You are on page 1of 15

Received: 26 October 2020 Revised: 16 January 2021 Accepted: 31 January 2021

DOI: 10.1002/amp2.10077

RESEARCH ARTICLE
Futures

Rapid pressure swing adsorption for small scale ammonia


separation: A proof-of-concept

Bosong Lin | I-Min Hsieh | Mahdi Malmali

Department of Chemical Engineering,


Texas Tech University, Lubbock,
Abstract
Texas, USA In a typical Haber-Bosch process, a gas stream containing 15–20 mol%
ammonia is obtained from the reactor effluent, and ammonia is then partially
Correspondence
Mahdi Malmali, Department of Chemical separated in a phase-changing condensation unit. When operating at lower
Engineering, Texas Tech University, pressure for distributed manufacturing, the single-pass conversion drops to less
807 Canton Avenue, Lubbock, TX 79409,
than 10 mol%, which makes the condensation more cost-intensive. A small
USA.
Email: mahdi.malmali@ttu.edu adsorber is proposed for concentrating ammonia through rapid pressure swing
adsorption (RPSA) that fits the small-scale processing. A mathematical model
is developed to evaluate the feasibility of the RPSA process for ammonia sepa-
ration with high recovery. The ideal adsorbed solution theory, based on the
Freundlich single-component isotherm model, is proposed to predict binary
isotherms for various commercially available adsorbents. The performance of
the RPSA-assisted adsorber is then studied at different process conditions for
concentrating ammonia. The effect of various operating variables such as
exhaust flow rate, cycle time, and feed pressure, is investigated. The proposed
numerical model shows that nearly pure ammonia can be continuously pro-
duced at optimized conditions, with more than 95% recovery. This low-
pressure RPSA-assisted adsorber can be used to design modular ammonia
devices for distributed manufacturing. Our proposed technology can be further
extended to concentrate other dilute gas mixtures, such as carbon dioxide.

KEYWORDS
ammonia separation, Haber-Bosch process, ideal adsorbed solution theory, low-pressure
manufacturing, modular adsorption processes, rapid pressures swing adsorption

1 | INTRODUCTION efficiency of the HB process and to lowering the capital


and operating expenses.[2] The state-of-the-art, large-scale
1.1 | Small-scale ammonia production ammonia process relies on steam methane reforming
(SMR) and cryogenic air sepration to produce hydrogen
The emergence of the Haber-Bosch (HB) process made a and nitrogen, respectively. Ammonia is then produced at
significant impact on the production of agricultural fertil- high temperatures (400–500 C) and high pressures (150–
izers over the last century. The increase in synthetic fer- 300 bar). High temperature is needed to accelerate reac-
tilizer production dramatically boosted food production tion kinetics while high pressure is required to increase
and is currently responsible for feeding nearly half of the single-pass conversion. Although extensively opti-
the world's population.[1] A century of research and mized, the SMR-based ammonia is still energy- and
development has been dedicated to improving the capital-intensive. In order to reduce the overall costs,

J Adv Manuf Process. 2021;3:e10077. wileyonlinelibrary.com/journal/amp2 © 2021 American Institute of Chemical Engineers 1 of 15
https://doi.org/10.1002/amp2.10077
2 of 15 LIN ET AL.

ammonia manufacturing takes place in large, centralized the underlying separation mechanisms: ammonia adsorp-
plants far from the end-users, with an approximate tion and ammonia absorption.[9,14–17] In either methods,
energy requirement of 28 GJ per ton of ammonia.[3] This the selected adsorbents and absorbents should have
key process consumes 2% of the global natural gas pro- excellent ammonia selectivity over other gases present in
duction and generates 1.2% of the global anthropogenic the multicomponent mixture (hydrogen and nitrogen),
CO2 emissions.[4,5] and sorbents should be regeneratable through pressure or
Recently, green ammonia produced from renewable temperature swings, or a combination thereof.[18] Pres-
resources has been considered as an alternative to the con- sure swing adsorption (PSA) methodologies developed in
ventional fossil-based technology. Not only is ammonia a the 1960s have been widely used for the separation and
pathway for distributed fertilizer production in remote purification of gas mixtures, and it consists of four
locations, but it is also proposed as a medium for short- major steps: pressurization, adsorption, blowdown, and
and long-term energy storage. It is particularly interesting purge.[19] For instance, Knaebel and Cussler established
because of its high hydrogen content (17.7 wt.%) and and proposed a multistep PSA scheme for ammonia sepa-
exceptional gravimetric energy density (3 kWh/kg).[6] ration, in which an enriched ammonia stream was used
Since renewable resources are available in small scale and to push the unabsorbed gas in the bed to obtain pure
are geographically distributed, distributed ammonia ammonia during desorption.[17] Later, more advanced
manufacturing in milder conditions has become an area of concepts were proposed which focused on combining
interest. Renewable and decentralized ammonia produc- PSA methodologies with a reaction. A reaction-
tion can be a cost-effective path for sustainable fertilizer adsorption process that incorporated a synthesis reactor
production and thermochemical energy storage of renew- and multiple adsorption units was demonstrated for
able resources based on the extensive knowledge available continuous ammonia synthesis and recovery.[20] Matito-
for this chemical process, its lower module costs, and sav- Martos et al. used Monte Carlo simulations and inte-
ings in transportation and distribution when deployed grated ammonia synthesis and adsorption, showing that
locally.[7–9] confinement of the ammonia synthesis reactor by adsor-
bents improved the ammonia synthesis conversion and
rates.[21] Some of the recent efforts toward enhanced
1.2 | State-of-the-art ammonia ammonia synthesis have been dedicated to low-pressure
separation technology ammonia manufacturing that can reduce capital and
operating costs.[10] Recently, an innovative, but simple,
The single-pass conversion in a conventional ammonia reaction-absorption process is proposed that enables
convertor is only about 15%–20%[2]; thus, to improve the ammonia production at 10–20 times lower pressure while
yield, it is necessary to separate ammonia from the reac- producing ammonia at rates comparable to those of con-
tor effluent and recycle unreacted nitrogen and hydrogen ventional process operating at high pressure.[22,23] A
back to the convertor. In conventional ammonia produc- major feature of the absorption-based separation is that
tion, ammonia is separated using phase-changing con- the absorber column, which is packed with metal halide
densation at high pressure (300 bar) and low temperature absorbents, can separate ammonia more completely at
(25 C) with water cooling. For lower pressure significantly lower ammonia concentrations.[9,24] Rapid
manufacturing (150 bar), a refrigeration system is separation achieved at fast recycle rates in the reaction-
required to achieve the lower condensation temperature absorption process can compensate for the reduction in
necessary to condense ammonia.[10] Since gas compres- the reaction rates due to lower pressure processing.
sion and refrigeration are two major contributors to the Unlike adsorption, which is a surface-based phenome-
total energy consumption in ammonia manufacturing, non, absorption relies on gas–solid reactions followed by
there is a critical need for designing separation processes bulk diffusion within the solid absorbents. Therefore, it
that reduce the process energy consumption.[2,10,11] One allows separation at higher temperature with 100% selec-
approach can be to replace phase-changing separation tivity for ammonia.[23,25]
with a more energy-efficient unit operation. It is widely
accepted that exploiting separations which obviate the
phase-changing can lead to energy savings up to one 1.3 | Scope of this study: Rapid pressure
order of magnitude.[12] swing adsorption-assisted adsorber
Few attention has been directed toward exploring
alternative separation methods for small-scale ammonia Extensive efforts have been reported on ammonia synthe-
manufacturing.[13] Unconventional methods for ammo- sis at milder conditions. Novel promoted Ru-based cata-
nia separation can be classified into two groups based on lysts can produce ammonia at pressures as low as 1 bar
LIN ET AL. 3 of 15

and 573 K.[26,27] Although operating at lower tempera- (compared to ammonia) in the adsorbents selected for
tures, the single-pass conversion is typically very low, this work. A report by Matito-Martos et al. suggests that
which makes it impractical to retrieve ammonia from the the loading of hydrogen on commonly used adsorbents
reactor effluent using conventional separation processes. such as 4A and Faujasite (FAU) is much lower than
PSA-based technologies can be applied for separation of that of nitrogen and almost negligible compared to that
dilute ammonia. Various PSA technologies, such as dual- of ammonia.[21] In addition, experimental studies on
reflux PSA and rectifying PSA, have been applied to con- hydrogen separation from gas mixtures using activated
centrate dilute gas streams.[28–31] Intensified PSA technol- carbon (AC) and zeolites suggests that hydrogen is the
ogies, such as rapid pressure swing adsorption (RPSA), least adsorbed component .[36,37] Therefore, for simplifi-
can be utilized to obtain shorter cycle times and a higher cation of the modeling approach, we only considered
productivity. RPSA has been widely applied in gas sepa- the mixture of nitrogen and ammonia for the RPSA
rations due to fast cycling and efficient separation.[32] For process.
instance, Turnock and Kadlec proposed rapid oscillating
adsorption for separating nitrogen from methane.[33]
Later, piston-driven RPSA was introduced by Eriksson[34] 2.1 | Single-component adsorption
and was then extended to dual-piston assisted RPSA by isotherm
Keller and Kuo.[35] This work is focused on evaluating a
conceptual scheme of RPSA-assisted adsorber for separat- Adsorption isotherms of NH3 and N2 are already
ing ammonia from a dilute gas mixture exiting the HB investigated and reported for various adsorbents. Sev-
reactor. eral zeolites (13X, 4A, 5A, and FAU), silica gel (Si),
The purpose of this work is to demonstrate the feasi- and AC were chosen as the candidates for RPSA stud-
bility of the RPSA-assisted adsorber, which can be even- ies in this work. The adsorption isotherms of pure
tually adopted for small-scale ammonia production. We NH3 and N2 at 298 K were obtained from the litera-
developed a detailed mathematical model for the pro- ture and are illustrated as markers in Figure 1.[21,38–
42]
posed technology, using binary adsorption isotherms The data points obtained from experimental and
determined by the ideal adsorbed solution theory simulation results can be well represented by the
(IAST). We investigated the performance of the RPSA Freundlich model:
process for ammonia separation in terms of several
important process variables. Results suggest that con- N i = k i P1=ni ð1Þ
centrated ammonia streams (+95 mol%) can be
obtained from the synthesis gas using ammonia-
selective adsorbents. The RPSA-assisted adsorber where Ni (mol/kg) and P (bar) are the amount of gas
enables continuous ammonia production when inte- adsorbed in the adsorbent and the equilibrium pres-
grated with a modular ammonia reactor for distributed sure, respectively. ki and ni are Freundlich isotherm
manufacturing. Our proposed RPSA process can be fur- constants for component i. After fitting the data from
ther extended and utilized for concentrating other the literature, ki and ni were found and are summa-
dilute gas mixtures (e.g., CO2). rized in Table 1 for the six adsorbents selected for this
study.
The adsorption isotherms of NH3 and N2 displayed in
2 | METHOD OLOG Y Figure 1 suggest that the adsorption capacity for NH3 is
significantly higher than that of N2 at the same tempera-
This modeling effort seeks to design a small RPSA- ture and pressure. Table 1 shows that the sorption uptake
based process for simultaneous separation and concen- capacity on 4A at 298 K reaches 9.83 mol NH3/kg and
tration of ammonia produced from a low-pressure HB 0.41 mol N2/kg at 1.5 bar and further increases to
reactor. The gas mixture in the reactor exhaust is com- 10.9 mol NH3/kg and 0.64 mol N2/kg at 3 bar. The 4A
posed of nitrogen and hydrogen in stoichiometric ratio zeolite adsorbent is interesting, as it shows the largest dif-
and approximately 10 mol% ammonia. For the simplic- ference between the ammonia and nitrogen equilibrium
ity of the modeling approach, we assumed that hydro- capacity at any specific condition. As displayed in Fig-
gen interaction with adsorbents is weak and similar to ure 1, 13X displays the largest ammonia uptake among
that of nitrogen. Hence, hydrogen was presumed to be six adsorbents at lower ammonia partial pressure. Hence,
approximated by nitrogen. Such assumption is not far 13X was chosen as the model adsorbent for this work to
from reality as prior literature suggests that both hydro- evaluate the performance of the RPSA process at different
gen and nitrogen are less competitively adsorbed process parameters.
4 of 15 LIN ET AL.

F I G U R E 1 Adsorption
isotherms of (A) NH3 and (B) N2 at
298 K. Data points were extracted
from references: NH3,[38] N2 -
13X,[39] 4A,[40] 5A,[41] Si,[42] AC,[39]
and FAU.[21] The fitted Freundlich
isotherm models displayed with
solid lines predict the uptake of
gases adsorbed on six adsorbents

T A B L E 1 Fitting parameters for


Adsorbent ki ni Uptake at 1.5 bar Uptake at 3.0 bar
the Freundlich isotherms of pure NH3
NH3[38] and N2 on 13X, 4A, 5A, Si, AC, and
13X 9.77 7.69 10.3 11.3 FAU at 298 K
4A 9.25 6.71 9.83 10.9
5A 7.91 8.76 8.28 8.97
Si 6.42 4.07 7.09 8.41
AC 4.40 1.61 5.66 8.71
FAU 1.73 2.06 2.11 2.95
N2
13X[39] 0.73 1.63 0.93 1.43
4A[40] 0.31 1.56 0.41 0.64
[41]
5A 0.76 3.19 0.86 1.07
[42]
Si 0.06 1.00 0.08 0.17
AC[39] 0.41 1.21 0.57 1.02
[21]
FAU 0.23 1.26 0.32 0.56

2.2 | Binary adsorption by IAST component i at the mixture sorption conditions (T and
π).[43] The spreading pressure can be obtained by inte-
The IAST model was applied to obtain the Freundlich grating Equation (5):
binary adsorption isotherms based on the Freundlich
single-component adsorption model.[43] Using the anal- Pði 0

 0  RT N i ðPi Þ
ogy of Raoult's law for vapor–liquid equilibria, the IAST π Pi = dPi ð5Þ
model leads to: A Pi
0

Pyi = Pi = x i P0i ðπ Þ for i = 1, 2 ð2Þ


where the spreading pressure π is a thermodynamic vari-
x1 + x2 = 1 ð3Þ able and can be calculated from the experimental adsorp-
tion isotherm. R is the gas constant, and A is the specific
y1 + y2 = 1 ð4Þ area of the adsorbent. Integration of Equation (5) at the
lower limit (p = 0 Pa) is undefined, so our calculations
used an approximation at p = 0.0001 Pa.
where xi and yi are mole fractions of component i in the For a given total pressure (P) and mole fraction in the
adsorbed phase and gas phase, respectively; P is the total gas phase of component i (yi), five equations were
pressure in the gas phase; Pi is the partial pressure of obtained from Equations (2), (3), and (5) for a binary sys-
component i; and P0i ðπ Þat the corresponding temperature. tem. This resulted in five unknowns—x1, x2, P01 , P02 , and
Thus, P0i ðπ Þ is the pure adsorbate vapor pressure for the π. Solving these equations numerically gave a solution
LIN ET AL. 5 of 15

for the unknowns, from which the total amount of the


adsorbate (Nt) can be calculated:

1 x1 x2
=   +  0 ð6Þ
N t N 1 P01 N 2 P2

  Nt is the total amount of gases adsorbed and


where
N i P0i is the amount of the component i adsorbed at the
spreading pressure π and temperature T, in the absence
of the other component in a binary mixture.
The relative volatility (α1,2) and selectivity coefficient
(S2,1) for binary mixtures is defined as:

y1=y 1
α1,2 = 2
= = S2,1 ð7Þ F I G U R E 2 Schematic of the RPSA-assisted adsorber. Feed gas
x 1=x S1,2
2
mixture of 10 mol% NH3 and 90 mol% N2 enters the adsorber in the
first half cycle (left-hand side). The less adsorbed gas (N2) leaves the
where α1,2 is equal to S2,1 which is defined as the selec-
bed from the exhaust end while the more adsorbed gas (NH3) is
tivity coefficient for the more adsorbed gas with respect enriched at the product end. A concentrated ammonia stream
to the other less adsorbed component. In this work, pure leaves the adsorber through the product line in the second half
adsorption isotherms of NH3 and N2 on six adsorbents at cycle (right-hand side). The blue line denotes adsorption while the
298 K were used to calculate Freundlich binary adsorp- orange line represents desorption. Two steps run consecutively in a
tion isotherms using the aforementioned equations. rapid manner

2.3 | RPSA process description 2.4 | Model equations

The RPSA-assisted adsorber consists of a fixed bed For the mathematic model of the RPSA-assisted adsorp-
column in which a certain component can be selec- tion process, the following assumptions are considered:
tively adsorbed and separated from a gas mixture.
Unlike a four-step Skarstrom cycle,[19] the RPSA- 1. Idea gas behavior is considered.
assisted adsorber involves only rapid adsorption and 2. An instantaneous equilibrium adsorption model is
desorption cycles inducing the separation of gas mix- considered.
ture. A schematic diagram of the RPSA-assisted 3. Axial dispersion is negligible (Pe > 100, see details in
adsorber is shown in Figure 2. A three-way solenoid Appendix A).[44]
valve at the adsorber inlet regulates the gas flow at 4. Adsorption equilibrium can be well predicted by
the feed and product ends, and an orifice at the Freundlich isotherm.
exhaust end allows adjustment of the exiting gas flow 5. The IAST model accurately predicts binary adsorp-
rate. A gas mixture of 10 mol% NH3 and 90 mol% N2 tion of NH3 and N2.
is supplied to the RPSA system at a certain pressure 6. Darcy's law characterizes the gas flow along the bed
during the first half cycle. Due to the small pressure (see details in Appendix B).
difference along the bed, the gas mixture flows from 7. Particle size and bed voidage are uniform for all the
the feed end toward the exhaust end with a small resi- adsorbent selected.
dence time in the adsorber. NH 3 is preferentially 8. Gas viscosity is independent of the pressure and
retained in the adsorbed phase at higher adsorber composition.
pressure while N 2 is being concentrated along the bed 9. Selectivity coefficient is constant.
to flow gradually toward the exhaust end. In the sec- 10. Separating ammonia from the mixture by RPSA is
ond half cycle, the three-way valve actuates to the assumed to be isothermal due to its small scale
product line where the adsorber pressure drops, and nature.
the ammonia in the adsorbent phase is released,
expanded, and flows toward the product end. The The model equations for the RPSA are already com-
rapid operation of these half cycles could eventually prehensively derived by Turnock and Kadlec.[33] The
separate the desired gas (NH 3) from the feed after a governing equations for the overall and single-
series of quick, continuous cycles. component material balances are given in Appendix C,
6 of 15 LIN ET AL.


which can be formulated in the dimensionless form p2F =p2a , 0 < θ≤1=2
shown below. Pð0,θÞ = ð13Þ
1, 1=2 < θ≤1
The dimensionless variables are defined as follows:
 
KA ∂PðX, θÞ V pa ∂Pð1, θÞ
2
P = PðX, θÞ = ½pðX, θÞ =p2a ð8Þ − = CQe +
2V m μLT ∂θ X =1 2RTλPð1, θÞ 1=2 ∂θ

X = l=L ð9Þ
For 0 ≤ X ≤ 1,
θ = t=λ ð10Þ
PðX, 0Þ = 1:0, for the 1st cycle: ð14Þ

where X and θ are the dimensionless length and time var- PðX, 0Þ for the mth cycle = PðX, 1Þ for the ðm −1Þth cycle:
iables, respectively.
The overall material balance can be given: The overall governing equation is a second-order par-
tial differential equation (PDE), which was converted
∂ 2 PðX, θÞ ∂PðX, θÞ into a set of ordinary differential equations (ODEs) using
= F ðθ Þ ð11Þ
∂X 2
∂θ the method of lines.[45] The set of ODEs was then solved
numerically by ODE45 in MATLAB® to obtain the pres-
where F(θ) is defined as a function of process parameters sure profile along the adsorber bed at any instant of the
formulated as: cyclic operation.
The oscillating pressure profile along the adsorber
 
TμV m L2 Aϕ 1 kW ð1=n − 1Þ 1 bed obtained from solving Equation (11) was used to cal-
F ðθ Þ = + p ð12Þ
T a KAλ RT P1=2 nL a Pð1 − 1=2nÞ culate the molar gas flow rate (qm) using Darcy's law,
outlined as follows:

where T is the temperature; μ is the gas viscosity, Vm is the T a p KA 1 dp


qm = − ð15Þ
molar volume; L is the bed length, Ta is the reference tem- T pa μ V m dl
perature, pa is the reference pressure, K is the permeability
of the bed, A is the cross-sectional area, λ is the cycle where dpdl represents the pressure gradient in the axial
time of the RPSA process, ϕ is the bed voidage, W is the direction of the adsorber column. Here, qm can be
weight of the adsorbent, and Nt is the total amount of obtained from the dimensionless discretized form as:
gases adsorbed. k and n are the Freundlich binary adsorp-
 
tion isotherm constants defined in Equation (1). They T a KA Pð1 −ΔX,θÞ −Pð1, θÞ
qm ð1,θÞ = ð16Þ
were obtained by solving the IAST model at various pres- 2Tpa μLV m ΔX
sures based on single-component adsorption isotherms at
298 K and the target feed composition. k and n for differ- where qm(1, θ) is the molar gas flow rate at the adsorber
ent adsorbent are summarized in Table 2. end (X = 1). P(1 − ΔX, θ) and P(1, θ) are the dimension-
The boundary and initial conditions used to solve the less pressure at the last two nodes of the adsorber outlet.
governing equation are given: The qm(X, θ) at any adsorber location can be calculated
from the integration of Equation (17):

ð1
T A B L E 2 Fitting parameters for the Freundlich binary ∂PðX, θÞdX
qm ðX,θÞ = qm ð1, θÞ + GðX, θÞ ð17Þ
isotherms for 10 mol% NH3-90 mol% N2 on 5A, 4A, 13X, Si, AC, ∂θ
and FAU at 298 K X

Adsorbent k n
where G(X, θ) can be calculated as follows:
5A 6.1 8.7
" #
4A 6.6 6.7 L Aϕpa kW pa 1=2
GðX, θÞ = + ð18Þ
13X 7.3 7.6 2λ RTPðX, θÞ1=2 n L PðX, θÞð1 − 1=2nÞ
Si 3.7 4.0
AC 1.2 1.5
Similarly, considering single component adsorption
FAU 0.6 1.9
from the gas mixture, the use of relative volatility (α)
LIN ET AL. 7 of 15

defined in Equation (7) can facilitate the computation of Similarly, the N2 Recovery and purity in the exhaust
single-component material balance in the system as line are defined:
follows:
Total moles of N2 in the exhaust line in one cycle
RN2 , E =
∂yðX, θÞ ∂yðX, θÞ Total moles of N2 in the feed line in one cycle
qm ðX, θÞ + f ðX, θÞ + hðX, θÞyðX, θÞ = 0
∂X ∂θ × 100%
ð19Þ ð26Þ

where f(X, θ) and h(X, θ) can be defined as a function of


Total moles of N2 in the exhaust line in one cycle
the process parameters: PN2 , E =
Total moles of gases in the exhaust line in one cycle
" # × 100%
1=n
1 ALϕPðX, θÞ1=2 pa pa PðX, θÞ1=2n Wkα
f ðX, θÞ = + ð27Þ
λ RT ½α − ðα−1ÞyðX,θÞ2
ð20Þ To evaluate the performance of various adsorbents
for NH3 recovery, the adsorbent productivity was
 
pa PðX, θÞð1=2n − 1Þ Wk
1=n
1 calculated[47]:
hðX, θÞ = −1
2λn α − ðα−1ÞyðX, θÞ
∂PðX, θÞ Productivity
∂θ Total moles of NH3 in the product line in one cycle
=
ð21Þ Volume of adsorbent × Total cycle time
The boundary and initial conditions are given: ð28Þ

yð0, θÞ = yf ð22Þ
3 | RESULTS A ND DISCUSSION
For 0 ≤ X ≤ 1:
The IAST was used to predict the binary adsorption iso-
yðX,0Þ = 1:0 for the 1st cycle: ð23Þ therms of six adsorbents for the NH3-N2 mixture. The iso-
therms were then incorporated into the process model.
yðX, 0Þ for the mth cycle = yðX, 1Þ for the ðm −1Þth cycle: The 13X adsorbent was chosen as the model adsorbent to
conduct a more comprehensive analysis of the effect of
The calculated cyclic steady-state pressure profile P important process variables on the performance of the
(X, θ) and the molar gas flow rate qm(X, θ) were used to proposed RPSA process.
calculate the gas phase composition y(X, θ). The gas
phase composition is a first-order parabolic partial differ-
ential equation that can be solved by the method of 3.1 | Binary adsorption
characteristics.[46]
Predicted binary isotherms for six adsorbents at different
pressures are shown in Figure 3. It is important to note
2.5 | Performance metrics that both NH3 and N2 equilibrium loadings are plotted
versus mole fraction of NH3 in the gas mixture for the
To evaluate the performance of the designed RPSA, the NH3 ease of comparison. A key finding from the binary iso-
recovery and purity in the product line are formulated: therms is that the adsorbents are more selective toward
ammonia and have a higher ammonia adsorption uptake
Total moles of NH3 in the product line in one cycle at similar conditions. This is favorable for our proposed
RNH3 , P =
Total moles of NH3 in the feed line in one cycle RPSA process, as ammonia is the desired compound to
× 100% be separated from the gas mixture. When considering the
ð24Þ composition of the gas effluent from an HB reactor
(10 mol% NH3), it can be found that the amount of
Total moles of NH3 in the product line in one cycle ammonia adsorbed on 13X will be about 200 times
PNH3 , P =
Total moles of gases in the product line in one cycle greater than that of nitrogen, indicating that nearly 99%
× 100% of adsorbed gas is ammonia. Among isotherms predicted
ð25Þ for the NH3-N2 binary system, 13X displayed the highest
8 of 15 LIN ET AL.

F I G U R E 3 Amount of NH3 and N2 on six adsorbents at 298 K and different pressures (A) 1.5 bar, (B) 3 bar, (C) 6 bar, and (D) 12 bar.
13X (green), 4A (red), 5A (black), Si (blue), AC (cyan), and FAU (magenta)

ammonia sorption capacity of 7.64 mol/kg. Similar to ammonia selectivity decreases as the ammonia mole frac-
13X, other adsorbents are more selective for ammonia at tion in the gas phase increases. For instance, the selectiv-
1.5 bar. Results indicate that the loading of ammonia in ity of 5A decreases from 5 × 104 to nearly one at 298 K.
the adsorbents decreases in the following order: This is primarily attributed to the fact that the nominator
13X > 4A > 5A > Si > AC > FAU. in Equation (7) remains constant, while the denominator
Table 2 summarizes the Freundlich binary isotherm exponential increases as ammonia concentration
parameters, k and n, for different adsorbents selected for increases. Overall, selectivity results suggest that FAU
this work. These parameters were obtained from solving and AC have relatively lower ammonia selectivity than
the IAST model at various pressures based on single- other adsorbents. The ammonia selectivity of these adsor-
component adsorption isotherms at 298 K and feed com- bents is in the following descending order:
position of 10 mol% NH3. The binary adsorption model 5A > 4A > 13X ≈ Si > AC ≈ FAU. It should be noted
for six adsorbents will be used in the simulation of the that there is a possibility that IAST overestimate the
proposed RPSA process. adsorption selectivity. Hence, experimental work on the
Another important observation from the predicted binary adsorption isotherms will be required to confirm
binary isotherms is that increasing the total pressure sacri- the predicted results.
fices the ammonia selectivity as the nitrogen uptake
increases more significantly, while leading to higher
adsorption capacities for both gases. Results suggest that 3.2 | RPSA cyclic profile
the increase in the pressure more notably influences the
NH3 and N2 adsorption capacities in AC. Overall, it can be The adsorber column design parameters for the proposed
deducted that lower pressure is more favorable for ammo- RPSA process are summarized in Table 3. The column
nia recovery. Therefore, we proceeded with 1.5 and 3 bar length was considered to vary between 0.05 and 0.09 m
as our target operating pressures, which are within the with a fixed diameter of 0.021 m. The column voidage
range we envision for small, distributed ammonia and average adsorbent size were estimated, and bed per-
production. meability was calculated from Kozeny–Carman equa-
For NH3-N2 binary system, the selectivity for the tion.[48] The effect of a few important variables, such as
more adsorbed gas was calculated using Equation (7), the exhaust flow rate and cycle time on the performance
and the results are displayed in Figure 4. Notably, the of the RPSA, were investigated.
LIN ET AL. 9 of 15

F I G U R E 4 Selectivity of
ammonia over nitrogen for six
adsorbents at (A) 1.5 and (B) 3 bar
and 298 K. 5A (black), 4A (red),
13X (green), Si (blue), AC (cyan),
and FAU (magenta)

T A B L E 3 Characteristics of the RPSA-assisted adsorber for


ammonia separation from a dilute gas mixture

Parameter Symbol Value/range


Column voidage ϕ 45%
−4
Adsorbent size (×10 m) dp 1.25
−11 2
Permeability (×10 m) K 2.50
Column diameter (m) D 0.021
Temperature (K) T 298
Parameters investigated
Exhaust flow (×10−6 m3/s) QE 4.75–11.0
Column length (m) L 0.05–0.09
Feed pressure (bar) P 1.5–3.0
Cycle time (s) λ 2.0–14

The set of PDEs proposed for the RPSA process in


Section 2.4 was solved for the six adsorbents selected,
with the base operating conditions being P = 1.5 bar,
T = 298 K, QE = 4.75 × 10−6 m3/s (1.93 × 10−4 mol/s), F I G U R E 5 Overview of the cyclic response of the proposed
λ = 10 s, L = 0.05 m, and yNH3 = 10 mol%. Figure 5 illus- RPSA-assisted adsorber packed with 13X at the base condition of
trates the overview of simulation results for the RPSA P = 1.5 bar, T = 298 K, QE = 4.75 × 10−6 m3/s, λ = 10 s, and
process using 13X adsorbent, during operation in three L = 0.05 m. (A) Pressure profile at the feed and exhaust ends;
consecutive cycles. Figure 5A displays the pressure pro- (B) Molar flow rate of the gas mixture at the feed and exhaust ends
(the red and blue lines denote the cyclic profile at the feed and
file within the adsorber at the feed and exhaust ends. The
exhaust end, respectively); and (C) average flow rate (magenta) and
pressure in the bed is controlled by switching the inlet
ammonia mole fraction (black) at the feed (product) end
three-way valve which periodically introduces the feed to
the adsorber during the first half cycle. Ammonia is pref-
erentially adsorbed at higher pressure during the first
half cycle, while nitrogen is being gradually transported increases and then decreases right after the three-way
to the exhaust end due to the pressure difference across valve switches. The pressure at the exhaust end reaches
the adsorber packed bed. The first half cycle is followed nearly 1.5 bar at a time constant of about 1 s (time con-
by the release of the pressure caused by switching the stant defined as the time required to achieve 63.2% of the
three-way valve to the product line. Ammonia is then overall change of a variable).
desorbed at lower pressure in the second half cycle. The Figure 5B shows the cyclic changes in the molar flow
pressure profile suggests that the pressure instantly rate of the gas mixture at the adsorber feed and exhaust
10 of 15 LIN ET AL.

ends. The pulse denotes the highest instantaneous molar indicating that ammonia can be completely separated from
flow rate of 0.065 mol/s at the beginning of the half cycle, the dilute gas mixture. It was found that the performance
followed by a rapid change in the flow at the inlet. As the of the RPSA-assisted adsorber improves at higher exhaust
gas mixture moves from the inlet toward the exit, the flow rates because more ammonia is introduced to the bed
molar flow rate decreases due to the shrinking pressure that is adsorbed on 13X. Once depressurized, more ammo-
gradient along the column, as displayed in Figure 5A. nia is subsequently released and collected in the product
Figure 5C displays the calculated average volumetric flow line. Therefore, it results in reduced nitrogen content in the
rate at the inlet of the absorber, which was obtained by column, hence avoiding the ammonia product being
integrating the flow rate over time. The data reveal that diluted with the nitrogen in the bulk phase. Increasing the
the average inlet flow rate of the pulse during the adsorp- cycle time also extends adsorption time, which is in favor
tion is about 1.12 × 10−5 m3/s, and it decreases to of the ammonia adsorption. Hence, the ammonia concen-
6.42 × 10−5 m3/s during desorption. The decrease in the tration will increase in the product line at longer
flow rate on the inlet side of the adsorber is because of cycle time.
the release of a portion of the nitrogen from the exhaust
line. A 10 mol% NH3 was fed to the adsorber, and the
ammonia purity in the product stream increased to about 3.4 | Effect of adsorber length
17.5% at the base conditions, at time that steady state was
achieved. One of the parameters with a major influence on the per-
formance of the proposed RPSA-driven adsorber is its vol-
ume. The major parameter of interest that should be
3.3 | Effect of exhaust flow and optimized is the adsorber length while the diameter was
cycle time

Figure 6 illustrates the effect of the exhaust volumetric flow


rate on the performance of the RPSA adsorber at different
cycling times. 13X adsorbent was used for this analysis.
Since 13X zeolite has excellent ammonia selectivity over
nitrogen, more than 95% of ammonia can be potentially
recovered from the gas stream at optimized conditions,
which can subsequently lead to producing a highly concen-
trated N2 gas (>99%) stream in the exhaust line. The purity
of NH3 in the product line and the recovered N2 in the
exhaust line versus the exhaust flow rate at different cycle
times are displayed in Figure 6A,B. It is apparent that
increasing the exhaust flow rate (from 4.75 × 10−6 to
1.1 × 10−5 m3/s) and the cycle time (from 2 to 14 s) results
in an increase in NH3 purity in the product line and N2 F I G U R E 7 Effect of the bed length on the performance of the
recovery in the exhaust line. The best performance can be RPSA-assisted adsorber in terms of NH3 purity in the product line
achieved when QE = 1.1 × 10−5 m3/s and λ = 14 s, which using 13X zeolite. (A) L = 0.05, (B) L = 0.07, and (C) L = 0.09 m at
concentrates ammonia from 10 mol% to above 99 mol%, T = 298 K, P = 1.5 bar, and QE = 4.75–11.0 m3/s

F I G U R E 6 Effect of exhaust
flow rate and cycle time on the
performance of the RPSA-assisted
adsorber using 13X zeolite.
(A) NH3 purity in the product line
and (B) N2 recovery in the exhaust
line, both operating at T = 298 K,
P = 1.5 bar, and L = 0.05 m
LIN ET AL. 11 of 15

kept constant. Figure 7 shows the effect of the adsorber concentration of ammonia. Hence, the feed pressure
length on the performance of the ammonia separation majorly influences the ammonia purity and nitrogen
using 13X adsorbent. We conducted the sensitivity analysis recovery. The effect of the feed pressure on the perfor-
on NH3 purity with respect to major design parameters mance of the RPSA-assisted adsorber is depicted in
and operating conditions, with the most important ones Figure 8. The ammonia purity in the product line and
being feed and exhaust flow rates, cycle time, and bed nitrogen recovery in the exhaust line are displayed with
length. To capture the effect of all these parameters in a respect to QE at different feed pressures. Results suggest
single graph, dimensionless form of the flow is used as the that separation efficiency is decreasing as the feed pres-
x-axis while having the effect of other major parameters, sure increases. The highest separation performance is
and results are illustrated Figure 7 for different bed lengths achieved when the feed pressure of 1.5 bar and large
and cycle times. The major emphasis is to study the effect exhaust flow rate are considered, as displayed in
of these parameters on NH3 purity. Figure 7A was adapted Figure 8A. Increasing the feed pressure to 2 and 3 bar
from Figure 6A at the condition of T = 298 K, P = 1.5 bar, results in a severe decrease in ammonia purity (30% and
and L = 0.05 m. As illustrated, the NH3 purity in the prod- 19%, respectively). The same trend is also observed in the
uct stream versus the exhaust-to-feed ratio (QE/QF) col- nitrogen recovery profile. The separation performance
lapses on a curve for all cycle times, with the highest decline at higher pressure which is due to the presence of
purity achieved with 14 s cycle time. Increasing the QE/QF more gases in the void fraction of the adsorber column at
leads to higher ammonia purity in the product line, since higher pressure. This decline in the ammonia purity in
almost all nitrogen is being directed toward the exhaust the product line is also partially attributed to the fact that
line. Thus, the RPSA separation is at its peak performance the increase in the feed pressure leads to a slight decrease
when exhaust flow approaches the feed flow: that is, when in the ammonia selectivity on 13X adsorbent, displayed
the amount of product produced is small. This results in in Figure 4A. For instance, the ammonia selectivity over
higher ammonia purity in the product line where ammo- nitrogen decreases from 1732 to 1093 as the feed pressure
nia is being desorbed in the second half cycle and leaving increases from 1.5 to 3 bar.
the adsorber through the product line. Note that
NH3 purity decreases from 99% to 47% and 32% when the
adsorber length increases from 0.05 m to 0.07 and 0.09 m 3.6 | Productivity of adsorbents
long at similar operating conditions, respectively. As the
adsorber length increases from 0.05 to 0.09 m, the purity Table 4 shows the comparison of the productivity of
obtained in the product line decreases due to adsorbents given that an adsorbent column can obtain
the ammonia product dilution by the nitrogen retained in nearly 99% NH3 with the ammonia recovery rate of
the adsorber bed. It is important to observe that the QE/QF 2.5 × 10−5 mol NH3/s. Notice that adsorbents need differ-
ratios are obtained from solving the set of PDEs. ent bed volumes for the same ammonia recovery rate via
the RPSA process, as summarized in Table 4. The
adsorber packed with FAU requires the largest volume,
3.5 | Effect of feed pressure that is, 3.5 × 10−5 m3, while the adsorber packed with 4A
is about twice smaller. Overall, the productivity of adsor-
Rapid pressure increase and decrease in the proposed bents follows this descending order 4A > Si > 13X >
adsorber is the primary mechanism of separation and 5A > AC > FAU. Nonetheless, it should be noted that

F I G U R E 8 Effect of feed
pressure on the performance of the
RPSA-assisted adsorber packed with
13X zeolite in terms of (A) NH3
purity in the product and (B) N2
recovery in the exhaust at
T = 298 K, L = 0.05 m, and λ = 10 s
12 of 15 LIN ET AL.

T A B L E 4 Comparison of bed volume and ammonia


productivity for different adsorbents. (a) Bed volume needed and
(b) productivity for different adsorber beds. Condition: T = 298 K,
P = 1.5 bar, λ = 14 s, and QE = 1.1 × 10−5 m3/s. The total amount
adsorbed was 3.5 × 10−4 mol per cycle

Adsorbent volume Productivity


Adsorbents (10−6 m3) (mol/s m3)
13X 10.2 2.43
4A 9.65 2.59
5A 13.3 1.89
Si 9.84 2.55 F I G U R E 9 Schematic of an RPSA-assisted ammonia
adsorption module with 20 adsorber mini columns packed with
AC 16.1 1.55
zeolite 4A. Each column is 0.021 m in diameter and 0.05 m in
FAU 35.0 0.71 length. Ten of the adsorbers are in the adsorption cycle (first half
cycle) while the other 10 are in the desorption cycle (second half
cycle)

the proposed RPSA-assisted adsorber with the specifica-


tions designed here is just for the modeling purpos and completely separated from the reactor effluent at nearly
has a relatively small throughput. atmospheric pressure, which can help with efforts
directed toward low-pressure ammonia production. The
proposed RPSA-assisted adsorber well fits the require-
3.7 | The potential application for ments for low-pressure ammonia production due to low
distributed ammonia manufacturing operating pressure and easy deployment.

The single-column, RPSA-assisted adsorptive separation


proposed in this study has the potential to be integrated 4 | CONCLUSION
with a low-pressure HB reactor for small, distributed
manufacturing of ammonia. Our proposed design An RPSA-assisted adsorber is proposed for the separa-
includes a number of adsorber columns (e.g., 20) bundled tion of dilute ammonia from the effluent of the HB
together. A schematic of the proposed integrated design reactor. For the proof-of-concept, a mathematical
is shown in Figure 9, in which each column is packed model is proposed to study the effect of different pro-
with zeolite 4A. This adsorbent was selected as it dis- cess variables on the performance of the RPSA-assisted
played the highest ammonia productivity, in comparison adsorber. The results suggest that high-purity ammonia
to other adsorbents considered in this study (see Table 4). can be obtained by adjusting the exhaust flow rate,
At any half cycle, half of the RPSA-assisted adsorber col- cycle time, feed pressure, and bed length. Six different
umns are in the adsorption (charging) mode while the adsorbents with relatively high ammonia selectivity
other half adsorber columns are in the desorption were studied, and it was found that 4A displayed the
(ammonia discharge) mode. When the three-way valve best performance (more than 95% ammonia purity).
switches, the feed flow will be directed to the other The proposed design can be further intensified for the
adsorber columns that were previously in the heat management of adsorption and desorption cycles
desorption mode. when multiple mini adsorbers are bundled together.
The zeolite 4A adsorber makes the proposed process The proposed RPSA-based separation has the potential
economically affordable due to the relatively lower cost. for implementation in other dilute gas mixture separa-
Each of the adsorbers displayed in Figure 9 can separate tions, such as carbon dioxide capture from the flue gas
2.5 × 10−5 mol NH3/s from the gas mixture on the basis given that adsorbents with high selectivity for CO2 are
of 3.5 × 10−4 mol NH3 per cycle per column (Table 4). identified. Since simulations were carried out based on
For NFUEL ammonia plants with a production rate of the predicted Freundlich binary isotherm models by
120 kg/h NH3,[49] the number of adsorber bundles IAST, there is a possibility that adsorbent capacity and
needed is about 4760 adsorber columns. The most prom- selectivity be overestimated. Experimental evaluation of
ising feature of using an RPSA-assisted process for binary adsorption will be useful in validating the pro-
ammonia separation is that dilute ammonia can be posed design.
LIN ET AL. 13 of 15

NO TATIO N ORCID
Latin symbols Mahdi Malmali https://orcid.org/0000-0001-5190-1261
A cross-sectional area of the pipe, m2
d diameter, m RE FER EN CES
D diameter of the column, m [1] J. W. Erisman, M. A. Sutton, J. Galloway, Z. Klimont, W.
k Freundlich isotherm model, constant Winiwarter, Nat. Geosci. 2008, 1, 636.
K permeability of bed, m2 [2] J. R. Jennings, Catalytic Ammonia Synthesis, Fundamentals
L total length of the column, m and Practice, Plenum Press, New York 1991.
[3] I. Rafiqul, C. Weber, B. Lehmann, A. Voss, Energy. 2005, 30,
m the mth cycle
2487.
n Freundlich isotherm model, constant [4] J. Nørskov, J. Chen, R. Miranda, T. Fitzsimmons, R. Stack, US
N amount of adsorbed gas, mol/kg DOE Office of Science, Published online, 2016.
pa reference pressure, bar [5] X. Liu, A. Elgowainy, M. Wang, Green Chem. 2020, 22, 5751.
pF feed pressure, bar [6] J. Guo, P. Chen, Chem 2017, 3, 709.
P dimensionless pressure variable [7] M. Reese, C. Marquart, M. Malmali, K. Wagner, E. Buchanan,
qm molar flow rate, mol/s A. McCormick, E. L. Cussler, Ind. Eng. Chem. Res. 2016, 55,
Q average flow rate, m3/s 3742.
[8] E. Cussler, A. McCormick, M. Reese, M. Malmali, J. Visualized
R gas constant, 8.314, J mol−1 K−1
Exp. 2017, 126, e55691.
S selectivity [9] M. Malmali, G. Le, J. Hendrickson, J. Prince, A. V. McCormick,
t time, s E. L. Cussler, ACS Sustainable Chem. Eng. 2018, 6, 6536.
T temperature of the column, K [10] B. Lin, T. Wiesner, M. Malmali, ACS Sustain. Chem. Eng.
Ta ambient temperature, K 2020, 3, 15517.
W weight of the adsorbent, g [11] A. Sánchez, M. Martín, J. Cleaner Prod. 2018, 178, 325.
x mole fraction in the adsorbed phase [12] D. S. Sholl, R. P. Lively, Nature 2016, 532, 435.
[13] C. Smith, A. K. Hill, L. Torrente-Murciano, Energy Environ.
X dimensionless length variable
Sci. 2020, 13, 331.
y mole fraction
[14] C. Y. Liu, Development of materials for ammonia separation-
storage in a small scale ammonia synthesis process. Ph.D.
Greek symbols thesis, Tokyo Institute of Technology, 2002.
α relative volatility [15] S. Sircar, Sep. Sci. Technol. 1990, 25, 1087.
λ cycle time, s [16] C. Y. Liu, K. Aika, Ind. Eng. Chem. Res. 2004, 43, 7484.
μ gas viscosity, Pa s [17] K. S. Knaebel, E. L. A. Cussler, Novel Pressure Swing Adsorp-
θ dimensionless time variable tion System for Ammonia Synthesis, Springer, Boston 1996,
p. 457.
ϕ porosity
[18] H. Yang, Z. Xu, M. Fan, R. Gupta, R. B. Slimane, A. E. Bland,
I. Wright, J. Environ. Sci. 2008, 20, 14.
Indices [19] D. M. Ruthven, S. Farooq, K. S. Knaebe, Pressure Swing
E exhaust Adsorption, Vol. 1, VCH Publishers, Weinheim 1994.
F feed [20] J. D. Beach, J. D. Kintner, A. W. Welch, Removal of Gaseous
g gas NH3 from an NH3 Reactor Product Stream, U.S. Patent
i component index 0339911 A1, 2018.
M Molar [21] I. Matito-Martos, J. García-Reyes, A. Martin-Calvo, D.
Dubbeldam, S. Calero, J. Phys. Chem. C 2019, 123, 18475.
p particle
[22] D. K. Ojha, M. J. Kale, A. V. McCormick, M. Reese, M.
P product Malmali, P. Dauenhauer, E. L. Cussler, ACS Sustainable
t total Chem. Eng. 2019, 7, 18785.
V volume [23] M. Malmali, Y. Wei, A. McCormick, E. L. Cussler, Ind. Eng.
Chem. Res. 2016, 55, 8922.
A U T H O R C ON T R I B U T I O NS [24] C. Smith, M. Malmali, C.-Y. Liu, A. V. McCormick, E. L.
Bosong Lin: Data curation; formal analysis; investigation; Cussler, ACS Sustainable Chem. Eng. 2018, 6, 11827.
[25] M. Malmali, M. Reese, A. V. McCormick, E. L. Cussler, ACS
writing-original draft. I-Min Hsieh: Conceptualization;
Sustainable Chem. Eng. 2018, 6, 827.
data curation; formal analysis; investigation; methodology;
[26] H. B. Chen, J. D. Lin, Y. Cai, X. Y. Wang, J. Yi, J. Wang, G.
validation; visualization. Mahdi Malmali: Conceptualiza- Wei, Y. Z. Lin, D. W. Liao, Appl. Surf. Sci. 2001, 180, 328.
tion; formal analysis; funding acquisition; investigation; [27] M. Hattori, T. Mori, T. Arai, Y. Inoue, M. Sasase, T. Tada, M.
project administration; resources; supervision; writing- Kitano, T. Yokoyama, M. Hara, H. Hosono, ACS Catal. 2018,
review & editing. 8, 10977.
14 of 15 LIN ET AL.

[28] M. Yoshida, J. A. Ritter, A. Kodama, M. Goto, T. Hirose, Ind. APP E NDIX A: AXIA L D IS P ER SI O N E FFE CT
Eng. Chem. Res. 2003, 42, 1795.
[29] A. D. Ebner, J. A. Ritter, AIChE J. 2002, 48, 1679. Peclet number (Pe) can be used to quantify the degree of
[30] T. S. Bhatt, G. Storti, R. Rota, Chem. Eng. Sci. 2015, 122, 34.
dispersion in the RPSA system.
[31] R. Weh, G. Xiao, M. A. Islam, E. F. May, Sep. Purif. Technol.
2020, 235, 116215.
[32] R. R. Vemula, M. V. Kothare, S. Sircar, AIChE J. 2016, 62, Pe = vg Lb =DL , ðA1Þ
1212.
[33] P. H. Turnock, R. H. Kadlec, AIChE J. 1971, 17, 335. where vg is the mean gas velocity along the adsorber col-
[34] R. Eriksson, Fractionating Apparatus, U.S. Patent 4169715A, umn, Lb is the column length, and DL is the axial disper-
1977. sion coefficient. DL can be estimated from the correlation
[35] G. Keller, C.-H. A. Kuo, Enhanced Gas Separation by Selective
given by Ruthven,[50]
Adsorption, EP0071964A1, 1981.
[36] F. V. S. Lopes, C. A. Grande, A. M. Ribeiro, J. M. Loureiro, O.
Evaggelos, V. Nikolakis, A. E. Rodrigues, Sep. Sci. Technol.
DL = γ 1 DM + γ 2 dp vg , ðA2Þ
2009, 44, 1045.
[37] Y. Park, Y. Ju, D. Park, C. H. Lee, Chem. Eng. J. 2016, 292, 348. where γ 1 and γ 2 are constants with typical values, 0.7 and
[38] J. Helminen, J. Helenius, E. P. Turunen, E. Paatero, J. Chem. 0.5, respectively. dp is the adsorbent particle diameter, and
Eng. Data 2001, 46, 391. the molecular diffusivity DM can be obtained from the
[39] D. P. Bezerra, R. S. Oliveira, R. S. Vieira, C. L. Cavalcante, Fuller, Schettler, and Giddings correlation,[51] which leads
D. C. S. Azevedo, Adsorption 2011, 17, 235.
to DM = 1.7 × 10−5 m2/s for the NH3–N2 binary system.
[40] E. D. Akten, R. Siriwardane, D. S. Sholl, Energy Fuels 2003,
17, 977.
Given that vg = 0.64 m/s, Lb = 0.05 m, and dp = 1.22
[41] P. B. Lederman, Adsorption of Nitrogen-Methane on Linde × 10−4 m, DL = 5.1 × 10−5 m2/s, Pe = 629. Since Pe >
Molecular Sieve Type 5A, Ph.D. Thesis, University of Michigan, 100, it can be concluded that the column operates under
1961. near plug flow conditions and the effect of axial disper-
[42] Z. Zhang, J. Guan, Z. R. Ye, D. Note, Adsorption 1998, 4, 173. sion on the system is very slight.[44] Therefore, axial dis-
[43] A. L. Myers, J. M. Prausnitz, AIChE J. 1965, 11, 121. persion can be omitted in the material balance.
[44] T. K. Perkins, O. C. Johnston, SPE J. 1963, 3, 70.
[45] W. E. Schiesser, The Numerical Method of Lines: Integration of
Partial Differential Equations, Elsevier, Amsterdam, 2012.
[46] H. S. Mickley, T. K. Sherwood, C. E. Reed, Applied Mathemat-
A P P E N D I X B : V A L I D I T Y O F D A R C Y ' S LA W
ics in Chemical Engineering, McGraw-Hill, New York, 1957.
[47] N. A. A. Qasem, R. Ben-Mansour, Appl. Energy 2018, 209, 190. In the RPSA process, the frictional pressure drop contri-
[48] B. R. Corrochano, J. R. Melrose, A. C. Bentley, P. J. Fryer, S. bution can be considerable compared to the simple PSA
Bakalis, J. Food Eng. 2015, 150, 106. that frictional pressure drop is typically negligible.[52,53]
[49] A. Patil, L. Laumans, H. Vrijenhoef, in Procedia Eng., Vol. 83,
2014, p. 322.
[50] D. M. Ruthven, Principles of Adsorption and Adsorption Pro-
cesses, John Wiley & Sons, New York, 1984.
[51] E. N. Fuller, P. D. Schettler, J. C. Giddings, Ind. Eng. Chem.
1966, 58, 18.
[52] E. M. Kopaygorodsky, V. V. Guliants, W. B. Krantz, AIChE J.
2004, 50, 953.
[53] V. R. Rao, S. Farooq, W. B. Krantz, AIChE J. 2009, 56, 354.
[54] R. B. Bird, E. N. Lightfoot, W. E. Stewart, Transport Phenom-
ena, John Wiley & Sons, New York, 2002.

How to cite this article: Lin B, Hsieh I-M,


Malmali M. Rapid pressure swing adsorption for
small scale ammonia separation: A proof-of-
concept. J Adv Manuf Process. 2021;3:e10077.
https://doi.org/10.1002/amp2.10077 F I G U R E B 1 Velocity, Reynolds number, and pressure profile
along the bed length. (Darcy's law vs. Ergun equation)
LIN ET AL. 15 of 15

Darcy's equation is typically valid for Re < 10 (laminar Boundary conditions for Equation (1,C1):
flow) whereas Ergun's equation can predict the pressure At l = 0,
drop and flux in wider range Re.[54] To verify that Darcy's

law is applicable to our case, we compared the pressure pF , 0 < t≤λ=2
pð0, t Þ = ðC4Þ
drop/velocity calculated by Darcy's law and Ergun equa- 1, λ=2 < t≤λ
tion, and results are illustrated in Figure B1. Our calcula-
tions suggest that Re within the packed absorber column
ranges from 3 to 10, thus the flow tends to be in laminar At l = L,
mode. Comparison of results suggests that both Ergun
equation and Darcy's law are capable of predicting flow V ∂pðL, tÞ Qp
= qm − ðC5Þ
characteristics within the range studied for RPSA process. RT a ∂t Vm
We adopted Darcy's law to characterize the interstitial
gas velocity based on the local pressure gradient, as Initial condition for Equation (1,C1) (for 0 ≤ l ≤ L):
suggested by prior reports.[52,53]
pðl,0Þ = pF , for the 1st cycle: ðC6Þ

A P P EN D I X C : D I M EN SI O N A L F O RM O F
pðl,0Þ for the mth cycle = pðl,λÞ for the ðm −1Þth cycle:
E Q U A T IO N S

Overall mass balance: Boundary conditions for Equation (2,C2):


For 0 ≤ t ≤ λ/2,
Aϕ ∂p ∂q W ∂N t
=− m− ðC1Þ
RT ∂t ∂l L ∂t yð0, t Þ = yF ðC7Þ

Component mass balance:


Initial condition for Equation (2,C2) (for 0 ≤ l ≤ L):
Aϕ ∂ ðypÞ ∂ ðyqm Þ W ∂ ðxN t Þ
=− − ðC2Þ
RT ∂t ∂l L ∂t yðl,0Þ = yF , for the 1st cycle: ðC8Þ

Momentum equation: yðl,0Þ for the mth cycle = yðl,λÞ for the ðm −1Þth cycle:

1 T a p KA ∂p All the parameters and variables are defined in the


qm = − ðC3Þ
V m T pa μ ∂l notation list.

You might also like