You are on page 1of 8

Electrochimica Acta 82 (2012) 384–391

Contents lists available at SciVerse ScienceDirect

Electrochimica Acta
journal homepage: www.elsevier.com/locate/electacta

Advanced alkaline water electrolysis


Stefania Marini a , Paolo Salvi a , Paolo Nelli a , Rachele Pesenti a , Marco Villa a,∗,1 , Mario Berrettoni b ,
Giovanni Zangari c,1 , Yohannes Kiros d,1
a
CSGI and Dipartimento Progettazione &Tecnologie, University of Bergamo, 24044 Dalmine, Italy
b
Department of Physical and Inorganic Chemistry, University of Bologna, 40196 Bologna, Italy
c
Materials Science & Engineering, University of Virginia, Charlottesville, USA
d
Department of Chemical Engineering and Technology, KTH, 100 44 Stockholm, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: A short review on the fundamental and technological issues relevant to water electrolysis in alkaline
Received 22 February 2012 and proton exchange membrane (PEM) devices is given. Due to price and limited availability of the
Received in revised form 2 May 2012 platinum group metal (PGM) catalysts they currently employ, PEM electrolyzers have scant possibilities
Accepted 2 May 2012
of being employed in large-scale hydrogen production. The importance and recent advancements in the
Available online 16 May 2012
development of catalysts without PGMs are poised to benefit more the field of alkaline electrolysis rather
than that of PEM devices. This paper presents our original data which demonstrate that an advanced
Keywords:
alkaline electrolyzer with performances rivaling those of PEM electrolyzers can be made without PGM
HER
OER
and with catalysts of high stability and durability. Studies on the advantages/limitations of electrolyzers
Water electrolysis with different architectures do show how a judicious application of pressure differentials in a recirculating
Non-noble metal catalysts electrolyte scheme helps reduce mass transport limitations, increasing efficiency and power density.
H2 economy © 2012 Elsevier Ltd. All rights reserved.

1. Introduction 1.1. The hydrogen evolution reaction (HER)

The water splitting reactions (Eq. (1)) is endothermic and there- The cathodic side of (1) is the HER, which can be written as:
fore requires energy which can be readily, but not uniquely,
2H2 O + 2e− → H2 + 2OH− (2a)
provided by the flow of an electric current through a suitable elec-
trochemical cell. and

H2 O + energy → H2 + 1/2O2 (1) 2H+ + 2e− → H2 (2b)

in an alkaline and acidic electrolyte, respectively. HER starts with


Despite its deceptive simplicity, the electrochemical splitting of the Volmer step, which involves the binding of atomic hydrogen to
water has a long, multifaceted and still ongoing history, which has the electrode at an adsorption site M
even today enormous relevance from both a scientific and techno-
logical standpoint. On one hand, a detailed understanding of the M + H2 O + e− → MHad + OH− (3a)
anodic and cathodic mechanisms of this reaction is needed if it + −
M + H + e → MHad (3b)
has to occur at high rates and with a minimum electrical energy
input; on the other hand, if the promises of H2 as a large scale and is completed with a desorption step which occurs either via
energy carrier are ever to be fulfilled, electrolyzers and fuel cells the Tafel reaction:
are needed which are inexpensive, reliable, compact, and with
considerable operational flexibility. We argue that, so far, the tech- MHad → 2M + H2 (4)
nological development has lagged much behind the important and or by the Heyrovsky reaction:
recent advances in our fundamental understanding of this reaction,
which we summarize below. H2 O + MHad + e− → M + H2 + OH− (5a)
+ −
MHad + H + e → H2 + M (5b)

∗ Corresponding author. The H-adsorption/desorption mechanism requires that hydro-


E-mail address: mvilla@unibg.it (M. Villa). gen binds well, but not too strongly, to the reaction site M: the
1
ISE member. exchange currents as a function of the M H bond strengths describe

0013-4686/$ – see front matter © 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.electacta.2012.05.011
S. Marini et al. / Electrochimica Acta 82 (2012) 384–391 385

a “volcano curve” having a maximum in correspondence of a bond- valence band states and occupied conduction band states; the latter
ing energy of ∼240 kJ/mol, near to that of Pt [1]. The H-adsorption number alone may also be a valid descriptor [10]. Rossmeisl et al.
energy is a good “descriptor” for HER, which guides the identifica- [11] performed DFT studies of the rutile-type oxides (TiO2 , RuO2 ,
tion of the most promising materials. The HER exchange current of IrO2 ) and Mn clusters in photosystem II assuming that the reaction
Pt is at least two orders of magnitude higher in acids than in alka- in acid (6b) begins with formation of an O O bond between one sur-
line electrolytes, including KOH, possibly because of the shorter face transition metal oxogroup (TM O) and one water molecule. By
Pt-Had distance in alkaline electrolytes suggested by theoretical comparing the reactions of water splitting in biological and elec-
estimates [2]. Subbaraman et al. [3] claim that Ni(OH)2 nanoclus- trochemical systems, Rossmeisl et al. [12] showed that ruthenium
ters on the Pt surface enhance HER rates in 0.1 M KOH by one order oxide is the best material among the inorganic compounds investi-
of magnitude; however, no theoretical explanation of such a syner- gated, but demonstrated also that the Mn clusters of photosystem
getic effect was attempted; furthermore, the long-term stability of II are more efficient than RuO2 toward OER, and that even better
Ni(OH)2 in a strongly reducing environment such as that occurring systems are theoretically possible.
at the cathode is not discussed. RuO2 and IrO2 , alone or in combination, are often considered
It may be expected that, by alloying metals with higher (e.g. Ni) the reference OER catalysts, but neither is “ideal”; thermodynamic
or lower (e.g. Mo) than optimal H-bond energy, the turnover rate calculations indicate that RuO2 binds oxygen a little too weakly
of the reaction centers and the intrinsic catalytic activity would (thus hindering oxidation of HO*) while IrO2 binds oxygen a lit-
increase. However, the evidence of these synergetic effects has tle too strongly (which limits formation of HOO*) [13]. The DFT
proven elusive for many years [4] due, in our opinion, to the fact that computations of Man et al. [9] would suggest that cobalt oxide,
the materials being compared were insufficiently characterized and Co3 O4 , is more reactive than RuO2 , a result supported only by recent
that the theoretical treatments involved were not sufficiently com- experimental data (with excess oxygen being a possible reason for
prehensive. the discrepancies with older data). In particular, the OER turnover
Despite the intrinsic difficulties in relating catalytic activity with frequency of gold-supported cobalt oxide in 0.1 M KOH has been
electronic-structure computations, the density functional theory estimated to be ∼40 times higher than that of “bulk” Co3 O4 [14].
(DFT) is now giving important contributions to the most relevant The reverse of the OER reactions (6a) and (6b) is the oxygen
advances in this area of electrocatalysis, and in particular to the reduction reaction (ORR) where molecular oxygen is reduced to
issue of synergetic effects in catalysis. Greeley et al. [5] for example water. ORR involves the same intermediates as the OER and is most
confirmed experimentally the prediction of a large scale combi- conveniently carried out in an alkaline environment [15] where
natorial screening that a BiPt surface alloy would be a better HER catalysts are available which are more active and stable than in
catalyst at pH 0 than Pt. Along the same lines, a possible low cost acid. It could have been expected that the same holds true for OER
replacement of Pt as HER catalyst in acidic solutions was suggested but a full experimental confirmation has just been published [16],
to be MoS2 in the form of vapor deposited nanoparticles [6]; exper- a reason being the mild instability of RuO2 and IrO2 in strong alkali.
iments and DFT simulations proved in fact that the performance of The OER activities of RuO2 , IrO2 and Co3 O4 at pH 0.1 and pH 14
MoS2 would be second only to those of platinum group metal cat- are generally comparable [16] but the Tafel slopes are smaller in
alysts. Even better performances at pH 0 have been obtained with alkali than in acid (e.g. 49 vs 59 mV/decade for IrO2 ) and the highest
amorphous MoS3 , which is easier to prepare than crystalline MoS2 activities are attained at high currents and high pH. The results of
by physical vapor deposition, but is still a long way from practical Minguzzi et al. [16] at small overpotentials are consistent with a
use [7]. The HER literature is probably the richest in electrochem- study of IrO2 and RuO2 nanoparticles by Lee et al. [17] who found
istry, but its practical usefulness is limited by the still prevailing slightly higher activities in acid than in base at very low currents.
trial and error approach in developing a new catalyst, the long-term
stability of which is seldom explored. 1.3. Technological and economical issues in electrolysis

1.2. The oxygen evolution reaction (OER) The equilibrium potential of 1.23 V for reaction (1) under stan-
dard conditions divided by the actual cell potential (Ecell ) at the
In alkaline and acidic electrolytes, respectively, the OER can be operating current density gives the conversion efficiency; the sys-
written as: tem efficiency is the ratio between the theoretical energy demand
to produce 1 N m3 of H2 at 1.23 V (3.0 kWh) and the energy required
2OH− → 1/2O2 + H2 O + 2e− (6a) by the overall system.

H2 O → 1/2O2 + 2e + 2H +
(6b) Most large-scale industrial electrolyzers are alkaline, and use
Ni-based electrodes. The actual situation does not seem to have
This reaction may occur through various pathways and adsorbed changed much from a 2004 NREL report [18] or even earlier reviews
intermediates (O*, HO*, HOO*) which makes the identification of a [19]; also the 2011 EERE interim updated report [20] is essentially a
“descriptor” somehow more difficult. Suntivich et al. [8] argue that copy of the 2007 version. In the most favorable situation of the fore-
the level of occupancy of the eg -symmetry electrons in the tran- court case (∼500 N m3 H2 /h) and with electric energy at 0.05$/kWh
sition metal (M) of a perovskite AMO3 determines the binding of the 2004 NREL report finds that the capital and energy costs to pro-
oxygen intermediates to M, and is an appropriate descriptor for duce 1 N m3 H2 are 0.11$ and 0.20$, respectively. However, with a
OER in alkali. These authors experimentally confirm the identifica- ∼50 times smaller “neighborhood” plant (∼10 N m3 H2 /h) the cap-
tion of the “best” compound: Ba0.5 Sr0.5 Co0.8 Fe0.2 O3−␦ (BSCF) within ital cost escalates to 1.1$ per N m3 H2 .
a family of AMO3 compounds, in substantial agreement with the Table 1 lists the main characteristics of water electrolyzers, both
theoretical overpotentials estimated by Man et al. [9] who identify of the acidic proton exchange membrane (PEM) and alkaline types.
SrCoO3 as the best catalyst among simple perovskites. Man et al. The “conventional” alkaline electrolyzers typically operate with cell
[9] believe that a suitable descriptor for the OER is the difference potentials near 2 V at current densities up to 300 mA/cm2 with
between the calculated free energy of reaction G◦ of the second a system efficiency near 60% [18] and demonstrated lifetimes in
and the first electron abstraction step (GO* ◦ − GHO* ◦ ); this quan- excess of 10 years and in some cases more than 20 years. The cap-
tity describes correctly experimental trends from the literature but ital cost per unit production rate is much size-dependent, and we
is of course model-dependent. A “classic” descriptor is the number have rounded the 665$/kW figure of a forecourt 2006 case study
of 3d electrons, eg + t2g , which has contributions from both occupied [20], which is roughly consistent with the findings of Ivy [18] for
386 S. Marini et al. / Electrochimica Acta 82 (2012) 384–391

Table 1
Typical figures of water electrolyzers.

Conventional alkaline PEM Advanced alkaline

Ecell ∼2 V 1.8 V <1.8 V


kWh (system)/N m3 H2 ∼5.0 4.3a <4.4
Current density (mA/cm2 ) >100 >1000 >500
Lifetime (years) >10 ∼5 >10
Cost ($/kW) >700 400a ?
a
[23].

alkaline electrolyzers. The capital cost is by far the largest contribu- (or super gravity) G-field limits the effects of the bubbles and
tion to the price of electrolytic hydrogen for alkaline plants below reduces Ecell at high currents [30]. The achievable power saving rel-
50 kW [18], which have price tags >3000$/kW, and an estimated ative to G = 1, is substantial at Ecell > 2.5 V and above 300 mA/cm2 ,
investment contribution >0.6$/N m3 H2 . This implies that capital but it hardly pays off due to the increased complexity of the system.
cost, rather than efficiency and price of electrical energy, is the The effects of a magnetic field upon Ecell are significant only at high
major factor limiting a distributed electrolytic hydrogen produc- currents (>1 A/cm2 ) and at relatively low efficiency, and have little
tion. In other words, we may be better off spending two times more technological implications [31]. The same seems to be true for the
energy per N m3 H2 if, at the same time, we halve the capital cost effects of ultrasounds and microwaves, which appear to have no
by doubling the power density. major effects on the overall efficiency.
Commercial PEM electrolyzers are available in smaller There is an incentive today to improve alkaline electrolysis over
sizes than alkaline electrolyzers, typically <10 N m3 H2 /h, with system scales larger than PEM electrolysis. This can be achieved in
Ecell ≈ 1.8 V at ∼1 A/cm2 . Substantially better performances (e.g. several ways:
1.6 V @1.1 A/cm2 ) have been reported for laboratory PEM cells at
80 ◦ C with about 2 mg/cm2 of precious catalysts Pt, Ru and Ir [21]. 1) Selection of better HER-OER catalysts; the advances reported
A lifetime of 5 or more years is considered to be achievable for PEM above lead to expectations that optimal non-PGM materials will
electrolyzers [22], whereupon lifetime certainly would be a func- be reliably identified; however, much development will still
tion of the operating temperature, resulting in increased efficiency be needed to control their morphologies and physicochemical
but decreased durability when temperature rises. Furthermore, properties, assess/improve their stability in alkali and set up high
increased catalyst loading would cause the cost of the device to volume production and preparation methods.
rise. Not only the price of the noble metals are high, but also system 2) Use system components and architectures that allow operation
complexity, difficulties in scaling up, and the high cost of the mem- at high temperatures and reduce system complexity, capital and
brane are considered to be the main disadvantages of PEM water operating costs.
electrolyzers. To our knowledge, the 2012 DoE goal of 400$/kW [23] 3) Use an anion exchange membrane (AEM) in place of the elec-
for the largest PEM electrolyzer underestimates current prices by trolyte, resulting in a compact system, similar to a PEM device
about one order of magnitude; laboratory-scale PEM electrolyzers and free of bubble effects, with a large choice of non-PGM cata-
come at about 15,000$/kW. lysts, with low resistance and high stability [32]. Alternatively,
The last column in Table 1 gives conservative estimates for adopt system architectures for liquid electrolyte devices which
the specifications of an advanced alkaline electrolyzer built with reduce bubble effects using a zero-gap configuration and/or gas
the gas diffusion electrodes we have been developing (see below). diffusion electrodes.
These figures can be substantially improved, in particular by incor-
porating an optimum combination of modified catalysts with the
In the following, we report on the selection of catalytic materials,
gas diffusion electrodes, whereby the performance gap between
mostly. Raney-Ni that have been found highly stable in alkali [33]
PEM and alkaline electrolyzers could be reduced or fully eliminated.
and investigate innovative system architectures and performances,
where gases are kept out of the intra-electrode space in order to
reduce bubble effects.
1.4. Strategies to improve water electrolysis

Efficiency and power density are the strong points of PEM elec- 2. Experimental
trolyzers. Their weak points are the price, and the limited global
availability of platinum group metals (PGM) which are used both at 2.1. Sample preparation and characterization
the cathode (Pt) and anode (Ir–Ru oxides), with or without surface
modifications [24–29]. The development of non-PGM catalysts for The electrode configuration consists of one or more layers,
PEM electrolysers and fuel cells is underway but no definite solu- which are prepared by mixing appropriate amounts of powders
tion is in sight. In particular, MoS3 in acid is a much less active HER with a binder (PTFE in suspension) and a volatile solvent (D70)
catalyst than Pt. Bio-inspired catalysts are likely to play a role in until a rubbery paste is obtained. The paste is flattened to layers
photo-electrochemical water splitting, but their use in electrolyz- a fraction of millimeter thick using a heavy rolling cylinder; the
ers with high power density will be more difficult. In short, it will layers are then pressed against a nickel net serving as the cur-
be a major challenge to maintain 1 A/cm2 current density with rent collector. Finally, the electrode assembly is heat-treated at
Ecell < 1.8 V in a PEM electrolyzer without PGM catalysts. 300 ◦ C for two hours under N2 at about one atmosphere. Assem-
The strength of conventional alkaline electrolyzers lies in the blies with the Ni net and up to three functional layers have been
long lifetime of the systems and their alternative non-PGM cata- produced: (i) the active layer is always present and consists of 5–10%
lysts, often in the form of Ni-based electrodes. On the other hand, PTFE, 70% Raney-Ni, 20% of a fine powder, either another catalyst
the main shortcoming of conventional alkaline electrolyzers is the and/or a conducting material; the typical catalyst loading is approx.
relatively low operating current density. This quantity is usually 50 mg/cm2 ; (ii) the gas diffusion layer consists of carbon black or Ni-
limited by bubble formation, which increases the electrolyte resis- powder bonded with ∼20% PTFE, (iii) a separator layer was made
tance and decreases the active area of the electrodes. A centrifugal with a ZrO2 powder (∼50 nm particle size) and a PTFE suspension,
S. Marini et al. / Electrochimica Acta 82 (2012) 384–391 387

Fig. 1. Different architectures of electrolyzers with: (a) immersion electrodes, (b) porous electrodes in the “zero-gap” configuration, (c) electrodes comprising Gas Diffusion
Layers (GDL) separating the gas compartments and the re-circulating electrolyte compartment (see text).

which bonded well with the active layer, occasionally replacing a type where the anode and cathode are on the two sides of a flat cur-
25 ␮m thick stretched polypropylene sheet (Celgard 2500) kindly rent collector to ease the serial connection among cells. H2 and O2
donated by Celgard (Charlotte, NC, USA). The Raney-Ni (A-2000, gas bubbles develop in separated electrolyte compartments, mostly
with 13%Fe, and A-7000, with 2%Mo) was supplied by Jonhson- between the electrodes where a gas impermeable membrane (the
Matthey. All electrochemical experiments have been carried out in “separator”) avoids mixing of hydrogen and oxygen. Bubbles reduce
a 28% KOH solution supplied by Carlo Erba (Milano, It). the active surface of the electrodes and contribute to the elec-
The electrode assemblies including the active layer and the Ni- trolyte resistance: Mostly because of this, high current densities
net and having both sides exposed to the electrolyte have been cannot be obtained but scaling up for large-scale volume H2 pro-
electrochemically characterized in cylindrical test cells made of duction is apparently easy. This configuration will not be further
plexiglas or PTFE with dimensions 3 cm in diameter and 5 cm in considered.
height, using a three electrode configuration. The working elec- The “zero-gap” configuration of an electrolyzer, shown in Fig. 1b,
trode (1 cm2 geometrical area) was spot-welded to a Ni-wire ideally excludes the bubbles from the intra-electrode space which
and fully surrounded by a large counter electrode made of Ni- is instead occupied, in our case, by a thin (<0.5 mm) cellulose
net (25 mesh). A Teflon Luggin capillary at the bottom of the felt, where the electrolyte is absorbed, confined and sandwiched
working electrode was connected through a plug in the wall of between two hydrophilic separators (Celgard membranes) tightly
the vessel with a reference electrode (commercial SCE or home pressed against anode and cathode. An anionic exchange mem-
made Hg/HgO in 28% KOH). A set of four cells in a thermal bath brane (AEM) [35] can be used in place of separators and confined
were simultaneously studied using a multichannel programmable electrolyte, making the system structurally identical to the cell of a
power supply in the current control mode with a multichannel PEM electrolyzer; the issues, in this case, are the resistivity and sta-
voltmeter and LABVIEW (National Instruments) software. Tafel bility of the AEM. Anode and cathode assemblies should be porous
plots have been obtained under different conditions and sur- and permeable to water or liquid electrolyte which thus reach the
face pre-treatments (e.g., polarity reversal, thermal cycles, slow “zero” gap between the electrodes. A full gas producing device may
charge–discharge and aging/stabilizing processes at a constant cur- have the electrolyte (or the water) circulating separately in the
rent density of ±100 mA/cm2 ). The “solution resistance” of each cell cathode and/or anode compartments as indicated in Fig. 1b, which
was estimated at room temperature and at several overpotentials may result in a heavy system requirements. However, in the lab-
with current interruption and/or electrochemical impedance spec- oratory practice, it is sufficient to put the entire epoxy-enclosed
troscopy (EIS) methods, using an Autolab PGSTAT-30 station (Echo assembly of the electrodes into a large, temperature-controlled
Chemie, NL). The data collected with the cylindrical test cells were KOH bath, which periodically is manually replenished or changed.
corrected for the solution resistance calculated at each tempera- The “zero-gap” configuration was used without an AEM membrane
ture from room temperature data and the known dependence of to estimate the performances of an electrolyser with minimal “solu-
the resistivity of 28% KOH with temperature [34]. tion resistance”.
Various compositions of the active catalysts both for the anode The scheme of electrolysis with gas diffusion electrodes is rep-
and cathode were characterized electrochemically in order to select resented in Fig. 1c. Each electrode assembly consists of a Ni-net,
the formulation yielding optimal performances and stability under gas diffusion layer (GDL), active layer, Celgard or ZrO2 separator
various operating conditions. Preliminary data have been pre- and is lodged in its own metal body (made of Ni or stainless steel)
sented elsewhere [33]. in good electrical contact with the Ni-net. The gas diffusion com-
partment is carved in this metal body along with ducts for gas
2.2. Architectures for electrolyzers and for purging the leaking electrolyte. The two metal bodies are
kept together by insulated screws and separated by a 1.5 mm thick
The configuration of a conventional alkaline electrolyzer is gasket, which defines the electrolyte compartment with the fac-
shown in Fig. 1a. The electrodes are non permeable and fully ing sides of anode and cathode assemblies. Temperature control is
immersed in the liquid electrolyte; they are often of the bipolar achieved with heating resistors applied to the metal bodies.
388 S. Marini et al. / Electrochimica Acta 82 (2012) 384–391

Fig. 2. Simplified schematics of pressure controls and electrolyte circulation system of the electrolysis cell with gas diffusion electrodes (see text).

Fig. 2 is a block diagram of the fluid circulation system of the full


electrolyzer, with gas diffusion electrodes shown as dashed lines.
The most critical part in the design of this system is the electrolyte
recirculation loop which, in our case, is achieved by controlling the
gas pressures over the source tank and the receiving tank, since no
peristaltic pump could meet the target range of pressures (up to
8 bar) and temperatures (up to 100 ◦ C). In normal operating condi-
tions, valve V1 is closed, V2 is open and the electrolyte flow through
the electrolyte compartment from the source tank to the receiving
tank at a rate determined by the difference PE-in − PE-out. When
the receiving tank is half full, a short KOH recovery cycle begins
automatically with V2 closed and the pressure over the recovery
tank ramped up above PE-in; V1 would then open until the recov-
ery tank is emptied. The KOH from the leakage tanks in the O2 and
H2 lines is recovered with a similar technique and the water con-
sumed in the electrolysis is periodically replaced in the source tank.
The corresponding circuits are not shown in Fig. 2.
Fig. 3. SEM image of a cathode with high concentration of Raney-Ni doped with Mo.
3. Results and discussion

3.1. Characterization of the active layers

The composition of one of the best cathodes was: 70% Mo-doped


Raney Ni (A7000), 10% MoO3 , 5% Cu, 5% graphite and 10% PTFE. It is
known that 1%, or less, of Mo in Raney Ni improves stability and HER
activity of this catalyst, and that even coarse mixtures of Ni alloys
and molybdenites may perform better than the individual catalysts
in HER [4]. A SEM image of this electrode after weeks of operation
is shown in Fig. 3; the micron-size crystals of the original MoO3
powder and the finer grains of Cu and graphite can be seen on or in-
between the larger (<30 ␮m) Raney-Ni particles; the PTFE binder is
not seen here since it eventually dissolves under the electron beam.
One of the best anodes (“L” composition) contained a mixture
of 70% Fe-doped Raney-Ni (A2000), 10% Fe-carbonyl, 10% Ni-210,
10% PTFE. Fig. 4 shows a SEM image of a large grain of Raney Ni
surrounded by sub-micron spheres of Fe-carbonyl along with the
wormlike motifs of Ni-210; some PTFE “glue” can be seen in the
top right corner of the picture. The pores of Raney-Ni cannot be
seen since they are in the 2–20 nm range. Also the morphology
of this electrode does not change appreciably upon months-long Fig. 4. SEM image of an anode with high concentration of Raney-Ni doped with Fe.

operation.
S. Marini et al. / Electrochimica Acta 82 (2012) 384–391 389

Fig. 5. Polarization behavior of cathode “J” at 25 ◦ C and at 60 ◦ C after treatment at Fig. 6. Polarization behavior of an anode at 25 ◦ C and at 60 ◦ C after activation at
−100 mA/cm2 for various times. 100 mA/cm2 followed by 2 h of polarity inversion.

Fig. 5 shows typical polarization data for one of the best cathodic (100 mA/cm2 ) and high (1 A/cm2 ) current densities recorded with
formulations (cathode “J”) collected with the cylindrical test cells the cylindrical test cells. Since the contributions of the solution
described in Section 2.1. The electrode was continuously kept at resistance have been subtracted, the difference between the anodic
−100 mA/cm2 except when the polarization curves were taken and cathodic curves represents the cell potential of an electrolyzer
by reading the stable voltages at different current steps. The first without ohmic losses from the electrolyte and separators. Relative
polarization curve at 25 ◦ C was recorded after few hours of polar- to the “zero gap” and GDL electrolyzers, diffusion losses and bub-
ization during which the overpotential decreased and apparently bles effects upon the active area are smaller in the cylindrical test
stabilized. This behavior is typical of all Raney-Ni cathodes and cell due to the direct contact of both sides of the active layer with
it is likely due to the partial reduction of the oxides/hydroxides the electrolyte. At 100 mA/cm2 , the cell potential shows 1.55 V and
layers formed at the surface of the grains when, during prepa- 1.45 V at 60 ◦ C and 84 ◦ C, respectively.
ration, the electrode is first exposed to air and then immersed Fig. 8 shows the long time behavior of a “zero-gap” 10 cm2 cell
in the KOH solution. When the temperature was raised to 60 ◦ C (Fig. 1b) made using the electrode formulations of Figs. 5 and 6,
while maintaining the current density at −100 mA/cm2 , the over- separated by a 0.5 mm thick cellulose felt inserted between two
potential dropped by ∼50 mV in less than an hour and then slowly Celgard separators. The cell potential at 100 mA/cm2 was above 2 V
lost another ∼50 mV in a couple of days to attain a stable value. and rather unstable in the first weeks of operation at room tem-
The average Tafel slopes in the 10–100 mA/cm2 interval showed perature because of incomplete wettability of the separators and
unusually low values (∼30 mV/dec) at 60 ◦ C while at 25 ◦ C were felt. The expected voltage of ∼1.8 V, close to the value of Fig. 7,
∼70 mV/dec. The decrease of the Tafel slope at low current densities was attained after 30 days of continuous operation. The tempera-
with increasing temperature is a common feature of all the inves- ture was then raised to 42 ◦ C, and the voltage stayed mostly below
tigated cathodes. Several treatment protocols included polarity 1.75 V for about 5 weeks. The electrolyte and the T-controller were
inversion and/or current interruption stages of few hours; in some changed after 83 days. Relative to the data of Fig. 7, the “zero-gap”
cathodes the recovery after re-starting electrolysis at −100 mA/cm2 cell at 42 ◦ C and 100 mA/cm2 has a voltage ∼100 mV higher, which
was very slow, requiring many hours, and in some cases was found is due to the ohmic resistance of the “zero-gap”, to imperfect wet-
to be only partial. ting of the active layers, and to larger gas bubbles effects. The initial
Fig. 6 presents selected polarization curves for one of the best
anodes (“L”) obtained with the cylindrical test cell. Again, the
overpotential at 100 mA/cm2 decreased by ∼50 mV when the tem-
perature was raised from 25 ◦ C to 60 ◦ C. In few hours of operation at
100 mA/cm2 and 60 ◦ C a modest increase of the overvoltage could
be observed; in the few hours after the first polarity inversion a
significant but temporary improvement was achieved. After polar-
ity inversion, the average Tafel slope in the current density range
10–100 mA/cm2 was about 55 mV/dec.
For all the anodes and cathodes consisting of an active layer on
a Ni-net, prolonged operation in the cylindrical test cell at high
current densities and repeated polarity inversions caused visible
damages with some loss of catalytic material. A more robust elec-
trode structure could be obtained by: (i) painting the active layer
with a diluted PTFE solution before heat-treatment, (ii) increas-
ing the PTFE content and (iii) incorporating the gas diffusion layer.
However, due to the change in the conductivity of the electrode, a
slight increase of the overvoltage at 100 mA/cm2 was noticed in all
these cases.
Fig. 7 shows the variation of the cathodic (cathode “J”) and Fig. 7. Voltage vs temperature at 1 A/cm2 and 100 mA/cm2 for an anode and a
anodic (anode “N”) voltages with temperature at intermediate cathode in test cells.
390 S. Marini et al. / Electrochimica Acta 82 (2012) 384–391

Fig. 8. Performance of Ecell for a “zero-gap” cell with time at 100 mA/cm2 .

Fig. 10. Cell voltages at different temperatures for an electrolyzer with 10 cm2
instability and the substantial fluctuations of the cell potential sug- gas diffusion electrodes and with an overpressure on the electrolyte compartment
pin = 2 bar.
gest that few large bubbles form between the separators and active
layers which persist for hours/days. The nature of the electrode
surfaces as well as diffusion and convection of dissolved gases are resistance of electrolyte and separators of configuration 1c can be
primary factors affecting the transfer of evolved gas to the bulk of estimated at ∼0.8  cm2 .
the solution or the hydrophilic separator [36]. The data of Fig. 8 and Fig. 10 shows cell voltages at different temperatures and at
our further tests suggest that these electrodes will deliver stable an overpressure on the electrolyte compartment of 2 bars, i.e.
performances in years of operation. the highest value tolerated by the gas diffusion electrodes with-
Fig. 9 shows the important role of an overpressure applied on out excessive KOH leaking. From Figs. 5 and 6, a cell voltage of
the electrolyte compartment to “squeeze out” the bubbles which 1.52 V at 300 mA/cm2 and 60 ◦ C would be predicted. By adding an
may form between the active layers and the separators. For this ohmic resistance contribution of 0.25 V (0.3 A × 0.80  cm2 ), a cell
experiment, the cell configuration in Fig. 1c with Celgard sepa- potential of about 1.8 V would therefore be expected instead of the
rators and gas diffusion electrodes has the same active layers as observed 2.05 V (see Fig. 10). The discrepancy is due to mass trans-
the electrodes of Fig. 7. For overpressures of the electrolyte com- port limitations (both of the gases and of the electrolyte) across the
partment relative to the gas compartments increasing from 0.2 to electrode assemblies, which also cause the sharp upward trend of
0.75 bar we observe a monotonic reduction of the cell potential by the overpotentials apparent from the data of Fig. 10 at current den-
∼400 mV; a hysteretic behavior is also observed between decreas- sities above 200 mA/cm2 where mass transport provides the main
ing and increasing pressure cycles. Above 1 bar of overpressure the contribution to the observed overpotentials.
cell potential shows a plateau, achieving 1.75 V at 100 mA/cm2 ;
this value is about 80 mV above the datum at 45 ◦ C of Fig. 7. At
this temperature and current density, diffusion losses are negligi- 4. Conclusions
ble; therefore, the contribution to the overvoltage due to the ohmic
Practical alkaline water electrolyzer with gas diffusion elec-
trodes including inexpensive and abundant electrocatalysts have
been constructed and were demonstrated to deliver high perfor-
mances and long-term stability. First, the active layers of cathodes
and anodes were optimized by selecting the formulations which
yield stable and low overpotentials at 60 ◦ C and 300 mA/cm2 in
our cylindrical test cells. Then we have studied how these formu-
lations perform in a “zero gap” electrolyzer cell and in a scheme
with gas diffusion electrodes. The architecture of GDE is believed
to be suitable for industrial H2 production; it attains performances
in par with the “zero gap” configuration used in laboratory con-
texts, and has relatively simple system requirements. As it may
have been expected, losses at high currents are substantially higher
than the sum of anode and cathode overpotentials obtained in our
test cells after subtraction of the contribution of the “solution resis-
tance”. However, our results indicates that the specifications of an
“advanced alkaline electrolyzer” can be achieved with our electrode
assemblies and the scheme of Fig. 1c. Furthermore, we believe that
better performances may be obtained by

(1) partially removing mass transport limitations through the use


of a more hydrophilic separator layer intimately bonded to the
active layer;
Fig. 9. Behavior of the cell potential Ecell of an electrolyzer with gas diffusion elec-
trodes and re-circulating electrolyte as a function of the overpressure (pin ) on the (2) decreasing the “solution resistance” from 0.8 to 0.3  cm2 , or
electrolyte compartment. less, by using better separators and current collectors, a thinner
S. Marini et al. / Electrochimica Acta 82 (2012) 384–391 391

electrolyte compartment and additives which enhance charge [5] J. Greeley, T.F. Jaramillo, J. Bonde, I. Chorkendorff, J.K. Nørskov, Nature Materials
transport within the active layers; 5 (2006) 909.
[6] T.F. Jaramillo, K.P. Jørgensen, J. Bonde, J.H. Nielsen, S. Horch, I. Chorkendorff,
(3) using catalysts with proper combination of stability, activity, Science 317 (2007) 100.
roughness, wettability and conductivity; partial substitution of [7] H. Vrubel, D. Merki, X. Hu, Energy Environmental Science (2012),
Raney Ni with Co3 O4 (the “ideal” OER catalysts) in micron-size http://dx.doi.org/10.1039/c2ee02835b.
[8] J. Suntivich, K.J. May, H. Gasteiger, J.B. Goodenough, Y. Shao-Horn, Science 334
grains may produce a modest decrease of overpotentials at high (2011) 1383.
temperatures and currents; [9] I.C. Man, H-Y. Su, F. Calle-Vallejo, H.A. Hansen, J.I. Martinez, N.J. Inoglu, J. Kitchin,
(4) adding pore-formers during electrode fabrication in order to T.F. Jaramillo, J.K. Nørskov, J. Rossmeisl, ChemCatChem 3 (2011) 1159.
[10] A. Vojvodic, J.K. Nørskov, Science 334 (2011) 1355.
increase the porosity of the structure and volume of internal
[11] J. Rossmeisl, Z.-W. Qu, H. Zhu, G.-J. Kroes, J.K. Nørskov, Journal of Electroana-
channels for gas transport. lytical Chemistry 607 (2007) 83.
[12] J. Rossmeisl, K. Dimitriesvski, P. Siegabahn, J.K. Nørskov, Journal of Physical
Chemistry C (2007) 18821.
An advanced alkaline electrolyzer with gas diffusion electrodes
[13] M.T.M. Koper, Journal of Electroanalytical Chemistry 660 (2011) 254.
and recirculating electrolyte will be more complex than a PEM [14] B.S. Yeo, A.T. Bell, Journal of the American Chemical Society 133 (2011) 5587.
electrolyzer, but its capital cost per kW and for any plant size is [15] J.S. Spendelow, A. Wieckowski, Physical Chemistry Chemical Physics 9 (2007)
potentially lower than a PEM system. With a catalyst loading of the 2654].
[16] A. Minguzzi, F-R.F. Fan, A. Vertova, S. Rondinini, A.J. Bard, Chemical Science 3
order of 50 mg/cm2 , our system requires about 100 g of catalyst per (2012) 217.
kW bringing the cost of our main catalyst (Raney Ni) below 10$/kW. [17] Y. Lee, J. Suntivich, K.J. May, E.E. Perry, Y. Shao Horn, Journal of Physical Chem-
For any plant size, manufacturing costs, rather than parts and mate- istry Letters 3 (2012) 399.
[18] J. Ivy, Summary of electrolytic hydrogen production, Milestone Comple-
rials, will be the main contributor to the capital cost. Furthermore, tion Report, NREL/MP-560-36734 (2004), http://www.nrel.gov/hydrogen/
large advanced electrolyzers will likely consist of identical modules pdfs/36734.pdf.
in parallel, which benefit maintenance and operating flexibility but [19] B.V. Tilak, P.W.T. Lu, J.E. Colman, S. Srinivasan, Electrolytic production of Hydro-
gen, in: J.O’M Bockris, B.E. Conway, E. Yeager, R.E. White (Eds.), Comprehensive
yield little saving when scaling-up the plant size. On the contrary, Treatment of Electrochemistry, vol. 2, Plenum Press, New York, 1981, p. 1.
traditional alkaline electrolyzers have capital and operating costs [20] http://www1.eere.energy.gov/hydrogenandfuelcells/mypp/pdfs/production.
which are strongly size-dependent, and drop below the energy cost pdf.
[21] W. Xu, K. Scott, International Journal of Hydrogen Energy 35 (2010) 12029.
for the biggest commercial systems. Many technical advances are
[22] P. Millet, D. Dragoe, S. Grigoriev, V. Fateev, C. Etievant, International Journal of
still needed to make this happen with advanced electrolyzers, but Hydrogen Energy 34 (2009) 4974.
no major stumbling block is perceived. [23] http://www.hydrogen.energy.gov/pdfs/hydrogen posture plan dec06.pdf.
[24] H. Colell, N. Alonso vante, C. Fischer, P. Bogdanoff, S. Fiechter, H. Tributsch,
Electrochimica Acta 39 (1994) 1607.
Acknowledgment [25] J. Cheng, H. Zhang, H. Ma, H. Zhong, Y. Zou, Electrochimica Acta 55 (2010) 1855.
[26] C. Baruffaldi, S. Cattarin, M. Musiani, Electrochimica Acta 48 (2003) 3921.
[27] E. Tsuji, A. Imanishi, K.-I. Fukui, Y. Nakato, Electrochimica Acta 56 (2011) 2009.
Research partly supported by CARIPLO grant 2010-0545 and a [28] B. Børresen, G. Hagen, R. Tunold, Electrochimica Acta (2002) 1819.
PRIN 2009 project. [29] T. Pauporté, F. Andolfatto, R. Durand, Electrochimica Acta 45 (1999) 431.
[30] H. Cheng, K. Scott, C. Ramshaw, Journal of the Electrochemical Society 149
(2002) D172.
References [31] H. Matsushima, D. Kiuchi, Y. Fukunaka, Electrochimica Acta 54 (2009) 5858.
[32] X. Wu, K. Scott, Journal of Materials Chemistry 21 (2011) 12344.
[1] S. Trasatti, Journal of Electroanalytical Chemistry 39 (1972) 163. [33] P. Salvi, P. Nelli, M. Viros, G. Zangari, G. Bruni, A. Marini, C. Milanese, Interna-
[2] W. Sheng, H.A. Gasteiger, Y. Shao-Horn, Journal of the Electrochemical Society tional Journal of Hydrogen Energy 36 (2011) 7816.
157 (2010) B1529. [34] R.J. Gilliam, J.W. Graydon, D.W. Kirk, S.J. Thorpe, International Journal of Hydro-
[3] R. Subbaraman, D. Tripkovic, D. Strmcnik, K-C. Chang, M. Uchimura, A.P. gen Energy 32 (2007) 359.
Paulikas, V. Stamenkovic, N.M. Markovic, Science 334 (2011) 1256. [35] X. Li, F.C. Walsh, D. Pletcher, Physical Chemistry Chemical Physics 13 (2011)
[4] A. Lasia, Hydrogen evolution reaction, in: W. Vielstich, A. Lamm, H.A. Gasteiger 1162.
(Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, [36] L.J.J. Janssen, C.W.M. Sillen, E. Barendrecht, S.J.D. van Stralen, Electrochimica
vol. 2, John Wiley & Sons, Ltd., Chichester, 2003, pp. 416–444. Acta 29 (1984) 633.

You might also like