Journal of Power Sources 497 (2021) 229864

Contents lists available at ScienceDirect

Journal of Power Sources

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

Voltage losses in zero-gap alkaline water electrolysis

J.W. Haverkort ∗, H. Rajaei
Process & Energy Department, Delft University of Technology, Leeghwaterstraat 39, 2628 CB Delft, The Netherlands

HIGHLIGHTS GRAPHICAL ABSTRACT

• Bubble and concentration overpotential

quantification for a zero-gap electrolyzer.
• An inactive electrode front explains the
anomalously large separator resistance.
• An additional ohmic drop arises tran-
siently, likely due to gas bubbles.
• Introducing a 0.2 mm gap strongly re-
duces the resistance.
• Local hydroxide depletion gives large
losses at low electrolyte concentrations.

ARTICLE INFO ABSTRACT

Keywords: Reducing the gap between the electrodes and diaphragm to zero is an often adopted strategy to reduce the
Alkaline water electrolysis ohmic drop in alkaline water electrolyzers for hydrogen production. We provide a thorough account of the
Zero-gap current–voltage relationship in such a zero-gap configuration over a wide range of electrolyte concentrations
Bubble Overpotential
and current densities. Included are voltage components that are not often experimentally quantified like those
Concentration overpotential
due to bubbles, hydroxide depletion, and dissolved hydrogen and oxygen. As is commonly found for zero-gap
Zirfon PERL
configurations, the ohmic resistance was substantially larger than that of the separator. We find that this is
because the relatively flat electrode area facing the diaphragm was not active, likely due to separator pore
blockage by gas, the electrode itself, and or solid deposits. Over an e-folding time-scale of ten seconds, an
additional ohmic drop was found to arise, likely due to gas bubbles in the electrode holes. For electrolyte
concentrations below 0.5 M, an overpotential was observed, associated with local depletion of hydroxide at
the anode. Finally, a high supersaturation of hydrogen and oxygen was found to significantly increase the
equilibrium potential at elevated current densities. Most of these voltage losses are shown to be easily avoidable
by introducing a small 0.2 mm gap, greatly improving the performance compared to zero-gap.

1. Introduction densities attractive. However, high current densities also make ohmic
losses relatively more important.
Modern alkaline water electrolyzers for hydrogen production are A way to reduce ohmic losses is to position the electrodes directly
designed for ever higher current densities. The cost of electrodes, adjacent to the separator in a so-called zero-gap configuration, inspired
diaphragms or membranes, and bipolar plates scale with the geomet- by fuel cell technology [1], see Fig. 1. It was first applied in a com-
rical electrode area, while the amount of hydrogen that is produced mercial electrolyzer by Zdansky–Lonza in the 1950s in a design that
is proportional to the current. This makes operating at high current was later taken over by Lurgi [2]. Since then, most manufacturers

∗ Corresponding author.
E-mail address: [email protected] (J.W. Haverkort).
URL: https://jwhaverkort.weblog.tudelft.nl/ (J.W. Haverkort).

https://doi.org/10.1016/j.jpowsour.2021.229864
Received 29 January 2021; Received in revised form 26 March 2021; Accepted 29 March 2021
Available online 9 April 2021
0378-7753/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
J.W. Haverkort and H. Rajaei Journal of Power Sources 497 (2021) 229864

Besides increasing the ohmic losses, trapped gas can also cause
Nomenclature corrosion deposits [13], increased separator deterioration through hot-
𝐴𝑅 Areal electronic resistance [Ωm2 ] spots [14], and voltage fluctuations [15]. The zero-gap configura-
𝑏𝑖 Tafel slope 𝑅𝑇 ∕𝛼𝑖 𝐹 [V] tion has been associated with increased separator damage and gas
cross-over [16].
𝑐 Electrolyte concentration [M]
Developed in the 1990s to replace asbestos diaphragms, a popular
𝐸 Voltage [V]
commercially available separator is Zirfon PERL produced by Agfa [17–
𝐸0 Onset potential in Eq. (7) [V] 21]. Owing to its polysulfone backbone, it is strong, and it has a
𝑗 Current density [A/m2 ] relatively low-cost [22]. The hydrophilic ZrO2 in Zirfon PERL improves
𝑗∗ Superficial exchange current density the wettability [14,17,23,24]. However, Ref. [23] finds a rather low
[A/m2 ] pore interconnectivity, bubble-point pressure, and wettability. This
𝑗eq Characteristic current density in Eq. (7) may cause incomplete wetting and perhaps even allow gas to enter
[A/m2 ] the separator and increase the resistance [5,25,26]. Relatively high
𝑙 Effective ohmic thickness in Eq. (4) [m] overpotentials, using Zirfon PERL in a zero-gap configuration, were
𝑙𝑠 Separator thickness ≈ (5 ± 0.5) ⋅ 10−4 [m] attributed to gas blanketing [26]. Similarly, use of the hydrophilic sep-
𝑇 Temperature ≈ 300 [K] arator material Celgard 2500 in a zero-gap configuration gave rise to an
anomalously high resistive overpotential of about 100 mV at 103 A/m2 ,
𝑡 Time [s]
attributed to ‘‘the ohmic resistance of the ‘‘zero-gap’’, to imperfect wetting
Constants of the active layers, and to larger gas bubbles effects’’ [25]. Application
of a modest overpressure of the order of 1 bar was found to strongly
𝐹 Faraday constant 96485.3329 [C/mol]
decrease these losses. [25,27]
𝑅 Gas constant 8.31446 [J/mol/K] Various papers consider the effect of gas bubbles in the traditional
Greek variables electrolyzer configuration with an electrode-separator gap; including
various reviews [28–31], experimental studies [32,33], and theoretical
𝛼 Charge transfer coefficient [–] analyses [33–35], mostly focusing on the additional ohmic drop intro-
𝜖 Separator porosity ≈ 0.5 ± 0.1 [–] duced by bubbles [14,34,36,37]. The additional overpotential due to
𝜖𝑒 Electrode open-hole fraction ≈ 1∕3 [–] bubbles reducing the reaction area was found to be relatively small, less
𝜂 Activation overpotential [V] than 5 mV at 103 A/m2 [38] on a small vertical cylindrical electrode in
𝜅 Electrolyte conductivity [S/m] an acidic electrolyte.
In the present work we add to the very little quantitative results that
𝜏 Separator tortuosity [–]
can be found on the effect of gas bubbles in zero-gap systems. We find
Subscripts that gas formation leads to a sizeable additional ohmic drop that builds
up transiently over the order of ten seconds. An often overlooked volt-
𝑎 Anode age loss that we also consider is the increase in equilibrium potential
𝑏 Bubbles due to dissolved hydrogen and oxygen. Under alkaline conditions the
𝑐 Cathode following redox reactions take place
𝑒 Electrodes
4H2 O + 4𝑒− → 4OH− + 2H2 , (cathode) (1)
𝑖 Index 𝑖 = 𝑎 or 𝑐 for anode or cathode
𝑡 Transient 4OH− → 2H2 O + O2 + 4𝑒− (anode) . (2)
eq Equilibrium Although the product gases have a very low saturation concentra-
lim Limiting, see Eq. (9) tion, especially at high electrolyte concentrations, they can be present
at very large supersaturations. Refs. [39,40] find hydrogen concen-
trations up to 120 mM, well over a hundred times the solubility.
The associated potential increases approximately logarithmically with
have adopted this strategy [3]. It is generally assumed that a zero gap increasing current density. This is explained by dissolved hydrogen
minimizes the ohmic losses [4]. However, the ohmic resistance in zero- that is removed from the surface by mass transfer, giving a so-called
gap configurations was consistently found to be much higher than that ‘gas evolution efficiency’ smaller than one [41,42]. For current den-
of the separator, awaiting a generally accepted explanation [5]. sities above 103 A/m2 , potential increases up to 70 mV have been
There has been some discussion in the literature about where gases measured [39,40].
evolve in a zero-gap configuration. Ref. [6] uses partially PTFE-covered There are several ways by which the transport of dissolved and
electrodes to conclude that the majority of the gas evolution occurs gaseous products away from the electrodes can be increased, most
at the electrode area facing the separator. Possibly, gas bubbles could obviously and effectively using flow. This may be flow-through the
escape here by capillary action through the fibers of the felt separator electrode, normal to the current [43], parallel to the current [44] or
and or a small unintentional gap. In contrast, Ref. [7] states that the flow-by, where slug flow was found to increase mass transfer [45].
electrode area facing the diaphragm will be largely inactive due to Promising alternatives include superaerophobic or superhydrophilic
coverage by a gas film. A schematic illustration of the resulting current electrodes [46], partially ‘liquid-free’ or ‘single irriguous’ operation [17,
lines is provided in, for example, Refs. [6,8] and Fig. 1. Ref. [9] stresses 47] or ‘bubble-free’ operation using hydrophobic diffusion layers [25,
the importance of a good wettability of the diaphragm to avoid this 48–50]. Somewhat more experimental ideas include the use of pressure
so-called gas blanketing effect. swings [51], ultrasound [52], or magnetic fields [53].
To avoid corrosion deposits and blanketing, Ref. [10] introduces A final, rarely considered, loss that we include is that due to hydrox-
filaments that are aligned with the upwards bubble movement and ide depletion, which is a reactant in Eq. (2) at the anode. Recent studies
have a preferred thickness of around 0.2 mm. Refs. [6,11] mention that showed that, using a separated anolyte and catholyte, global depletion
electrode openings of similar magnitude are required to avoid trapping of hydroxide of the anolyte can occur, leading to a limiting current
of hydrogen bubbles, and larger ones for oxygen bubbles. In Ref. [12] density [54,55]. Here we show that at low electrolyte concentrations
a gap of at least this magnitude is used in testing various diaphragms there can also be local depletion, giving rise to very large concentration
and membranes. overpotentials.

2
J.W. Haverkort and H. Rajaei Journal of Power Sources 497 (2021) 229864

Fig. 1. An idealized schematic of possible low current density streamlines through the
separator (light blue) between two electrodes (dark gray) at 𝑡 = 0 immediately after
the current is switched on and no bubbles (light gray) have yet formed (left) compared
to some time later where the current lines have to go around the generated bubbles
(right). Disclaimers: anode and cathode holes will not generally be this aligned; at
higher current densities and in 3D the streamlines can stay closer to the separator; gas Fig. 2. The cell configuration showing the white separator, black electrodes, and
formation between the separator and diaphragm is speculative. transparent PMMA layers. Shown is the ‘almost-zero-gap’ configuration with two
0.2 mm thick nylon spacers visible between the electrodes and the separator. Squares,
the size of the electrodes, have been cut out of the spacers. The blue copper wire,
together with a similar one at the other side, was used as a pseudo-reference electrode,
Simple semi-empirical fitting models are often used to predict the and to measure the electrolyte potential drop over the separator. The power supply and
behavior of laboratory cells [37,56] as well as advanced water elec- voltmeters are connected with separate crocodile clamps to the parts of the electrodes
sticking out from the top.
trolyzers [57,58], so-called for their zero-gap, active electrodes, ele-
vated temperatures and often elevated pressures. Here we extend such
empirical relations to hold over a wide range of electrolyte concen-
trations and current densities by including the effect of bubbles, dis- of current density. No significant differences between the ascending
solved gas products, and concentration polarization due to hydroxide and descending runs were found in the cell voltage. The cell voltage
depletion. was recorded every second using a Madgetech RFVolt2000 A digital
voltmeter with an accuracy of 0.05%, using separate clamps directly
2. Materials and methods attached the electrodes. Although the cell voltage became stable well
within a minute, for the below plots we used the value after 100 s, using
Fig. 2 shows the used cell configuration. A sub-microporous poly- the average of the values obtained during the ascending and descending
sulfone-supported ZrO2 -based separator (Zirfon PERL, Agfa-Gevaert runs. To further test the reproducibility we performed an additional run
NV) with a thickness of 500 ± 50 μm was used, with a porosity of at a different occasion and location, in descending current density or-
0.5±0.1 [59]. This diaphragm was sandwiched between two equal der, using a battery cycling system (Biologic, BCS-815, < 0.7% accuracy
0.5 mm thick expanded metal electrodes of 10 cm2 (Permascand AB) in current).
with main active components Ruthenium and Nickel oxides [27,60,61]
and containing eye-shaped holes of approximately 3 × 1 mm. These
layers were bolted together between two 6 mm thick PMMA plates. 3. Model
The assembly was immersed in a beaker of 13 cm diameter, filled
with electrolyte to a height of about 8 cm to cover the electrode and We introduce the following semi-empirical expression to model the
separator area. ( √ )𝑛 cell voltage
We used conductivities 𝜅𝑛 = 1.445 2 S/m from 𝜅1 = 2.04 S/m 𝑗𝑙
to 𝜅11 = 65.4 S/m, by varying the potassium hydroxide (KOH) concen- 𝐸cell = 𝐸eq + + 𝐴𝑅𝑗 + 𝜂. (4)
𝜅
tration from 𝑐 = 0.08 M to 6 M using the formula [62]
Here 𝐸eq is the equilibrium cell voltage obtained immediately after
𝑐
𝜅 = − 204𝑐 − 0.28𝑐 2 + 0.5332𝑐𝑇 + 20720 switching off the current density 𝑗. The electrolyte losses are described
𝑇 by Ohm’s law using an effective length-scale 𝑙. Additional linear losses
+ 0.1043𝑐 3 − 3 ⋅ 10−5 𝑐 2 𝑇 2 , (3) that do not scale inversely proportional to the electrolyte conductivity
with 𝜅 in S/m, 𝑐 in M, inserting 𝑇 = 300 K. We note that the original 𝜅 are included using an areal resistance 𝐴𝑅. Finally, the activation
Ref. [63] has −204.1𝑐 as the first term. overpotential 𝜂 = 𝜂𝑐 + 𝜂𝑎 + 𝜂lim where, with 𝑖 = 𝑐 and 𝑎,
The anodic and cathodic Tafel slopes and exchange current densities ( )
𝑗
were determined from an experiment using a 6 M electrolyte concen- 𝜂𝑖 = 𝑏𝑖 asinh . (5)
2𝑗∗𝑖
tration, measuring the potential difference between the electrodes and
that of a copper wire, inserted into the adjacent electrolyte as shown Here the Tafel slope on a natural logarithm basis reads 𝑏𝑖 = 𝑅𝑇 ∕𝛼𝑖 𝐹 ,
in Fig. 2. More details can be found in the supplementary data of with 𝛼𝑖 the charge transfer coefficient. The cathodic activation over-
Appendix A. potential 𝜂𝑐 is concentration independent. The anode adds, besides
A constant current was applied using a BK Precision 9151 with an a concentration-independent part 𝜂𝑎 , a concentration overpotential
accuracy of 0.1%+ 15 mA. The used current densities were distributed 𝜂lim . This part is due to the dependence of the anodic half-reaction,
√
from 𝑗 ≈ 10 A/m2 up to 𝑗 = 104 A/m2 , multiplying with 2 for Eq. (2), on the hydroxide concentration. For symmetric charge trans-
every subsequent value. Using the same geometric progression as the fer,(the anodic concentration-dependent
) Butler–Volmer equation 𝑗 =
𝜂𝑎 +𝜂lim 𝜂 +𝜂
conductivity allowed us to compare equal 𝑗∕𝜅 at different electrolyte 𝑐𝑎 − 𝑎 𝑏 lim
𝑗∗𝑎 𝑐
𝑒 𝑏𝑎 −𝑒 𝑎 gives for the total anodic activation losses
concentrations, without interpolation. Each current density was main-
tained for 120 s, followed by 300 s of zero current, after which a higher ( )
𝑗 𝑐𝑎
current was applied. The same series was repeated in descending order 𝜂𝑎 + 𝜂lim ≈ 𝑏𝑎 asinh . (6)
2𝑗∗𝑎 𝑐

3
J.W. Haverkort and H. Rajaei Journal of Power Sources 497 (2021) 229864

Fig. 4. The measured cell voltage increase for 𝑐 = 6 M (solid lines) can be fitted well
Fig. 3. The cell voltage, measured within a fraction of a second after the current has with Eq. (8) (dashed lines — approximately overlapping). The fitting parameters 𝑙𝑡 ≈ 1.1
been switched off. An average over all conductivities is provided, excluding data points mm and 𝑡𝑏 ≈ 10 s for the highest current density both only very slightly increase with
deviating more than 0.5 V from the dashed line, representing Eq. (7). The inset plot decreasing current density, see Appendix A.
gives the IR-corrected overpotentials measured relative to the copper wire visible in
Fig. 2, after subtracting the activation overpotentials. The dashed lines in the inset
𝑗
graph indicate ln 1+𝑗∕𝑗 times 𝑅𝑇 ∕𝐹 (60 mV/dec) and 𝑅𝑇 ∕2𝐹 (30 mV/dec) for anode
eq
and cathode, respectively. separator porosity 𝜖 ≈ 0.5 ± 0.1 and tortuosity 𝜏 ≈ 1.55 − 2.84 [18]
combine to give a much lower effective thickness 𝑙𝑠 𝜏∕𝜖 ≈ 1.5 − 3 mm.
Part of the ohmic drop, 𝑗𝑙𝑡 ∕𝜅 with 𝑙𝑡 ≈ 1 mm, was found to arise
transiently and will be discussed in the next section. The remaining
In writing Eq. (6) we assumed that the average hydroxide concentra-
𝑙−𝑙𝑡 = 3.6 mm corresponds to a tortuosity 𝜏 ≈ 3.6, or MacMullin number
tion at the anode surface, 𝑐𝑎 , tends to the average bulk electrolyte
𝜏∕𝜖 = 7.2, still above all experimental determinations of the separator
concentration 𝑐 for low overpotentials 𝜂𝑎 ≲ 𝑏𝑎 .
tortuosity.
Based on our experimental results, we introduce the following para-
This situation is typical for zero-gap configurations, as argued in
metrization for the equilibrium voltage
Ref. [5]. A conductivity of 𝜅 = 138 S/m at 80 ◦ C and 30 w% electrolyte
( )
7𝑅𝑇 𝑗 concentration would give an areal resistance of (𝑙 − 𝑙𝑡 )∕𝜅 = 0.26 Ωcm2
𝐸eq = 𝐸0 + ln , (7)
4𝐹 1 + 𝑗∕𝑗eq or 𝑙∕𝜅 = 0.33 Ωcm2 , both within the range of values listed there. In
Ref. [5] simulations are discussed in which the electrode area that
see Section 4.1 for its justification. In the next section we will com-
faces the separator was effectively made inactive by using a very high
pare all semi-empirical expressions introduced here, with experimental
gas fraction inside the gap. For gas fractions in the bulk ranging from
results.
0 to 0.6, areal resistances ranging from 0.194 to 0.361 Ωcm2 were
obtained. The measured ohmic drop can thus potentially be explained
4. Experiments
by a largely inactive electrode frontal area. To support this hypothesis,
we found that the ohmic losses were not significantly impacted when
4.1. Equilibrium voltage
a thin layer of epoxy glue (two-component epoxy resin and hardener,
Bison Kombi) was applied to the front of both anode and cathode.
The functional form of Eq. (7) is inspired by the Nernst equation for
Inactivity of the frontal electrode area without the epoxy layer present
the reactions (1) and (2) assuming ideal unity activity coefficients [56,
may arise, for example, due to blocking of the separator pores by a
64]. The prefactor 7∕4 derives from the sum of the stoichiometries
gas film, compression of the electrode, or reaction deposits [10]; a
1∕2 and 1∕4 of hydrogen and oxygen, respectively, plus an additional high local supersaturation of product gases in the poorly mixed region
∼ 𝑅𝑇 ∕𝐹 that is introduced to provide a better fit with the current- near the separator can further reduce the reactivity. However, the exact
interrupt data of Fig. 3. This Tafel slope was also found at low current origin requires further investigation.
densities 𝑗 < 𝑗∗𝑎 at the anode, as shown in the inset of Fig. 3. A
possible cause may be an anodic side-reaction in which Ruthenium 4.2.2. Bubble losses
oxide corrodes to give dissolved RuO2− 4
[65], see also Appendix A and Fig. 4 shows how, after the current is switched on, an additional
Section 5 for more information and discussion, respectively. voltage drop arise that is linear in the current density. This can be
Eq. (7) can be derived assuming that the concentration of dissolved attributed to an ohmic drop in the electrolyte, since it was also observed
products involved in the Nernst equation is proportional to the current between two copper wires placed on either side of the separator, see
density. With a constant mass transfer coefficient, dissolved oxygen Fig. 2 and Appendix A. The associated e-folding time scale 𝑡𝑏 ∼ 10 s is
and hydrogen concentrations that are much higher than the solubility smaller than the diffusive time-scale of the separator [55]. The diffusion
become proportional to the current density [66]. This can also be seen potential would also lead to a decrease in the cell voltage [55]. We note
from the measurements in Refs. [39,40] where the cathode potential that recently a similar potential increase was observed in proton ex-
1
was found to increase with 𝑅𝑇 2𝐹 ln 10
≈ 30 mV upon every decade of change membrane electrolysers [67,68]. Very recently, the time-rate of
increase in current. For 𝑗 ≳ 10 A/m2 the potential flattened, similar
3
change in the voltage of an alkaline electrolysis cell with a hydrophobic
to what can be seen in Fig. 3. We take this into account through the diffusion layer was used as a measure of bubble formation [69].
denominator 1 + 𝑗∕𝑗eq in Eq. (7). A time-scale of seconds correspond to that associated with the
growth and detachment of bubbles at lower current densities and
4.2. Ohmic losses the formation of a haze of small bubbles at higher current densities,
primarily at the cathode.
4.2.1. Electrolyte ohmic losses Inside the electrode openings, shielded from natural convection
The best fit of the ohmic term 𝑗𝑙∕𝜅 in Eq. (4) to the experimental flows, bubbles can accumulate for several seconds before being swept
data was obtained with 𝑙 = 4.6 mm. This is a surprisingly large value away, as shown in Fig. 5 and the accompanying videos in the sup-
given that the separator has a thickness of only 𝑙𝑠 = 0.5 mm. The plementary data of Appendix A. In Ref. [24], bubbles attached to the

4
J.W. Haverkort and H. Rajaei Journal of Power Sources 497 (2021) 229864

Table 1
Fitting parameters used in Eqs. (4)–(7), and (9). We assumed a constant temperature
of 𝑇 = 300 K throughout. The used electrodes have an open hole fraction of 𝜖𝑒 ≈ 1∕3
and a thickness 𝑙𝑒 ≈ 0.5 mm. The indicated Tafel slopes 𝑏𝑖 are on a natural logarithm
base. .
𝐸0 1.38 V 𝑙 4.6 mm
𝑗eq 200 A/m2 𝑙𝑡 1.1 mm
𝐴𝑅 17 μΩm2 𝑏lim 1 V
𝑗∗𝑐 800 A/m2 𝑙lim 0.8 mm
𝑗∗𝑎 200 A/m2 𝜅lim 1.63 S/m
𝑏𝑐 52 mV/e 𝑝 0.1
𝑏𝑎 40 mV/e 𝑞 1.77

𝑗 ≳ 2 ⋅ 103 A/m2 . Therefore, for the highest current densities, the rear
of the electrode can be considered largely inactive. Since we already
established that the front of the electrodes is also mostly inactive, we
find that in a zero-gap configuration the side walls of the electrode
openings will provide the majority of the reaction area. Fig. 5 and the
videos in the supplementary data of Appendix A do indeed seem to
show that most of the gas bubbles originate from these locations.
In Refs. [54,55] we actually already noted, but could not explain,
a transient increase in potential. We modeled the ohmic resistance by
adding to the electrode thickness roughly the electrode size on both
sides of the separator, using 𝜏 = 1.62 giving 𝑙 = (1.5 mm)𝜏∕𝜖 = 4.9
mm, slightly more than the 𝑙 = 4.6 mm we find here. In Ref. [70] we
discuss also the data from a repeat experiment that is best fitted with
𝑙 = 4 mm. These results shows that 𝑙 is not exactly reproducible between
experiments. It has been speculated that nano bubbles trapped inside
the separator may play a role in increasing the resistance [5,25,26].
However, such bubbles would also decrease the effective diffusion coef-
Fig. 5. Oxygen (top) and hydrogen (bottom) bubbles approximately 𝑡 = 3.7 s after the ficients inside the separator. To calculate the limiting current associated
current is switched on. After this time, primarily at the cathode, the visibility strongly with global hydroxide depletion, we needed no such correction in
deteriorates. An electrolyte concentration of 𝑐 = 6 M and a container with a distance of Refs. [54,55] and used an effective separator thickness (0.5 mm)𝜏∕𝜖 =
8 mm between the electrode and wall of the container was used for improved visibility. 1.62 mm, using 𝜏 = 1.62, with reasonable agreement. A much lower
Small bubbles seem to be preferentially generated at the rims, while larger ones grow
tortuosity could thus be used compared to that required to describe
at the top of the electrode holes. Bubbles on the leeward bottom left rim are released
while those at the top windward right rim tend to slide upwards. See Appendix A for the ohmic drop. This may be explained by an additional electrolyte
videos of each image. flux, in parallel to the one required for the current. Contrary to the
current density [55], such an electrolyte flux can be assisted by ad-
vective mixing due to bubbles [72] and the natural convection flows
diaphragm were found mostly near the top of the electrode openings they induce. As a result, inside the separator this additional flux may
where they grew by diffusion. The final bubble size was found to be more evenly distributed than the current density. This therefore
primarily depend on the separator material. provides some evidence that the additional resistance we find in a
Over the course of several seconds, the electrolyte became cloudy zero-gap configuration, both that almost immediately present and that
from a haze of small accumulating bubbles. From Fig. 5, this can arising transiently, arises external to the electrode.
be seen to happen primarily near the cathode. A simple differential
4.2.3. Electrode ohmic losses
equation for the gas fraction 𝜀 with a constant production term, and a
Subtracting from the measured cell voltage fits of all the other losses
removal term that is proportional to the gas fraction, reads along with
shows a remaining loss of about 0.17 V at 104 A/m2 that is approxi-
its solution
( ) mately linear in current density and independent of ionic conductivity,
𝑑𝜀 𝜀 −𝜀
= ∞ → 𝜀 = 𝜀∞ 1 − 𝑒−𝑡∕𝑡𝑏 . (8) see Appendix A. Bubble-associated increased activation losses cannot
𝑑𝑡 𝑡𝑏
explain these observations as they would require unrealistically high
This solution is shown to accurately fit the data in Fig. 4. The time-scale surface coverages [35].
𝑡𝑏 ≈ 10 s only weakly decreases with increasing current density, see Measuring the voltage drop over a narrow strip of electrode of
Fig. SI.2 in Appendix A. With a constant 𝜀∞ , this admittedly simplified known length and width, we measured a surprisingly high effective
model can quantitatively describe the transient ohmic losses shown in electrode resistivity of 𝜌𝑒 ≈ 2 ⋅ 10−6 Ωm. The electrode strip visible at
Fig. 4. Using the Bruggemans correction factor [70,71] for spherical the top of Fig. 2 has a width 𝑤 ≈ 1.5 cm and height ℎ ≈ 4.5 cm up to
( )1.5
bubbles, the measured 𝑙𝑡 ≈ 𝑙𝑡0 ∕𝜖𝑒 1 − 𝜀∞ ≈ 1 mm requires the ionic the position of the clamps. This gives a resistance 𝜌𝑒 ℎ∕𝑙𝑒 𝑤 ≈ 12 mΩ,
current to travel, for example, through a gas fraction of 𝜀∞ = 0.55 over or electrode-area-specific resistance 𝐴𝑅 ≈ 1.2 ⋅ 10−5 Ωm2 , which is only
a distance of 𝑙𝑡0 = 0.1 mm or through 𝜀∞ = 0.29 over a distance of slightly below the fitted value in Table 1. The small difference can be
𝑙𝑡0 = 0.2 mm. These values correspond well to, and may therefore be accounted for by additional electronic losses from the average distance
explained by, the typical diameter and surface coverage of a layer of that the current travels within the active part of the electrodes.
adhering bubbles [33]. Detached bubbles will further add to the ohmic
drop. 4.3. Activation overpotentials
Considering the electrode as a non-tortuous porous medium with
porosity 𝜖𝑒 ≈ 1∕3 and Tafel slope 𝑏, most of the current will be gener- 4.3.1. Kinetic overpotentials
ated within a distance 𝜅𝜖𝑒 𝑏∕𝑗 [70]. Even for the highest conductivity In Eqs. (4)–(6) we wrote the overpotential 𝜂 = 𝜂𝑐 +𝜂𝑎 +𝜂lim as the sum
used, this distance is already smaller than the electrode thickness for of cathodic and anodic activation overpotentials, where we separate

5
J.W. Haverkort and H. Rajaei Journal of Power Sources 497 (2021) 229864

Fig. 6. The activation overpotential 𝜂 = 𝐸cell −𝑗𝑙∕𝜅 −𝐴𝑅𝑗 for conductivities 𝜅 > 20 S/m
for which the concentration overpotential 𝜂lim is negligible. The inset plot shows the
𝐼𝑅-corrected overpotentials 𝜂𝑎 and 𝜂𝑐 relative to the copper wire shown in Fig. 2, after Fig. 7. The cell voltage recorded with the set-up of Fig. 2 for conductivities 𝜅1 = 2.04
𝑗
√
subtracting ln 1+𝑗∕𝑗 times 𝑅𝑇 ∕𝐹 (60 mV/dec) and 𝑅𝑇 ∕2𝐹 (30 mV/dec) for anode and S/m, 𝜅2 = 2𝜅1 = 2.89 S/m, etcetera. The dashed line gives 𝐸0 + 𝑗𝑙∕𝜅 using the para-
eq
cathode, respectively. meters of Table 1. The inset plot gives the hydroxide depletion concentration overpo-
tential 𝜂lim = 𝐸cell − 𝐸eq − 𝜂𝑎 − 𝜂𝑐 − 𝑗𝑙∕𝜅 − 𝐴𝑅𝑗, where the gray dashed lines give the
parametrization of Eqs. (6)–(9).

the latter into a concentration-independent activation overpotential 𝜂𝑎

and a concentration overpotential 𝜂lim due to local anodic hydroxide
4.3.3. Concentration overpotential
depletion.
( ) Fig. 7 shows the recorded steady-state cell voltages as a function
The total activation overpotential 𝜂 = 𝐸cell − 𝐸eq + 𝑗𝑙∕𝜅 + 𝐴𝑅𝑗 , of the ratio 𝑗∕𝜅. Most of the data collapses to a similar linear trend
obtained using Eq. (7) and the fitting parameters in Table 1, is shown corresponding to ohmic losses. Some of the spread arises because the
in Fig. 6. Only the highest four conductivities are displayed, for which activation overpotentials depend on current density 𝑗 rather than 𝑗∕𝜅.
𝜂lim is negligible. The lowest conductivities 𝜅 ≤ 𝜅6 or 𝑐 < 0.5 M show pronounced
Relative to a copper wire we measured 𝜂𝑐 and 𝜂𝑎 shown in the inset additional losses 𝜂lim that are shown in the inset plot and can be
plot of Fig. 2. A cathodic Tafel slope 𝑏𝑐 = 𝑅𝑇 ∕𝛼𝑐 𝐹 ≈ 52 mV/e was attributed to local hydroxide depletion. This data can be fit reasonably
found or, multiplying with ln 10 ≈ 2.3, 120 mV/dec. This corresponds well using Eq. (6) with
to a charge transfer coefficient of 𝛼 = 1∕2, suggesting a Volmer 𝑐𝑎 ( ( )𝑝𝑏 ∕𝑏 )−1∕𝑝
= 1 + 𝑗∕𝑗lim lim 𝑎 ,
rate-determining step. The cathodic effective exchange current density 𝑐
𝑗∗𝑐 ≈ 800 A/m2 is high, owing to the excellent catalytic properties for 2𝐹 𝐷− 𝑐 ( ( )𝑞 )
𝑗lim = 1 + 𝜅∕𝜅lim . (9)
the hydrogen evolution reaction of the used electrodes. The obtained 𝑙lim
anodic Tafel slope 𝑅𝑇 ∕𝛼𝑎 𝐹 = 40 mV/e, or 92 mV/dec, corresponds Here 𝑝, 𝑞, 𝑙lim , and 𝜅lim area given in Table 1 and 𝐷− ≈ 5.3 ⋅ 10−9 m2 /s
to an anodic charge transfer coefficient of 𝛼𝑎 ≈ 0.65. The anodic is the hydroxide ion diffusivity [77]. The factor two in Eq. (9) derives
exchange current density 𝑗∗𝑎 ≈ 200 A/m2 was found to be lower than from a contribution of migration to the current density that, in steady
the cathodic one. Note that the validity of using a copper pseudo- state, is equal to that of diffusion [54,55]. The current density 𝑗lim
reference electrode has not been established previously for the used here is a characteristic value for which the hydroxide concentration at
conditions. The activation overpotentials shown in the main Fig. 6 were the anode surface starts to deplete locally. It differs from the limiting
obtained from the cell voltage and therefore do not depend on, but current density in Refs. [54,55] where the hydroxide concentration
show good agreement with, the measurements using this reference. depleted in the entire anolyte. Overlimiting currents [78] beyond 𝑗lim
While this is encouraging, this is no thorough proof that copper can are possible and for 𝑗 ≫ 𝑗∗𝑎 , 𝑗lim Eqs. (5), (6), and (9) give 𝜂lim =
always be accurately and stably be used as a reference in alkaline water 𝑏lim ln 𝑗 𝑗 . This logarithmic increase is clearly visible in the inset plot
lim
electrolysis. of Fig. 7. The effective length-scale 𝑙lim ≈ 0.8 mm in Eq. (9) corresponds
to a distance 𝑙lim 𝜖𝑒 ≈ 0.3 mm, of the order of the electrode thickness.
As is further illustrated in Appendix A, the concentration polarization
4.3.2. Resistive effects 𝜂lim develops transiently over a diffusive time-scale of the order of a
Often much lower Tafel slopes are measured on Ruthenium ox- minute.
ides [73], although typically at very low current densities. Could the
relatively high values we find perhaps be a result of transport limita- 4.4. Almost-zero gap
tions? A doubling in current density leads to a halving of the current
penetration thickness and reactive surface area, leading to a doubling of Using the parameters of Table 1 in Eqs. (4)–(7) along with Eq. (9), a
the effective Tafel slope [70,74,75]. Two-dimensional simulations show maximum relative difference with the experimental data below 6% was
something similar to this one-dimensional theoretical expectation [76]. obtained, and below 2% for the highest four conductivities. We show
This would imply that we observe twice the intrinsic kinetic Tafel an example of such a fit for 𝑐 = 6 M in Fig. 8 in which also the various
terms in Eq. (4) are indicated.
slopes. However, in this case the effective exchange current density
√ We see that at the highest current density, more than half of the
is predicted to be proportional to 𝜅 [70,75], something that is not
overpotential comes from the separator ohmic losses 𝑗𝑙∕𝜅. To further
observed in Fig. 6. It is presently unclear why no Tafel slope doubling
investigate the partly responsible inactivity of the electrode area facing
can be observed.
the separator, we introduced a small gap of 0.2 mm between the
The front of the electrode is likely always within the current pen- electrodes and the separator. Such almost-zero gap configurations have
etration thickness 𝜅𝑏𝑖 ∕𝑗, but was shown to not significantly contribute also been referred to as near zero-gap [79,80] or nearly zero gap [12].
to the reactivity. Alternatively, the two-dimensional nature of the elec- We used nylon spacers, visible in Fig. 2, from which a hole with the
trodes and the spatial gas distribution invalidates the one-dimensional size of the electrodes was cut out. Fig. 8 shows the resulting polariza-
prediction of Tafel slope doubling. This warrants further investigation. tion curve, using a 𝑐 = 6 M electrolyte, along with that of the original

6
J.W. Haverkort and H. Rajaei Journal of Power Sources 497 (2021) 229864

𝑅𝑇 ∕𝐹 of the 7𝑅𝑇 ∕4𝐹 found with current-interrupt measurements, has

to be attributed to measurement inaccuracy.
The non-uniqueness of the obtained fitting parameters illustrates the
difficulty of untangling the various losses. Therefore, it is not surprising
that there is an almost complete lack in literature of measurements
of bubble-induced losses in a zero-gap configuration. We did not find
any evidence of increased activation losses due to bubbles covering the
surface, an effect that presumably is relatively small for the current
densities ≤ 104 A/m2 studied.
The ohmic resistance was found to be much larger than that of
the separator, consistent with the area between the electrode and the
separator being inactive. This was confirmed by an experiment in which
Fig. 8. The cell voltage recorded with the set-up of Fig. 2 showing that a gap of 0.2 mm
between the electrodes and the separator (yellow triangles) significantly reduces the
this part was covered with epoxy, giving a virtually indistinguishable
ohmic losses compared to the configuration without such a small gap (blue squares and polarization curve. This inactivity may be explained by blockage of the
black crosses). Covering with epoxy the area of the electrode that faces the separator separator pores by gas, mechanical forces, and or deposits, requiring
seems to slightly increase the activation losses at low current densities, but does not further investigation.
significantly alter the ohmic losses. The solid lines give the various losses described by
Eqs. (4)–(7) and Eq. (9) using the parameters of Table 1.
We also found a sizeable transient increase in the ohmic losses
that can plausibly be attributed to gas bubbles, likely mostly inside
the electrode holes, although we cannot with certainty exclude the
possibility that also some gas may enter the separator. The less than
zero-gap configuration. The shown fits use the same parameters except
perfect reproducibility of these losses can be attributed to the stochastic
that 𝑙 = 4.6 mm for the zero-gap configuration is replaced by 3 mm
nature of bubbles and local flow conditions.
for the almost-zero-gap configuration. This strong reduction in ohmic
A small 0.2 mm gap sufficed to remove a large part of the bubbles
drop can be partially explained by the almost absent transient losses,
losses, greatly improving upon the performance over a zero-gap config-
see Appendix A. The applicable 3 mm is however even below 𝑙−𝑙𝑡 = 3.6
uration. Further studies may show alternative means to avoid bubble
mm, despite the additional 2 × 0.2 mm = 0.4 mm gap. This can be
losses, like electrolyte flow, increased pressure, modified material wet-
explained by the more homogeneous current distribution that results
ting properties, and tailored electrode shapes. There are, however,
when the electrode front becomes active.
several additional advantages to the simple solution of a small gap,
Using 𝜅 = 138 S/m for 80 ◦ C and 30 w%, the area-specific resistance
including decreased separator damage, and likely strongly reduced gas
(3 mm) ∕𝜅 = 0.22 Ωcm2 is lower than all of the ‘zero-gap’ results listed
cross-over. We observed a small decrease in equilibrium potential in the
in Ref. [5], while simulations with a 0.2 mm gap and a high gas fraction
presence of a gap, which can be attributed to improved electrolyte ad-
of 0.6 showed a similar resistance of 0.19 Ωcm2 .
vection decreasing the dissolved gas concentrations. A lower dissolved
We note that a positive effect of a small gap has been suggested in
gas concentration at the separator surface will strongly decrease gas
the conclusions of Ref. [6] but was not quantified. Ref. [13] introduced cross-over through the separator, increasing the operational window of
a smaller gap < 0.15 mm, resulting in a much smaller decrease in losses electrolyzers.
then found here.
Upon introducing the small gap we also found a small, perhaps not
CRediT authorship contribution statement
significant, decrease of ∼ 10 mV in the equilibrium potential at low
current densities. This can be explained by enhanced advection that
J.W. Haverkort: Conceptualization, Methodology, Validation, For-
transports out dissolved products.
mal analysis, Investigation, Writing - original draft, Visualization, Su-
pervision, Project administration, Funding acquisition. H. Rajaei: Soft-
5. Conclusions and discussion ware, Validation, Formal analysis, Investigation, Writing - review &
editing. Visualization.
We provided detailed measurements of the cell voltage of a lab-scale
zero-gap alkaline water electrolysis cell over a wide range of current Declaration of competing interest
densities 10 A/m2 ≤ 𝑗 ≤ 104 A/m2 and electrolyte concentrations
0.08 M ≤ 𝑐 ≤ 6 M.
The authors declare that they have no known competing finan-
Using additional input from reference electrode and current- cial interests or personal relationships that could have appeared to
interrupt data, the cell voltage was split into equilibrium, activation, influence the work reported in this paper.
and ohmic components according to Eq. (4). The activation overpoten-
tials could be described well with the Butler–Volmer equation provided
Appendix A. Supplementary data
that, for concentrations 𝑐 ≲ 0.5 M, the concentration polarization due to
local anodic hydroxide depletion was taken into account. We provided
Supplementary material related to this article can be found online
a purely empirical equation (9) to describe these losses, which likely
at https://doi.org/10.1016/j.jpowsour.2021.229864. The following is
strongly depends on the specific electrode and separator properties, as
the supplementary data related to this article:
well as local flow conditions.
The equilibrium potential 𝐸eq was found to increase with increasing • Additional data and graphs
current density to above 1.6 V, as shown in Fig. 3. Apart from a high • All videos:
onset potential we found a high slope 𝜕𝐸eq ∕𝜕 ln 𝑗 ≈ 7𝑅𝑇 ∕4𝐹 ≈ 0.1
V/dec below 𝑗eq ≈ 200 A/m2 , of which only 3𝑅𝑇 ∕4𝐹 can be explained – Anode, 102 A/m2
by dissolved oxygen and hydrogen. At current densities well below the – Anode, 103 A/m2
anodic exchange current density 𝑗∗𝑎 ≈ 200 A/m2 an anodic Tafel slope – Anode, 104 A/m2
𝑅𝑇 ∕𝐹 was found that may be related to a parasitic side-reaction of the – Cathode, 102 A/m2
Ruthenium-based electrode. Since 𝑗eq ≈ 𝑗∗𝑎 , we cannot fully exclude – Cathode, 103 A/m4
the possibility that these losses are kinetic in nature. In this case, about – Cathode, 104 A/m2

7
J.W. Haverkort and H. Rajaei Journal of Power Sources 497 (2021) 229864

References [32] J. Eigeldinger, H. Vogt, The bubble coverage of gas-evolving electrodes in a

flowing electrolyte, Electrochim. Acta 45 (2000) 4449.
[1] R.L. Costa, P.G. Grimes, Electrolysis as a Source of Hydrogen and Oxygen, Tech. [33] R.J. Balzer, H. Vogt, Effect of electrolyte flow on the bubble coverage of vertical
Rep, Allis-Chalmers Mfg. Co., Milwaukee, 1967. gas-evolving electrodes, J. Electrochem. Soc. 150 (2002) E11.
[2] A.T. Kuhn, Industrial Electrochemical Processes, Elsevier Publishing Company, [34] N. Nagai, M. Takeuchi, T. Kimura, T. Oka, Existence of optimum space between
1971. electrodes on hydrogen production by water electrolysis, Int. J. Hydrogen Energy
[3] J. Divisek, Water electrolysis in a low and medium temperature regime, in: Elec- 28 (2003) 35.
trochemical Hydrogen Technologies-Electrochemical Production and Combustion [35] J. Dukovic, C.W. Tobias, The influence of attached bubbles on potential drop and
of Hydrogen, Elsevier, Oxford, 1990. current distribution at gas-evolving electrodes, J. Electrochem. Soc. 134 (1987)
[4] R. Phillips, A. Edwards, B. Rome, D.R. Jones, C.W. Dunnill, Minimising the ohmic 331.
resistance of an alkaline electrolysis cell through effective cell design, Int. J. [36] C.W. Tobias, Effect of gas evolution on current distribution and ohmic resistance
Hydrogen Energy 42 (2017) 23986. in electrolyzers, J. Electrochem. Soc. 106 (1959) 833.
[5] M.T. de Groot, A.W. Vreman, Ohmic resistance in zero gap alkaline electrolysis [37] J. Rodríguez, E. Amores, CFD modeling and experimental validation of an
with a Zirfon diaphragm, Electrochim. Acta (2020) 137684. alkaline water electrolysis cell for hydrogen production, Processes 8 (2020) 1634.
[6] V. Kienzlen, D. Haaf, W. Schnurnberger, Location of hydrogen gas evolution on [38] J.A. Leistra, P.J. Sides, Hyperpolarization of gas evolving electrodes—I. Aqueous
perforated plate electrodes in zero gap cells, Int. J. Hydrogen Energy 19 (1994) electrolysis, Electrochim. Acta 32 (1987) 1489.
[39] S. Shibata, Supersolubility of hydyogen in acidic solution in the vicinity of
729.
hydrogen-evolving platinum cathodes in different surface states, Denki Kagaku
[7] H. Wendt, Neue konstruktive und prozeßtechnische Konzepte für die
oyobi Kogyo Butsuri Kagaku 44 (1976) 709.
Wasserstoff-Gewinnung durch Elektrolyse, Chem. Ing. Tech. 56 (1984) 265.
[40] S. Shibata, The concentration of molecular hydrogen on the platinum cathode,
[8] H. Vogt, G. Kreysa, Electrochemical reactors, in: Ullmann’s Encyclopedia of
Bull. Chem. Soc. Japan 36 (1963) 53.
Industrial Chemistry, 2000.
[41] H. Vogt, The rate of gas evolution of electrodes—I. An estimate of the efficiency
[9] P. Gallone, L. Giuffré, G. Modica, Developments in separator technology for
of gas evolution from the supersaturation of electrolyte adjacent to a gas-evolving
electrochemical reactors, Electrochim. Acta 28 (1983) 1299.
electrode, Electrochim. Acta 29 (1984) 167.
[10] J. Divisek, P. Malinowski, Electrolyzer with sandwich arrangement of diaphragm
[42] L. Müller, M. Krenz, K. Rübner, On the relation between the transport of
and electrodes and method of producing the sandwich arrangement, 1988, US
electrochemically evolved Cl2 and H2 into the electrolyte bulk by convective
Patent 4, 773, 982.
diffusion and by gas bubbles, Electrochim. Acta 34 (1989) 305.
[11] J. Fischer, H. Hofmann, G. Luft, H. Wendt, Fundamental investigations and
[43] F. Yang, M.J. Kim, M. Brown, B.J. Wiley, Alkaline water electrolysis at 25 a
electrochemical engineering aspects concerning an advanced concept for alkaline
cm- 2 with a microfibrous flow-through electrode, Adv. Energy Mater. (2020)
water electrolysis, AIChE J. 26 (1980) 794.
2001174.
[12] J. Brauns, J. Schönebeck, M.R. Kraglund, D. Aili, J. Hnát, J. Žitka, W. Mues,
[44] H. Rajaei, A. Rajora, J.W. Haverkort, Design of membraneless gas-evolving
J.O. Jensen, K. Bouzek, T. Turek, Evaluation of diaphragms and membranes as
flow-through porous electrodes, J. Power Sources 491 (2021) 229364.
separators for alkaline water electrolysis, J. Electrochem. Soc..
[45] I. Dedigama, P. Angeli, K. Ayers, J.B. Robinson, P.R. Shearing, D. Tsaoulidis,
[13] J. Divisek, P. Malinowski, Diaphragm for alkaline electrolysis and process for
D.J.L. Brett, In situ diagnostic techniques for characterisation of polymer
manufacture of diaphragm, 1987, US Patent 4, 636, 291.
electrolyte membrane water electrolysers–flow visualisation and electrochemical
[14] M.J. Lavorante, C.Y. Reynoso, J.I. Franco, Water electrolysis with Zirfon® as
impedance spectroscopy, Int. J. Hydrogen Energy 39 (2014) 4468.
separator and NaOH as electrolyte, Desalin. Water Treat. 56 (2015) 3647.
[46] W. Xu, Z. Lu, X. Sun, L. Jiang, X. Duan, Superwetting electrodes for gas-involving
[15] A. Manabe, H. Domon, J. Kosaka, T. Hashimoto, T. Okajima, T. Ohsaka, Study on
electrocatalysis, Acc. Chem. Res. 51 (2018) 1590.
separator for alkaline water electrolysis, J. Electrochem. Soc. 163 (2016) F3139.
[47] S. Dutta, Technology assessment of advanced electrolytic hydrogen production,
[16] A. Manabe, M. Kashiwase, T. Hashimoto, T. Hayashida, A. Kato, K. Hirao, I.
Int. J. Hydrogen Energy 15 (1990) 379.
Shimomura, I. Nagashima, Basic study of alkaline water electrolysis, Electrochim.
[48] P. Tiwari, G. Tsekouras, K. Wagner, G.F. Swiegers, G.G. Wallace, A new class
Acta 100 (2013) 249.
of bubble-free water electrolyzer that is intrinsically highly efficient, Int. J.
[17] P. Vermeiren, W. Adriansens, J.P. Moreels, R. Leysen, Evaluation of the Zirfon®
Hydrogen Energy 44 (2019) 23568.
separator for use in alkaline water electrolysis and Ni-H2 batteries, Int. J. [49] M. Koj, J. Qian, T. Turek, Novel alkaline water electrolysis with nickel-iron gas
Hydrogen Energy 23 (1998) 321. diffusion electrode for oxygen evolution, Int. J. Hydrogen Energy 44 (2019)
[18] J. Rodríguez, S. Palmas, M. Sánchez-Molina, E. Amores, L. Mais, R. Cam- 29862.
pana, Simple and precise approach for determination of ohmic contribution of [50] O. Winther-Jensen, K. Chatjaroenporn, B. Winther-Jensen, D.R. MacFarlane,
diaphragms in alkaline water electrolysis, Membranes 9 (2019) 129. Towards hydrogen production using a breathable electrode structure to directly
[19] M. Schalenbach, W. Lueke, D. Stolten, Hydrogen diffusivity and electrolyte separate gases in the water splitting reaction, Int. J. Hydrogen Energy 37 (2012)
permeability of the Zirfon PERL separator for alkaline water electrolysis, J. 8185.
Electrochem. Soc. 163 (2016) 80. [51] M.M. Bakker, D.A. Vermaas, Gas bubble removal in alkaline water electrolysis
[20] J. Brauns, T. Turek, Alkaline water electrolysis powered by renewable energy: with utilization of pressure swings, Electrochim. Acta 319 (2019) 148.
A review, Processes 8 (2020) 248. [52] M.-Y. Lin, L.-W. Hourng, Ultrasonic wave field effects on hydrogen production
[21] P. Trinke, P. Haug, J. Brauns, B. Bensmann, R. Hanke-Rauschenbach, T. Turek, by water electrolysis, J. Chin. Inst. Eng. 37 (2014) 1080.
Hydrogen crossover in pem and alkaline water electrolysis: mechanisms, direct [53] H. Matsushima, D. Kiuchi, Y. Fukunaka, Measurement of dissolved hydrogen
comparison and mitigation strategies, J. Electrochem. Soc. 165 (2018) F502. supersaturation during water electrolysis in a magnetic field, Electrochim. Acta
[22] S. Sevda, X. Dominguez-Benetton, K. Vanbroekhoven, T.R. Sreekrishnan, D. Pant, 54 (2009) 5858.
Characterization and comparison of the performance of two different separator [54] J.W. Haverkort, H. Rajaei, Electro-osmotic flow and the limiting current in
types in air–cathode microbial fuel cell treating synthetic wastewater, Chem. alkaline water electrolysis, J. Power Sources Adv. 6 (2020) 100034.
Eng. J. 228 (2013) 1. [55] J.W. Haverkort, Modeling and experiments of binary electrolytes in the presence
[23] H. In Lee, D.T. Dung, J. Kim, J.H. Pak, S. k. Kim, H.S. Cho, W.C. Cho, C.H. Kim, of diffusion, migration, and electro-osmotic flow, Phys. Rev. A 14 (2020) 044047.
The synthesis of a Zirfon-type porous separator with reduced gas crossover for [56] D. Jang, H.-S. Cho, S. Kang, Numerical modeling and analysis of the effect
alkaline electrolyzer, Int. J. Energy Res. 44 (2020) 1875. of pressure on the performance of an alkaline water electrolysis system, Appl.
[24] A.S. Strub, G. Imarisio, Hydrogen As an Energy Vector: Proceedings of the Energy (2021) 116554.
International Seminar, Held in Brussels, 12–14 February 1980 Springer Science [57] Ø. Ulleberg, Modeling of advanced alkaline electrolyzers: a system simulation
& Business Media, 2013. approach, Int. J. Hydrogen Energy 28 (2003) 21.
[25] S. Marini, P. Salvi, P. Nelli, R. Pesenti, M. Villa, M. Berrettoni, G. Zangari, Y. [58] A. Roy, S. Watson, D. Infield, Comparison of electrical energy efficiency of
Kiros, Advanced alkaline water electrolysis, Electrochim. Acta 82 (2012) 384. atmospheric and high-pressure electrolysers, Int. J. Hydrogen Energy 31 (2006)
[26] M.R. Kraglund, Alkaline Membrane Water Electrolysis with Non-Noble 1964.
Catalysts (Ph.D. thesis), Technical University of Denmark, 2017. [59] P. Haug, M. Koj, T. Turek, Influence of process conditions on gas purity in
[27] R. Martinez Picazo, Optimal Electrode Material Study for an Alkaline Electrolyser alkaline water electrolysis, Int. J. Hydrogen Energy 42 (2017) 9406.
Integrated Into a Micro-Scale Methanol Plant (Master’s thesis), 2020. [60] T. Shimamune, E. Zimmerman, C. Andreasson, Electrode, 2009, US Patent 7,
[28] K. Zeng, D. Zhang, Recent progress in alkaline water electrolysis for hydrogen 566, 389.
production and applications, Prog. Energy Combust. Sci. 36 (2010) 307. [61] D. Pletcher, X. Li, S. Wang, A comparison of cathodes for zero gap alkaline water
[29] C.A.C. Sequeira, D.M.F. Santos, B. Šljukić, L. Amaral, Physics of electrolytic gas electrolysers for hydrogen production, Int. J. Hydrogen Energy 37 (2012) 7429.
evolution, Braz. J. Phys. 43 (2013) 199. [62] C. Henao, K. Agbossou, M. Hammoudi, Y. Dubé, A. Cardenas, Simulation tool
[30] X. Zhao, H. Ren, L. Luo, Gas bubbles in electrochemical gas evolution reactions, based on a physics model and an electrical analogy for an alkaline electrolyser,
Langmuir 35 (2019) 5392. J. Power Sources 250 (2014) 58.
[31] A. Angulo, P. van der Linde, H. Gardeniers, M. Modestino, D.F. Rivas, Influence [63] R.J. Gilliam, J.W. Graydon, D.W. Kirk, S.J. Thorpe, A review of specific
of bubbles on the energy conversion efficiency of electrochemical reactors, Joule conductivities of potassium hydroxide solutions for various concentrations and
4 (2020) 555. temperatures, Int. J. Hydrogen Energy 32 (2007) 359.

8
J.W. Haverkort and H. Rajaei Journal of Power Sources 497 (2021) 229864

[64] C. Lamy, P. Millet, A critical review on the definitions used to calculate the [72] H. Vogt, K. Stephan, Local microprocesses at gas-evolving electrodes and their
energy efficiency coefficients of water electrolysis cells working under near influence on mass transfer, Electrochim. Acta 155 (2015) 348.
ambient temperature conditions, J. Power Sources 447 (2020) 227350. [73] J. Yu, Q. He, G. Yang, W. Zhou, Z. Shao, M. Ni, Recent advances and prospective
[65] S. Holmin, L.-Å. Näslund, Á.S. Ingason, J. Rosen, E. Zimmerman, Corrosion of in ruthenium-based materials for electrochemical water splitting, ACS Catal. 9
ruthenium dioxide based cathodes in alkaline medium caused by reverse currents, (2019) 9973.
Electrochim. Acta 146 (2014) 30. [74] J. Divisek, H. Schmitz, A bipolar cell for advanced alkaline water electrolysis,
[66] H. Vogt, The concentration overpotential of gas evolving electrodes as a multiple Int. J. Hydrogen Energy 7 (1982) 703.
problem of mass transfer, J. Electrochem. Soc. 137 (1990) 1179. [75] L.G. Austin, H. Lerner, The mode of operation of porous diffusion electrodes—I.
[67] D. Guilbert, G. Vitale, Dynamic emulation of a pem electrolyzer by time constant Simple redox systems, Electrochim. Acta 9 (1964) 1469.
based exponential model, Energies 12 (4) (2019) 750. [76] Y. Nishiki, K. Aoki, K. Tokuda, H. Matsuda, Secondary current distribution in
[68] A. Hernández-Gómez, V. Ramirez, D. Guilbert, B. Saldivar, Cell voltage static- a two-dimensional model cell composed of an electrode with an open part, J.
dynamic modeling of a pem electrolyzer based on adaptive parameters: Appl. Electrochem. 16 (1986) 291.
development and experimental validation, Renewable Energy 163 (2021) 1508. [77] J. Newman, K.E. Thomas-Alyea, Electrochemical Systems, John Wiley & Sons,
[69] G. Tsekouras, R. Terrett, Z. Yu, Z. Cheng, G.F. Swiegers, T. Tsuzuki, R. Stranger, 2012.
R.J. Pace, Insights into the phenomenon of ‘bubble-free’ electrocatalytic oxygen [78] E.V. Dydek, B. Zaltzman, I. Rubinstein, D.S. Deng, A. Mani, M.Z. Bazant,
evolution from water, Sustain. Energy Fuels 5 (3) (2021) 808. Overlimiting current in a microchannel, Phys. Rev. Lett. 107 (2011) 118301.
[70] J.W. Haverkort, A theoretical analysis of the optimal electrode thickness and [79] A. Vilanova, P. Dias, J. Azevedo, M. Wullenkord, C. Spenke, T. Lopes, A. Mendes,
porosity, Electrochim. Acta 295 (2019) 846. Solar water splitting under natural concentrated sunlight using a 200 cm2
[71] B. Tjaden, S.J. Cooper, D.J.L. Brett, D. Kramer, P.R. Shearing, On the origin and photoelectrochemical-photovoltaic device, J. Power Sources 454 (2020) 227890.
application of the bruggeman correlation for analysing transport phenomena in [80] C.W. Raetzsch, H.J.F. Van, H. Cunningham, Diaphragm cell, 1975, US Patent 3,
electrochemical systems, Curr. Opin. Chem. Eng. 12 (2016) 44. 910, 827.