i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/he

Pressurized PEM water electrolysis: Efficiency

and gas crossover

Maximilian Schalenbach a,*, Marcelo Carmo a, David L. Fritz a,

Jürgen Mergel a, Detlef Stolten a,b
a
Forschungszentrum Jülich GmbH, Institute of Energy and Climate Research,
IEK-3: Electrochemical Process Engineering, 52425, Germany
b
RWTH Aachen University, Germany

article info abstract

Article history: In this study the influence of cathodic and anodic pressures during polymer electrolyte
Received 27 June 2013 membrane water electrolysis on the gas crossover is simulated and compared to in-situ
Received in revised form measurements of the anodic hydrogen content at differential and balanced pressure
31 August 2013 operation. The efficiency losses due to the reduced Faraday efficiency caused by crossover,
Accepted 4 September 2013 ohmic loss of the membrane and pressurized hydrogen and oxygen evolution are esti-
Available online 10 October 2013 mated. Therefore, the correlated dependencies on the current density, membrane thick-
ness, anodic and cathodic pressures, membrane conductivity and permeabilities are
Keywords: quantified. In addition, pressurized electrolysis is compared to adiabatic and isothermal
Hydrogen production subsequent compression in focus of efficiency. The outcome of this study can be utilized as
Water electrolysis a powerful computational tool to optimize the membrane thickness with respect to the
High pressure PEM electrolysis operating pressures.
Efficiency modeling Copyright ª 2013, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights
Nafion permeability reserved.

1. Introduction water electrolysis correlated to the interplay of membrane

A clean and emission free technique to produce hydrogen by
renewables can be provided by means of water electrolysis
[1e5]. Compared to alkaline water electrolysis, polymer elec-
trolyte membrane (PEM) water electrolysis is a more promising
technology due to its higher efficiency, gas purity, current den-
sity and dynamic power range [1e3,6e14]. Furthermore, the
produced hydrogen can be pressurized during electrolysis due to
the low gas permeability and high mechanical resistance of the
membrane [1,9,15e18]. However, only a few studies concerning
the influence of the pressure on gas purities and efficiency
during PEM water electrolysis have been published [19e22].
Therefore, in this study we estimate the efficiency loss for PEM
thickness, gas crossover, current density and anodic/cathodic
pressure. Furthermore, a discussion comparing pressurized
electrolysis at 30 bar and subsequent hydrogen compression is
provided under efficiency considerations.
1.1. Thermodynamics of PEM water electrolysis
The entire chemical energy content of water is described by its
enthalpy
dH ¼ dG þ TdS ; (1)
which is the sum of Gibb’s free energy (dG) and heat
energy (TdS). Thermodynamically, for the electrochemical

* Corresponding author. Tel.: þ49 2461 619802; fax: þ49 2461 618161.
E-mail address: m.schalenbach@fz-juelich.de (M. Schalenbach).
0360-3199/$ e see front matter Copyright ª 2013, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijhydene.2013.09.013
14922 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3

Nomenclature s Conductivity, S cm1

x Drag coefficient, dimensionless
A Partial pressure enhancement factor, bar cm2 A1
a Activity, dimensionless Superscripts and subscripts
c Concentration, mol l1 0 Standard conditions
D Diffusion coefficient, cm2 s1 4st Four stage adiabatic compression process
d Membrane thickness, mm ad adiabatic
dG Gibb’s free energy, kJ mol1 an Anode
dH Enthalpy, kJ mol1 cat Cathode
dQ Heat, kJ mol1 cell Cell
dS Entropy, kJ mol1 K1 comp Compressor
E Potential, V dif Diffusion
F Faraday constant, 96485.3 C mol1 dp Differential pressure
j Current density, A cm2 drag Electro-osmotic drag
n Amount of substance, mol en Entire
p Pressure, bar f Final
R Gas constant, 8.314 J mol1 K1 F Faraday
Rarea Area resistance, U cm2 i Initial
S Solubility coefficient, mol bar1 cm3 id Ideal
T Temperature, K iso Isothermal
UU Ohmic drop at the membrane, V j Sum index
UN Nernst Voltage, V pro Production
V Volume, m3 re Real
W Work, kJ mol1 rev Reversible
U Voltage
Greek symbols
g Adiabatic index, dimensionless Abbreviations
G Production rate density, mol s1 cm2 IrO2 Iridium oxide
ε Permeability, mol s1 cm1 bar1 PEM Polymer electrolyte membrane
h Efficiency, dimensionless Pt/C Carbon supported platinum
F Flux density, mol s1 cm2 SHE Standard hydrogen electrode
r Mole density, mol m3

decomposition of water, the heat energy and the voltage cor-


1 J J
responding to Gibb’s free energy, denoted as reversible voltage Erev ¼  159:6 T þ 2:847  105 (3)
2F K mol mol
dG
Erev ¼ ; (2) To accelerate the reaction kinetics at the catalysts’ in-
2F
terfaces and therefore reduce overvoltages related to activa-
have to be applied, where F ¼ 96485.3 C/mol denotes the
tion losses, one usually aims for higher operation
Faraday constant. At standard conditions the reversible
temperatures, which consequently increase degradation is-
voltage equals 1.23 V. The voltage correlated to the enthalpy
sues [25,26]. From a thermodynamic point of view, electrolysis
(assuming that no energy is provided in the form of heat) is
at high temperature is favorable since the heat coming from
referred to the thermoneutral voltage of 1.48 V at standard
overvoltages can be used to reduce the reversible voltage
conditions. With applied voltages above the reversible voltage
(eq (3)). Usually temperatures around 80  C are considered to
but below the thermoneutral voltage, the heat of the system
be the best compromise between a reasonable lifetime and
has to be consumed to achieve electrolysis, in order to balance
good performance [13]. Therefore, in the present study, the
the required enthalpy. In this case additional heat input into
operating temperature is assumed to be 80  C.
the system is necessary, for example, as utilized by solid oxide
The thermodynamic potentials for the half reactions at the
electrolysis at higher temperatures [23]. When more voltage is
anode,
supplied than required to reach the enthalpy, the excess en-
ergy (the overvoltage) is converted from electricity to heat. 1
H2 O/ O2 þ 2Hþ þ 2e (4)
The heat produced from the overvoltage must be removed in 2
order to prevent overheating. Overvoltages directly reduce the
and the cathode,
efficiency and are caused by activation losses (slow reaction
kinetics at the electrodes), ohmic losses (resistivities related to 2Hþ þ 2e /H2 (5)
proton and electron transport) and mass transport losses.
The reversible voltage for the electrochemical decomposi- are temperature and pressure dependent as described by their
tion of liquid water is approximately linear in temperature [24]. respective Nernst equations,
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3 14923

2 pffiffiffiffiffiffiffiffiffiffiffiffiffi
! electrolysis [11]. Despite their poor chemical stability, hydro-
RT aðHþ Þ aðO2 Þ
Ean
¼ E0 þ ln (6) lysis issues (polymer dissolution in water) [11,33e37] and
2F aðH2 OÞ
about a tenfold reduction in specific proton conductivity
! compared to PFSA membranes [28], they are interesting for
2
RT aðHþ Þ high pressure electrolysis operation due to their reduced gas
Ecat ¼ Ecat
0 þ ln (7)
2F aðH2 Þ permeability [25,32].
Usually the membrane is laminated in between the elec-
where a denotes the activities of the involved substances. The
trodes (the catalyst layers) [10], building a construction that is
0 Þ is
standard potential of the half reaction at the cathode ðEcat
commonly denoted as catalyst coated membrane (CCM).
identified with the potential of the standard hydrogen elec-
Platinum supported on carbon (Pt/C) for the cathode and
trode (SHE) of 0 V. At the anode, the standard potential (E0)
iridium oxide for the anode are the state-of-the-art catalysts
corresponds to the reversible voltage for the decomposition of
[1,3,6]. This work is focused on extruded Nafion (DuPont)
water vs SHE at standard conditions.
with an equivalent weight of 1100 as the PEM [38], which is
Since hydrogen and oxygen have low solubilities in water
currently the most investigated and commonly used PFSA
[27], their activities can be approximated by their concentra-
membrane [1,27].
tions (c) normalized to the reference standard concentration
As anodic current collectors, most commonly sintered ti-
(c ¼ 1 mol/l).
tanium bodies are used [1]. In the case of differential pressure
a ¼ c=c0 (8) electrolysis, where the cathodic pressure is larger than the
anodic pressure, the CCM is forced onto the anodic current
The concentrations of the diluted gases are proportional to collector. Due to the stiffness of the sintered titanium body
their partial pressures (p) as described by Henry’s law, and small pore sizes (520 mm [39]), the CCM can withstand
high differential pressures, as examined in the literature [18].
c¼pS (9)

with respect to a negligible pressure dependence of the solu- 1.3. Safety issues due to gas crossover
bility S for pressures up to 100 bar [27].
Due to the presence of liquid water at the electrodes, the For pressurized PEM water electrolysis, safety issues due to
water activity can be approximated as unity. The thermody- gas crossover have been thoroughly investigated in the liter-
namic voltage required for the electrochemical decomposition ature [17,20,40]. Extensive mixing of the product gases due to
of water is described by the Nernst equation for acid gas crossover must be avoided, especially at low current
electrolysis, densities, where hydrogen and oxygen production rates are
0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 reduced. The hydrogen concentration at the anode side can
RT @pH2 SH2 pO2 SO2 A especially achieve dangerous levels (lower explosion limit
UN ¼ E an
E cat
¼ E0 þ ln (10)
2F c03=2 4 mol% H2 in O2 [41]). Consequently, the usual safety limit is
regarded to be 50% of the lower explosion limit, instilling a
derived from the half reactions at anode (eq. (6)) and cathode safety factor of 2.
(eq. (7)). Hence, electrolysis at a cathodic hydrogen pressure of To circumvent hazardous anodic gas mixtures, platinized
y
pH2 compared to a lower hydrogen pressure pHx2 at equal oxy- current collectors can be inserted, where permeated hydrogen
gen pressure is accompanied by a thermodynamic voltage can be catalytically reacted with oxygen or eventually elec-
enhancement, owed to the pressurized evolution of hydrogen: trochemical oxidized to protons, which could permeate
! through the membrane and be reduced at the cathode [20,40].
    RT y
pH2
y
DU ¼ UN pH2  UN pHx2 ¼ ln (11) This approach has not been studied yet under degradation
2F pHx2
issues with respect to platinum dissolution [34,36,42,43] at
anodic potentials (>1.23 V). Employing external catalytic gas
1.2. Characteristics of PEMs recombiners to reduce the gas impurity has also been
explored [20,40]. Although these approaches might handle
PEMs are strong acids that provide a large proton charge car- safety issues, gas crossover still lowers the Faraday efficiency
rier density, which crucially determines the proton conduc- (even if hydrogen oxidation at the anode takes place)
tivity [26,28]. Most commonly perfluorosulfonic acid (PFSA)
Qid
membranes (Nafion, fumapem) are used for PEM fuel cell hF ¼ ; (12)
Qre
and electrolysis applications [1,6] due to their superior proton
conductivity [29,30] and durability [31]. However, gases and where Qid denotes the ideal charge for the production of a
water can permeate through these membranes [25,27,32], given amount of hydrogen and Qre denotes the required
leading to hydrogen at the oxygen side and vice versa, a charge [6,9]. In addition, parasitic currents can reduce the
phenomenon commonly referred to as gas crossover. As later Faraday efficiency.
discussed, the ratios of specific proton conductivity and gas Until now, the deliberations have been dedicated to the
permeabilities of the membrane are decisive for the cell thermodynamics of PEM water electrolysis in order to derive
performance. the Nernst equation. Moreover, characteristics of PEMs and
Hydrocarbon membranes, which consist of a low cost non- safety issues due to gas crossover have been resumed. In the
fluorinated organic chain backbone with sulfonated func- following, a simulation estimating the gas crossover, Faraday
tional groups [28] have also been investigated for PEM water efficiency, and voltage efficiency will be presented in order to
14924 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3
computationally optimize the membrane thickness for PEM
water electrolysis with respect to the applied current density
and anodic as well as cathodic pressure. The hydrogen per-
meabilities used for the simulation will be obtained by fits of
the model to experimentally measured anodic hydrogen
contents, while the oxygen permeability, proton conductivity
and solubilities of hydrogen and oxygen are taken from the
literature [27].
2. Experimental
In-situ measurements of the anodic hydrogen content during
electrolysis were performed at a cell temperature of 80  C. An
electrolysis cell with an active area of 25 cm2 was equipped
with a Nafion N117 membrane coated with a cathodic Pt/C
and an anodic IrO2 catalyst layer. An individual water circu-
lation loop was used for each electrode. The measurement
range was limited to the safety threshold for the anode gas
impurity of 2% H2 in O2, which could only be slightly exceeded.
The hydrogen content of the anodic exhaust gas was
measured either with a K1550 gas analyser (HITECH In-
struments) or a XNX Transmitter (Honeywell). While the first
sensor measures the heat conductivity of the analyzed gas
(hydrogen has a larger heat conductivity than oxygen), the
second sensor measures the heat resulting from the catalyzed
reaction of both gases to water. The water content of the
measured gas was first condensed, then further reduced by
purging the gas through silica gel (Merck-chemicals).
3. Results and discussion
A model for the gas crossover during PEM water electrolysis
will be presented and compared to in-situ measurements. The
efficiency loss correlated to the estimated gas crossover,
ohmic resistance of the membrane and voltage enhancement
due to pressurized hydrogen and oxygen evolution will be
calculated. Furthermore, pressurized electrolysis and elec-
trolysis at atmospheric pressure with subsequent compres-
sion will be compared under efficiency considerations with
respect to isothermal and adiabatic compressor efficiencies.
Therefore, the results of the simulation will be exemplified at
hydrogen pressure of 30 bar, which is the minimum hydrogen
pressures that is considered to be suitable for methanation
processes and the direct integration into the natural gas grid
[22]. All stated pressures are referred to as absolute pressures.
The properties of Nafion membranes will be applied to the
model.
3.1. Simulation fundamentals
Based on Faraday’s law, the production rate density of
hydrogen
j cathode. During the transport, the gases can be diluted into
GH2 ¼
2F
during electrolysis can be calculated, where j denotes the
current density. Since the production of an oxygen molecule
requires twice the electrons compared to hydrogen (eq (4)), the
production rate density of oxygen ðGO2 Þ it is half that of
hydrogen.
By Fick’s law, the respective flux densities
Dc Dp
Fdif ¼ D ¼ εdif (14)
d d
of oxygen and hydrogen across the membrane of thickness
d correlated to the diffusion can be estimated (Fig. 1C). The
solubility (S ¼ εdif/D) of a gas in the membrane connects its
diffusion coefficient (D) and correlated permeability (εdif) [27],
as well as its partial pressure (p) and concentration (c) ac-
cording to Henry’s law (eq (9)). The diffusion of oxygen and
hydrogen during PEM water electrolysis across the membrane
is owed to the thermal movement of the molecules and the
concentration differences (Dc) at the anodic and cathodic
catalyst layer of the respective gases.
To reduce the enhancement of the Nernst voltage (eq (10))
due to the pressurization of oxygen and the concomitant
hazards [20,40], PEM water electrolysis can be performed at
smaller anodic than cathodic pressures [6]. The resulting dif-
ferential pressure between the anodic and cathodic catalyst
layer is an additional driving force for the hydrogen perme-
ation across the membrane (Fig. 1D), which is taken into ac-
count by the differential pressure flux density of hydrogen,
H2  pO2
pcat an
dp dp
FH2 ¼ εH2 (15)
d
dp
with the corresponding permeability εH2 . The partial pressures
of hydrogen in the cathodic and oxygen in the anodic catalyst
Fig. 1 e Schematic sketch showing transport mechanisms
that lead to gas crossover during PEM water electrolysis.
(A): Proton flux of the water electrolysis. (B): Due to the
proton flux, water and dissolved oxygen and hydrogen can
be electro-osmotically dragged from the anode to the
(13)
the surrounding water. (C): Diffusion of oxygen and
hydrogen. (D): Permeation due to differential pressure. (E):
Catalytic reaction of hydrogen and oxygen on the cathodic
platinum catalyst.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3
layer are denoted by pcat an
H2 and pO2 . The differential pressure is
expected to partly push the oxygen inside the membrane back
to the anode and could therefore reduce the oxygen crossover.
As later discussed, this effect could not be implemented into
the simulation.
In addition to the permeation mechanisms of the gases
discussed above, water and diluted gases can be transported
through the membrane by convection due to the electrolysis’
proton flux (Fig. 1B), a phenomenon commonly referred to as
electro-osmotic water drag [27,44]. The experimentally
determined temperature dependent dimensionless drag
coefficient
1
x ¼ 0:0134  T þ 0:03 (16)
K
of Nafion immersed in liquid water [45] corresponds to the
number of water molecules dragged along with each proton.
Hydrogen and oxygen diluted in water are thus carried along
with the proton flux across the membrane. The dissolved
oxygen at the anode is assumed to be carried by the perme-
ating water to the cathode. Thus, the resulting oxygen drag
flux density
pan
drag j O2 SO2
FO2 ¼ x (17)
F cðH2 OÞ
equals the product of the current density divided by the
Faraday constant, drag coefficient and ratio of diluted oxygen
(eq (9)) per water molecule at the anode. Immersed in water at
100  C, Nafion membranes weigh 38% more with a swelling of
15% compared to the dried state at room temperature [38]. The
dried material’s weight density [38] is 1.97 kg/l, while soaking
and swelling with water lead to a weight density of rwet ¼ 1.38/
1.15rdry. Using the water content of 11/1.38 ¼ 28%, the
water concentration c(H2O)¼0.28  rwet/MH2O z 37 mol/l in-
side Nafion membranes during electrolysis can be roughly
estimated. An alternative approach for the calculation of
c(H2O) can be realized by the amount of water molecules per
sulfonic group of the membrane. At 80  C about 17 water
molecules for each functional group [27] are taken up by
Nafion, which has an equivalent weight of EW ¼ 1100 g/mol.
A sulfonic group concentration of cðSO 3 Þ ¼ rwet =EW leads to a
water concentration of cðH2 OÞ ¼ 17  cðSO 3 Þ ¼ 43 mol=l. The
mean of both rough estimations is used for the calculation of
the oxygen drag flux density (eq (17)). During the transport
through the membrane, the oxygen concentration of the
transported water can decrease over the distance, as oxygen
dissolves into the surrounding water. This effect could not be
quantified, since the rates for the dissolution could not be
Table 1 e Numeric values of the parameters used for the model.
Gas A (bar cm2/A) εdif (1011 mol/cm s bar)
H2 2.4a 4.65a
O2 2.8c 2b
a
Obtained by the fits in Fig. 3.
b
From Ref. [27].
c
Calculated by a and b according to equation (22).
14925
estimated. By neglecting the dissolution of oxygen during the
transport by the electro-osmotic drag mechanism, equation
(17) can be interpreted as a maximum estimation of the oxy-
gen flux density due to the electro-osmotic water drag
mechanism.
The hydrogen flux density resulting from the water drag
could also only be estimated by its maximum contribution. In
contrast to oxygen, the electro-osmotic drag of hydrogen re-
duces the gas crossover since the hydrogen permeation from
cathode to anode has the opposite direction of the proton flux
(Fig. 1). The amount of diluted hydrogen at the cathode side of
the membrane is assumed to be the largest inside the mem-
brane, which determines the maximum possible contribution
of the drag mechanism for hydrogen. Similar to the calcula-
tions of the oxygen drag flux density, the maximum contri-
bution due to the hydrogen drag flux density
drag j pcat
H2 SH2
FH2 ¼  x (18)
F cðH2 OÞ
can be estimated, while the negative sign means the opposite
direction to the permeation in terms of Fick’s law (eq (14)) and
differential pressure (eq (15)).
The ratio of the flux densities of hydrogen diffusion and
convection due to water drag is not affected by the cathodic
hydrogen pressure (eqs (27) and (18)). For oxygen, this ratio is
independent on the anodic pressure (eqs (20) and (17)). The
largest contributions of the estimated drag mechanism to the
gas crossover have to be expected at high current densities
and thick membranes, where the contributions to the cross-
over of the diffusion mechanism is small (Fig. 4). Up to current
densities of 2 A/cm2 and a membrane thicknesses of 300 mm,
the estimated maximum hydrogen flux density due to water
drag is about tenfold smaller than that of the estimated
diffusion mechanism, using the parameters stated in Table 1.
For the oxygen crossover, this estimated ratio of the contri-
butions due to diffusion and electro-osmotic water drag cor-
responds to a factor of about three. Hence, the drag current
densities are not taken into account for the calculation of the
gas crossover, since only their maximum contributions could
be estimated and were found to be marginal compared to that
of the diffusion mechanisms. Thus, the entire hydrogen
permeation flux density across the membrane is assumed to
be given by the sum of the diffusion flux density (eq (14)) and
the permeation flux density due to differential pressure
(eq (15)):
H2  pH2
pcat dp pH2  pO2
an cat an
H2 ¼ εH2
Fen þ εH2
dif
(19)
d d
εdp (1011 mol/cm s bar) S (107 mol/cm3 bar)
2a 0.72b
e 0.8b
14926 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3

In this model, the partial pressures of hydrogen in the above) the influence of differential pressure on the oxygen
anodic catalyst layer and oxygen in the cathodic catalyst layer permeation across the membrane is neglected, the partial
are assumed to be independent on the gas crossover. pressure enhancement of oxygen at the anode is assumed
Furthermore, the influence of pressure on the solubility and to be
permeability is neglected, which is below 1% for the assumed
O2 ¼ p
pan þ AO2 j ;
an
(23)
conditions [27].
dp
H2 and εH2 are obtained by fits of the
The permeabilities εdif an
where p denotes the total pressure at the anodic gas outlet.
simulation to in-situ measurements of the anodic hydrogen Besides gas transport, supersaturation due to gas evolution
content (Fig. 3, Table 1), as later discussed. At the cathodic [46e49] could further lead to a partial pressure enhancement
platinum catalyst, oxygen that comes across the membrane of the produced gases in catalyst layers.
can be electrochemically reduced or catalytically reacted with The anodic partial pressure of hydrogen,
hydrogen forming water. Hence, it is assumed that the
cathodic oxygen content plays a negligible role. In addition, a Fen
H2
H2 ¼ pH2
pan cat
(24)
measurement of the cathodic oxygen content is not a measure GO2 þ Fen
H2

for the entire oxygen permeation across the membrane.

is proportional to the fraction of permeated hydrogen (eq (19))
Accordingly, a reduction of the oxygen crossover due to the
to the entire anodic gas composition. Using this approach,
influence of differential pressure could neither be measured
back diffusion of hydrogen from the anode to the cathode in
nor be implemented into the simulation. Despite this influ-
terms of Fick’s law (eq (14)) can be included, leading to an
ence, the oxygen permeation across the membrane is
entire hydrogen permeation flux density of
approximated by the diffusion mechanism according to Fick’s
 Fen

law (eq (14)),
H2 1  GO þFen
H2
pcat
H2  pO2
pcat an
H2 dp
Fen
H2 ¼ εH2 þ εH2 :
dif 2
(25)
pan
O2 d d
O2 ¼ εO2
Fen dif
(20)
d
This quadratic equation has two solutions
assuming a negligible partial pressure of cathodic oxygen. The

pcat pan
permeability of Nafion membranes at 80  C for the oxygen Fen
GO2 dp
H2 ¼ 1  εH2 
H 2 O 2
2 GO2 d
diffusion (Table 1) is obtained from the literature (Table 1). At
0 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
balanced pressure, the contributions of the permeation due to u pcat pcat pan (26)
B u εdif
H 2 þε dp H 2 O 2 C
differential pressure are negligible, so that the diffusion B1u
t1 þ 4 2 2 C A;
H d H2 d
@ pcat pan
mechanism dominates the crossover. GO2 1εH
dpH 2 O 2
2 GO d
2
Caused by the gas evolution of hydrogen in the cathodic
and oxygen in anodic catalyst layer, the produced gases have of which only the positive solution is of a physical nature. In
to be transported out of the catalyst layers to the respective case of a far larger oxygen production rate compared to the
gas outlets. Since the production rate densities of the gases are entire hydrogen permeation flux density, the square root can
pffiffiffiffiffiffiffiffiffiffiffiffi
proportional to the current density (eq (13)), a partial pressure be approximated by its Taylor series ð 1 þ xz1 þ x=2Þ,
enhancement of hydrogen in the cathodic catalyst layer with resulting in equation (19) with a negligible hydrogen partial
the same dependence in the current density is assumed. The pressure at the anode:
permeation of hydrogen out of the cathodic catalyst layer to
pcat H2  pO2
pcat an
the cathodic gas outlet is assumed to be affected by the dif- Fen
H2
H2 ¼ ε þ εdp
dif
(27)
ferential pressure in the same manner as the permeation d d
through the membrane, leading to partial pressure enhance- To this point, the local partial pressure enhancements in
ment of hydrogen in the cathodic catalyst layer of the catalyst layers due to gas evolution were assumed to be
  proportional to the current density. The local partial pressures
dp
εdif
H2 pH2 þ εH2
cat
H2  pO2
pcat an
of the evoluted gases in the catalyst layers were assumed to be
pcat
H2 ¼p
cat
þ AH2 j; (21)
εdif cat
H2 pH2
larger than at respective gas outlets, which results in an
enhanced crossover (eq (27)) and thermodynamic potential
where pcat denotes the total pressure at the cathodic gas outlet
according to the Nernst equation (eq (10), Henry). Further-
and AH2 the fit parameter for the partial pressure enhance-
more, the equations (26) and (20) for the respective perme-
ment of hydrogen in the cathodic catalyst layer. This fit
ation flux densities of hydrogen and oxygen across the
parameter depends on the structure, thickness and perme-
membrane were derived. Finally, the percental anodic
ability of the catalyst layer. For oxygen at the anode, the same
hydrogen content can be calculated by
effect occurs. Assuming, that the catalyst layers are of same
thickness and similar structure, the fit parameter for the Fen
H2
H2 in O2 ¼  100% : (28)
partial pressure enhancement of oxygen in the anodic catalyst GO2 þ Fen
H2
layer is assumed to be
  1 3.2. Comparison of the model with in-situ
AO2 z εdif
H2 =εO2
dif
AH2 ; (22)
2 measurements

according to a smaller permeability and half the production In the following it will be shown, that hydrogen is not elec-
rate of oxygen compared to hydrogen. Since (as discussed trochemically oxidized at the anodic iridium oxide catalyst
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3
layer. Subsequently, the simulated anodic hydrogen content
will be compared to in-situ measurements in order to evaluate
dp
the presented model and fit the stated parameters εdif H2 , εH2
and AH2 .
3.2.1. Hydrogen oxidation at the anodic catalyst layer
In PEM fuel cells a reduction of the hydrogen content at the
oxygen side can occur due to the electrochemical oxidation of
hydrogen at the platinum catalyst layer. In contrast to PEM
fuel cells, usually iridium oxide (IrO2) is used as the anodic
catalyst for PEM water electrolysis [1,3]. For the simulation of
the anodic hydrogen content presented above (eq (28)), a
reduction of the anodic gas impurity due to an electro-
chemical hydrogen oxidation at the anode is not included. To
verify this approach, it was experimentally examined, if
hydrogen oxidation at the anode catalyst can take place dur-
ing electrolysis. Therefore, a PEM water electrolysis cell con-
sisting of a Nafion N117 membrane with a IrO2 catalyst layer
on one side and a Pt/C catalyst layer on the other side was
used as electrochemical hydrogen pump. The anode (þ) was
purged with either humidified hydrogen or nitrogen, while the
cathode () was purged with hydrogen. An electrochemical
hydrogen oxidation at the anode and reduction at the cathode
leads to a current (Fig. 2).
Three different variations of the setup were used: First, the
Pt/C catalyst layer acted as anode with an applied DC voltage
of 0.05 V. By purging with nitrogen, a DC-current below
0.5 mA/cm2 was measured, while purging with hydrogen
yielded a current of about 18 mA/cm2, which can be attributed
to the electrochemical oxidation of hydrogen at the anode and
reduction of protons at the cathode. Second, the voltage and
gas supply was switched, so that the IrO2 side acted as anode
and a voltage of 0.6 V was applied. Again, the anode was
purged with either hydrogen or nitrogen, while the measured
current stayed constant below 0.5 mA/cm2. At last, a voltage of
1.5 V was applied to the same setup in order to simulate
electrolysis potentials. Due to electrolysis, a current of
approximately 9 mA/cm2 was measured, independent
whether the anode was purged with hydrogen or nitrogen.
Fig. 2 e Current densities of a hydrogen pump with either
Pt/C (blue line) or IrO2 (red and green line) as anodic
catalyst. The anode was purged either with humidified
nitrogen or hydrogen. The current at 1.5 V corresponds to
electrolysis. Not any hydrogen oxidation at IrO2 could be
measured. (For interpretation of the references to color in
this figure legend, the reader is referred to the web version
of this article.)
14927
Hence, it can be assumed, that no noticeable oxidation of
hydrogen on IrO2 occurs.
3.2.2. In-situ measurements of the anodic hydrogen content
The anodic hydrogen content of a PEM water electrolyzer with
a Nafion N117 membrane was measured in-situ at the anodic
gas outlet. Therefore pressures of 6 bar at the cathodic and
either 6 bar or 1 bar at the anodic gas outlet were applied.
Sensors based on the measurement principle of catalytic
combustion and heat conductivity were used to determine the
hydrogen content of the anodic gas (Fig. 3). The measured
hydrogen content with the first sensor is at low oxygen pro-
duction rates (partial load operation) smaller than the mea-
surements obtained using the heat conductivity sensor
(Fig. 3). Since the catalytic reaction of both gases to water
consumes twice as much hydrogen as oxygen, the composi-
tion of the gas mixture is influenced towards lower hydrogen
contents by the sensor that makes use of the catalytic com-
bustion. With approaching partial load operation, the influ-
ence of the catalytic combustion on the gas composition
increases (Fig. 3). Accordingly, the measurements with the
heat conducting sensor are considered to be more trust-
worthy, so that the simulation is verified with these
measurements.
In the dry state, the used Nafion N117 membrane is about
183 mm thick [38]. Due to swelling of approximately 14% of the
membrane under electrolysis conditions [38], a membrane
thickness of 209 mm is assumed for the comparison between
the experimental measurements and the simulation. In order
to determine the parameters εdif H2 and AH2 , the simulation (eq
(28)) was fitted to measurements obtained under balanced
dp
pressure. To determine the remaining parameter εH2 , the
simulation was fitted to the differential pressure measure-
ment while the other two parameters were held constant. The
obtained parameters are shown in Table 1.
Summarizing, it has been shown, that no observable
hydrogen oxidation during electrolysis at the anodic IrO2
2.5
2.0
H2 in O2 (%)
1.5
1.0
0.5 μ
0.0
0.0 0.5 1.0 1.5 2.0
Fig. 3 e In-situ measurements of the anodic hydrogen
content with sensors analyzing the heat conductivity and
the heat evolution due to catalytic combustion of the gas
mixture. At 80  C cell temperature and 6 bar cathodic
pressure and either 6 bar (balanced) or 1 bar (differential)
anodic pressure. The simulated results excluding and
including the mass transport of the produced gases in the
catalyst layers and the corresponding partial pressure
enhancements (eqs (21) and (23)) are plotted.
14928 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3

catalyst takes place. The decisive parameters for the hydrogen 25

crossover were fitted to the measurements of the anodic
20
hydrogen content, while the permeation coefficient correlated
to the oxygen diffusion was obtained from literature.

H2 in O2 (%)
15 μ

10
3.3. Simulated hydrogen crossover
5
μ
For the simulation of the hydrogen crossover, the permeabil-
0
ities (ε) and partial pressure enhancement coefficients (A) 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 2.0
shown in Table 1 are used. In Fig. 4, the simulated anodic
hydrogen content for membrane thicknesses from 50 mm to
300 mm is shown. Using thicker membranes, the hydrogen Fig. 5 e Simulated anodic hydrogen content (eq (28)) in
crossover due to diffusion decreases (eqs (14), (20) and (26)). dependence of the current density at 30 bar cathodic
With increasing current densities, the enhanced oxygen pro- pressure and balanced (left) and differential (right) pressure
duction rate leads to a reduced anodic hydrogen content (eq operation. The dashed black line corresponds to the safety
(28)). At balanced pressure, the simulated oxygen perme- limit of the gas mixture.
ation across the membrane is approximately by a factor of
H2 =εO2 ¼ 2:3 smaller than the one of oxygen, according to
εdif dif

equations (20) and (26).
In Fig. 5, the simulated anodic hydrogen contents at 30 bar
balanced and differential pressure operation are shown. With
enhanced pressures, the crossover due to diffusion increases
(eqs (14), (20) and (26)). Furthermore, at differential pressures,
where the cathodic pressure is larger than the anodic pres-
sure, the hydrogen permeation across the membrane is
enhanced (eqs (15) and (26)). Thus, at balanced pressure, the
hydrogen crossover is about one third smaller compared to
the same cathodic pressure at differential pressure, while the
oxygen permeation across the membrane is about 30 times
larger (eq (20)). At partial load operation, the safety limit of 2%
hydrogen in oxygen is exceeded with both, balanced and dif-
ferential pressure operation for membranes thinner than
300 mm.
Thus far, the hydrogen crossover has been estimated uti-
lizing the hydrogen permeation flux density (eq (26)) based on
Fick’s law of diffusion (eq (14)) and differential pressure (eq
(15)). The gas transport by electro-osmotically dragged water
has been estimated to be at least tenfold smaller for hydrogen
(one third for oxygen) related to the diffusion mechanism.
Concerning safety issues, cathodic pressures can lead to
problematic anodic hydrogen contents, which have to be
2 gas crossover, influence of anodic and cathodic pressures on
μ sponding overvoltages of other sources are not within the
H2 in O2 (%)
handled separately to ensure a partial load operation without
any hazardous explosive mixtures of both gases.
3.4. Faraday and voltage efficiency losses
The cell efficiency of a water electrolyzer is given by the
product of Faraday efficiency (eq (12)) and voltage efficiency
(hU):
hcell ¼ hF hU (29)
While gas crossover dominates the Faraday efficiency, the
voltage efficiency is affected by activation losses, mass
transport losses and ohmic losses. Furthermore, as shown
before, pressurized electrolysis enhances the required ther-
modynamical voltage according to the Nernst equation (eq
(10)). The membrane is the sole component that influences
both, Faraday and voltage efficiency, due to its gas perme-
ability and ohmic proton resistance, which are dependent on
the material, temperature and water content of the consid-
ered membrane [11,27,28]. Thus, an optimization of the
membrane thickness (with respect to its permeabilities and
specific proton conductivity) is crucial to minimize the effi-
ciency losses during electrolysis.
In the following considerations, the efficiency loss due to
the Nernst voltage (eq (10)) and proton resistivity of the
membrane will be discussed. Efficiency losses and corre-

scope of this research and will not be considered, since their

1 influence on the membrane thickness optimization is regar-
ded to be negligible.
The Faraday efficiency (eq (12)) can be expressed as func-
tion of the flux densities across the membrane and production
μ rates of the gases:
0
0.0 0.5 1.0 1.5 2.0 per
FH2
per
FO2
hF ¼ 1  2 (30)
GH2 GH2

Fig. 4 e Simulated anodic hydrogen content (eq (28)) in As discussed before, oxygen that comes across the mem-
dependence of the current density for 1 bar balanced brane from the anode to the cathode is assumed to be elec-
pressure and membranes thicknesses from 50 mm to trochemically reduced or catalytically reacted with hydrogen
300 mm. forming water. Hence, oxygen that crosses the membrane is
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3
assumed to consume twice the amount of hydrogen, taken
into account by a factor of two for the oxygen permeation in
equation (30). In terms of DC electronic contributions, the
conductivity of the membrane is not taken into account, since
the material is regarded to be electrically insulating. Addi-
tionally, the ionic conductivity of pure water (<105 S/cm)
inside the membrane is neglected, since the proton conduc-
tivity of Nafion is more than four decades larger [29].
The ohmic drop
UU ¼ j Rarea
due to the through plane area resistance of the membrane for
protons (Rarea) is proportional to the current density. The
proton conductivity of Nafion immersed in water at 80  C is
reported in the literature [28,29] and the mean of the found
values (s ¼ 0.2 S/cm) is used for the further calculations. In
addition, isotropic conductivities in the through-plane and in-
plane direction are assumed [30], leading to an estimated area
resistance of
d permeation, the simulated efficiency loss at balanced pres-
Rarea ¼ :
s sure is larger than at differential pressure.
PEM water electrolysis usually does not require any addi-
tional heating due to the heat coming from overvoltages.
Consequently, the heat energy (T dS) is not taken into account
for the estimation of the efficiency loss. Thus, the voltage
efficiency
Erev
hU ¼
U N þ UU
is normalized to the temperature dependent reversible
voltage (eq (3)), including the ohmic losses due to the mem-
brane resistance and the voltage enhancement due to pres-
surized electrolysis by the Nernst voltage (eq (10)). For the
complete estimation of the cell efficiency, activation losses,
mass transport losses and losses due to further resistances
can be implemented into this simulation by adding their
corresponding overvoltages (according to [18,50e52]) to the
denominator of the voltage efficiency.
In Fig. 6, the simulated cell efficiency loss (1hcell) at at-
mospheric pressure operation is plotted. The efficiency loss
due to gas crossover dominates at partial load, while the
30
20
μ 30
1- η (%)
10
μ
0 0
0.0 0.5 1.0 1.5
Fig. 6 e Simulated efficiency loss due to crossover and
ohmic resistance of the membrane in dependence of the
current density for 1 bar balanced pressure and membrane
thicknesses from 50 mm to 300 mm.
14929
ohmic drop is dominant towards large current densities. To
reduce the Faraday efficiency loss due to crossover, the
membrane thickness can be enhanced, which increases the
ohmic loss. Thus, a compromise between Faraday losses and
ohmic losses has to be found by a variation of the membrane
thickness, with respect to the pressures, desired current
density range and the safety threshold of 2% H2 in O2.
In Fig. 7, the simulated efficiency loss at 30 bar for both,
balanced and differential pressure electrolysis is plotted. If the
cathodic hydrogen pressure is far larger than the anodic ox-
(31) ygen pressure, the oxygen permeation across the membrane
will be far smaller than the hydrogen permeation, so the latter
dominates the efficiency loss correlated to the gas crossover.
Thus, the influence of differential pressure on the oxygen
crossover is neglected, as previously discussed. Besides an
enhanced thermodynamic voltage according to the Nernst
equation (eq (10)) for balanced compared to differential pres-
sure operation, the oxygen permeation is more 30 times larger
while the hydrogen permeation is roughly one third smaller at
balanced pressure. Accordingly, owed to the enhanced oxygen
(32)
Concerning the membrane material, the ratios of the
membrane’s specific conductivity and permeabilities are
crucial for the entire loss, while the membrane thickness is
only a parameter that has to be optimized. For example, hy-
drocarbon membranes could reduce the efficiency loss
compared to Nafion, if their ratios of specific proton con-
dp
H2 , εO2 and εH2 ) are larger.
ductivity (s) and permeabilities (εdif dif
(33)
3.5. Pressurized electrolysis vs. subsequent compression
The computational results in Fig. 7 have shown, that towards
high cathodic hydrogen pressures during electrolysis thicker
membranes are required to reduce the efficiency loss due to
crossover. However, thicker membranes are accompanied by
an enhanced ohmic drop, which reduces the voltage effi-
ciency. Below pressures of 100 bar and above temperatures of
30  C, the deviation in the pressure-volume diagram for iso-
therms between the ideal gas law (pV ¼ nRT) and Van der
Waals approximation of real gases is below 3% in pressure and
molar volume (V/n). Thus, hydrogen as an ideal gas assump-
tion (and therefore the Nernst equation) holds valid for the
considered pressures and is used in the following.
40
μ
1- η (%)
20
10 μ
2.0 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 2.0
Fig. 7 e Simulated efficiency loss due to crossover, ohmic
resistance of the membrane and pressurized hydrogen and
oxygen evolution in dependence of the current density at
30 bar balanced (left) and differential (right) pressure.
14930 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3
3.5.1. Isothermal compression
During pressurized electrolysis, hydrogen is produced under
isothermal conditions at constant pressure. The required en-
ergy for electrolysis at a higher hydrogen pressure compared
to a lower hydrogen pressure can be related to an isothermal
electrochemical compression, according to a voltage
enhancement (eq (11)). In terms of isothermal mechanical
compression, this voltage enhancement can be understood as
work
Zf


Vi pf
Wiso ¼ p dV ¼ nRT ln ¼ nRT ln ; (34)
Vf pi
i
where n denotes the amount of substance. The work during
compression has to be applied against the internal pressure of
the gas and is therefore counted as negative. Since the inter-
nal energy of the gas during isothermal process is constant,
the negative of the heat exchange to the environment
(dQ ¼ Wiso) balances the spent work by an enhancement of
the gases entropy
dQ
dS ¼ : (35)
T
At lower temperatures isothermal compression requires
less thermodynamic work compared to higher temperatures
(eq (34)). Novel techniques, as the hydrogen compression by
an ionic liquid could result in nearly isothermal low temper-
ature hydrogen compression.
To include the isothermal efficiency of a compressor hiso
comp
into the simulation presented above, the entire consumed
energy during subsequent compression is expressed by a
corresponding voltage (Ucomp) in terms of an equivalent elec-
trochemical hydrogen compression. Therefore, equation (11)
is divided by hiso
comp . By adding this voltage to the denomina-
tor of the voltage efficiency equation (33), the voltage effi-
ciency loss of electrolysis with subsequent isothermal
compression
Erev
hU ¼
UN þ UU þ Ucomp
can be determined.
At a cathodic pressure of 30 bar and anodic pressure of
1 bar, electrolysis with Nafion membranes is accompanied by
a simulated loss of at least 15% (Fig. 7). Whereas at atmo-
spheric pressures and a membrane thickness of 50 mm, the
estimated loss is below 12% for current densities from
0.2 A/cm2 to 2 A/cm2 (Fig. 8). The simulated entire efficiency
loss of electrolysis using a 50 mm thick membrane at atmo-
spheric pressure with additional isothermal subsequent
compression at 30  C for compressor efficiencies from 40% to
100% is shown in Fig. 8. Therefore, initial pressures of 1 bar are
assumed for the modeling of the efficiency losses of both,
electrolysis and isothermal subsequent compression. Since
the mole density of hydrogen is dependent on the tempera-
ture, the fraction of final and initial pressure decreases (ac-
cording to the ideal gas law) towards lower temperatures,
when compression to the same mole density is considered. In
terms of the stated example, the final pressure of 30 bar at
80  C corresponds to 30 bar  303.15 K/353.15 K ¼ 25.75 bar at
30  C. In this case, only Wiso(30  C)/Wiso(80  C) ¼ 82% of the
η
15
1- η (%)
10 η
5
0.0 0.5 1.0 1.5 2.0
Fig. 8 e Black line: Simulated efficiency loss for electrolysis
at 1 bar balanced pressure and 80  C cell temperature with
a 50 mm membrane. Colored lines: With additional
efficiency loss (eqs (36) and (29)) by subsequent isothermal
compression to 25.75 bar at 30  C, which equals 30 bar at
80  C. Compressor efficiencies from 40% to 100% in 10%
steps.
thermodynamic work for the compression at 80  C has to be
applied by compression at 30  C. If isothermal compression
efficiencies of at least 40% at 30  C compression temperature
can be achieved, probably lower cell efficiency losses over the
entire current density range compared to pressurized elec-
trolysis at 30 bar with Nafion membranes can be reached by
atmospheric pressure electrolysis with thin membranes and
subsequent compression (Fig. 7 and 8). The efficiency loss with
the discussed subsequent compression depends only slightly
on the current density, offering an enhanced efficiency at
dynamic operation compared to pressurized electrolysis. Be-
sides efficiency, safety issues due to gas crossover can be
avoided at low pressure electrolysis (Fig. 4).
3.5.2. Adiabatic compression
By using mechanical compression techniques, the heat ex-
(36)
change with the environment during compression can be
rather small, so that nearly adiabatic processes take place.
Thus, the temperature increases during compression, while
the entropy stays nearly constant. Accordingly, for adiabatic
processes more thermodynamic work is required compared to
isothermal processes, when the gas is compressed to the
same mole density. The thermodynamic work for adiabatic
compression of hydrogen as an idealized diatomic gas (5 de-
grees of freedom) is given by equation (37), where Ti,ad and pi,ad
denote the initial temperature and pressure at the beginning
of the process and pf,ad the final pressure.
!g1 !
g
5 pf;ad
Wad ¼  nRTi;ad 1 (37)
2 pi;ad
The adiabatic index g is defined by the ratio of the heat
capacities at constant pressure and constant volume (g ¼ 1.4
for an idealized diatomic gas). Due to a mixture of heat ex-
change with the environment and entropy enhancement of
real compression processes, the value of the index is in be-
tween the isothermal and adiabatic idealization (1 < g < 1.4).
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3
At the end of an idealized adiabatic process, the compressed
gas has a temperature of:
!g1
g
pf;ad
Tf;ad ¼ Ti;ad
pi;ad
According to the enhanced temperature, higher pressures
of the compressed gases are required to achieve the same
mole density as with low temperature isothermal compres-
sion. As discussed before, an ideal gas has the same density
when the ratio of pressures and temperatures are equal
(eq (39)). In the following, pf,iso and Tiso denote the pressure
and temperature of an isothermal process which leads to the
same mole density as the considered adiabatic process.
pf;iso Tiso
¼ (39)
pf;ad Tf;ad
By inserting equation (38) into equation (39) the pressures of
both, adiabatic and isothermal processes can be related to
each other, with respect to the same mole density of the
compressed gas at the end of the processes.

g
Ti;ad 1g
pf;ad ¼ g
pf;iso pi;ad
Tiso
Thus, during an adiabatic process, the required pressure to
achieve a defined mole density can be depicted as a function
of the corresponding isothermal pressure. By inserting
equation (40) into equation (37) the work of the adiabatic
compression is expressed as a function of the corresponding
isothermal pressure.
!g1
  5 Tgi;ad pgf;iso pg g isothermal compressor efficiency of 40% at 30  C, adiabatic
i;ad
Wad pf;iso ¼  nRTi
2 Tgiso
To reduce the temperature enhancement during an adia-
batic compression and the resulting additional work, various
stages for compression can be applied. In between these
stages the compressed gas can be cooled. The entire work
required for those stages is additive,
X Wjad
Wad;en ¼ ; (42)
j
had
j Towards higher pressures than the considered 30 bar, the
where Wjad denotes the thermodynamic adiabatic work
(eq (41)) and had
j the mechanical efficiency of the respective
adiabatic compression stages. By the ratio of equations (34)
and (42), the efficiency of an adiabatic process related to an
idealized isothermal compression can be determined, in order
to obtain the isothermal compressor efficiency hiso comp of an
adiabatic process.
Wiso ðTÞ
comp ¼
hiso
Wad;en ðTi Þ
In the following, an idealized four stage adiabatic
compression with cooling to 30  C at 3, 10 and 30 bar absolute
pressures in between the stages is assumed and the corre-
sponding isothermal compressor efficiency hiso comp is deter-
mined (Fig. 9). According to equations (38) and (40), the
approximation of an idealized gas still holds valid, while the
deviations in the adiabatic index of g ¼ 1.4 due to pressure and
temperature enhancement are negligible [53].
14931
15 1.0
10
(38)
W (kJ/mol)
0.9
ηcomp
iso
5
0.8
0
0 10 20 30 40 50
Fig. 9 e Left scale: Required thermodynamic work for
idealized isothermal compression (at 30  C and 80  C) and
adiabatic one and four stage compression (initial
temperature at each stage: 30  C). Right scale:
Thermodynamic efficiency (eq (43)) of idealized 4 stage
adiabatic compression compared to idealized isothermal
compression at 30  C and 80  C.
(40)
The assumed four stage mechanical compression has
thermodynamically about 80% (92%) of the isothermal effi-
ciency at 30  C (80  C) in the considered pressure range (Fig. 9).
In Fig. 8 it is shown, that with respect to a final hydrogen
pressure of 30 bar at 80  C, pressurized electrolysis is slightly
less efficient compared to electrolysis under atmospheric
pressure with subsequent compression using an isothermal
compressor efficiency of 40% at 30  C. Related to the
!
1 (41) compressor efficiencies of 50% for each of the assumed adia-
batic four stage compression lead to a similar performance
comp ¼ 50%  80% ¼ 40%Þ. Since the adiabatic index is
ðhiso
assumed to be larger than in real systems, the comparison
between subsequent adiabatic compression and pressurized
electrolysis is slightly for the benefit of the latter.
Concluding, the efficiency loss of pressurized electrolysis
at 30 bar with Nafion membranes can be undercut, if com-
pressors with the aforementioned efficiencies are realizable.
crossover enhances and further reduces the efficiency of
pressurized electrolysis, so that subsequent compression gets
more attractive in terms of efficiency. Depending on the effi-
ciencies of available mechanical compressors, a compromise
between pressurized electrolysis and subsequent compres-
sion may be reasonable.
(43) 4. Summary
In this work the influence of cathodic and anodic pressure
during PEM water electrolysis with Nafion membranes on the
gas crossover and the corresponding efficiency loss was
simulated. Therefore, diffusion processes and permeation due
to partial pressure differences between the anode and cathode
were estimated to dominate the hydrogen and oxygen cross-
over by exceeding the contributions to the gas crossover by
the electro-osmotic water drag. Experimentally, no hydrogen
14932 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3
oxidation at the anodic iridium oxide catalyst layer could be
observed. The simulated anodic hydrogen content could pre-
cisely be fitted to in-situ measurements at balanced and dif-
ferential pressure operation. By using the current dependency
of the measured anodic hydrogen content, the local partial
pressure enhancements in the catalyst layers due to gas
evolution could be estimated. At balanced pressure, the
hydrogen permeation was estimated to be about one third
smaller compared to differential pressure. The efficiency loss
correlated to the calculated gas crossover, Nernst equation
and ohmic proton resistance of the membrane, was estimated
with respect to the membrane thickness, cathodic and anodic
pressure and current density. Compared to differential pres-
sure operation, the enhanced oxygen permeation at balanced
pressure was estimated to lead to larger efficiency losses. For a
final hydrogen pressure of 30 bar, atmospheric pressure
electrolysis with isothermal (adiabatic) compressor effi-
ciencies of at least 40% (50%) was estimated to exceed the
efficiency of pressurized electrolysis. By adapting the utilized
parameters (proton conductivity, permeabilities, partial
pressure enhancement factors and temperature), the model
can be used for all types of proton exchange membranes in
water electrolysis applications.
Acknowledgment
This research has been supported by the German Federal
Ministry of Economics and Technology under Grant No.
03ESP106A. We thank Michael Hehemann, Stefanie Fischer
and Daniel Holtz for their support concerning the
measurements.
references
[1] Carmo M, Fritz DL, Mergel J, Stolten D. A comprehensive
review on PEM water electrolysis. Int J Hydrogen Energy
2013;38:4901e34.
[2] Mergel J, Carmo M, Fritz D. Status on technologies for
hydrogen production by water electrolysis. In: Stolten D,
Scherer V, editors. Transition to renewable energy systems.
Weinheim: Wiley-VCH; 2013. p. 423e50.
[3] Smolinka T, Rau S, Hebling C. Polymer electrolyte membrane
(PEM) water electrolysis. In: Stolten D, editor. Hydrogen
energy. Wiley-VCH; 2010. p. 271e88.
[4] Barbir F. PEM electrolysis for production of hydrogen from
renewable energy sources. Sol Energy 2005;78:661e9.
[5] Edwards PP, Kuznetsov VL, David WIF. Hydrogen energy. Phil
Trans R Soc A 2007;365:1043e56.
[6] Smolinka T. Water electrolysis. In: Garche J, editor.
Encyclopedia of electrochemical power sources. Elsevier B.V.;
2009. p. 394e413.
[7] Su H, Bladergroen BJ, Pasupathi S, Linkov V, Ji S. Performance
investigation of membrane electrode assemblies for
hydrogen production by solid polymer electrolyte water
electrolysis. Int J Electrochem Soc 2012;7:4223e34.
[8] Marshall A, Børresen B, Hagen G, Tsypkin M, Tunold R.
Hydrogen production by advanced proton exchange
membrane (PEM) water electrolysers - reduced energy
consumption by improved electrocatalysis. Energy
2007;31:431e6.
[9] Ursúa A, Gandı́a LM, Sanchis P. Hydrogen production from
water electrolysis: current status and future trends. IEEE
2012;100:410e26.
[10] Xu W, Scott K. The effects of ionomer content on PEM water
electrolyser membrane electrode assembly performance. Int
J Hydrogen Energy 2010;35:12029e37.
[11] Wei G, Xu L, Huang C, Wang Y. SPE water electrolysis with
SPEEK/PES blend membrane. Int J Hydrogen Energy
2010;35:7778e83.
[12] Antonucci V, Di Blasi A, Baglio V, Ornelas R, Matteucci F,
Ledesma-Garcia J, et al. High temperature operation of a
composite membrane-based solid polymer electrolyte water
electrolyser. Electrochim Acta 2008;53:7350e6.
[13] Ayers KE, Capuano C, Anderson EB. Recent advances in cell
cost and efficiency for PEM-based water electrolysis. ECS
Trans 2012;41:15e22.
[14] Ma L, Sui S, Zhai Y. Investigations on high performance
proton exchange membrane water electrolyzer. Int J
Hydrogen Energy 2009;34:678e84.
[15] Millet P, Ranjbari A, de Guglielmo F, Grigoriev S, Auprêtre F.
Cell failure mechanisms in PEM water electrolyzers. Int J
Hydrogen Energy 2012;37:17478e87.
[16] Millet P, Ngameni R, Grigoriev SA, Mbemba N, Brisset F,
Ranjbari A, et al. PEM water electrolyzers: from electrocatalysis
to stack development. Int J Hydrogen Energy 2010;35:5043e52.
[17] Millet P, Mbemba N, Grigoriev SA, Fateev V, Aukauloo A,
Etiévant C. Electrochemical performances of PEM water
electrolysis cells and perspectives. Int J Hydrogen Energy
2011;36:4134e42.
[18] Marangio F, Santarelli M, Pagani M, Calii Quaglia M. Direct
high pressure hydrogen production: a laboratory scale PEM
electrolyser prototype. ECS Trans 2009;17:555e67.
[19] Grigoriev SA, Kalinnikov AA, Millet P, Porembsky VI,
Fateev VN. Mathematical modeling of high-pressure PEM
water electrolysis. J Appl Electrochem 2009;40:921e32.
[20] Grigoriev SA, Porembskiy V, Korobtsev S, Fateev V,
Auprêtre F, Millet P. High-pressure PEM water electrolysis
and corresponding safety issues. Int J Hydrogen Energy
2011;36:2721e8.
[21] Norman T, Contact P. Unitized design for home refueling
appliance for hydrogen generation to 5,000 psi. DOE
Hydrogen and Fuel Cells Program; 2012. p. 122e5. FY 2012
Annual Progress Report.
[22] Bensmann B, Hanke-Rauschenbach R, Peña Arias IK,
Sundmacher K. Energetic evaluation of high pressure PEM
electrolyzer systems for intermediate storage of renewable
energies. Electrochim Acta 2013. http://dx.doi.org/10.1016/
j.electacta.2013.05.102 [in press].
[23] Nguyen VN, Fang Q, Packbier U, Blum L. Long-term tests of a
Jülich planar short stack with reversible solid oxide cells in
both fuel cell and electrolysis modes. Int J Hydrogen Energy
2013;38:4281e90.
[24] Barin I, Platzki G. Thermochemical data of pure substances.
3rd ed.; 1995. p. 795e6. Weinheim.
[25] Sethuraman VA, Weidner JW, Haug AT, Protsailo LV.
Durability of perfluorosulfonic acid and hydrocarbon
membranes: effect of humidity and temperature. J
Electrochem Soc 2008;155:B119e24.
[26] Sethuraman VA, Weidner JW, Haug AT, Motupally S,
Protsailo LV. Hydrogen peroxide formation rates in a PEMFC
anode and cathode. J Electrochem Soc 2008;155:B50e7.
[27] Ito H, Maeda T, Nakano A, Takenaka H. Properties of Nafion
membranes under PEM water electrolysis conditions. Int J
Hydrogen Energy 2011;36:10527e40.
[28] Kopitzke RW, Linkous CA, Anderson HR, Nelson GL.
Conductivity and water uptake of aromatic-based proton
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 8 ( 2 0 1 3 ) 1 4 9 2 1 e1 4 9 3 3
exchange membrane electrolytes. J Electrochem Soc
2000;147:1677e81.
[29] Slade S, Campbell SA, Ralph TR, Walsh FC. Ionic conductivity
of an extruded Nafion 1100 Ew series of membranes.
J Electrochem Soc 2002;149:A1556e64.
[30] Silva R, De Francesco M, Pozio A. Tangential and normal
conductivities of Nafion membranes used in polymer
electrolyte fuel cells. J Power Sources 2004;134:18e26.
[31] Laconti A, Liu H, Mittelsteadt C, McDonald R. Polymer
electrolyte membrane degradation mechanisms in fuel cells
- findings over the past 30 years and comparison with
electrolyzers. ECS Trans 2006;1:199e219.
[32] Zhang L, Ma C, Mukerjee S. Oxygen permeation studies on
alternative proton exchange membranes designed for elevated
temperature operation. Electrochim Acta 2003;48:1845e59.
[33] Borup B, Meyers J, Pivovar B, Kim YS, Mukundan R,
Garland N, et al. Scientific aspects of polymer electrolyte fuel
cell durability and degradation. Chem Rev 2007;107:3904e51.
[34] de Bruijn FA, Dam VAT, Janssen GJM. Review: durability and
degradation issues of PEM fuel cell components. Fuel Cells
2008;8:3e22.
[35] Liu H, Coms FD, Zhang J, Gasteiger HA, Laconti AB. Chemical
degradation: correlations between electrolyzer and fuel cell
findings. In: Büchi FN, Inaba M, Schmidt TJ, editors. Polymer
electrolyte fuel cell durability. Springer ScienceþBusiness
Media; 2009. p. 71e118.
[36] Gittleman CS, Coms FD, Lai YH. Membrane durability:
physical and chemical degradation. In: Mench MM,
Kumbur EC, Veziroglu TN, editors. Polymer electrolyte fuel
cell degradation. Elsevier Inc.; 2012. p. 15e88.
[37] Collier A, Wang H, Ziyuan X, Zhang J, Wilkinson D.
Degradation of polymer electrolyte membranes. Int J
Hydrogen Energy 2006;31:1838e54.
[38] DuPont Fuel cells. DuPontTM Nafion PFSA membranes.
http://www.fuelcells.dupont.com.
[39] Grigoriev SA, Millet P, Volobuev SA, Fateev VN. Optimization
of porous current collectors for PEM water electrolysers. Int J
Hydrogen Energy 2009;34:4968e73.
[40] Grigoriev SA, Millet P, Korobtsev S, Porembskiy V, Pepic M,
Etiévant C, et al. Hydrogen safety aspects related to high-
pressure polymer electrolyte membrane water electrolysis.
Int J Hydrogen Energy 2009;34:5986e91.
[41] Janssen H, Bringmann J, Emonts B, Schroeder V. Safety-
related studies on hydrogen production in high-pressure
electrolysers. Int J Hydrogen Energy 2004;29:759e70.
14933
[42] Young AP, Stumper J, Knights S, Gyenge E. Ionomer
degradation in polymer electrolyte membrane fuel cells.
J Electrochem Soc 2010;157:B425e36.
[43] Liu H, Gasteiger HA, Laconti A, Zhang J. Factors impacting
chemical degradation of perfluorinated sulfonic acid
ionomers. ECS Trans 2006;1:283e93.
[44] Zawodzinski TA, Davey J, Valerio J, Gottesfeld S. The water
content dependence of electro-osmotic drag in proton-
conducting polymer electrolytes. Electrochem Acta
1995;40:297e302.
[45] Onda K, Murakami T, Hikosaka T, Kobayashi M, Notu R, Ito K.
Performance analysis of polymer-electrolyte water
electrolysis cell at a small-unit test cell and performance
prediction of large stacked cell. J Electrochem Soc
2002;149:A1069e78.
[46] Tanaka Y, Kikuchi K, Saihara Y, Ogumi Z. Bubble
visualization and electrolyte dependency of dissolving
hydrogen in electrolyzed water using solid-polymer-
electrolyte. Electrochim Acta 2005;50:5229e36.
[47] Tanaka Y, Uchinashi S, Saihara Y, Kikuchi K, Okaya T,
Ogumi Z. Dissolution of hydrogen and the ratio of the
dissolved hydrogen content to the produced hydrogen in
electrolyzed water using SPE water electrolyzer. Electrochim
Acta 2003;48:4013e9.
[48] Kikuchi K, Tanaka Y, Saihara Y, Maeda M, Kawamura M,
Ogumi Z. Concentration of hydrogen nanobubbles in
electrolyzed water. J Colloid Interface Sci 2006;298:914e9.
[49] Kikuchi K, Nagata S, Tanaka Y, Saihara Y, Ogumi Z.
Characteristics of hydrogen nanobubbles in solutions
obtained with water electrolysis. J Electroanal Chem
2007;600:303e10.
[50] Fritz D, Carmo M, Mergel J, Stolten D. An efficient numerical
model with validation of a PEM electrolyzer. In: 4th European
PEFC and H2 Forum 2013.
[51] Santarelli M, Medina P, Calı̀ M. Fitting regression model and
experimental validation for a high-pressure PEM
electrolyzer. Int J Hydrogen Energy 2009;34:2519e30.
[52] Marangio F, Santarelli M, Cali M. Theoretical model and
experimental analysis of a high pressure PEM water
electrolyser for hydrogen production. Int J Hydrogen Energy
2009;34:1143e58.
[53] U.S. of Commerce on behalf of the United States of America.
http://webbook.nist.gov/chemistry/fluid/2011.