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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7

Available online at www.sciencedirect.com

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

Cell failure mechanisms in PEM water electrolyzers

P. Millet a,*, A. Ranjbari a, F. de Guglielmo a,c, S.A. Grigoriev b, F. Auprêtre c

a
Institut de Chimie Moléculaire et des Matériaux, UMR CNRS n
8182, Université Paris Sud, Bât 410, 91405 Orsay Cedex, France
b
Hydrogen Energy and Plasma Technology Institute, National Research Center “Kurchatov Institute”, 123182 Moscow Russia,
Kurchatov sq. 1, Russia
c
Compagnie Européenne des Technologies de l’Hydrogène, Zone Industrielle de la Prairie, 10 Rue de la Prairie,
91140 Villebon-sur-Yvette, France

article info abstract

Article history: PEM water electrolysis offers an efficient and flexible way to produce “green-hydrogen”
Received 24 February 2012 from renewable (intermittent) energy sources. Most research papers published in the open
Received in revised form literature on the subject are addressing performances issues and to date, very few infor-
2 June 2012 mation is available concerning the mechanisms of performance degradation and the
Accepted 5 June 2012 associated consequences. Results reported in this communication have been used to
Available online 12 July 2012 analyze the failure mechanisms of PEM water electrolysis cells which can ultimately lead
to the destruction of the electrolyzer. A two-step process involving firstly the local perfo-
Keywords: ration of the solid polymer electrolyte followed secondly by the catalytic recombination of
Hydrogen hydrogen and oxygen stored in the electrolysis compartments has been evidenced. The
Water electrolysis conditions leading to the onset of such mechanism are discussed and some preventive
PEM measures are proposed to avoid accidents.
Safety Copyright ª 2012, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights
reserved.

1. Introduction a function of both operating temperature and pressure. DG

becomes negative (i.e. the dissociation of water molecules
Hydrogen is currently produced from natural hydrocarbons becomes a spontaneous process) at very high temperature
(natural gas, oil and coal), using steam reforming or gasifica- (>2500 K), opening the way to the direct thermal dissociation
tion processes. As a result, large amounts of carbon dioxide of water into hydrogen and oxygen. However, the develop-
are emitted in the atmosphere (0.3e0.4 m3CO2/m3H2) [1] and ment of technologies operating in this elevated temperature
CO2 is considered as responsible for the increasing green- range remains very challenging, both in terms of materials
house effects and associated climate changes. In this context, and reactor design and the reaction is not feasible (at least
hydrogen produced from water offers a CO2-free alternative using conventional materials and technology) even from
path and a lot of R&D is carried out on the subject in view of concentrated solar energy sources. This is why, solar chem-
the so-called “hydrogen economy”. istry focuses on thermochemical cycles which proceed in
The water splitting reaction (dissociation of water mole- several steps and enables hydrogen generation at moderate
cules into molecular hydrogen and oxygen) is a non-- temperatures which are manageable by using more conven-
spontaneous and endo-energetic process: At standard tional engineering materials [3].
conditions (298.15 K, 101.3 kPa), the standard Gibbs free- Besides thermal processes, electricity can also be used to
energy change DG
¼ þ237.19 kJ mol1 [2]. DG is also decompose water. Water electrolysis offers a convenient and

* Corresponding author. Tel.: þ33 1 69 15 48 12.

E-mail address: [email protected] (P. Millet).
0360-3199/$ e see front matter Copyright ª 2012, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijhydene.2012.06.017
 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7 17479

flexible way to produce hydrogen. The total amount of energy a PEM cell/stack. Most research papers published in the liter-
bound in 1 mol of water is given by the enthalpy of formation ature on PEM water electrolyzers are addressing performances
of water (in standard conditions, DH
¼ þ285 kJ mol1). To issues. Indeed, the improvement of cell efficiency at
dissociate water molecules, part of DH can be supplied as increasing operating current density values remains a critical
thermal energy, with a maximum of DQ ¼ T. DS (where DS is challenge. Some reports are available concerning the long
the entropy change associated with the water splitting reac- term behavior of PEM electrolyzers. Single PEM electrolysis
tion), and the remaining part (DG) as electricity. Thus, the total cells using Nafion as solid electrolyte have been run with
energy can be supplied by a combination of electricity and steady performances over long periods of time (>20,000 h) [12]
heat and the amount of electricity can be reduced by but problems with excessive levels of hydrogen in the oxygen
increasing the operating temperature. This is an advantage in product stream have been reported on larger systems [13].
some cases, in particular when high-temperature heat is They have been ascribed to the thinning of the membrane
available as a “by-product” [4]. From a practical viewpoint, during operation and as a result, the membrane has been
different water electrolysis technologies are available for identified as the weakest part of a PEM electrolysis cell
operation at different temperatures. At low temperature regarding long term performances. Concerning safety issues,
(room e 100
C), alkaline and polymer-electrolyte membrane some information is also available, in particular those asso-
(PEM) technologies can be used. They differ mainly by the pH ciated to high pressure operation [14], but results are mostly
and the nature of the electrolyte. In the higher temperature related to risk analysis and prevention issues. The failure
range (800e1000
C), the less-mature solid oxide water elec- mechanism of PEM cells, a topic of major interest, is seldom
trolysis (the reverse of solid oxide fuel cell) can be used. treated in the open literature. The purpose of this paper is to
Results reported in this communication pertain to the low- provide an insight on such mechanism and to show that, at
temperature PEM water electrolysis process [5], which is least in some extreme cases, it can lead to stack destruction.
a sister-technology of PEM fuel cell technology [6e8] (H2/O2 Some recommendations are provided to monitor operational
PEM fuel cells were initially developed at the dawn of the US risks and improve the durability of the equipment.
space program, in view of electricity production in zero-
gravity environments). Although the large scale production
of hydrogen by PEM water electrolysis has been considered 2. Experimental section
from both technical and financial viewpoints in the early
eighties [9], the aim of most R&D programs since that time was 2.1. Membrane-electrodes assemblies
to develop and optimize small (up to several Nm3 H2/hour)
electrolysis units for the generation of molecular oxygen for Results presented in this communication were obtained using
breathing purpose in anaerobic environments. Civilian appli- a PEM stack operating at a maximum pressure of a few bars
cations appeared in the late XXth century when the world and delivering a maximum of 100 Nliter H2/hour. Rectangular
energy situation urged for the development of a non-fossil (10  7.5 ¼ 75 cm2) membrane e electrode assemblies (MEAs)
fuel economy. It turned out that PEM technology was well- were prepared using conventional (commercially available)
suited for operation using intermittent and fluctuating elec- materials. Perfluoro-sulfonic acid polymer (Nafion 117, Ion-
tricity sources and was therefore called to play a major role in Power Inc.) was used as solid polymer electrolyte. Platinum-
the management of renewable energy sources. However, group metals (PGM) were used as electrocatalysts. Carbon
domestic and industrial applications both require cheap and (Vulcan, Cabot Co.)-supported metallic platinum (with
reliable technologies. To reduce investments costs and meet a loading of approximately 0.4 mg cm2 of Pt) was used at the
market requirements, the possibility of increasing operating cathode for the hydrogen evolution reaction (HER) and
current density up to several A cm2 has been investigated. In a loading of approximately 2.0 mg cm2 of metallic iridium
state-of-art technology, a PEM water electrolyzer can now was used at the anode for the promotion of the oxygen
operate at 1e2 A cm2 with a high-heating-value (HHV) effi- evolution reaction (OER) [15]. The MEAs were prepared as
ciency close to 75e80% and the multiple A cm2 range is now follows. Suspensions of PGM catalyst particles in isopropanol
accessible. In addition, the possibility of delivering pressur- were mixed with a 5 wt.% alcoholic solution of Nafion (Sig-
ized hydrogen up to several hundred bars without additional maeAldrich Co.) and sonicated at 60
C. The catalytic inks
energy cost to facilitate hydrogen storage in pressurized thus obtained were repeatedly sprayed over the membrane
containers was demonstrated [10,11]. On the less positive side, surface in order to obtain a homogeneous coverage of the
there are increasing operational risks (explosion of H2/O2 gas active surface.
mixtures) associated with such severe operating conditions.
Although highly desirable, higher operating current density, 2.2. PEM cell
temperature and pressure must not lead to a less reliable
technology and for practical applications (both stationary and Main components of PEM water electrolysis cells are (Fig. 1) :
intermittent), enhanced electrochemical performances (1) the proton exchange membrane onto which are platted (2a,
should remain very stable and the lifetime of PEM water 2b) two electrocatalytic layers; (3a, 3b) the porous current
electrolyzers should be in the upper range of the 104e105 h collectors (1.2 mm thick plates of sintered titanium powder
time interval. are commonly used with an optimum open porosity of 40%
In this context, it is important to learn more about degra- [16]); (4a, 4b) a 2 mm thick cell spacer (rectangular grooves,
dation processes and to have a better understanding of the machine-made in the thickness of bipolar plates, or titanium
mechanisms which can ultimately lead to the destruction of grids used as spacers) used to manage a void were liquid water
17480 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7

operation at atmospheric pressure is pictured in Fig. 2. For

operation at higher pressures, a pressurization vessel is
required and stainless steel tubes are used to circulate water
and liquidegas biphasic mixtures in the circuitry. The stack
itself is clamped between two thick (1.5 cm) aluminium end-
plates. Individual water admission inlets are provided at the
bottom of one end-plate for the cathodic and anodic cell
compartments. Before operation, it was checked (using
helium) that there is no gas leak between the different cells.
During operation, liquid water is mechanically pumped
through both anodic and cathodic compartments. Liquidegas
mixtures are collected through outlets disposed on top of the
end-plates. Biphasic liquidegas mixtures (water e hydrogen
on the cathodic loop and water e oxygen on the anodic loop)
are then separated in gravity-separators and the water is re-
cycled back to the electrolyzer. The pressure drop across the
water outlet is such that liquid water pumped into the cell
compartments is homogeneously distributed over the entire
Fig. 1 e Schematic cross-section of a PEM water electrolysis active area of the cell. This is a prerequisite to insure
cell. 1 e solid polymer electrolyte; 2 e electrocatalytic a homogeneous distribution of current lines along the stack
layers; 3 e porous current collectors; 4 e bipolar plates; and to obtain a good overall efficiency.
5 e contacts points between current collectors and
electrocatalytic layers (d e average distance between
2.4. Stack characteristics
contact points between current collector and catalytic
layers); 6 e current collectors made of sintered powder
2.4.1. Charge density
particles. (a e anodic, c e cathodic).
Distribution of current lines inside the stack is a critical issue.
Heterogeneous current lines may be the initial step of a failure
mechanism. Therefore, before operation, it is important to
is circulated, and (4a, 4b) 2 mm thick titanium bipolar plates
check that within the stack, each individual PEM cell experi-
used to separate individual electrolysis cells in the stack.
ences a similar electrical environment. As discussed
Carbon-based gaskets (Viton fluoroelastomer from DuPont
elsewhere [17], cyclic voltammetry can be used for that
Co., not represented) are used as cell sealants. They usually
purpose. By measuring the total coulombic charge associated
can sustain an operating pressure of a few bars. In a PEM cell,
with each cell (Fig. 3), it is possible to obtain information about
electric current flows from one bipolar plate (4a) to the next
the homogeneity of the electrical environment for each MEA
one (4c). Liquid water molecules are transferred from the
and therefore about the distribution of current lines along the
anodic bipolar plate (4a) across the anodic current collector
stack during operation. As can be seen from Fig. 3, the values
(3a) up to the anode (2a) where they are oxidized into oxygen,
obtained for the seven cells of the stack are bracketing the
electrons and protons. Gaseous oxygen is then transferred
reference value measured on a 23 cm2 MEA used as bench-
back to the anodic bipolar plate across the anodic current
mark value. There are differences from one cell to the other
collector. Solvated protons are driven across the membrane (1)
but the discrepancy of datapoints remains limited. This is an
down to the cathode (2c) where they are reduced into
hydrogen which is then transported across the cathodic
current collector (3c) up to the cathodic bipolar plate (4c).
Water solvent molecules are released at the cathode (so-called
electro-osmotic flow of water). Current collectors of large
porosity will facilitate gas removal from the interfaces but will
also increase the ohmic resistance of the plates and introduce
additional parasite ohmic losses at contact points between
current collector and catalytic layers (front sides) and between
current collectors and channels (backsides). Therefore, an
optimization of the geometry of the pore structure is required
in terms of overall porosity and pore size distribution. As
discussed in Ref. [16], an open porosity of 40% is an adequate
value.

2.3. PEM electrolyzer

The PEM electrolyzer used in the experiments contains seven

individual (75 cm2 active area each) cells stacked in series in Fig. 2 e Photograph of the PEM water electrolyzer for
a filter press configuration. The configuration used for operation at atmospheric pressure.
 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7 17481

Fig. 3 e Charge densities measured on each cell of the PEM

water electrolyzer [17].

indication that the stack is homogeneously tightened and that

current lines are homogeneously distributed.

2.4.2. Electrochemical performances

Before investigating degradation processes, it is necessary to
record polarization curve of individual MEAs inside the stack
in order to check that electrochemical performances are
similar. As can be seen from Fig. 4a (where individual cell
voltages are plotted as a function of current density), an
inflexion point is sometimes observed at higher current
densities (ca. 400 mA cm2 in Fig. 4). The situation is generally
encountered when the stack is insufficiently tightened. As
a result, a screening effect develops. Its origin can be attrib- Fig. 4 e (a) Polarization curves recorded on individual PEM
uted either to the formation of an insulating film of gas cells at 80
C. (b) Polarization curves recorded on the stack
between the catalytic layer 2a or 2c and the porous current at 80
C.
collector 3a or 3c, or to a local drying of the membrane due to
a non-sufficient supply of liquid water. From a practical
viewpoint, this is of course a limiting current density beyond insure a constant supply of feed water to the anode and
which it is not possible to operate. However, when the stack is symmetrical operating conditions. The process was remotely
appropriately tightened, this screening effect disappears and monitored. Individual cell voltages were recorded on a regular
it is possible to increase the operating current density up to time basis as well as temperature of input and output water
much larger values (several A cm2). At an operating flows and hydrogen content on the oxygen circuit. Before
temperature of 80
C, a stack efficiency close to 80% (based on operation, several potential failure scenarios had been antic-
the high-heating-value of the reaction, i.e. on the energy ipated and the monitoring software was able to detect some
required to split liquid water) was obtained at 1 A cm2 specific dysfunctions during operation. In particular, faulty
(Fig. 4-b). In such conditions, the electrolyzer is operating in cells could be detected by comparing the value of their cell
a satisfactory way and there is no indication of potential voltage to reference values. A reduced cell voltage suggests
dysfunction. In particular, it is assumed that current lines are a possible cell short-circuit and a larger cell voltage indicates
homogeneously distributed between anode and cathode and an increased cell resistivity and/or an insufficient supply of
that the mean temperature inside the stack is also water to the anode. The hydrogen content in the anodic
homogeneous. liquidegas separator was also monitored. During the runs, the
power supply was automatically shut-down within seconds if
any of these criteria of dysfunction was met for any cell. It is
3. Results and discussion of course difficult, from an experimental viewpoint, to obtain
valuable information on an irreversible degradation process.
3.1. Experimental operating conditions This is because there is in general no clear forerunning sign
that can be used to detect an emerging dysfunction and
Experiments have been made to assess the ability of the because this is usually a slow process which suddenly
electrolyzer to operate continuously at high (1.8e2.0 A cm2) diverges and can have dramatic consequences for the integ-
current density. Thermostated liquid water (80
C) was rity of the electrolyzer. Nevertheless, in some situations, it has
continuously pumped through each cell compartment to been possible to stop the electrolyzer before total destruction
17482 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7

and to gain information on the failure mechanism. Results are presence of PTFE is the indication of a significant increase of
presented and discussed in the following sections. local temperature which could be ascribed to the spontaneous
and exothermic recombination of H2 and O2 into water. It can
be concluded that the run was stopped just after the perfo-
3.2. Step 1 of the failure mechanism: Perforation of the ration of the membrane and before the mixing of hydrogen
MEA and oxygen gases stored in the cell. It was also checked that
bipolar plates were not perforated.
3.2.1. Experimental result
A critical situation was met experimentally with the stack of 3.2.2. Discussion
Fig. 2 during a run at elevated (1.8 A cm2) current density. The The fact that the solid polymer electrolyte remains the
failure was detected automatically by the continuous moni- weakest part of a PEM cell has already been identified and
toring of individual cell voltage. The voltage of one cell reported in the literature [13]. Potentially, there are different
suddenly fell down below a threshold value of 1.5 V, sug- causes which may lead to the perforation of the membrane
gesting the onset of an electrical short-circuit. A burst of and to the subsequent mixing of hydrogen and oxygen gases.
hydrogen concentration in the oxygen production was also The most significant ones are listed in this section although at
detected and the power supply was shut-down immediately this stage (and this is certainly a limitation of this preliminary
to prevent the destruction of the stack. It was therefore work), the discussion remains mostly speculative.
possible to dismantle the stack and to get information about A first hypothesis is that there might be some mechanical
the failure mechanism at an early stage. It turned out that the origins. However, it seems unlikely that the forced circulation
membrane was perforated (Fig. 5). A hole was found in the of water in the anodic and cathodic compartments can lead to
upper area of the MEA, close to the exhaust pipe, and the the perforation of the membrane by mechanical friction. As
membrane was significantly put out of shape. The chemical can be seen from Fig. 1, main water flows are circulating along
analysis of the white deposit observed in the vicinity of the the bipolar plates (4-a and 4-b) and not directly against the
hole revealed that it mainly contained PTFE, suggesting the membrane surface. Also, the electro-osmotic drag of water
onset of a kind of local melting. It is known that Nafion from the anode to the cathode is not sufficient to cause
ionomer tolerates relatively high temperatures, with mechanical friction. However, the soft membrane material is
a melting point over 200
C [18]. According to DuPont (the clamped between two hard metallic current collectors. If the
manufacturer of the ionomer) a maximum operating stack is not homogeneously tightened, then there is a risk of
temperature of 190
C is possible when membranes are perforation. Post-mortem analysis of MEAs commonly show
bonded to reinforcing materials (this is not the case here) or the presence of some titanium particles deeply embedded
are used with liquids (this is the case here). A lower temper- inside the polymer. Surface prints and deformations which
ature of 150
C is quoted when the membrane is used in gas can ultimately lead to the short-circuit of the cell are also
dryers with dry gases and is not reinforced. Thermal decom- observed. However, as discussed in section 2.4.1, this was not
position of Nafion requires much higher temperatures. The the case in the present set of experiments. Cyclic voltammetry
measurements have shown that the cells were not short-
circuited. Therefore, it can be concluded that the perforation
occurred during electrolysis and that the hole is not the result
of a mechanical degradation process.
A second hypothesis is that the perforation was the result
of so-called “hot spots” due to the uneven distribution of
current lines over the active area. This assumption is sup-
ported by the fact that the failure occurred during operation at
high current density. If, for any reason (for example non-
homogeneous distribution of pressure forces at the surface
of the cells), there is a surface-area where the current density
is substantially larger than in others, then this is where the
temperature will increase due to the Joule effect. Although it
has been shown that heat dissipation in a PEM water elec-
trolysis cell is not a limiting factor, even at current densities of
several A cm2 [19], it is known that a Nafion membrane
soaked in water above 110
C loses all its mechanical strength
and falls apart. Also, the increase in the degree of swelling of
the membrane can lead to significant dimensional changes
and generate some internal constraints which can reduce its
mechanical strength. All these effects can contribute to the
perforation of the membrane.
A third hypothesis is the local drying of the membrane
due to insufficient amounts of water supplied to the anode.
As can be seen from Fig. 1, two opposite flows are taking
Fig. 5 e Photograph of the MEA after perforation. place at the same time across the anodic current collector
 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7 17483

(3a) : liquid water is supplied to the anode and gaseous hydrogen stream (the difference of concentration usually
oxygen is produced at the anode. The driving forces for these observed is due to the difference of H2 and O2 diffusion coef-
transfers are the gradients of concentrations across the ficients). The mechanism of gas permeability is the Fickian
current collector and capillary forces. Although the amount diffusion of dissolved gas across the membrane and the
of liquid water required to fuel the reaction is of limited driving force is the gradient of chemical potential set across
volume (e.g., at 1 A cm2, liquid water is supplied to the the membrane during operation. Therefore, flows of perme-
anode at a rate of 0.1 mm3 cm2 s1; at the same time, ating gases are proportional to gas mobility, membrane
gaseous oxygen is produced from the anode at a rate of thickness and to the difference of pressure set across the
58 Nmm3 cm2 s1), it becomes more and more difficult to membrane. Usually, contamination levels increase with
supply sufficient amounts of liquid water to the anode as the operating pressure. A critical step is reached when either the
current density increases. As a result, the electrical resis- hydrogen concentration in the oxygen anodic production line
tance of the membrane in this area will increase and in turn, reaches a threshold value of 4.1 vol.% or when the concen-
the temperature will also increase due to the Joule effect. tration of oxygen in the cathodic hydrogen production line
There is few if any information available in the literature reaches a threshold value of 4 vol.%, these figures corre-
about this phenomenon and it is difficult to know whether sponding to respectively the lower explosion limits of the H2/
such heterogeneities exist at the micro- or macroscopic scale. O2 gas mixtures. Such concentration levels can also be
However, as discussed in the experimental section, a direct reached gradually inside the stationary gaseous clouds which
consequence of local drying is the formation of a screening may form in the cells. Such effects and associated contin-
effect (see section 2.4.2) which can be detected by monitoring gency plans have been described and analyzed in details
the cell voltage. Our experiments have been made at elsewhere [14,20].
constant cell voltage. If a screening effect had occurred, then A fifth hypothesis is a process of membrane thinning as
the total current would have decreased. This has not been already reported elsewhere [13]. As a result, hot spots can
observed and it can be concluded that membrane drying due form and gas cross-permeation effects can increase. However,
to inappropriate water supply is not involved in the process such degradation process has been observed on electrolysis
of membrane perforation. stack after z15,000 h of intermittent operation at 1 A cm2
A fourth hypothesis is the formation of a hydro- and 80
C. This suggest that this is a rather a long-term process
geneoxygen atmosphere in at least one cell compartment, although operation at a higher current density (i.e. 1.8 A cm2
before the perforation of the membrane. The biphasic liquide as reported in this paper) could fasten the process and can be
gas mixtures in circulation in the PEM electrode compart- used as an ageing test.
ments contain mainly liquid water and a dispersion of small It is difficult at this stage to tell which, among the above-
gas bubbles (the mean bubble diameter is close to the diam- listed mechanism, is responsible for the perforation of the
eter of the pores i.e. up to a few microns). Although the liq- membrane. Would the same degradation processes take
uidegas separation takes place outside the PEM stack in place on the longer term if the cell was operated at lower
liquidegas separators (gravitation is used as driving force), current densities remains to be determined and additional
convection is such inside the stack that liquidegas separation experimental works is still required to obtain a better
can also take place (at least partly) directly inside the PEM cell understanding. From these results, it can only be concluded
compartments. As a result, a stationary gaseous cloud can that a critical step is reached once the membrane is perfo-
form and remain trapped on top of the cell compartments, rated. What can happen next? The perforation of the
close to the exhaust pipes where the hole shown in Fig. 5 membrane (Fig. 5) led to a cell short-circuit which was easily
formed (the formation of such gaseous clouds has been detected. However, this is not because a hole is formed that
observed experimentally using transparent model electrolysis the cathodic and anodic current collectors are immediately
cells). The subsequent catalytic recombination of H2 and O2 short-circuited. It is therefore expected that on some occa-
(promoted by PGM catalysts at the surface at the membrane) sions, it will be difficult to detect the accident. Even when the
can liberate sufficient heat to burn the membrane. It is membrane is perforated, the unaltered surface can still play
therefore necessary to understand how such a H2/O2 atmo- its role of cell separator. The cell voltage is maintained and
sphere can form. The fact is that the polymer electrolyte is not the gas production can proceed. This is probably why this
fully gastight. This is not due to the existence of pinholes or degradation mechanism is quickly and irreversibly diverging,
microscopic pores (the quality and homogeneity of commer- is difficult to detect and difficult to stop. Once a hole is
cial membrane products is such now a days that this is formed, the next step is the formation of a reactive gas
certainly not the main cause of gas transport across the mixture (gaseous hydrogen and oxygen) in the vicinity of the
membrane). In fact, this is due to gas cross-permeation hole and the subsequent combustion of the two gases.
phenomena which are known to take place. Under normal Hydrogen and oxygen produced in stoichiometric propor-
conditions of operation, hydrogen and oxygen produced at the tions at the anode and the cathode can mix and the catalysts
cathode and the anode of the PEM cell are properly separated coated onto the membrane (at least platinum used at the
by the membrane which acts as a solid electrolyte (ionic cathode for the HER) can promote their direct chemical
conductor) and as a cell separator. However, H2 and O2 solu- recombination. The process is spontaneous and exothermic.
bility in perfluoro-sulfonic polymers are such that gas In turn, the energy is released to the surroundings. This is
permeability across the membrane is possible. As a result, a divergent process. As discussed in the next section, this can
trace amounts of hydrogen are measured in the oxygen lead in some limiting cases to the total destruction of the
stream and trace amounts of oxygen are measured in the electrolyzer.
17484 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7

3.3. Step 2 of the failure mechanism: Hydrogeneoxygen

combustion

3.3.1. Experimental result

There are other experimental situations for which any
forerunning technical failure (such as the change in the cell
voltage or a lower gas purity mentioned in the previous
section) cannot be detected and for which the process can
rapidly lead (within only a few seconds) to the total destruc-
tion of the stack. This is what happened in another run at high
current density (1.8 A cm2). When the automaton used to
monitor the system detected a failure situation (sudden
decrease of current) and shut down the power supply, it was
already too late to avoid the destruction of the stack. As can be
seen from Figs. 6 and 7, the combustion of non-metallic cell
components (MEAs and cell sealants) was total, leading to the
formation of carbon deposits over the titanium bipolar plates.
The electrolysis stack was destroyed by combustion within
a few seconds.
Indirect information on the combustion process has been
gained by examining the stainless steel tubes used to circulate
water and liquidegas biphasic mixtures in the circuitry. The
effects of the combustion were not limited to the stack itself.
As can be seen from Figs. 8 and 9, metallic elbow connectors
located at a distance of ca. 20 cm away from the stack were Fig. 7 e Photograph of the separating plate.
neatly perforated during the combustion which originated in
the stack. The section of the hole measured on these elements
is equal to the internal section of the tube. This is an indica- gas mixture is saturated with water vapor, the temperature of
tion that a H2/O2 combustion flame formed inside the stack (at the flame is still expected to be sufficiently high to perforate
least in one cell) and then propagated along the tubing section the metallic components. An explosion would have produced
up to the first elbows in the circuitry and acted as a H2/O2 torch mechanical distortions inside the circuitry which were not
flame. The flame temperature of a stoechiometric H2/O2 gas observed.
mixture is known to be close to 3200
C [21]. Even if the H2/O2
3.3.2. Discussion
Once the solid polymer electrolyte is perforated, the second
step of the failure mechanism should logically be the mixing
of increasing amounts of gaseous hydrogen and oxygen in the
vicinity of the hole and the subsequent combustion of

Fig. 8 e Photograph of a stainless steel fitting drilled by

Fig. 6 e Photograph of the MEA. a H2/O2 flame formed inside the PEM stack.
 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7 17485

hardly be sufficient to raise the temperature of the membrane

by ca. 25
C (the estimation is made by assuming that the
medium is isotropic; no distinction is made between the
fluoro-carbon backbone and the aqueous phase of the
membrane). A temperature of 190
C would be reached for an
operating pressure of 7.4 bar. Therefore, it is necessary to
consider the total amount of gas stored in the stack to explain
the combustion of the electrolyzer, not only the local
concentration. The maximum amount of hydrogen stored in
a 75 cm2 PEM cell at atmospheric pressure is 16.5 N cm3.
According to the literature, H2eO2 torches can produce high
temperature (3200
C) flames by consuming few (2e5 cm3)
atmospheric hydrogen per second [22]. Therefore, there is
enough hydrogen in a 75 cm2 PEM cell to fuel such a flame for
Fig. 9 e Photograph of a stainless steel nut drilled by 5e10 s and enough hydrogen in the seven cell stack to fuel
a H2/O2 flame formed inside the PEM stack. such a flame for 30 s to 1 min. Commercial H2/O2 torches are
using a nozzle to slow down the combustion. In the PEM cell,
hydrogen in oxygen. The combustion reaction is a sponta- the combustion is probably faster. Nevertheless, there is
neous process over a large range of operating temperature. enough energy in a PEM cell operating at atmospheric pres-
The combustion is not favored by the negative entropy change sure to destroy the entire cell by combustion of the non-
of the reaction (due to the reduction of mole numbers) but the metallic cell components. In spite of the assumption used
reaction is largely exothermic (the enthalpy change is almost for the calculation (there is no liquid water inside the cells)
constant over the entire temperature range) and hence, DG is and even if the combustion of water-saturated hydrogen-
continuously negative. Although the operating temperature of oxygen gas mixtures provides less energy than the combus-
the PEM stack is low (<100
C) and although hydrogen and tion of dried gases, the amount of energy stored inside the
oxygen are saturated with water vapor, the presence of finely stack is sufficient to destroy the entire stack.
divided platinum catalytic particles on each side of the poly-
mer electrolyte can promote the recombination. Therefore,
the main risk for the electrolyzer is the catalytic recombina- 4. Prevention issues and recommendations
tion of hydrogen and oxygen stored inside the PEM cells and
even in the whole electrolysis unit. It should be noted here The following technical measures are recommended to
that in stationary conditions (constant current density), there reduce operational risks in electrolysis stack.
is the same amount of hydrogen and oxygen in each cell
compartment because the volume of each cell compartment is 4.1. Prevention measures for the electrolyzer
the same and the pressure is usually the same. Thus, the gas
composition in the PEM cell is sub-stoechiometric in hydrogen Some recommendations for high pressure water electrolysis
fuel, the combustion of which is the limiting factor. Therefore, are already available in the literature [23]. From the results
to evaluate the risk, it is necessary to determine the hydrogen presented here, there are different risk issues for a PEM water
(fuel) amount inside the cell as a function of operating current electrolyzer operating at high (>1.0 A cm2) current density.
density and pressure. This is not a trivial task because Different recommendations can be made to prevent the onset
a biphasic liquid e gas mixture is present in each compart- of irreversible degradation processes. First, circular cell
ment. A limiting case corresponding to a maximum energy geometry should be preferred over rectangular geometry when
value is obtained when there is no liquid inside the cathodic designing a PEM water electrolysis stack to avoid the risk of
compartment. The total amount of gas stored in the cells is formation of gaseous atmospheres in the upper part of the cell,
proportional to the size of the cells, to the internal geometry close to the cell outlet. Also, for the same reason, cell outlets
and to operating pressure. It is therefore possible to calculate should be located in the upper part of electrode compartments.
the amount of energy stored in each cell as a function of Second, it is necessary to measure the charge density for each
operating conditions (in particular as a function of current cell before operation and to determine whether there are risks
density). or not of uneven distribution of current lines. Third, the elec-
From geometrical considerations (Fig. 1), the internal trolyzer should operate at constant cell voltage and not
volume Vcell of each electrolysis compartment expressed per constant current density. If for any reason the impedance of
cm2 of active area is Vcell z 0.22 cm3/cm2. The maximum any cell increases, then the current dissipated in the stack will
amount of hydrogen stored in the PEM cell is obtained when decrease. It is also highly desirable to monitor individual cell
there is no liquid inside the cathodic compartment. This voltages : an increase in cell voltage can be related to inap-
amount is a direct function of operating pressure. At atmo- propriate water distribution in the anodic compartment and
spheric pressure (the pressure at which results reported in a decrease in cell voltage can be due to a cell short-circuit and
this communication were obtained), the maximum amount of the formation of a hole in the MEA which in turn can lead to the
hydrogen is 7.5  106 mol cm2, corresponding to formation of highly reactive H2/O2 gas mixtures. Of course,
a maximum energy of 2.14 J cm2 (HHV). If that amount of a continuous monitoring of the hydrogen content in oxygen
energy was released as heat to the surroundings, this would and of the oxygen content in hydrogen is also recommended.
17486 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7

If the electrolyzer is operating at high pressure, gas cross- current densities. Some preliminary experiments have
permeation effects across the solid polymer membrane revealed that current densities as high as 10 A cm2 could be
should be reduced as much as possible (for example by incor- reached. Besides practical interest (reduction of capital
porating inorganic fillers in the polymer or by using thicker expenses), experiments in this current density range
membranes), and the catalytic recombination of H2 and O2 combined with high (>100
C) operating temperatures are
should be actively promoted (for example by implementation expected to accelerate degradation kinetics and provide
of internal gas recombiners inside the PEM cells). a better understanding of ageing processes. Finally, some
synergies can be expected between R&D programs on H2/O2
4.2. Prevention measures for the electrolysis unit PEM fuel cells and PEM water electrolyzers since both tech-
nologies are using similar materials in quite similar operating
The discussion on operational risks can be extended to the conditions.
entire electrolysis unit where larger amounts of gases are
stored (in particular in the liquidegas separators) and where
explosion hazards are more severe. There are a number of
safety requirements for the closed-premises in which Acknowledgement
hydrogen leaks can potentially occur. In particular, there is
a need for a continuous ventilation on the production site in Financial support from the European Commission within the
order to maintain the hydrogen concentration to values less 6th Framework Programme (GenHyPEM, STREP no 019802,
than the 4 vol.% corresponding to the lower explosion limit of 2005e2009) is gratefully acknowledged.
H2/O2 gas mixtures. Ventilation should be made with an inert
gas to reduce the inflow of oxygen which could participate in
the ignition or explosion of hydrogen. references

5. Conclusion [1] Muradov N. Hydrogen via methane decomposition: an

application for decarbonization of fossil fuels. International
PEM water electrolysis is considered as an efficient and flex- Journal of Hydrogen Energy 2001;26:1165e75.
ible technology for the production of “green-hydrogen” from [2] Mortimer CE. Chemistry. 4th ed.; 1986. Belmont.
renewable (intermittent) energy sources. Whereas most [3] Roeb M, Neises M, Säck J-P, Rietbrock P, Monnerie N, Dersch J,
et al. Operational strategy of a two-step thermochemical
research papers published on the subject are addressing
process for solar hydrogen production. International Journal
performances issues, there is also a need to identify and of Hydrogen Energy 2009;34:4537e45.
analyze the mechanisms of performance degradation and the [4] Kroschwitz JI, Howe-Grand M, editors. Kirk-Othmer
associated consequences. This is a topic seldom treated in the encyclopedia of chemical technology. 4th ed., vol. 13.
open literature. Nevertheless, it is critical to assess the ability New-York: Wiley VCH; 1995.
of the technology to operate safely on the long-term and to [5] Grubb WT Jr., US Patent No. 2,913,511, 17 November 1959.
[6] Barbir F. PEM fuel cells, theory and practice. Elsevier
identify potential operational risks. Results reported in this
Academic Press; 2005.
communication have been used to analyze the failure mech-
[7] Grubb Jr WT. Batteries with solid ion-exchange electrolytes.
anisms which can lead to the destruction of a PEM electro- Journal of the Electrochemical Society 1959;106:275e9.
lyzer. Although it remains difficult to establish a clear [8] Millet P, Andolfatto F, Durand R. Design and performances of
relationship between operating conditions and degradation a solid polymer electrolyte water electrolyzer. International
processes (several postulated degradation mechanisms listed Journal of Hydrogen Energy 1996;21:87e93.
in this paper are hypothetical and would require experimental [9] Solid polymer electrolyte water electrolysis technology
development for large scale hydrogen production. General
confirmation), it is shown that there are some experimental
Electric Company; 1981. DOE Report DOE/ET/26 202-1.
situations (in particular when the cell is operated at elevated [10] Millet P, Ngameni R, Grigoriev SA, Mbemba N, Brisset F,
current density) which can lead to the perforation of the Ranjbari A, et al. PEM water electrolyzers: from
membrane. Therefore, the major risk for PEM electrolysers is electrocatalysis to stack development. International Journal
due to the possible catalytic recombination of hydrogen and of Hydrogen Energy 2010;35:5043e52.
oxygen stored inside the electrolysis compartments. Accord- [11] Millet P, Ngameni R, Grigoriev SA, Fateev VN. Scientific and
ing to our experience, internal H2/O2 combustion prevails over engineering issues related to PEM technology: water
electrolyzers, fuel cells and unitized regenerative systems.
explosion. At this stage, additional insights are still required to
International Journal of Hydrogen Energy 2011;36:4156e63.
clearly identify the causes of degradation processes because [12] Scherer GG, Devantay H, Oberlin R, Stucki S. Wasserstoff und
there is no clear forerunning sign that could be used to detect Ozonerzeugung durch Membrel-Wasserelektrolyse.
dysfunctions and because the failure of a PEM cell is usually Dechema Monographien 1985;98:407e15.
a slow process which suddenly diverges over a very short [13] Stucki S, Scherer GG, Schlagowski S, Fischer E. PEM water
period of time of only a few seconds. In spite of these limita- electrolysers: evidence for membrane failure in 100 kW
demonstration plants. Journal of Applied Electrochemistry
tions, some preventive measures that can be used to avoid
1998;28:1041e9.
accidents have been listed and some recommendations are
[14] Grigoriev SA, Millet P, Korobtsev SV, Porembskiy VI, Pepic M,
proposed. In terms of perspectives, a better understanding of Etievant C, et al. Hydrogen safety aspects related to high-
irreversible degradation processes could be obtained by pressure polymer electrolyte membrane water electrolysis.
investigating the behavior of MEAs operating at very high International Journal of Hydrogen Energy 2009;34:5986e91.
 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 7 ( 2 0 1 2 ) 1 7 4 7 8 e1 7 4 8 7 17487

[15] Grigoriev SA, Millet P, Fateev VN. Evaluation of carbon- [19] Millet P. Water electrolysis using EME technology:
supported Pt and Pd nanoparticles for the hydrogen temperature profile inside a Nafion membrane during
evolution reaction in PEM water electrolyzers. Journal of electrolysis. Electrochimica Acta 1991;36(2):263e7.
Power Sources 2008;177:281e5. [20] Grigoriev SA, Porembskiy V, Korobtsev S, Fateev VN,
[16] Grigoriev SA, Millet P, Volobuev SA, Fateev VN. Optimization of Aupretre F, Millet P. High pressure PEM water electrolysis
porous current collectors for PEM water electrolyzers. and corresponding safety issues. International Journal of
International Journal of Hydrogen Energy 2009;34(11):4968e73. Hydrogen Energy 2011;36:2721e8.
[17] Millet P, Mbemba N, Grigoriev SA, Fateev VN, Aukauloo A, [21] North American combustion handbook. North American
Etievant C. Electrochemical performances of PEM water Mfg. Co; 1983.
electrolysis cells and perspectives. International Journal of [22] Viall E. In: Gas torch and thermit welding. McGraw-Hill; 1921.
Hydrogen Energy 2011;36:4134e42. [23] Marangio F, Santarelli M, Calı M. Theoretical model and
[18] Lage LG, Delgado PG, Kawano Y. Thermal stability and experimental analysis of a high pressure PEM water
decomposition of Nafion membranes with different cations. electrolyser for hydrogen production. International Journal
Journal of Thermal Analysis and Calorimetry 2004;75:521e30. of Hydrogen Energy 2009;34:1143e58.