Journal of Membrane Science 614 (2020) 118483

Contents lists available at ScienceDirect

Journal of Membrane Science

journal homepage: http://www.elsevier.com/locate/memsci

A catalytic composite membrane reactor system for hydrogen production

from ammonia using steam as a sweep gas
Yongha Park a, Junyoung Cha a, Hyun-Taek Oh b, Taeho Lee a, Sung Hun Lee c, Myung Gon Park c,
Hyangsoo Jeong a, Yongmin Kim a, Hyuntae Sohn a, Suk Woo Nam a, d, Jonghee Han a, d,
Chang Won Yoon a, e, **, Young Suk Jo a, *
a
Center for Hydrogen and Fuel Cell Research, Korea Institute of Science and Technology (KIST), 5 Hwarang-ro 14-gil, Seongbuk-gu, Seoul, 02792, Republic of Korea
b
Department of Chemical and Biomolecular Engineering, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul, 03722, Republic of Korea
c
Wonik Materials Co., 30 Yangcheong 3-gil, Ochang-eup. Chengwon-gu, Cheongju-si, Chungcheongbuk-do, 28125, Republic of Korea
d
Green School, Korea University, Anam-ro 145, Seongbuk-gu, Seoul, 02841, Republic of Korea
e
KHU-KIST Department of Converging Science and Technology, Kyung Hee University, 26 Kyungheedae-ro, Dongdaemun-gu, Seoul, 02447, Republic of Korea

A R T I C L E I N F O A B S T R A C T

Keywords: For catalytic reactions involving H2 extraction, the membrane reactor is an attractive option for enhancing the
Catalytic membrane reactor equilibrium and kinetics while eliminating excessive purification steps. In this study, a steam carrier adopted
Pure hydrogen production composite membrane reactor system is developed to produce pure H2 (>99.99%) from ammonia with high H2
Ammonia decomposition
productivity (>0.35 mol-H2 g−cat1 h− 1) and ammonia conversion (>99%) at a significantly reduced operating
Steam sweep gas
Membrane reactor process demonstration
temperature (<723 K). Coupling of a custom developed palladium/tantalum composite metallic membrane and
ruthenium on lanthanum-doped alumina catalysts allowed stable operation of the membrane system with sig­
nificant mass transfer enhancement. Various reactor assemblies involving as-fabricated membranes and catalysts
are experimentally compared to suggest the optimal configuration and operating conditions for future applica­
tions. Steam is adopted as a sweep gas, presenting efficient H2 recovery (>91%) while replacing conventionally
utilized noble carrier gases that require additional gas separation processes. The steam carrier presents similar
membrane reactor performance to that of noble gases, and the water reservoir used for steam generation acts as
an ammonia buffer via scrubbing effects. Finally, electricity generation is demonstrated using a commercial fuel
cell along with process simulation, substantiating potential of the proposed membrane system in practical ap­
plications for H2 production from ammonia and on-site power generation.

1. Introduction being actively investigated.

The renewable energy share is growing as a mean to mitigate the
global climate crisis and maintain sustainability. While renewable en­
ergy has great potential, its unpredictable nature presents challenges for
direct integration into the grid. Additionally, either 1) it has a small
energy production scale or 2) large production sites are too distant from
energy utilization sites. Therefore, a large amount of renewable energy
remains untapped, with limited logistics to store and transport the en­
ergy efficiently and economically. To resolve these temporal and spatial
discrepancies between the supply and demand of renewable energy,
energy carriers with high energy densities and economic feasibilities are
Recently, ammonia has been identified as one of the energy carriers
that allow efficient storage of renewable energy in a liquid form at a mild
pressure under atmospheric conditions (0.8 MPa, 293 K [1]). When
liquefied, it has a high energy density of 22.5 MJ kg− 1 (higher heating
value), with high gravimetric and volumetric H2 contents of 17.6 wt%
and 108 g L− 1 at 293 K [2], respectively. Additionally, well-established
storage and distribution infrastructure makes ammonia an economically
feasible mass-scale energy carrier. While ammonia synthesis is
energy-intensive, efforts have been directed toward making the process
“greener” by adopting water electrolysis in conjunction with a conven­
tional Haber Bosch process [3] or by electrochemical synthesis [4]. This

* Corresponding author.
** Corresponding author. Center for Hydrogen and Fuel Cell Research, Korea Institute of Science and Technology (KIST), 5 Hwarang-ro 14-gil, Seongbuk-gu, Seoul,
02792, Republic of Korea.
E-mail addresses: cwyoon@kist.re.kr (C.W. Yoon), yjo@kist.re.kr (Y.S. Jo).

https://doi.org/10.1016/j.memsci.2020.118483
Received 24 April 2020; Received in revised form 2 July 2020; Accepted 4 July 2020
Available online 25 July 2020
0376-7388/© 2020 Elsevier B.V. All rights reserved.
Y. Park et al.
green ammonia can be shipped from the renewable energy producer to
the renewable energy consumer, whereby it can be cracked into H2 and
N2. H2 production from ammonia, shown in Fig. 1, is a well-known
catalytic process involving catalytic chemical reaction and H2 purifica­
tion steps.
For the ammonia decomposition reaction, Ru is known to be the most
active metal, and various support materials, such as alumina [5,6],
mesoporous silica [7,8], activated carbon [9], metal oxides [10], and
carbon nanotubes [11] are being explored as effective support materials
[12,13]. Additionally, there have been significant efforts to replace Ru
with non-noble metals as active sites owing to its cost and scarcity [5,14,
15]. Even after decades of research, a large amount of work is being
performed to reduce the reaction temperature and the metal loading of
expensive Ru. After the catalytic NH3 decomposition reaction, there are
two purification steps required to obtain H2 with a high purity: removal
of unreacted NH3 and separation of N2. Even at a high temperature and
extensive catalyst loading, there remains a small amount of unreacted
NH3 in the product stream [16], which can be detrimental to a
proton-exchange membrane fuel cell (PEMFC) having a trace level (~10
ppm) of NH3 tolerance [17,18]. While there are numerous reports on the
development of catalysts for ammonia decomposition, few studies have
investigated the whole process, including H2 purification. Kojima et al.
studied the effects of Li-exchange X-type zeolite (Li-X) as an adsorbent
material for unreacted ammonia [19], and Dolan et al. purified the
product stream from ammonia decomposition using a Pd-coated tubular
V membrane to get >99.99% pure H2 [20]. Our group recently
demonstrated operation of a 1-kW-class fuel cell using H2 produced by
an ammonia decomposition reformer, along with the removal of NH3
using 13X zeolite, resulting in <1 ppm of NH3 in the product stream
[16].
Finally, after the removal of NH3, high-purity H2 can be produced by
employing pressure swing adsorption (PSA) or membrane processes to
separate N2, which accounts for 25 vol.% of the produced gas. This
adsorption-based H2 purification steps can comprise a significant
portion of the fuel reforming process, leading to high H2 costs [21].
While it is possible to operate a PEMFC without separating N2 (as it is
harmless to the fuel-cell stack), this reduces the H2 partial pressure and
prohibits H2 recirculation, adversely affecting the stack performance.
Also, the N2 content in the product stream will reduce the energy density
of the compressed H2 produced from ammonia to be used in the H2
refueling station.
The catalytic reaction and H2 purification can be conducted sepa­
rately, but this makes the NH3 decomposition process complex and
energy-intensive. To address this issue, membrane reactor concepts have
been suggested. Membrane reactors are particularly beneficial for the
NH3 decomposition reaction, for several reasons: (i) high-purity H2 can
be produced with no extra extensive NH3 purification or N2 separation
steps, allowing power generation using fuel cells with a low NH3 toler­
ance or use as a pure H2 source for H2 charging stations; (ii) enhanced
NH3 decomposition reaction kinetics can be achieved with in situ
extraction of H2, as described by the Temkin–Pyzhev equation [22],
reducing the reaction temperature; (iii) the thermodynamic equilibrium
can be shifted, which is becoming important as the catalyst performance
Fig. 1. Ammonia as a renewable energy carrier: H2 production process from ammonia.
2
Journal of Membrane Science 614 (2020) 118483
is constantly being improved, reducing the reaction temperature,
potentially to the point where the equilibrium NH3 decomposition
conversion starts to drop significantly (below 98% at 623 K [12]). While
theoretical benefits of CMR are widely accepted, it is important to note
that the increase in conversion and catalytic activity might be marginal
in some cases. Detailed analyses are thus required to truly benefit from
the adoption of CMR. In the following, the proposed membrane reactor
concepts for NH3 decomposition are reviewed, with references to studies
performed from 1994 to the present.
Early studies on NH3 decomposition membrane reactors focused on
application for the reduction of NOX emissions. Collins et al. designed a
membrane reactor using a Pd-ceramic membrane with Ni/Al2O3 cata­
lysts. The NH3 conversion increased from 53% (packed bed reactor,
denoted as PBR) to 94% (catalytic membrane reactor, denoted as CMR)
at 873 K [23]. Gobina et al. modeled an NH3 membrane reactor using
experimental H2 permeation data of Pd–Ag membrane and an NH3
decomposition kinetic model for Ni/Al2O3 catalysts [24], and Abashar
et al. performed simulations to determine the equilibrium shift via in situ
H2 extraction [25]. Later studies focused more on the use of NH3 as a H2
carrier, to produce pure H2 as an energy carrier. Zhang et al. demon­
strated the production of 99.96% pure H2 with a recovery yield of 77%
using an electroless plated Pd membrane with a Ni/La–Al2O3 catalyst
[26]. García-García et al. fabricated a CMR utilizing a Pd membrane and
Ru/C catalysts for producing high-purity H2, in an attempt to exceed the
equilibrium conversion at a temperature lower than 653 K [27]. Israni
et al. adopted a hollow fiber membrane reactor with Pd nanopore
membranes and Ni catalysts with 10% conversion increase with CMR
[28].
While most studies utilized Pd membranes, Li et al. fabricated a
bimodal CMR by doping Ru on a silica membrane support, which served
as a catalytic support layer. The ammonia conversion was significantly
higher for the bimodal configuration (95% at 723 K) than for the
monomodal configuration (45% at 723 K) [29]. In a follow-up study, Li
et al. examined the effects of the γ-Al2O3 impregnation amount, pore
structure, and catalytic activity of Ru on the performance of a bimodal
membrane reactor. The purity of the permeate H2 stream was 84%, with
4.5% NH3 [30]. While the silica membrane is one of viable membrane
options for NH3 CMRs, its low H2 selectivity necessitates an additional
adsorptive device to remove NH3 and N2. More recent works again
adopted advanced Pd membrane fabrication and catalyst synthesis
techniques. Itoh et al. fabricated an NH3 CMR using a 0.2-mm-thick
tubular Pd membrane with a Ru/SiO2 catalyst and achieved a 15%
conversion increase compared to a PBR at 723 K [31]. Rizzuto et al.
utilized a commercial CMR with a Pd membrane and Ru catalyst in an
attempt to identify the operation condition where equilibrium conver­
sion could be achieved with no use of sweep gas [32]. Zhang et al.
produced >99.7% pure H2 using a CMR with a 6-μm-thick Pd film
deposited on a Ru-impregnated yttria-stabilized zirconia tube, allowing
almost complete NH3 conversion at temperatures as low as 673 K [33].
In addition to experimental studies, a few simulation studies have been
performed, with developed models coupling the NH3 reaction kinetics,
gas diffusion in the catalyst bed, and H2 permeation through the mem­
brane via the solution-diffusion mechanism [34–37].
Y. Park et al.
Previous studies clearly indicate the effectiveness of coupling H2
separation using a membrane with NH3 decomposition to increase NH3
conversion. However, for the practical application of the NH3 CMR
technology, further investigation of various aspects is required. First, the
optimal membrane types for reducing the usage of Pd and increasing the
purity of the produced H2 must be identified. Then, the productivity or
throughput of the CMR must be improved to increase the H2 production
per Ru metal and membrane area. Finally, the use of a vacuum pump or
inert gas carrier in the permeate side must be avoided, or there must be a
novel way to increase the H2 partial pressure difference between the
feed and permeate sides. Our group recently investigated the viability of
a Pd coated body-centered cubic (V, Nb, Ta) composite dense metallic
membranes for high-purity H2 production from mixed gas separation
[38,39], and applied it to an NH3 decomposition, benefiting from the
matched operation region between H2 permeation and the NH3
decomposition reaction [40]. However, to further substantiate the
application of the composite membrane reactor concept, enhanced mass
transfer for higher H2 productivity, performance analysis over wide
range of operation conditions, and removal of commonly used inert
sweep gases (e.g. N2, Ar, etc), were mandatory.
In this study, new CMR configurations utilizing a custom fabricated
Pd/Ta/Pd composite membrane and optimized Ru/La–Al2O3 pellet
catalysts are developed to substantially increase the H2 productivity and
recovery yield while reducing the Ru catalyst loading for better utili­
zation of the limited membrane surface area and Ru catalysts. Two CMR
configurations and one PBR configuration are compared at different
temperatures, pressures, NH3 feed rates, and sweep gas flowrates to
determine the optimal operation conditions for an NH3 CMR adopting
dense metallic composite membranes. Then, steam is adopted as a sweep
gas to reduce the permeate-side H2 pressure while acting as an NH3 trap
in the case of membrane failure or leakage. Finally, as-produced H2 from
the CMR was directly fed to a small-scale PEMFC to demonstrate onsite
H2 and electricity generation applications, and a membrane process
simulation was further conducted to balance the heat requirement of the
steam generation and endothermic NH3 decomposition.
2. Experimental
2.1. Fabrication of membranes and catalysts
For Pd/Ta/Pd composite membrane fabrication, dense Ta tubes
(>99.95%) with an outer diameter of 6.35 mm and a wall thickness of
0.25 mm were purchased from Koralco Co., (Republic of Korea). The
lengths of the Ta tubes tested were 100 and 50 mm for higher NH3 feed
and lower NH3 feed, respectively. To deposit the thin Pd layer on the
outside and inside surfaces of the Ta tubes, electroless plating (ELP) was
employed. The detailed procedure of the ELP method adopted in this
study was presented in previous reports [38,39,41]. The thickness of the
Pd layer on the Ta tube was controlled by changing the plating time, and
the thickness of the coating layer was determined via scanning electron
microscopy (SEM, Inspect F, FEI). After the Pd plating on the Ta tube,
heat treatment was performed under a N2 atmosphere at 773 K for 2 h to
form a dense Pd layer on the Ta surface and to reduce the voids between
Pd and Ta layers for better adhesion. The crystallinity analysis was
conducted using X-ray diffraction spectroscopy.
For the catalyst preparation, ruthenium chloride hydrate
(RuCl3∙xH2O, 40.2 wt% Ru, Sigma–Aldrich), lanthanum nitrate hexa­
hydrate (La(NO3)3⋅6H2O, Crown Guaranteed Reagents), and γ-Al2O3 1/
8′′ pellet (Alfa Aesar) were purchased and used as received, unless stated
otherwise. The γ-Al2O3 1/8′′ pellet was pulverized and sieved to
appropriate pellet sizes (1–2 mm) considering the inner diameter of the
Pd/Ta/Pd tubular membrane. La (target concentration of 10 mol%) was
doped onto the γ-Al2O3 as a promoter of the catalyst via immersion in a
lanthanum nitrate solution followed by a synthesis condition introduced
in a previous study [16]. Subsequently, Ru was deposited on La-γ-Al2O3
supports via the wet impregnation method. The supports were immersed
3
Journal of Membrane Science 614 (2020) 118483
in an RuCl3∙xH2O solution, followed by drying in a vacuum oven. A
quantitative analysis of the metal contents in the Ru/La–Al2O3 catalyst
was performed using inductively coupled plasma optical emission
spectroscopy (720 ICP-OES, Agilent Technologies, Inc.). The results
indicated a composition of approximately 0.65 wt% Ru and 10 mol% La
ions.
2.2. Design and operation of CMR
A schematic of the system used in both the PBR for the catalytic-
activity test and the CMR is shown in Fig. 2. The detailed characteris­
tics of the apparatus and operating conditions are presented in Table 1.
The system was automatically operated and recorded by an AIO/DIO
PLC interface and a control program (LabVIEW, National Instruments
Corp., USA).
The ammonia decomposition reaction was conducted with a
concentric double-pipe reactor configuration: the outside shell was a 3/
4′′ stainless-steel tube, and the inside was a 1/4′′ dummy Ta membrane
tube for the PBR experiments and a 1/4′′ Pd/Ta/Pd composite mem­
brane for the CMR experiments. It was confirmed that H2 permeation
through Ta tube was not detected without Pd coatings in a temperature
range tested. For a fair comparison between the PBR and CMR opera­
tions, the reactor geometries were kept the same (Fig. S1 shows the
membrane assembly). Ru(0.65 wt%)/La(10 mol%)-Al2O3 catalysts were
packed in a dummy Ta tube or a Pd/Ta/Pd composite membrane, and
the pressure drops between the inlet and outlet streams under various
reaction conditions were confirmed to be insignificant. The feed-side
flowrates were controlled by electrical mass flow controllers (MFCs,
F–201C, Bronkhorst, High-Tech), and an electrical backpressure
controller (EPC, P–702C, Bronkhorst, High-Tech) was used to maintain a
constant feed pressure under varying reactant flowrates. To supply the
sweep gas to the permeate side, an MFC for the N2 carrier and a high-
pressure liquid chromatography pump (Series II Legacy HPLC pump,
TELEDYNI SSI) for the steam carrier were utilized. To evaporate
deionized (DI) water and preheat the steam supplied to the permeate
side, a 1-m-long 1/16′′ stainless-steel tube was installed in the electric
furnace. The flowrate of the separated H2 through the Pd/Ta/Pd com­
posite membrane (permeated side) was measured by a mass flow meter
(MFM, F-201 CL, Bronkhorst, High-Tech, Netherlands) after the flow
passed the gas–liquid separator and silica adsorbents (Silica gel blue,
medium 5–10 mesh, Samchun Chemical Co.). The H2 permeation rate
was determined according to the total flowrate of the permeate side
measured via MFMs, excluding the N2 or steam sweep gas flowrate. All
the MFCs and MFMs used in this study were calibrated using film flow
meters (SF–1U/2U, Horiba, Japan).
In the experimental system, the temperature was controlled by an
electrical tube furnace, and three thermocouples were installed at the
feed inlet, the permeate, and the outer wall of a double pipe reactor to
ensure that the experimental temperature was in the desired range. The
pressures of both the feed (catalytic reaction bed) and permeate sides
were measured using a pressure transducer (PSAN, Autonics; range:
0–10 bar; accuracy within 0.1% of full scale). Pneumatic valves were
installed and automatically activated according to the operating con­
ditions and designated flow directions. Check valves and ball valves
were installed at the proper positions to prevent reverse flow and
physical impacts on the measuring instruments. To analyze the con­
centration of the unreacted NH3 from the reactor, retentate gases were
introduced to the NH3 analyzer (IR type (0–6000 ppm), Airwell +7,
KINSCO technology, Republic of Korea) with 7 L min− 1 of dry air as a
diluent.
2.3. Performance evaluation
The performance of dense metallic composite or alloy membrane
separation processes is generally evaluated with regard to the ability to
transport H2 through the membrane, e.g., the flux and permeability, and
Y. Park et al. Journal of Membrane Science 614 (2020) 118483

Fig. 2. Process flow diagram (PFD) of the PBR and CMR.

reactor is assembled, a minor leak may occur, depending on the fittings

Table 1
Characteristics of the CMR and operating conditions.
Membrane Materials
Membrane type
Ta tube outer diameter [mm]
Ta tube thickness [μm]
Ta tube Length/Effective
permeation length [mm]
Effective permeation area
[mm2]
Pd deposition time [min]
Pd layer thickness [μm]
Catalytic Materials
Catalyst
Particle pellet size [mm]
Packed catalyst amount [g]
Operating conditions
Pressure [barabsolute]
Temperature [K]
Feed flowrate [cm3 min− 1]
Sweep gas flowrate [cm3
min− 1]
or gaskets used to seal the membrane feed chamber from the permeate
chamber or the outside of the reactor. In our study, Swagelok ferrules
Configuration 1 (for Configuration 2 (for were adopted (with the optimized fitting torque between the membrane
higher NH3 feed) lower NH3 feed)
and the ferrules) to avoid failure due to H2 embrittlement while ensuring
leakage-free operation of the reactor system. Swagelok fittings were
Pd/Ta/Pd composite membrane (tubular)
tightened using a torque wrench under constant torque and then tested
6.35
250
for leakage, to find the minimum torque assuring leakage-free operation.
100/76.0 50/26.0 The results of impurity analyses of the as-produced H2 are presented in
Section 3.4. The NH3 conversion, H2 recovery yield, and H2 productivity
1396.0 477.6 were calculated via the following equations.
40 FRetentate, NH3 FFeed, NH3 (1 − x)
Approximately 1–2 CNH3 = = (2)
FRetentate FFeed, NH3 (1 + x) + FAir − FPermeate, H2
( )
Ru(0.65 wt%)/La(10 mol%)-Al2O3 FFeed, NH3 − CNH3 FFeed, NH3 +FAir − FPermeate, H2
NH3 conversion, 100x[%]= ( ) ×100
1–2 FFeed, NH3 1+CNH3
1.0 0.38
(3)
FPermeate, H2
1–5 H2 recovery yield[%] = × 100 (4)
673–773 1.5FFeed, NH3 x
20–100
0–100 [ ] 1.5FFeed, NH3 x
Productivity mol hr− 1 g− 1 = (5)
Catalyst (or metal) amount

where, CNH3 , FRetentate,NH3 , FRetentate , FFeed,NH3 , FAir , FPermeate,H2 , and x

by the quality (e.g., purity) of the produced H2. The H2 permeation rate,
represent the concentration of ammonia in the retentate gas measured
permeability, and H2 purity, along with a flow diagram including the
by the NH3 analyzer, the unreacted NH3 flowrate in the retentate, the
mass balance of the CMR used in this study, are presented in Fig. S1. The
total flowrate of the retentate gas, the NH3 feed flowrate, the diluent
permeability, which is defined as a combination of the diffusivity (D)
(air) flowrate, the H2 permeation rate, and the ammonia conversion
and solubility (S), was calculated, assuming the conventional solution-
rate, respectively. NH3 conversion was calculated based on the
diffusion mechanism in bulk Ta. The H2 flux, which is an important
measured NH3 concentration in the retentate stream (detailed deriva­
performance metric for the metallic membrane, was determined by the
tion shown in Appendix 1 in Supplementary Information).
H2 permeation rate divided by the effective permeation area shown
Permeated H2 and sweep gas were directly introduced to the com­
below
mercial PEMFC (H-12, Horizon Fuel Cell Technologies, Japan). The
Fpermeate, H2 concentration of NH3 on the permeate side was determined using an
H2 flux = (1)
Effective permeation area ultraviolet–visible (UV–vis) spectrophotometer (Cary 100, Agilent Tech,
USA) after the permeated gas was dissolved for 2 h in 10 mL of a 10 mM
The H2 produced from dense metallic membranes with a thickness
sulfuric acid solution with Nessler’s reagent.
greater than a certain level is usually ultrapure. However, when the

4
Y. Park et al.
The CMR was operated at a pressure up to 5 bar and in the tem­
perature range of 673–773 K, with a gas hourly space velocity (GHSV)
ranging from 1200 to 6000 cm3 g−cat1 h− 1. Prior to each operation, the Ru/
La–Al2O3 catalyst was reduced at 773 K under H2:N2 (50:50 cm3 min− 1)
conditions in reactors for >2 h to ensure removal of Cl ions which might
affect the catalytic activity. The experimental reproducibility of the
CMR operation was confirmed to be within 5% by performing experi­
ments in triplicate under the same conditions. Finally, to evaluate the
theoretical energy and mass balance of the developed CMR process,
Aspen Plus (Aspen Technology Inc., USA) was used as a process simu­
lation tool. For calculations of the thermodynamic properties, the Pen­
g–Robinson equation of state was applied [42].
3. Results and discussion
3.1. Characterization and performance analyses of composite membrane
and catalyst
The Pd/Ta/Pd composite membrane used in the membrane reactor
was fabricated via ELP, and the amount of Pd deposited on the Ta was
estimated according to the weight change after the ELP. After ELP on
100-mm-long tubular Ta, approximately 0.38 g of total weight increase
was observed, including various chemical compounds from the ELP
solutions. The actual thickness of the Pd layer and the surface mor­
phologies were determined via SEM, as shown in Fig. 3a. Approximately
1.3–2.3 μm of Pd was uniformly deposited on the both feed and
permeate surface of the Ta tube, with good adhesion after the heat
treatment. After repeated ELP, a repeatable thickness and weight gain
from the Pd was confirmed under the given plating conditions.
Accordingly, it was considered that a Pd/Ta/Pd composite membrane
having an average Pd thickness of approximately 1.5 μm was obtained.
XRD peaks of the bare Ta and as-fabricated Pd/Ta/Pd membrane
confirmed deposition of metallic Pd on Ta (Fig. S2). The differences in
the thicknesses of the plated Pd between the inside and outside of the Ta
tube were possibly due to concentration gradients formed in the plating
bath, requiring further optimization of the plating methods.
In this study, the H2 permeation of the fabricated Pd/Ta/Pd com­
posite membrane (Config. 1, Table 1) was tested using a H2(75%)/
N2(25%) binary gas mixture, simulating the gas compositions when
ammonia feed is fully converted into H2 and N2. The H2 permeation rate
through the membrane significantly depends on the difference in H2
partial pressures on the feed and permeate sides, as well as permeability
determining bulk diffusion of H atoms across the bulk Ta layer by
solution-diffusion mechanism [43] (detailed mechanism described in
Fig. S3). According to Sieverts’ law, the H2 permeation rate increases
linearly with a constant exponential factor of the pressure difference.
The numerical value of the exponential factor in Sieverts’ law depends
on the transport mechanism in the membrane; 0.5 for bulk diffusion and
Fig. 3. (a) SEM image of cross section and surface of the composite membrane with digital picture of the membrane assembly and (b) H2 permeation performance of
the fabricated Pd/Ta/Pd composite membrane under 200 cm3 min− 1 flow of a H2(75%)/N2(25%) binary gas mixture with a feed pressure of up to 5 bar at 773 K with
a sweep N2 flowrate of 100 cm3 min− 1.
5
Journal of Membrane Science 614 (2020) 118483
1.0 for mass transfer control. As the bulk diffusion dominant hydrogen
transport is widely accepted for Pd coated body-centered cubic (V, Nb,
Ta) composite membranes [43] and experimentally verified in previous
studies by our group [38,39], 0.5 for exponential factor was adopted in
this study as well.
The flux of the membrane and recovery yield reached 10.34 cm3
min− 1 cm− 2 and 96.3% at 5 bar of feed pressure, respectively, and both
of them significantly increased at a higher operating pressure (Fig. 3b).
They tended to converge above a feed pressure of 4 bar, as the feed
flowrate was fixed at 200 cm3 min− 1, but more H2 can permeate with
higher feed rates. H2 recovery yield did not increase further, owing to
mass transfer limit incurred by increasing partial pressure of N2 with
increasing H2 recovery yield above 95%. With pure H2, the calculated
permeability and permeance values were 6.03 × 10− 8 mol m− 1 s− 1
Pa− 0.5 and 3.05 × 10− 7 mol m− 2 s− 1 Pa− 1 at 773 K. The permeability was
much higher than that of the Pd–Ag membrane (1.6–2.4 × 10− 8 mol
m− 1 s− 1 Pa− 0.5) and comparable to those of other Pd/Ta/Pd membranes
reported in previous studies (6–20 × 10− 8 mol m− 1 s− 1 Pa− 0.5) at similar
temperature range [40,43]. Additionally, no embrittlement issues were
observed in the operating range tested, owing to the optimized torque
fitting around the membrane tube as described in the experimental
section and previous method by our group [39,40].
2NH3 → N2 + 3H2 (6)
Ammonia can decompose to hydrogen (Eq. (6)), and equilibrium
conversion of ammonia to hydrogen can be achieved to >98% at 623 K
[12]. However, a relevant catalyst is necessary to lower the activation
energy, promoting the dehydrogenation reaction from ammonia. While
Ru is known to be the most active metal for NH3 decomposition, catalyst
supports were also known to play an important role in promoting cat­
alytic activity. Previous studies indicated that La-doped Al2O3 catalysts
increase the electron density of Ru via strong metal to support interac­
tion and form Lewis acid sites on the catalyst surface [16,40]. Promoting
effects of alkali metals (e.g. La, Mg, K) on catalysts performance
improvement, in terms of activity and durability, have been previously
reported [44,45]. In this study, a Ru(0.65 wt%)/La(10 mol%)-Al2O3
pellet was synthesized using the wet impregnation method. Detailed
characterizations of the synthesized catalyst were reported in the liter­
ature [16]. Fig. 4 shows the catalytic activity of the synthesized pellet
catalyst for ammonia decomposition under different GHSVs (NH3
flowrate divided by the packed catalyst amount) and temperatures. The
activity of the catalyst increased with the temperature and decreased
with the increasing GHSV. The ammonia conversion was approximately
99% at >798 K and 1 bar under all the GHSV conditions. Based on the
results obtained, the kinetically limited region of the synthesized cata­
lysts was <773 K, where the membrane reactor can be beneficial for
increasing the NH3 conversion.
Y. Park et al. Journal of Membrane Science 614 (2020) 118483

Fig. 4. Catalytic activity of the ammonia decomposition reaction on Ru(0.65

wt%)/La(10 mol%)-Al2O3. The reaction was conducted in the temperature
range of 623–823 K (increments of 25 K) with a GHSV of 1200–3000 (in­
crements of 600 cm3 g−cat1 h− 1).

3.2. Performance of composite membrane reactor: optimal feed flowrate

and pressure

The performance of the CMR for ammonia decomposition was

evaluated with respect to the inlet ammonia flowrate (or GHSV), tem­
perature, and NH3 feed pressure. To increase the pressure difference
between the retentate and permeate sides, a 100 cm3 min− 1 N2 carrier
flow was introduced to permeate side in some cases. Critical perfor­
mance metrics, i.e., the H2 permeation rate and NH3 conversion of each
CMR operation (Config. 1, Table 1) at a pressure of 5 bar, are presented
in Fig. 5a and b. As shown in Fig. 5a, the H2 permeation rate increased
rapidly with the increasing temperature and GHSV, and the H2 perme­
ation rate gradually converged above a certain GHSV, except at 773 K.
The performance of the H2 permeation at 773 K had the potential to
increase with the GHSV. In binary H2/N2 separation, as shown in Fig. 3,
the composite membrane produced >144 cm3/min of H2 at a H2 partial
pressure of 3.75 bar on the retentate side. Therefore, when the GHSV
and partial pressure of H2 on the retentate side are increased at 773 K,
the H2 permeation rate of the CMR can reach a value similar to that for Fig. 5. (a) H2 permeation rate, (b) NH3 conversion, and (c) H2 recovery yield of
H2/N2 separation. the composite membrane reactor (Config. 1, Table 1) at temperatures ranging
As shown in Fig. 5b, the NH3 conversion gradually decreased with from 673 to 773 K (increments of 25 K) under a feed pressure of 5 bar with
GHSVs ranging from 1200 to 6000 (increments of 600 cm3 g−cat1 h− 1), corre­
the increasing GHSV, and the conversion was >99% even at a GHSV of
sponding to an ammonia flowrate of 20–100 cm3 min− 1, and with a sweep N2
4800 cm3 g−cat1 h− 1 at 773 K. Additionally, up to a GHSV of 3000 cm3 g−cat1
flowrate of 100 cm3 min− 1.
h− 1, the NH3 conversion was >99% at both 748 and 773 K. Furthermore,
>99% conversion was achieved at a GHSV of 1200 cm3 g−cat1 h− 1 at all the
of H2 was recovered at 773 K. However, the recovery yields rapidly
temperatures except for 673 K. Compared with previous results,
decreased with the increasing GHSV, despite increasing the H2 pro­
considering the even lower Ru content on the catalyst and the same
duction amount, as shown in Fig. 5a. Fig. S4b presents a comparison of
composite membranes, the CMR geometry with a shorter distance from
the H2 flowrates on the permeate and retentate sides according to the H2
the catalysts to the surface of the membrane used in the present study
production amount at 723 K. As the NH3 feed flowrate increased, both
enhanced the mass transfer of H2 to the membrane side and reduced the
the amount of H2 produced and the amount of H2 permeated increased.
concentration gradient on the catalytic side (shown in Fig. S1). CMR
However, the residence time in the catalytic bed decreased owing to the
operations of 673 K were not possible at a GHSV of >3600 cm3 g−cat1 h− 1,
high GHSV, causing a majority of the produced H2 to be discharged
owing to the measurement limits of the ammonia analyzer employed.
through the retentate stream rather than being separated. According to
These results suggest higher NH3 conversion at lower operating tem­
the foregoing results, the GHSV range of 2400–3000 cm3 g−cat1 h− 1 was the
peratures and higher GHSVs, indicating significant improvements
optimal condition in the developed CMR configuration.
compared to the previous study of our group [40] and other previous
Furthermore, compared to PBR results in Fig. 4, the H2 productivity
literatures. This was further confirmed by evaluating the partial pres­
was significantly improved with CMR at the identical temperature and
sures in the feed and permeate sides (Fig. S4a).
GHSV despite at higher pressures (Fig. 5a). H2 productivities of PBR and
Fig. 5c shows the H2 recovery yields of the CMR operations. For all
CMR are 40 and 70 mmol-H2 g−cat1 h− 1, respectively, at 663 K and at GHSV
the CMR operations, the H2 recovery yield reached its maximum value
of 3000 cm3 g−cat1 h− 1, presenting more than 63% improvements
and then decreased with the increasing GHSV. The GHSV range of
benefitting from the in situ H2 extraction. However, the improvement of
2400–3000 cm3 g−cat1 h− 1 exhibited the highest recovery values in all the
H2 productivity of CMR decreased to 7% at 773 K, implying that the
temperature ranges tested. All the recovery yields were >85%, and 91%

6
Y. Park et al.
effectiveness of CMR is more prominent at lower temperature. Then, H2
productivity was compared to results of previous studies including our
group’s. H2 productivity was measured to be 350 mmol-H2 g−cat1 h− 1 at
773 K under 5 bar. H2 productivity was under 200 mmol-H2 g−cat1 h− 1 for
literatures using CMR with Ru or Ni catalysts at optimum operating
conditions [27,28,46]. Considering specific Ru or Ni metal loading, H2
1
productivity values are 10,700 and 54,800 mmol-H2 g−catmetal h− 1 at 673
K and 773 K under 5 bar, respectively, about 30 times higher than that of
CMR using Ni catalyst and more than 2 times higher than that of CMR
using Ru catalyst.
Fig. 6a and b show the effect of the operating pressure on the per­
formance of CMR with respect to that of PBR. Because the GHSV of 3000
cm3 g−cat1 h− 1 was the optimal condition for CMR operation, the perfor­
mance was evaluated with respect to the operating pressure of the CMR
under this condition. As mentioned previously, a pressurized feed is
preferred for both H2 extraction and recovery yield improvement owing
to the increased driving force for H2 permeation. H2 recovery yield
increased with the increasing operating pressure, as shown in Fig. 6a,
and they tended to increase marginally above a feed pressure of 3 bar, as
explained by H2 flux trends shown in Fig. S5a. Furthermore, considering
vapor pressure of ammonia (6.2 bar at 283 K) charged for trans­
portation, the operation pressure of 5 bar for the CMR is deemed
appropriate. In the CMR operation at 723 K, the H2 recovery yield was
improved from 77.5% to 88.7% when the pressure increased from 1 to 5
bar, and under other temperature conditions, the recovery yields
increased by 8.1%–12.9% with the increasing pressure.
While a pressurized feed is beneficial for H2 production in a mem­
brane reactor, it negatively impacts the ammonia decomposition in the
catalytic reaction. Fig. 6b shows the NH3 conversion in the PBR as a
Fig. 6. (a) H2 recovery yield of CMR at temperatures ranging from 673 to 773 K
(increments of 25 K) and (b) NH3 conversion of the PBR and CMR (Config. 1,
Table 1) at temperatures ranging from 673 to 773 K (increments of 50 K) under
a feed pressure of 5 bar with a GHSV of 3000 cm3 g−cat1 h− 1 (corresponding to an
ammonia flowrate of 50 cm3 min− 1), and with a sweep N2 flowrate of 100 cm3
min− 1. A complete set of data comparing PBR and CMR at temperature from
673 to 773 K with increments of 25 K is shown in Figs. S5b–c.
7
Journal of Membrane Science 614 (2020) 118483
function of the pressure in the temperature range of 673–773 K and at
pressures up to 5 bar. The conversion decreased significantly with an
increase in the pressure. The conversion at 773 K and 5 bar was similar
to that at 748 K and 1 bar, and the other results also exhibited a decrease
in the conversion from 6.7% to 17.7% with the increasing pressure.
However, the reaction kinetics can be enhanced with in situ removal of
the H2 by reducing the average H2 concentration in the catalyst bed. The
performance of the CMR was compared with that of the PBR under the
identical conditions (Fig. 6b), and a conversion improvement was
apparent even under a comparatively high GHSV. It has to be noted that
PBR eventually needs to be operated at pressurized conditions with PSA.
With in situ H2 extraction using the membrane reactor, the conversion
increased with the pressure under the identical conditions used for the
PBR. Increasing the partial pressure of NH3 through in situ H2 extraction
enhanced the NH3 adsorption on the catalyst, improving the reaction
kinetics. This implies that an increase in the H2 permeation/extraction
from the CMR not only increased the H2 recovery but also improved the
NH3 conversion. At a reaction temperature of 773 K, the conversion was
>99% regardless of the pressure. The NH3 conversion was improved by
>36.7% compared with that of the PBR at temperature of 723 K and
pressure of 5 bar.
In addition to the operating GHSV, pressure, and temperature of the
CMR, the performance of the membrane reactor may vary according to
the effective area of the composite membrane and the amount of cata­
lysts packed in the membrane reactor. In Fig. 7a and b, the H2 flux and
NH3 conversion for two different CMR configurations (Config. 1 and
Config. 2, Table 1) are compared over different GHSVs. The two con­
figurations exhibited similar trends of the flux and conversion with the
increasing GHSV, but Config. 2 exhibited a significantly higher H2 flux
than Config. 1. Considering theoretical permeability of Ta membranes
Fig. 7. (a) H2 flux and (b) NH3 conversion for two different CMR configurations
(Config. 1 and Config. 2, Table 1) at temperatures ranging from 723 to 773 K
(increments of 25 K) under a feed pressure of 5 bar with GHSVs ranging from
1200 to 6000 (increments of 600 cm3 g−cat1 h− 1 for Config. 1 and increments of
1580 cm3 g−cat1 h− 1 for Config. 2), corresponding to an ammonia flowrate of
20–100 cm3 min− 1 with a sweep N2 flowrate of 100 cm3 min− 1.
Y. Park et al. Journal of Membrane Science 614 (2020) 118483

reported in the literature, the flux of the membrane did not reach its
maximum value, and there was room to increase the H2 flux if the partial
pressure of H2 was increased at a higher GHSV with higher catalyst
loading. The difference in the H2 flux between the two configurations
also affected the ammonia conversion: the higher NH3 conversion of
Config. 2 compared with that of Config. 1 was due to the higher H2 flux
for Config. 2 under the same GHSV, which resulted in faster catalytic
reaction kinetics and more effective removal of H2.
The performance of the two CMR configurations according to the
operating pressure was tested at a fixed NH3 flowrate (Fig. 8a). Despite
the approximately threefold difference in the effective permeation area
between the two configurations, the difference in the H2 permeation rate
was only approximately 80% at 773 K and 5 bar. The H2 permeation rate
of the CMR can be increased at higher GHSV, and the area of the
membrane was used more efficiently in Config. 2. The ammonia con­
version for Config. 2 was compared with that under the same GHSV
condition of the PBR in Fig. 8b, and the tendency was similar to the
comparison in Fig. 6b. However, in contrast to the case of Fig. 6b, where
the conversion increased with the pressure, the conversion remained
almost constant with the increasing pressure. As mentioned previously,
although removing the H2 can improve the reaction kinetics, in the high-
GHSV region, the ammonia conversion is not significantly increased,
owing to the low concentration of ammonia.
Fig. 9 shows a comparison of the H2 recovery yields for both the
tested CMR configurations. The H2 recovery yield with respect to the
GHSV exhibited similar trends for both configurations, but for Config. 2,
it decreased significantly after the GHSV reached 4700 cm3 g−cat1 h− 1. In
contrast, for Config. 2, the H2 recovery yield increased with the pressure,

Fig. 9. H2 recovery yield for two different CMR configurations (Config. 1 and
Config. 2) at temperatures ranging from 723 to 773 K (increments of 25 K) with
a sweep N2 flowrate of 100 cm3 min− 1 (a) under a feed pressure of 5 bar with
GHSVs ranging from 1200 to 6000 (increments of 600 cm3 g−cat1 h− 1 for Config. 1
and increments of 1580 cm3 g−cat1 h− 1 for Config. 2), corresponding to ammonia
flowrate of 20–100 cm3 min− 1, and (b) under feed pressures ranging from 1 to
5 bar at an NH3 feed rate of 50 cm3 min− 1 (GHSV: 3000 cm3 g−cat1 h− 1 for Config.
1 and 7900 cm3 g−cat1 h− 1 for Config. 2).

similar to the tendency of Fig. 8a. Increasing H2 permeation rate with

higher pressure strongly affected the H2 recovery yield, as shown in
Fig. 8a. The study on different CMR configurations clearly shows the
tradeoff relations between H2 production rate, H2 recovery, and NH3
conversion, as described above. These need to be considered in a large
scale system for process optimization, with reaction kinetic and mass
transfer simulations as conducted by previous studies [31,33,37].

3.3. Effects and benefits of steam as sweep gas in permeate stream

A sweep gas such as N2 or Ar on the permeate side facilitates the

Fig. 8. (a) H2 permeation rates for the two different CMR configurations
(Config. 1 and Config. 2) and (b) NH3 conversion for PBR and CMR (Config. 2)
at temperatures ranging from 723 to 773 K (increments of 25 K) under feed
pressures ranging from 1 to 5 bar with an ammonia feed rate of 50 cm3 min− 1
(GHSV: 3000 cm3 g−cat1 h− 1 for Config. 1 and 7900 cm3 g−cat1 h− 1 for Config. 2),
and with a sweep N2 flowrate of 100 cm3 min− 1.
permeation of H2 by increasing the driving force of the H2 transport
across the membrane, improving the H2 recovery yield as well as the
ammonia conversion. Figs. S6a–b shows the performance variation ac­
cording to the N2 sweep gas flowrate in CMR Config. 1 at a pressure of 5
bar and a GHSV of 3000 cm3 g−cat1 h− 1. H2 flux, ammonia conversion, and
H2 recovery yield at all temperatures significantly increased with the N2
sweep gas flowrates, and the increment of each performance metric was
marginal with a further increase in the sweep gas flowrates. Thus, the

8
Y. Park et al.
performance of the CMR can be improved through the use of N2 sweep
gas, but the sweep gas must eventually be separated via an additional
unit process, such as pressure swing adsorption, to obtain high-purity
H2.
To solve this problem, steam was applied as an alternative sweep gas,
and high-purity H2 was produced via relatively simple gas–liquid phase
separation. There are few studies which adopted steam as a sweep gas in
H2 extraction reactions [47,48], but its application to NH3 decomposi­
tion membrane reactor has not been previously reported. Fig. 10a shows
the performance variation of the CMR with different steam sweep gas
amounts at a pressure of 5 bar and GHSV of 6000 cm3 g−cat1 h− 1, compared
with the case where N2 sweep gas was used under the same conditions.
At a high GHSV, the performance of the CMR increased with the sweep
gas flowrates, similar to the trends presented in Fig. S6, and the per­
formance difference between the two sweep gases was insignificant.
Again, the benefits of using a sweep gas were apparent (increased H2
recovery yield, NH3 conversion, and H2 permeation rate). In particular,
when the sweep gas was used, the H2 recovery rate was improved by
approximately 5.5% compared with the case where the sweep gas was
not used. Additionally, experiments were conducted with the steam
sweep gas of 100 cm3 min− 1 at different feed pressures (Fig. 10b), and
the trends of the performance increase were similar to those in Fig. S6.
When the CMR was operated at a high GHSV (>5000 cm3 g−cat1 h− 1), as in
the case of Config. 2, the H2 recovery yield was highly dependent on the
operating pressure.
For a membrane reactor, leak-free scalability is important, and there
must be a way to cope with unexpected failures of the system. Leakage is
particularly critical for the NH3 decomposition membrane, as the pres­
ence of NH3 in H2 gas is detrimental to PEMFC operation, which is the
Fig. 10. H2 permeation rate, H2 recovery yield, and NH3 conversion (a) for
sweep gas flowrates ranging from 0 to 100 cm3 min− 1 with increments of 25
cm3 min− 1 at 773 K and 5 bar with a GHSV of 6000 cm3 g−cat1 h− 1 and (b) for feed
pressures ranging from 1 to 5 bar at 773 K with a GHSV of 6000 cm3 g−cat1 h− 1
with a sweep steam flowrate of 100 cm3 min− 1.
9
Journal of Membrane Science 614 (2020) 118483
major application of the produced H2. Using the Nessler method,
quantitative titration of NH3 was conducted along with UV–vis spec­
troscopy. In Scheme 1, it was confirmed that the produced H2 contained
approximately 4.2 ppm of ammonia with steam sweep gas. However, the
amount of ammonia in the permeated product decreased to approxi­
mately 1.8 ppm after the material passed through a gas–liquid separator
containing 300 cm3 of DI water with simple bubbling to mimic the effect
of a scrubber. Thus, a gas–liquid separator can act as a reservoir for
steam sweep gas recirculation and prevent the poisoning of the PEMFC
by removing trace amounts of ammonia from the product to ensure that
it is fuel-cell-grade. Additionally, CMR operation under prolonged time
of 65 h with steam sweep gas showed no significant decrease in the
performance (Scheme 1). For further practical applications of the sug­
gested system, long term durability test of the system would be useful, as
stability of the Pd/Ta/Pd membrane can be problematic under high
temperature operations [49–51]. Thermal stability of Pd-based mem­
branes is known to be affected by inter-diffusion [52] and surface
morphology change, so strategies for stability improvement needs to be
studied further by adopting diffusion barrier [53,54] or by alloying [55].
To generate power using the as-produced H2, a commercial
laboratory-scale PEMFC (12-W-class) was fed by the product H2 gas of
the CMR. As shown in Fig. 11a, the I–V curves were measured and
compared with those for pure H2 and H2 produced by the CMR with N2
or steam sweep gases. Even though the same amount of H2 was intro­
duced into PEMFC, significant differences in the I–V curves were
observed. When N2 was used as the sweep gas in the CMR, the voltage of
the PEMFC was approximately 2.5% lower than that when pure H2 was
used, owing to the lower H2 partial pressure on the anode side. However,
the voltage increased slightly when the H2 produced with steam sweep
gas was used. Because the H2 produced after passing through the
gas–liquid separator was humid, the PEMFC performance was increased.
The effects of anode humidification on proton conductivity of mem­
brane utilized in the current PEMFC are well established in previous
studies [56,57]. To validate the effect of humidification, pure H2 from
the gas cylinder was provided to PEMFC with and without bubbling as
schematic shown in Fig. 11b, and the performance increase with hu­
midification was similar as shown in Fig. 11a. Therefore, using steam as
the sweep gas is beneficial for onsite power-generation applications of
the composite membrane reactor.
3.4. Energy and mass balance evaluation of composite membrane reactor
The CMR process developed in this study requires the heat source to
replenish thermal energy owing to the endothermic reactions of NH3
decomposition and the latent heat of vaporization for the generation of
steam (as a sweep gas). Furthermore, utilities such as air and water are
necessary to complete the process. A process flow diagram (PFD) of the
CMR for ammonia decomposition and H2 production is presented in
Fig. 12 based on the experimental results at 773 K and 5 bar with GHSV
of 6000 cm3 g−cat1 h− 1 using and sweep steam flowrate of 100 cm3 min− 1.
To design the CMR, the model was divided into two units: a reformer for
ammonia decomposition and a membrane separator, and both of them
was assumed as steady-state, isothermal, and isobaric conditions. In the
PFD, the product gas stream (S1) of NH3 decomposition in a reformer is
linked to the membrane to purify the H2. The permeated H2 with steam
sweep gas from the membrane (S2) is introduced to gas-liquid (G-L)
separator after cooling down by a heat exchanger (S3). The retentate gas
after pass through the heat exchanger (S4) is mixed with air and supply
to the burner (S5).
The energy consumption and production of the process were evalu­
ated with respect to the process units. i) The heat loss due to the
endothermic reaction of the ammonia decomposition reaction was
referred to as Q1. ii) To approach the reaction temperature of the sweep
gas after the heat exchanged by retentate and permeate flows, an
additional heat exchanger to replenish the energy (Q2) was applied. iii)
The retentate gas after the heat exchanger was combusted with air to
Y. Park et al. Journal of Membrane Science 614 (2020) 118483

Scheme 1. NH3 concentration in the permeate gas before and after passing through the vapor–liquid separator with long term testing of the CMR at conditions listed
in the scheme. Cross-sectional morphology of the fresh and spent membrane, after >60 h of testing, from previous study of our group [40].

amounts of water were considered. The amount of H2 produced by the

CMR (M1) was divided by the splitter to be used for the power gener­
ation by the PEMFC (M2) or the heat source of the process (M3). The
water consumed for generating the steam sweep gas (M4) included the
water recovered from the gas–liquid separator (M5) and the byproduct
of the electrochemical reaction in PEMFC (M6).
Table S1 presents the values of the mass and heat stream according to
the foregoing descriptions. Analyzing the energy balance using the
simulation tool revealed that a total energy of 5.285 W was required for
ammonia decomposition and steam generation through the heat
exchanger. However, when the retentate gas was used as a fuel for the
combustor, approximately 3.2 W of heat energy could be replenished for
the process. To balance the total caloric value in the process, some of the
H2 produced by the gas–liquid separator was utilized as a heat source.
When the retentate gas and approximately 16.7% of the generated H2
were used as the heat source of the process, they covered the entire
amount of energy required for the process. Additionally, if the generated
H2 is used for power generation for the PEMFC, it can produce
approximately 6.7 W of electricity.
The PEMFC generates water as a byproduct, which can be utilized for
the sweep gas of the CMR. As shown in the stream information of
M4–M6 in Table S1, the amount of water produced in the PEMFC is
similar to the amount required by the CMR, and most of the water can be
recovered through the gas–liquid separator. This is another benefit of
utilizing steam as a sweep gas in the proposed process. The present
simulation was performed to evaluate the feasibility of the process by
estimating the energy and mass balance; further process simulations and
experimental studies are needed to estimate the system efficiency at the
full scale.
Fig. 11. (a) I–V curve of the commercial PEMFC operated using pure H2 (68
cm3 min− 1) from a gas cylinder, as-produced H2 (68 cm3 min− 1) from the CMR
4. Conclusions
operated at 773 K, 5 bar, GHSV of 3000 cm3 g−cat1 h− 1 with N2 and steam sweep
gas (b) Normalized I–V curve with respect to open circuit voltage (OCV) using
the dry H2 and humidified H2 streams.
A Pd/BCC composite membrane was fabricated to be adopted for an
NH3 decomposition CMR. The as-fabricated membrane had a perme­
ability of 6.03 × 10− 8 mol m− 1 s− 1 Pa− 0.5, with a H2 flux of 10.3 cm3
supply the heat energy (Q3) required in the process. iv) The H2 produced
min− 1 cm− 2. These values were comparable to previously reported
by the process was partially fed to the combustor to replenish the energy
values for dense metallic membranes using BCC metals, and the mem­
(Q4). v) To generate electrical power, the produced H2 was used as a
brane was operated with high stability and without embrittlement is­
source for the conjugated PEMFC (Q5). For the material balance, the
sues. Then, Ru catalysts were synthesized, and their catalytic activity
amount of H2 produced by the CMR and the consumed/recovered
was evaluated at different temperatures, pressures, and GHSVs. Using

10
Y. Park et al.
Fig. 12. PFD of the composite membrane reactor for ammonia decomposition and H2 production.
the as-fabricated membranes and catalysts, different configurations of
membrane reactors were developed and operated in a region where
further conversion increase could not be achieved solely by changing the
operating conditions of a PBR.
Comprehensive experiments were performed at temperatures
ranging from 673 to 773 K, under feed pressures ranging from 1 to 5 bar,
and with GHSVs ranging from 1200 to 6000 cm3 g−cat1 h− 1. For the as-
developed CMR, the NH3 conversion was >99% with a GHSV of 4800
cm3 g−cat1 h− 1 at 773 K, and >99% conversion was maintained up to a
GHSV of 3000 cm3 g−cat1 h− 1 at temperatures as low as 748 K. H2 pro­
ductivity improved more than 63% with CMR, with respect to PBR,
benefitting from the in situ H2 extraction. Calculated H2 productivity per
catalyst loading indicated more than 2 times higher values than that of
previous studies of CMR using Ru catalyst, showing a significant
improvement in the H2 productivity compared with previous results. A
H2 recovery yield of >85% was observed over a wide range of GHSVs,
and the maximum H2 recovery yield of 91% was achieved at 773 K. With
an increase in the feed pressure from 1 to 5 bar at 723 K, the H2 recovery
yield improved from 77.5% to 88.7%, accompanied by a significant
increase in the H2 flux.
While the NH3 conversion decreased with increasing pressure in the
case of the PBR, it increased with pressure under the same operating
conditions for the CMR. A significant increase in the NH3 conversion was
achieved under pressurized conditions (even when compared to the NH3
conversion using the PBR at the atmospheric pressure) at all the tem­
perature ranges tested. Different CMR configurations with different
membrane lengths and catalyst loadings were further compared with
regard to the GHSV, pressure, temperature, and sweep gas effects. The
two configurations exhibited similar performance trends, but the
configuration with a higher H2 flux resulted in higher NH3 conversion.
Considering that the membrane flux can be increased further, the H2
productivity can be improved with better catalyst design or higher
catalyst loading per unit membrane area achieved by changing the ge­
ometry of the membrane reactor.
Then, steam was adopted as a sweep gas, yielding comparable per­
formance to cases using conventional sweep gases such as N2. However,
use of steam was beneficial for the following reasons. (i) The steam
recirculation reservoir can act as an NH3 scrubber, reducing the NH3
content in the H2 stream. (ii) The reservoir can act as an NH3 buffer with
an in situ concentration measurement device in the case of membrane
system failure. (iii) Wet H2 can be produced, eliminating the need for a
humidifier, and the DI water reservoir can be replenished using water
11
Journal of Membrane Science 614 (2020) 118483
produced by the PEMFC. Finally, a process simulation was conducted,
confirming that waste energy from the integrated process can provide
the energy required to generate the steam and heat needed for endo­
thermic NH3 decomposition.
A higher H2 productivity, a lower operating temperature, faster ki­
netics, and improved H2 purification for NH3 decomposition are
demonstrated based on the developed CMR system in this study. How­
ever, it has to be noted that the benefits could be marginal in some
operating conditions; the adoption of CMR concept is meaningful in
specific operating conditions. Therefore, further analyses would be
required to assess potential benefits of CMR in terms of process inten­
sification at higher scales. The proposed system utilizing steam sweep
gas can widely be applied to the other membrane reactor configurations
with different membrane materials as well as to hydrogen extraction
from other liquid hydrogen carriers including methanol, ethanol, formic
acid, liquid organic hydrogen carriers, and so forth.
Credit author statement
YP: Conceptualization, Investigation, Writing – Original Draft,
Formal analysis.
JC: Conceptualization, Investigation.
HO: Investigation.
TL: Methodology.
SL: Resources.
MP: Resources.
HJ: Data curation.
YK: Visualization.
HS: Validation.
SN: Project administration.
JH: Resources.
CY: Funding Acquisition, Project administration, Writing – Review­
ing & Editing.
YJ: Conceptualization, Supervision, Writing – Original Draft, Writing
– Reviewing & Editing.
Declaration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Y. Park et al.
Acknowledgements
This work was supported financially by a National Research Foun­
dation (NRF) grant funded by the Korean Government (Ministry of
Science, ICT & Future Planning) [grant number NRF-
2019M3E6A1064611], by a Korea Institute of Energy Technology
Evaluation and Planning (KETEP) grant funded by the Korean Govern­
ment (Ministry of Trade, Industry & Energy) [grant number
20183010042020], and by the KIST Institutional Program of the Korea
Institute of Science and Technology [grant numbers 2E30520,
2E30202].
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.memsci.2020.118483.
References
[1] A. Klerke, C.H. Christensen, J.K. Nørskov, T. Vegge, Ammonia for hydrogen
storage: challenges and opportunities, J. Mater. Chem. 18 (2008).
[2] T. He, P. Pachfule, H. Wu, Q. Xu, P. Chen, Hydrogen carriers, Nat. Rev. Mater. 1
(2016).
[3] R.F. Service, Liquid sunshine, Science 361 (2018) 120–123.
[4] F. Zhou, L.M. Azofra, M. Ali, M. Kar, A.N. Simonov, C. McDonnell-Worth, C. Sun,
X. Zhang, D.R. MacFarlane, Electro-synthesis of ammonia from nitrogen at ambient
temperature and pressure in ionic liquids, Energy Environ. Sci. 10 (2017)
2516–2520.
[5] T.E. Bell, L. Torrente-Murciano, H2 production via ammonia decomposition using
non-noble metal catalysts: a review, Top. Catal. 59 (2016) 1438–1457.
[6] A.M. Karim, V. Prasad, G. Mpourmpakis, W.W. Lonergan, A.I. Frenkel, J.G. Chen,
D.G. Vlachos, Correlating particle size and shape of supported Ru/gamma-Al2O3
catalysts with NH3 decomposition activity, J. Am. Chem. Soc. 131 (2009)
12230–12239.
[7] Y. Li, L. Yao, Y. Song, S. Liu, J. Zhao, W. Ji, C.T. Au, Core-shell structured
microcapsular-like Ru@SiO2 reactor for efficient generation of CO(x)-free
hydrogen through ammonia decomposition, Chem. Commun. (Camb) 46 (2010)
5298–5300.
[8] X. Li, W. Ji, J. Zhao, S. Wang, C. Au, Ammonia decomposition over Ru and Ni
catalysts supported on fumed SiO2, MCM-41, and SBA-15, J. Catal. 236 (2005)
181–189.
[9] L. Li, Z.H. Zhu, G.Q. Lu, Z.F. Yan, S.Z. Qiao, Catalytic ammonia decomposition over
CMK-3 supported Ru catalysts: effects of surface treatments of supports, Carbon 45
(2007) 11–20.
[10] X. Ju, L. Liu, P. Yu, J. Guo, X. Zhang, T. He, G. Wu, P. Chen, Mesoporous Ru/MgO
prepared by a deposition-precipitation method as highly active catalyst for
producing COx-free hydrogen from ammonia decomposition, Appl. Catal. B
Environ. 211 (2017) 167–175.
[11] S.F. Yin, B.Q. Xu, W.X. Zhu, C.F. Ng, X.P. Zhou, C.T. Au, Carbon nanotubes-
supported Ru catalyst for the generation of COx-free hydrogen from ammonia,
Catal. Today 93–95 (2004) 27–38.
[12] S.F. Yin, B.Q. Xu, X.P. Zhou, C.T. Au, A mini-review on ammonia decomposition
catalysts for on-site generation of hydrogen for fuel cell applications, Appl. Catal.
Gen. 277 (2004) 1–9.
[13] S.-F. Yin, Q.-H. Zhang, B.-Q. Xu, W.-X. Zhu, C.-F. Ng, C.-T. Au, Investigation on the
catalysis of COx-free hydrogen generation from ammonia, J. Catal. 224 (2004)
384–396.
[14] P. Xie, Y. Yao, Z. Huang, Z. Liu, J. Zhang, T. Li, G. Wang, R. Shahbazian-Yassar,
L. Hu, C. Wang, Highly efficient decomposition of ammonia using high-entropy
alloy catalysts, Nat. Commun. 10 (2019) 4011.
[15] L. Huo, B. Liu, H. Li, B. Cao, X.-c. Hu, X.-p. Fu, C. Jia, J. Zhang, Component synergy
and armor protection induced superior catalytic activity and stability of ultrathin
Co-Fe spinel nanosheets confined in mesoporous silica shells for ammonia
decomposition reaction, Appl. Catal. B Environ. 253 (2019) 121–130.
[16] J. Cha, Y.S. Jo, H. Jeong, J. Han, S.W. Nam, K.H. Song, C.W. Yoon, Ammonia as an
efficient CO X -free hydrogen carrier: fundamentals and feasibility analyses for fuel
cell applications, Appl. Energy 224 (2018) 194–204.
[17] F.A. Uribe, S. Gottesfeld, T.A. Zawodzinski, Effect of ammonia as potential fuel
impurity on proton exchange membrane fuel cell performance, J. Electrochem.
Soc. 149 (2002).
[18] Y.A. Gomez, A. Oyarce, G. Lindbergh, C. Lagergren, Ammonia contamination of a
proton exchange membrane fuel cell, J. Electrochem. Soc. 165 (2018) F189–F197.
[19] H. Miyaoka, H. Miyaoka, T. Ichikawa, T. Ichikawa, Y. Kojima, Highly purified
hydrogen production from ammonia for PEM fuel cell, Int. J. Hydrogen Energy 43
(2018) 14486–14492.
[20] K.E. Lamb, D.M. Viano, M.J. Langley, S.S. Hla, M.D. Dolan, High-purity H2
produced from NH3 via a ruthenium-based decomposition catalyst and vanadium-
based membrane, Ind. Eng. Chem. Res. 57 (2018) 7811–7816.
12
Journal of Membrane Science 614 (2020) 118483
[21] G. Bernardo, T. Araújo, T. da Silva Lopes, J. Sousa, A. Mendes, Recent advances in
membrane technologies for hydrogen purification, Int. J. Hydrogen Energy 45 (12)
(2019) 7313–7338, 4.
[22] M. Temkin, V. Pyzhev, Kinetics of ammonia synthesis on promoted iron catalysts,
Acta Physicochim. URSS 12 (1940) 217–222.
[23] J.P. Collins, J.D. Way, Catalytic decomposition of ammonia in a membrane reactor,
J. Membr. Sci. 96 (1994) 259–274.
[24] E.N. Gobina, J.S. Oklany, R. Hughes, Elimination of ammonia from coal
gasification streams by using a catalytic membrane reactor, Ind. Eng. Chem. Res.
34 (1995) 3777–3783.
[25] M.E.E. Abashar, Y.S. Al-Sughair, I.S. Al-Mutaz, Investigation of low temperature
decomposition of ammonia using spatially patterned catalytic membrane reactors,
Appl. Catal. Gen. 236 (2002) 35–53.
[26] J. Zhang, H. Xu, W. Li, High-purity COx-free H2 generation from NH3 via the ultra
permeable and highly selective Pd membranes, J. Membr. Sci. 277 (2006) 85–93.
[27] F.R. García-García, Y.H. Ma, I. Rodríguez-Ramos, A. Guerrero-Ruiz, High purity
hydrogen production by low temperature catalytic ammonia decomposition in a
multifunctional membrane reactor, Catal. Commun. 9 (2008) 482–486.
[28] S.H. Israni, B.K.R. Nair, M.P. Harold, Hydrogen generation and purification in a
composite Pd hollow fiber membrane reactor: experiments and modeling, Catal.
Today 139 (2009) 299–311.
[29] G. Li, M. Kanezashi, T. Tsuru, Highly enhanced ammonia decomposition in a
bimodal catalytic membrane reactor for COx-free hydrogen production, Catal.
Commun. 15 (2011) 60–63.
[30] G. Li, M. Kanezashi, H.R. Lee, M. Maeda, T. Yoshioka, T. Tsuru, Preparation of a
novel bimodal catalytic membrane reactor and its application to ammonia
decomposition for COx-free hydrogen production, Int. J. Hydrogen Energy 37
(2012) 12105–12113.
[31] N. Itoh, A. Oshima, E. Suga, T. Sato, Kinetic enhancement of ammonia
decomposition as a chemical hydrogen carrier in palladium membrane reactor,
Catal. Today 236 (2014) 70–76.
[32] E. Rizzuto, P. Palange, Z. Del Prete, Characterization of an ammonia decomposition
process by means of a multifunctional catalytic membrane reactor, Int. J. Hydrogen
Energy 39 (2014) 11403–11410.
[33] Z. Zhang, S. Liguori, T.F. Fuerst, J.D. Way, C.A. Wolden, Efficient ammonia
decomposition in a catalytic membrane reactor to enable hydrogen storage and
utilization, ACS Sustain. Chem. Eng. 7 (2019) 5975–5985.
[34] M.R. Rahimpour, A. Asgari, Production of hydrogen from purge gases of ammonia
plants in a catalytic hydrogen-permselective membrane reactor, Int. J. Hydrogen
Energy 34 (2009) 5795–5802.
[35] M.Á. Gómez-García, I. Dobrosz-Gómez, J. Fontalvo, J.M. Rynkowski, Membrane
reactor design guidelines for ammonia decomposition, Catal. Today 191 (2012)
165–168.
[36] M.E.E. Abashar, Ultra-clean hydrogen production by ammonia decomposition,
J. King Saud Univ.- Eng. Sci. (2016).
[37] M.E.E. Abashar, The impact of ammonia feed distribution on the performance of a
fixed bed membrane reactor for ammonia decomposition to ultra-pure hydrogen,
Int. J. Hydrogen Energy 44 (2019) 82–90.
[38] Y.S. Jo, C.H. Lee, S.Y. Kong, K.-Y. Lee, C.W. Yoon, S.W. Nam, J. Han,
Characterization of a Pd/Ta composite membrane and its application to a large
scale high-purity hydrogen separation from mixed gas, Separ. Purif. Technol. 200
(2018) 221–229.
[39] C.H. Lee, Y.S. Jo, Y. Park, H. Jeong, Y. Kim, H. Sohn, C.W. Yoon, S.W. Nam, H.
C. Ham, J. Han, Unconventional hydrogen permeation behavior of Pd/BCC
composite membranes and significance of surface reaction kinetics, J. Membr. Sci.
595 (2020).
[40] Y.S. Jo, J. Cha, C.H. Lee, H. Jeong, C.W. Yoon, S.W. Nam, J. Han, A viable
membrane reactor option for sustainable hydrogen production from ammonia,
J. Power Sources 400 (2018) 518–526.
[41] P.P. Mardilovich, Y. She, Y.H. Ma, M.-H. Rei, Defect-free palladium membranes on
porous stainless-steel support, AIChE J. 44 (1998) 310–322.
[42] B. Lee, J. Park, H. Lee, M. Byun, C.W. Yoon, H. Lim, Assessment of the economic
potential: CO -free hydrogen production from renewables via ammonia
decomposition for small-sized H2 refueling stations, Renew. Sustain. Energy Rev.
113 (2019).
[43] M.D. Dolan, Non-Pd BCC alloy membranes for industrial hydrogen separation,
J. Membr. Sci. 362 (2010) 12–28.
[44] J. Ni, R. Wang, F. Kong, T. Zhang, J. Lin, b. Lin, K. Wei, Highly efficient Ru-Ba/AC
catalyst promoted by magnesium for ammonia synthesis, Chin. J. Catal. 32 (2011)
436–439.
[45] C.F. Huo, B.S. Wu, P. Gao, Y. Yang, Y.W. Li, H. Jiao, The mechanism of potassium
promoter: enhancing the stability of active surfaces, Angew Chem. Int. Ed. Engl. 50
(2011) 7403–7406.
[46] J. Liu, X. Ju, C. Tang, L. Liu, H. Li, P. Chen, High performance stainless-steel
supported Pd membranes with a finger-like and gap structure and its application in
NH3 decomposition membrane reactor, Chem. Eng. J. 388 (2020).
[47] H. Wang, P. Kölsch, T. Schiestel, C. Tablet, S. Werth, J. Caro, Production of high-
purity oxygen by perovskite hollow fiber membranes swept with steam, J. Membr.
Sci. 284 (2006) 5–8.
[48] Y. Zhang, Q. Sun, X. Gu, Pure H2production through hollow fiber hydrogen-
selective MFI zeolite membranes using steam as sweep gas, AIChE J. 61 (2015)
3459–3469.
[49] R.E. Buxbaum, T.L. Marker, Hydrogen transport through non-porous membranes of
palladium-coated niobium, tantalum and vanadium, J. Membr. Sci. 85 (1993)
29–38.
Y. Park et al.
[50] K.S. Rothenberger, B.H. Howard, R.P. Killmeyer, A.V. Cugini, R.M. Enick,
F. Bustamante, M.V. Ciocco, B.D. Morreale, R.E. Buxbaum, Evaluation of tantalum-
based materials for hydrogen separation at elevated temperatures and pressures☆,
J. Membr. Sci. 218 (2003) 19–37.
[51] V.N. Alimov, A.O. Busnyuk, M.E. Notkin, E.Y. Peredistov, A.I. Livshits, Hydrogen
transport through V–Pd alloy membranes: hydrogen solution, permeation and
diffusion, J. Membr. Sci. 481 (2015) 54–62.
[52] D.J. Edlund, J. McCarthy, The relationship between intermetallic diffusion and flux
decline in composite-metal membranes: implications for achieving long membrane
lifetime, J. Membr. Sci. 107 (1995) 147–153.
[53] Y. Huang, R. Dittmeyer, Preparation and characterization of composite palladium
membranes on sinter-metal supports with a ceramic barrier against intermetallic
diffusion, J. Membr. Sci. 282 (2006) 296–310.
13
Journal of Membrane Science 614 (2020) 118483
[54] Y. Huang, R. Dittmeyer, Preparation of thin palladium membranes on a porous
support with rough surface, J. Membr. Sci. 302 (2007) 160–170.
[55] E. Yan, R.N. Min, P. Zhao, R.D.K. Misra, P.R. Huang, Y.J. Zou, H.L. Chu, H.
Z. Zhang, F. Xu, L.X. Sun, Design of Nb-based multi-phase alloy membranes for
high hydrogen permeability and suppressed hydrogen embrittlement, J. Membr.
Sci. 595 (2020).
[56] H. Sun, G. Zhang, L.-J. Guo, S. Dehua, H. Liu, Effects of humidification
temperatures on local current characteristics in a PEM fuel cell, J. Power Sources
168 (2007) 400–407.
[57] S. Kim, I. Hong, Effects of humidity and temperature on a proton exchange
membrane fuel cell (PEMFC) stack, J. Ind. Eng. Chem. 14 (2008) 357–364.