Chemical Engineering Science 154 (2016) 61–71

Contents lists available at ScienceDirect

Chemical Engineering Science

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

Charge coup de fouet phenomenon in soluble lead redox flow battery

Mahendra Nandanwar, Sanjeev Kumar n
Department of Chemical Engineering, Indian Institute of Science, Bangalore 560012, India

H I G H L I G H T S

Voltage dip after initial peaking of charging voltage, and slow return to normalcy.

Effective conductivity is obtained using a percolation model with a critical fraction.

Voltage drop across surface deposits included in model through reduced over-potential.

Insulating regions on electrode expand and contract with discharge/charge cycle.

Charging Coup de fouet increases with increase in depth of discharge in previous cycle.

art ic l e i nf o a b s t r a c t

Article history: The charge coup de fouet phenomenon, known in the context of lead-acid battery, refers to the presence
Received 15 February 2016 of a voltage dip shortly after charging of a fully discharged battery begins. While the attempts to relate
Received in revised form magnitude of coup de fouet phenomena with the state of health of battery have appeared in the literature,
30 June 2016
the phenomena continue to be poorly understood. The soluble lead redox flow battery (SLRFB), with
Accepted 3 July 2016
Available online 7 July 2016
potential for energy storage at large scale at low cost, also displays similar features. We report in this
work our modeling and experimental efforts aimed at understanding charge coup de fouet phenomenon
Keywords: in natural convection driven SLRFB. We present a model that incorporates the presence of non-con-
Flow battery ducting PbO in deposits through a percolation type model for conductivity. The associated potential drop
Effective conductivity
across the deposits is incorporated in the model through reduced overpotential available for driving
Percolation model
Butler–Volmer kinetics. The complete model with coupled natural convection induced by non-uniform
Overpotential
Natural convection concentration of Pb ions in electrolyte successfully captures charge coup de fouet phenomenon, and ex-
Soluble lead plains the measured variation of its magnitude with the depth of discharge in the previous cycle. The
model explains our earlier observation that during the discharge process, a receding zone of deposits is
seen only on cathode but not on anode. The approach used is applicable to electrochemical systems in
which solid conducting matrix evolves to non-conducting or poorly conducting and vice-versa, with a
change in composition.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction decrease slowly and steadily, as expected of discharging voltage

Coup de fouet phenomena are reported in the literature in the
context of conventional lead-acid batteries. When a fully dis-
charged battery is charged, the voltage first rises sharply in a ra-
ther short time, following which is decreases, goes through a
comparatively shallow minimum, and then begins to rise slowly
and steadily, as expected of charging voltage for a battery. Simi-
larly, when a fully charged battery is discharged, the battery vol-
tage first drops sharply in a rather short time, following which it
rises, goes through a rather broad maximum, and then begins to
n
Corresponding author.
E-mail address: sanjeev@chemeng.iisc.ernet.in (S. Kumar).
for a battery. These two phenomena, shown schematically in Fig. 1,
are named charge coup de fouet and discharge coup de fouet re-
spectively. Depending on the operating conditions, magnitude of
the voltage dip and voltage rise in charge and discharge coup de
fouet phenomena varies in range of 10–50 mV, and lasts from a few
seconds to a couple of minutes (Delaille et al., 2006; Pascoe and
Anbuky, 2002). These phenomena find use in battery diagnosis
through their correlation with state of health (SOH) of a battery
(Bose and Laman, 2000; Pascoe and Anbuky, 2002; Pascoe et al.,
2002), an indicator of its ability to perform well in charge-dis-
charge cycles. The SOH tells a user when to replace a battery. The
coup de fouet phenomena are also correlated to state of charge
(SOC) of a battery (Pascoe and Anbuky, 1999) which indicates how

http://dx.doi.org/10.1016/j.ces.2016.07.001
0009-2509/& 2016 Elsevier Ltd. All rights reserved.
62 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71
Fig. 1. Schematics of (a) charge coup de fouet and (b) discharge coup de fouet observed in conventional lead-acid battery.
much more charge can be put in it.
Berndt and Voss (1965) has studied charge coup de fouet in
conventional lead-acid battery, and suggested that a peak in po-
tential could arise from insulating PbSO4 formed on discharge
products in lead-acid battery. Delaille et al. (2006) have contested
the above explanation. According to them, end-of-discharge does
not correspond to complete coverage of electrode by an insulating
layer of PbSO4. The authors instead proposed mass transport
limitations to be the reason for charge coup de fouet. No mathe-
matical model appears to have captured the peak associated with
charge coup de fouet yet.
The recently proposed membraneless soluble lead-acid redox
flow battery (SLRFB) with only one electrolyte flow loop (Hazza
et al., 2004, 2005; Li et al., 2009; Pletcher and Wills, 2004, 2005;
Pletcher et al., 2008ab) is a promising candidate for large scale
storage of energy, as it can potentially offer a number of ad-
vantages over the other widely used redox flow batteries (Verde
et al., 2013; Nandanwar and Kumar, 2014). The active species Pb2 þ
in SLRFB comes from lead methanesulfonate salt [(CH3SO3)2Pb:
PbMS] dissolved in aqueous solution of methanesulfonic acid
[CH3SO3H: MSA]. The redox reactions involved are:
charge
Negative electrode: Pb2 + + 2e− H ooooooooI Pb (s )
discharge
discharge
Positive electrode: PbO2 (s) + 4 H+ + 2e− H ooooooooI Pb2 + + 2H2 O
charge
charge
Overall reaction: 2Pb2 + + 2H2 O H ooooooooI Pb (s ) + PbO2 (s) + 4H+
discharge
In charge cycle, lead ions are deposited from electrolyte on
cathode and anode in the form of Pb and PbO2 respectively. The
same are redissolved as Pb2 þ ions in discharge cycle. The electrode
reactions involved in soluble lead battery appear much simpler
than those involved in the conventional lead-acid battery. The
measurements however show similar charge coup de fouet and
discharge coup de fouet phenomena (Verde et al., 2013).
In this work, we investigate charge coup de fouet in SLRFB ex-
perimentally, and develop a quantitative model to understand it.
The latter is necessary as the existing models do not capture it. The
experiments are carried out in a standard rectangular cell, with no
external flow, but in the presence of induced natural convection,
explained in latter sections. We begin in the next section with a
review of the current understanding of SLRFB, as this is needed for
model development. This is followed by sections covering ex-
perimental details, model development, and results.
2. Current state of understanding of SLRFB
The continued investigations on SLRFB over the last ten years
have brought up a number of their characteristics which need to be
explained by models. Hazza et al. (2004) experimented with
10 cm  10 cm electrodes with inter-electrode separation of 1.2 cm
and found anode to be the limiting one. Collins et al. (2010) reported
that charging in the first cycle occurs at a nearly constant voltage,
followed by two step charging in the subsequent cycles. Beck and
Bhn (1975) suggested that reduction of PbO2 during discharge leads
to its electro-dissolution and also formation of an insoluble deposit
which oxidizes more readily than Pb2 þ ions in the next charge cycle
to form PbO2 deposits. The second cycle charging therefore starts at a
lower voltage. Pletcher and Wills (2005) carried out EDAX analysis of
insolubles on anode and found Pb and O in ratio close to 1:1, sug-
gesting PbO as insolubles. Oury et al. (2012) carried out mass change
⎫
( − 0.126 V /SHE )⎪
⎪
⎪
( + 1.455 V /SHE )⎬
⎪
⎪
( + 1.58 V /SHE ) ⎪
⎪
⎭ (1)
investigations at anode using quartz crystal micro-balance to confirm
the presence of solid state reaction on anode, and suggested for-
mation of non-stoichiometric PbOx. The authors suggested that such
non-stoichiometric compounds may come from modification of in-
termediate compound Pb(OH)2 formed during PbO2 to Pb2 þ con-
version. These findings provided basis for Shah et al. (2010) to pro-
pose the following side reaction at anode to run concurrently to the
main reactions given by Eq. (1):
charge
PbO + H2 O H ooooooooI PbO2 + 2e− + 2H+.
discharge (2)
Shah et al. (2010) developed the first model for SLRFB. The
model considered Nernst-Planck equations to govern transport of
 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71
ions in the flow field prevailing in their flow-through cell, Butler–
Volmer kinetics for electron-transfer reactions, and uniform de-
position/dissolution of solid Pb and PbO2 on cathode and anode.
An empirical expression similar to the Butler–Volmer equations
but involving surface concentrations of solid deposits was pro-
posed by them to account for the side reaction given by Eq. (2).
The model captured the onset of two step charging from second
charge cycle onwards very well. It required much smaller con-
centration of hydrogen ions than that present in the system to
capture rapid decrease in discharge voltage towards the end (Na-
danwar and Kumar, 2014). Bates et al. (2014) used similar equa-
tions to study the effect of temperature and electrolyte con-
centration, while Oury et al. (2014) investigated deposition of PbO2
alone in honeycomb shaped anode.
Nandanwar and Kumar (2014) argued for varying mass transfer
resistance along the length of an electrode due to the growing
concentration boundary layer, and accounted for non-uniform
deposition and dissolution of PbO2 and Pb deposits in their model.
Their model equations and the form used for Butler–Volmer ki-
netics are the same as those of Shah et al. (2010). The electro-
dissolution of PbO2 in their model is permitted to continue in-
dependent of the buildup of the non-conducting PbO locally. In the
charge cycle, the oxidation of solid PbO and Pb2 þ in the electrolyte
commences simultaneously to form PbO2 everywhere on the
electrode. The model, with some refitting of parameters proposed
by Shah et al. (2010) for reaction kinetics, successfully captures
charge-discharge characteristics and rapid decrease in discharge
Fig. 2. Schematic of (a) closed rectangular cell with electrodes mounted on walls used to carry out experiments, and (b) the computational grid used to carry out simulations.
63
voltage towards the end of discharge cycle through partial baring
of electrodes. It also predicts that alternating flow direction can
enhance performance of a SLRFB.
Nandanwar and Kumar (2016a) found that natural convection
induced by difference in concentration of Pb2 þ ions near the
electrode surface and the bulk of the electrolyte plays a dominant
role in the performance of a soluble lead battery with no external
circulation. This has been modeled by Nandanwar and Kumar
(2016b) by using a concentration difference driven body force
term in the equations of motion in the model of Nandanwar and
Kumar (2014). The augmented model of Nandanwar and Kumar
(2016b), with all the parameter values same as those of Nandan-
war and Kumar (2014), very well explains the experimental data
obtained by them in a narrow rectangular cell with no external
flow, shown schematically in Fig. 2.
The conductivity of electrodes and deposits in all the previous
models is considered to be infinitely large compared to that of the
electrolyte. These models therefore do not require potential dis-
tribution in solid phase to be determined. On a microscopic level,
the assumption of high conductivity of deposits holds well for
cathode as current passes either through Pb deposits or graphite
electrode, both of which are highly conducting. The situation at
anode is different though. The assumption of high conductivity of
anode holds for the first charge cycle as the deposit consists of
only highly conducting PbO2. In the first discharge cycle, some of
the PbO2 is electro-dissolved and some other is converted into
non-conducting PbO solid (Eq. (2)). Oury et al. (2012) monitored
64 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71

mass change at anode during discharge cycle using quartz crystal We adopt the model of Nandanwar and Kumar (2016b) which
micro-balance and impedance spectra. Their studies confirmed incorporates natural convection caused by electrochemical reac-
PbO2 to PbO conversion and formation of an insulating layer that tion on electrode surface and non-uniform current density on it as
eventually stops discharge process altogether. the base model, and incorporate in it electrical resistance of solid
The assumption of highly conducting deposits on anode irre- deposits on electrodes (Euler et al., 1980). The model set consists
spective of the proportion of non-conducting PbO in it appears to of transport equations for Pb2 þ , H þ , and ( CH3 SO−3 ), velocity field,
be an extreme one. In this work, we seek to incorporate in the
electrode kinetics, and relevant initial and boundary conditions.
model of Nandanwar and Kumar (2016b) changes occurring in
The commonly made assumptions, stated and discussed in detail
conductivity of deposits on anode, and study its impact on charge-
earlier (Shah et al., 2010; Nandanwar and Kumar, 2014) are: excess
discharge profiles, particularly in the context of charge coup de
water, dilute solution theory (Newman and Thomas-Alyea, 2004),
fouet which is not explained by any of the existing models.
laminar flow, and no gas evolution at electrodes. The transport
equation for ionic species are given by
3. Experimental ∂ci
= − ∇·Ni i = H+, Pb2 + , CH3 SO−3
∂t (4)
The experiments were carried out in a rectangular cell
where flux Ni is
(4 cm  1.5 cm  7 cm) made of transparent perspex, housing two
graphite electrodes (3 cm  3 cm) located symmetrically on the zi ci Di F
Ni = − Di ∇ci − ∇ϕ + ci u.
opposite walls, with an inter-electrode gap of 1.5 cm. We used RT (5)
smaller electrodes than those reported in the literature so as to
The electro-neutrality and charge conservation require
facilitate CFD simulations to be completed in reasonable time
(about 24 h). We expect electrode size to leave charge coup de ∑ zi ci = 0,
fouet phenomenon investigated in this work unaffected. The ex- i (6)
perimental procedure adopted is as follows. Briefly, the cell was
completely filled with electrolyte solution consisting of 0.5 M
methanesulfonic acid and 0.5 M lead methanesulfonate, and ∇·j = 0; j = F ∑ zi Ni .
aligned vertically. No additives were used. Electrochemical mea- i (7)
surements were carried out using Autolab Potentiostat/Galvano-
The time dependent velocity field u in Eq. (4) is obtained by sol-
stat-Model 30, in the absence of any external convection. The
battery cell was first galvanostatically charged (Banerjee et al., ving the Navier–Stokes equation
2013) for 900 s followed by relaxation (open circuit) for 300 s. The ∂u
ρ + ρ (u ·∇) u = − ∇p + μ∇2u + F
battery was then discharged galvanostatically. Unless stated ∂t (8)
otherwise, a cut-off voltage of 1.2 V and current density of
and the continuity equation for an incompressible fluid
200 A/m2 were used. The discharge phase was followed by a re-
laxation period of 300 s before starting the next charging phase. A ∇·u = 0. (9)
magnetic stirrer was used to homogenize electrolyte during the
relaxation phase (open circuit condition) to eliminate stratification The body-force F in Eq. (8) is given by Gebhart and Pera (1971)
resulting from charging/discharging. All the experiments were
F = gρ ∑ βi (ci0 − ci )
performed at room temperature. i (10)
þ
for i being species Pb2 þ and H . Here, c0i
is the initial concentra-
4. The model tion and βi is the expansion coefficient defined as

1 ⎛ ∂ρ ⎞
Euler et al. (1980) have proposed that when volume fraction of βi = ⎜ ⎟ .
ρ ⎝ ∂ci ⎠T , P, c j ≠ i (11)
conducting material in a matrix becomes less than a ‘critical limit’,
the conducting pathways cease to exist and conductivity of the The subscripts T, P, and cj indicate quantities held constant.
solid matrix drops to zero. Above the critical limit, the solid mix- The mass balance for deposition of solids (Csi) due to the local
ture behaves like a good conductor and below like an insulator.
electrochemical reaction, reflected by local current density (j), is
The conductivity of such a mixture (se), as given by effective
given by
medium theory (Euler et al., 1980), is
s s
dC PbO 2 dCPbO
1 ⎛ ⎞ at anode, 0<y<L = j PbO 2 /2F; = − jPbO /2F
σe = ⎜Q + Q2 + 2 (z − 2) σa σ b ⎟ dt dt (12)
z − 2⎝ ⎠ (3)

where
s
dCPb
⎛z ⎞ ⎛z ⎞ at cathode, 0<y<L = jPb /2F
Q = ⎜ A a − 1⎟ σa + ⎜ Ab − 1⎟ σ b, dt (13)
⎝2 ⎠ ⎝2 ⎠
The thickness of the deposits formed is several orders of magni-
Aa and Ab are fractional volumes of components a (PbO) and b
tude smaller than the inter-electrode gap. We therefore assume
(PbO2), sa and sb are specific conductivities of components, and z
is coordination number (number of contacts per particle). Re- the latter to remain unchanged. The local current density at anode
cently, Gandhi (2015) has incorporated formulation of Euler et al. jan is equal to jPbO2 + jPbO . The current density for the main reaction
(1980) to model charge/discharge of conventional lead-acid bat- on anode (Eq. (1)) is given by the classical Butler–Volmer reaction
teries. The author found that conductivity of the matrix plays a kinetics (Shah et al., 2010; Bates et al., 2014). We use here, without
major role in determining discharge capacity of battery. The au- refitting any parameters, the expression provided by Nandanwar
thor did not comment on coup de fouet phenomena. and Kumar (2014).
 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71 65

⎧ ⎛ ⎞⎛ ⎞⎡ ⎤ Vcell = V+ − V− (20)
⎪ ⎜ c 2+ ⎟ ⎜ cH + ⎟ ⎢ ⎛ 0.5Fη+ ⎞ ⎛ − 0.5Fη+ ⎞ ⎥
⎪ 2Fk PbO 2 ⎜ Pb
⎟⎜ ⎟ ⎢ exp ⎜ ⎟ − exp ⎜ ⎟ ⎥ for Case 1
jPbO2 = ⎨ ⎜ c ref2 + ⎟ ⎜ c ref+ ⎟ ⎢ ⎝ RT ⎠ ⎝ RT ⎠ ⎥ where V− is fixed at zero as the reference voltage.
⎪ ⎝ Pb ⎠⎝ H ⎠⎣ ⎦
⎪ An exact incorporation of spatially and temporally varying
⎩ 0, for Case 2 (14)
electrical resistance of deposits requires determination of voltage
The complexities associated with interconversion of solid PbO2 profile in the composite solid phase. The voltage on the surface of
to solid PbO and vice versa and the need for hydrogen ions in the deposits is coupled to the spatially distributed current density
electrolyte to participate in this reaction necessitate the use of an through rate expressions for electrochemical reactions. This in-
empirical approach. Shah et al. (2010) have proposed a form si- creases the computational complexity many fold. In view of the
milar to Butler–Volmer kinetics expression to model solid-solid high conductivity of the underlying graphite electrode and rela-
reaction. It involves surface concentrations of the deposited solids. tively small thickness of the deposit layer, we propose the fol-
The same approach is used by others subsequently (Bates et al., lowing. We treat the original electrode material (graphite) to be at
2014; Nandanwar and Kumar, 2014). In the present work, we have uniform potential everywhere, and consider the potential drop
used the expression provided by Nandanwar and Kumar (2014) required across the deposit, which supports the local current, to
without refitting any parameters: reduce the surface potential to which the reacting species are
exposed.
⎛ C s ⎞ν ⎛ ⎞
0.9Fη+ ⎟ In other words, the potential drop across the deposits reduces
jPbO = 2Fkf ⎜⎜ sPbO ⎟ exp ⎜
, ref ⎟ ⎜ RT ⎟ the overpotential available otherwise to drive Butler–Volmer ki-
⎝ CPbO ⎠ ⎝ ⎠
netics. We suitably modifying the value of overpotential on anode.
s ⎛ ⎞
CPbO 2 cH −0.9Fη+ ⎟ The voltage drop ΔVs required to support the flow of current from
exp ⎜⎜
+
− 2Fkb ⎟
s, ref
c ref+ the site of reaction on deposit surface to the underlying graphite
CPbO 2 H ⎝ RT ⎠ (15)
electrode, in the limit of one dimensional approximation for thin
The coefficient 0.9 used in the exponential terms is consistent with deposits, is
the constraint that in a multi-step reaction, a part of the energy lt
made available by over-potential is dissipated (Schmickler and ΔVs = jan × Rs = jan
σe (21)
Santos, 2010). The overpotential for the side reaction on anode is
considered to be the same as that for the main reaction, and any where jan = jPbO2 + jPbO and Rs = lt /σe is the resistance offered by
deviation is absorbed in constants kf and kb (Shah et al., 2010). The deposits of thickness lt with composition dependent electrical
value of exponent ν controls the onset of second step during conductivity se. The local overpotential available to drive the
charging from second cycle onwards, and is not expected to be an electrochemical reactions is suitably modified from η þ to η′+ as
integer for a non-elementary reaction.
The Butler–Volmer kinetics at cathode is given by (Shah et al.,
( )
η+′ = V+ − ΔVs − ϕ − E± = η+ − ΔVs (22)
2010; Bates et al., 2014; Nadanwar and Kumar, 2014): The initial and boundary conditions needed to simulate the
⎧ ⎛ ⎞⎡ ⎤ model, for the computational domain shown in Fig. 2, are as fol-
⎪ c 2+ ⎛ Fη ⎞ ⎛ −Fη− ⎞ ⎥
⎪ 2FkPb ⎜ Pb ⎟ ⎢ exp ⎜ − ⎟ − exp ⎜ ⎟ for Case 1 lows. The boundary conditions at anode and cathode are:
jPb =⎨ ⎜ c ref ⎟ ⎢ ⎝ RT ⎠ ⎝ RT ⎠ ⎥
⎪ ⎝ Pb2 + ⎠ ⎣ ⎦ 2F (NPb2 + ·n) = jPbO 2 F (NH + ·n) = 2jPbO 2 + jPbO 0<y<L (23)
⎪
⎩ 0, for Case 2 (16)

with all the parameter values held same as those provided by 2F (NPb2 + ·n) = jPb (NH + ·n) = 0 0<y<L (24)
Nandanwar and Kumar (2014). The two cases appearing in Eqs.
(14) and (16) are devised to prevent continued dissolution from a where n is the outward normal to surface. The applied average
local region even after all the active solids (Pb and PbO2) are current density at anode satisfies
electro-dissolved. These cases are: L
1
japp =
L
∫0 jan dy .
(25)
1. Case 1: (i) charging and Cis ≥ 0, and (ii) discharging and Cis > 0
2. Case 2: discharging and Cis = 0 The no-flux boundary condition ( j·n = 0) was applied at all other
boundaries. The initial conditions for concentrations of ionic spe-
where i stands for PbO2 on anode and Pb on cathode. cies in the electrolyte domain are specified as
For uniform potential in electrode and deposits, which corres-
c Pb2 +(x, y , t = 0) = c 0 2 ; c H +(x, y , t = 0) = c H0 +.
ponds to zero electrical resistance everywhere in solid phases, η þ Pb + (26)
and η  in Eqs. (14)–(16) are given by The initial concentration of solid species deposited on electrodes,
s (x = 0, y , t = 0) , C s
CPb , and CPbO (x = W , y, t = 0),
η± = V± − ϕ± − E±, (17) PbO2 (x = W , y , t = 0)
were set to zero. The initial velocity field u in the domain was also
where V± are electrode potentials, ϕ± are potentials in electrolyte specified to be zero everywhere. The no-slip condition was em-
next to the electrodes, and E± are Nernst potentials, given by (at ployed at all solid-electrolyte interfaces. The reference pressure
anode) pref was specified at the top left corner of the numerical domain.
RT ⎛ ⎞ The system of equations along with the initial and boundary
E+ = E+0 − ⎜ ln c Pb2 + − ln c H + ⎟. conditions can be solved in principle to obtain model predictions.
2F ⎝ ⎠ (18)
The presence of the unknown local flux jan in the exponential
and (at cathode) terms in Butler–Volmer equations makes the model highly cou-
pled and implicit. Even for the simplified 2-d configuration and
RT
E− = E−0 + ln c Pb2 +, relatively small size electrodes used in the work, the model re-
2F (19)
quires extensive computational effort in COMSOL to obtain pre-
The cell voltage is defined as dictions. We therefore simplified the model equations further
66 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71

using the approximations discussed below. We rewrite Eqs. (14) Table 1

and (15), with η′+ replacing η þ , as Values of electrochemical parameters used for simulation of SLRFB.

⎡ ⎛ 0.5Fη ⎞ ⎛ −0.5FγjPbO Rs ⎞ Parameter Value Parameter Value

jPbO2 = i0 ⎢ exp ⎜ +
⎟ exp ⎜ 2
⎟
⎢⎣ ⎝ RT ⎠ ⎝ RT ⎠ D Pb2 + a 0.61e  9 m2/s kPbO2 6.25e  5 mol/m2/s
⎛ −0.5Fη ⎞ ⎛ 0.5FγjPbO Rs ⎞ ⎤ D H+b 5.66e  9 m2/s kPb 2.75e  3 mol/m2/s
− exp ⎜ +
⎟ exp ⎜ 2
⎟⎥ DCH3SO− c 2
2.0e  10 m /s kf 2.5e  4 mol/m2/s
⎝ RT ⎠ ⎝ RT ⎠ ⎥⎦ (27) 3
z Pb2 + 2 kb 2.5e  6 mol/m2/s

where z H+ 1 ν 1.4
zCH3SO− 1 s, ref
CPbO 1 mol/m2
⎛ ⎞⎛ ⎞ 3
c 2 + ⎟ ⎜ cH + ⎟ 3
1 mol/m2
i0 = 2Fk PbO 2 ⎜ Pb c0 500 mol/m s, ref
CPbO
Pb2 +
⎜ c ref ⎟ ⎜ c ref+ ⎟ 2
⎝ 2+ ⎠⎝ H ⎠
Pb c 0+
H
500 mol/m3 c ref2 + 500 mol/m3
Pb
T 300 K c ref+ 500 mol/m3
and H
pref 100 kPa E+0 1.56 V
⎛ C s ⎞ν ⎛ ⎞ ⎛ ⎞
0.9Fη+ ⎟ −0.9FλjPbO Rs ⎟
jPbO = 2Fkf ⎜⎜ sPbO ⎟ exp ⎜ exp ⎜⎜ μ 0.001 Pa s E−0 d  0.13 V
, ref ⎟ ⎜ ⎟ ⎟
⎝ CPbO ⎠ ⎝ RT ⎠ ⎝ RT ⎠ ρ0 1169 kg/m3 β Pb2 + 2.573e  4 m3/mol
F 96485 C/mol β H+ 3.328e  5 m3/mol
s
CPbO ⎛ ⎞ ⎛ ⎞
2 cH −0.9Fη+ ⎟ 0.9FλjPbO Rs ⎟
exp ⎜⎜ exp ⎜⎜
+ R 8.314 J/mol/k L 0.03 m
− 2Fkb s, ref ⎟ ⎟ W 0.015 m
CPbO c ref+
2 H ⎝ RT ⎠ ⎝ RT ⎠ (28)
a
Hazza et al. (2004).
Here, 1/γ and 1/λ represent fractional contribution of currents jPbO2 b
Soudijn (2012).
c
and jPbO to the total current jan, and satisfy the constraint: Telfah et al. (2010).
d
Zhang et al. (2011).
1 1
+ =1
γ λ (29) potential needed for the current to flow corresponds to an as-
along with γ, λ > 1. Further simplification was carried out by ex- sumed ratio of current carried by the two reactions (through fixed
panding exponential terms involving deposit resistance Rs and values of λ and γ). The values of these variables were changed
retaining only the first two terms. The new Butler–Volmer equa- later, while satisfying the constraint ( λ−1 + γ −1 = 1), to see their
tions for jPbO2 and jPbO take the following form: impact on model predictions.
The value of local deposit thickness lt, which also appears on
⎡ ⎛ 0.5Fη ⎞ ⎛ −0.5Fη+ ⎞ ⎤
i0 ⎢ exp ⎜ RT + ⎟ − exp ⎜ ⎟⎥ the right hand side through Rs, is uniquely related to the local and
⎣ ⎝ ⎠ ⎝ RT ⎠ ⎦ time varying values of C PbO2 and CPbO. In order to keep the equa-
jPbO2 =
0.5Fi γR ⎡ −0.5Fη+ ⎤
(
0.5Fη
1 + RT0 s ⎣⎢ exp RT + + exp ) RT (
⎦⎥ ) (30)
tions explicit, we have followed the same approach for lt as well,
and fixed its value to 250 μm. We will establish in the next section
and that the model predictions are not sensitively dependent on the

⎛ C s ⎞ν ⎛ ⎞ s
CPbO ⎛ ⎞
0.9Fη+ ⎟ c + −0.9Fη+ ⎟
2Fkf ⎜⎜ sPbO ⎟ exp ⎜ − 2Fkb s, ref2 Href exp ⎜⎜
, ref ⎟ ⎜ ⎟ ⎟
⎝ CPbO ⎠ ⎝ RT ⎠ CPbO 2 c H + ⎝ RT ⎠
jPbO =
⎛ C s ⎞ν ⎛ ⎞ s
CPbO ⎛ ⎞
0.9Fη+ ⎟ 2 c H 0.9FλRs ⎜ −0.9Fη+ ⎟
⎟ 0.9FλRs exp ⎜
+
1 + 2Fkf ⎜⎜ sPbO
, ref ⎟ ⎜ RT ⎟ + 2 Fk b s, ref
exp ⎜ RT ⎟
⎝ CPbO ⎠ RT ⎝ ⎠ CPbO c ref+ RT
2 H ⎝ ⎠ (31)

Eqs. (30) and (31) differ from the classical Butler–Volmer
equation in the denominator which represents increased over-
potential required to provide for ohmic drop across deposits. These
equations allow us to go from one limiting case to other. When the
resistance offered by the deposits is zero (Rs ¼0), the above
equations become identical to those used earlier in the literature
(Eqs. (14)–(16)). When the resistance offered by deposits ap-
proached a large value, due to the presence of large proportion of
PbO, the current density approaches zero locally.
The above two equations are still implicit as γ = jan /jPbO2 and
λ = jan /jPbO appear on the right hand side, and require highly
iterative and intensive computations even for the simplified 2-D
geometry (because the cell voltage needs to be first guessed at
every time instant). To obtain explicit equations, we assumed the
values of γ and λ to be the same and equal to 2. Thus, the re-
sistance Rs offered by the deposits is computed based on the
proportion of PbO and PbO2 present on it locally, but the over-
values of λ, γ, and lt over a wide range.
The above explicit equations for reaction kinetics were solved along
with the other model equations using COMSOL4.3 under galvanostatic
conditions (Nandanwar and Kumar, 2016b). Boundary layer grid was
used near electrode surfaces where stiff gradients of velocity and
concentrations are expected. A triangular mesh was used in the rest of
the numerical domain. A second order accurate discretization scheme
was used for concentrations and potential in the electrolyte domain.
P1þP1 discretization (Gresho et al., 1999) was used for velocity and
pressure fields. Streamline diffusion and crosswind diffusion methods
were used to stabilize numerical solution. Mixing during the relaxa-
tion period was simulated by setting buoyant force F to zero in Eq. (8)
and by increasing diffusion coefficients artificially for all the species to
let the concentration gradients in the domain die out. Mass con-
servation in this process was monitored.
In summary, we have developed a model which accounts for
natural convection, spatially non-uniform deposition/dissolution
 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71 67

Table 2 literature (Bode, 1977). As the molar volumes of PbO2 (25.5 cm3/mol)
Values of natural convection related parameters used for simulation of SLRFB. and PbO (23.4 cm3/mol) are nearly the same, we replaced volume
proportions (Aa and Ab) in Eq. (3) with the respective molar pro-
Parameter Value Parameter Value
portions. The approximation introduces less than 10% error in the
μ 0.001 Pa s ρ0 1169 kg/m3 value of effective conductivity (se).
β H+ 3.328  10  5 m3/mol β Pb2 + 2.573  10  4 m3/mol The model, with composition dependent conductivity of de-
W 0.015 m L 0.03 m posits, was used to simulate voltage vs. time profile for current
density of japp = 200 A/m2. The experimental measurements under
these conditions are reported earlier (Nandanwar and Kumar,
Table 3
2016b). The deposits formed at the end of the first charge cycle are
Values of parameters of conductivity model used for simulation of SLRFB.
highly conducting PbO2 on anode and Pb on cathode. The model
Parameter Value Parameter Value therefore makes identical predictions as those obtained with the
model of Nandanwar and Kumar (2016b) based on infinite con-
z 8 lt 250 (μm) ductivity of deposits. We will henceforth refer to the latter pre-
sa 1.25  10  12 (S/m) sb 5  105 (S/m)
dictions as those from the base model with no conductivity effects.
Since the predictions of the earlier model are in good agreement
on electrode surface, and composition dependent conductivity of with the measurements, we have not shown any comparisons for
deposits on anode. It includes in it, over and above the model of the first charge cycle. Fig. 3 instead shows model predictions for
Nandanwar and Kumar (2016b), the effect of finite conductivity of the first discharge cycle (bottom) and the second charge cycle
deposits on anode which varies from highly conducting to non- (top) for z¼8 (the number of contacts per particle). The corre-
conducting with evolving composition of deposits. The new model sponding set of predictions from the base model are also shown.
has coordination number z alone as an unknown parameter. The The figure shows that the two sets of predictions for the first
values of all the other parameters in expressions for Butler–Volmer discharge cycle are nearly identical for the most part, except to-
kinetics and Nernst potential are taken from the literature. Al- wards the end, when the rapidly changing conductivity of deposits
though the deposit thickness lt, λ, and γ can be calculated using the on anode makes a difference. The overprediction seen in the early
model, these are introduced as parameters. The composition of stages for both the models is similar to the overprediction seen in
deposits which causes orders of magnitude change in resistance is the conventional lead-acid battery, which is explained by the over-
determined exactly from the model. potential required for nucleation of new phase. We conjecture a
similar explanation for SLRFB.
The predictions of the two models for the second charge cycle
5. Results and discussion (Fig. 3, top) are quite different at short times. In the base model,
the battery starts to charge at the voltage required to oxidize PbO
The model equations were solved for the rectangular cell shown (Nandanwar and Kumar, 2014). In the early stages of the second
in Fig. 2 with no external flow. The parameter values used to simulate charge cycle, most of the transferred charge on anode comes from
the model are shown in Tables 1–3. The parameter values are the the conversion of PbO to PbO2 which occurs as at lower potentials.
same as those reported by Nandanwar and Kumar (2014) for flow- With depletion of PbO, the transferred charge starts to come from
through SLRFB, also used by Nandanwar and Kumar (2016b) to deposition of Pb2 þ as PbO2, and the battery starts to charge at
successfully simulate natural convection in battery cell. The values of higher voltage. The charging voltage predicted by the base model
expansion coefficients βi are measured experimentally by Nandan- thus slowly increases with time, from the beginning.
war and Kumar (2016b). The conductivity values are taken from the The incorporation of composition dependent conductivity of
deposits in the present model leaves anode at the end of the
previous discharge cycle with very poorly conducting deposits.
The battery cell therefore needs higher voltage in the next charge
cycle to start the charging process. The formation of conducting
PbO2 even in small amount increases the conductivity of deposits.
The charging voltage rapidly comes down and joins the charging
voltage for the base model at around 100 s. The model based ex-
planation for charge coup de fouet phenomenon is the poor con-
ductivity of deposits towards the end of the previous discharge
cycle, and rapid increase in conductivity of deposits early in the
next charge cycle.

5.1. Parameter sensitivity

The model consists of parameters z, lt, γ, and λ. The latter two

Fig. 3. A comparison of experimentally measured cell voltage vs. time for first
discharge cycle (bottom) and second charge cycle (top) with the predictions ob-
tained using the base model which did not consider conductivity effects and the
present model which incorporates the effect of finite conductivity of deposits on
electrodes.
satisfy the constraints λ−1 + γ −1 = 1, and λ , γ > 1. The sensitivity of
model predictions to these parameter values needs to be ascer-
tained. Fig. 4(a) shows insensitivity of predictions to coordination
number z for the most part and a slight increase in discharge time
(and charge efficiency) with an increase in coordination number:
91.1% for z ¼6, 94.1% for z¼ 8, and 95.1% for z ¼10. The coordination
number z is visualized as the number of current carrying contacts
a particle has with its neighbors. The typical values for z are: 4 for
diamond packing, 6 for simple cubic packing, 8 for body centered
packing, and 12 for face centered and hexagonal closed packing
(Euler et al., 1980). The value of z for complex mixtures needs to be
68 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71

To understand the insensitivity of model prediction to these

parameters, we calculated voltage drop across a 250 μm thick
deposit film as a function of percentage of PbO2 in it, for two
current densities using Eq. 21. The results presented in Fig. 5 show
that at moderate fraction of PbO2, the required voltage drop is less
than 0.1 mV even at a current density of 1000 A/m2, which is
negligibly small compared with the typical values of overpotential
in the range of 100 mV (Fig. 1). The required voltage drop increases
steeply to infinity for both the current densities as the fraction of
PbO2 in deposits approaches the critical limit of 25% PbO2. Over a
short range of composition, the resistivity Rs increases by a factor
of 1015 and the deposits change from being highly conducting to
non-conducting. The value of ΔVs is thus either very low (ON) or
very high (OFF), which is largely controlled by the value of Rs
(through se). The values of parameters γ, λ, and lt therefore have
very little effect on model predictions.
In view of the extreme variation of Rs, Eqs. (30) and (31) can be
further approximated to:
⎡ ⎛ 0.5Fη+ ⎞ ⎛ −0.5Fη+ ⎞ ⎤
i0 ⎢ exp ⎜ ⎟ − exp ⎜ ⎟⎥
⎣ ⎝ RT ⎠ ⎝ RT ⎠ ⎦
jPbO2 =
1 + R¯ s (32)

and
⎛ C s ⎞ν ⎛ ⎞ s
C PbO ⎛ ⎞
0.9Fη+ ⎟ 2 c H+ −0.9Fη+ ⎟
2Fkf ⎜⎜ PbO ⎟⎟ exp ⎜ − 2Fkb exp ⎜
s , ref ⎜ ⎟ s , ref ref ⎜ ⎟
⎝ CPbO ⎠ ⎝ RT ⎠ C PbO c +
2 H ⎝ RT ⎠
jPbO = ,
1 + R¯ s (33)

Fig. 4. Effect of parameter values on discharge characteristics of battery, as pre-
dicted by the present model.
( (
where R̄s is the scaled resistance =Rs / RT /kf F2 . The predictions))
obtained using the above equations are also shown in Fig. 4(b). The
good agreement seen among all the predictions establishes that
charge coup de fouet is largely controlled by the model used to
represent rapid decrease in conductivity of deposits with forma-
tion of PbO.

5.2. Distribution of current and solids on electrodes

The inclusion of finite conductivity of deposits greatly influ-

ences model predicted composition of deposits on anode. Fig. 6
shows concentration profiles of PbO2 and PbO residue at the end of
the first discharge cycle for the base model and the present model.
The base model predicts formation of a patch which has no PbO2
in it. The present model in comparison predicts the presence of
PbO2 everywhere on anode. The PbO2 and PbO concentrations in
the upper half are non-uniform—larger near the edge due to the

Fig. 5. Voltage drop across 250 μm thick film of deposit as a function of % PbO2 in
the deposits at two current densities.

obtained experimentally. Gandhi (2015) has used z¼ 8 in his

model of lead-acid battery. The measured efficiency of 92.0% in the
present case supports a z value in the range of 6–8. We have
therefore used z ¼8 in all the subsequent model predictions.
Simulations were next carried out to test sensitivity of the
model to values of γ (and λ). The results presented in Fig. 4(b) for
three widely different values of γ (10, 2, and 100/99) show that the
discharge profiles are identical for the three cases. A ten fold in-
Fig. 6. The predicted concentration profiles of PbO2 and PbO deposits (in mol/m2)
crease and decrease in deposit layer thickness lt also did not on anode at the end of first discharge cycle. The predictions are obtained using the
change the predicted discharge profiles. base model (without the conductivity effects included in it) and the present model.
 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71
Fig. 7. Variation of ratio [PbO2]/[PbO] (top) and associated conductivity of the
mixture (bottom) on anode at various times during discharge cycle.
edge effect (Nandanwar and Kumar, 2014), but the fraction of PbO2
in deposits is uniform at 0.25 (the ratio of PbO2 to PbO is 0.33), the
value at which deposit conductivity becomes zero for z¼ 8. The
early termination of discharge cycle at the same cut-off voltage
leaves significantly larger PbO2 residue on anode. The amount of
PbO formed on anode is the same for the two models, because the
formation rate is high in the beginning.
The charge efficiency for the first charge-discharge cycle as
predicted by the present model (94.1%) is therefore smaller than
that for the base model (97.4%), and closer to the experimentally
measured value of 92.0%. Interestingly, the charge efficiency ob-
tained for the present model is far larger than 75% (corresponding
to 25% PbO2 that remains trapped in residue) because a substantial
part of PbO2 dissolves as Pb2 þ as well.
Fig. 7 (top) shows the variation of ratio of surface concentration
of PbO2 to PbO as a function of distance from the bottom of anode
at different times during discharge. The corresponding con-
ductivity profile of deposits is shown at the bottom. The con-
ductivity of the solid deposits lies in the range of 105 S/m for ratio
[PbO2]/[PbO] greater than 0.33. It drops quite abruptly by 15 or-
ders of magnitude for the value of ratio approaching 0.33. A por-
tion of the electrode thus turns into insulator. The voltage drop
seen earlier in Fig. 3 beyond 600 s is due to the increase in current
density on the remaining part of anode to sustain galvanostatic
discharge. The non-conducting part of the electrode still has a
significant amount of PbO2 trapped in it. Thus, although the parts
of the electrode are conducting and non-conducting and are also
separated clearly, PbO2 and PbO residue are present all over the
electrode at the end of the discharge cycle. It may not be possible
to visually distinguish the two regions. This possibly explains the
visually uniform texture/morphology of anode reported by Nan-
danwar and Kumar (2016b), which could not be explained by their
model based on zero resistance of deposits. Both the models
predict formation of a bare patch on cathode on the upper side,
which is validated by the measurements (Nandanwar and Kumar,
2016b).
The step decrease in conductivity of deposits makes local cur-
rent density zero. Thus, towards the end of the first discharge
cycle, a receding current peak moves towards the bottom (left in
the figure), as shown in Fig. 8 (bottom). The movement is similar
what is predicted by Nandanwar and Kumar (2014) for flow-
69
Fig. 8. Movement of location of peak current with time during first discharge cycle
(bottom) and charge second charge cycle (top), as predicted by the present model.
through SLRFB. The dynamics of charging in the next cycle, in
comparison, is quite different and interesting. When the con-
ductivity effects are not taken into account, the entire electrode
surface participates in chemical reactions and contributes to the
total current from the beginning itself (Nandanwar and Kumar,
2014, 2016b). With only a part of the anode remaining conducting
at the end of the previous discharge cycle, the present model
predicts a peak in current density in the next charge cycle at t¼0
at the interface of the conducting and non-conducting regions on
anode, as shown in Fig. 8 (top). The peak moves up rapidly to the
top (right in the figure) and reaches the upper edge of the elec-
trode in less than 100 s, to mark conversion of the entire insulating
part of anode to conducting one. The movement is faster initially.
Conversely, in the discharge cycle, the peak movement is slower in
the beginning. The latter is explained in detail by Nandanwar and
Kumar (2014). The explanation for the former is provided below.
The conversion of non-conducting zone to conducting zone
requires addition of conducting PbO2 to it. This can be achieved
through deposition of Pb2 þ as PbO2 or conversion of local PbO to
PbO2. Both the reactions require conduction of electrons and H þ
ions next to the reaction site for Butler–Volmer kinetics. The
highest concentration of H þ is in the electrolyte rising up, past the
end point of the conducting zone, because H þ ions are produced
on the conducting part of anode. For a fixed applied current,
smaller the area of the conducting part, larger is the current
density on it and higher is the concentration of H þ in the rising
liquid past the end point of the conducting zone. The conversion of
non-conducting part to conducting one occurs at this point which
moves up with time. As the surface area of the conducting part
increases with time, the current density on it decreases and so
does the concentration of H þ in the rising liquid past the end
point of conducting zone. The expansion of the conducting zone
therefore slows down with time.
5.3. Effect of the depth of discharge (DOD)
It is known that the magnitude of charge coup de fouet in
conventional lead acid batteries is influenced by the depth of
discharge (DOD) in the previous cycle. In order to see if a SLRFB
shows similar behavior, we experimentally studied charging be-
havior after discharging the battery in the previous cycle to dif-
ferent cut-off voltages. The experiments were conducted in the
rectangular cell, described earlier, in the absence of any forced
70 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71
Fig. 9. Effect of depth of discharge on magnitude of charge coup de fouet observed
in individual experiments when cut-off voltage for previous discharge cycle is
varied.
convection. The battery was first charged for 900 s, followed by
five minutes of relaxation with gentle mixing. In the discharge
cycle, different cut-off voltages were used to terminate the cycle. A
five minute relaxation period was provided with gentle mixing,
before the onset of the next charge cycle. The results obtained for
three cut-off voltages (1.2 V, 1.3 V and 1.35 V) in separate experi-
ments are presented in Fig. 9(a). The figure shows that the mag-
nitude of charge coup de fouet (difference between the peak and
the plateau voltage) increases with a decrease in cut-off voltage in
the previous cycle, in agreement with the reported observations
for the conventional lead-acid battery (Delaille et al., 2006). The
magnitudes of charge coup de fouet are 0.047 V, 0.026 V, and 0.01 V
for cut-off voltages of 1.2 V, 1.3 V, and 1.35 V respectively in the
previous cycle.
The measurements do not show any regular pattern in charging
voltage at later times, however. This could be caused by varying
external contact resistance when new connections are made for
separate experiments. In order to establish that the observed effect
is due to anode only, a pre-deposited cathode (for 3600 s at
200 A/m2) was used. It also eliminated any complications arising
out of partial baring of cathode during the experiments. A single
experiment without removing the contacts was performed next: a
Fig. 10. The model predicted effect of depth of discharge in previous discharge
cycle on the magnitude of charge coup de fouet in the second charge cycle.
charge cycle of 900 s, followed by 5 min of relaxation, followed by
discharge cycle with decreasing cut-off voltages, and then char-
ging. The cut-off voltages for the first, second, and third discharge
cycles were 1.35 V, 1.3 V, and 1.2 V respectively. The results ob-
tained are presented in Fig. 9(b). The figure shows similar trend as
before, with improved control on the long time response during
charging. These measurements establish that charge coup de fouet
effect is related to anode, and its magnitude increases with de-
crease in cut-off voltage in the previous cycle.
The present model was simulated with different cut-off voltages
to compare model predictions with observations. Fig. 10 shows the
observed increase in the magnitude of charge coup de fouet as cap-
tured by the model. The experimentally measured magnitudes of
charge coup de fouet are about 0.03 V and 0.05 V for cutoff voltages of
1.3 V and 1.2 V respectively. The model predicted values of 0.03 V and
0.08 V respectively are obtained without fitting any parameters. Gi-
ven the complex nature of the underlying phenomenon, the agree-
ment obtained between the predictions and the observations is
considered good. The experimentally observed steep rise in voltage
in a very short time interval in the beginning of the experiment is a
well known feature (Budevski et al., 2000; Guo and Searson, 2010)
and is not attempted by the present set of models. It is believed to
arise due to the overpotential required for formation of nuclei of the
depositing material on foreign substrate.
6. Conclusions
The charge coup de fouet phenomenon, in which charging vol-
tage of a fully discharged battery goes through a minimum, is
systematically investigated in this work in the context of the so-
luble lead redox flow battery. The experiments show that an in-
crease in the depth of discharge, effected by decreasing the cut-off
voltage for termination of discharge cycle, increases the magni-
tude of charge coup de fouet. The models available in the literature
in which the conductivity of deposits is assumed to be infinitely
large fail to explain the observed phenomenon. A new model is
developed which incorporates in it variation of conductivity of
deposits with composition, from highly conducting to non-con-
ducting when conducting pathways cease to percolate beyond a
critical volume fraction of non-conducting phase. The orders of
magnitude change in conductivity of a thin layer of deposits on
conducting graphite electrodes is used to develop an approxima-
tion that permits model predictions to be obtained without de-
termining voltage profile in the composite solid phase. The model
explains the observed charge coup de phenomenon and its var-
iation with depth-of-discharge well.
 M. Nandanwar, S. Kumar / Chemical Engineering Science 154 (2016) 61–71
The model is general and applicable to situations involving a
conducting electrode turning itself into a non-conducting one due
to the formation of an insulating phase, such as in the conven-
tional lead-acid battery.
Acknowledgments
The authors are grateful to Prof. K.S. Gandhi for many useful
discussions. The authors also acknowledge financial support re-
ceived from the Department of Science and Technology, New
Delhi, India through its IRHPA scheme.
Appendix A. Supplementary data
Supplementary data associated with this article can be found in
the online version at http://dx.doi.org/10.1016/j.ces.2016.07.001.
References
Banerjee, A., Saha, D., Guru Rao, T.N., Shukla, A.K., 2013. A soluble-lead redox flow
battery with corrugated graphite sheet and reticulated vitreous carbon as po-
sitive and negative current collectors. Bull. Mater. Sci. 36 (1), 163–170.
Bates, A., Mukerjee, S., Lee, S.C., Lee, D., Park, S., 2014. An analytical study of a lead-
acid flow battery as an energy storage system. J. Power Sources 249, 207–218.
Beck, F., Bhn, H., 1975. Bleidioxid als lsungselektrode. Berichte der Bunsenge-
sellschaftfr physikalische Chemie, vol. 79(3), pp. 233–244.
Berndt, D., Voss, E., 1965. The voltage characteristics of a lead-acid cell during
charge and discharge. In: Collins, D.H. (Ed.), Batteries 2: Research and devel-
opment in non-mechanical electrical power sources. In: Proceedings of the 4th
International Symposium held at Brighton, September 1964, Pergamon Press,
pp. 17–27.
Bode, H., 1977. Lead-Acid Batteries', first edition. J. Wiley& Sons, New York
(translated by RJ Brodd and KV Kordesch).
Bose, C.S., Laman, F.C., 2000. Battery state of health estimation through coup de
fouet. In: Preceedings of the Telecommunications Energy Conference, 2000.
INTELEC. Twenty-second International, IEEE, pp. 597–601.
Budevski, E., Staikov, G., Lorenz, W.J., 2000. Electrocrystallization: Nucleation and
growth phenomena. Electrochim. Acta 45 (15), 2559–2574.
Collins, J., Kear, G., Li, X., Low, C.T.J., Pletcher, D., Tangirala, R., Stratton-Campbell, D.,
Walsh, F.C., Zhang, C., 2010. A novel flow battery—A lead acid battery based on
an electrolyte with soluble lead(II): VIII. The cycling of a 10 cm  10 cm flow
cell. J. Power Sources 195, 1731–1738.
Delaille, A., Perrin, M., Huet, F., Hernout, L., 2006. Study of the coup de fouet of lead-
acid cells as a function of their state-of-charge and state-of-health. J. Power
Sources 158 (2), 1019–1028.
Euler, K.J., Kirchhof, R., Metzendorf, H., 1980. Electronic conductivity of battery
active masses–a limiting factor? J. Power Sources 5, 255–262.
Gandhi, K.S., 2015. Role of electrical resistance of electrodes in modeling of dis-
charging and charging of flooded lead-acid batteries. J. Power Sources 277,
124–130.
Gebhart, B., Pera, L., 1971. The nature of vertical natural convection flows resulting
from the combined buoyancy effects of thermal and mass diffusion. Int. J. Heat.
Mass. Transf. 14, 2025–2050.
Gresho, P.M., Sani, R.L., Engelman, M.S., 1999. Incompressible Flow and the Finite
Element Method. John Wiley and Sons, Chichester, pp. 457–470.
71
Guo, Lian, Searson, Peter C., 2010. On the influence of the nucleation overpotential
on island growth in electrodeposition. Electrochim. Acta 55 (13), 4086–4091.
Hazza, A., Pletcher, D., Wills, R., 2004. A novel flow battery—A lead acid battery
based on an electrolyte with soluble lead(II): I. Preliminary studies. Phys. Chem.
Chem. Phys. 6, 1773–1778.
Hazza, A., Pletcher, D., Wills, R., 2005. A novel flow battery—A lead acid battery
based on an electrolyte with soluble lead(II): IV. The influence of additives. J.
Power Sources 149, 103–111.
Li, X., Pletcher, D., Walsh, F.C., 2009. A novel flow battery—A lead acid battery based
on an electrolyte with soluble lead(II): VII. Further studies of the lead dioxide
positive electrode lead acid batterybased on an electrolyte with soluble lead(II)
Part VII. further studies of the lead dioxidepositive electrode. Electrochim. Acta
54, 4688–4695.
Nandanwar, M.N., Kumar, S., 2014. Modelling of effect of non-uniform current
density on the performance of soluble lead redox flow batteries. J. Electrochem.
Soc. 161 (10), A1602–A1610.
Nandanwar, M.N., Kumar, S., 2016a. A soluble lead redox flow battery with no
externalflow during charging/discharging. J. Power Sources, Submitted for
publication.
Nandanwar, M.N., Kumar, S., 2016b. Soluble lead redox (natural convection) flow-
battery. J. Electrochem. Soc., Submitted for publication
Newman, J., Thomas-Alyea, K.E., 2004. Electrochemical Systems, 3rd edition. John
Wiley & Sons, Hoboken, N.J., p. 672.
Oury, A., Kirchev, A., Bultel, Y., 2014. A numerical model for a soluble lead-acid flow
battery comprising a three-dimensional honeycomb-shaped positive electrode.
J. Power Sources 246, 703–718.
Oury, A., Kirchev, A., Bultel, Y., Chainet, E., 2012. PbO2/Pb2 þ cycling in methane-
sulfonic acid and mechanisms associated for soluble lead-acid flow battery
applications. Electrochim. Acta 71, 140–149.
Pascoe, Phillip E., Anbuky, Adnan H., 2002. The behaviour of the coup de fouet of
valve-regulated lead-acid batteries. J. Power Sources 111 (2), 304–319.
Pascoe, P.E., Anbuky, A.H., 1999. Estimation of vrls battery capacity using the ana-
lysis of the coup de fouet region. In: Proceedings of the Telecommunication
Energy Conference, 1999. INTELEC'99. The 21st International, IEEE, pp. 114–122.
Pascoe, P.E., Sirisena, H., Anbuky, A.H., 2002. Coup de fouet based VRLA battery
capacity estimation. In: Proceedings of the Electronic Design, Test and Appli-
cations. The First IEEE International Workshop on, IEEE, pp. 149–153.
Pletcher, D., Wills, R., 2004. A novel flow battery—A lead acid battery based on
anelectrolyte with solublelead(II): II. Flow cell studies. Phys. Chem. Chem. Phys.
6, 1779–1785.
Pletcher, D., Wills, R., 2005. A novel flow battery—A lead acid battery based on an
electrolyte with soluble lead(II): III. The influence of conditions on battery
performance. J. Power Sources 149, 96–102.
Pletcher, D., Zhou, H., Kear, G., Low, C.T.J., Walsh, F.C., Wills, R., 2008a. A novel flow
battery—A lead-acid battery based on an electrolyte with soluble lead(II): VI.
Studies of the lead dioxide positive electrode. J. Power Sources 180, 621–629.
Pletcher, D., Zhou, H., Kear, G., Low, C.T.J., Walsh, F.C., Wills, R., 2008. A novel flow
battery—A lead-acid battery based on an electrolyte with soluble lead(II):
Studies of the lead negative electrode. J. Power Sources 180, 630–634.
Schmickler, W., Santos, E., 2010. Interfacial Electrochemistry, 2nd edition. Springer-
Verlag, Heidelberg, pp. 145–150.
Shah, A.A., Li, X., Wills, R.G.A., Walsh, F.C., 2010. A mathematical model for the
soluble lead-acid flow battery. J. Electrochem. Soc. 157 (5), A589–A599.
Soudijn, M., 2012. Proton Transport in Aqueous Ionic Solutions. University of Am-
sterdam, Master’s thesis, 53-54.
Telfah, A., Majer, G., Kreuer, K.D., Schuster, M., Maier, J., 2010. Formation andmo-
bility of protonic charge carriers in methyl sulfonic acidwater mixtures: a
modelfor sulfonic acid based ionomers at low degree of hydration. Solid State
Ion. 181, 461–465.
Verde, M.G., Carroll, K.J., Wang, Z., Sathrum, A., Meng, Y.S., 2013. Achieving high
efficiency and cyclability in inexpensive soluble lead flow batteries. Energy
Environ. Sci. 6, 1573–1581.
Zhang, C.P., Sharkh, S.M., Li, X., Walsh, F.C., Zhang, C.N., Jiang, J.C., 2011. The per-
formance of a soluble lead-acid flow battery and its comparison to a static lead-
acid battery. Energy Convers. Manag. 52, 3391–3398.