016
potential wind power is far from areas of high population. As a Revised: July 24, 2016
result, wind is sometimes called “stranded energy”.3 At the Accepted: July 25, 2016
same time, much of the stranded wind energy is where Published: July 25, 2016

© 2016 American Chemical Society 8922 DOI: 10.1021/acs.iecr.6b01880

Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

process changes, we have decided not to focus on the reaction breakthrough of the absorber’s packed bed, we have a perfect
chemistry, feeling that this has been so carefully studied over separation of ammonia from the gas mixture. Hence, before
the last century that the chances of big improvements at any breakthrough, the absorption resistance is effectively zero.
scale are small.11−14 Instead, we are investigating three other However, after breakthrough, the production rate can become
aspects of this process. First, we are investigating synthesis at dominated by the increasing absorption resistance.
lower pressure, because such pressure could reduce process The presence of three resistances in series is observed in
complexity and make a small-scale process less expensive and many rate processes, where the amount produced is propor-
less hazardous. Second, we are investigating ammonia-selective tional to the overall driving force divided by the total
absorbents to reduce the need for recycle. Third, we are resistance.23 The total resistance is the sum of the resistances
beginning to explore processes that avoid recycle altogether. of reaction, absorption, and recycle. Phrased in other terms, the
In this paper, we limit the discussion just to the first aspect: total resistance is a harmonic average of the speeds of these
studies of lower pressure synthesis. After a summary of the three steps, so that the slowest speed has the biggest effect on
process itself, we report measurements of chemical kinetics to the rate of ammonia production. If the reactor is operating at a
show that at low pressures and a range of temperatures, we get temperature that is too low, the reactor resistance will be
the same results as others working under these conditions. We dominating: if the absorber is undersized, the absorption
then report the measurements of ammonia production in a resistance can be largest, and if the recycle rate is too low to
reaction-separation process. Calcium chloride absorbents are take full advantage of the other unit operations, the recycle
employed, where absorption occurs not just on the surface of resistance will be most important. We will use this model later
the solid, but also by diffusion into the solid. Absorption and in this paper to discuss the experiments and, hence, to explore
desorption15 of ammonia on chlorides of alkaline-earth metals how the reaction−separation process can be made more
have been studied in numerous applications, such as chemical productive.
heat pumps,16 ammonia storage,4,17,18 indirect hydrogen Linearized Chemical Reaction Rate Constant. Before
storage,19,20 and enhanced ammonia synthesis.21 We use the we make this analysis, we want to review how the linearized
measured combination of reaction and absorption to estimate reaction rate constant can be estimated from earlier studies.
the feasibility of ammonia synthesis at lower pressures. The reaction rate is most often correlated using the Temkin−
Overall Production Rate. Here, we employ a model that Pyzhev8,9,24 equation:
we developed in our earlier work6 for the reaction−separation
process with recycle, which guides the analysis of our
⎛ P P 1.5 ⎞ ⎛ P ⎞
experiments. This theory helps us to understand the complex
r = k1⎜⎜ 2 2 ⎟⎟ − k 2⎜⎜ A1.5 ⎟⎟
N H
behavior of the reaction−separation process, and we feel that
⎝ PA ⎠ ⎝ PH2 ⎠ (2)
this theoryspecifically, the time constants of the reaction,
separation, and transportmakes it easier to compare the where k1 and k2 are forward and reverse reaction rate constants,
findings of our reaction−separation process. In the following, respectively; and PN2, PH2, and PA are the partial pressures of
we give more details about the model that eases the
nitrogen, hydrogen, and ammonia, respectively. This equation
comprehension of the theory and the model. More details on
is rewritten by defining the following variable:
this model can be found elsewhere.6
In this simplified model, we assume that the process is in PA* − PA
steady state, and that the reactor and absorber are each well- X=
PN*2 (3)
mixed. (The steady-state assumption is accurate only at the
beginning of our experimental test.) The well-mixed Linearization using the Taylor series for small values of X
assumption implies that the average nitrogen concentration in simplifies this expression:
the reactor is equal to the nitrogen concentration at the exit.22
While this is untrue if most of the nitrogen entering the reactor ⎛ P* P* 1.5 ⎞⎡ PN* ⎤
9
r = k1⎜⎜ 2 2 ⎟⎟⎢ + 2 ⎥(X − X *)
N H
reacts in a single pass, it is much closer to being almost true
⎠⎢⎣ 4 PA* ⎥⎦
*
when only a small fraction of the nitrogen reacts per pass, ⎝ PA
which will be the case here. However, this will not interfere
⎛ * ⎞⎡ P * ⎤
with our findings, as will be discussed later. The model predicts P 1
+ k 2⎜⎜ A1.5 ⎟⎟⎢ 2 − ⎥(X − X *)
N
the rate of ammonia production (a) to be defined as
⎝ PH*2 ⎠⎢⎣ PA 4 ⎥⎦
*
xA* − xA0
a= = kR′ (X − X *) (4)
1 1 1 − xA*
+ + (1)
kR VR kabPS
t m
where kR′ has the dimension of moles of ammonia per catalyst
where x*A and x0A are the mole fractions of ammonia at reaction volume per time. The kR′ parameter is related to kR because
equilibrium and in the absorber, respectively; kR and kab are the ⎡ ⎤
PA* − PA xA
linearized chemical reaction rate constant and the absorption X= = 4⎢xA* − 2
⎥
mass-transfer coefficient, respectively; Pt is the operating PN*2 ⎣ (1 − xA*) ⎦ (5)
pressure; S is the surface area of the absorber; and m is the
total molar flow rate. In this result, the term (1/kRVR) is the Hence, when the conversion is close to zero, which is the case
resistance to ammonia production due to the chemical reaction. in our analysis, then
Similarly, (1/kabPtS) and (1 − x*A )/m are the resistances of the kR′ ≈ 4kR (6)
absorber and of the recycle loop, respectively.
In this model, the absorber’s performance is assumed to where kR′ is the corrected linearized reaction rate constant after
behave as a first-order rate process. In reality, before the the change of variables. Equation 4 shows that the linearized
8923 DOI: 10.1021/acs.iecr.6b01880
Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

Figure 1. Schematic diagram of the apparatus. The gas mixture is recirculated between the reactor and absorber using a piston pump.

reaction rate constants kR′ and kR obtained here are functions of controllers were connected to a National Instrument Labview
concentration. When xA approaches x*A , k′R decreases to zero; card (Austin, TX), with the data being recorded every second.
however, when xA approaches zero, kR′ increases to its Labview controlled the mass flow controllers (Model
maximum value, representing only the forward reaction rate. SLA5850S, Brooks Instruments, Hatfield, PA) installed in the

■
apparatus inlet, which injected known volumetric flows of
EXPERIMENTAL SECTION reactant gases. The inlet and outlet of the system were
equipped with needle valves.
Materials. Anhydrous CaCl2 in granular form (>7 mm)
Procedure. Three grams of catalyst, loaded in the reactor at
with 93% purity (CAS No. 10043-52-4, Lot No. SLBL2770 V)
room temperature, was reduced by pumping hydrogen gas
was purchased from Sigma−Aldrich (St. Louis, MO). The
through the catalyst bed while slowly increasing the reactor
reactant gases N2 and H2 with ultrahigh purity were purchased
temperature. Hydrogen gas was pumped through the catalyst
from Matheson (New Brighton, MN). We employed a
bed at the flow rate of 0.5 standard liters per minute (SLPM).
prereduced nonstoichiometric ferrous oxide catalyst (wustite)
The hydrogen reacted with the coating layer, forming water
with promoters (AmoMax-10 RS, Clariant, Charlotte, NC).
The catalyst is provided in the form of irregularly shape vapor. A multiramp temperature profile heated the catalyst very
granules, with a nominal size range of 1.5−3 mm, and is slowly to ensure that the water vapor concentration during the
stabilized with an oxygen-rich protective layer. activation is <1000 ppm. The temperature ramp required ∼27
Apparatus. The experimental apparatus, a schematic h for the temperature to increase from room temperature to
diagram of which is shown in Figure 1, was built using 723 K. After the reactor reached 723 K, hydrogen flow
Swagelok (Chaska, MN) 316 stainless steel tubing and fittings continued through the reactor for at least 24 h. After this initial
of 6 mm inner diameter. The reactor was 0.15 m long. Catalyst reduction, the apparatus was always kept under nitrogen.
particles were ground from their produced size until smaller We performed two sets of experiments. During the first set,
(<1 mm) before loading the reactor. The absorber was a 0.3-m- which we refer to as “reaction experiments”, we were only
long stainless steel tube, with an inner diameter of 0.022 m, interested in the kinetics of ammonia synthesis, so the absorber
bundled with heating tape. The absorber and heating tape was bypassed and the gas mixture circulated only through the
bundle were insulated with silica woven insulation (AVS reactor. We investigated the effect of operating pressure on the
Industries, New Castle, DE). A variable piston pump (Model reaction rate constants. In the second set of experiments, which
PW2070N, PumpWorks, Inc., Minneapolis, MN) circulated the we refer to as “reaction−separation experiments”, the gas
gas mixture between the reactor and the absorber. mixture was circulated between the reactor and absorber using
A ceramic heater (Model CRFC-36/115-A, Omega, the pump. This set of experiments was similar to the
Stamford, CT), equipped with a multiramp proportional− conventional Haber−Bosch process, except that we replaced
integral−differential (PID) controller (Omega, Model the condenser with an absorber. In both sets of experiments, we
CN96211TR) controlled the temperature of the reactor. The utilized the pressure reading data log to calculate the ammonia
inlet and outlet gas temperatures and the temperature on the reaction rates and conversion.
outer surface of the reactor were measured using Type K The reaction experiments utilized a three-way valve to direct
thermocouples connected to a signal conditioner (Omega, the circulation loop toward the bypass line, without the
Model DRG-SC-TC). The system pressure was recorded using absorber. The catalyst remained unchanged throughout the
a pressure transducer (Model 50426877, WIKA, Lawrenceville, experiments. The hydrogen and nitrogen were fed to the
GA) with a 0−10 V dc output. The experiments were carried reactor with the ratio of 3:1. We stopped the flow and closed
out in a recirculation batch mode. All instruments and the inlet valve once the system reached to the desired pressure.
8924 DOI: 10.1021/acs.iecr.6b01880
Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

After pump started to recirculate the gas mixture with the flow of ammonia reaction at each operating temperature. The rate of
rate of 0.7 mL/s, we began to record pressure data. These reaction at 740 K is much faster than the rate at 660 K, and we
conditions are summarized in Table 1. shall determine if these rates are consistent with the literature.25
The reaction rate is initially fast, but quickly slows at longer
Table 1. Conditions for Reaction and Reaction−Separation times, because the reverse reaction rate accelerates as the partial
Experiments pressure of ammonia increases.
The half-lives of reaction, as an alternative representation of
Experimental Conditions the reaction rates, are shown as an Arrhenius plot versus
reaction reaction−separation temperature in Figure 2b. To determine the half-life from the
reaction temperature (K) 620−740 700 pressure−time profiles, the equilibrium conversion and final
absorption temperature (K) 460 pressure at each specific temperature were calculated, following
initial pressure (bar) 74 55 the analysis of Gillespie et al.25 The time at which the half-life
H:N ratio 3:1 3:1 operating pressure occurred then was extracted. The half-lives
pump flow rate (mL/s) 0.7 0.7 range from 0.15 h to 12 h. This variation in the half-life of
reaction/reaction-absorption time (h) 3 3 reaction is evident in the pressure profiles as well. For instance,
desorption temperature (K) 600 in the test at 740 K, the pressure reaches equilibrium after
purge flow (SLPM N2) 100 ∼2000 s, while the pressure at 660 K continues to decrease
desorption time (h) ∼20 until 5000 s. This affirms the common conclusion for this
system, that higher temperature is advantageous for reaction
rate, while lower temperature favors the thermodynamic
Unless otherwise mentioned, the reaction−separation experi- equilibrium.
ments were at 700 K. The three-way ball valve included the The operating pressure also affects the reaction kinetics, but
absorber in the recirculation loop. The absorber was initially to a lesser degree than the temperature does. The data in Figure
filled with ∼70 g of fresh anhydrous CaCl2 and was always kept 3, for a variety of temperatures and pressures, illustrate this.
at temperatures greater than 460 K. The absorber was purged
with dry nitrogen for 24 h at 673 K before starting the first test.
The absorber and catalyst were unchanged throughout the
experiments. An initial pressure of 55 bar was used, and the
apparatus was fed with hydrogen and nitrogen gas mixture with
the ratio of 3:1. The pump rates of 0.7 mL/s recirculated the
gas mixture for 3 h. After each test, the absorber was
regenerated by heating to 600 K, decreasing the system
pressure to 2−3 bar, and pumping nitrogen through the
absorber at 0.1 SLPM for 20 h. Details of these experiments are
given in Table 1.

■ RESULTS
Reaction Experiments without Absorption. To under-
stand the reaction kinetics with our catalyst, we carried out
several experiments to investigate the effect of operating Figure 3. Arrhenius plot of reverse rate of reaction versus 10 000/T.
temperature and pressure. For example, Figure 2a reports the Results indicate a strong dependency of the reverse rate constant on
pressure versus time for the catalytic reaction of ammonia temperature, but a smaller variation with pressure.
initially at 74 bar and reaction temperatures of 660 and 740 K.
The slopes of the pressure profiles represent the apparent rates This figure reports the reverse reaction rate constant in the
form of an Arrhenius plot. We chose reverse reaction rate
constants rather than our more directly measured forward rate
constants, because most previous literature8,26 on catalytic
reaction of ammonia is rather focused on disassociation of
ammonia, so more data are available on reverse rate constants.
The reverse reaction rate constant is simply proportional to the
forward reaction rate constant
k1
K p2 =
k2 (7)
and has the following stoichiometry:
1 3
N2 + H 2 ⇔ NH3
2 2 (8)
where Kp is the equilibrium reaction constant calculated from
Figure 2. Effect of reaction temperature on (a) kinetics of reaction and Gillespie et al.,25 which includes fugacities that can alter the
(b) half life of reaction. Higher temperature increases the rates of equilibrium reaction constant by a few percent.
reaction; however, at higher temperature, the mole fraction of The reverse reaction rate constants in Figure 3 exponentially
ammonia at equilibrium decreases. increase with temperature, but change less with pressure.
8925 DOI: 10.1021/acs.iecr.6b01880
Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

Specifically, the rate constant increases by a factor of 10 000 Hence, the absorption rates into the CaCl2 structure can
when the operating temperature is increased from 600 K to 730 change. Thus, we repeated the reaction−absorption and then
K; however, the rate constant at 650 K varies by only a factor of desorption cycles until we obtained reproducible results. The
2 when the operation pressure is increased from 20 bar to 69 data are illustrated in Figure 5a for selected cycles. The total
bar. At 740 K, there is an order-of-magnitude difference in the pressure, our measure of reaction−separation, is linear with
reaction rate constants at 20 and 69 bar, which is certainly time during the first cycle. The pressure changes increase
significant but less than the variation with temperature. Lower notably in subsequent cycles, consistent with a structural
operating pressures give larger reaction rate constants while change in the CaCl2. The initial production rates (slopes) are
higher operating pressures give lower ones. The tests at 20 and relatively fast, and the rates then go through a transition region;
28 bar show less variation, probably because the times to reach finally, they slow to a rate very similar to that of the first cycle
the operating pressures are smaller and more reproducible. for all cycles. The initial rates get faster after desorption of each
Nevertheless, the experimental errors for reaction engineering cycle, and take longer to slow down. If we plot the reciprocal of
in this range are typical (cf. Figure 4). the apparent ammonia capacity after 10 000 s versus the
reciprocal of the cycle number, we get a straight line, as shown
in Figure 5b. The time period of 10 000 s chosen here is
arbitrary, but it illustrates some of the implications of Figure 5a.
The intercept in Figure 5b corresponds to the inverse capacity
for an infinite number of cycles (that is, for the maximum
change in the solid geometry). In practice, we always used
samples subjected to seven or more cycles, which are thus
within ∼10% of the maximum absorption that is obtainable at
the operating temperature.
Effect of Reaction Temperature. To determine whether the
chemical reaction is the rate-limiting step, we carried out tests
with reaction temperatures of 680, 700, 720, and 740 K, while
keeping all other conditions unchanged. The pressure−time
graphs for these tests are shown in Figure 6. The results show
Figure 4. Comparison of the reverse reaction rate constants with
literature. Reverse reaction rate constants are independent of operating
pressure.

These reaction rate measurements are compared with those

reported in the literature27,28 in Figure 4. This shows an
Arrhenius plot of the Temkin reverse reaction rate constant k2,
both as reported in the literature and as found from this study.
Our work displays a systematic negative deviation, most likely
due to experimental error associated with the time to start our
experiment. Again, these data and those of others all show that
the reaction rate constants vary much less with pressure than
with temperature.
Reaction−Separation Experiments. Effect of Absorbent
History. Earlier work says that the absorption−desorption of Figure 6. Effect of reaction temperature on the production rates pf
ammonia on CaCl2 absorbent can generate a nanoporous reaction/absorption. The small effect on the overall kinetics of the
structure with a Brunauer−Emmett−Teller (BET) surface area reaction temperature shows that reaction kinetics is not the rate-
that is up to 10 times larger than was there originally.20,29 limiting step of the combined reaction−absorption.

Figure 5. Effect of absorption history of the CaCl2 on the production rate of reaction/absorption: (a) the production rate increases until the seventh
cycle, and (b) the accessible volume of the CaCl2 increases gradually.

8926 DOI: 10.1021/acs.iecr.6b01880

Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

that the initial overall rate is almost the same, independent of Figure 8a, may indicate that two competing mechanisms are
the temperature. Later, the rates at higher catalyst temperatures controlling the absorption of ammonia. First, the diffusion
are systematically slower. This implies that the overall kinetics coefficients in the solid CaCl2 increase with temperature.
are not controlled by catalytic reaction. It is also likely that, with Second, the capacity of the absorbent decreases with temper-
small reactor residence time and effective absorber operation, ature: for instance, the equilibrium absorption capacity of the
the reverse reaction for ammonia synthesis is not significant CaCl2 at 325 and 470 K at an ammonia partial pressure of 1 bar
because the ammonia mole fraction is kept low. is 4 and 1.5 mol of ammonia per mole of CaCl2, respectively.20
Effect of Initial Pressure. The initial rate of reaction is As Figure 8a shows, there is no significant effect of temperature
proportional to the system operating pressure, as shown in on absorption: the rate is greatest at 460 K, smaller at 510 K,
Figure 7. The operating pressure has two synergistic effects on and somewhat larger at 600 K. The apparent production rate
per capacity versus the absorber temperature shown in Figure
8b is approximately constant, implying that diffusion and
absorption capacity counterbalance. Earlier experiments on the
absorption of ammonia without reaction in magnesium chloride
also show only a weak absorption dependence on temper-
ature.30
Effect of Recycle Flow Rate. A key route to reducing the
total resistance to ammonia synthesis is to increase the fluid
velocity in the system, as shown in Figure 9. While the flow of
gas has an insignificant effect on the forward or reverse
chemical rate constants, it may affect both the absorption
resistance and the recirculation resistance. At the beginning of
the test, the packed-bed absorber works as a perfect separator.
Thus, the absorption time constant is equal to zero until the
breakthrough point. We can then infer that the film mass-
Figure 7. Higher initial operating pressure enhances the production transfer resistance around the particle is not important. We
rates of absorption and apparent capacity of the absorbent.
expect that the resistance to absorption will gain more
importance once the breakthrough occurs. As a result, we
rates of the reaction−separation experiments. First, the catalytic suspect that the resistance of the recycle flow, which is
reaction rate is higher at higher operating pressure. Second, the represented by the term (1 − x*A /m) in eq 1, influences the
larger ammonia concentration in the gas stream leaving the overall rate significantly. This is true: faster pumping accelerates
reactor increases the driving force for the absorption of the process, as Figures 9a and 9b show. Even increasing the
ammonia into CaCl2 crystals. At longer times, the behavior is flow rates by a factor of 12 does not yet suggest an asymptote in
more complex, for example, when the conversion curves for Figure 9b.
initial pressures of 55 and 69 bar cross. These complexities If we could maintain almost-complete removal of the
probably come from factors like the onset of reverse reaction ammonia in the absorber, increasing the recycle flow decreases
rate, increased mass-transfer resistance, and changes in the single-pass conversion of ammonia. In other words, faster
absorbent structure. This more-complex behavior is beyond recycle flow accelerates forward reaction and decelerates reverse
our current objective. Instead, we will focus on maintaining the reaction, because of the lower ammonia concentration in the
high initial rate seen here with appropriate choice of absorbent reactor. However, the fast recirculation rates required may
and recycle operation. compromise the reactor and absorber temperatures. In our
Effect of Absorber Temperature. The effect of absorber case, increasing the pump flow rates to more than 3 mL/min
temperature on the rate of ammonia production, displayed in reduced the reactor temperature because of the heat loss to the

Figure 8. Absorption temperature effect on production rates: (a) pressure versus time graphs at different absorption temperatures and (b) apparent
rates versus absorption temperatures. Higher absorption temperature increases the production rate constant but decreases the driving force. The
apparent rates and capacities shown are calculated for the initial 1000 and 10 000 s, respectively. (In panel (b), r′ [mmol NH3/s] represents the
production rate, and C is the apparent absorbent capacity [mmol NH3/mmol CaCl2].)

8927 DOI: 10.1021/acs.iecr.6b01880

Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

Figure 9. (a) Effect of pumping flow rates on production rates, and (b) initial rates of absorption in different pumping flow rates. Results indicate
that film resistance surrounding the particle is not important. The production rates shown are calculated from the initial 1000 s.

gas flowing out of the reactor, so none of those data are shown relatively fast ammonia synthesis rates (unchanged for up to
here. This must be a concern in future process design. 5000 s). This suggests that high production rates at pressures as
Figure 10 shows the production rates at different time low as 25 bar are viable when ammonia is removed from the
intervals for different pumping rates. Results suggest that during system efficiently. The absorption separates the synthesized
ammonia from the reaction environment and reduces the
reverse reaction.
However, while we have developed a reaction−separation
process with enhanced rates and conversion, we do not
understand the detailed performance of each unit operation.
We have investigated the impact of different operating
conditions. For instance, we find that the reaction temperature
has a minor effect on the process, while increased recycle
pumping shows significantly increased production. In the
following paragraphs, we discuss this performance in more
detail.
To do so, we employ the model summarized in the theory
above to understand the rate constants for each unit at different
operating conditions. The complete reaction conditions and
Figure 10. Effect of recycle flow rate on the production rate in
calculated times for each unit are summarized in Table 2. The
different time intervals of the test. Absorber column breakthrough first column in this table identifies the experimental category.
leads to the formation of the asymptote at different flow rates; larger Columns 2−5 identify the reaction conditions, including the
recycle flow results in faster absorber breakthrough. operating pressure, reaction temperature, absorption temper-
ature, and pumping flow rate. Columns 6 and 7 give the
the initial 1000 s, the transport resistance in eq 1, represented measured time constants for the reaction and the recycle
as (1 − xA*/m), is much larger than the reaction resistance in eq obtained from eq 1, respectively. When the absorber is
1, which is defined as [1/(kRVR)]. Because the regenerated removing all ammonia from the gas flow before the
packed bed absorber is removing the ammonia completely, the breakthrough of the bed, the absorption time constant is
absorption resistance is zero; therefore, the recycle flow is zero. The values of these three rate constants merit careful
controlling the reaction−separation process, which leads to the consideration, because they are the key by which the process
linear increase in the production rate. At times between 1000 can be improved. The longest time constant corresponds to
and 4000 s, the absorber’s packed bed breakthrough starts to slowest step in the reaction−separation and, hence, controls the
appear, so the absorber is no longer capable of complete overall rate of the process under the conditions studied.
removal of the ammonia from the gas stream. By contrast, at Table 2 shows that the recycle flow is the rate-controlling
longer times in the range of 4000−5000 s, we observe smaller step before breakthrough of the packed bed absorber. The time
production rates at higher recycle flows; this lower production constant for the recycle is 100 times larger than the reaction
rate is attributed to the appearance of the breakthrough point time constant. Thus, our reactive separation process can be
and (partial) exhaustion of the bed. Note that at higher recycle improved significantly by faster recycle flows. However, this is
flow, we had larger production rates, which led to faster loading not a trivial change, because controlling the operating
of the absorbents. This is consistent with our earlier parameters of the reaction−separation systems becomes
observation of the slower slopes at ∼5000 s. difficult.

■ DISCUSSION
The results shown above confirm the viability of the reaction−
These results still guide speculation about the design of a
small, efficient process which includes the synthesis and
simultaneous removal of the ammonia at reduced pressure. In
absorption process for the enhanced production of ammonia at the widely verified rate equations, there are three factors that
significantly lower operating pressures. For example, in one set directly affect the reaction rate: the temperature, which affects
of measurements, we obtained more than 80% conversion with all the reaction rate constants; the partial pressures of reactants,
8928 DOI: 10.1021/acs.iecr.6b01880
Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

Table 2. Examples of Process Conditions and Time Constantsa

pressure (bar) reaction temperature (K) absorption temperature (K) pumping flow rate (L/h) 1/(kRVR) (× 10−3 h/mol) 1/m (h/mol)
Absorption History Test
55 700 460 2.5 5.8 0.18
Reaction Temperature Test
55 680 460 2.5 16 0.18
55 700 460 2.5 5.8 0.18
55 720 460 2.5 2.2 0.18
55 740 460 2.5 1.1 0.18
Pressure Test
55 700 460 2.5 5.8 0.18
69 700 460 2.5 5.8 0.14
83 700 460 2.5 5.8 0.12
Absorption Temperature Test
55 700 460 2.5 5.8 0.18
55 700 510 2.5 5.8 0.18
55 700 600 2.5 5.8 0.18
Pumping Flow Rate Test
55 700 460 1 5.8 0.46
55 700 460 2.5 5.8 0.18
55 700 460 10 5.8 0.04
a
The largest time constant, for the absorption, is the step which most affects the overall rate.

Figure 11. Schematic diagram of a conventional reactor (top) and a reaction−separation reactor with absorption (bottom). In the reaction−
separation design, green, blue, red arrows shown represents the bed in process, transition, and regeneration, respectively.

which influence the forward reaction rate; and the partial Table 3. Experimental Conditions of the Nielsen Experiment
pressure of the ammonia, which influences the reverse rates. If
parameter value
we can remove ammonia from the reaction chamber efficiently,
then the disassociation of ammonia through the reverse P 304 bar
reaction no longer proceeds rapidly. Thus, we can obtain T 765 K
Vcat 2.5 cm3
respectable rates of reaction at reduced pressure if we keep the
Lcat 130 mm
concentration of ammonia low with an efficient separation.
space velocity 94200 1/h
We now compare the performance of a conventional reaction
outlet % NH3 19.7
with a reaction−separation process. One attractive design for
the reaction−separation would be mixed catalyst and absorber
in the same bed. However, the catalyst can be poisoned by
water and other impurities on the absorbent. Another attractive
ammonia per hour. We assume that the absorber works perfect
design would be a conventional reactor divided into short
and the gas stream leaving the absorption sections has no
segments with absorption beds between them. Our results ammonia.
show that faster pumping increase production, but higher To illustrate this, we use the rate equation proposed by
recycle flows may be difficult in practical applications. Still, the Nielsen et al.27 (eq 9) for the synthesis of ammonia. We choose
new process has promise. this form of rate equation because the reaction rate does not
To explore this promise, we compare the reactor used by diverge to infinity when the ammonia partial pressure is zero
Nielsen et al. with that proposed here using the specifics shown (although we acknowledge that the authors tested this rate only
schematically in Figure 11.27 From Neilsen’s extensive down to PA = 0.2 bar):
experiments, we choose the one summarized in Table 3,
which is capable of producing 1.8 moles of ammonia per hour. ⎛ PA 2 ⎞
k 20⎜PN2K a 2 − ⎟
In our simplified reaction−separation process, we can set the ⎝ PH2 3 ⎠
mole fraction of ammonia at the outlet of each catalytic section r=
⎡ ⎛ PA ⎞⎤
2α
to <1% and then determine the required initial operating ⎢⎣1 + K3⎜⎝ PH ω ⎟⎠⎥⎦
pressure, which satisfies the production rate of 1.8 moles of 2 (9)

8929 DOI: 10.1021/acs.iecr.6b01880

Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

Figure 12. Comparison of rate of ammonia synthesis and ammonia mole fraction along the length of the reactor. We need ∼17 catalytic sections,
followed by absorption, to maintain the concentration of ammonia at <1% in this design.

where Pi is the partial pressure of component i; Ka and K3 are divided by the total initial charge in the system) is shown. The
equilibrium and reaction rate constants; and α and ω are the black data points (square markers) represent data from the
parameters of ammonia reaction rate. In our case, we used α = reaction test at 723 K and an initial operating pressure of 55
0.64 and ω = 1.564. Other constants of this equation, in bar; red data points (circle markers) represent data from the
accordance with these parameters, can be found elsewhere.27 As result for the reaction test at 723 K and an initial operating
the equation shows, the concentration of ammonia product pressure of 55 bar; blue data points (rectangular markers)
does have a significant influence on the rate of ammonia represent data from the result for the reaction−separation test,
production. where the reaction occurs at 723 K, absorption occurs at 460 K,
Our data show that such a process is capable of producing and the initial operating pressure is 55 bar, and the pump was
1.8 mol of ammonia per hour at 25 bar, i.e., at pressures ∼12 also set to a value of 0.7 mL/s. The data points are averaged for
times lower than the pressure used in the Nielsen experiment. each 50 s. The slopes of these curves represent the apparent
Figure 12 shows the rate of ammonia production and mole rates of ammonia synthesis.
fraction of ammonia along the length of the reactor for two When the absorber is introduced to the batch−recycle−loop
example designs. system, the rate of production remains comparable to the initial
As shown, the rate of ammonia production in conventional rate seen in the absence of the absorber. The absorber is
reactor rapidly decreases at the entrance of the reactor. By the managing to keep the product concentration above the catalyst
time the mole fraction of ammonia reaches 5%, the rate of low, so the rate is comparable to just the forward rate. Doing
reaction decreased significantly. Alternatively, if we alternate so, it permits us to attain rates comparable to just the forward
catalytic and absorption sections, the reaction rate in our rate. Figure 13 clearly indicates that, after a short period, the
simplified reaction−separation process is high while operating reaction−separation test system is showing faster apparent
at 25 bar. In principle, we can make ammonia rapidly and at rates, compared to the reaction test system. While during the
lower pressure. We look forward to exploring the practical initial stage (t ̅ = 100 s/mol), the reaction test shows the fastest
utility of the science that we have now demonstrated. rate at 90 bar, the reaction test at 55 bar displays a slower rate,
The proof-of-the-concept for the viability of enhanced similar to that observed in the reaction−separation test.
ammonia synthesis at reduced pressure via reaction−separation However, the apparent rates change significantly after a while
process is displayed in Figure 13. The dimensionless pressure (t ̅ > 100 s/mol). Comparison of reaction test at initial pressures
(P/P0) versus normalized time (t,̅ which is defined as time of 90 and 55 bar with the reaction−separation test at an initial
pressure of 55 bar shows that the rates of the reaction tests
decrease very quickly while the apparent rate of the reaction−
separation test remains unchanged. For instance, at t ̅ = 2500 s/
mol, the apparent rate of the reaction−separation test is four
times larger than the apparent rate of the reaction test, although
the operating pressure is almost half. At t ̅ > 5000 s/mol, the
reaction tests approach equilibrium while the apparent rate of
the reaction−separation test is still unchanged at pressures as
low as 25 bar. While the concept of enhanced ammonia
synthesis using metal chlorides has been demonstrated in our
earlier work,25 here, we show the viability of ammonia synthesis
at much reduced pressure with comparable rates.

■ CONCLUSIONS
In this study, we have reported the application of a reactive−
Figure 13. Comparison of reaction and reaction−separation tests at absorption process for enhanced production of ammonia at
different operating pressures. Greater apparent reaction rate is reduced operating pressure. Absorption on calcium chloride
observed via the reaction−separation test at lower operating pressures. provides an efficient separation for the removal of ammonia,
8930 DOI: 10.1021/acs.iecr.6b01880
Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

even at temperatures close to the reaction temperature. Both (5) Wojcik, A.; Middleton, H.; Damopoulos, I.; Van Herle, J.
our analysis and model predictions indicate that both reaction Ammonia as a Fuel in Solid Oxide Fuel Cells. J. Power Sources 2003,
and separation rates are fast and recycle flow is the rate-limiting 118, 342.
step in our proposed process. (6) Reese, M.; Marquart, C.; Malmali, M.; Wagner, K.; Buchanan, E.;

■
McCormick, A.; Cussler, E. L. Performance of a Small-Scale Haber
Process. Ind. Eng. Chem. Res. 2016, 55, 3742.
AUTHOR INFORMATION (7) Dyson, D. C.; Simon, J. M. Kinetic Expression with Diffusion
Corresponding Author Correction for Ammonia Synthesis on Industrial Catalyst. Ind. Eng.
*Tel.: 612-625-1596. Fax: 612-626-7246. E-mail: cussler@umn. Chem. Fundam. 1968, 7, 605.
(8) Temkin, M.; Pyzhev, V. Kinetics of Ammonia Synthesis on
edu.
Promoted Catalysts. Acta Physiochim. USSR 1940, 12, 327.
Notes (9) Annable, D. Application of the Temkin Kinetic Equation to
The authors declare no competing financial interest. Ammonia Synthesis in Large-Scale Reactors. Chem. Eng. Sci. 1952, 1,

■
145.
(10) Horiuti, J.; Takezawa, N. The Mechanism of Catalyzed
ACKNOWLEDGMENTS
Synthesis of Ammonia in the Presence of Doubly Promoted Iron
This work was primarily supported by the Minnesota Catalyst. J. Res. Inst. Catal., Hokkaido Univ. 1960, 170−187.
Environment and Natural Resources Trust Fund as recom- (11) Stoltze, P. Surface Science as the Basis for the Understanding of
mended by the Legislative-Citizen Commission on Minnesota the Catalytic Synthesis of Ammonia. Phys. Scr. 1987, 36, 824.
Resources (LCCMR), and by the MNDrive, an initiative of the (12) Ertl, G. Primary Steps in Catalytic Synthesis of Ammonia. J. Vac.
University of Minnesota. Joshua Prince is also acknowledged Sci. Technol., A 1983, 1, 1247.
for assisting with the rate constant calculations. (13) Ertl, G. Surface Science and CatalysisStudies on the

■
Mechanism of Ammonia Synthesis: The P. H. Emmett Award
Address. Catal. Rev.: Sci. Eng. 1980, 21, 201.
NOMENCLATURE (14) Catalytic Ammonia Synthesis: Fundamentals and Practice;
a = rate of ammonia production (mol/h) Springer Science & Business Media: New York, 2013.
C = apparent absorbent capacity (mmol NH3/mmol CaCl2) (15) Jones, M. O.; Royse, D. M.; Edwards, P. P.; David, W. I. F. The
k = absorption initial rate Structure and Desorption Properties of the Ammines of the Group II
k1 = forward rate of reaction constant Halides. Chem. Phys. 2013, 427, 38.
(16) Lebrun, M.; Spinner, B. Models of Heat and Mass Transfers in
k2 = reverse rate of reaction constant
Solidgas Reactors Used as Chemical Heat Pumps. Chem. Eng. Sci.
kab = absorption mass-transfer coefficient (mol/m2 Pa h) 1990, 45, 1743.
kR = linearized catalytic reaction rate constant (mol/L h) (17) Sharonov, V. E.; Aristov, Y. I. Ammonia Adsorption by MgCl2,
Ka = equilibrium constant in Nielsen rate expression CaCl2 and BaCl2 Confined to Porous Alumina: The Fixed Bed
Kp = reaction equilibrium constant Adsorber. React. Kinet. Catal. Lett. 2005, 85, 183.
K3 = reaction rate constant (18) Liu, C. Y.; Aika, K. Ammonia Absorption into Alkaline Earth
m = number of moles of gas mixture per hour in the recycle Metal Halide Mixtures as an Ammonia Storage Material. Ind. Eng.
flow (mol/h) Chem. Res. 2004, 43, 7484.
Pi = partial pressure of component i (19) van Hassel, B. A.; Karra, J. R.; Santana, J.; Saita, S.; Murray, A.;
S = total surface area of the absorber Goberman, D.; Chahine, R.; Cossement, D. Ammonia Sorbent
r = rate of reaction Development for On-Board H2 Purification. Sep. Purif. Technol.
r′ = apparent production rate (mmol NH3/s) 2015, 142, 215.
VR = volume of catalyst (m3) (20) Sørensen, R. Z.; Hummelshøj, J. S.; Klerke, A.; Reves, J. B.;
Vegge, T.; Nørskov, J. K.; Christensen, C. H. Indirect, Reversible
x*A = ammonia mole fraction at equilibrium
High-Density Hydrogen Storage in Compact Metal Ammine Salts. J.
xi = mole fraction of component i in the recycle stream
Am. Chem. Soc. 2008, 130, 8660.
x0A = mole fraction of ammonia in the absorber (21) Himstedt, H. H.; Huberty, M. S.; McCormick, A. V.; Schmidt, L.
Subscripts D.; Cussler, E. L. Ammonia Synthesis Enhanced by Magnesium
i = component i Chloride Absorption. AIChE J. 2015, 61, 1364.
A = ammonia (22) Kumar, A.; Daoutidis, P. Nonlinear Dynamics and Control of
H = hydrogen Process Systems with Recycle. J. Process Control 2002, 12, 475.
(23) Baldea, M.; Daoutidis, P. Dynamics and Nonlinear Control of
N = nitrogen
Integrated Process Systems; Cambridge University Press: Cambridge,
α, ω = parameters of the ammonia reaction rate

■
U.K., 2012.
(24) Guacci, U.; Traina, F.; Ferraris, G. B.; Barisone, R. On the
REFERENCES Application of the Temkin Equation in the Evaluation of Catalysts for
(1) Russo, M. V. The Emergence of Sustainable Industries: Building the Ammonia Synthesis. Ind. Eng. Chem. Process Des. Dev. 1977, 16,
on Natural Capital. Strateg. Manag. J. 2003, 24, 317. 166.
(2) Henderson, A. R.; Morgan, C.; Smith, B.; Sørensen, H. C.; (25) Gillespie, L. J.; Beattie, J. A. The Thermodynamic Treatment of
Barthelmie, R. J.; Boesmans, B. Offshore Wind Energy in EuropeA Chemical Equilibria in Systems Composed of Real Gases. I. An
Review of the State-of-the-Art. Wind Energy 2003, 6, 35. Approximate Equation for the Mass Action Function Applied to the
(3) Piwko, R.; Osborn, D.; Gramlich, R.; Jordan, G.; Hawkins, D.; Existing Data on the Haber Equilibrium. Phys. Rev. 1930, 36, 743.
Porter, K. Wind Energy Delivery Issues [Transmission Planning and (26) Emmett, P. H.; Kummer, J. T. Kinetics of Ammonia Synthesis.
Competitive Electricity Market Operation. IEEE Power Energy Mag. Ind. Eng. Chem. 1943, 35, 677.
2005, 3, 47. (27) Nielsen, A.; Kjaer, J.; Bennie, H. Rate Equation and Mechanism
(4) Klerke, A.; Christensen, C. H.; Nørskov, J. K.; Vegge, T. of Ammonia Synthesis at Industrial Conditions. J. Catal. 1964, 3, 68.
Ammonia for Hydrogen Storage: Challenges and Opportunities. J. (28) Larson, A. T.; Tour, R. S. Contributions to the Study of
Mater. Chem. 2008, 18, 2304. Ammonia CatalystsIV. Chem. Metall. Eng. 1922, 26 (14), 647.

8931 DOI: 10.1021/acs.iecr.6b01880

Ind. Eng. Chem. Res. 2016, 55, 8922−8932
Industrial & Engineering Chemistry Research Article

(29) Hummelshøj, J. S.; Sørensen, R. Z.; Kustova, M. Y.;

Johannessen, T.; Nørskov, J. K.; Christensen, C. H. Generation of
Nanopores during Desorption of NH3 from Mg(NH3)6Cl2. J. Am.
Chem. Soc. 2006, 128, 16.
(30) Huberty, M. S.; Wagner, A. L.; McCormick, A.; Cussler, E.
Ammonia Absorption at Haber Process Conditions. AIChE J. 2012,
58, 3526.

8932 DOI: 10.1021/acs.iecr.6b01880

Ind. Eng. Chem. Res. 2016, 55, 8922−8932