CFD modeling of multiphase flow in an alkaline water

electrolyzer
Citation for published version (APA):
Zarghami, A., Deen, N. G., & Vreman, A. W. (2020). CFD modeling of multiphase flow in an alkaline water
electrolyzer. Chemical Engineering Science, 227, Article 115926. https://doi.org/10.1016/j.ces.2020.115926

Document license:
TAVERNE

DOI:
10.1016/j.ces.2020.115926

Document status and date:

Published: 14/12/2020

Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please
follow below link for the End User Agreement:
www.tue.nl/taverne

Take down policy

If you believe that this document breaches copyright please contact us at:
[email protected]
providing details and we will investigate your claim.

Download date: 26. Nov. 2024

 Chemical Engineering Science 227 (2020) 115926

Contents lists available at ScienceDirect

Chemical Engineering Science

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

CFD modeling of multiphase flow in an alkaline water electrolyzer

A. Zarghami a, N.G. Deen a, A.W. Vreman a,b,⇑
a
Power and Flow Group, Department of Mechanical Engineering, Eindhoven University of Technology, Eindhoven, the Netherlands
b
Nouryon, Research Development and Innovation, Process Technology, Deventer, the Netherlands

h i g h l i g h t s

Euler-Euler simulation of the flow in a water electrolyzer for a range of current densities.

Comparison of the simulated local gas fraction with experimental data from literature.

Development of user-defined functions in Fluent was required in order to satisfy the momentum balance of the gas phase.

Demonstration of the important role of the turbulent dispersion force.

The effect of the turbulence model on the gas distribution.

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

Article history: The hydrodynamic properties of gas–liquid flows in water electrolyzers are of great practical interest
Received 20 March 2020 since the local distribution of gas influences the amount of electrical energy required to produce hydro-
Received in revised form 13 June 2020 gen. We used the Euler-Euler model to simulate the multiphase flow in a water electrolyzer and com-
Accepted 19 June 2020
pared the results to existing experimental data, for a range of current densities. Our study shows that
Available online 5 July 2020
if only the drag force and buoyancy force are incorporated in the model, the spreading of the gas layers
formed at the electrodes is not accurately predicted. By adding the turbulence dispersion force to the
Keywords:
model, reasonable agreement with the experimental data could be obtained for the higher current den-
Electrolysis
Computational fluid dynamics
sities. The turbulence dispersion had to be implemented via user-defined functions, in order to obtain
Gas volume fraction results that satisfied the momentum balance. In addition the effect of different turbulence models on
Hydrogen evolution the turbulent dispersion was investigated.
Turbulent dispersion Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction is a fundamental building block for the manufacture of ammonia

(Lattin et al., 2006), and hence fertilizers, and of methanol, used
Hydrogen, as an energy carrier, is expected to play a key role in in the manufacture of many polymers. Refineries, where hydrogen
future energy systems of the world. It owes its popularity to the is used for the processing of intermediate oil products, are another
increase in the energy costs caused by the declining availability of area of use (Mueller-Langer et al., 2007). Moreover, hydrogen is
oil reserves, production and supply (McGlade, 2012) and also to used in buildings and power industries, where it could be mixed
the concerns about global warming and climate changes, which with natural gas or combined with CO emissions to produce syngas
are blamed on man-made carbon dioxide (CO2) emissions associ- (Liu et al., 2010). Furthermore, hydrogen energy used by the trans-
ated with fossil fuel use (Norby and Luo, 2004). Hydrogen is classi- port sectors is growing, where it can provide low-carbon mobility
fied as a clean fuel as it emits nothing except water at the point of through fuel-cell electric vehicles (Cheng et al., 2007).
use. Also, it can be produced using renewable energy (e.g., wind, Such a wide-ranging hydrogen consumption requires large
solar, hydropower, etc.) which makes it quite attractive (Steinfeld, scale hydrogen production. Hydrogen is usually produced by gasi-
2002). fication and reforming of heavy oil (Sato et al., 2003), gasification of
Also, hydrogen is used as raw material in the chemical industry, coal and petroleum coke (Trommer et al., 2005) and reforming of
and also as a reductor agent in the metallurgic industry. Hydrogen natural gas (Turner, 2004). Although water electrolysis is a mature
technology and is one of the simplest ways to produce hydrogen
(and oxygen), it still contributes to only a small fraction ( 4%)
⇑ Corresponding author at: P.O. Box: 513, 5600MB, TU/e, Eindhoven, the of the world hydrogen production (De Souza et al., 2007). However
Netherlands. in comparison to other methods, water electrolyser has the advan-
E-mail address: [email protected] (A.W. Vreman). tage of producing extremely pure hydrogen (> 99:9%), ideal for

https://doi.org/10.1016/j.ces.2020.115926
0009-2509/Ó 2020 Elsevier Ltd. All rights reserved.
2 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926

some high value-added processes such as the manufacturing of increase efficiency (by applying high temperature and high pres-
electronic components. Moreover, water electrolysis can be pow- sure operation), durability and safety of current electrolyzers.
ered by renewable energy sources which results in zero CO2 emis- An alkaline-water electrolyzer is a type of electrochemical cell
sion. Hence applications of water electrolysis are mostly limited to that is characterized by having two electrodes (namely anode
small-scale applications where large-scale hydrogen production and a cathode) operating in a liquid alkaline electrolyte solution
plants are not accessible or economical to use, including marine, of potassium hydroxide (KOH) or sodium hydroxide (NaOH). In
rockets, spacecrafts, electronic industry and medical applications industrial alkaline water electrolysis, 20–40% KOH or NaOH aque-
(Santos et al., 2013; Zeng and Zhang, 2010). The challenges for ous solution is used as the electrolyte instead of pure water to
expanding the use of water electrolysis are to reduce energy con- overcome high resistivity of water to electricity. Alkaline elec-
sumption and cost and maintenance, and, on the other hand, to trolyzers operate via transport of hydroxide ions (OH) through

Fig. 1. a) Schematic of the cathodic half-cell, where hydrogen is produced, when three electrodes are active (i.e. l = 12 cm), and b) selected boundary conditions and
computational parameters for simulations.
 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926 3

the electrolyte from the cathode to the anode with hydrogen bub- and turbulent viscosity in the system to better understand the phy-
bles being generated at the cathode and oxygen bubbles at the sics of the problem. Finally, recommendations for a proper simula-
anode (see Fig. 1). tion of hydrogen generation in an electrochemical cell are
The performance of an alkaline-water electrolyzer is closely presented.
linked to the hydrodynamic characteristics of the gas–liquid flow The rest of paper is structured as follows. The details of the
in each cell. The presence of bubbles is known to cause local turbu- numerical model are presented in Section 2. Results are presented
lence, which is very efficient in mixing and local distribution of the in Section 3, following by concluding remarks in Section 4.
species. The bubbles form a curtain of increasing thickness along
the vertical electrode (see Fig. 1). The rising bubbles also accelerate
the electrolyte flow near the electrode which in turn pronounces 2. The CFD model
the convective transport of electrochemically active species. On
the other hand, in a zero-gap electrolysis cell the bubbles attached 2.1. Physical case
to the electrode reduce the effective electrode surface area (Vogt,
2012) while in nonzero-gap configurations the rising bubbles act In this paper the experimental data achieved by Riegel et al.
as moving electrical insulators, thus affecting the current density (1998) is used for validation purposes. The experimental setup
distribution and increasing the ohmic drop across the cell (Hreiz consisted of two compartments, i.e. the cathodic and anodic com-
et al., 2015). Hence, the hydrodynamic behaviour of the two- partments, which were separated by a diaphragm. The gases were
phase flow in an electrolysis cell can have a considerable effect evolved on ten electrode pairs mounted in the upper section of the
on the cell efficiency. electrolyser. Fig. 1a shows a sketch of the cathode cell where
It is worth mentioning that the bubble coverage, i.e. the fraction hydrogen was produced. The size of each electrode is 4 cm and
of the electrode surface covered by adhering bubbles, is an impor- they could be activated one by one, from top to bottom, indicating
tant operation parameter affecting the performance of the cell. Actu- that the length of the cathode varied from l ¼ 4 cm to l ¼ 40 cm for
ally, bubbles adhering to an electrode surface insulate a part of the different experiments. Also, measurements were performed for
surface making it inactive in the electrochemical reaction, so that various current densities, i.e. i = 500, 1500, 3250 and 6250 A/m2.
the current density and the surface overpotential at the bubble- The width of the compartment was D ¼ 8 mm and the potassium
free fraction of the surface is increased in case the total current is hydroxide solution (KOH) as the electrolyte was pumped through
controlled to be constant. On the other hand, detaching bubbles the electrolyser at uin ¼ 0:69 m/s. The gas fraction profile was mea-
from the electrode induce microconvection in the boundary layer sured at the upper end of the channel, just above the top electrode
intensifying mass transfer (Vogt, 2012; Vogt and Balzer, 2005). (see line A  A in Fig. 1a). At that height a row of nine small plat-
Therefore, detailed investigation of bubble dynamics, phase inum electrodes was mounted on a line perpendicular to the chan-
interactions and gas hold-up is crucial for understanding the nel wall, so that, through the Maxwell equation, the gas fraction
mechanism and enhances the performance of an electrochemical profile on that line was determined from the measured profile of
cell. There are numerous studies in the literature investigating var- the electrical resistance.
ious aspects of two-phase flow hydrodynamics in electrochemical
cells (Hreiz et al., 2015; El-Askary et al., 2015; Abdelouahed et al., 2.2. Governing equations
2014a; Abdelouahed et al., 2014b). However, owing to the high gas
fraction, many key features of the multiphase flow field cannot be In this paper, the ANSYS FluentÒ 19.1 package is used for simu-
captured by the common optical techniques. Hence, computational lating the multiphase flow. The flow is considered Newtonian, vis-
fluid dynamics (CFD) is also used for studying complex multiphase cous, incompressible and isothermal as the physical properties of
flow in electrochemical cells (Mat et al., 2004; Hreiz et al., 2015; the phases remain constant. It is assumed that the operating pres-
Aldas, 2004; Hawkes et al., 2009; Alexiadis et al., 2011; Charton sure is 1 bar and the electrolyte is 30% KOH aqueous solution and
et al., 2010). ql ¼ 1250 kg/m3 and ll ¼ 0:9  103 kg/m.s. Also, the gas phase is
However, despite many interesting CFD studies available in the considered as a mixture of hydrogen and water vapor with
literature, validation of CFD results (such as hydrogen volume frac- qg ¼ 0:21 kg/m3 and lg ¼ 2  105 kg/m.s (Haug et al., 2017;
tion and width of the gas hold-up in a cell) with equivalent exper- Takeuchi et al., 2008; Walker, 1971).
imental data is quite rare. Hence, there is as yet no consensus on In the plane channel the transition from laminar to turbulent
the most capable and suitable method for simulating gas–liquid flow regime occurs at critical Reynolds number which is defined
flow in an electrolyzer. In this way, ability of CFD models for pre- as Recr ¼ ql uin D=ll  2100. Since the Reynolds number of the liq-
dicting the width of the hydrogen bubbles curtain, which is an uid in the cathode, Re  7600 is considerably higher than the crit-
important feature of the flow, is a significant criterion for accuracy ical Reynolds number, the flow regime corresponding to the above
and reliability of the model. Accurate prediction of the bubble cur- mentioned experimental conditions is assumed to be turbulent. In
tain spreading is quite challenging since if no bubble dispersion/- order to represent the multiphase flow behavior we apply an Euler-
transverse migration term is added to the CFD model, the Euler model which utilizes a set of momentum and continuity
distribution of local gas fraction cannot be predicted accurately equations for each phase. The governing continuity and momen-
(Caire et al., 2009; Abdelouahed et al., 2014b). tum equations according to the Eulerian model are (van der Hoef
In this paper we aim to establish a CFD model to simulate mul- et al., 2006; van der Hoef et al., 2008)
tiphase flow in an alkaline-water electrolysis cell and compare the
results with the experimental data. More particularly we focus on @ ag qg
þ r  ðag qg ug Þ ¼ rV;g ; ð1Þ
gas fraction and width of the bubble curtain in a cell. For this pur- @t
pose the influences of drag and turbulent dispersion forces acting
on the bubbles are analyzed to see the effect of each parameter @ ag qg ug
þ r  ðag qg ug ug Þ ¼ ag qg g  ag rp þ r  ðag Tg Þ þ S ð2Þ
on the flow pattern. To validate the modelling approach, numerical @t
results are confronted to experimental data. The effects of inter- for the dispersed phase, and
phase forces on the accuracy/stability of the CFD model as well
@ al ql
as simulation results are also investigated. We also study some þ r  ðal ql ul Þ ¼ 0; ð3Þ
flow field parameters such as slip velocity, total amount of gas @t
4 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926

@ al ql ul the turbulent viscosity lturb

2
is defined by ql C l kl =el , where
þ r  ðal ql ul ul Þ ¼ al ql g  al rp þ r  ðal Tl Þ  S ð4Þ l
@t C l ¼ 0:09 and kl ¼ 12 trðRl Þ.
for the continuous phase, where subscripts g and l refer to gas (dis- The last term S in the momentum equations of the phases rep-
persed) and liquid (continuous) phase, respectively. The symbols resents the interphase momentum transfer which include the
q; l; a and u represent density, dynamic viscosity, void fraction forces exerted on/by the dispersed phase. We consider this term
and velocity vector respectively. Note that, both phases are as a summation of drag (Fd ) and turbulence dispersion (Ftd ) forces:
assumed to have a constant density. Also, rV;g is a volumetric mass
S ¼ Fd þ Ftd : ð9Þ
source that is only active in a thin layer adjacent to the electrode, so
that no gas inlet boundary condition at the electrode is needed (see These forces play the most important role in predicting the dis-
Section 2.4). Note that the pressure, p, is shared by the two phases tribution of the gas volume fraction. The wall-normal component
in the Eulerian model. Also, the volume fractions are assumed con- of the force balance determines the establishment of radial distri-
tinuous functions of space and time and their sum is equal to one, butions of the bubbles.
i.e. al þ ag ¼ 1.
It can be seen that the Euler-Euler model has two continuity 2.2.1. Drag force
and two momentum equations for the two phases to calculate vol- The drag force represents a resistance to the movement of gas
ume fraction and velocity fields of each phase and a shared pres- bubbles and acts in the opposite direction of the bubble–liquid slip
sure field. In this model, the phase volume fractions are assumed velocity. The drag force is expressed as:
to be continuous functions of space and time and their sum is equal
to one. In other words, as the densities of phases are assumed to be 3 C d ql ag
Fd ¼  jug  ul jðug  ul Þ; ð10Þ
constant, the continuity equations provide governing equations for 4 db
the phases volume fractions (i.e. ag and al ), while both equations where ug  ul represents the rise (or terminal, relative) velocity vec-
together with the constraint al þ ag ¼ 1 lead to a Poisson equation tor between two phases, db is the bubble diameter, C d is the drag
from which the pressure is calculated. coefficient, for which the so-called Schiller and Naumann correla-
In the governing equations, Tl denotes the stress tensor, which tion (Schiller, 1933) is employed here:
includes both viscous and turbulent stresses. Using Stokes’ hypoth- (
esis for the second coefficient of viscosity, it is described for e.g. the 24
ð1 þ 0:15Re0:687 Þ for Re 6 1000
Cd ¼ Re ; ð11Þ
viscous part of the liquid phase as 0:44 for Re > 1000
1
Tl ¼ 2lmol
l ðDl  trðDl ÞIÞ  ql R l ; ð5Þ where Re is the Reynolds number of the bubble calculated on the
3
basis of relative velocity as Re ¼ ql jug  ul jdb =ll . The Schiller and
where lmol is the molecular dynamic viscosity, tr represents the Naumann model is a model for spherical particles. Since in the pre-
trace of matrix, I is the unit tensor and Dl is the strain rate tensor sent case, the Eotvos number is low due to the small size of the bub-
which is defined as bles, the bubbles are nearly spherical.

Dl ¼ 0:5ðrul þ ðrul ÞT Þ: ð6Þ

2.2.2. Turbulent dispersion force
In Eq. 5, Rl is the Reynolds stress tensor which is defined in Turbulent fluctuations produce randomness in the relative
terms of the turbulent fluctuating velocities as Rl ¼< u0l u0l >, where velocity between phases, which leads to oscillations in the drag
<> makes the involved averaging operation explicit. This term can force. While these fluctuations in the streamwise direction are
be modeled either by using the Boussinesq eddy viscosity hypoth- low compared to the mean drag and buoyancy force, they bring a
esis or by solving the Reynolds stress transport equation. In this considerable impact on the redistribution of bubbles in the span-
work, the Reynolds stress equation (RSE) turbulence model is wise direction. This effect can be expressed as a turbulent disper-
applied for modeling the stress tensor. This model accounts for ani- sion force acting on the gas phase which signifies the turbulent
sotropy of the Reynolds stresses and is therefore often able to offer dispersion of the bubbles by the random motion of continuous
better accuracy than isotropic eddy viscosity-based turbulence phase eddies. This is caused by the combined action of turbulent
models. Since the concentration of the gas phase is rather low, eddies and interphase drag which is derived by the volume fraction
the dispersed turbulence model is used. Thus, the transport equa- gradient and tend to flatten the volume fraction distribution (Ishii
tions for turbulence quantities are only solved for the liquid phase, and Hibiki, 2010).
while Rg , the Reynolds stress tensor of the dispersed (gas) phase, is Burns et al. (2004) have derived a model for the turbulent dis-
simply proportional to Rl and the proportionality factor is com- persion force based on Favre averaging of the drag term, which is
puted according to the Tchen theory (Chan-Mou, 2013). The trans- given by:
port equation for the Reynolds stress tensor is calculated as (Pope,  
3 Cd lturb 1 1
2001; Wilcox, 2006) Ftd ¼  ag jug  ul j l þ rag ; ð12Þ
4 db Sctd ag al
@ðal ql Rl Þ

þ r  ðal ql ul  Rl Þ ¼ r  al ðlmol þ C s lturb Þr  R l where Sctd is the Schmidt number of turbulent dispersion with
@t

l l
 ð7Þ
þal ql Pl þ /l  3 el I ;
2 adopted value of 0.9. The proportionality to the gradient of the vol-
ume fraction in turbulence dispersion equation causes transport of
and the turbulent dissipation rate, el , is given as gas from regions of high concentration to regions of low concentra-
@ðal ql el Þ
 tion. Physically this transport is due to local (turbulent) fluctuations
þ r  ðal ql ul el Þ ¼ r  al ðlmol þ C e lturb Þrel of the velocities. Therefore this term is called the turbulent disper-
@t

l l
 ð8Þ
þal ql kel C e;1 12 trðPl Þ  C e;2 el ; sion model.
l

where C s ; C e ; C e;1 and C e;2 are equal to 0.25, 0.15, 1.44 and 1.92 2.2.3. Buoyancy force
respectively. The tensor Pl ¼ 2Rl  D is the production by main The buoyancy is an upward force exerted by a fluid on bubbles
shear, while the tensor /l represents the pressure-strain model for- in a gravity field. In fluids, pressure increases with depth; hence,
mulated by Gibson and Launder (1978) and Launder (1989). Also, when a bubble rises in a fluid, the pressure exerted on its bottom
 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926 5

surface is higher than the pressure exerted on its top surface. This wall function approach is applied (Pope, 2001; Launder and
difference in the pressure leads to a net upward force which Spalding, 1983), which means that a model for the shear stress
opposes the gravity force. It is remarked that the buoyancy force close to the wall is applied as wall stress condition in the
acting on the bubbles is not included in S but included in the gov- momentum equation.
erning equations (see Eq. 2) as Fb ¼ ag qg g  ag rp. This term is sig- In order to simulate the bubble generation at the electrodes, a
nificant and positive in the vertical direction, while it is small in volumetric mass source for the gas phase is applied in a region
the horizontal direction. adjacent to the left wall with a width equal to the diameter of
the bubbles, i.e. w = 150 lm for i ¼ 500 A/m2 and w = 100 lm for
i P 1000 A/m2. Note that for each case the mesh is stretched in
2.3. Bubble size
such a way that the width of the first cell becomes equal to the
diameter of the bubbles. Thus the hydrogen produced by the reac-
The diameter of bubbles is a required input parameter. Hence,
tion at the surface of electrode is prescribed as a volumetric mass
we have to prescribe a representative bubble size in order to
source and not as a surface mass source (or horizontal inlet veloc-
numerically solve the problem. The size distribution of bubbles
ity). It should be highlighted that the latter was also tried, but it
at the electrode depends on several parameters, such as number
appeared to interfere with or deactivate the wall shear stress
of nucleation sites, surface wettability, etc. Various experimental
model needed in the standard wall function approach. Therefore,
studies (Hreiz et al., 2015; Abdelouahed et al., 2014b;
we have chosen the other route, the use of a volumetric source
Boissonneau and Byrne, 2000; Hreiz et al., 2015; Vogt, 2012) have
term. The volumetric source term is calculated as:
shown that there are three major phenomena that may change the
bubble size significantly: (1) change in hydrostatic pressure Mg F g
applied to the bubbles during their rise, (2) mass transfer phenom-
rV;g ¼ C ; ð13Þ
w
ena and incorporation of dissolved gas from the electrolyte and (3)
where M g and F g ¼ i=Fz represent the molar mass (kg/mol) and
bubble coalescence. However, the first phenomenon is only signif-
icant if the operating pressure is very low or if the cells height mole flow (mol/m2s) of the gas mixture respectively, i is the current
exceeds a few meters. Considering the operating conditions and density, F ¼ 96487 A.s/mol is the Faraday constant, and z ¼ 2 is the
geometry of the selected case study, we can safely ignore this phe- number of electrons involved in the electrochemical reaction
nomenon. The second phenomenon is also neglected because (2H2O + 2e ! H2+2OH). Also, C is a factor that accounts for the
inclusion of mass transport of dissolved hydrogen is beyond the mole fraction vH2 O and vH2 in the gas phase at T ¼ 80  C. The water
scope of this paper. vapor pressure in the alkaline solution is approximately 0.26 bar, so
Flow visualizations have shown that (Abdelouahed et al., that vH2 O ¼ 0:26 and vH2 ¼ 0:74 (Haug et al., 2017; Takeuchi et al.,
2014b; Hreiz et al., 2015) bubble coalescence occurs in the vicinity 2008; Walker, 1971). Therefore:
of the electrodes only, and mainly among the bubbles that are not vH2 O þ vH2 1
yet detached from the surface. However, the high ionic strength of C¼ ¼ ¼ 1:35: ð14Þ
electrolyte solutions, which is measure of the total concentration
vH2 vH 2
of ions in solution, is known to limit the significance of this phe- According to the above assumptions, the molar weight of the gas
nomenon. For instance, it has been observed that a significant mixture is Mg ¼ 0:0062 kg/mol and the volumetric source term
number of bubbles do not coalesce and maintain their initial small for current density i = 1500 A/m2 is equal to rV;g = 0.652 kg/m3s.
size (Boissonneau and Byrne, 2000; Hreiz et al., 2015). Hence, the
Note that in order to reduce the number of simulations that are
phenomenon is far from being prevalent and we assume that there
required in this work, we simulate the electrolyzer with ten active
is no bubble coalescence and break-up. Therefore, a monodisperse
electrodes. Then, the distribution of void fraction above the first
bubble size distribution is a reasonable assumption for the purpose
electrode is compared with the experimental data for one pair of
of this work.
active electrodes, the profile of void fraction above the second elec-
Haug et al. (2017) measured the averaged bubble size as a func-
trode is compared with the experimental data for two (pairs of)
tion of current density for an alkaline-water electrolyzer at 80  C
active electrodes and so on. Our investigations showed that there
with  30% KOH aqueous solution as the electrolyte. They showed
are no significant differences between the results of the selected
that the mean cathodic bubble size diameter firstly grows from
strategy and a strategy with multiple computational domains, in
170 to 220 lm in the range from i ¼ 100 to 300 A/m2 and then which each computational domain contains an electrode with a
decreases to an approximately constant value for i > 1000 A/m2. length equal to the total length of the number of active electrodes
Hence, according to data available in Haug et al. (2017), we set considered, whose total electrode length is precisely equal to the
db ¼ 150 lm for i ¼ 500 A/m2 and db ¼ 100 lm for i P 1000 A/m2. total length of number of active electrodes considered. Hence, we
define a parameter, h, that represents the height along the 40 cm
2.4. Boundary conditions high electrode in the simulation, but at the same time h also repre-
sents the distance from the bottom of the lowest active electrode
The detailed geometry of the channel is shown in Fig. 1b. The in the experiment to the top of the highest active electrode, where
velocity-inlet boundary condition is applied for the bottom experimental gas fraction was measured. Thus (simulation) results
boundary of the computational domain as a uniform velocity at height h ¼ 4N cm are compared with (experimental) results
profile with ul;y ¼ 0:69 m/s, ul;x ¼ 0; ug;y ¼ 0 and ug;x ¼ 0. The measured just above N active electrodes (which have a total length
inlet boundary is plced at a distance 12:5D ¼ 10 cm below the of l ¼ h). Also, the local gas layer thickness, dg , is defined as a hor-
first electrode. We verified that this entry length was sufficiently izontal distance from the cathode surface to the point at which the
large to obtain a fully developed mean velocity profile just gas void fraction is 0.001.
below the first electrode. At the outlet boundary (at the top), One of the important factors affecting multiphase flow beha-
a constant pressure is prescribed. In order to prevent upstream viour is the slip velocity of bubbles which determines the momen-
effects of the outlet boundary condition on the fluid flow, the tum transfer between phases. In this work, the slip velocity is
boundary is placed 10 cm above the top electrode, so that the calculated as Us ¼< ug  ul >, where <> indicates time averaging.
total height of the computational domain is 60 cm. All other Also, the total gas fraction is determined by the volume average of
boundaries are set as no-slip boundaries, for which the standard ag as:
6 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926

Z
1 present problem. Three types of stretched meshes with different
ag;tot ¼ ag dV; ð15Þ
V number of nodes in the cross-section direction, i.e. N x ¼ 15; 30
and 60 were used for i ¼ 1500 A/m2. The mesh size in the vertical
where V is the volume of the entire flow domain. (or streamwise) direction is uniform and is equal to Dy ¼ 2 mm,
Dy ¼ 1 mm and Dy ¼ 0:5 mm for N x ¼ 15; 30 and 60, respectively.
2.5. Mesh dependency test The predicted gas volume fraction along h ¼ 24 cm is shown in
Fig. 2. As shown in the figure, the numerical results with N x ¼ 60
Since mesh refinement is a very important factor for determin- and N x ¼ 30 agree well with each other (differences less than
ing an accurate solution, numerical tests were performed to deter- 1%). Hence, the selected mesh size with N x ¼ 30 and Dy ¼ 1 mm
mine the grid size for nearly grid-independent solutions for the is a proper choice. This mesh is also used for the other current den-

Fig. 2. Predicted gas volume fraction at h ¼ 24 cm with various grid sizes for i = 1500 A/m2.

Fig. 3. Distributions of a) streamwise (Fy ) and b) lateral (Fx ) components of force densities acting on bubbles along the cross-section of the channel at h ¼ 32 cm when forces
are selected as the default implementations in Fluent.
 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926 7

sities, except that for i ¼ 500 A/m2 the stretching factor was dient, convective and viscous terms are negligible. Any deviation
slightly modified (as explained in Section 2.4). from zero can be considered as error in the momentum balance.
It is worth mentioning that the so-called QUICK scheme is cho- Fig. 3 shows that the momentum balance in streamwise direc-
sen for spatial discretization of the governing equations. Also, the tion is not fully satisfied. Note that all simulation were well con-
first-order implicit time stepping method is applied for temporal verged with residuals of all equations well below 105 . The
discretization. maximum error in the streamwise direction reaches 7% of the drag
force in the vicinity of the electrode surface. However, there is a
2.6. Model verification significant momentum imbalance in the lateral direction: the error
reaches roughly 20% of the drag force in the region close to the
As we mentioned earlier, the Reynolds stress equation (RSE) electrode. Note that the error is calculated simply as the summa-
model is applied as the turbulence model. The effects of various tion of the Buoyancy, drag and turbulent dispersion forces which
turbulence models on the simulation results will be discussed in should be very close to zero to show the satisfy the momentum
the last part of Section 3. Here, the drag and the turbulence disper- balance.
sion forces are considered as the major interaction forces between It is mentioned that the amount of imbalance in the lateral
the phases. At first, these two forces were simply selected as they direction is not the same for all cross-sections and it gets larger
are available by default in Fluent. In order to verify the internal as the gas hold-up increases either by increasing the electrode size
consistency of the model we check the momentum balance for or by increasing the current density. Furthermore, it is mentioned
the gas phase (see Eq. 2). For this purpose, the distributions profiles that the imbalance was found to be at least as large when the RSE
of the force components along the cross section of the channel at model was replaced by the k  e or SST k  x model.
h ¼ 32 cm with i ¼ 1500 A/m2 are plotted in Figs. 3 and 4. The con- After doing extensive studies, we concluded that the turbulence
vective, pressure gradient and viscous terms in the horizontal dispersion force is not implemented as the formula presented in
momentum equation are negligible compared to the forces and the Fluent manual or there is a kind of error in implementing this
therefore not shown. It has been verified that the simulation force in the source code of Fluent. To overcome this issue, we
reached the steady state and that the time derivatives in the implemented user-defined-functions (UDFs) for both the Schiller-
momentum balance are negligible. Hence, the buoyancy, the drag Naumann drag and the Burns turbulent dispersion forces according
and the turbulent dispersion forces are the remaining terms. For to equations described in the Fluent manual. Fig. 4 shows that the
the gas momentum balance to be satisfied the summation of these momentum balance is satisfied perfectly if the UDFs are used as
three forces should be very close to zero, because the pressure gra- momentum source terms and the default interfacial forces in Flu-

Fig. 4. Distributions of a) streamwise (Fy ) and b) lateral (Fx ) components of force densities acting on bubbles along the cross-section of the channel at h ¼ 32 cm when forces
are applied by using UDFs in Fluent.
8 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926

ent are switched off: the sum of the forces is very close to zero at forces, the default Fluent settings for the lift force, the wall lubrica-
all points of both the streamwise and the lateral profiles. An tion and virtual mass forces. We added these forces to the model
equally good force balance was found when the UDF implementa- one by one. Our study showed that the wall lubrication force has
tion was used in combination with the k  e or SST k  x turbu- insignificant influence on the spreading of the gas fraction in the
lence model. channel. However, when the lift force or virtual mass force was
Furthermore, we have performed simulations in which we added to the model, these forces affected the steady state gas frac-
switched on, in addition to the drag and turbulent dispersion tion profiles in a spurious manner, as the steady state results

Fig. 5. Comparison of the predicted gas volume fraction with experimental data (Riegel et al., 1998) as a function of number of electrodes for i = 1500 A/m2.

Fig. 6. Streamwise component of the liquid velocity for different number of electrodes and i = 1500 A/m2.
 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926 9

Fig. 7. Horizontal, U s;x , and vertical, U s;y or streamwise, components of the slip velocity at h ¼ 24 cm for i ¼ 1500 A/m2.

Fig. 8. Effect of drag only (S ¼ Fd ) and drag and turbulent dispersion forces (S ¼ Fd þ Ftd ) on distribution of gas layer along the cross-section of the channel at h ¼ 24 cm in the
channel and i ¼ 1500 A/m2.

Fig. 9. Comparison between numerical volume fraction distribution of gas along the channel cross-section with the experimental data (Riegel et al., 1998) for different
current densities .and h ¼ 12 cm.
10 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926

became dependent on the time step. These spurious effects did not width along the streamwise direction of the flow. For a larger num-
disappear when the time step was reduced. Because of these ber of active electrodes (larger h in the simulations), the gas frac-
issues, we decided to ignore these forces and include only the tion is larger. This enhances the effect of buoyancy, so that the
two most dominant forces, the drag and turbulent dispersion force, liquid velocity profile tends more toward the surface of the
implemented as UDFs (as described above). electrodes.
The distributions of streamwise (Fy ) and lateral (or cross-
stream, Fx ) components of drag and non-drag force densities acting
3. Results
on bubbles were shown in Figs. 4a and 4b respectively. The nega-
tive sign of Fd;y indicates that the drag force acts in the negative y
The results of our simulations are compared with the experi-
direction. Due to the concentration of the gas bubbles in the vicin-
mental data of Riegel et al. (1998), who performed experiments
ity of the cathode, Fd;y shows a peak value close to the electrode
for multiple current densities and for a multiple number of active
and then decreases gradually within the gas layer. The negative
electrodes. Simulations of some of these experiments were also
sign of Fd;x reveals that the gas bubbles experience a wall-
presented in Mat et al. (2004) and El-Askary et al. (2015). However,
directed force. This is caused by the positive horizontal slip veloc-
in these references, the numerical results predicted for one active
ity (see Fig. 7). The horizontal slip velocity is approximately the
electrode (with the length of 4 cm) were compared with the exper-
same as the horizontal gas velocity because the horizontal liquid
imental data obtained for three active electrodes (with the total
velocity is negligible (absolute value less than 0.0002 at h ¼ 24
length of 12 cm). Hence, a new attempt to simulate these experi-
cm). As shown in Fig. 7, the magnitude of the horizontal slip veloc-
ments is clearly justified. It is worth mentioning that in Mat
ity is significant compared to the vertical one. The positive hori-
et al. (2004) a steady-state laminar flow has been solved for simu-
lating two-phase mixture of the liquid and gas in a cell. Also, a
mass diffusion term has been added to the continuity equation
to model bubble dispersion. At the electrode surface, the velocity
inlet boundary condition has been applied where the horizontal
velocity component of gaseous phase is calculated using Faraday’s
law. In El-Askary et al. (2015), the Euler-Euler model coupled with
k  e turbulent model has been used for simulating gas–liquid
flow. The authors have applied Butler–Volmer type boundary con-
dition for generating gas bubbles at the electrode.
In the next subsection, we show results for our base case, which
corresponds to a current density and the RSE turbulence model. In
the second subsection we show results for different current densi-
ties, while in the third subsection we include results for other tur-
bulence models.

3.1. Results for the base case

Fig. 5 shows the predicted gas volume fraction distribution

along the channel cross-section as a function of number of active
electrodes for a current density i ¼ 1500 A/m2. The experimental
data of Riegel et al. (1998) are also included. It can be seen that
the CFD results are generally in reasonably good agreement with
the measurements. The agreement becomes better if the number
of active electrodes increases. The gas volume fraction attains a
maximum value at the cathode and then decreases gradually along
the cross-section of the channel as the channel becomes free of gas
near the diaphragm surface. From this figure we also see that the
gas volume fraction increases with the number of electrodes. This
effect is more visible in close distances from the electrode. Also, the
width of the gas volume fraction profile increases by activating a
larger number of electrodes, which is due to the mixing and diffu-
sion of the gas phase along the cell as a result of included non-drag
forces.
The streamwise component of the liquid velocity as a function
of number of electrodes is shown in Fig. 6. At the end of the
entrance region, just below the electrodes (i.e. at h ¼ 0), the veloc-
ity profile is symmetric and resembles a fully developed turbulent
channel flow profile. Above this entrance region, electrodes are
present, and by generating the gas they change the liquid velocity
profile, which becomes asymmetric. The maximum shift toward
the left, towards the location of the electrode surface at x ¼ 0. In
other words, the velocity near the surface of the electrodes is
enhanced, because the bubbles accelerate the fluid. The bubbles
are driven by the buoyancy force caused by the density difference Fig. 10. Contour plots of the gas volume fraction in a cell with three active
between the phases. The rising bubbles, which are generated along electrodes for different current densities, from left to right panels:
the entire surface of the electrodes form a curtain of increasing i ¼ 500; i ¼ 1500; i ¼ 3250 and i ¼ 6250 A/m2, respectively.
 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926 11

zontal slip velocity (or horizontal gas velocity) is caused by the different current densities along the cross-section of the channel
positive turbulent dispersion force. with three electrodes (i.e. h ¼ 12 cm). One can see that there is a
Fig. 4b shows that the lateral component of the turbulent dis- reasonable agreement between the CFD results and the experimen-
persion force density, Ftd;x , has a considerable value with a positive tal data, apart from i ¼ 500 A/m2. It is clear that by increasing the
sign, whereas its streamwise component is negligible. The turbu- current density the agreement between the CFD results and the
lent dispersion force has an important influence on the lateral experimental data improves.
gas fraction profiles in the electrolyzer as it expands the width of As the electrochemical reaction rate is proportional to the cur-
the gas layer. rent density, by increasing the current density the gas production
The lateral component of the buoyancy force density is negligi- rate rises, resulting in higher void fraction at the cathode surface.
ble, however the streamwise component, Fb;y , is significant. The However, the void fraction at the cathode surface is not propor-
balance between the buoyancy and the drag force determines the tional to the increase in the current density due to thickening of
slip velocity of the bubbles. Since Fb;y is proportional to the gas vol- the gas layer on the electrode. Increasing the current density
ume fraction, this force decreases with increasing distance from results in a higher gas volume fraction, which on turn leads to
the electrode surface. more turbulent dispersion (the turbulent dispersion force is pro-
To further discuss the influence of turbulent dispersion forces portional to the gas volume fraction), so that bubbles move away
on the hydrodynamics of the flow, we compare the distribution from the electrode and the gas layer expands.
of the gas layer for two cases, i.e. S ¼ Fd and S ¼ Fd þ Ftd , along This is illustrated by Fig. 10, which shows contours of the gas
the same cross-section. It is observed from Fig. 8 that by including volume fraction in the region h 6 12 cm, for different current den-
only the drag force, i.e. S ¼ Fd , the thickness of gas layer shrinks, sities. If the current density increases, more gas is produced at the
which is due to the absence of the lateral component of Ftd that electrode and dispersed by the turbulence, so that both the local
moves the bubbles away from the electrode. Hence, it means that gas fraction as the total gas hold-up increase. The gas layer starts
effects of Ftd on the hydrodynamics of the flow are quite important. to grow at the leading edge at h ¼ 0. We observe that the thickness
Inclusion of this force is required to predict the spreading of the gas of the gas layer as function of h has a small growth rate at low cur-
layer sufficiently with reasonable accuracy. rent density, but also that the growth rate increases if the current
density increases. For the highest current density shown, the gas
3.2. Effect of current density layer has spread across the entire channel width after three elec-
trodes (h ¼ 12 cm).
In this part, the effect of current density on the gas fraction dis- To quantify this further, we define the thickness of the gas layer,
tribution is discussed. Fig. 9 shows gas volume fraction profiles for dg , by the location where the gas fraction is less than 0.001. The

Fig. 11. The effect of height along the cathode on the thickness of the gas layers at different current densities.

Fig. 12. Effect of current density on total volume fraction of gas in the electrolyzer.
12 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926

thickness of the gas layer as function of height is shown in Fig. 11, 3.3. Effect of turbulence models
also for different current densities. The growth rate of the gas layer
is the slope of dg in this figure. As seen in Fig. 11, the thickness after In this section, we briefly present a comparative analysis of a
nine electrodes (h ¼ 36 cm) is equal to 8 mm for i = 500 A/m2, few commonly used turbulence models for predicting the gas vol-
whereas the gas layer reaches the diaphragm surface (i.e. ume fraction in the channel. For this purpose, we use the standard
dg =8 mm) for h P 28 cm, h P 20 cm and h P 12 cm for i = 1500, k  e and the so-called SST k  x models for turbulence modeling
i = 3250 and 6250 A/m2, respectively. The predicted total volume and compare the results with the RSE model.
fraction of the gas, ag;tot in the cathode cell is shown in Fig. 12, Distributions of the gas void fraction along the cross-section of
which depicts that the total gas hold-up, the gas fraction averaged the channel at h ¼ 32 cm predicted by the selected turbulence
over the entire computational domain, is equal to models are shown in Fig. 13a. Generally, the k  e model predicts
ag;tot ¼ 0:5%; 1:1%; 2:4% and 4:6% for i = 500, 1500, 3250 and a slightly lower gas fraction compared to the RSE model which is
6250 A/m2, respectively. most evident at the electrode surface. This difference becomes lar-

Fig. 13. a) Distribution of gas void fraction across h ¼ 32 predicted by different turbulence models for i ¼ 1500 A/m2 and b) streamwise velocity of the gas phase and c)
turbulent viscosity profiles along the cross-section of the channel.
 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926 13

ger at higher current densities. In contrast, the SST k  x model Acknowledgment

predicts a much larger value for the gas fraction at the electrode
compared to RSE model. Far away from the electrode, the SST This work was done as a part of the Alkaliflex project funded by
k  x predicts a lower gas fraction than the k  e model, so that Rijksdienst Voor Ondernemen Nederland (RVO) and Nouryon and
both these models produce larger deviations from the experimen- as part of the TKI project ‘‘Mass transfer in water electrolysis”, also
tal data than the RSE model does. funded by RVO and Nouryon. We thank M.T. de Groot for writing
The streamwise component of the gas velocity for the selected the Alkaliflex proposal and leading that project, and we thank A.
turbulence models is shown in Fig. 13b. Compared to the RSE M. Meulenbroek, M.T. de Groot and J. van der Schaaf for
and k  e models, the gas velocity predicted by the SST k  x discussions.
model tends more toward the electrode surface due to the larger
amount of gas predicted in the vicinity of the electrode. The distri-
bution of the turbulence viscosity of the liquid phase predicted by References
each model is shown in Fig. 13c. It can be seen that the SST k  x Abdelouahed, L., Hreiz, R., Poncin, S., Valentin, G., Lapicque, F., 2014a.
model results in smaller turbulence viscosity which in turn leads to Hydrodynamics of gas bubbles in the gap of lantern blade electrodes without
lower dispersion of the gas phase and consequently a stronger forced flow of electrolyte: Experiments and cfd modelling. Chem. Eng. Sci. 111,
255–265.
accumulation of gas near the electrode as shown in Fig. 13a.
Abdelouahed, L., Valentin, G., Poncin, S., Lapicque, F., 2014b. Current density
distribution and gas volume fraction in the gap of lantern blade electrodes.
Chem. Eng. Res. Des. 92, 559–570.
4. Conclusions Aldas, K., 2004. Application of a two-phase flow model for hydrogen evolution in an
electrochemical cell. Appl. Mathe. Comput. 154, 507–519.
Alexiadis, A., Dudukovic, M., Ramachandran, P., Cornell, A., Wanngård, J., Bokkers, A.,
Accurate prediction of the gas volume fraction distribution is 2011. Liquid–gas flow patterns in a narrow electrochemical channel. Chem. Eng.
critical for understanding the hydrodynamics of the multiphase Sci. 66, 2252–2260.
flow in electrolyzers. A wide variety of CFD modeling options is Boissonneau, P., Byrne, P., 2000. An experimental investigation of bubble-induced
free convection in a small electrochemical cell. J. Appl. Electrochem. 30, 767–
available, and there is not yet consensus on what the most suit- 775.
able model is. Many available models suffer from a lack of gener- Burns, A.D., Frank, T., Hamill, I., Shi, J.-M., et al., 2004. The favre averaged drag model
ality and are highly dependent on tunable coefficients. In this for turbulent dispersion in eulerian multi-phase flows. In: 5th International
Conference on Multiphase Flow, ICMF, vol. 4, ICMF, pp. 1–17.
paper we tried to predict and validate the hydrodynamics of
Caire, J., Espinasse, G., Dupoizat, M., Peyrard, M., 2009. A hydraulic model to
gas–liquid flow in an electrolyser by including the effects of most simulate the hydrodynamics of a fluorine electrolyser. WIT Trans. Eng. Sci. 65,
important inter-phase forces, i.e drag, and turbulent dispersion 23–34.
Chan-Mou, T., 2013. Mean Value and Correlation Problems Connected with the
forces. Our study showed that the turbulence dispersion force
Motion of Small Particles Suspended in a Turbulent Fluid. Springer.
is of major importance for simulating the gas–liquid flow in an Charton, S., Janvier, J., Rivalier, P., Chaınet, E., Caire, J.-P., 2010. Hybrid sulfur cycle
electrolyser, as without the turbulence dispersion force, the for H2 production: A sensitivity study of the electrolysis step in a filter-press
model is not able to predict the spreading of the gas layer in cell. Int. J. Hydrogen Energy 35, 1537–1547.
Cheng, X., Shi, Z., Glass, N., Zhang, L., Zhang, J., Song, D., Liu, Z.-S., Wang, H., Shen, J.,
the cell correctly. Using user-defined versions for the drag and 2007. A review of pem hydrogen fuel cell contamination: Impacts, mechanisms,
turbulence dispersion forces in the model, we computed gas void and mitigation. J. Power Sources 165, 739–756.
fractions across the channel for different heights of the electrode De Souza, R.F., Padilha, J.C., Gonçalves, R.S., De Souza, M.O., Rault-Berthelot, J., 2007.
Electrochemical hydrogen production from water electrolysis using ionic liquid
and also for various current densities, and we obtained fairly as electrolytes: towards the best device. J. Power Sources 164, 792–798.
good qualitative agreements with experimental data from litera- El-Askary, W., Sakr, I., Ibrahim, K., Balabel, A., 2015. Hydrodynamics characteristics
ture. The user-defined implementation of the forces was intro- of hydrogen evolution process through electrolysis: Numerical and
experimental studies. Energy 90, 722–737.
duced in order to satisfy the gas momentum balance in Fluent. Gibson, M., Launder, B., 1978. Ground effects on pressure fluctuations in the
This balance was not satisfied when the standard built-in forces atmospheric boundary layer. J. Fluid Mech. 86, 491–511.
in Fluent were used. As the turbulence dispersion model relies Haug, P., Kreitz, B., Koj, M., Turek, T., 2017. Process modelling of an alkaline water
electrolyzer. Int. J. Hydrogen Energy 42, 15689–15707.
on the model for the turbulent viscosity, a comparison of results
Hawkes, G., O’Brien, J., Stoots, C., Hawkes, B., 2009. 3D CFD model of a multi-cell
obtained with different turbulence models was included. As a high-temperature electrolysis stack. Int. J. Hydrogen Energy 34, 4189–4197.
next step for the presented research, it is important to validate Hreiz, R., Abdelouahed, L., Fuenfschilling, D., Lapicque, F., 2015. Electrogenerated
bubbles induced convection in narrow vertical cells: A review. Chem. Eng. Res.
the Eulerian model at high gas volume fractions, but the main
Des. 100, 268–281.
problem is to find/obtain detailed experimental data for these Hreiz, R., Abdelouahed, L., Fuenfschilling, D., Lapicque, F., 2015. Electrogenerated
conditions. bubbles induced convection in narrow vertical cells: Piv measurements and
euler–lagrange cfd simulation. Chem. Eng. Sci. 134, 138–152.
Ishii, M., Hibiki, T., 2010. Thermo-fluid Dynamics of Two-phase Flow. Springer
Science & Business Media.
CRediT authorship contribution statement Lattin, W., Utgikar, V.P., 2006. Transition to hydrogen economy in the united states:
a 2006 status report. Int. J. Hydrogen Energy 32, 3230–3237.
Launder, B.E., 1989. Second-moment closure and its use in modelling turbulent
A. Zarghami: Data curation, Formal analysis, Investigation, industrial flows. Int. J. Numer. Meth. Fluids 9, 963–985.
Methodology, Software, Validation, Visualization, Writing - origi- Launder, B.E., Spalding, D.B., 1983. The numerical computation of turbulent flows.
In: Numerical Prediction of Flow, Heat Transfer, Turbulence and Combustion,
nal draft. N.G. Deen: Funding acquisition, Methodology, Resources, Elsevier, 1983, pp. 96–116.
Supervision, Writing - review & editing. A.W. Vreman: Conceptu- Liu, K., Song, C., Subramani, V., 2010. Hydrogen and Syngas Production and
alization, Formal analysis, Funding acquisition, Investigation, Purification Technologies. John Wiley & Sons.
Mat, M.D., Aldas, K., Ilegbusi, O.J., 2004. A two-phase flow model for hydrogen
Methodology, Resources, Software, Supervision, Validation, Writ-
evolution in an electrochemical cell. Int. J. Hydrogen Energy 29, 1015–1023.
ing - review & editing. McGlade, C., 2012. A review of the uncertainties in estimates of global oil resources.
Energy 47, 262–270.
Mueller-Langer, F., Tzimas, E., Kaltschmitt, M., Peteves, S., 2007. Techno-economic
assessment of hydrogen production processes for the hydrogen economy for the
Declaration of Competing Interest short and medium term. Int. J. Hydrogen Energy 32, 3797–3810.
Norby, R.J., Luo, Y., 2004. Evaluating ecosystem responses to rising atmospheric Co2
The authors declare that they have no known competing finan- and global warming in a multi-factor world. New Phytol. 162, 281–293.
Pope, S.B., 2001. Turbulent flows.
cial interests or personal relationships that could have appeared Riegel, H., Mitrovic, J., Stephan, K., 1998. Role of mass transfer on hydrogen
to influence the work reported in this paper. evolution in aqueous media. J. Appl. Electrochem. 28, 10–17.
14 A. Zarghami et al. / Chemical Engineering Science 227 (2020) 115926

Santos, D.M., Sequeira, C.A., Figueiredo, J.L., 2013. Hydrogen production by alkaline van der Hoef, M.A., Ye, M., van Sint Annaland, M., Andrews, A., Sundaresan, S.,
water electrolysis. Química Nova 36, 1176–1193. Kuipers, J., 2006. Multiscale modeling of gas-fluidized beds. Adv. Chem. Eng. 31,
Sato, S., Lin, S.-Y., Suzuki, Y., Hatano, H., 2003. Hydrogen production from heavy oil 65–149.
in the presence of calcium hydroxide? Fuel 82, 561–567. van der Hoef, M.A., van Sint Annaland, M., Deen, N., Kuipers, J., 2008. Numerical
Schiller, L., 1933. A drag coefficient correlation. Zeit. Ver. Deutsch. Ing. 77, 318–320. simulation of dense gas-solid fluidized beds: a multiscale modeling strategy.
Steinfeld, A., 2002. Solar hydrogen production via a two-step water-splitting Annu. Rev. Fluid Mech. 40, 47–70.
thermochemical cycle based on zn/zno redox reactions. Int. J. Hydrogen Energy Vogt, H., 2012. The actual current density of gas-evolving electrodesnotes on the
27, 611–619. bubble coverage. Electrochim. Acta 78, 183–187.
Takeuchi, J., Satake, S.-I., Morley, N.B., Kunugi, T., Yokomine, T., Abdou, M.A., 2008. Vogt, H., Balzer, R., 2005. The bubble coverage of gas-evolving electrodes in
Experimental study of mhd effects on turbulent flow of flibe simulant fluid in stagnant electrolytes. Electrochim. Acta 50, 2073–2079.
circular pipe. Fusion Eng. Des. 83, 1082–1086. Walker, R.D. Jr., 1971. A study of gas solubilities and transport properties in fuel cell
Trommer, D., Noembrini, F., Fasciana, M., Rodriguez, D., Morales, A., Romero, M., electrolytes.
Steinfeld, A., 2005. Hydrogen production by steam-gasification of petroleum Wilcox, D.C., 2006. Turbulence Modeling for CFD. Dcw Industries, Inc., La Canada,
coke using concentrated solar poweri. thermodynamic and kinetic analyses. Int. Ca.
J. Hydrogen Energy 30, 605–618. Zeng, K., Zhang, D., 2010. Recent progress in alkaline water electrolysis for hydrogen
Turner, J.A., 2004. Sustainable hydrogen production. Science 305, 972–974. production and applications. Prog. Energy Combust. Sci. 36, 307–326.