Separation and Purification Technology 261 (2021) 118235

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Modelling and simulation of H2-H2O bubbly flow through a stack of three

cells in a pre-pilot filter press electrocoagulation reactor
Miguel A. Sandoval a, b, Rosalba Fuentes a, Tzayam Pérez a, Frank C. Walsh c, José L. Nava d, *,
Carlos Ponce de León c
a
Universidad de Guanajuato, Departamento de Ingeniería Química, Norial Alta S/N, 36050 Guanajuato, Guanajuato, Mexico
b
Universidad de Santiago de Chile USACH, Facultad de Química y Biología, Departamento de Química de los Materiales, Laboratorio de Electroquímica Medio
Ambiental, LEQMA, Casilla 40, Correo 33, Santiago, Chile
c
Electrochemical Engineering Laboratory, Energy Technology Research Group, Engineering Sciences, University of Southampton, Highfield, Southampton, SO17 1BJ,
United Kingdom
d
Universidad de Guanajuato, Departamento de Ingeniería Geomática e Hidráulica, Av. Juárez 77, Zona Centro, 36000 Guanajuato, Guanajuato, Mexico

A R T I C L E I N F O A B S T R A C T

Keywords: Computational fluid dynamics simulations were carried out to describe the hydrodynamic characteristics of a
Biphasic turbulent flow two-phase bubbly flow in a filter-press flow reactor stack of three cells, which is typically used in electro­
Current distribution coagulation (EC). The hydrogen evolution reaction (HER) took place at the cathode; dissolution of aluminium
Euler-Eulerian approach
occurred at the anode. The fundamental transport equations of momentum and electrical potential were
Hydrogen evolution reaction
Multi-electrode stack
simultaneously solved to simulate the H2-H2O flow. Continuous (H2O) and dispersed phase (H2) velocity fields
were modelled via the Euler-Eulerian approach, using the biphasic Reynolds Averaged Navier-Stokes (RANS)
equations and the standard k − ε turbulence model. The influence of volumetric flow rate (1.7 ≤ Q ≤ 15 cm3 s− 1)
and applied current density (–28 ≤ j ≤ –5 mA cm− 2) was systematically addressed to calculate the fraction of
dispersed phase and current distribution along the electrodes. The evolved H2 bubbles were transported away
from the electrode by the liquid flow. The dispersion of H2 through the electrode gap showed a modest bubble
curtain profile due to the liquid flow rate. A homogeneous current distribution along the electrode length was
experienced due to the geometrical design of the electrochemical cell and the low degree of H2 dispersion. The
velocity profiles of the H2-H2O mixture were different in each cell due to the change of flow direction. H2 bubbles
increased the velocity of the liquid phase but the gas fraction of such bubbles resulted in a higher pressure drop.
Good agreement between theoretical and experimental residence time distribution curves was achieved; the
experimental aluminium dose released by the anode agreed well with the simulations.

1. Introduction counter electrode process is usually the hydrogen evolution reaction

(HER).
Filter-press electrochemical reactors with vertical parallel plate In vertical parallel plate electrode reactors, electrode off gases are
electrodes are commonly used in industrial practice due to their effec­ released into the electrolyte as small, dispersed bubbles, whose behavior
tiveness in a diverse range of applications such as electrosynthesis of and significance in the vicinity of the electrode depend on the current
chemicals, metal ion removal, energy storage, environmental remedia­ density [4]. When their local concentration exceeds saturation, bubbles
tion, and drinking water treatment [1–3]. Many industrial processes form by heterogeneous nucleation. Once the bubbles reach a sufficient
involve gas bubble generation as the main, side, or counter-electrode size, they leave the electrode surface, moving upwards under buoyancy
reaction [4]. For example, in the chloro-alkali process, chlorine and to form a two-phase layer near the electrode, namely, the ‘bubble cur­
hydrogen gases are produced at the anode and cathode, respectively. In tain’ [4,5]. The gas released into the interelectrode gap in the cell affects
electrocoagulation (EC), anodic dissolution of metal takes place and the the liquid flow distribution and ionic transport to the electrode surfaces,

* Corresponding author.
E-mail addresses: [email protected], [email protected] (M.A. Sandoval), [email protected] (R. Fuentes), [email protected] (T. Pérez), F.
[email protected] (F.C. Walsh), [email protected] (J.L. Nava), [email protected] (C. Ponce de León).

https://doi.org/10.1016/j.seppur.2020.118235
Received 26 September 2020; Received in revised form 29 November 2020; Accepted 16 December 2020
Available online 25 December 2020
1383-5866/© 2020 Elsevier B.V. All rights reserved.
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

which can have a considerable effect on cell performance [6]. Moreover, simultaneously solved. In the same manner, Al3+ concentrations were
the liquid–gas mixture in the electrode gap determines the apparent obtained solving the averaged diffusion-convection equation. The pre­
electrolytic conductivity and the current and potential distributions [7]. sent study systematically addressed the influence of applied current
The later also affects the current efficiency and energy consumption density and volumetric flow rate on the dispersed phase fraction and
during electrolysis [8]. current distribution along the electrodes. Experimental aluminium ion
Two main approaches have been used to simulate liquid–gas flow in concentration and residence time distribution were compared with CFD
electrochemical reactors and interactions between gas and liquid pha­ simulations.
ses: the Euler-Eulerian (EE) and Euler-Lagrangian (EL) methods [9]. The In this work, the formation of the aluminium floc, such as Al(OH)3(s)
latter solves Newton’s second law of motion for the location of an in­ and Al2O3(s), is not modeled, since the reactor operates in the continuous
dividual fluid particle to identify its possible pathways, in the former mode and coagulation-flocculation is carried out outside the reactor, in a
approach, the phases are treated as a continuum providing average flocculator-clarifier. The modelling of pollutant removal forms part of a
quantities [10]. The modelling and simulation of two phase (gas–liquid) future research program. The modelling of bubble rupture and coales­
flow in electrochemical reactors have been carried out by means of EL cence are beyond the scope of this paper; future research should
and EE computational fluid dynamics (CFD) approaches using com­ consider these important effects.
mercial and open software, which provides clear information and an
understanding of biphasic phenomena [11–18]. Euler-Eulerian and 2. Methods and materials
Euler-Lagrangian CFD models have been used to characterize the
fundamental physics of turbulent gas–liquid bubble flows [6] but it is 2.1. Description of the reactor and electrode reactions
important to emphasize that theoretical studies of gas–liquid flow in
multi-electrode filter press flow electrolysers has not been published in Fig. 1 shows the FM01-LC cell with a stack of three cells; the detailed
the literature. description of the reactor can be consulted elsewhere [19,23]. It is
In our previous work, CFD simulations were carried out for single- important to highlight that during the EC process, the massive produc­
phase flow in a pre-pilot filter-press flow reactor using a stack of three tion of H2 bubbles at the cathode has a negative effect on the process as
cells (16 cm × 4 cm × 0.6 cm each cell), where the variation of the cross- they cause breaking of the flocs [24]. The characterization of H2-H2O
sectional area induces velocity field modifications in an irregular flow opens an opportunity to optimize the reaction environment in
manner, generating jet flows, vortex and rotational flow structures, so- future cells.
called turbulent eddies [19]. Turbulence models can be used to accu­ The electrochemical reactions can be concisely stated (potentials
rately simulate flow patterns in complex geometries [13,18–22]. being stated with respect to the standard hydrogen electrode (SHE)):
This paper involves the numerical simulation of H2 gas evolution at Hydrogen gas is evolved at the aluminium cathodes:
the cathode and release of Al3+ ions at the anode, in a filter-press flow
3H2 O + 3e− →1.5H2 + 3OH − E = − 0.830 V vs. SHE (1)
reactor stack composed of three cells, which is typically used in elec­
trocoagulation (EC) [23]. The biphasic flow system is considered simply while anodic dissolution generates aluminium cations:
as H2-H2O. The two-phase flow is solved by the biphasic Reynolds
Al(s) →Al3+ + 3e− E = − 1.662 V vs. SHE (2)
Averaged Navier-Stokes (RANS) equations with the k-ε turbulence
model through the EE approach. As the rate of the HER depends on the in preference to the oxygen evolution reaction (OER).
local current distribution, the momentum and Laplace equations were

Fig. 1. Sketch of the FM01-LC reactor with a four-electrode stack of three undivided cells. The electrodes are switched in monopole configuration.

2
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

H2 O→2H+ + 0.5O2 + 2e− E = 1.229 V vs. SHE (3) manometer, using a water/gel mixture as the fluid. The manometer was
3+ connected by silicone tubing to a T-piece located as close as possible to
Both H2 and Al are convectively transported away from the elec­
the reactor inlet or outlet.
trode surfaces to the bulk solution by the electrolyte flow provided by a
centrifugal pump.
The electrodes were mounted in a vertical position and they were 3. Formulation of the numerical simulation
electrically connected in a monopolar configuration. Table 1 shows the
geometrical parameters of the filter-press multielectrode stack reactor. The H2-H2O bubbly flow is produced on one hand by the HER at the
Four aluminium plate electrodes (99.7% purity), which were used as cathode, and by the volumetric liquid inflow rate (Q) imposed on the
anodes and cathodes, were fitted between polypropylene separators multielectrode stack. In this context, the momentum and Laplace
acting as fluid distributors to form a stack of three cells. equations must be solved simultaneously to simulate the biphasic flow
The electrolytic solution flowed upwards past the vertical electrodes characteristics. Continuous (H2O) and dispersed phase (H2) velocity
within three parallel rectangular flow channels, Fig. 1. The electrolyte fields were modelled via the Euler-Eulerian approach as described
was fed through a 1.3 cm diameter circular tube which acted on the front below.
face of the first fluid distributor, at the back of the third cell. Each fluid The volumetric liquid inflow rates (Q) studied here were 8.3, 10,
distributor contains five inlet manifolds which take the electrolyte 11.7, 13.3 and 15 cm3 s− 1 allowing mean linear liquid flow velocities (u,
across the interelectrode gap (0.6 cm). At the top, five outlet manifolds evaluated using the cross-sectional area of the three parallelepiped flow
direct the electrolyte out through a 1.3 cm diameter circular tube. The channels) of 1.2, 1.4, 1.6, 1.8, and 2.1 cm s− 1. These mean linear flow
total length of the bottom and upper linking pipe (between poly­ velocities correspond to Reynolds numbers (Re = udh/υ) ranging from
propylene separators and electrodes) was 7 and 10 cm, respectively. 125 to 219 for the continuous phase, where dh is the hydraulic diameter
of the rectangular channel and υ is the kinematic viscosity of the elec­
trolyte. In the multi-electrode stack reactor, there are abrupt changes in
2.2. Experimental details. the cross-sectional area and the direction of the flow, which create
variations of the velocity field in a random fashion. In that connection,
2.2.1. Residence time distributions experiments. the H2-H2O flow was simulated solving the biphasic RANS equations
Residence time distributions (RTD) measurements were performed with the k-ε turbulence model via the EE approach, even at such low Re
at volumetric inflow rates comprised between 8.3 ≤ Q ≤ 15 cm3 s− 1 values (〈2 5 0). Such an approach was guided by previous literature
during the electrolysis trials performed at constant current density of − [12,18–22].
7 mA cm− 2, where the H2 bubbles are generated at the cathode by means The following assumptions were considered to model and simulate
of Eqn (1). Drinking water containing 1 mg dm− 3 NaOCl at pH 7 with an the two-phase flow: i) an Euler-Eulerian CDF approach to solve the
electrolytic conductivity value of 540 µS cm− 1 was used in RTD mea­ average volume fraction, in our case, average dispersed fraction less
surements. Tracer injections of 1 cm3 of Na2SO4 (1 M) were made at the than 3.9%, ii) two continuous and fully interpenetrating, and incom­
reactor entrance and the conductivity was determined at the reactor exit pressible phases, iii) the Euler-Eulerian CDF approach solves transport
with an Oakton meter (ECTestr type). equations for the turbulence quantities using a mixture averaged tur­
The stimulus–response technique was applied to determine the RTD bulence model (RANS, standard k − ε turbulence), iv) the drag model
of the mixture H2-H2O. A schematic diagram of the system used for the (phases interchange momentum) is described by the Shiller-Naumann
experimental RTD and pressure drop measurements is described else­ method which is valid for rigid spheres, i.e. not consider coalescence
where [19]. 1 cm3 of Na2SO4 (1 M) was injected by a syringe at 1.5 cm and break up phenomena, v) the dynamic viscosity of Krieger type is
from the reactor inlet, with an injection time of approximately 1 s. At the used which is appropriate for dispersed phase that not form any pure
reactor outlet, the Na2SO4 concentration was quantified online by the phase region. A brief description of the model with key equations, based
conductivity meter reading the conductivity values every second. This on these assumptions, is given in Section 3.2.
technique was sufficiently sensitive to capture the electrolyte conduc­
tivity changes at the reactor outlet.
3.1. Current distribution model.
2.2.2. Aluminium ion concentration
Aluminium concentrations were quantified by atomic absorption In dilute solutions, the current density vector, j, can be calculated
(after addition of sulfuric acid to reach pH = 2) using a Perkin Elmer from the local potential gradient, according to Ohm’s law [25]:
AAnalyst™ 200 atomic absorption spectrometer at a wavelength of
(4)
309.27 nm and with a detection limit of 0.1 mg L− 1.
j = − keff ∇ϕ

where keff is the effective electrolytic conductivity which, in the case of

2.2.3. Pressure drop experiments
biphasic flow, depends on the pure electrolyte conductivity, k0, modified
Pressure drop measurements were performed under similar experi­
by the fraction of the dispersed phase (αg ) according to the Bruggeman
mental conditions to that employed for RTD trials. The pressure drop
equation [26]:
measurements were performed through a 60-cm U-tube vertical glass
keff = k0 (1 − αg )3/2 (5)
Table 1
Geometrical parameters of the electrochemical reactor stack containing three
The electric potential, ϕ at any point in the cell can be described by
undivided cells. the Laplace equation [20]:
Volume*, V /cm3 130 ∇∙(keff ∇ϕ) = 0 (6)
Cell width, B /cm 4.0 Hydrogen gas released at the cathode, reaction (1), and the elec­
Cell thickness, S /cm 0.6
Cell length, L /cm 16.0
trodissolution of aluminium at the anode, reaction (2), are represented
Number of cells 3 by a secondary current distribution model, since water and aluminium
Equivalent diameter of flow cell thickness, dh = 2BS/(B + S) /cm 1.04 are in excess. Therefore, the boundary conditions used to solve equation
Anode area, in each cell, in contact with solution /cm2 64 (6) are expressed as follows:
Cathode area, in each cell, in contact with solution /cm2 64
At the cathode and anode, the overpotential is adequately related to
*Including the volume of the cell connecting pipes plus fluid manifolds. the magnitude of local current density through the Tafel approximation:

3
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

( )
∂ϕ − η 3.2.1. Biphasic turbulence model
− keff = − jo,c exp (7)
∂ξ bc In this paper, the standard κ − ε turbulence model was used to
calculate the fluid flow turbulence [28,29]. The transport of the turbu­
( )
∂ϕ η lence quantities (the turbulent kinetic energy, κ, and the turbulent en­
− keff = jo,a exp (8)
∂ξ ba ergy dissipation rate, ε) which are based on the mixture velocity,
[ ( )]
um = (1/ρm ) ul αl ρl + ug αg ρg where ρm is the mixture density (ρm =
where ξ is the distance normal to the electrode surface, jo,c and jo,a are αl ρl + αg ρg ) and μm is the volume averaged mixture viscosity (μm =
the exchange current densities for the cathodic and anodic reactions,
αl μl + αg μg ), stated according the Eqs. (15)–(16).
while bc and ba are the cathodic and anodic Tafel slopes, respectively.
(( ) )
Finally, the local overpotential (η) at the electrodes describes the local μ
ρm um ∙∇κ = ∇∙ μm + T ∇κ + Pκ − ρm ε (15)
current density and is related to other potentials by: σκ
η = ϕ m − ϕ − E0 (9) (( ) )
μ ε ε2
ρm um ∙∇ε = ∇∙ μm + T ∇ε + Cε1 Pκ − Cε2 ρm (16)
where ϕ is the potential of the solution adjacent to the electrode and, ϕm σε κ κ
and E0 are the metal and equilibrium electrode potentials.
where Pκ is the energy production term (Pκ =
For the insulating walls, zero flux: [ ( ) ]
∂ϕ μT ∇um : ∇um + (∇um )T − 23(∇∙um )2 − 23ρm κ∇∙um ), and Cµ (0.09),
− keff =0 (10)
∂ξ Ce1 (1.44), Ce2 (1.92), σk (1), σ ε (1.3) are dimensionless constant values
Variables ϕ and αg were solved simultaneously by coupling the obtained by data fitting over a wide range of turbulent flow conditions
Ohm’s law, using an Euler-Eulerian biphasic approach, as explained [30].
below. The resulting turbulent viscosity is defined as:

κ2
μT = ρ m C μ (17)
ε
3.2. The Euler-Eulerian CFD approach.
The total stress tensors for each phase are:
( )
A statistical description of multiphase flow is useful to characterize 2 2
σ l = (μl + μT ) ∇ul + (∇ul )T − (∇ul )I − ρl κI (18)
the dispersed phase particles. In statistical theories of single-phase tur­ 3 3
bulent flow, the Eulerian velocity is represented by a random vector
( )
field. Unlike single-phase flow, velocity and pressure fields, even in ( )T 2( ) 2
σ g = (μg + μT ) ∇ug + ∇ug − ∇ug I − ρ κI (19)
laminar multiphase flow, show variability and are significantly repre­ 3 3 g
sented by random fields [27]. An analogous methodology can be
assumed for biphasic (gas–liquid) flows. However, it is necessary to where I is the unity tensor, μ is the dynamic viscosity, and μT is the
specify the location and shape of the dispersed phase (hydrogen bub­ turbulent viscosity defined according the k − ε turbulence model. The
bles). In the Euler–Eulerian approach, both gas and liquid phases are notation ()T indicates the transpose of ∇u, which should not be confused
treated as a continuum, which can interpenetrate with the other fluid with a turbulent suffix.
phase. Finally, assuming mixture turbulence, the transport equation for the
The theoretical determination of the distribution of bubbles, and dispersed phase fraction is:
mixture velocity in the inter-electrode space were calculated by solving, ( )
∇∙ αg ug = ∇∙(Dmg ∇αg ) (20)
simultaneously, the Laplace, and the biphasic RANS equations with the
k-ε turbulence model. For all simulations, an isothermal system was
where Dmg ( = μT /(ρm Scb )) is the turbulent diffusion coefficient (addi­
supposed and, consequently, the energy equation was not considered. In
tional diffusion produced by turbulent vortexes) which is defined from
this paper, the mass transfer between phases and chemical reactions
the turbulent viscosity and the turbulent bubble Schmidt number (Scb =
were disregarded.
0.35) [31].
The momentum and continuity equations for continuous and
dispersed phases in the steady state are:
3.2.2. Viscosity model and interphase momentum transfer.
[ ]
∂ Empirical and analytical models for the dynamic viscosity of the two-
αl ρl (ul ) + ul ∇∙(ul ) = − αl ∇P + αl ρl g + αl ∇∙σ l + Mi,l (11) phase mixture (μlg ) have been developed usually as a function of the
∂t
dispersed volume fraction. It is assumed that turbulence effects in both
∂ continuous and dispersed phases can be modeled by solving the rela­
(α ρ ) + ∇∙(αl ρl ul ) = 0 (12)
∂t l l tionship that describe the turbulence in the mixture. A simple dynamic
[ ] mixture viscosity expression that covers a wide range of particle con­
∂( ) ( )
centrations is the Krieger type model [32]:
αg ρg u + ug ∇∙ ug = − αg ∇P + αg ρg g + αg ∇∙σ g + Mi,g (13)
∂t g
( )− μg +0.4μ
2.5αg,max μ +μ l
αg g l
∂( ) ( ) μlg = μl 1 − (21)
αg ρg + ∇∙ αg ρg ug = 0 (14) αg,max
∂t
where αg,max is the maximum packing limit (equal to 1 for bubbles).
where u is the averaged velocity vector, P is the pressure shared by two
As described in Section 3.2, there are several interaction forces be­
phases, ρ is the density, σ is the total stress tensor and Mi is the interface
tween the two phases during the interphase momentum exchange.
momentum exchange, which is usually divided into several different
Nevertheless, the main interaction force is due to the drag force caused
components, such as drag force, virtual mass forces, interfacial pressure,
by the slip between phases [33]. The interphase force contemplated in
lift force, and the Basset force [6]. The relationship between the volume
this study is the drag force between liquid and gas phases, while other
fractions can be expressed as: αl = 1 − αg . The subscripts l and g in the
forces (virtual mass force, turbulent dispersion force, lift force, and the
equations are referred to water (continuous phase) and hydrogen
Basset force) were neglected. It is reported that adding these forces did
(dispersed phase), respectively.

4
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

not bring any obvious refinement to the simulation results instead, only electron stoichiometry (=3) according to Eq. (1), and F is the Faraday
provokes convergence difficulties [13]. Bubble phenomena, such as constant (96485 C mol− 1).
bubble breakup and coalescence, are not considered in the mathematical • At the anode, the boundary condition is considered in Section 3.4.1.
model proposed here. Future research should consider these • A normal stress equal to the pressure at the outlet,
[ ( ) ]
phenomena. − P +(μ + μT ) ∇um + (∇um )T ∙n = − P0 ∙n, where P0 is the
The drag force between phases added to the momentum equations
pressure at the electrolyte exit. This equation represents that the
(11, 13), is represented by [34]:
( ) turbulent characteristic of whatever is outside the computational
MDl = − MDg = β ug − ul (22) domain is guided by the flow inside the computational domain [38].
Such assumption is physically reasonable as long as relatively small
where MD is the drag force and β is the drag force coefficient: amounts of fluid enter the system. Moreover, at the electrolyte outlet,
⃒ ⃒ ∇ε∙n = 0 and ∇κ∙n = 0.
3 CDl,g
β= αg αl ρl ⃒ug − ul ⃒ (23) • For all other boundaries, a velocity u+ given by Eq. (26). The value of
4 db
y+ was set at 11.1 after validating the solution at different values of
where db is the bubble diameter and CDl,g is the drag coefficient which is y+ and step sizes. This value is in the turbulent region, where the
turbulent stresses and fluxes are more important [39].
determined by the Schiller-Naumann correlation [33]:
24 ( )
CDl,g = 1 + 0.15Re0.687 , Rep ≤ 1000 (24)
Rep p
3.3. Simulation of the residence time distribution
In general, the drag coefficient is a function of the particle Reynolds
The tracer mass transport inside the multielectrode stack, in tran­
number (Rep), defined by [35]:
sient and turbulent regime, was simulated using the averaged con­
ρl db ug − ul vection–diffusion equation (Eq. (27)). The mass transport was simulated
Rep = (25)
μl at different volumetric liquid inflow rates of 8.3, 10, 11.7, 13.3 and 15
cm3 s− 1.
Electrogenerated bubbles are not similar. The bubble diameter is
affected by factors such as physicochemical properties (e.g., superficial ∂C
= − um ∙∇C + ∇∙(D + DT )∇C (27)
tension), operating parameters (e.g., type of electrode) and hydrody­ ∂t
namic conditions such as the geometry of the reactor [4]. The Euler-
Eulerian CFD approach assumes homogeneous bubble distribution where C is the average concentration of tracer, t is the time, D is the
within each control volume. The bubble diameter employed in this diffusion coefficient of the tracer, and DT is the turbulent diffusivity
paper was set at a typical value of 50 × 10− 2 cm, which was obtained which can be obtained from the turbulent Schmidt number (ScT =
from experimental studies and has been widely used for numerical μT /(ρm DT )) described by the Kays–Crawford model [19,29] according to
simulations [4,16]. Eq. (28). The mixture velocity vector is obtained by the solution of
momentum equations.
3.2.3. Boundary conditions (
1 0.3 μ
(
μ
)( (
ρm D
) ) )− 1
The turbulence model is applicable for disordered flows, and even for ScT = + √̅̅̅̅̅̅̅̅̅̅ T − 0.3 T 1 − exp − √̅̅̅̅̅̅̅̅̅̅
2ScT∞ ScT∞ ρm D ρm D 0.3μT ScT∞
small Reynolds numbers, as is the case study in this paper, where the
(28)
fluid flow in the multielectrode stack is disturbed. In this context, the
near-wall zones, where the velocity decreases promptly, are inaccessible where the Schmidt number at infinity (ScT∞) takes a value of 0.85 [28].
by this model. To overcome this problem, the logarithmic wall function In this paper, Na2SO4 was employed as a tracer, the diffusion coefficient
is employed. of sulphate ions, DSO2− = 0.08 × 10–9 m2 s− 1 [40].
4

u+ = 2.5lny+ + 5.5 (26)

3.3.1. Boundary conditions.
where u+ is the normalized velocity component corresponding to the If we consider a perfect mixing condition both before the inlet and
turbulent layer, y+ is the dimensionless distance from the wall, y+= after the outlet of the reactor, the boundary, and the initial conditions to
√̅̅̅
ρuτy/μ, where uτ is the friction velocity (uτ = C1/4
μ k) and y is the dis­ solve the Eq. (27) can be summarized as follows:
tance from the wall [36].
The boundary conditions to solve Eqs. (4), (11)–(19) are: • At the reactor inlet, − n∙N = u0 C0 , where N is the flux of tracer, C0
and u0 are the initial tracer concentration and the inlet liquid ve­
• At the inlet, αg = 0 with a normal liquid velocity ul = − u⋅n, where n locity. Tracer injection was simulated in the time interval by a
is the unit normal vector and u is the mean linear inflow velocity. In Gaussian pulse function, this function was varied from 3 to 4 s using a
this simulation, the initial values of κ0 and ε0 were obtained from the standard deviation, σ from 1.5 to 2.7. In this study, an initial con­
turbulent intensity (IT), and the turbulent length scale (LT), through centration of sodium sulfate of 1000 mol m− 3 (1 M), was employed as
3/4 a tracer. A Gaussian pulse function was used to simulate the tracer
the following simple forms: κo = (3/2)(ul IT )2 and εo = Cμ κ3/2 /LT .
pulse injection:
The turbulent intensity for fully turbulent flows has dimensionless
( )
values between 0.05 and 0.1, while the turbulent length scale can be 1 − (t− to )2
obtained as a function of the pipe radius by means of LT = 0.07r, C = C0 √̅̅̅̅̅ e 2σ2 (29)
σ 2π
where r is the inlet pipe radius [37]. In this work, IT and LT were
defined at 0.05 and 0.0455 cm, respectively. t > 0 were used for numerical simulations.
• At the cathodes: H2 bubble formation occurs, by means of a dispersed Before the tracer injection (t = 0), the tracer concentration is zero, C0
phase mass flux (ṁd = j⋅m
F⋅z ), with ug = 0, where j is the adjacent local
= 0.
( )
− 1 − 2
current density (C s m ) obtained from Eq. (7) ( − jo,c exp −bcη ), m • At the electrolyte outlet, n∙( − (D + DiT )∇C +um ∙∇C ) = 0, (zero
is the molar mass of hydrogen gas (2.02 × 10–3 kg mol− 1), z is the mass flux).
• For all other boundaries, − n∙N = 0.

5
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

RTD curve, E(t), is obtained from data by normalization, the curve Multiphysics® 5.5 on a computer with two Intel® Xeon ™ 2.30 GHz
describes the tracer distribution in certain time periods at the electrolyte processor, 96 GB of RAM, and 64-bit operating system. Table 2 resumes
outlet [41]: the parameters, electrolyte, and gas properties, employed in the
simulations.
C(t)
E(t) = ∫ ∞ (30) Based on the results of the sensitivity analysis, a fine element size was
C(t) dt
0 set in the computational domain; this mesh built 207273 elements. It is
important to highlight that a high-quality mesh at walls after the
where C(t) is the time-dependent concentration response taking from Eq.
refinement was considered due to the biphasic turbulent model
(27).
employed herein. The simulation run times of Laplace and momentum
A dimensionless function E(θ) can be defined as:
equations were around 720 min depending on the inflow rate and the
E(θ) = τE(t) (31) current density. The simulations of the mass transport problem lasted
around 240 min. The solver employed was iterative, GMRES, and a
This function can be plotted vs. the dimensionless time θ = t/τ,
relative tolerance of accuracy of the CFD simulations applied a
where τ is the spatial residence time (V/Q), and V is the electrolyte
convergence criterion below 1 × 10–5.
volume inside the reactor. In this study, V = 130 cm3, this volume in­
cludes the volume where tracer was injected and where its output was
4. Results and discussion
recorded. More details of the simulation domain can be consulted else­
where [19].
4.1. Simulation of the current distribution

3.4. Simulation of the anodic dissolution of aluminium Fig. 2 displays a cross-sectional plot (in the plane x-y) of the current
distribution on the cathode and anode located in the second cell, at
In this paper, the formation of aluminium flocs are not included in average current density of jave = –7 mA cm− 2 and Q = 8.3 cm3 s− 1. A
the model, since the reactor operates in continuous mode, and the quasi-uniform distribution is observed at the entire electrode surface
coagulation-flocculation processes take place at the exit of the multi­ area except in the vicinity near to the beginning and end, being more
electrode stack, in a flocculator-clarifier [24]. Moreover, the modelling evident at the last one. Fig. 3a) and b) show the normalized current
of the pollutant’s removal, was beyond of the scope of this paper. In this distribution profiles along the x-coordinate from the cathode and anode,
context, we only simulate the concentration of Al3+ considering no respectively, at different average current densities (–7, –14 and –28 mA
chemical reaction and negligible effects of the electrical field. Therefore, cm− 2) at Q = 8.3 cm3 s− 1. As can be seen, at the cathode, the electro­
the mass transport was simulated by using the Eq. (27), in the steady generated hydrogen bubbles affected the current distribution in a
state, at different volumetric liquid inflow rates of 1.7, 3.3, 5, 6.7, and greater fashion (in the order of 10–3) than in the anode (in the order of
8.3 cm3 s− 1. The diffusion coefficient of aluminium ions, D3+ Al = 1.01 × 10–4), close to the exit of the cell, at x/L = 1. As jave increased, at the
10–9 m2 s− 1 [42]. cathode, the current distribution increases at the exit of the cell, at x/L
= 1. This is attributable to the increase in hydrogen bubbles concen­
3.4.1. Boundary conditions tration, as discussed below, giving values of the dispersed phase of
The boundary conditions to solve the Eq. (27) are expressed as: approx. 0.039, 0.076, and 0.144, at current densities of –7, –14, and –28
mA cm− 2, respectively (Table 3). The quasi-uniform secondary current
• At the anodes, the electrogenerated Al3+ occurs at the anode surfaces distribution on the anode was due to the geometry of the reactor. Similar
involves a mass flux , m˙Al = F⋅zj , where j is the local current density
Al
results were found at the cathode and anode surfaces in the first and
1
(C s− m− 2) positively coupled and obtained from Eq. (8),
( )
(=jo,a exp bηa ), and zAl is the electron stoichiometry (=3) according Cathode Anode
to Eq. (2). At this contour u = 0. End
• At the cathode, see boundary condition in Section 3.2.3.
• At the electrolyte outlet, n∙( − (D + DiT )∇C +um ∙∇C ) = 0.
• At all other boundaries, − n∙N = 0.

3.5. Simulation

The Laplace and momentum equations were solved simultaneously

in 3D by the finite element method, using the software COMSOL
j/jave

Table 2
Simulation parameters including liquid and gas properties.
Electrolyte temperature, T/K 298.15
Kinematic viscosity of electrolyte, υ /m2 s− 1 1 × 10–6
Global diffusion coefficient of SO2–
4 , DSO2−
4
/m2 s− 1
[40] 0.08 × 10–9
Global diffusion coefficient of Al3+, DAl/m2 s− 1 [42] 1.01 × 10–9
Dynamic viscosity of electrolyte, μ/Pa s− 1 0.001
Hydrogen density, ρg/kg m− 3 0.083
Bubble diameter, db/cm [4,16] 50 × 10–2
Density of electrolyte, ρl/kg m− 3 1000 x
Electrolytic conductivity, k0/μS cm− 1 540
Anode open circuit potential, Ea, 0 vs SHE/V − 0.68 y Beginning
Cathode open circuit potential, Ec, 0 vs SHE/V 0.57
Cathodic Tafel slope, bc/V dec–1 0.12 Fig. 2. Surface plot (plane xy) of the secondary current distribution on the
Exchange current density, j0/A m− 2 48.5
cathode and anode surface located in cell 2 at jave = –7 mA cm− 2 and Q = 8.3
Anodic Tafel slope, ba/V dec–1 0.28
cm3 s− 1.

6
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

Fig. 3. Influence of the normalized current distribution profiles along the cathode (a) and anode (b) in the x-coordinate at Q = 8.3 cm3 s− 1. Influence of the mean
linear inflow rate along the cathode (c) and anode (d) at jave = − 7 mA cm− 2. The profiles were constructed at y = 2 cm.

Table 3
Averaged dispersed fraction values inside the FM01-LC reactor at different applied current densities and volumetric flows rates.
Volumetric flow rates / cm3 s− 1

1.7 3.33 5 8.33 10 11.7 13.33 15

− 2
jave/mA cm Averaged dispersed fraction

− 5 0.14 0.074 0.049 0.028 0.022 0.019 0.016 0.014

− 6 0.17 0.088 0.058 0.033 0.027 0.023 0.02 0.017
− 7 0.19 0.1 0.068 0.039 0.031 0.026 0.023 0.02
− 14 0.35 0.193 0.131 0.076 0.062 0.052 0.045 0.039
− 28 0.59 0.34 0.239 0.144 0.118 0.1 0.087 0.075

third cells (not shown here). the upper linking pipe acts as a fluid mixer. The mixture velocity
Fig. 3c) (cathode) and 3d) (anode) displays the normalized current magnitude field shows values slightly higher compared to the velocity
distribution on the electrode surfaces at different Q (11.7, 13.3 and 15 magnitude field obtained for a single-phase flow (water) found in a
cm3 s− 1), at jave = –7 mA cm− 2. The current distribution did not depend previous paper by our group [19], which is attributed to the H2 bubbles
significantly on the flow rate, at the cathode and anode, due to the small concentration exerting a drag force. The velocity profiles of the mixture
variation of the dispersed phase values (0.026, 0.023 and 0.020, at 11.7, (H2-H2O) were constructed to analyze the hydrodynamic behavior in­
13.3 and 15 cm3 s− 1, respectively), shown in Table 3. Similar behavior side the three cells within the stack.
was noted at the anode and cathode surfaces of cells 1 and 3 (not Fig. 5 shows the mixture velocity magnitude profiles in the cell width
shown). (y − coordinate) taking at different heights along the cell (x-coordinate),
these profiles were constructed at a depth of 0.3 cm (z − coordinate) in
4.2. Simulation of H2-H2O bubbly flow the inter-electrode gap. The inset in Fig. 5 (at x = 15 cm), displays the
sensitivity analysis bearing in mind the mixture velocity magnitude at
Fig. 4 shows the velocity magnitude field of the H2-H2O mixture the reactor outlet (just in the middle of the pipe). The mixture velocity
was considered unchanged using a fine mesh (207,223 elements).
inside the multielectrode stack at Q = 13.3 cm3 s− 1 and jave = –7 mA
cm− 2; these values were chosen to reflect operating parameters used in At 0.3 cm height, random velocities appeared. However, maximum
values are observed due to the jet flows originated by the inlet manifolds
electrocoagulation [43]. The H2-H2O mixture starts to flow at the
parallelepiped zone, where the electrodes are located (cell length of 16 reaching velocities from 0.19 to 4.2 cm s− 1. The magnitude of the
mixture velocity profiles was different at each cell due to the turbulence
cm); the electrochemical cells initiates at the end of the inlet manifold.
At the electrolyte inlet, the velocity field showed the effects of the created by the change of flow direction and by the electrogenerated
bubbles at each cathode. The mixture velocity magnitude increases ac­
common manifold; these effects are associated with the variations in the
cross-sectional area. Afterwards, in the three-parallelepiped zones, the cording the following order cell 1 < cell 2 < cell 3. The latter was
originated from turbulence generated by the change of flow direction in
mixture velocity remains heterogeneous along the x-coordinate. Higher
velocity magnitudes can be seen in cell 3, which is more evident when the third cell, see bottom flow contour in Fig. 4. From 3 to 8 cm height,
the mixture velocity profiles are attenuated at each cell presenting the
the flow contours are displayed (inset of Fig. 4). At the electrolyte outlet,
the flow contours show that the cell linking pipe causes a spiral flow maximum values at the end of each cell width. After building several
velocity profiles between 8 < x < 15 cm, it is possible to confirm that a
pattern where the maximum values are found at the center, it means that

7
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

Fig. 4. Mixture (H2-H2O) velocity magnitude field inside the stack, at Q = 13.3 cm3 s− 1, j = − 7 mA cm− 2
and bubble diameter db = 50 × 10− 2
cm. The insets
illustrate the flow contours. 70 levels were plotted at the inlet and the exit of the stack.

Fig. 5. Mixture velocity magnitude profiles along the channel width (in the y-coordinate) at different heights in the x-coordinate, evaluated at 0.3 cm of depth in the z-
coordinate. The inset shows the sensitivity analysis using element size from extremely coarse to extremely fine. Data taken from the simulation trials in Fig. 4.

quasi-plug flow pattern is achieved at 15 cm height in each cell. This [44,45]. Moreover, the hydrogen bubbles generated are dispersed
flow pattern was the same for the other Q comprised between 1.7 and 15 throughout the inter-electrode space. Fig. 6 shows the volumetric bubble
cm3 s− 1, at j = –7 mA cm− 2. fraction (percentage) across the channel width (y-coordinate) at a 0.3
In vertical parallel plate reactors, as the flow progresses, the cm depth (z-coordinate), at different heights in the x-coordinate. The
dispersed fraction in the vicinity of the hydrogen-producing cathodes dispersed phase fraction profiles show that the H2 fraction increases in
increases in the upward direction due to the accumulation of the bubbles the upward direction (x-coordinate) due to the accumulation of void

8
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

H2 bubbles with liquid flow rate.

4.3. Theoretical and experimental analysis of RTD

The validation of the mathematical modelling developed in this

study was carried by biphasic RTD experiments which were compared
with RTD obtained from CFD simulations. Fig. 8 shows the comparison
between the experimental and modeled RTD curves at different volu­
metric flow rates comprised between 8.3 ≤ Q ≤ 15 cm3 s− 1 during the
cathodic production of H2 bubbles. Simulated RTD curves describe
adequately the experimental behavior of experimental RTD. As the
mean volumetric flow rate increases, the dimensionless curves, E(θ) − θ,
become slimmer, and the maximum values of E(θ) become larger,
appearing at θ < 1. This means that some tracer elements leave the
reactor earlier than the average residence time; these biphasic RTD ex­
periments compared with the monophasic RTD assays, reported in a
previous paper [19], confirm that hydrogen bubbles slightly increase the
velocity of the liquid phase. At the lowest volumetric flow rate, the
agreement between the calculated RTD by the model and the experi­
mental RTD results is poor. This might be explained due to the bubble’s
characteristics (size, shape) and its interaction (coalescence or breakup)
possibly originated along the cell and inside the upper cell linking pipe.
Additionally, the RTD curves tend to the left with flow rate and the tail
decreases as the volumetric flow rate increases; this indicates a better
trend towards a continuous mixing flow pattern. The velocity profiles
behavior showed in Fig. 4 would suppose a quasi-plug flow pattern in
the RTD curves (i.e., obtaining the maximum values of the RTD curves at
θ = 1). However, the upper connecting pipe acted as a fluid mixer,
modifying the shape of the RTD curves.

4.4. Theoretical and experimental analysis of the electrodissolution of

aluminium

Fig. 6. Dispersed phase fraction profiles along the channel width (in the y- A comparison between the experimental aluminium ion concentra­
coordinate) at different heights (in the x-coordinate) across each cell and eval­ tion with those values obtained from the numerical simulations is shown
uated at 0.3 cm of depth (in the z-coordinate). Data taken from the simulation in Fig. 9. The experimental aluminium ion concentration values were
trials in Fig. 4. obtained from our previous work [23]. The aluminium concentration
profiles are displayed at different jave of –5, –6, and –7 mA cm− 2 as a
fraction (hydrogen bubbles); this explains the higher values observed of function of the volumetric inflow rate (Q = 1.7, 3.3, 5, 6.7, and 8.3 cm3
j/jave at the exit of the cell, x/L > 0.8 (Fig. 3a) and c. As seen in Fig. 6, s− 1). It is worth mentioning that the electrodissolution of aluminium, at
there are two different dispersed fraction profiles. The hydrogen bubbles the anodes (Eq. (2)), occurs at the same magnitude of jave but with
concentration inside the cell 1 and cell 3 recorded the higher void current flow in the opposite direction. The numerical aluminium con­
fraction values compared with cell 2. This is attributed to the position of centrations were obtained at the electrolyte outlet of the filter-press
the cathode, which is located at the opposite side of the cell and, reactor. As can be seen in Fig. 9, the experimental aluminium dose
consequently, the position of the inlet manifolds. In accordance with the was partially different to the numerical one at some flow rates (i.e. 1.7,
mixture velocity profiles, and the change of the flow direction, the 3.3, and 8.3 cm3 s− 1). However, this discrepancy was reduced at volu­
accumulation of hydrogen bubbles in cell 3 is less than the cell 1 because metric flow rate of 5 cm3 s− 1. The discrepancies might be attributed to
as the mixture velocity increases the bubble build up decreases. Fig. 7a chemical dissolution of the anodes at the beginning of the experimental
shows the H2 bubbles fraction along each cell (x-coordinate) through the trials.
inter-electrode gap. At the lowest section, the void fraction observed in
cell 1 and 3 are quite similar due to the position of the cathodes.
4.5. Pressure drop
However, in cell 3, hydrogen bubbles concentration decreased slightly
because of the mixture velocity values. Since dispersed phase (H2) ap­
In H2-H2O bubbly flow, the hydrodynamics improved due to the
pears at the cathodes, an apparent bubble curtain spread was observed.
appearance of the H2 bubbles formed on the cathode. However, an
Dahlkild reported that the widening of bubbles curtains is dependent on
increased bubble fraction results in an increased pressure drop hence
the bubble diameter [46]. In addition, Wedin and Dahlkild observed that
higher pumping costs. Several experimental results have proposed that
as the dispersed fraction increases the flow rate does and, consequently,
the pressure drop (ΔP) is a logarithmic function of the flow velocity,
the bubble curtain is thicker [47]. At higher inflow velocities, the
expressed in terms of the Reynolds number via an empirical power law:
hydrogen bubbles curtain is less evident due to both the diminution of
dispersed fraction, Table 3, and the quicker evacuation of hydrogen ΔP = aReb (33)
bubbles. Fig. 7b displays the dispersed phase fraction field inside the
The experimental and theoretical logarithmic-logarithmic plots of
stack, the volumetric average αg values in cell 1, 2, and 3 were 2.68%,
pressure drop as a function of the Reynolds’ number for the two-phase
1.85%, and 2.31%, respectively. Table 3 shows that the average
flow in the multi-electrode stack are shown in Fig. 10. On the same
dispersed fraction increases with current density due to increasing H2
graph, studies reported in the literature are shown for single-phase flow
bubbles but decreases with volumetric inflow rate due to rapid release of
in the multi-electrode stack [19] and single-phase flow for empty single

9
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

a) b)

4
Dispersed phase fraction / %

3.5
3
2.5
2

αg / %
1.5
cell 1
1 cell 2
0.5 cell 3
cathodes
0
0 5 10 15

Channel length / cm x
y z

Fig. 7. a) Dispersed phase fraction profiles along the height (x-coordinate) of each cell at 2 cm of width in the y-coordinate. The profiles are referred to the thickness
of each cell. b) Dispersed phase fraction field inside the stack. Data taken from the simulation trials in Fig. 4.

Fig. 8. Comparison between theoretical (—) and experimental (●) RTD curves at different inflow volumetric rates are shown. The experimental biphasic RTD curves
were performed during the electrolysis of water to yield H2 at j = − 7 mA cm− 2.

10
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

Table 4
Experimental values of pressure drop over the FM01-LC reactor (using the
empirical power law ΔP = aReb).
Configuration a value ×102 b Reference
Pa value

Stack of three undivided cells (two-phase 0.030 3.49 This work

flow)
Stack of three undivided cells (single-phase 0.028 2.88 [19]
flow)
Empty unit cell 0.69 1.39 [48]
Filled unit cell with PTFE turbulence 1.69 1.54 [48]
promoter type D

with flow pattern) increases in comparison with the value obtained for
the stack of three cells in single-phase flow, the one empty cell, and the
cell with turbulence promoter. The experimental pressure drop was
higher to that obtained in single-phase flow owing to the dispersed
fraction of the electrogenerated H2 bubbles. Values of b increase in the
order: empty unit cell (single-phase flow) < filled unit cell with turbu­
lence promoter (single-phase flow) < stack of three cells (single-phase
flow) < stack of three cells (biphasic flow), from values of 1.39 to 3.49,
Table 4. This confirms that, in such filter press flow cell arrangements,
the turbulent flow predominates (considering that b = 1 for fully laminar
flows, and b → 2 when the flow becomes turbulent). The higher values of
b (2.88 and 3.49 during the monophasic and biphasic flows, respec­
tively) corroborate that flow dispersion (i.e., variations of the velocity in
a random fashion, apparition of jet flows, vortex and rotational flow
structures) is provoked by the abrupt changes in the cross-sectional area
and the direction of the flow. Close agreement was attained between the
experimental pressure drop and CFD simulation; the small pressure
variation is attributed to the experimental method used to measure
pressure drop. Finally, the energy required to impose the biphasic fluid
flow becomes greater than the obtained in a single-phase flow, with a
higher pumping power requirement.
Fig. 9. Comparison of experimental and numerical aluminium concentrations
at different current densities (jave = 5, 6, and 7 mA cm− 2) as a function of the 5. Conclusions
inflow volumetric rates (Q = 1.7, 3.3, 5, 6.7, and 8.3 cm3 s− 1).
A theoretical analysis of the electrogeneration of hydrogen bubbles
was carried out using a multi-coupled numerical simulation consisting in
secondary current distribution, the two-phase (H2-H2O) flow and mass
transport by convection–diffusion in a pre-pilot filter press flow elec­
trolyser with a stack of three cells. At vertical, parallel electrode plates,
the secondary current distribution was uniform along the working
electrodes and the counter electrodes.
The Euler-Euler turbulent two-phase model successfully described
the complex H2O-H2 flow behavior in the stack of three cells. Several
types of flow behaviors were seen in each cell. At the entrance of each
channel, CFD simulations showed a high degree of dispersion in the
mixture velocity while an apparent plug flow pattern was developed at
lengths beyond 12 cm. An increment of mixture velocity magnitude was
observed in the following order cell 1 < cell 2 < cell 3. This latter is due
to the turbulence caused by the change of flow direction in the third cell.
Fig. 10. Logarithmic plot of the pressure drops versus Reynolds number for the Dispersed phase profiles constructed along each cell and across the inter-
FM01-LC reactor containing a stack of three empty cells in two-phase flow (O; electrode space showed a pseudo bubble curtain profile. Here, the
this work, experimental) and (●; this work, CFD simulation); this is compared bubbles released from the cathode are transported away from the elec­
with a single-phase flow in the multi-electrode stack reactor (□) [19], an empty trode by the velocity of the liquid. H2 bubbles increases the velocity of
single cell (Δ) [48], and a filled single cell with PTFE turbulence promoter type the liquid phase but the void fraction of such bubbles also increases
D (◊) [48]. For the experiments performed here the pressure drop was per­ pressure drop. The simulated biphasic RTD curves were in good agree­
formed during the electrolysis of water to evolve H2 at j = − 7 mA cm− 2; the ment with the experimental ones. Finally, the model predicted reason­
same conditions were used for CFD simulations.
ably well the anodic dissolution of aluminium. Further studies should
consider the concomitant coagulation-flocculation and their
cell and filled single cell with turbulence promoter [48] are plotted. mechanisms.
Table 4 summarizes the experimental parameters reported in the liter­ The model proposed in this paper could be extended to other
ature for single-phase flow using different flow configurations in com­ biphasic systems such as O2-H2O, Cl2-H2O, among others, over a wider
parison to those obtained in this work. In this paper, the a value range of operating conditions.
(associated with the geometry) decreases and the b index (associated

11
M.A. Sandoval et al. Separation and Purification Technology 261 (2021) 118235

CRediT authorship contribution statement [19] M.A. Sandoval, R. Fuentes, F.C. Walsh, J.L. Nava, C. Ponce de León, Computational
fluid dynamics simulations of single-phase flow in a filter-press flow reactor having
a stack of three cells, Electrochim. Acta 216 (2016) 490–498.
Miguel A. Sandoval: Conceptualization, Data curation, Formal [20] T. Pérez, C. Ponce de León, F.C. Walsh, J.L. Nava, Simulation of current
analysis, Investigation, Validation, Software, Writing - original draft. distribution along a planar electrode under turbulent flow conditions in a
Rosalba Fuentes: Validation, Writing - review & editing. Tzayam laboratory filter-press flow cell, Electrochim. Acta 154 (2015) 352–360.
[21] E.P. Rivero, M.R. Cruz-Díaz, F.J. Almazán-Ruiz, I. González, Modelling the effect of
Pérez: Formal analysis, Software, Writing - review & editing. Frank C. non-ideal flow pattern on tertiary current distribution in a filter-press-type
Walsh: Conceptualization, Formal analysis, Writing - review & editing. electrochemical reactor for copper recovery, Chem. Eng. Res. Des. 100 (2015)
José L. Nava: Conceptualization, Validation, Supervision, Writing - 422–433.
[22] A. Frías-Ferrer, J. González-García, V. Sáez, C. Ponce de León, F.C. Walsh, The
review & editing, Funding acquisition. Carlos Ponce León: Formal effects of manifold flow on mass transport in electrochemical filter-press reactors,
analysis, Writing - review & editing. AIChE J. 54 (2008) 811–823.
[23] M.A. Sandoval, R. Fuentes, J.L. Nava, O. Coreño, Y. Li, J.H. Hernández,
Simultaneous removal of fluoride and arsenic from groundwater by
Declaration of Competing Interest electrocoagulation using a filter-press flow reactor with a three-cell stack, Sep. Pur.
Technol. 208 (2019) 208–216.
The authors declare that they have no known competing financial [24] M.A. Sandoval, R. Fuentes, J.L. Nava, I. Rodríguez, Fluoride removal from drinking
water by electrocoagulation in a continuous filter press reactor coupled to a
interests or personal relationships that could have appeared to influence flocculator and clarifier, Sep. Purif. Technol. 134 (2014) 163–170.
the work reported in this paper. [25] A.D. Villalobos-Lara, T. Pérez, A.R. Uribe, J.A. Alfaro-Ayala, J.J. Ramírez-
Minguela, J.I. Minchaca-Mojica, CFD simulation of biphasic flow, mass transport
and current distribution in a continuous rotating cylinder electrode reactor for
Acknowledgments
electrocoagulation process, J. Electroanal. Chem. 858 (2020), 113807.
[26] Ph. Mandin, J. Hamburger, S. Bessou, G. Picard, Modelling and calculation of the
Miguel A. Sandoval is grateful to CONACYT (Mexico) for granting current density distribution evolution at vertical gas-evolving electrodes,
the postdoctoral scholarship, No. 386022. J.L. Nava acknowledges Electrochim. Acta 51 (2005) 1140–1156.
[27] S. Subramaniam, Lagrangian-Eulerian methods for multiphase flows, Prog. Energ.
Universidad de Guanajuato (Mexico) for financial support through the Combust. 39 (2013) 215–245.
project No. CIIC 113/2020. [28] M.R. Cruz-Díaz, E.P. Rivero, F. Almazán-Ruiz, Á. Torres-Mendoza, I. González,
Design of a new FM01-LC reactor in parallel plate configuration using numerical
simulation and experimental validation with residence time distribution (RTD),
References Chem. Eng. Process. 85 (2014) 145–154.
[29] L. Vázquez, A. Alvarez-Gallegos, F.Z. Sierra, C. Ponce de León, F.C. Walsh,
[1] F.F. Rivera, C. Ponce de León, J.L. Nava, F.C. Walsh, The filter-press FM01-LC Simulation of velocity profiles in a laboratory electrolyser using computational
laboratory flow reactor and its applications, Electrochim. Acta 163 (2015) fluid dynamics, Electrochim. Acta 55 (2010) 3437–3445.
338–354. [30] P.S. Bernard, J.M. Wallace, Turbulent flow: Analysis, measurement and prediction,
[2] F.C. Walsh, L.F. Arenas, C. Ponce de León, Developments in plane parallel flow first ed., John Wiley & Sons, New Jersey, 2002.
channel cells, Curr. Opin. Electrochem. 16 (2019) 10–18. [31] Y. Tominaga, T. Stathopoulos, Turbulent Schmidt numbers for CFD analysis with
[3] L.F. Arenas, C. Ponce de León, F.C. Walsh, The versatile plane parallel electrode various types of flow field, Atmos. Environ. 41 (2007) 8091–8099.
geometry, J. Electrochem. Soc. 167 (2020), 023504. [32] Y. Gorb, O. Mierk, L. Rivkind, D. Kuzmin, Finite element simulation of three-
[4] R. Hreiz, L. Abdelouahed, D. Fünfschilling, F. Lapicque, Electrogenerated bubbles dimensional particulate flows using mixture models, J. Comput. Appl. Math. 270
induced convection in narrow vertical cells: A review, Chem. Eng. Res. Des. 100 (2014) 443–450.
(2015) 268–281. [33] K.M. Abd Ali, CFD Simulation of bubbly flow through a bubble column, IJSER 5
[5] R. Hreiz, L. Abdelouahed, D. Fünfschilling, F. Lapicque, Electrogenerated bubbles (2014) 904–910.
induced convection in narrow vertical cells: PIV measurements and Euler-Lagrange [34] A.R. Khophkar, A.R. Rammohan, V.V. Ranade, M.P. Dudukovic, Gas-liquid flow
CFD simulation, Chem. Eng. Sci. 134 (2015) 138–152. generated by a Rushton turbine in stirred tank vessel: CAPRT/CT measurements
[6] A. Alexiadis, M.P. Dudukovic, P. Ramachandran, A. Cornell, J. Wanngard, and CFD simulations, Chem. Eng. Sci. 60 (2005) 2215.
A. Bokkers, Liquid–gas flow patterns in a narrow electrochemical channel, Chem. [35] L. Zhongqiu, L. Baokuan, Scale-adaptive analysis of Euler-Euler large eddy
Eng. Sci. 66 (2011) 2252–2260. simulation for laboratory scale dispersed bubbly flows, Chem. Eng. J. 338 (2018)
[7] Ph. Mandin, A.A. Aissa, H. Roustan, J. Hamburger, G. Picard, Two-phase 465–477.
electrolysis process: From the bubble to the electrochemical cell properties, Chem. [36] H.K. Versteeg, W. Malalasekera, An introduction to computational fluid dynamics:
Eng. Process. 47 (2008) 1926–1932. The finite volume method, second ed., Prentice Hall, London, 1995.
[8] F.F. Rivera, C. Ponce de León, F.C. Walsh, J.L. Nava, The reaction environment in a [37] J. Szekely, Fluid flow phenomena in metals processing, Academic Press Ind, New
filter-press laboratory reactor: the FM01-LC flow cell, Electrochim. Acta 161 York, 2012.
(2015) 436–452. [38] D.C. Wilcox, Turbulence Modelling for CFD, DCW Industries Inc, California, 1998.
[9] L.F. Castañeda, F.F. Rivera, T. Pérez, J.L. Nava, Mathematical modelling and [39] H. Schlichting, Boundary-Layer Theory, seventh ed., MC Graw-Hill, New York,
simulation of the reaction environment in electrochemical reactors, Curr. Opin. 1979.
Electrochem. 16 (2019) 75–82. [40] O. Annunziata, J.A. Rard, J.G. Albright, L. Paduano, D.G. Miller, Mutual diffusion
[10] S. Ich Ngo, Y. Lim, Multiscale Eulerian CFD of chemical processes: A review, coefficients and densities at 298.15 K of aqueous mixtures of NaCl and Na2SO4 for
ChemEngineering. 4 (2020) 23. six different solute fractions at a total molarity of 1.500 mol dm− 3 and of aqueous
[11] B. Ashraf Ali, S. Pushpavanam, Analysis of unsteady gas–liquid flows in a Na2SO4, J Chem. Eng. 45 (2000) 936–945.
rectangular tank: Comparison of Euler-Eulerian and Euler-Lagrangian simulations, [41] H.S. Fogler, Elements of chemical reaction engineering, fourth ed., Prentice Hall,
Int. J. Multiphas. Flow. 37 (2011) 268–277. New Jersey, 2005.
[12] J.A. Ramírez, A. Rodríguez, F.F. Rivera, F. Castañeda, Experimental study and [42] J. Lu, Z. Wang, X. Ma, Q. Tang, Y. Li, Modelling of the electrocoagulation process: a
mathematical modelling of two-phase flow with a Eulerian approach in a study on the mass transfer of electrolysis and hydrolysis products, Chem. Eng. Sci.
continuous gas evolving electrochlorinator, Chem. Eng. Res. Des. 144 (2019) 165 (2017) 165–176.
538–549. [43] M.A. Sandoval, R. Fuentes, A. Thiam, R. Salazar, Arsenic and fluoride removal by
[13] H. Wang, X. Jia, X. Wang, Z. Zhou, J. Wen, J. Zhang, CFD modelling of electrocoagulation process: A general review, Sci. Total Environ. 753 (2021),
hydrodynamic characteristics of a gas–liquid two-phase stirred tank, Appl. Math. 142108.
Model. 38 (2014) 63–92. [44] K. Aldas, N. Pehlivanoglu, M.D. Mat, Numerical and experimental investigation of
[14] J. Ding, X. Wang, X. Zhou, N. Ren, W. Guo, CFD optimization of continuous stirred- two-phase flow in an electrochemical cell, Int. J. Hydrog. Energy 33 (2008)
tank (CSTR) reactor for biohydrogen production, Bioresource. Technol. 101 (2010) 3668–3675.
7005–7013. [45] K. Aldas, Application of a two-phase flow model for hydrogen evolution in an
[15] L. Abdelouahed, G. Valentin, S. Poncin, F. Lapicque, Current density distribution electrochemical cell, Appl. Math. Comput. 154 (2004) 507–519.
and gas volume fraction in the gap of lantern blade electrodes, Chem. Eng. Res. [46] A.A. Dahlkild, Modelling the two-phase flow and current distribution along a
Des. 92 (2014) 559–570. vertical gas-evolving electrode, J. Fluid. Mech. 428 (2001) 249–272.
[16] K. Wadaugsorna, S. Limtrakula, T. Vatanathama, P.A. Ramachandran, [47] R. Wedin, A.A. Dahlkild, On the transport of small bubbles under developing
Hydrodynamic behaviours and mixing characteristics in an internal loop airlift channel flow in a buoyant gas-evolving electrochemical cell, Ind. Eng. Chem. Res.
reactor based on CFD simulation, Chem. Eng. Res. Des. 113 (2016) 125–139. 40 (2001) 5228–5233.
[17] L. Abdelouahed, R. Hreiz, S. Poncin, G. Valentin, F. Lapicque, Hydrodynamics of [48] C.J. Brown, D. Pletcher, F.C. Walsh, J.K. Hammond, D. Robinson, Studies of space-
gas bubbles in the gap of lantern blade electrodes without forced flow of average mass transport in the FM01-LC laboratory electrolyser, J. Appl.
electrolyte: Experiments and CFD modelling, Chem. Eng. Sci. 111 (2014) 255–265. Electrochem. 23 (1993) 38–43.
[18] L. Castañeda, René Antaño, Fernando F. Rivera, José L. Nava, Computational fluid
dynamic simulations of single-phase flow in a spacer-filled channel of a filter-press
electrolyzer, Int. J. Electrochem. Sci. 12 (2017) 7351–7364.

12