Article

pubs.acs.org/IECR

Performance of a Small-Scale Haber Process

Michael Reese,† Cory Marquart,† Mahdi Malmali,‡ Kevin Wagner,‡ Eric Buchanan,† Alon McCormick,‡
and Edward L. Cussler*,‡
†
West Central Research and Outreach Center, 46352 State Highway 329, Morris, Minnesota 56267, United States
‡
Department of Chemical Engineering and Materials Science, University of Minnesota, 421 Washington Avenue SE # 151,
Minneapolis, Minnesota 55455, United States

ABSTRACT: This work identifies a benchmark for the

performance of a small-scale ammonia synthesis plant powered
by wind energy. The energy used is stranded, far from urban
centers but near locations of fertilizer demand. The wind
energy drives the pressure swing absorption of air to make
nitrogen and the electrolysis of water to make hydrogen. These
are combined in the small-scale continuous Haber process to
synthesize ammonia. The analysis of runs of the small plant
presented in this article permits an assessment of how the
current production rate is controlled by three resistances:
catalytic reaction, ammonia separation by condensation, and
recycling of unreacted gas. The measured catalytic reaction
rates are consistent with separate experiments on chemical kinetics and with published reaction mechanisms. The condensation
rates predicted are comparable with literature correlations. These rate constants now supply a rigorous strategy for optimizing
this scaled-down, distributed ammonia plant. Moreover, this method of analysis is recommended for future small-scale,
distributed manufacturing plants.

■ INTRODUCTION
Since the early 1900s, the supply of natural fertilizers has not
operate at a larger scale. For processes based on fossil fuels,
such larger scales make good sense.
allowed food production adequate for the growing population. In contrast, in this work, we focus on developing plants that
Synthetic ammonia, first developed by Fritz Haber and will not be based on fossil fuels drawn from a tanker or a
commercialized by Carl Bosch, has provided the synthetic pipeline. Such small plants use renewable wind energy to make
nitrogen fertilizer required for significant increases in food small amounts of ammonia for a local market and, so, represent
production around the world.1,2 After a century of intensive distributed manufacturing, rather than the centralized manu-
development, this Haber−Bosch process is still the basis for facturing chosen for hydrocarbon feedstocks.12 We have three
producing more synthetic fertilizer. It is one of the greatest objectives in this effort: First, we want to see how the small
technical innovations known. In addition, in the middle of the plant performs as a basis for future technical improvements.
20th century, Norman Borlaug’s research improved wheat Second, we want to explore new ways to separate the ammonia
varieties, resulting in additional food production. This “Green made. Third, we want to study where such small plants should
Revolution” required vastly increased amounts of nitrogen be located. This article is concerned only with the first of these
fertilizer, with an associated increase in fossil-fuel consumption objectives, the small ammonia plant. A companion work will
and carbon emissions.3 Ammonia is also discussed as an report efforts on new separations and plant location.13
environmentally friendly route to the storage of wind energy.4,5 Together, these works are a first step toward a benchmark
Thus, the Haber−Bosch process and the Green Revolution for judging how much advantage these small plants can offer.
have led to a new challenge for the 21st century: to continue To see why such an environmentally respectful process might
increasing food production while using clean energy, avoiding make sense, consider the U.S. wind resources and fertilizer
release of greenhouse gases, and supporting farming for the demand shown in Figures 1 and 2, respectively.14,15 The wind
next generation.6,7 resources in Figure 1 are greater in the more heavily shaded
We are exploring technologies to make fertilizer using wind regions of the map. These resources are greatest in regions with
power rather than fossil fuels. Such wind-driven ammonia low population densities, so the energy they promise is often
plants will be much smaller than those currently used.8,9 The called “stranded energy”: energy that is too far from major
plants will use wind to make electricity and then to make
ammonia locally. This implies process development in a Received: January 5, 2016
different direction than that practiced in the traditional Revised: February 27, 2016
chemical industry.10,11 Such traditional development centers Accepted: March 8, 2016
on building bigger plants, which are cheaper because they Published: March 8, 2016

© 2016 American Chemical Society 3742 DOI: 10.1021/acs.iecr.5b04909

Ind. Eng. Chem. Res. 2016, 55, 3742−3750
Industrial & Engineering Chemistry Research Article

Figure 1. Wind resources in the United States. The greatest potential sources of wind power, which are darker in the figure, are in areas with small
populations. As such, they represent so-called stranded energy that can be used to make ammonia. (Figure adapted from ref 14.)

Figure 3. Schematic drawing of the Haber process small-scale plant.

Hydrogen and nitrogen fed to the plant are mixed with recycled gases
and compressed. They are preheated using gases exiting the reactor
and then pass over the catalyst in the reactor. Then, the ammonia in
this effluent is condensed and harvested.
Figure 2. Ammonia demand in the United States. The regions of
greatest ammonia demand, shown as the darker areas, roughly coincide recycled gases, and are fed to a compressor. The resulting gas
with the areas of greatest potential wind energy. (Figure adapted from mixture flows through one side of a heat exchanger to a heater
ref 15.)
and then into the reactor. The reaction, catalyzed by a
conventional ammonia synthesis catalyst, increases the amount
population centers to be used without excessive losses in of ammonia in the exiting gas mixture, which is cooled when it
transmission. At the same time, the fertilizer demand, that is, flows through the other side of the heat exchanger. The gas
where ammonia is most needed (see Figure 2), is in roughly the mixture is then further cooled to condense liquid ammonia,
same regions of the country as the stranded energy. This is one which is removed in the separator. Gases that are not
reason why projects converting wind to ammonia might make condensed are returned to the compressor.
sense: The energy is there, but it will not be used directly by In the remainder of this article, we report a model and data
concentrated populations for electricity. for this process to investigate whether they are consistent with
We have already demonstrated the promise of farm-scale, the optimal production rates. We then discuss how these results
wind-based, environmentally benign ammonia synthesis in a can guide our search for optimized small-scale, distributed
unique facility, namely, the Renewable Hydrogen and ammonia production.
Ammonia Pilot Plant, the first local farm-to-coop scale system
for fertilizer production. An analysis of the first technical data
from this facility, dedicated in summer 2013 at the West
■ THEORY
To guide our analysis of the results found from the small plant,
Central Research and Outreach Center (WCROC) in Morris, we derive an approximate model based on the schematic
MN, is reported here. It is a scaled-down version of the drawing shown in Figure 3. In this derivation, we assume that
conventional ammonia synthesis process; however, rather than the process is at steady state and that the reactor and condenser
using fossil fuels, it uses wind energy. gases are well-mixed, so that the concentrations exiting these
In other respects, this small-scale process is conventional, as regions are the same as those present in these regions.16,17 This
shown schematically in Figure 3. Hydrogen and nitrogen enter implies that the average nitrogen concentration in the reactor
the process in the upper right-hand corner, are combined with equals the nitrogen concentration at the exit. Although this is
3743 DOI: 10.1021/acs.iecr.5b04909
Ind. Eng. Chem. Res. 2016, 55, 3742−3750
Industrial & Engineering Chemistry Research Article

Figure 4. Simplified block diagram of the small plant. The contents of the reactor and condenser are assumed to be well-mixed, as justified in the
text.

certainly untrue if most of the nitrogen entering the reactor where kc is the mass-transfer coefficient between the bulk
reacts in a single pass, it is much more nearly true when little vapor; xA is the ammonia mole fraction; x0A is the equilibrium
reacts in a single pass, so that the amount recycled is large mole fraction at the cold condenser wall, equal to the saturation
relative to the amount reacting. As we will show, this is the case vapor pressure divided by the total pressure; and Ac is the
for this plant. condenser’s interfacial area. Like kRVR, the quantity kcAc has
To put these ideas on a more quantitative basis, we define a dimensions of moles per unit time; unlike kRVR, however, it
as the number of moles per unit time of liquid ammonia depends only weakly on temperature and more strongly on
removed from the condenser and m as the molar flow rate of physical factors such as flow and surface geometry. The
gas mixture exiting the separator. The recycle stream then has approximation of mass transfer as a first-order process is much
xAm moles of ammonia per unit time, xHm moles of hydrogen less severe than the corresponding approximation of reaction
per unit time, and xNm moles of nitrogen per unit time, where kinetics. Nevertheless, kc might be a weak function of xA − x0A
xi represents the mole fraction of component i in the recycle because of free convection.
stream. The stream entering the reactor has xAm moles of We now combine eqs 1 and 2 to find the ammonia
ammonia per unit time, xHm + 3a/2 moles of hydrogen per unit production a in terms of the known concentrations x*A and x0A,
time, and xNm + a/2 moles of nitrogen per unit time; the which are known at the temperature and pressure of the reactor
stream leaving the reactor contains xAm + a moles of ammonia and condenser, respectively. From eq 2
per unit time, xHm moles of hydrogen per unit time, and xNm a
xA = xA0 +
moles of nitrogen per unit time. Figure 4 presents a simplified kcAc (3)
block diagram of the small-scale ammonia synthesis plant.
We next supplement these mass balances with rate equations The concentration in the recycle, xA, is always larger than
for the reactor and the condenser. For the chemical kinetics in that found from the saturation vapor pressure x0A. From eq 1,
the reactor, the rate expression for the balance on ammonia is we write
b⎛ ⎞
⎡ ⎛ x 0 m + a ⎞⎤ ⎜ 1 + 2 − 1⎟
4c
a=
0 = −a + kR VR ⎢xA* − ⎜ A ⎟⎥ 2⎝ b ⎠
⎢⎣ ⎝ m + a ⎠⎥⎦
(4)
(1)
where
where kR is the linearized apparent catalytic reaction rate ⎛ 1 1 − xA* ⎞
b = mkR VR ⎜ ⎟
1
constant, VR is the volume of catalyst, x*A is the ammonia mole + +
fraction at equilibrium, and the quantity in parentheses is the ⎝ kR VR kcAc m ⎠ (5)
ammonia mole fraction actually in the reactor. If x*A equals this and
actual mole fraction, the reaction is at equilibrium. The details
of the linearization and of finding the numerical values of the c = mkR VR (xA* − xA0 ) (6)
rate constant are shown in the Results section below.
This is the key result of the model.
This equation is a major approximation, not only because of
The physical meaning of this result can be clarified if one
the assumption of good mixing in the reactor. It also implicitly
recognizes that the term 1/kRVR is the resistance due to the
assumes that the reaction is first-order. However, this is not
chemical reaction. Similarly, 1/kcAc and (1 − x*A )/m are the
true: For example, in the widely quoted Temkin−Pyzhev resistances of the condenser and of the recycle loop,
equation,18−21 the forward rate of reaction depends nonlinearly respectively. Moreover, the ratio c/b2 in eq 4 can be rearranged
on the partial pressures of ammonia, nitrogen, and hydrogen. as
Here, by approximating this rate as a single quantity (i.e., kRVR)
⎡ ⎤⎛ ⎞
⎢ (xA* − xA0 )/m ⎥⎜ ⎟
with dimensions of moles per unit time, we acknowledge that 1
c
=⎢
1 − xA* ⎥⎜ 1 − xA* ⎟
this rate constant will vary not only with temperature and kR VR

⎜ 1 + m ⎟⎠
⎣⎢ kRVR + kcAc + m ⎥⎦⎝ kRVR +
catalyst volume, but also with mole fractions in the reactor and b2 1 1 1
kcAc (7)
with the reactor pressure.
The second balance, on the condenser, is also at steady state The numerator of the first term in brackets on the right-hand
side of eq 7 is the driving force for the reaction. As long as the
0 = a − kcAc(xA − xA0) (2) ammonia concentration in the inlet stream to the reactor (x0A)
3744 DOI: 10.1021/acs.iecr.5b04909
Ind. Eng. Chem. Res. 2016, 55, 3742−3750
Industrial & Engineering Chemistry Research Article

is less than the ammonia saturation mole fraction at the reactor These gases, now at a pressure of about 145 bar, flow to a 20
pressure and temperature (x*A ), the production of ammonia kW heater (Sep-Pro Systems Inc., Houston, TX), where they
takes place. The denominator of the first term in brackets on can be heated to reactor temperature.
the right-hand side is the total resistance normalized by the The hot gases then flow to the reactor (Consolidated Inc.,
recycle, and the term in parentheses on the right-hand side of Gary, IN), which is 2.23 m long and 0.203 m in diameter and is
eq 7 is the fraction of the total resistance due to the reaction. sealed with a metal O-ring made of stainless steel. Tightening
To see how these different factors affect ammonia this O-ring to avoid leaks requires considerable care using
production, it is helpful to examine the result of a Taylor hydraulic torqueing of the bolts that secure the top flange to the
series expansion of the square root in eq 4, which gives reactor body. Within the reactor, the catalyst is contained in an
annular basket (Consolidated Inc., Gary, IN) that is 1.8 m long
xA* − xA0 and has a 0.152-m outer diameter. To load the catalyst, the top
a=
1 1 1 − xA* flange is removed, and the catalyst is poured into the basket.
+ + (8)
kR VR kcAc m The reagent gas feeds into the lumen of this annular space and
flows radially out through the catalyst. The particular catalyst
This behavior is observed in many rate processes with recycle used is AmoMax-10 (Sud-Chemie, Louisville, KY). The gases
and separation: The amount produced is proportional to the exiting the reactor flow back through the shell side of the heat
overall driving force divided by the total resistance.17 The total exchanger and, so, preheat the reactive feed.
resistance is the sum of the resistances of reaction, These gases now flow to the condenser (Sep-Pro Systems
condensation, and recycle. Phrased in other terms, it is a Inc., Houston, TX), which has one tube that is 8.4 m long and
weighted harmonic average of the speeds of these three steps, 0.013 m in diameter, giving a condenser surface area of about
so that the lowest speed has the greatest effect on the rate of 0.34 m2. A refrigeration skid (Sep-Pro Systems Inc., Houston,
ammonia production. If the reactor is undersized or operating TX) drives the refrigerant, R-404A (Copeland, Sidney, OH),
at a temperature that is too low, the reactor resistance will through an 8.4-m-long, 0.025-m-diameter outer shell around
dominate. If the condenser is undersized, the condenser the condenser tube. It operates at about −20 °C. The resulting
resistance will dominate, and if the recycle rate is too low to gases exit from the condenser at a temperature of about −17 °C
take full advantage of the unit operations, the recycle resistance and flow to the separator, a flash drum. The flash drum
will dominate. We use this model and these results to analyze separates the ammonia liquid. Ammonia separated in the flash
the experimental plant runs described next and, hence, to drum exits the plant through a solenoid valve and flows to a
explore how this small-scale plant can be made more larger storage tank. The ammonia collected is stored in an 11.7
productive. In pursuing this case study, we recommend this m3 storage tank (WestMor Industries, Morris, MN) at a
procedure for the analysis of other such small-scale, distributed pressure of 10 bar. The uncondensed gases are mixed with the
processes. incoming feed and recycled back into the compressor to start

■ EXPERIMENTAL SECTION
Plant Details. The ammonia synthesis plant that is the key
the process again.
The piping used for this plant, most of which was purchased
from Swagelok (Chaska, MN), is made from two different
to these experiments is shown schematically in Figure 3. The materials: carbon steel and stainless steel. Embrittlement with
entire system is powered by electricity either produced by a hydrogen is always a problem. All tubing is fixed inside the pilot
1.65 MW wind turbine (Vestas, Aarhus, Denmark) or from the skid. Unless the tubing is stressed, the risk of leaking is very
local utility (Ottertail Power Company, Fergus Falls, MN). small. However, the corrosive nature of ammonia worsens the
Nitrogen is produced by an Innovative Gas Systems NS-10 situation. Typically, in the tubing through which the hot gas
pressure swing adsorption system (Grosseto, Italy). The mixture is transported between heat exchanger and reactor,
nitrogen it produces, which is greater than 99.9% pure, is stainless steel 316L is employed because it is more resistant
stored at ambient temperature at a pressure of 165 bar in 18 toward corrosion by hot ammonia. In the tubing through which
tanks, each with a volume of 0.05 m3 (Norris Cylinder the low-temperature gas mixture is transported among the heat
Company, Longview, TX). Hydrogen, which is generated by a exchanger, separator, and compressor, carbon steel is used.
Proton Onsite Hogen H6 system (Wallingford, CT), is over Carbon steel has a higher resistance toward high pressure and a
99.9% pure after being dried across phosphorus pentoxide. It is longer lifetime in the presence of hydrogen.
stored at ambient temperature and 165 bar pressure in 54 0.05 The chief difficulties in startup were plumbing of the vent
m3 tanks identical to those used for nitrogen. Both of these ports on the mixed-gas compressor and leaks throughout the
systems function properly; occasionally, the hydrogen produc- skid. Holes on the side and bottom of the reactor, which had
tion was marginally less than the amount required for the been drilled for testing, also leaked and, so, were welded shut
synthesis. with a certified R-stamp welder. Some of the equipment also
The nitrogen and hydrogen gases flow through an orifice had material that was not compatible with ammonia and had to
meter (Imperial Flange & Fitting Company Inc., Los Angeles, be replaced or repaired. Significant effort was required to
CA) into the main system, where they are combined with understand and program the control systems.
recycled gases coming from the separator. These mixed gases, Now that the small plant is capable of producing extended
at 83 bar, enter an RIX Industries (Benicia, CA) 4VX1BG-1.7 runs at steady state, we are able to subject its performance to
compressor. The exiting gases leave at a pressure of about 145 the analysis described in the Theory section.
bar and flow to a train of four double-pipe heat exchangers, Laboratory Kinetic Experiments. To estimate the reactor
where they are warmed using the gases that are discharged from resistance, we needed to test the catalyst used in the small
the reactor. The inner pipe of each heat exchanger (Sep-Pro plant. This was accomplished using a laboratory-scale reactor
Systems Inc., Houston, TX) has an inside diameter of 0.013 m with the same catalyst. The reactant gases N2 and H2, of
and a length of 20.3 m, giving a total interfacial area of 0.83 m2. ultrahigh purity, were purchased from Matheson (New
3745 DOI: 10.1021/acs.iecr.5b04909
Ind. Eng. Chem. Res. 2016, 55, 3742−3750
Industrial & Engineering Chemistry Research Article

Brighton, MN). The stainless steel reactor was tubular, 0.38 m

long and 6.4 mm in diameter. The first and last 0.13-m sections
of the reactor were filled with stainless steel wool, and the
catalyst was packed in the middle. The catalyst, the same as in
the small plant, with a nominal size range of 1.5−3 mm, is also
stabilized with an oxygen-rich protective layer. The reactor used
an Omega ceramic heater (CRFC-36/115-A, Stamford, CT)
equipped with an Omega multiramp proportional−integral−
derivative (PID) controller (CN96211TR) to control the
temperature. The inlet and outlet gas temperatures and the
reactor surface temperature were measured using K-type
thermocouples connected to an Omega signal conditioner
(DRG-SC-TC). The system pressure was recorded using a
WIKA pressure transducer (50426877, Lawrenceville, GA) with Figure 5. Reactor temperature and pressure in an example plant run.
0−10 V dc output. After the transients in the first 3 days, the plant process operates near
The experiments were carried out in circulation batch with a steady state. The recycle flow is also constant in time.
variable piston pump (model PW2070N, PumpWorks,
Minneapolis, MN). Data were recorded every second using a pressure and the recycle flow. Such a lack of correlation might
National Instruments LabVIEW program that also controlled reflect control of the synthesis by chemical reaction, as explored
the mass flow controllers, which injected known volumetric in the Discussion section below.
flows of reactant gases into the reactor. Before each run, the The feeds of nitrogen and hydrogen are in a ratio of 3:1, the
system was tested for leaks for 3 h, using nitrogen at 1500 psi. ratio consistent with the reaction’s stoichiometry. Although
To remove the oxidized protective layer and activate the these data show steady operation over days, the feeds are
catalyst, the system was pressurized at 200 psi with 500 cm3 subject to considerable variation over hours, as shown in Figure
(STP)/min hydrogen. The temperature was slowly raised for 6. The fluctuations in nitrogen flow rate, which are smaller,
27 h to reach to 723 K. The hydrogen flowed through the
catalyst bed for at least 24 h at this temperature. After
activation, the system was kept under nitrogen.

■ RESULTS
Our analysis centers on five runs carried out in the small-scale
plant for times ranging from 4 to 43 days. The runs had reactor
exit temperatures of about 565 K and pressures of about 115
bar. More specific parameters are given in Table 1. Although

Table 1. Basic Data for Five Process Runsa

start date Tb (K) Pb (bar) run (days) N2 feed (mol/h)
07/14/2014 569 112 4 63
09/02/2014 575 72 4 62
Figure 6. Nitrogen and hydrogen flows into the plant in an example
09/07/2014 563 124 30 41
plant run. These flows, averaged over time, are in the stoichiometric
10/29/2014 565 117 33 47 ratio of 3:1.
01/09/2015 557 128 43 35
a
Values are averages for the entire run. bMeasured at the reactor exit. might be caused by changes in the pressure swing adsorption
unit, which are not completely damped out by the storage tank
the runs differed slightly, they produced similar results; for for nitrogen. The larger fluctuations in hydrogen are due
illustration, we discuss in detail only the run begun on October primarily to the operation of the electrolysis unit. Occasionally,
29, 2014 (highlighted in bold in Table 1), judging it to be the electrolysis unit appeared not to supply hydrogen at the rate
typical. needed by the small plant. When this happened, we reduced
The temperature and the recycle obtained for the run of 10/ the nitrogen flow rate. Still, even with these occasional
29/2014 are shown in Figure 5. During the first 3 days, there variations, the temperature, pressure, and recycle rate remained
were excursions not only in temperature and pressure but also relatively constant, as shown by the data in Figure 5.
in recycle rate. Additionally, we observed a small spike in We now turn to the implications of the data in Figures 5 and
pressure on days 18−21, mainly due to the difficulty associated 6 using the model developed in the Theory section. These are
with controlling the operation of the small plant. As a result, we shown in Table 2 for the earlier and later periods of the
discarded the data obtained on days 1−3 and 18−21 and experiment begun on 10/29/2014. The second column in
concentrated our analysis on two divided periods on days 4−18 Table 2 includes the average feed rates of nitrogen and
and days 21−33. hydrogen. There is no ammonia in this fresh feed. The third,
The data in Figure 5 show that the small plant is operating fourth, and fifth columns in the table give the flows at different
near steady state. The pressure is almost constant, varying by points within the process. These internal flows are greater than
less than 5%. The recycle rate is also nearly constant, fluctuating the amounts of nitrogen and hydrogen actually fed, as is
by 7%. The temperature shows greater excursions, by as much characteristic of most ammonia synthesis plants. To get fast
as 50 °C. These excursions seem to be uncorrelated with the kinetics, the reactor temperature and pressure should be high.
3746 DOI: 10.1021/acs.iecr.5b04909
Ind. Eng. Chem. Res. 2016, 55, 3742−3750
Industrial & Engineering Chemistry Research Article

Table 2. Gas Flows (mol/h) within the Steady-State System (10/29/2014 Runa)b
new feed flow into reactor flow out of reactor flow out of condenser product
Day 4−Day 18
N2 40 (0.22) 1100 (0.25) 1000 (0.23) 1000 (0.24) 0
H2 140 (0.78) 3100 (0.70) 3000 (0.70) 3000 (0.71) 0
NH3 0 230 (0.05) 320 (0.07) 230 (0.05) 90
total ∼180 ∼4400 ∼4300 ∼4200
Day 21−Day 33
N2 47 (0.25) 1100 (0.24) 1100 (0.24) 1100 (0.24) 0
H2 140 (0.75) 3400 (0.74) 3300 (0.72) 3300 (0.74) 0
NH3 0 64 (0.02) 160 (0.04) 64 (0.01) 90
total ∼190 ∼4600 ∼4600 ∼4500
a
Average temperature, 565 K; average pressure, 113 bar. bValues in parentheses are mole fractions.

However, the high temperature producing the fast kinetics also

means the conversion at equilibrium is limited. This means that
large amounts of unreacted gases must be separated from the
ammonia produced, which, in this case, is removed by
condensation. In addition to the unreacted hydrogen and
nitrogen, there will also be a substantial amount of ammonia
that is not removed in the condenser. As a result, the total
recycle stream here is about 20 times the new feed stream, as
values in the Table 2 show. This is true in both the experiments
during days 4−18 and the experiments during days 21−33.
Regrettably, the plant does not currently contain the
instrumentation to measure these flows more directly. Although
the orifice meters measuring the nitrogen and hydrogen flows
are operating at Reynolds numbers of about 5000, where
calibration is difficult, we believe that the orifice measurements Figure 7. Laboratory-measured vs literature-calculated reaction
are reliable. The flow of the recycled gases is 70% of the kinetics at 723 K. The rates measured from laboratory kinetic
maximum flow possible in this pump, consistent with the experiments (shown by markers) in this work agree with those
design. estimated from the literature19−22 (solid line).
In the future, we plan to modify the equipment to allow
higher temperatures, to ensure that the nitrogen and hydrogen to be detailed. Indeed, we have already made an early economic
flows are accurate, and to measure the concentrations in the analysis of these data.11 We have used these results to start to
recycle stream. Part of the point of the analysis in this article is identify where small plants for distributed ammonia production
to check our intuition that (a) the current production rate is should be located.13
limited by the reaction and (b) the other components of the Nonetheless, the core of this article is the exploration of the
system will be able to accommodate increased production if the technical limitations and opportunities for this small-scale
reaction rate is increased. technology. We have a sound start, but there is a shortfall in the
Finally, before we discuss these results, we need to report our intended ammonia production. We assume that this is probably
laboratory measurements of ammonia kinetics compared with the result of the limited temperature achieved in the reactor.
those predicted from literature sources, summarized as the We reached a temperature of about 565 K, whereas the design
Temkin−Pyzhev equation. This comparison is given in Figure target was 670 K. Had we been able to achieve this higher
7, where our measurements at the temperatures and pressures temperature, we might expect a corresponding increase in
shown are compared with those expected from other ammonia productionas long as the separator and the recycle
references.19−22 The figure shows that the agreement is good can keep up with the increased production. We discuss this
(R2 = 0.98). We use these values in the estimates of the exploration with some rigor next.
different resistances to ammonia synthesis in the discussion of Our immediate efforts to extend this technology combine
the small-plant performance that follows. these preliminary experiments with the analysis suggested by

■ DISCUSSION
The results reported in the preceding section show that the
the model developed in the Theory section. This model,
summarized in eq 4 or 8, predicts that the ammonia synthesis
by the entire process is proportional to an overall rate constant
small plant works and can produce extended runs of data times an overall driving force. The overall rate constant is a
logging at steady state. It makes ammonia with wind energy. It harmonic average of the three rate constants of the reaction, the
does so with nitrogen made by the pressure swing adsorption of ammonia separation, and the recycle. The overall driving force
air and with hydrogen from the electrolysis of water. The is the ammonia concentration that would exist at equilibrium,
pressures used are near those specified in the design, although xA*, minus the ammonia concentration in the vapor in the
the reactor temperature is about 100 K less than the condenser, x0A.
temperature of a conventional reactor as originally sought. The values of the three rate constants merit careful
Still, these early results supply a template on how to proceed. consideration, as they are the key by which the process will
They should allow the costs for small-scale ammonia synthesis be improved. For instance, it would not make sense to take
3747 DOI: 10.1021/acs.iecr.5b04909
Ind. Eng. Chem. Res. 2016, 55, 3742−3750
Industrial & Engineering Chemistry Research Article

measures to increase the reactor temperature if we found that

the overall production rate is limited by the separator capacity.
In each case, we linearize the rate constants for small deviations
from equilibrium.
Linearized Rate Constant for Reaction. The rate
constant for the reaction is the most complex. This reaction
rate is most often correlated using the Temkin−Pyzhev
equation18−21
pN pH1.5 pA
r = k1 2 2
− k2
pA pH1.5 (9)
2

This equation can be rewritten by defining the variable Figure 8. Linearized reaction rate vs temperature. The reaction rate
values, obtained from laboratory kinetic experiments, extrapolated to
PA* − PA the plant reactor temperature of 565 K, were used in estimating the
X=
PN* 2 (10)
characteristic times in Table 3.

and for stoichiometric feed with no purge

affected both by the rate of mass transfer on the reactive gas
X PN − PN*2 PH − PH*2 side and by the heat transfer on the refrigerant side. At present,
= 2 = 2
2 PN*2 PH*2 (11) we have no basis for determining which of these rates is more
important. From the Chilton−Colburn analogy,22 we suspect
Linearization using the Taylor series for small values of X that they will be similar in magnitude. As a result, we decided to
simplifies this substitution to yield make our first estimates based on heat transfer, because the
PN*2PH1.52 * ⎛ P* ⎞ P* ⎛
correlations are better developed.
P*
r = k1 ⎜⎜1 + 5 + N2 ⎟⎟X − k 2 A ⎜⎜1 − N2 Our estimate of this rate, then, begins with the overall heat-
PA* ⎝ 4 PA* ⎠ PH1.52 * ⎝ PA* transfer coefficient U

3⎞
1/U = 1/ho + 1/hw + 1/hj (14)
− ⎟⎟X
4⎠ (12)
where U and hj are overall and individual heat-transfer
coefficients, respectively, and the subscript j refers to the
Subtracting the concentrations at equilibrium gives external (o), wall (w), and internal (j) surfaces. The gas stream
entering the condenser is expected to be at 333 K (recorded
PN*2PH1.52 * ⎛ 9 P* ⎞ P* ⎛ P* ⎞
⎜⎜ + N2 ⎟⎟(X − X *) + k 2 A ⎜⎜ N2 − 1 ⎟⎟
from the plant temperature indicator), with a mass flow rate of
r = k1 38 kg/h.23 In the condenser, this flow is cooled on the shell side
PA* ⎝ 4 PA* ⎠ PH1.52 * ⎝ PA* 4⎠
with R-404A refrigerant, which is expected to have an external
(X − X *) = kR′ (X − X *) (13)
heat-transfer coefficient ho of about 2000 W/m2·K.24 Because
the condenser tubes are constructed of 4-mm-thick carbon steel
where kR′ is a linearized overall rate constant with dimensions of (SA-333-6), the wall heat-transfer coefficient is over 50000 W/
moles of ammonia per catalyst volume per time, which is m2·K. The piping for the condenser is a 12.7-mm schedule 80
different from kR in eq 1 by a factor of 4. We used two different pipe, with a wall thickness equal to about 4 mm. The heat-
approaches to calculate this linearized reaction rate constant for transfer coefficient for ammonia condensing inside the tube is
the plant reactor’s temperature. The first approach was to about 4000 W/m2·K. Thus, the overall heat-transfer coefficient
calculate the linearized rate constant for each temperature, plot is equal to 1300 W/m2·K. Previous literature supports this
the logarithms of these constants versus the reciprocal of estimate for ammonia condensation.25 Because the condenser
temperature, and use the data to extrapolate to find the rate wall is at the refrigerant temperature (250 K), the heat of
constant of the plant process at 565 K. The resulting Arrhenius vaporization of ammonia is 22.8 kJ/mol, and the total area for
plot is shown in Figure 8. This first approach gives a value for heat transfer is 0.343 m2, the rate constant for evaporation is
kR of 0.67 mol/L·h. Because the catalyst volume is 12 L, the
UAΔT 1300 × 0.343 × [333 − (250)] × 3600
characteristic rate constant, kRVR, is 14 mol/h. The second kcA = =
approach was to plot the Arrhenius equation for k1 and k2 and ΔH vap 22.8 × 103
extrapolate to find the values at the plant temperature. Then,
= 5850 mol/h (15)
we used the extrapolated values to calculate k1 and k2 at 565 K
and substituted these values into eq 13 to determine the Again, the apparatus contains 113 mol of gas mixture, so the
linearized kR. Although this approach gave a value of 0.23 mol/ characteristic time for the condenser is 0.020 h. This is much
L·h, smaller than that of the first method, the fit of the plots was shorter than the reaction time.
poorer. As a result, we used the first approach to calculate the Characteristic Rate Constant for Recycle. The third rate
values used in our analysis. Finally, the small-plant system constant that affects the process is that associated with the
contains n0 = 113 mol of gas mixture actually present in the recycle pump. The equilibrium mole fraction at the operating
system. Thus, the characteristic time for the reaction is 14 h. pressure and temperature of 565 K and 117 bar, respectively, is
Characteristic Rate Constant for Condensation. We 0.58. Because the data show recycle flow rates of about 4400
next turn to the second rate constant of the small-plant process, mol/h, the pumping characteristic time for the 10/29/2014 run
that associated with the condenser. The condensation rate is is defined as
3748 DOI: 10.1021/acs.iecr.5b04909
Ind. Eng. Chem. Res. 2016, 55, 3742−3750
Industrial & Engineering Chemistry Research Article

Table 3. Characteristic Process Properties for a Stoichiometric Ratio of Nitrogen to Hydrogen of 1:3
07/14/2014 09/02/2014 09/17/2014 10/29/2014 01/09/2015
T (K) 569 575 563 565 557
p (bar) 112 72 124 117 128
m (mol/h) 3300 2200 4400 4400 4400
n0 (mol) 110 68 120 110 120
x*A 0.56 0.53 0.58 0.55 0.60
x0A 0.033 0.047 0.019 0.016 0.013
condenser T (K) 400 360 420 410 420
n0/kRVR (h) 2.4 0.92 4.5 3.5 7.5
n0/kcAc (h) 0.019 0.012 0.021 0.020 0.022
n0(1 − xA*)/m (h) 0.014 0.015 0.011 0.011 0.011

n0(1 − xA*) 113(1 − 0.58) However, we note that the separation or recycle capacity
pumping time = = = 0.011 h might well be judged limiting in other cases, and these
m 4400
(16) operations are felt to influence broad sections of the chemical
industry.17
This is smaller than both the reaction time and the heat- This simple method of analysis, we suggest, should be useful
transfer time. for many current efforts to produce small-scale, distributed
Comparing Characteristic Rate Constants As Charac- manufacturing processes. We note that the principles guiding
teristic Time Constants. To explore the meaning of these downscaling are not as familiar as those for upscaling, in large
times more completely, we turn to the summary of results in part because the economic driving forces can be unusual (e.g.,
Table 3. This table shows more complete data for the five runs stranded power, zero-carbon incentives) and also in part
mentioned in Table 1, identified with the dates in the column because the capital cost is often governed by custom
headings. The results for the experiment started on 10/29/ manufacturing. Thus, it is useful to analyze prototype small
2014 are those detailed above and shown in bold. However, all plants to optimize operation.
five experiments give similar results. Rows 1 and 2 in the table
give the temperature and pressure of the experiments. Rows 3
and 4 give the moles per hour recycled in the process (m) and
■ CONCLUSIONS
In this work, we have presented a benchmark for the
the total number of moles actually present in the reactor, n0,
performance of a small-scale ammonia synthesis with a zero-
respectively. Rows 5 and 6 give the mole fractions at carbon footprint. Stranded wind energy supplies the energy for
equilibrium of ammonia in the reactor and in the condenser, this process, and the feed is provided from renewable resources.
respectively; the difference between these values is the overall Such a process can satisfy fertilizer demand in distant locations
driving force for the process, as summarized in eq 8. Row 7 with limited access to conventional energy resources. Generally,
gives the temperature in the condenser. Most importantly, rows controlling small-scale processes is difficult, and this process is
8−10 give the times for the three rates of reaction, not operating under its best conditions. To further understand
condensation, and recycle, respectively. We use the total the ammonia synthesis in our small-scale Haber−Bosch plant
number of moles of gas in the plant to fashion these as and determine the optimal conditions for operation, we
characteristic times. The longest time, which corresponds to the simplified the ammonia production process to three main
slowest step, controls the overall rate of the process; this is the steps operating in series: reaction, separation, and recycle. This
characteristic time associated with the principal resistance to simplified model successfully predicts the performance of
the overall plant production rate. different units and gives information regarding each region.
Under the process conditions used so far, the unit operation Our laboratory-scale kinetic studies along with our linearized
dominating the rate of production is the reactor. Upon reaction rate show that the chemical reaction is rate-limiting.
reflection, this is not surprising. A review of the operating Our analysis indicates that the catalytic reaction resistance of
logs shows that the reactor operating temperatures used in the the currently operating plant is at least 3 orders of magnitude
runs to date have been substantially below those characteristic larger than the condensation and recycle resistances. Our
of the use of this catalyst in conventional plants (ca. 700 K). kinetic studies support the suggestion that increasing the
We also see that the other two characteristic times, of the reaction temperature to approximately 700 K would lead to a
condenser and of the recycle, are 3 orders of magnitude reaction resistance of the same order of magnitude as the
smaller: These units are quite capable of handling increased condensation resistance. In this instance, we have shown that
production rates. If the reaction temperature could be increased there is no reason to be concerned that the increased
to 700 K, the characteristic time of the reaction rate constant production rate might not be accommodated by the separator
would equal that estimated for the condenser, and the plant or the recycle capacities. Such a straightforward model might
performance would be closer to optimal. This optimum might help to explain the design of other small-scale processes in the
be found at lower temperatures if, as expected, the condenser future.
time includes resistance to mass transfer in the ammonia, as
well as heat-transfer resistance in the refrigerant; only the latter
contribution has been considered so far.
In this case study, our analysis suggests that the recycle and
■ AUTHOR INFORMATION
Corresponding Author
separation capacities of the current small plant are more than *Tel.: 612-625-1596. Fax: 612-626-7246. E-mail: cussler@umn.
adequate to accommodate increased reactor performance. edu.
3749 DOI: 10.1021/acs.iecr.5b04909
Ind. Eng. Chem. Res. 2016, 55, 3742−3750
Industrial & Engineering Chemistry Research Article

Notes (5) Patil, A.; Laumans, L.; Vrijenhoef, H. Solar to Ammonia − Via
The authors declare no competing financial interest. Proton’s NFuel Units. Procedia Eng. 2014, 83, 322.
(6) Aguirre-Villegas, H. A.; Larson, R.; Reinemann, D. J. From waste-
■ ACKNOWLEDGMENTS
This work was primarily supported by MNDrive, an initiative of
to-worth: Energy, emissions, and nutrient implications of manure
processing pathways. Biofuels, Bioprod. Biorefin. 2014, 8, 770.
(7) Gilbert, P.; Alexander, S.; Thornley, P.; Brammer, J. Assessing
the University of Minnesota; and by the Minnesota Environ- economically viable carbon reductions for the production of ammonia
ment and Natural Resources Trust Fund as recommended by from biomass gasification. J. Cleaner Prod. 2014, 64, 581.
the Legislative-Citizen Commission on Minnesota Resources (8) Bartels, J. R. A Feasibility Study of Implementing an Ammonia
(LCCMR). Additional support came from the Dow Chemical Economy. M.S. Thesis, Iowa State University, Ames, IA, 2008.
Company, Midland, MI. The Wind Power (2014) map was (9) Tunå, P.; Hulteberg, C.; Ahlgren, S. Techno-economic
created by the National Renewable Energy Laboratory for the assessment of nonfossil ammonia production. Environ. Prog.
U.S. Department of Energy with data provided by AWS Sustainable Energy 2013, 33, 1290.
(10) Tallaksen, J.; Bauer, F.; Hulteberg, C.; Reese, M.; Ahlgren, S.
TruePower.
Nitrogen fertilizers manufactured using wind power: Greenhouse gas

■ NOMENCLATURE
a = moles of ammonia produced (mol/h)
and energy balance of community-scale ammonia production. J.
Cleaner Prod. 2015, 107, 626.
(11) Tiffany, D. G.; Reese, M.; Marquart, C. Economic Evaluation of
A = condenser interfacial area (m2) Deploying Small to Moderate Scale Ammonia Production Plants in
b, c = quadratic equation parameters Minnesota Using Wind and Grid-Based Electrical Energy Sources;
hi = heat-transfer coefficient of component i Minnesota Corn Research and Promotion Council: Shakopee, MN,
m = moles of gas mixture per hour in the recycle flow (mol/ 2015. http://tinyurl.com/ammonia-from-wind (Accessed February 3,
2015).
h)
(12) Morgan, E.; Manwell, J.; McGowan, J. Wind-powered ammonia
k1 = rate constant of the forward reaction fuel production for remote islands: A case study. Renewable Energy
k2 = rate constant of the reverse reaction 2014, 72, 51.
kc = heat-transfer coefficient (mol/h·m2) (13) Allman, W. A.; Daoutidis, P. Ammonia Supply Chains: A New
k′R = linearized catalytic reaction rate using eq 13 (mol/L·h) Framework for Renewable Generation with a Case Study for
kR = linearized catalytic reaction rate using eq 1 (mol/L·h) Minnesota. To be presented at the 26th European Symposium on
n0 = total number of moles of gas present in the small-scale Computer Aided Process Engineering, Portorož, Slovenia, Jun 12−15,
plant 2016.
Pi = partial pressure of component i (14) National Renewable Energy Laboratory, (NREL) Wind
U = overall heat transfer Resource Map. 2014 - data provided by AWS TruePower. Available
at http://www.nrel.gov/gis/wind.html. (Accessed September 17,
VR = volume of catalyst (m3)
2015)
X = fractional driving force in a normalized mole fraction (15) Planted Corn Acreage by County. National Agricultural
scale Statistics Service (NASS), Washington, DC, 2014.
x*A = ammonia mole fraction at equilibrium (16) Kumar, A.; Daoutidis, P. Nonlinear dynamics and control of
x0A = mole fraction of ammonia at the condenser wall process systems with recycle. J. Process Control 2002, 12, 475.
xi = mole fraction of component i in the recycle stream (17) Baldea, M.; Daoutidis, P. Dynamics and Nonlinear Control of
Subscripts Integrated Process Systems; Cambridge University Press: Cambridge,
A = ammonia U.K., 2012.
(18) Temkin, M. I.; Pyzhev, V. Kinetics of ammonia synthesis on
c = condenser promoted iron catalyst. Acta Physiochim. USSR 1940, 12, 327.
H = hydrogen (19) Annable, D. Application of the Temkin kinetic equation to
i = component i ammonia synthesis in large-scale reactors. Chem. Eng. Sci. 1952, 1, 145.
j = internal (20) Guacci, U.; Traina, F.; Ferraris, G. B.; Barisone, R. On the
N = nitrogen Application of the Temkin Equation in the Evaluation of Catalysts for
o = outer side the Ammonia Synthesis. Ind. Eng. Chem. Process Des. Dev. 1977, 16,
w = wall 166.
Superscript (21) Jennings, J. R. Catalytic Ammonia Synthesis: Fundamentals and
Practice; Plenum Press: New York, 1991.
0 = phase equilibrium (22) Cussler, E. L. Diffusion: Mass Transfer in Fluid Systems;
* = equilibrium Cambridge University Press: New York, 2009.
Greek Letters (23) Cavallini, A.; Censi, G.; Col, D. D.; Doretti, L.; Rossetto, L.;
ΔHvap = heat of vaporization (J/mol) Longo, G. Heat Transfer Coefficients of HFC Refrigerants during
ΔT = temperature difference in the condenser Condensation at High Temperature Inside an Enhanced Tube.
Presented at the 9th International Refrigeration Conference; West
■ REFERENCES
(1) Stoltzenberg, D. Fritz Haber: Chemist, Nobel Laureate, German,
Lafayette, IN, 2002; Paper 563.
(24) Green, D. W. Section 2: Physical and Chemical Data. In Perry’s
Chemical Engineers’ Handbook; Perry, R. H., Green, D. W., Eds.;
Jew; Chemical Heritage Press: Philadelphia, 2004. McGraw-Hill: New York, 2007.
(2) Charles, D. Master Mind: The Rise and Fall of Fritz Haber, the (25) Geankoplis, C. J. Transport Processes and Unit Operations; PTR
Nobel Laureate Who Launched the Age of Chemical Warfare; Ecco: New Prentice Hall: Engelwood Cliffs, NJ, 1993.
York, 2005.
(3) Energy Efficiency and CO2 Emissions in Ammonia Production;
International Fertilizer Industry Association: Paris, France, 2009.
(4) Lan, R.; Irvine, J. T. S.; Tao, S. Ammonia and related chemicals as
potential indirect hydrogen storage materials. Int. J. Hydrogen Energy
2012, 37, 1482.

3750 DOI: 10.1021/acs.iecr.5b04909

Ind. Eng. Chem. Res. 2016, 55, 3742−3750