Hamid JPS 2023

Journal of Power Sources 565 (2023) 232756
Contents lists available at ScienceDirect
Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour
A bottom-up, multi-scale theory for transient mass transport of redox-active

species through porous electrodes beyond the pseudo-steady limit
Md Abdul Hamid a, Kyle C. Smith a, b, c, d, *
a
Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, 61801, USA
b
Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, 61801, USA
c
Computational Science and Engineering Program, University of Illinois at Urbana-Champaign, Urbana, IL, 61801, USA
d
Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, IL, 61801, USA
H I G H L I G H T S G R A P H I C A L A B S T R A C T
• Bottom-up theory is presented for tran

sient multi-scale modeling of flow
batteries.
• Pore-scale transfer functions: spectral
Sherwood number and advective flux.
• Transient polarization matches experi
ment, while conventional model fails.
• Dominant pore-scale transport physics
change with frequency based on Péclet
number.
• Operation map shows over-limiting
current regime for short current pulses.
A R T I C L E I N F O A B S T R A C T
Keywords: New theory is presented for the dynamic response of redox-active electrolyte flowing through porous electrodes
Flow battery under time-dependent applied current. This is done by introducing certain frequency-dependent transfer func
Concentration polarization tions (TFs) that incorporate pore-scale transport physics. One TF – dubbed the spectral Sherwood number –
Transfer function
extends the film law of mass transfer (FLoMT) to transient conditions. Another TF captures the acceleration/
Over-limiting current
suppression of solute advection that results from pore-scale velocity/concentration gradients. Numerical results
are shown for the frequency-dependent TFs of solid cylinders in crossflow to represent porous electrodes
commonly used in flow batteries (FBs). Spectral regions are observed where a transition from lagless response to
semi-infinite Warburg response occurs with increasing frequency. The embedding of these TFs into an up-scaled
model is also formulated to obtain the time-domain response of FBs. Without adjustable parameters this model
predicts polarization in agreement with transient FB experiments, despite systematic overprediction by the
conventional FLoMT model. Analysis of concentration polarization and reactant concentration is also used to
construct non-dimensional maps of operational space. These predictions show that fast current fluctuations are
sustained even when current exceeds the limiting current expected from the time-invariant FLoMT without solute
advection suppression/acceleration, suggesting implications for electrochemical conversion and separations
devices in addition.
* Corresponding author. Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, 61801, USA.
E-mail address: kcsmith@illinois.edu (K.C. Smith).
https://doi.org/10.1016/j.jpowsour.2023.232756
Received 10 May 2022; Received in revised form 13 December 2022; Accepted 29 January 2023
Available online 9 March 2023
0378-7753/© 2023 Elsevier B.V. All rights reserved.
M.A. Hamid and K.C. Smith Journal of Power Sources 565 (2023) 232756
1. Introduction frequency-based formalism to predict transient dynamics of solid-state

diffusion processes [39] and to model electrochemical impedance
Renewable energy sources are projected to comprise approximately spectra [40–42]. In this article we introduce a new theory for modeling
60% of US electricity generation by 2050, with wind and solar power transient pore-scale mass transfer with simultaneous convection and
contributing approximately 2000 and 1000 TW-h respectively [1]. redox reactions using a frequency-response formalism for the first time
Electricity generation from wind and solar power plants requires that enables direct up-scaling of the coupling between pore-scale con
installation of large-scale energy-storage devices to overcome the centration polarization and temporal variations of current supply and
mismatch between electricity demand and intermittent supply arising demand. In this article, we do this by defining a new spectral Sherwood
from the stochastic nature of wind and the diurnal cycle of solar inso ̃ that relates average surface flux 〈j〉 (i.e.,
number transfer function 〈Sh〉 s
lation [2]. Redox flow batteries (RFBs) have drawn attention due to their reaction rate) to the associated concentration difference at the
ability to independently scale power and energy-storage capacity
pore-scale Δc = cs − 〈c〉 in the frequency domain: 〈Sh〉
̃ ≡ 〈j〉 (b /D)/(cs −
s
because of the modularity of their reactors that dictate power capability
〈c〉), where A is the continuous-time Fourier transform of quantity A, b is
and of their tanks that dictate energy-storage capacity. Here,
a characteristic pore length-scale, and D is diffusivity.
redox-active liquid electrolyte is circulated through RFB electrodes to
The rate capability of electrochemical devices is determined by the
and from tanks, forming a closed loop. Despite advances in RFB elec
amount of electric current produced and various mechanisms of polar
trolyte chemistry [3–7], the surface chemistry of their electrodes [8–15],
ization. In RFBs and other devices using flow-through electrodes (e.g.,
and the design of flow fields [16–20], limited understanding has been
electrochemical deionization [43–45]), continuous supply of fresh
established to date concerning the impact of the inherently transient and
electrolyte to electrode pores and the simultaneous evacuation of spent
irregular charge/discharge scenarios that RFBs are likely to experience
electrolyte from electrode pores is facilitated by bulk flow, i.e., advec
on renewable-powered grids and micro-grids. For instance, past RFB
tive flux. The advective flux of soluble species is affected by their
models have adopted a film law of mass transfer using a
depletion or accumulation near reacting interfaces where such species
time-independent mass transfer coefficient to predict pore-scale con
are respectively consumed or produced. Previous models of electro
centration polarization and associated energy losses (e.g., in Refs. [16,
chemical devices using flow-through electrodes have adopted a
21–25]).
macro-homogeneous advection-reaction-dispersion(-migration) (ARD
Here, we posit that understanding of such concentration polarization
(M)) theory, building on earlier geochemical theory and practice [46],
must start from the bottom-up at the scale of the electrode pores at the
that implicitly assumes that the local volume-averaged solute concen
surfaces on which redox reactions take place, in concert with the flow of
tration is advected by the bulk superficial velocity of electrolyte [24,25,
electrolyte and the transport of soluble species occurring therein.
43–45]. In contrast, previous modeling of unconventional RFBs using
Because all three modes of soluble species transport (diffusion, advec
mixed-conducting suspensions flowing through open channels – rather
tion, and migration) contribute to total flux, the time variation of cur
than flowing liquid electrolyte through porous electrodes – showed that
rent and cell voltage – induced by the instantaneous power supplied to
coulombic and energy efficiency can be enhanced by uniformizing
or demanded from the cell incorporating that electrode – drives the
advective transport via the control of interfacial slip [47], rheology
associated transport of species in solution via redox reactions at elec
[47–49], and fluid pulsation [47,48,50], the effects of which are not
trode surfaces. Here, the power spectral density (PSD) distribution
captured by conventional ARD(M) theory. Studies motivated by other
associated with any given power-supply or -demand entity indirectly
applications have previously modeled the effects of interfacial reactions
captures their transient and stochastic qualities using a frequency-based
on advection in porous media by accounting for the difference between
representation [26,27]. Wind power experiences an especially wide PSD
mean solute velocity and mean solvent velocity caused by the in
distribution relative to other renewable sources [26], stemming from the
homogeneity in solute concentration across pores [51,52]. The predic
occurrence of gusts (~1 Hz [26]) and daily weather (~0.1–1 mHz). For
tion of such was accomplished in the long-time limit using the method of
the sake of comparison, a typical aqueous all-vanadium RFB (diffusivity
moments [53–55], where an adjoint eigenvalue problem coupled to a
D ∼ 10− 6 to 10− 7 cm2 /s [28–31]) with carbon-felt electrode (fiber
closure problem was solved numerically to obtain the mean solute ve
diameter b ∼ 10 to 20 μm [32,33]), has a characteristic frequency fc =
locity, reactivity coefficient, and dispersion tensor in a self-consistent
D/b2 ∼ 0.025 to 1 Hz. Further, the low diffusivity of active species in
manner [55]. In contrast, the approach that we use in this article is to
non-aqueous electrolytes (e.g. D = 10− 10 cm2 /s in deep eutectic solvent characterize the integral covariance between the pore-scale velocity
[34]) can decrease fc by orders of magnitude. Overlap between the field and the pore-scale concentration field obtained from solutions to
ranges of fc with the PSD distribution of wind, therefore, suggests that the Fourier-transformed mass conservation equations (MCEs), inspired
the frequency-dependent response of RFBs is likely to play an important by rigorous theory of pore-scale averaging [56]. We ultimately express
role in the efficient utilization of wind energy. this covariance using a non-dimensional average advective-flux transfer
Despite its potential importance for grid-scale energy storage with ⃒
function 〈N〉 ≡ (1 /|us ⃒)(us 〈c〉 − ε〈uc〉)/(cs − 〈c〉) that quantifies the de
RFBs, limited theoretical understanding of solution-phase frequency
viation of average advective flux ε〈uc〉 from the product of average
response has been established at the pore-scale in modeling and exper
concentration and superficial velocity us 〈c〉. In all instances bolded
imentation. Beyond the film-law of convection mass transfer, time-
symbols are vector quantities.
domain pore-scale simulations have been conducted using the lattice
In this article we present new theory for the transient response of
Boltzmann method [35,36], but rigorous up-scaling strategies have yet
redox-active electrolyte in porous electrodes using frequency-dependent
to be developed to enable the simulation of transient transport phe
transfer functions (TFs) that we show are readily up-scalable. These TFs
nomena at scales larger than representative volume elements. In
link the Fourier transform of output parameters (reaction rates and
contrast, a recent approach used Laplace transformation to analyze the
advective flux deviation) to that of input parameters (surface concen
transient redox and diffusion of vanadium-based active-species in stag
tration and volume-averaged concentration). We use these TFs to cap
nant electrolytes at carbon-felt electrodes during cyclic voltammetry
ture the microscopic response of RFB electrodes in the frequency domain
[37]. We also recently introduced theory [38] to capture the effect of
under varying current input at the reactive interface. Later these
charge/discharge at the pore-scale for active species undergoing
frequency-dependent TFs are embedded into an up-scaled model (using
simultaneous diffusion and advection in the pseudo-steady limit, which is
porous electrode theory) to obtain the macroscopic response of RFBs. A
approached when the characteristic time period for a charge/discharge
schematic representation of the approach is shown in Fig. 1. In Sec. 2 the
cycle τc/d is large relative to the characteristic diffusion time-scale (i.e.,
definitions of these TFs are derived by Fourier transformation of the
τc/d ≫b2 /D). Previous work on stationary batteries has used a corresponding mass conservation equations (MCEs), which are first
2
defined in the time domain. Then the macro-scale MCEs based on porous yield spectral Sherwood number and advective flux transfer functions.
electrode theory are presented, where pore-scale responses are incor The macro-scale model incorporates concentration polarization and
porated using the TFs derived earlier. In Sec. 3 the developed transient advective flux within electrode pores using state-space approximations
model is compared with the experiments using transient operation of a of the aforementioned pore-scale transfer functions under time-varying
ferro/ferri-cyanide based flow cell with microchannels engraved in applied current. Though the present macro-scale model neglects certain
graphite foil. Finally, the obtained pore-scale and macro-scale MCEs are effects (superficial-velocity gradients, ion migration, and ion disper
solved numerically for different porosity (ε) and Péclet number (Pe) sion), the means by which we incorporate pore-scale transfer functions
values using electrodes comprised of cylinders in crossflow. The results into it serves as a template for their incorporation into macro-scale
obtained from pore-scale simulations are presented in Sec. 4. In Sec. 5 models using such effects, as in Ref. [24]. Derivations used to obtain
numerically obtained macroscopic responses of RFBs under step impulse the above models and the numerical methods used to solve the gov
variation of input current are presented for two different porosity elec erning equations at both scales are in the supplementary information.
trodes. For interested readers derivations for the conditions of validity of
this theory, a proof of the equivalence between the limit of vanishing 2.1. Governing equations for pore-scale transport
frequency and the pseudo-steady limit, numerical model implementa
tion, and model verification are presented in the supplementary In this section we present the governing equations for electrolyte
information. containing a soluble redox couple undergoing redox at electrode sur
faces. The mass conservation equations (MCEs) and associated boundary
2. Methods conditions (BCs) governing the evolution of the pore-scale concentration
field within electrode pores are presented in the time domain and are
Here we present theory for the modeling of pore-scale transfer transformed into the frequency domain to derive transfer functions (TFs)
functions in the frequency domain and for their incorporation into a for reactive advective flux. We note that similar and analogous equa
macro-scale model of RFB electrodes in the time domain. We do this by tions govern the transport of other electrochemical, chemical, and
defining the pore-scale problem in the time domain and by subsequently thermal systems, but we focus here on the specific application of the
converting it to the frequency domain using the Fourier transform to present theory to RFBs.
Fig. 1. (a) Illustration of a redox flow battery with flow-through porous electrodes containing electrolyte with close-up view at the micro-scale pores where redox-
active species R and O are dissolved. Transient current with arbitrary time-dependence is mapped to the frequency domain using Fourier transformation to obtain a
spectral density for current, and frequency-dependent transfer functions (TFs) are used to capture the microscopic response of the RFB in the frequency domain,
which are later embedded into a macroscopic model. (b) Macroscopic model formulation using porous electrode theory, where the dynamic response is captured
through variation of polarization magnitude |Δc| and average reactant concentration 〈creac 〉 with time for a step impulse input current. (c) Schematic of transient RFB
operation based on the dynamic response of |Δc| and 〈c〉 inside a porous RFB electrode.
3
2.1.1. Time-domain governing equations frequency and the latter of which 〈N〉 captures the integral covariance
For electrochemical systems that use heterogenous electron transfer between pore-scale velocity and the Fourier-transformed concentration
reactions involving a soluble redox couple (RzR ⇌OzO + e− ), as in RFBs, fields.
the consumption of reduced species R is coupled with the production of
oxidized species O. We neglect the electromigration contributions to the 2.1.2.1. Fourier transformed governing equations. Fourier transformation
net flux provided that the transference number of any participating of the MCEs and the BCs expressed in Eq. (1) results in a steady, linear
soluble species is sufficiently small [38]. Under such condition we show PDE that governs the Fourier transformed species concentration c while
subsequently that the theory that we presented previously [38] can be the solid/solution interface is subjected to a time-varying sinusoidal
extended to incorporate arbitrary time-variation of the surface concen surface concentration with angular frequency ω = 2πf in rad/s, where f
tration ci,s (t) of species i when the associated Damköhler numbers is regular frequency in Hz. In general, Fourier transformed functions are
(Dared/ox = kred/ox b/D) for reduction and oxidation reactions are suffi ∫+∞
ciently large: Dared ≫min[1, exp(− Fη /RT)] and Daox ≫min[1,exp(Fη /RT)], denoted here by an overbar (i.e., F {A(t)} = A(ω) = A(t)exp( −
where η is a characteristic overpotential and RT/F is the thermal voltage. − ∞
The derivation of these criteria using scaling analysis is presented in Sec. jωt)dt, where A(t) is an arbitrary function of time). The Fourier trans
A of the supplementary information. Under such conditions the con formed version of Eq. (1) is given by:
servation of mass for species i is written as:
∇⋅(u∎ ci − Di ∇ci ) + jωci = 0 ci |s = ci,s (ω) (2)
∂ci
+ ∇⋅(u∎ ci − Di ∇ci ) = 0 ci |s = ci,s (t) (1a) √̅̅̅̅̅̅̅
∂t where j = − 1 is the imaginary unit. Frequency response of the corre
sponding system’s concentration field is captured by a transfer function
Here, ci is the concentration of soluble redox-active species i,1 which
Ci (ω, r) defined with ci (ω, r) as its output quantity and with ci,s (ω) as its
depends on position r and time t, Di is its corresponding Fickian diffusion
input quantity:
coefficient, and u∎ is the molar-volume averaged fluid-velocity vector.2
Further, we assume that the characteristic streamwise concentration Ci (ω, r) ≡
ci (ω, r)
(3)
difference is negligible relative to the characteristic pore-scale concen ci,s (ω)
tration difference through the microstructure of porous electrode, as in
Substituting ci (ω, r) = ci,s (ω)Ci (ω, r) into Eq. (2) results in a time-
our previous work [38]. Such an assumption enables us to use periodic
independent PDE for Ci with a frequency-independent interfacial
BCs at the boundaries of the repeat units of periodic microstructures:
condition:
ci (r, t) = ci (r + R, t) (1b)
∇⋅(u∎ ℂi − Di ∇ℂi ) + jωℂi = 0 ℂ i |s = 1 (4)
where R is any integer-multiple linear combination of lattice vectors For DO = DR inspection reveals that the dynamic responses of R and
defining the microstructure’s periodicity. By analyzing the time- O are identical (CR = CO ). To establish a necessary condition to ensure
dependent dynamics of RFBs cycled under galvanostatic conditions we the coupled production and consumption of R and O we invoke reaction
showed previously that the validity of such a condition depends on the stoichiometry:
respective magnitudes of the Sherwood number 〈Sh〉, Péclet number Pe, ⃒ ⃒ ⃒ ⃒
and Fourier number Fo = Dt/b2 [38]. For the limit of vanishing Fo this ∂cR ⃒ ∂cO ⃒⃒ ∂ℂR ⃒⃒ ∂ℂO ⃒⃒
DR ⃒⃒ = − DO ⇒ D c = − D c (5)
condition requires 〈Sh〉≪(Lr /b)/Fo, which allows the use of the periodic ∂n s ∂ n ⃒s R R,s
∂n ⃒s O O,s
∂n ⃒s
BC expressed in Eq. (1b) for high Pe flows where the associated char From this expression we deduce that cR,s = − cO,s = ci,s enforces the
acteristic frequency of the transient current is large. Under such condi conjugation condition exactly when DR = DO is satisfied. We note that
tions solute transport is controlled by a fast-changing surface changing the sign of ci,s does not change the associated PDEs and BCs
concentration, and the effect of streamwise concentration differences is (Eq. (4)) that govern Ci (ω, r). Thus, identical PDEs for O and R auto
negligible in the entrance region where the local concentration field matically enforce the coupled production/consumption of active species
develops. In this section, we invoke periodic boundary conditions in an when posed in terms of Ci (Eq. (4)).
ad hoc manner motivated by the arbitrariness of the surface condition We now elaborate on the physical significance of the transfer func
imposed. However, after defining the spectral Sherwood number and the tion Ci (ω, r). Since the interface condition that is controlled by ci,s cor
nondimensional advective flux transfer function, we derive a new cri responds to a sinusoidal, periodic disturbance signal in the time domain,
terion for this assumption to be valid using scaling analysis in the fre Eq. (4) governs the propagation of this disturbance throughout the
quency domain (see Sec. B of the supplementary information). electrolyte solution. The transient response at a location r is captured by
the magnitude and phase of Ci that respectively represent the fractional
2.1.2. Frequency-domain governing equations and pore-scale transfer concentration change and phase lag with respect to the change made in
functions surface concentration ci,s . For example, at an interface the conditions
Using the time-domain governing equations from Sec. 2.1.1, we |Ci | = 1 and θCi = 0 imply that the concentration there tracks the
show that certain transfer functions (TFs) can be used to succinctly imposed cs (t) boundary condition instantly. On the contrary Ci = 0 at
capture the transient response of output parameters to both the surface- some arbitrary location suggests that the interfacial disturbance induced
and volume-averaged solute concentrations. To do so, the MCEs that by time-varying ci,s is not experienced there. Hereafter, we drop the
govern concentration fields are reduced to a set of linear partial differ subscript i from the various parameters of interest because the various
ential equations (PDEs) in the frequency domain using the continuous- species of interest exhibit identical governing equations for Ci .
time Fourier transform. Using this approach, we define TFs for the
average surface flux and the average advective flux, the former of which
2.1.2.2. Spectral sherwood number. The steady response of solution un
we show is a kind of spectral Sherwood number 〈Sh〉 ̃ that depends on
dergoing convective mass transfer with simultaneous reactions in
porous media has historically been captured using an overall Sherwood
number 〈Sh〉 = 〈hm 〉b/D, with 〈hm 〉 being the overall mass transfer co
1
Throughout this article i as a subscript indicates either species R or O, while efficient that relates the average surface flux 〈j〉s = D〈∂c /∂n|s 〉 due to
i otherwise indicates current density.
2
We model the bulk electrolyte as isochoric, such that the pore-scale molar-
volume averaged fluid-velocity u∎ is divergence free (i.e., ∇⋅ u∎ = 0) [38].
4
reactions to a driving concentration difference Δc, according to the film assumptions that are needed to decompose 〈u∎ c〉 into an arbitrary sum of
law of mass transfer: 〈j〉s = 〈hm 〉Δc.3 The operators 〈…〉 and 〈…〉s are products between average and deviatoric u∎ and c fields:
respectively integral averages over solution volume and solid/solution ∫ ∫ ∫
interfaces contained in a representative volume element. To our 〈W〉 ≡ − covar(u∎ , c) ≡ u∎ dvf c dvf − u∎ c dvf = (us / ε)〈c〉 − 〈u∎ c〉
knowledge, however, similar theory has yet to be developed that cap vf vf vf
tures the transient response of solutions undergoing convective mass (7)

transfer in porous media. It is our goal presently to predict average re
action rates vis-à-vis average surface flux subjected to an arbitrary time- Here, vf is the interstitial fluid volume occupied by electrolyte. Further,
varying surface concentration to enable up-scaling in volume-averaged/ once 〈W〉 is known we posit that the average advective-flux can be used
macro-homogeneous formulations (cf. Ref. [56]) in a manner that is directly as a kind of constitutive equation within local volume-averaged or
consistent with transient pore-scale transport phenomena. macro-homogeneous formulations by recognizing that 〈u∎ c〉 = uεs 〈c〉 −
To do this we derive a transfer function in the frequency domain that 〈W〉, which is an alternative to the common assumption of 〈u∎ c〉 = uεs 〈c〉
extends the concept of the Sherwood number to transient conditions. that is valid only when covar(u∎ , c) = 0.
This particular Sherwood number, which we call a spectral Sherwood The utility of 〈W〉 in volume-averaged formulations motivates pre
number 〈Sh〉,
̃ is a transfer function in the frequency domain that captures dicting its response to transient impulses. In the spirit of modeling such
the correlation between average surface flux 〈j〉s and the characteristic effects subject to arbitrary variations of interfacial conditions, we define
concentration difference Δc = cs − 〈c〉: a corresponding advective-flux transfer function 〈N〉 in terms of 〈W〉 and
Δc = cs − 〈c〉.
b 〈j〉s − 〈∂C/∂n* 〉 〈G〉
〈̃
Sh〉(ω) ≡ = = (6) ε 〈W〉
D cs − 〈c〉 1 − 〈C〉 1 − 〈C〉 〈N〉 ≡ ⃒ (8)
|us ⃒ cs − 〈c〉
Here, the nondimensional normal coordinate is n* = n/ b, and G(ω, r) is
a gradient transfer function defined as G(ω, r) ≡ (b /D)(js /cs ) = − Analysis of Eq. (7) shows that 〈W〉 scales in proportion to the charac
teristic interstitial velocity us /ε. Thus, we normalize 〈N〉 by |us |/ε, and
〈∂C /∂n* 〉 with input as surface concentration cs , rather than Δc = cs −
we decompose it into streamwise (〈N‖ 〉 = 〈N〉⋅̂e ‖ ) and transverse
〈c〉.4 With 〈Sh〉
̃ known from pore-scale simulation, the average surface
(〈N⊥ 〉 = 〈N〉⋅̂e ⊥ ) components, where ̂e ‖ and ̂e ⊥ are streamwise and
flux 〈j〉s can be determined in the time domain as
transverse unit vectors with respect to us :
− 1
〈j〉s = (D/b)F {〈Sh〉(cs − 〈c〉)}, according to Eq. (6).
̃
( )
For the sake of consistency with the definition of the Sherwood 〈W ‖ 〉 〈ℂ〉 ε 〈u∎‖ ℂ〉
〈N‖ 〉(ω) ≡ = − ⃒ (9a)
number obtained in the pseudo-steady limit [38], we confirm presently |us |
[cs − 〈c〉]
1 − 〈ℂ〉 |us ⃒ 1 − 〈ℂ〉
ε
that the spectral Sherwood number 〈Sh〉
̃ obtained in the limit of van
( )
ishing frequency is identical to the conventional 〈Sh〉 in the 〈W ⊥ 〉 ε 〈u∎⊥ ℂ〉
〈N⊥ 〉(ω) ≡ =− ⃒ (9b)
pseudo-steady limit: lim 〈Sh〉
̃ = 〈Sh〉 . A detailed proof demonstrating
PSL
|us |
[cs − 〈c〉] |us ⃒ 1 − 〈ℂ〉
ω→0 ε
this property of 〈Sh〉

̃ is shown in Sec. C of the supplementary informa
We note that 〈N‖ 〉 and 〈N⊥ 〉 are respectively expressed in terms of
tion. This result thus provides intuition for the broader meaning of the pore-scale velocity components that are parallel u∎‖ and transverse u∎⊥ to
pseudo-steady limit in the context of frequency response. As we later
the superficial velocity vector us . While Eq. (9) alone is sufficient to
demonstrate using numerical modeling, the limit of vanishing frequency
capture the vectorial nature of 〈N〉 in two-dimensional systems, in
is equivalent to the pseudo-steady limit because concentrations and
practice two distinct 〈N⊥ 〉 components would be needed along orthog
fluxes exhibit lagless tracking (i.e., vanishing phase difference) with
onal directions to capture such effects for three-dimensional systems.
time variations of surface concentration in that limit.
With 〈Nj 〉 known along each direction j the corresponding component of
average advective flux 〈u∎ c〉⋅̂ e j can be determined in the time domain as
2.1.2.3. Advective-flux transfer function. We now define a transfer [( ) { }]
e j 〈c〉 − |us |F − 1 〈Nj 〉(cs − 〈c〉) .
〈u∎ c〉⋅̂e j = (1/ε) us ⋅̂
function that quantifies the transient response of average advective flux
to the pore-scale concentration difference Δc = cs − 〈c〉. Previously
derived volume-averaging theory for pore-scale transport suggests that 2.2. Macro-scale time-domain governing equations
the local volume-averaged advective flux 〈u∎ c〉 should be decomposed
into terms that account for the average contribution to velocity and Here we present governing equations for the macro-scale transport of
concentration, as well as those that result from the deviation of those redox-active species using porous electrode theory. In such a formula
parameters from their averages [56,57]. In contrast, we characterize tion the porous electrode is considered as a superposition of two con
〈u∎ c〉 presently by defining the integral covariance vector 〈W〉 for ve tinua, one representing the liquid phase and the other representing the
locity u∎ and concentration c, thus avoiding the introduction of ad hoc solid phase, where spatially varying quantities are included using their
volume averaged values. While a dearth of past work using porous
electrode theory has been done, we refer interested readers to the
seminal work where its basic postulates were first presented [58]. We
3
Established theory and experiment [73–75] have previously shown that 〈 first express a macroscopic mass balance equation by embedding the TFs
Sh〉 principally depends on Péclet number Pe = |us |b/D, where us is superficial 〈Sh〉
̃ and 〈N〉 to account for their time-dependent pro
velocity, and on the Schmidt number Sc = ν/D, where ν is kinematic viscosity. duction/consumption rates and enhanced/retarded advection rates due
In porous materials undergoing forced convective heat transfer, the Nusselt
to their transient redox at interfaces. As both TFs are obtained in the
number 〈Nu〉 = 〈h〉b/k, with overall heat transfer coefficient 〈h〉 and thermal
frequency domain, their inverse Fourier transforms are taken and
conductivity k, is equivalent to 〈Sh〉 by virtue of the formal analogy between the
implemented in a consistent manner. Further, we neglect the
equations governing heat and mass transfer when the Prandtl number Pr = ν/ α
with thermal diffusivity α and the thermal Péclet number Pet = |us |d/ α are macro-scale diffusion of reactive species by considering negligible con
respectively used in place of Sc and Pe. centration variation across the reactor lengths parallel and perpendic
4
We note that cs /(∂c/∂n)|s has previously been referred to as the diffusion ular to the macroscopic flow direction. In the limit of vanishing
impedance under potential modulation [41,76], and it can be obtained from the macro-scale diffusion, the up-scaled mass conservation for species i
reciprocal of the gradient transfer function G as G(ω, r) ≡ − 1 /b[(∂c/∂n)|s /cs ]. can be expressed in terms of the macroscopic volume averaged
5
concentration 〈ci 〉 as: which arise from the statistical distribution of electrode cross-section
area and the uncertainty in active-species diffusivity. For such short
∂〈c 〉 { ( )} aD { ( )}
ε i +∇⋅us 〈ci 〉− |us |∇⋅F − 1 〈N〉 cs,i − 〈c〉i = i F − 1
〈̃
Sh〉 cs,i − 〈c〉i pulses diffusion is inherently transient with concentration variations
∂t si b
penetrating from interfaces into bulk to an extent that is much shorter
(10)
than the characteristic pore dimension (~100 μm), thus validating the
Here, 〈ci 〉(r, t) is local volume-averaged concentration obtained by transient aspects of the present B-UTM model. For large impulse dura
averaging ci over a finite volume larger than a periodic unit representing tion (τ ≥ 10) where solute transport occurs through the entirety of
the porous microstructure. The source term at the right-hand side of Eq. pores, we observe that both the CM and B-UTM models produce polar
(10) accounts for the transient volumetric production/consumption rate ization variation with time that deviates qualitatively from experiment,
of species that is coupled to the transient current density 〈i〉s as likely as a result of factors that both models neglect including entrance
region effects and macroscopic diffusion, dispersion, and ion conduc
a〈i〉 /(si F) = (a)〈ji 〉 = aD /(si b)F − 1 {〈Sh〉(c
̃ s,i − 〈c〉 )}. Further, the
s s i tion. Despite this, we note that B-UTM shows less deviation from
term |us |∇⋅F − 1 {〈N〉(cs,i − 〈c〉i )} accounts for the deviation (accelera experiment than CM. We attribute this outcome for B-UTM to its ac
tion/suppression) in reactive species advection rate through the porous counting of the suppression/acceleration arising from the in
electrode. We modeled the reaction at the solid/solution interface by homogeneity of pore scale concentration and velocity fields, validating
using Butler-Volmer (B–V) kinetics (Eq. (11)), which describes the its inclusion in the present model.
current-overpotential relationship considering symmetric reaction ki Modeled and experimental time-averaged polarization for the
netics for both oxidation and reduction: various conditions tested are shown in Fig. 2. The experimental data
[ (
Fη
) (
Fη
)] shows that average polarization depends not only on the applied peak
i = i0 exp − exp − (11) current ratio irp , but also on the pulse duration τ. The CM model fails to
2RT 2RT
capture the dependence of average polarization on impulse duration τ.
Here, i is the current density at the electrode surface, i0 is the exchange Further, this inability to capture transient response causes CM predic
current density, and RT/F is the thermal voltage. The reaction over tion to deviate substantially at high current ratio where it over-predicts
potential η is expressed using the solid phase potential φs , solution the polarization up to 400%, and it prevents such models from simu
(electrolyte) phase potential φe , and the equilibrium potential φeq as: lating RFBs with near-limiting (irp ∼ 1) and over-limiting (irp > 1) con
( ) ( ) ditions that have been observed in previous studies [63–66]. On the
η = φs − φe − φeq = φs − φe −
RT
ln
cO,s
(12) other hand, the present B-UTM model is not only capable of simulating
F cR,s the near-limiting and over-limiting operations, but it also produces more
Further, the mass conservation equation for a well-mixed tank accurate predictions at all irp and τ values tested.
without any consumption/production of species is expressed by Eq.
4. Pore-scale simulations versus frequency
(13), where 〈ci 〉T is the volume averaged concentration of species i inside
the tank.
Using the frequency-based theory introduced in Sec. 2.1, we ob
∂〈ci 〉T tained results for numerical simulations performed by applying that
+ ∇⋅us 〈ci 〉T = 0 (13)
∂t theory to study the transient response of redox-active electrolyte flowing
through a porous electrode comprised of infinite cylinders undergoing
The assumption well-mixed refers to the fact that the local concen
crossflow as a representative geometry for carbon fiber felt commonly
tration field inside the tank is uniform.
used in RFBs. Here, pore-scale simulation results are presented to show
the characteristics of the frequency-dependent TFs. These simulations
3. Experimental validation of bottom-up transient model
are used to investigate the effect of microstructure porosity, superficial
velocity, and frequency, where different cases are created by varying
In this section we compare model predictions based on Section 2 with
their corresponding non-dimensional parameters to make the results
experimental results obtained from the transient response of an elec
extensible to other conditions, provided that geometric, hydrodynamic,
trolyte solution containing the Fe(CN)4− 3−
6 / Fe(CN)6 couple at 50% state- and mass-transfer similarity criteria are satisfied. The non-dimensional
of-charge that undergoes reversible redox via Fe(CN)3−
6 + frequency f * ≡ ωb2 /2π D is the ratio of the characteristic time-scale for
e− ⇋ Fe(CN)4−
6 . This redox couple was chosen for its facile reaction ki diffusion (τd = b2 /D) to the time-scale associated with the angular fre
netics (rate constant of ks = 0.1 ∼ 1 cm/s [59–62]), so as to produce quency ω (τc/d = 2π/ω). For f * ≪1 we find that pore-scale mass transfer
negligible kinetic overpotential in the present experiments. The exper is consistent with that of the pseudo-steady limit, while higher fre
imental reactor consisted of two graphite foil electrodes (Ceramaterials) quencies produce a finite temporal lag in response that we show is a
laser-engraved (Trotec Speedy Flexx 400 fiber laser) with ~100 μm wide signature of transient response. For the present simulations we choose
parallel microchannels that were separated by a non-selective separator the characteristic length b as the cylinder diameter d. Subsequently, we
(NP030 from STERLITECH Corporation). Electrolyte containing 15 mM present results for the local-concentration transfer function C(ω, r), the
active species and 3 M KCl supporting salt was pumped through the
overall spectral Sherwood number 〈Sh〉(̃ ω), and the parallel/transverse
microchannels in the electrodes. To probe the electrodes’ transient
components of the average advective-flux transfer function 〈N‖/⊥ 〉(ω).
response current was applied using a square wave with half-period or
pulse-width τ, as shown in Fig. D2b. Experimental methods and model We solved the transformed MCE (Eq. (4)) numerically in the fre
implementation are provided in the supplementary information. quency domain to obtain C(ω, r) for a two-dimensional porous micro
Figure D3 in Sec. D of the supplementary information shows the time structure. The porous electrodes investigated here are regular arrays of
variation of IR-corrected cell voltage with non-dimensional time. In unconsolidated cylinders of infinite length and diameter d (Fig. 3a).
contrast with our experimental results, the conventional model (CM) – Simulations were performed over a domain defined by the unit cell
that uses a time-invariant film law of mass transfer and that neglects shown in Fig. 3b–c. We present simulated results for the transient
solution advection suppression/acceleration – shows instantaneous response of solution within microstructures having ε = 0.546 and ε =
jumps in polarization when applied current is switched with a suffi 0.874 as representative cases in the limits of low and high porosity
ciently small pulse width (τ < 10). On the contrary the present bottom- respectively. Inline flow with superficial velocity us parallel to the x-axis
up transient model (B-UTM) captures the dynamic response in good was simulated with Pe ranging from 10− 2 to 22.8 × 103 .
agreement with experiment for such pulse widths, the deviations of The electrolyte solution flowing through the microstructure’s pores
6
Fig. 2. Variation of average concentration polarization with applied current ratio irp at different pulse width τ. (a) τ/tr = 0.0057, (b) τ/tr = 0.0128 and (c) τ/ tr =
0.0494. tr is the mean residence time of solute (tr = Lr /us ).
Fig. 3. Schematic of the modeled periodic domain: (a) regular, unconsolidated array of infinite circular cylinders with bulk electrolyte flow through interstitial pores,
(b) designation of a repetitive unit cell with a circular cylinder at the cell’s center, and (c) enlarged view of a repetitive unit cell annotated with dimensions. A given
microstructure’s porosity ε results in a certain minimum pore dimension bt at the pore throat shown in (c). Spatial distribution of |C| at different nondimensional
frequencies for (d) ε = 0.546 with Pe = 0.019 (first row) and Pe = 380 (second row), and (e) ε = 0.874 with Pe = 0.01 (first row) and Pe = 300 (second row). The
bulk flow direction is horizontal from left to right. The values at the center of each distribution show the corresponding nondimensional frequency f * .
is assumed as a Newtonian, isochoric solution with constant dynamic numerical implementation, and model verification are shown in Sec. E
viscosity, while the flow field is modeled as steady, creeping flow, as of the supplementary information.
justified by a small pore-scale Reynolds number: Re = Pe/Sc≪ 1, where
Sc is Schmidt number. The pore-scale velocity field u∎ is obtained from
4.1. Concentration
numerical solution of a time-independent, inertia-free version of the
vorticity transport equation (VTE) that is written in terms of the stream
function ψ as ∇4 ψ = 0 [67]. The VTE is solved with appropriate BCs Figure 3d–e shows the spatial distribution of |C| =
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
imposed on cylinder surfaces (i.e., no-slip, impermeability, and no-lift)
(Re[C])2 + (Im[C])2 as a function of non-dimensional frequency for
and at the unit-cell boundaries (i.e., periodic jump/fall). For a finite
high and low porosity values with either high or low Péclet number. At
frequency (ω > 0) the corresponding C(ω, r) field is obtained by solving
Eq. (4) with periodic BCs applied at unit-cell boundaries, along with a sufficiently high frequency (f * ≫200) these distributions reveal that
Dirichlet BC Cs = 1 at cylinder surfaces. Details of discretization, most solution volume remains unaffected by the disturbance induced by
interfacial reactions for all Pe and ε studied and that variations of |C| are
7
focused near solid/solution interfaces. This effect is caused by the concentration fluctuation starts at the interface.
comparatively slow rate of diffusion relative to the frequency at which When the characteristic diffusion time-scale is comparable with the
the surface concentration is cycled in such a limit. As a result, the time-scale for cycling at the solid/solution interface (f * = 1 ∼ 200), |C|
disturbance occurring at the solid/solution interface is unable to prop varies in space in different ways depending on what the predominant
agate substantially into bulk solution before a new cycle of mass-transfer mode is. For Pe≪1, the spatial distribution of |C| is sym
Fig. 4. Variation of transfer functions (a) 〈Sh〉

̃ and (b) 〈N‖ 〉 with frequency for ε = 0.546 (left panel) and ε = 0.874 (right panel). The upper and lower row show
magnitude and phase of the respective TFs.
8
metric with respect to the x and y axes, and |C| approaches an axisym 0.874.
metric distribution for high-porosity electrodes due to the weak in In between Regions I and III, we observe Pe-specific transitions from
teractions between adjacent cylinders. For low-porosity electrodes a low to high frequency (Region II) as a result of the characteristic diffu
four-peaked distribution is observed as a result of strong inter-cylinder sion time-scale being comparable to the time period of boundary
interactions that are evidenced by peak positions coinciding with the cycling. Within this region the magnitude and the phase of 〈Sh〉̃ show
surface positions at which neighboring cylinders have closest proximity. frequency dependence that is distinguishable for different extremes of
For high Pe cases the spatial distribution of |C| becomes skewed along Péclet number. For systems with low Péclet number (Pe≪1) a smooth,
the streamwise direction as a result of the predominant role of advection monotonic variation of phase is observed with frequency as a result of its
mass transfer in that limit. In the limit of vanishing frequency (f * ≪ 1) transition from diffusion-controlled transport at low frequency to
the distribution of |C| approaches a uniform value of unity (i.e., |C| = 1) frequency-controlled transport at high frequency. On the other hand, for
as a result of the lagless tracking of surface concentration by the entire systems with high Péclet number (Pe≫1) the variation of phase with
concentration field in that limit. However, we note that such an attribute frequency is non-monotonic because of its transition from advection-
does not imply an infinite mass transfer coefficient, as evidenced by our controlled transport to frequency-controlled transport.
subsequent investigation of the spectral Sherwood number.
In the pseudo-steady limit our previous simulations revealed that the
effect of advective mass transport is evidenced by the shape of the 4.3. Advective-flux transfer function
concentration profile at the midplane of the cylinder and along the di
rection transverse to the streamwise direction [38]. To determine how Solute can either deplete or accumulate adjacent to such interfaces
frequency influences such profiles the non-dimensional concentration due to the finite reaction rates occurring at solid/solution interfaces,
profiles are shown in Sec. F of the supplementary information, and they thus reducing the effectiveness of advective transport near such in
suggest that response in the low-frequency limit is controlled primarily terfaces because of the no-slip condition. To characterize such effects, in
by the dominant transport mechanism in the solution’s bulk (pseudo-s Sec. 2.1.2.3 we introduced a non-dimensional advective-flux transfer
⃒
teady diffusion and/or advection), while response in the high-frequency function 〈N〉 ≡ − covar(u∎ , c)ε/[|us ⃒(cs − 〈c〉)] that quantifies the depar
limit is controlled by transient diffusion through boundary layers of so ture covar(u∎ , c) of average advective-flux 〈u∎ c〉 in the frequency domain
lution that exhibit frequency-specific thickness adjacent to interfaces. from its value 〈u∎ 〉〈c〉 that is expected for homogeneous concentration
and velocity fields. When a microstructure comprised of a square array
4.2. Spectral sherwood number of cylinders is subjected to horizontal flow along the x-axis, velocity and
concentration fields are produced that show reflection symmetry about a
line directed in the streamwise direction that also passes through the
We now present results for the overall spectral Sherwood number 〈
cylinder’s centroid. For such cases to which we restrict the present
Sh〉
̃ ≡ b〈j〉 /[D(cs − 〈c〉)], which relates the average surface flux due to
s investigation, the transverse component of 〈N〉 is identically zero by
reactions to the concentration difference that drives it in the frequency virtue of this symmetry. Thus, in Fig. 4b we present simulated results for
domain. Fig. 4 shows the variation with frequency of 〈Sh〉 ̃ magnitude the variation with frequency of the streamwise component of 〈N〉 at
and phase for different Pe and porosity values. The variation of 〈Sh〉 ̃ different Pe and porosity values, denoted as 〈N‖ 〉.
shows three distinct regions in the frequency domain, as annotated in At high frequencies, |〈N‖ 〉| decreases with increasing frequency by
Fig. 4 At high frequency (Region III) the magnitude of 〈Sh〉 ̃ scales in following a power law correlation with exponent − 0.5, while a negli
√̅̅̅̅
proportion to ω while phase approaches π/4, consistent with the gible effect of Pe is observed because of frequency-controlled transport
scaling expected for a semi-infinite Warburg impedance element and in that limit. The decrease in |〈N‖ 〉| with increasing frequency can be
consistent with our observations from Fig. 3d–e that compact diffusion attributed to the fact that the steep concentration gradients that are
layers form in the vicinity of the solid/solution interfaces in the high- produced at interfaces at high frequency (Figs. F1c and d) promote
frequency limit. In this part of the frequency spectrum surface- supply of reactant as well as the removal of product. In this frequency
region the phase of 〈N‖ 〉 approaches − π /4. At low frequencies the
concentration fluctuations alone control 〈Sh〉,
̃ not those of average
magnitude of 〈N‖ 〉 approaches a constant, non-zero value that depends
concentration. These features result in minimal effect of Pe on 〈Sh〉
̃ at
on porosity, while its phase approaches zero, consistent with the lagless
high frequencies, and as a result we refer to this region as having
response that is a hallmark of the pseudo-steady limit. Similar to our
frequency-controlled transport (FCT). We define a lower bound for this
region as the frequency where phase reaches within 5% of its asymptotic observations for 〈Sh〉,
̃ 〈N‖ 〉 shows smooth, monotonic variation for the
value π/4, corresponding to non-dimensional frequencies of f * = 104 transition from diffusion-controlled transport to frequency-controlled
transport at low Péclet number (Pe≪1), while non-monotonic varia
and 103 for microstructures with respective porosities of ε = 0.546 and
tion is apparent for the transition from advection-controlled transport to
0.874.
frequency-controlled transport at high Péclet number (Pe≫200).
At low frequency (Region I) the magnitude of 〈Sh〉 ̃ approaches a
constant non-zero value as frequency vanishes, while its phase ap
5. Macro-scale simulations versus time
proaches zero simultaneously. Such a limiting phase value shows that
frequency response is lagless in this limit, an effect that results from the
Here we present simulations of the macro-scale response of an RFB
relatively fast mass transfer in comparison with the time-scale of
subjected to step impulse current variation by solving the TF embedded
boundary condition cycling. Hence, in this limit the dominant mode of
macro-scale MCEs from Sec. 2.2 (Eqs. (10)–(13)) in the time domain. We
pore-scale mass transfer, not the driving frequency, determines the
assume that an ideal separator isolates the high- and low-potential sides
magnitude of 〈Sh〉.
̃ In this limit systems with Pe≪1 experience diffusion-
of its reactor to eliminate crossover of reactive species. Thus, we limit
controlled transport (DCT), and systems with Pe≫1 experience our study to one side of the reactor (i.e., a half cell) where a reduced
advection-controlled transport (ACT), consistent with our previous species R is oxidized to produce O due to a positive current applied at
modeling [38] in the pseudo-steady limit. We define the upper bound for solution/electrode interfaces. Assuming a parallel flow configuration
this region of the frequency domain as the lowest frequency that exhibits and assuming negligible macro-scale concentration gradients in the
a value of 〈Sh〉
̃ that is 5% higher than its value in the pseudo-steady transverse direction, a one-dimensional model is sufficient to capture
limit, corresponding to non-dimensional frequencies of f * = 3.4 and the temporal and spatial variation of 〈ci 〉 along the reactor’s length Lr
0.8 for microstructures with respective porosities of ε = 0.546 and (Fig. 5b). We discretized each electrode into 50 finite volumes (FVs)
9
Fig. 5. (a) Schematic of an RFB with electrolyte flowing through both sides of the reactor, (b) one-dimensional model showing the upper half cell with finite volumes,
and (c) step impulse variation of applied current density with time shown as current ratio ir . Variation of maximum local concentration polarization |Δc|lm with time
at different τ /tr for (d) ε = 0.546 and (e) ε = 0.874.
over which the local concentrations are averaged to obtain 〈ci 〉 (see time τth , resulting in a certain Péclet number for low (Pe = 22.8 × 103 )
Fig. 5b). Each tank was represented by a single FV with a time- and high (Pe = 4 × 103 ) porosity electrodes using Pe = Qactual d/Ar D. In
dependent average concentration 〈ci 〉T . The step impulse variation of practice, τth was determined by Faradaic balance from the theoretical
the applied current is presented non-dimensionally by defining a current charge-storage capacity of the redox-active electrolyte and the magni
ratio ir where the applied current density iapp is normalized by the tude of the applied current. A kinetic rate constant that produces a
limiting current density ilim . The limiting current density is calculated as Damköhler number of 414 (Da = kd/D) was used to ensure that reaction
ilim = FD〈cR 〉ini 〈Sh〉/d, where 〈Sh〉 is the time-independent Sherwood polarization was negligible compared to the concentration polarization.
number corresponding to PSL. The step impulse variation is created by For all cases the tank and the reactor were initialized with identical
instantaneously increasing ir to a peak-current ratio irp (from a base electrolyte solution comprised of 90.9% reactant (R) and 9.1% product
current ratio irb = 0.1) for a time τ, after which ir returns to its initial (O) and were charged for 160 s with irb = 0.1 before the step impulse
value. Both τ and irp were varied to simulate different cases of RFB was applied. The time ranges presented were adjusted so that the
operation with high- (ε = 0.874) and low- (ε = 0.546) porosity step-impulse current started at t = 0.
electrodes. We used a self-consistent iterative algorithm to solve Eqs. (10)–(13)
For all cases simulated, the nondimensional flow rate β was at least numerically, where 〈ci 〉 and 〈ci 〉T were calculated from the discretized
16 to prevent reactant depletion at the reactive interface due to insuf mass conservation equations written for each FV inside the reactor (Eq.
ficient electrolyte flow [68]. Here, β is the actual flow rate Qactual (10)) and the tank (Eq. (13)). The inverse Fourier transforms appearing
normalized by the stoichiometric flow rate Qstoic , where Qstoic = in Eq. (10) were calculated using a state-space model for each TF by
(Vtank +Vreactor ) /τth depends on the electrolyte volume in the tank (Vtank ) recognizing that the system is linear time-invariant [39,69]: 〈TF〉 =
and the reactor (Vreactor ) and on the theoretical charging/discharging C(jωI − A)− 1 B + D, where 〈TF〉 is a rational approximation to a given
10
irrational TF and A, B, C, and D are system matrices/vectors. Since 〈Sh〉

̃ is section, advection is the dominant mode of transport because Pe≫1 was
an improper transfer function due to its divergence at infinite frequency chosen. As we subsequently show, the dynamic response of such systems
is dictated by (i) the characteristic time-scale of the transient input and
(see Fig. 4a), we determined a state-space model for its reciprocal 〈Z〉
̃ =
(ii) the average time solute spends inside of the reactor. Hence, different
1/〈Sh〉.
̃ To obtain the poles pi and residues ri for rational approximations
cases of impulse duration are characterized by their corresponding τ/tr
Z〉 and 〈N‖ 〉, a vector-fitting algorithm was used [70–72]. Additional
to 〈̃ values, where tr is the fluid’s mean residence time defined by tr =
⃒ ⃒
details of the model’s discretization, self-consistent iterative algorithm, Lr /⃒u∎s ⃒. Further, we present system response versus t/τ, which repre
state-space models, vector fitting, and verification are included in Sec. G sents time as a fraction of impulse duration. Fig. 5 shows that the CM
of the supplementary information. model always predicts a lagless response, as evidenced by a rectangular
impulse variation of |Δc|lm synchronously with the applied current.
5.1. Concentration polarization across electrode pores Building on these results, we constructed a non-dimensional opera
tion map that shows the maximal concentration polarization over time
We used the macro-scale numerical model to analyze the transient versus pulse duration for different peak current ratios irp (Fig. 6). Three
concentration polarization inside of the RFB reactor. The corresponding different regimes of impulse duration are apparent. First, for τ/tr > 102
time variation of local maximum concentration polarization |Δc|lm (t) = the B-UTM and CM models produce identical results as a result of
⃒ ⃒ pseudo-steady response. Further, pore-scale transport is dominated by
max (⃒ci,s (t) − 〈ci 〉(r, t)⃒) is shown in Fig. 5 for different porosity ε, peak-
bulk advection. This is a consequence of the fluid’s average residence
current ratio irp , and impulse duration τ values. We focus on |Δc|lm
time being much shorter than impulse duration, resulting in each solute
because the reactor is most likely to deplete where and when |Δc| is
molecule experiencing elevated current for less than 1 % of the im
maximum. Numerical solutions were also obtained assuming time-
pulse’s duration. In this regime irp alone dictates concentration polari
independent 〈Sh〉 and negligible solute acceleration/suppression (i.e., 〈
zation, while further increase of τ/tr has neglibible effect. Second, for
N〉 = 0) as a conventional model (CM) against which to benchmark the
τ/tr < 10 the B-UTM varies significantly from CM as a result of
present bottom-up transient model (B-UTM). For all cases studied in this
Fig. 6. Variation of time-maximal concentration

polarization |Δc|tmlm with τ/tr at different peak-
current ratio values irp for (a) low-porosity (ε =
0.546, Pe = 22800) and (b) high-porosity (ε =
0.874; Pe = 4000) electrodes. The corresponding irp
values are overlaid on each curve. CM predictions
for irp < 1 are shown as blue-dashed lines. Cases are
only presented for conditions that did not produce
reactant depletion at any instant and location. (For
interpretation of the references to colour in this
figure legend, the reader is referred to the Web
version of this article.)
11
frequency-controlled transport. This is a consequence of the fluid’s concentration and velocity while CM neglects it. Consequently, the B-
average residence time being similar to impulse duration, resulting in UTM model exhibits abrupt temporal changes of 〈〈ci 〉〉 that track syn
each solute molecule experiencing elevated current for more than 10% chronously with the abrupt change in current that occurs when the
of the impulse’s duration. In this regime concentration polarization is current pulse begins, whereas the CM model produces less abrupt and
dependent on both irp and τ/tr . Of particular note is the modest con less significant temporal changes of 〈〈ci 〉〉. For the B-UTM the sudden
centration polarization that is achievable even for over-limiting condi increase in current increases concentration polarization instantly via
tions (Fig. 6a and b, upper left corners) despite CM predicting inoperable finite 〈Sh〉,
̃ which accelerates the advection rate of reactant, causing a
conditions. Third, for 10 < τ/tr < 102 a transition regime occurs where sudden decrease in 〈〈creac 〉〉 via non-zero 〈N‖ 〉. Similarly, increased
B-UTM predictions near τ/tr ≈ 10 deviate from CM, but gradually merge concentration polarization suppresses the advection rate of product,
with CM for τ /tr →102 . causing 〈〈cprod 〉〉 to increase instantly. In general, we also note that the B-
This operation map also contains a regime of vanishing concentra UTM model retains less reactant and more product compared than CM at
tion polarization (|Δc|tm
lm /〈creac 〉
ini
≈ 0) that corresponds to highly effi a given instant in time as a result of the transient experienced when
cient RFB operation with uniform reaction-rate along the streamwise current is initially turned on (see Sec. H in the supplementary
direction. Such operation is achieved by long duration charging/dis information).
charging with very low current density. Another regime with very high Our earlier results (Figs. 5 and 6) suggest that in the limit of short
concentration polarization (|Δc|tm lm /〈creac 〉
ini
≈ 1) is produced for very pulse duration the acceleration/suppression effect is negligible, while
short duration charging/discharging with high current density that re transient mass transport effects are substantial. The bottom-up transient
sults in non-uniform reaction-rate along the streamwise direction. Under model shows that such properties of the short pulse duration limit
such conditions cycling terminates prematurely during local reactant enable RFB operation with over-limiting current (irp > 1). The opposite
depletion (Fig. 6a and b, gray zones with boundaries marked by cross scenario is observed in the limit of long pulse duration, where the ac
symbols). In this regime the B-UTM model predicts reactant depletion at celeration/suppression effect is significant, and the transient mass
much lower τ/tr compared to the CM model due to the increased transport effect is negligible. To evaluate the relative influence of con
advection rate of the reactant from the reactor, as elucidated centration polarization and reactant availability we show the variation
subsequently. of 〈〈〈creac 〉〉〉t /Δctm
lm in Fig. 8, where 〈〈〈creac 〉〉〉t is the time-averaged
reactant concentration inside the reactor calculated by averaging 〈
〈creac 〉〉 from t/τ = 0 to t/τ = 4. In the limit of short pulse duration, the B-
5.2. Reactant availability through electrode pores
UTM model predicts higher 〈〈〈creac 〉〉〉t /Δctm lm values compared to the CM
Apart from pore-scale concentration polarization between electrode model. The difference between B-UTM and CM in this limit is a direct
surfaces and concentrations in the bulk of electrode pores, the local result of transient mass transport effects that depends on both τ/tr and ir
volume-averaged concentration within electrode pores 〈ci 〉 exhibits where system response is controlled by transient diffusion through the
macro-scale variation along the streamwise direction. Thus, the deple concentration boundary layer adjacent to the reactive interface. On the
tion of 〈ci 〉 can cause premature termination of cycling. To quantify this other hand, in the limit of long pulse duration the B-UTM produces
effect we determined the macro-scale average of solute concentration by consistently lower 〈〈〈creac 〉〉〉t /Δctm
lm than CM as a result of its dominant
∫
averaging 〈ci 〉 over each electrode as 〈〈ci 〉〉 = 1/Vr 〈ci 〉dV, where Vr is acceleration/suppression effect. For a given peak-current ratio this ef
electrode pore volume. While Fig. 7 reveals that short duration impulses fect remains constant in the limit of long pulse duration. Further, system
(τ/tr < 10) produce negligible temporal change in 〈〈ci 〉〉 for all cases of response of the B-UTM in this limit is dictated either by the advection of
irp both for the B-UTM and for CM, disparate responses are observed for solute by bulk electrolyte flow (Pe≫1) or by pseudo-steady diffusion
across pores (Pe≪1). From Fig. 8 it is also evident that high-porosity
the two models for long duration impulses (τ/tr > 102 ). For long im
electrodes exhibit increased transient mass transport effects and less
pulses system response is lagless for both models, but the B-UTM ac
acceleration/suppression effects compared to low-porosity electrodes.
counts for the impact of local reaction rate on the covariance of
Fig. 7. Time variation of volume-averaged concentration of reactant and product 〈〈ci 〉〉 for (a) irp = 0.3 and (b) irp = 0.8 inside an electrode of ε = 0.546 (left panels)
and ε = 0.874 (right panels). The upper and lower half of the broken vertical axis belong to the reactant and product species respectively.
12
Fig. 8. Variation of time-averaged reactant con

centration 〈〈〈creac 〉〉〉t with τ /tr at different peak-
current ratio irp for (a) a low-porosity electrode
(ε = 0.546, Pe = 22800) and (b) a high-porosity
electrode (ε = 0.874; Pe = 4000). The numbers
shown on the curves represent corresponding irp
values. The results from CM with irp < 1 are also
shown by broken blue lines in the figures. (For
interpretation of the references to colour in this
figure legend, the reader is referred to the Web
version of this article.)
Thus, the use of high-porosity electrodes is advantageous for transient without ad hoc introduction of a separate velocity for solute. For solu
operations, along with their low pressure-drop benefits. tions undergoing electrochemical reactions in porous electrodes these
transfer functions provide fundamental insights into the mechanisms
6. Conclusions that could limit achievable current densities, thermodynamic effi
ciencies, device resilience, and cycle life. Beyond their potential signif
A bottom-up multi-scale modeling approach is presented to study the icance to electrochemical systems, these concepts will also find use in
transient response of redox flow batteries with time-dependent current understanding the coupling between reactions and convection processes
input. The transient pore-scale mass transport physics, including con in porous media, as well as in understanding transient thermal transport
centration polarization and the resulting deviation of advection rate of via formal analogies between heat and mass transfer.
reactive species, are captured via frequency dependent transfer func Both qualitative and quantitative comparison between experiment
tions (TFs), which are incorporated into an up-scaled model through and bottom-up transient model (B-UTM) predictions are presented in
volume averaging. These transfer functions include the spectral Sher contrast with the conventional constant Sherwood number model,
wood number 〈Sh〉̃ that enables determination of the average flux on where the B-UTM model not only showed substantial improvement in
reactive surfaces 〈j〉s subject to a driving concentration difference at the capturing transient response, but also provides a means to simulate flow
pore-scale Δc = cs − 〈c〉, as well as an advective flux transfer function 〈 cell operations with short impulses of near-limiting and over-limiting
N〉 that captures the deviation of the total advective flux vector 〈uc〉 from applied current.
its value expected in the absence of covariance between u and c (i.e., Aside from their future use, we demonstrated the utility of our
〈u〉〈c〉), subject to the same driving force Δc. frequency-based theory by simulating such transfer functions using a
The spectral Sherwood number is shown for the first time to extend simplified microstructure containing a regular array of solid cylinders in
the conventional concept of a Sherwood number to transient conditions. horizontal crossflow. The effects of porous electrode microstructure,
Also, the advective flux transfer function describes the relative accel superficial velocity, and boundary cycling frequency were respectively
eration/suppression of solute advection rates relative to bulk flows, investigated by varying porosity ε, Péclet number Pe, and nondimen
13

sional frequency f * . From the variations of 〈Sh〉

̃ and 〈N‖ 〉 with frequency provided by National Science Foundation.
at different ε and Pe we deduced three different spectral regions of
Data availability
response. In the low-frequency region (f * ≪1) transport is primarily
controlled by the prevailing pseudo-steady mass-transfer mode: diffusion
Data will be made available on request.
and advection respectively for Pe≪1 and Pe≫1. In this spectral region,
the magnitude of 〈Sh〉
̃ and 〈N‖ 〉 approach an asymptotic non-zero value
Acknowledgements
with vanishing phase lag. In the high frequency limit (f * ≫ 1) transport is
controlled by transient diffusion within boundary layers, showing strong This research was supported by the US National Science Foundation
dependence of each transfer function on frequency. In this spectral re (Award no. 1931659) and the Department of Mechanical Science and
gion, the magnitude of 〈Sh〉
̃ and 〈N‖ 〉 respectively increase and decrease Engineering at the University of Illinois at Urbana-Champaign. We thank
according to a power law with exponent ±1/2, while their phases Irwin Loud for fabricating and characterizing the structure of the elec
approach ± π/4 in agreement with semi-infinite Warburg response. trodes used for experimental validation that were fabricated at Siebel
We embedded these TFs into an up-scaled model using porous elec Center for Design, University of Illinois at Urbana-Champaign (UIUC).
trode theory to investigate the dynamic response of a redox flow battery Structural characterization was carried out in part in the Materials
operating under step impulse current input. The results obtained suggest Research Laboratory Central Research Facilities, UIUC. We thank Vu Do
that the dynamic response of such systems is dictated by the charac for assistance with flow-cell setup and electrochemical characterization.
teristic time-scale of transient current variation. For fast fluctuations the We thank Professor Joseph Bentsman for discussion concerning the
effect of transient mass transfer is substantial, and the system response is mathematical properties of transfer functions.
found to lag. For such operation significant reduction in concentration
polarization is achieved with 60%–70% reduction for an applied current Appendix A. Supplementary data
of 90% of the limiting current. Finally, we demonstrated that short
duration charging/discharging of a redox flow battery with over- Supplementary data to this article can be found online at https://doi.
limiting current is possible, which provides a means to increase the org/10.1016/j.jpowsour.2023.232756.
utilization of charge capacity of existing flow-based electrochemical
devices that are operated under transient conditions. Furthermore, References
while operating under fluctuations of low characteristic frequency, such
systems reach a pseudo-steady limit with negligible transient mass [1] W. Cole, N. Gates, T. Mai, D. Greer, P. Das, 2019 Standard Scenarios Report: A U.S.
Electricity Sector Outlook, Natl. Renew. Energy Lab., No. NREL/TP-6A20-74110,
transport effect. Here, the accelerated advection of reactant species from
2019, pp. 8–12, https://www.nrel.gov/docs/fy20osti/74110.pdf.
the reactor to the tank causes earlier (in time) depletion at the reactive [2] W. Wang, Q. Luo, B. Li, X. Wei, L. Li, Z. Yang, Recent progress in redox flow battery
interfaces. research and development, Adv. Funct. Mater. 23 (2013) 970–986, https://doi.
org/10.1002/adfm.201200694.
These findings motivate the exploration of related effects using the
[3] G.L. Soloveichik, Flow batteries: current status and trends, Chem. Rev. 115 (2015)
presently introduced transfer functions to simulate the charging current 11533–11558, https://doi.org/10.1021/cr500720t.
input that is typical of renewable power sources, as well as to model such [4] Y. Matsuda, K. Tanaka, M. Okada, Y. Takasu, M. Morita, T. Matsumura-Inoue,
effects with different electrode microstructures. Additional physico A rechargeable redox battery utilizing ruthenium complexes with non-aqueous
organic electrolyte, J. Appl. Electrochem. 18 (1988) 909–914, https://doi.org/
chemical effects not shown here (e.g., macro-scale migration and 10.1007/BF01016050.
diffusion of redox-active species and supporting-electrolyte ions and [5] Q. Liu, A.E.S. Sleightholme, A.A. Shinkle, Y. Li, L.T. Thompson, Non-aqueous
macro-scale superficial velocity gradients induced by the use of vanadium acetylacetonate electrolyte for redox flow batteries, Electrochem.
Commun. 11 (2009) 2312–2315, https://doi.org/10.1016/j.elecom.2009.10.006.
serpentine, interdigitated, or other flow fields) are readily coupled with [6] M. Yao, H. Senoh, S.I. Yamazaki, Z. Siroma, T. Sakai, K. Yasuda, High-capacity
the present B-UTM by adopting the two transfer functions of interest to organic positive-electrode material based on a benzoquinone derivative for use in
express the interfacial redox-reaction rate and the average advection rechargeable lithium batteries, J. Power Sources 195 (2010) 8336–8340, https://
doi.org/10.1016/j.jpowsour.2010.06.069.
rate in place of those determined by the conventional film law of mass [7] B. Huskinson, J. Rugolo, S.K. Mondal, M.J. Aziz, A high power density, high
transfer in previous porous-electrode theory formulations of RFB models efficiency hydrogen-chlorine regenerative fuel cell with a low precious metal
(see Ref. [24]). Further, the similarity of the time-domain governing content catalyst, Energy Environ. Sci. 5 (2012) 8690–8698, https://doi.org/
10.1039/c2ee22274d.
equations motivates the use of the transfer functions defined here in
[8] B. Sun, M. Skyllas-Kazacos, Modification of graphite electrode materials for
other electrochemical contexts using flow-through electrodes, including vanadium redox flow battery application-I. Thermal treatment, Electrochim. Acta
electrochemical separations and energy conversion. Increased accuracy 37 (1992) 1253–1260, https://doi.org/10.1016/0013-4686(92)85064-R.
[9] B. Sun, M. Skyllas-Kazacos, Chemical modification of graphite electrode materials
of the model is also possible by additionally including macro-scale
for vanadium redox flow battery application-part II. Acid treatments, Electrochim.
dispersion, ion conduction, entrance region effects, and non-uniform Acta 37 (1992) 2459–2465, https://doi.org/10.1016/0013-4686(92)87084-D.
surface concentration effects. [10] H. Kabir, I.O. Gyan, I. Francis Cheng, Electrochemical modification of a pyrolytic
graphite sheet for improved negative electrode performance in the vanadium redox
flow battery, J. Power Sources 342 (2017) 31–37, https://doi.org/10.1016/j.
CRediT authorship contribution statement jpowsour.2016.12.045.
[11] H.R. Jiang, W. Shyy, M.C. Wu, R.H. Zhang, T.S. Zhao, A bi-porous graphite felt
electrode with enhanced surface area and catalytic activity for vanadium redox
Md Abdul Hamid: Conceptualization, Methodology, Formal anal flow batteries, Appl. Energy 233–234 (2019) 105–113, https://doi.org/10.1016/j.
ysis, Software, Validation, Writing – original draft, Writing – review & apenergy.2018.10.033.
editing. Kyle C. Smith: Conceptualization, Methodology, Formal anal [12] Z. He, M. Li, Y. Li, J. Zhu, Y. Jiang, W. Meng, H. Zhou, L. Wang, L. Dai, Flexible
electrospun carbon nanofiber embedded with TiO2 as excellent negative electrode
ysis, Supervision, Project administration, Funding acquisition, Writing –
for vanadium redox flow battery, Electrochim. Acta 281 (2018) 601–610, https://
review & editing. doi.org/10.1016/j.electacta.2018.06.011.
[13] P.C. Ghimire, R. Schweiss, G.G. Scherer, N. Wai, T.M. Lim, A. Bhattarai, T.
D. Nguyen, Q. Yan, Titanium carbide-decorated graphite felt as high performance
negative electrode in vanadium redox flow batteries, J. Mater. Chem. A. 6 (2018)
Declaration of competing interest
6625–6632, https://doi.org/10.1039/c8ta00464a.
[14] H. Gursu, M. Gencten, Y. Sahin, Preparation of sulphur-doped graphene-based
The authors declare the following financial interests/personal re electrodes by cyclic voltammetry: a potential application for vanadium redox flow
lationships which may be considered as potential competing interests: battery, Int. J. Electrochem. Sci. 13 (2018) 875–885, https://doi.org/10.20964/
2018.01.71.
Kyle C. Smith reports financial support was provided by National Sci [15] D.M. Kabtamu, A.W. Bayeh, T.C. Chiang, Y.C. Chang, G.Y. Lin, T.H. Wondimu, S.
ence Foundation. Md Abdul Hamid reports financial support was K. Su, C.H. Wang, TiNb 2 O 7 nanoparticle-decorated graphite felt as a high-
14
performance electrode for vanadium redox flow batteries, Appl. Surf. Sci. 462 [41] T. Jacobsen, K. West, Diffusion impedance in planar, cylindrical and spherical
(2018) 73–80, https://doi.org/10.1016/j.apsusc.2018.08.101. symmetry, Electrochim. Acta 40 (1995) 255–262, https://doi.org/10.1016/0013-
[16] J.D. Milshtein, K.M. Tenny, J.L. Barton, J. Drake, R.M. Darling, F.R. Brushett, 4686(94)E0192-3.
Quantifying mass transfer rates in redox flow batteries, J. Electrochem. Soc. 164 [42] J. Bisquert, G. Garcia-Belmonte, F. Fabregat-Santiago, P.R. Bueno, Theoretical
(2017) E3265–E3275, https://doi.org/10.1149/2.0201711jes. models for ac impedance of finite diffusion layers exhibiting low frequency
[17] J.Q. Chen, B.G. Wang, H.L. Lv, Numerical simulation and experiment on the dispersion, J. Electroanal. Chem. 499 (1999) 112–120, https://doi.org/10.1016/
electrolyte flow distribution for all vanadium redox flow battery, Adv. Mater. Res. S0022-0728(99)00346-0.
236–238 (2011) 604–607. https://doi.org/10.4028/www.scientific.net/AMR.23 [43] K.C. Smith, R.D. Dmello, Na-ion desalination (NID) enabled by Na-blocking
6-238.604. membranes and symmetric Na-intercalation: porous-electrode modeling,
[18] D.S. Aaron, Q. Liu, Z. Tang, G.M. Grim, A.B. Papandrew, A. Turhan, T. J. Electrochem. Soc. 163 (2016) A530–A539, https://doi.org/10.1149/
A. Zawodzinski, M.M. Mench, Dramatic performance gains in vanadium redox flow 2.0761603jes.
batteries through modified cell architecture, J. Power Sources 206 (2012) [44] K.C. Smith, Theoretical evaluation of electrochemical cell architectures using
450–453, https://doi.org/10.1016/j.jpowsour.2011.12.026. cation intercalation electrodes for desalination, Electrochim. Acta 230 (2017)
[19] R.M. Darling, M.L. Perry, The influence of electrode and channel configurations on 333–341, https://doi.org/10.1016/j.electacta.2017.02.006.
flow battery performance, J. Electrochem. Soc. 161 (2014) A1381–A1387, https:// [45] S. Liu, K.C. Smith, Quantifying the trade-offs between energy consumption and salt
doi.org/10.1149/2.0941409jes. removal rate in membrane-free cation intercalation desalination, Electrochim. Acta
[20] X. Ke, J.M. Prahl, I.J.D. Alexander, J.S. Wainright, T.A. Zawodzinski, R.F. Savinell, 271 (2018) 652–665, https://doi.org/10.1016/j.electacta.2018.03.065.
Rechargeable redox flow batteries: flow fields, stacks and design considerations, [46] D.L. Parkhurst, C.A.J. Appelo, User’s Guide to PHREEQC (Version 2): A Computer
Chem. Soc. Rev. 47 (2018) 8721–8743, https://doi.org/10.1039/c8cs00072g. Program for Speciation, Batch-Reaction, One-Dimensional Transport, and Inverse
[21] D. Schmal, J. Van Erkel, P.J. Van Duin, Mass transfer at carbon fibre electrodes, Geochemical Calculations, Denver, CO, 1999.
J. Appl. Electrochem. 16 (1986) 422–430, https://doi.org/10.1007/BF01008853. [47] K.C. Smith, Y.-M. Chiang, W. Craig Carter, Maximizing energetic efficiency in flow
[22] A.A. Shah, M.J. Watt-Smith, F.C. Walsh, A dynamic performance model for redox- batteries utilizing non-Newtonian fluids, J. Electrochem. Soc. 161 (2014)
flow batteries involving soluble species, Electrochim. Acta 53 (2008) 8087–8100, A486–A496, https://doi.org/10.1149/2.011404jes.
https://doi.org/10.1016/j.electacta.2008.05.067. [48] F.Y. Fan, W.H. Woodford, Z. Li, N. Baram, K.C. Smith, A. Helal, G.H. McKinley, W.
[23] X. You, Q. Ye, P. Cheng, The dependence of mass transfer coefficient on the C. Carter, Y.-M. Chiang, Polysulfide flow batteries enabled by percolating
electrolyte velocity in carbon felt electrodes: determination and validation, nanoscale conductor networks, Nano Lett. 14 (2014) 2210–2218, https://doi.org/
J. Electrochem. Soc. 164 (2017) E3386–E3394, https://doi.org/10.1149/ 10.1021/nl500740t.
2.0401711jes. [49] T. Wei, F.Y. Fan, A. Helal, K.C. Smith, G.H. Mckinley, Y.-M. Chiang, J.A. Lewis,
[24] V.P. Nemani, K.C. Smith, Assignment of energy loss contributions in redox flow Biphasic electrode suspensions for Li-ion semi-solid flow cells with high energy
batteries using exergy destruction analysis, J. Power Sources 447 (2020), 227371, density, fast charge transport, and low-dissipation flow, Adv. Energy Mater.
https://doi.org/10.1016/j.jpowsour.2019.227371. (2015), 1500535, https://doi.org/10.1002/aenm.201500535.
[25] J.S. Newman, K. Thomas-Alyea, Electrochemical Systems, third ed., Prentice Hall, [50] Z. Li, K.C. Smith, Y. Dong, N. Baram, F.Y. Fan, J. Xie, P. Limthongkul, W.C. Carter,
Englewood Cliffs, N.J, 2004. Y.M. Chiang, Aqueous semi-solid flow cell: demonstration and analysis, Phys.
[26] J. Leadbetter, L.G. Swan, Selection of battery technology to support grid-integrated Chem. Chem. Phys. 15 (2013) 15833–15839, https://doi.org/10.1039/
renewable electricity, J. Power Sources 216 (2012) 376–386, https://doi.org/ c3cp53428f.
10.1016/j.jpowsour.2012.05.081. [51] D.A. Edwards, M. Shapiro, H. Brenner, Dispersion and reaction in two-dimensional
[27] J. Leadbetter, L. Swan, Battery storage system for residential electricity peak model porous media, Phys. Fluids A Fluid Dyn. 5 (1993) 837–848, https://doi.org/
demand shaving, Energy Build. 55 (2012) 685–692, https://doi.org/10.1016/j. 10.1063/1.858631.
enbuild.2012.09.035. [52] B.B. Dykaar, P.K. Kitanidis, Macrotransport of a biologically reacting solute
[28] E. Sum, M. Rychcik, M. Skyllas-kazacos, Investigation of the V(V)/V(IV) system for through porous media, Water Resour. Res. 32 (1996) 307–320, https://doi.org/
use in the positive half-cell of a redox battery, J. Power Sources 16 (1985) 85–95, 10.1029/95WR03241.
https://doi.org/10.1016/0378-7753(85)80082-3. [53] H. Brenner, Dispersion resulting from flow through spatially periodic porous
[29] S. Zhong, M. Skyllas-Kazacos, Electrochemical behaviour of vanadium(V)/ media, Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 297 (1980) 81–133, https://
vanadium(IV) redox couple at graphite electrodes, J. Power Sources 39 (1992) 1–9, doi.org/10.1098/rsta.1980.0205.
https://doi.org/10.1016/0378-7753(92)85001-Q. [54] H. Brenner, P.M. Adler, Dispersion resulting from flow through spatially periodic
[30] X.G. Li, K.L. Huang, S.Q. Liu, L.Q. Chen, Electrochemical behavior of diverse porous media II. Surface and intraparticle transport, Philos. Trans. R. Soc. A Math.
vanadium ions at modified graphite felt electrode in sulphuric solution, J. Cent. Phys. Eng. Sci. 307 (1982) 149–200, https://doi.org/10.1098/rsta.1982.0108.
South Univ. Technol. (English Ed. 14 (2007) 51–56, https://doi.org/10.1007/ [55] D.A. Edwards, M. Shapiro, H. Brenner, Dispersion and reaction in two-dimensional
s11771-007-0011-6. model porous media, Phys. Fluids A Fluid Dyn. 5 (1993) 837, https://doi.org/
[31] T. Yamamura, N. Watanabe, T. Yano, Y. Shiokawa, Electron-transfer kinetics of 10.1063/1.858631.
Np3+/Np4+, NpO 2+/NpO22+, V2+/V 3+, and VO2+/VO2+ at carbon [56] W.G. Gray, A derivation of the equations for multi-phase transport, Chem. Eng. Sci.
electrodes, J. Electrochem. Soc. 152 (2005) A830–A836, https://doi.org/10.1149/ 30 (1975) 229–233.
1.1870794. [57] S. Whitaker, The transport equations for multi-phase systems, Chem. Eng. Sci. 28
[32] J. González-García, P. Bonete, E. Expósito, V. Montiel, A. Aldaz, R. Torregrosa- (1973) 139–147, https://doi.org/10.1016/0009-2509(73)85094-8.
Maciá, Characterization of a carbon felt electrode: structural and physical [58] J. Newman, W. Tiedemann, Porous-electrode theory with battery applications,
properties, J. Mater. Chem. 9 (1999) 419–426, https://doi.org/10.1039/ AIChE J. 21 (1975) 25–41, https://doi.org/10.1002/aic.690210103.
a805823g. [59] J. Luo, A. Sam, B. Hu, C. DeBruler, X. Wei, W. Wang, T.L. Liu, Unraveling pH
[33] K. Kato, K. Kano, T. Ikeda, Electrochemical characterization of carbon felt dependent cycling stability of ferricyanide/ferrocyanide in redox flow batteries,
electrodes for bulk electrolysis and for biocatalyst-assisted electrolysis, Nano Energy 42 (2017) 215–221, https://doi.org/10.1016/J.
J. Electrochem. Soc. 147 (2000) 1449–1453, https://doi.org/10.1149/1.1393376. NANOEN.2017.10.057.
[34] Q. Xu, L. Qin, H. Su, L. Xu, P. Leung, C. Yang, H. Li, Electrochemical and transport [60] K. Gong, F. Xu, J.B. Grunewald, X. Ma, Y. Zhao, S. Gu, Y. Yan, All-soluble all-iron
characteristics of V(II)/V(III) redox couple in a nonaqueous deep eutectic solvent: aqueous redox-flow battery, ACS Energy Lett. 1 (2016) 89–93, https://doi.org/
temperature effect, J. Energy Eng. 143 (2017), 04017051, https://doi.org/ 10.1021/acsenergylett.6b00049.
10.1061/(ASCE)EY.1943-7897.0000484. [61] N.G. Tsierkezos, A. Knauer, U. Ritter, Thermodynamic investigation of
[35] G. Qiu, A.S. Joshi, C.R. Dennison, K.W. Knehr, E.C. Kumbur, Y. Sun, 3-D pore-scale ferrocyanide/ferricyanide redox system on nitrogen-doped multi-walled carbon
resolved model for coupled species/charge/fluid transport in a vanadium redox nanotubes decorated with gold nanoparticles, Thermochim. Acta 576 (2014) 1–8,
flow battery, Electrochim. Acta 64 (2012) 46–64, https://doi.org/10.1016/j. https://doi.org/10.1016/J.TCA.2013.11.020.
electacta.2011.12.065. [62] P. Kulesza, T. Jedral, Z. Galus, The electrode kinetics of the Fe(CN)63− /Fe(CN)
[36] M.D.R. Kok, R. Jervis, T.G. Tranter, M.A. Sadeghi, D.J.L. Brett, P.R. Shearing, J. 64− system on the platinum rotating electrode in the presence of different anions,
T. Gostick, Mass transfer in fibrous media with varying anisotropy for flow battery J. Electroanal. Chem. Interfacial Electrochem. 109 (1980) 141–149.
electrodes: direct numerical simulations with 3D X-ray computed tomography, [63] V.V. Nikonenko, N.D. Pismenskaya, E.I. Belova, P. Sistat, P. Huguet, G. Pourcelly,
Chem. Eng. Sci. 196 (2019) 104–115, https://doi.org/10.1016/j.ces.2018.10.049. C. Larchet, Intensive current transfer in membrane systems: modelling,
[37] T. Tichter, D. Andrae, J. Mayer, J. Schneider, M. Gebhard, C. Roth, Theory of cyclic mechanisms and application in electrodialysis, Adv. Colloid Interface Sci. 160
voltammetry in random arrays of cylindrical microelectrodes applied to carbon felt (2010) 101–123, https://doi.org/10.1016/J.CIS.2010.08.001.
electrodes for vanadium redox flow batteries, Phys. Chem. Chem. Phys. 21 (2019) [64] D. Deng, E.V. Dydek, J.-H. Han, S. Schlumpberger, A. Mani, B.Z.M.Z. Bazant,
9061–9068, https://doi.org/10.1039/c9cp00548j. Overlimiting current and shock electrodialysis in porous media, Langmuir 29
[38] M.A. Hamid, K.C. Smith, Modeling the transient effects of pore-scale convection (2013) 16167–16177.
and redox reactions in the pseudo-steady limit, J. Electrochem. Soc. 167 (2020), [65] P. Grathwohl, D. Lerner, Natural attenuation of organic pollutants in groundwater,
013521, https://doi.org/10.1149/2.0212001jes. J. Contam. Hydrol. 53 (2001) 173–174, https://doi.org/10.1016/s0169-7722(01)
[39] X. Hu, S. Stanton, L. Cai, R.E. White, Model order reduction for solid-phase 00165-6.
diffusion in physics-based lithium ion cell models, J. Power Sources 218 (2012) [66] G. Yossifon, H.-C. Chang, Selection of nonequilibrium overlimiting currents:
212–220, https://doi.org/10.1016/j.jpowsour.2012.07.007. universal depletion layer formation dynamics and vortex instability, Phys. Rev.
[40] J. Huang, Diffusion impedance of electroactive materials, electrolytic solutions and Lett. 101 (2008), 254501.
porous electrodes: Warburg impedance and beyond, Electrochim, Acta 281 (2018) [67] Y. Jaluria, K. Torrance, Computational Heat Transfer, Hemisphere Publishing
170–188, https://doi.org/10.1016/j.electacta.2018.05.136. Corporation, 1986.
15
[68] V.P. Nemani, K.C. Smith, Uncovering the role of flow rate in redox-active polymer Microw. Wireless Compon. Lett. 18 (2008) 383–385, https://doi.org/10.1109/
flow batteries: simulation of reaction distributions with simultaneous mixing in LMWC.2008.922585.
tanks, Electrochem. Acta. 247 (2017) 475–485, https://doi.org/10.1016/j. [73] J.P. Sørensen, W.E. Stewart, Computation of forced convection in slow flow
electacta.2017.07.008. through ducts and packed beds-III. Heat and mass transfer in a simple cubic array
[69] X. Hu, S. Stanton, L. Cai, R.E. White, A linear time-invariant model for solid-phase of spheres, Chem. Eng. Sci. 29 (1974) 827–832, https://doi.org/10.1016/0009-
diffusion in physics-based lithium ion cell models, J. Power Sources 214 (2012) 2509(74)80201-0.
40–50, https://doi.org/10.1016/j.jpowsour.2012.04.040. [74] P.W. Appel, J. Newman, Application of the limiting current method to mass
[70] B. Gustavsen, A. Semlyen, Rational approximation of frequency domain responses transfer in packed beds at very low Reynolds numbers, AIChE J. 22 (1976)
by vector fitting, IEEE Trans. Power Deliv. 14 (1999) 1052–1059, https://doi.org/ 979–984, https://doi.org/10.1002/aic.690220605.
10.1109/61.772353. [75] P.S. Fedkiw, J. Newman, Mass-transfer coefficients in packed beds at very low
[71] B. Gustavsen, Improving the pole relocating properties of vector fitting, IEEE Trans. Reynolds numbers, Int. J. Heat Mass Tran. 25 (1982) 935–943, https://doi.org/
Power Deliv. 21 (2006) 1587–1592, https://doi.org/10.1109/ 10.1016/0017-9310(82)90069-2.
TPWRD.2005.860281. [76] A. Caprani, C. Deslouis, S. Robin, B. Tribollet, Transient mass transfer at partially
[72] D. Deschrijver, M. Mrozowski, T. Dhaene, D. De Zutter, Macromodeling of blocked electrodes: a way to characterize topography, J. Electroanal. Chem. 238
multiport systems using a fast implementation of the vector fitting method, IEEE (1987) 67–91, https://doi.org/10.1016/0022-0728(87)85166-5.
16

Hamid JPS 2023

Uploaded by

Hamid JPS 2023

Uploaded by

Journal of Power Sources 565 (2023) 232756

Contents lists available at ScienceDirect

Journal of Power Sources

A bottom-up, multi-scale theory for transient mass transport of redox-active

• Bottom-up theory is presented for tran

1. Introduction frequency-based formalism to predict transient dynamics of solid-state

tures the transient response of solutions undergoing convective mass (7)

this property of 〈Sh〉

Fig. 4. Variation of transfer functions (a) 〈Sh〉

irrational TF and A, B, C, and D are system matrices/vectors. Since 〈Sh〉

Fig. 6. Variation of time-maximal concentration

Fig. 8. Variation of time-averaged reactant con

sional frequency f * . From the variations of 〈Sh〉

You might also like

Hamid JPS 2023

Uploaded by

Hamid JPS 2023

Uploaded by

Journal of Power Sources 565 (2023) 232756

Contents lists available at ScienceDirect

Journal of Power Sources

A bottom-up, multi-scale theory for transient mass transport of redox-active

• Bottom-up theory is presented for tran­

1. Introduction frequency-based formalism to predict transient dynamics of solid-state

tures the transient response of solutions undergoing convective mass (7)

this property of 〈Sh〉

Fig. 4. Variation of transfer functions (a) 〈Sh〉

irrational TF and A, B, C, and D are system matrices/vectors. Since 〈Sh〉

Fig. 6. Variation of time-maximal concentration

Fig. 8. Variation of time-averaged reactant con­

sional frequency f * . From the variations of 〈Sh〉

You might also like

• Bottom-up theory is presented for tran

Fig. 8. Variation of time-averaged reactant con