w13-46B6/81/040455-15some/o

PerppmonRcss Ltd.

A COMPARISON BETWEEN FLOW-THROUGH AND FLOW-

BY POROUS ELECTRODES FOR REDOX ENERGY STORAGE

JAMES A. TRAINHAM) and JOHN NEWMAN +

Materials and Molecular Research Division, Lawrence Berkeley Laboratory and Department of Chemical
Engineering, University of California Berkeley, CA 94720, U.S.A.

(Received 3 August 1979; in reotiedfirm 13 May 1980)

Abstract-An economic comparison is made between electrode configurations for flow redox battery
applicationr (i) the flow-through configuration (current parallel to the fluid flow) and (ii) the flow-by
configuration {current perpendicular to the fluid flow). Steady-state computer models aredeveloped for each
electrode system. These models are used to predict current density, cell voltage, and power density over a
complete cycle (charge and discharge). The economic comparison is made by optimizing each configuration
with respect to an objective function appropriate for redox battery applications. Only the variable costs are
considered. The resultsof the optimization show that the flow-by configuration is superior. The flow-through
configuration not only yields a lower return on investment, but it is impractical due to a requirement of
extremelv low flow rates IRe -z 0.0011. Its failure is due to current flow (and ohmic potential drop) in the same
direction-as the fluid flow. ’

NOMENCLATURE superficial current density in the matrix,

A/&m’
specific inrerfacial area, cm-’ superficial current density in pore-solution
concentration of the reactant species, in the phase, A/cm1
flow-by analysis it is referred to as the reduced transfer current for reaction j per unit in-
species, mole/c& terfacial area, A/cm’
concentration of the oxidized species in a redox = s,i/nFve,,, dimensionless superficial cur-
reactio- mole/cm3 rent density to an electrode
concentration of the product species in a redox total current density to an electrode due to
reaction, mole/cm3 metal-deposition reaction, A/cm2
reactant species concentration in the side total current to an electrode, A
reaction, mole/cm3 current density on charge, A/cm’
concentration at electrode exit, mole/cm3 current density on discharge, A/cm2
cost per unit volume, $/m’ pore-wall flux of species i, mole/cm2-s
cost per superficial-electrode area, S/m’ averagemass-transfer coefficient between flow-
capital and fixed costs, $/year-m3 ing solution and electrode surface, cm/s
fixed costs, $/year-m2 thickness of porous electrode, cm
variable costs, S/year-m’ separator thickness in flow-by porous elec-
daytime power costs, .$/W trode, cm
nighttime power costs, $/W gap thickness of flow-through porous elec-
downstream counterelectrode, downstream trode, cm
current collector symbol for the chemical formula of species i
downstream counterelectrode, upstream cur- number of electrons transferred in an electro-
rent collector chemical reaction
axial dispersion coefficient of the reactant number of electrons transferred in reaction j
species, cm’/s hydrogen or chlorine partial pressure, atm
molecular diffusion coefficient of the reactant P&let number, v/aD,
spies, cm2/s = ( - sRio,, ,eT/nFkmcR,)t +a.~lac~, dimension-
effective diffusion coefficient of the reactant less parameter describing the relative import-
species, cm2/s ance of the anodic term in a metal deposition
= (DR + 0,)ok,/v2. dimensionless parameter reaction, and a redox reaction
describing the relative importance of axial
diffusion and dispersion
, dimension-
Faraday’s constant, 96487 C/equiv less parameter describing the relative import-
superficial current density to an electrode, ance of the ohmic potential drop
A/cm2 (for the flow-by system i varies in the
direction of fluid flow) = - sRi,,,,rlnFk,eacsF~uIRT
exchange current density for reaction j, A/cm2
( --~k,C~/ISR~q~,rs~)e’,S’ucR, dimensionless
exchange current density for reaction j at a
parameter descrtbmg the relative importance
reference compomtlon ci, rrf, A/cm2
of the forward term in the side reaction
= (- ~~i,~,,~/nFk,c~~)ss+~c~/~<aexp[ - (Q
l Present address: E. I. du Pant de Nemours & Co., Inc. + a,,)AUF/RTl, dimensionless parameter de-
Polymer Products Department, Experimental Station, scribing the’relative rate of the backward term
WTmington, DE 19898, U.S.A. in the side reaction
To whom correspondence should be addressed. dimensionless parameter which character&s
456 JAMES A. TRAINHAU AND JOHNNEWMAN

the ohmic potential drop in the pore-solution exponent in composition dependence of ex-
phase, see equation (30) change current density
P6 dimensionless parameter which characterizes porosity or void volume
the ohmic potential drop in the matrix phase, quasi-electrostatic potential, V
see equation (3 1) electrostatic potential in matrix phase, V
p7 = La/k,,, ratio of mass transfer coetlicients quasi-electrostatic potential in the pure so-
for reactant and product series in a redox lution phase.
reaction ohmic potential drop across porous electrode,
P; dimensionless parameter describing the re- V
lative importance of the cathodic term in a estimate of ohmic potential across porous
metal deposition reaction and a rcdox reaction, electrode at the limiting current, V
see equation (23) I,Lga&,. ohmic drop across gap in flow
same physical significance as P2 only for an through porous electrode, V
anodic reactant, see equation (24) = QI, -0, - Uj, surface overpotential for re-
same physical significance as P3, see equation action j, V
(25) = 9, - m2 local overpotential, V
same physical significance as P,, see equation dimensionless local overpotential for the ca-
(26) thodic reactant
same physical significance as Ps, see equation dimensionless local potential for the anodic
(33) reactant
same physical significance as P,, see equation effective conductivity of solution, mho/cm
(34) intrinsic conductivity of solution, mho/crn
Yij cathodic reaction order effective conductivityofthe separator, mho/cm
r the effect of interest, depreciation and taxes, viscosity, g/cm-s.
yr-’
R universal gas constant, 8.3143 J/mole-K
Re = Y/~Y or v/$av, Reynolds number
s on-stream factor, days operating/365 days or upstream feed
operating cycles/year wall or surface
%j stoichiometric coefkient of species i in elec- reactant or reduced species
trode reaction j product species
SR stoichiometric coefficient of reactant species oxidized species
SP stoichiometric coefficient of product species
T absolute temperature, K
r, charging time, s
rr discharge time, s INTRODUCTION
u open-circuit cell potential. V
VD upstream counterelectrode. downstream cur- In the post “energy crisis” years the ability of electric
rent collector utilities to meet peak power demands economically has
VU upstream counterelectrode, upstream current become increasingly acute. One answer to thisproblem
collector
is to store off-peak power and to recover it during high
ujw open-circuit potential for reaction j at local
demand periods.(power load-leveling). Energy storage
wall composition relative to a reference elec-
trode of a given kind, V requirements for electric utilities may be met in several
lJ; standard electrode potential for reaction j. V ways. The numerous advantages of flow-redox bat-
teries (Thaller[ 1,2], Beccu[3], and Kangro[4,5] have
V: standard electrode potential for reaction k, V
been discussed by Yao and Birk[B], and Warshay and
Vl standard electrode potential of reference Wright[‘I] ). Some particularly attractive features are
electrode, V
listed below:
AU = V, - V,, differencein open-circuit cell pot-
entials of the side reaction and primary re- (1) Unlimited cycle life regardless of the depth of
action at the reference composition, V discharge because redox reactions do not affect the
U superficial fluid velocity, cm/s.
morphology of the electrode material.
VOP potential of the cathode current collector re-
lative toasaturated calomel referenceelectrode
(2) Simple electrode reactions.
placed in the dilute product stream, V (3) High temperatures not required.
( ~Of%m maximum value of VOP at which appreciable (4) Minimal environmental impact.
side reaction does not occur, V (5) Readily available reagents and materials of
A V, VOP - CJ: + qa,, v construction.
x distance through porous electrodein the direc-
tion of fluid flow, cm Preliminary capital cost estimates and size estimates
xak,/v dimensionless d’istance through porous for an iron-titanium [Ti(III) 1Ti(IV) 11Fe(III)I Fe(II)]
Y
electrode for the flow-through analysis, in the redox battery indicate economic viability for the
flow-by analysis it is the distance in the direc- system (Warshay and Wright[7]). Many other redox
tion of current flow, cm couples have been suggested (Thaller[l, 21, Beccu[3],
valence or charge number of species i Gieyer, Swette, and Cah111[8], and Kangro[4,5]). Of
, the iron-chromium system [Cr(III)I
Cr(II)I IFe(III)IFe(II)] appears to be the most
Greek letters promising. Recently, the NASA-Lewis Research
= ak,,,/o, reciprocal of penetration depth at the Center, using an iron-chromium cell, has achieved a
limiting current, cm- * power density of approximately 915 W per square
anodic transfw coefficient for reaction j meter of membrane separator (Gahn and
cathodic transfer coefficient for reaction j Hagedorn[9]).
 A comparison between flow-through and flow-by porous electrodes 457

Two basic electrode configurations have been pro- (1) the flow-through configuration (current parallel
posed for flow-redox batteries. Kangro[4,5] has sug- to the fluid flow);
gested an electrode configuration where the current (2) the flow-by configuration (current perpendi-
and fluid flow are in parallel (flow-through cular to the fluid flow). The basis for comparison is the
configuration, see Fig. la and b). Thaller[l, 21 and return on investment based on operating costs.
Beccu[3] have suggested a cell configuration where the
current is perpendicular to the fluid flow and a
separator is essential (flow-by configuration, see Fig. FLOW-THROUGH POROUS ELECTRODES
1~)).The flow-through configuration (Fig. la and b) has
the advantage over the flow-by configuration (Fig. Ic) The analysis for a redox reaction is similar to that
that a separator is not required; however, a major presented by Trainham and Newman[ll, 121 for a
disadvantage of this configuration can be seen in Fig. metal-deposition reaction. However, if the dimen-
lb. Consider an iron-titanium cell. On discharge the sional analysis also presented by Trainham and
unreacted Ti(II1) and Fe(TII) would be lost according Newman[ll, 121 is to be used, separate analyses must
to the reaction be considered for both a cathodic limiting reactant and
an anodic limiting reactant.
Ti(II1) + Fe(III)+ Ti(IV) + Fe(I1) (1)
when the streams are allowed to mix after flowing Cathodic reactant. The Butler-Volmer equation
through the porous electrodes (NewmanClO]). may be written as (Newman and Tiedemann[13] and
The purpose of this paper is to compare, on an Trainham and Newman[ 1 l] )
economic basis, the performance of the two con-
figurations for redox-energy storage. As mentioned
above, Warshay and Wright[7] have made preliminary
capital cost estimates for the two-tank, flow-by
configuration. The flow-by configuration is usually
considered to be either a two-tank or a four-tank
system. The flow-through configuration is primarily a where I’,, rsTis the value of the exchange current density
three-tank system. The two-tank flow-by configu- at refer&ce compositions cR,ref and cp,rcf. The re-
ration can also be operated with a time varying ference electrode used in measuring @‘zis of the same
superficial velocity, whereas, the other systems cannot kind as the redox reaction, and the concentration of the
due to their single-pass operation (Thaller[l]). (By reactant and product species in the reference electrode
single-pass operation we mean the reactants on each compartment, cRr and cp,, and the reference con-
half cycle pass through the electrode only once.) centrations c=.,~~and ~~,~~r,have been set equal to cR,
The perpective here is to compare the steady-state and c,, (the upstream feed concentration of the redox
performance of two operational redox storage systems: couple i _It IS
. also assumed that sR = - 1 and sp = 1.
If axial diffusion is small compared to axial disper-
1Separator or Gop sion or if the diffusion coefficients of the reactant and
(o) ,-[Link].“. Elec+rodes product species ate equal, the local concentration
the product species within the poresis simply related to
of

the local reactant concentration within the pores and

the feed concentration

t
cp = c,j+cpl-cR. (3)
Also, for a redox couple, the wall fluxes of the
Direction of Flow reactant and product species are equal and opposite

(bl k,,&, - CRr) = - k,&, - CP,. 1. (41

One can now write cp, in terms of cR and car with the
Current use of (3) as
V V
EL- cfi
-_=1+%_!$ I+!! _kmR?!!K* (5)
1 ‘=PJ (
cPJ IIIP > km, cPJ
Direction Of Flow
Equation (5) allows the explicit dependence of the
4v tv product species concentration on the local reaction
(cl ‘- ‘- rate to be removed from (2) with the result:

iID- ? t
Directhe Of Flow
Current
4, =ioRmf{
[l+$(cRj-‘a(1
-2)
-2cRw.)]expt&q)

-$exp (-gq)},
Fig. 1. various configurations of electrode piacement relative
to the direction of fluid flow. (a) and (b) are the flow-through
configuration on charge and discharge. (c) is the flow-by where q =~l-@z. Only one differential material
configuration. balance is now requited-that for the cathodic re-
458 JAMESA. TRAINHAM AND JOHN NEWMAN

actant. If the cathodic reactant R participates only in

the primary reaction, then
e-(1tF) PI exp CC1
+ ~,d+drl’f

.
%R = -z jRn =$ k,(c, - cRw). (7)
-1+exp(t13+P,P,/e~/expC(~ +%R/%Rh'l
(13)

The elimination of cRWbetween (6) and (7) results in the Charge balance
equation d2$
dy2=
jRn =

1
(%R + %RIF

RT rl

(8)
The dimensionless parameters D’, P,, P2, P,, and P4
A material balance equation for the case of a cathodic me the same as those defined by Trainham and
limiting reactant is then obtained by substituting the Newman[l I]. The additional parameters P, and 8,
above equation for i,, into the continuity equation are given by
(Newman and Tiedemann[l3])
P7 = k,RlkP, (15)
d*r.,
od~=e(D,+DJg + ah,, (9) &, = %‘/IcRf. 06)

Anodic reactant. For an anodic reactant, equation

with the result (2) is expressed as

--exp
CPf

1 Equation (5) can then be used to eliminate c,/c,[in

@,,R + %-R) F
RT q.
(17). By following the same procedures used in derivmg
(lo), a material balance equation for an anodic limiting
(10) reactant is obtained:
If the same assumptions are made about the side
reaction as made by Trainham and Newman[ 1 l] (ie, v+(Da+D )=*R
R dx=
mass transfer resistance is unimportant and no explicit
dependence of the rate on composition), an equation
can be derived for the local overpotential:
-=
d’v --a

Consequently, the charge balance equation becomes

R- + 1
k%R %R)F,,

I 13
CRT RT

kmRcRj
IIF 1 (IX.R +&F,,
- --exp
SR'oR,d [
GRF
RTq
1
+rexp
l&=Pf RT
(11)

Equations (10) and (11) can be expressed in dimen-

sionless form by introducing the dimensionless coor-
dinate y = ak,,,x/v, the dimensionless concentration 0
= c,/c,,, and the dimensionless overpotential $: (19)
Equations (18) and (19) can now be put into forms
‘tFk&Rf identical to (13) and (14). This is accomplished by
exp(rj’) = -Y exp(%RFdRT) (12)
sR=oR,d introducing into (18) and (19) a dimensionless local
overpotential based on the anodic reactant-compare
Material balance ivith (12) for the cathodic reactant:
d0 nFk,c,/
_=&! expW3 = _ exP(--d,RFtl/RTh (20)
dy dy’ SRfoR,rsf
 A comparison between flow-through and flow-by porous electrodes 459

along with the dimensionless coordinate y and the

dimensionless concentration 8 to yield COUth8kClmde
c Scporata ol gap
Poraus clecinxle

xk-
Material balance CWlCt+d
LID
” v

t cunent cotlector
okschon of flW

P; exp( -Q&‘/Q)
w ”

IDI3 +
Direction of flow
Cuncnt
Y

UU

e- (I +e)T expC(1+a&,Rh*7 l

+ 1 + exph*?+ PIPd%pPC(l ++Jdtl*‘l ’

(22) DD
where the dimensionless parameters are defined as
bimc?ion of flow
(23)

Fig. 2. Various configurationsofcounterelectrode placement

and current collector placementrelativeto the direction of the
(24) fluidRow:(i) upstreamcounterelectrode,downstream current
collector (UD). (ii) downstream counterelectrode, upstream
current collector (DU), (iii) upstream counterelectrode, up-
stream current collector (VU), (iv) downstream counter-
electrode, downstream current collector (DD). UD and UU
are charging configurations, whereas DU and DD are dis-
charge configurations. (See Trainham and Newman[12].)
(261
These parameters correspond to the dimensionless which are the Danckwerts[ 143, yehner-Wilhelm[ 151
variables defined previously by Ttainham and conditions when axial diffusion and dispersion are
Newman[l l] for the cathodic reactant (eg P, for the included. The conditions on q’ and q+’ depend on
cathodic reactant corresponds to P; for the anodic electrode configuration and may be determined from
reactant). Ohm’s Law

Boundary conditions. (13) and (14) or (21) and (22) 64

-= d(% -%I = 4 I i2
(29)
can be solved simultaneously for 0 and q’, or 8 and q+‘, dx dx Q K’

respectively, four boundary conditions are required for

and are tabulated in Table 1. On charge, the conditions
each electrode configuration shown in Fig. 2. Far 0, the
for either the UD configuration or the UV con-
following conditions are applicable
figuration are applicable. On discharge, the conditions
for either the DU configuration or the DD con-
(27) figuration apply[12]. The dimensionless parameters
P, and P, for a cathodic reactant which arise in the
and analysis are related to Pz
d8
-=Oaty=aL p,= -- UP2
dy U+K

Table 1. Current and potential boundary conditions for various electrode configurations

Electrode
Configuration ii (0) i,(L) i2 (0)

UD 0 --i --i 0 -i/n i/a P5Xf -P,P

DU i 0 0 i i/x PJ’ -p,1*
uu i 0 -i 0 0 -P,If 0

DD 0 -i 0 i 0 0 &I*
460 JAMES A. TRAINHAMANDJOHN NEWMAN

and electrode, respectively; subject to the DU configuration

and the following restrictions:
P, = -~- Kp, (31) a. The feed concentrations for each electrode on
U+K
charge and discharge are the same as the product
so that concentrations on charge;
-Pz= p,+p,. b. The superficial velocity on discharge u1 = u&t,,,
(32)
where u,,+ = u, _ and t,/t, = 1 in the calculations
For an anodic reactant, the conditions on q*’ are of the which follow;
same form as those given in Table 1 for q’, except that c. The discharge current density was arbitrarily set
P5 and P, should be replaced by P; and Pk, which are at I, = 0.95 I,. (Note that 1; and I, are negative
defined as for the cathode and positive for the anode);
up’, 6. The cell voltage on discharge is then calculated
p;=--
UfK) from the results of 5, which again includes the ohmic
drop in the gap between the electrodes.
and
From the above results the power on charge and
KP; discharge are calculated: B, = I,V, and P, = idVd. In
p;= -~
U+K’ the optimization procedure which follows in a later
section I,, id, and v are optimized with respect to an
so that objective function appropriate for this system. 4 is
-p; = p;+pk. (35) optimized with the condition that cell operation on
discharge is at the power maximum. This choice
follows from the optimization because energy-is lost
Cycling procedure. A calculation procedure is due to mixing in the gap between the electrodes by (1).
needed so that system behavior can be predicted over a
complete cycle (ie, charge and discharge). The follow-
ing procedure was used for tHe flow-through
configuration. FLOW-BY POROUS ELECTRODES

For a flow-by porous electrode system, the macro-

Table 2. Variables set on charge and conditions used in the
calculations. + = positive electrode, - = negative electrode, scopic variations in concentration, potential, and cur-
c = charge, d = discharge rent are two-dimensional. Therefore, rigorous pre-
diction of the behavior of the flow-by system requires
Variables Conditions the solution of coupled, elliptic partial differential
set on used in equations. Alkire and Ng[16, 171 have modeled the
charge calculations cylindrical geometry of the flow-by configutition.
Their model, which was specifically applied to
L +,L- Electrode lengths for the po- I., =L_ metal-ion removal, included two-dimensional con-
sitive electrode and the ne-
centration variations in the absence of axial diffusion
gative electrode
and dispersion, and complete Butler--Volmer kinetics.
Ccrr c’, Superficial velocities for the “+ = u-, The simplifying assumptions made by Alkire and
positive electrode and the Ng[l6, 173 reduced the system of elliptic partial
negative electrode
differential equations to one parabolic partial differen-
Current density on charge I, -z 0 for tial equation for the concentration and one ordinary
cathode differential equation for the potential.
Concentration of reactants P+< = c-r However, for an economic comparison between the
on charge in the feed tank flow-by system and the flow-through system, a simp-
lified “engineering” model of the flow-by system was
Concentration of products c +d = L’_d
on charge in the feed tank at equilibrium sufficiently complex. The engineering model developed
with c +c and for the flow-by porous battery included the following
c-< characteristics:

1. Each electrode exhibits linear polarization

behavior; this is typical of a “practical” redox system
1. The variables listed in Table 2 were set; because the exchange current densities are large.
2. Equations (21) and (22) were solved for the 2. The positive electrode and the negative electrode
positive electrode (subject to the UD configuration are both oflength L (in the direction of fluid flow) with
boundary conditions); isotropic porosities E+ and E_ and specific surface
3. Equations (13)and (14) were then solved for the areas a, and a_ which remain constant in time; for a
negative electrode (subject to the UD configuration redox system no morphological changes occur in the
boundary conditions); electrode material.
4. The cell voltage on charge was then evaluated from 3. The concentration variations in the direction of
the results of 2 and 3, which includes the ohmic drop in current flow are negligible.
the gap between the electrodes given by A@,, 4. The variation of potential in direction of fluid flow
= &&,!K; is negligible; this will be a good approximation except
5. Equations (21) and (22) and then (13)and (14) were near the fluid inlet.
solved for the positive electrode and the negative 5. The hydrodynamics are characterized by super-
 A comparison between flow-through and flow-by porous electrodes 461

fluid flow
Y-
Equation (37) has been written for a positive value of

). membrane i i on discharge and a negative value of i on charge. In

addition, i varies at every cross section in the direction
of fluid flow. The first term and the second term in
brackets represent the total potential loss in each
electrode in the direction of current flow, a com-
bination of surface overpotential for linear kin-
etics and ohmi potential drop (Newman and
Tiedemann[13]). The last term in brackets represents
the ohmic potential drop across the separator.
In equation (37), Y’ is the dimensionless exchange
current density for an electrode (Newman and
Tiedemann[ J 33 )

v2 = (c!,+u<)--- (38)

It is assumed to have a composition dependence of the

form

v- Y.
where i, rrr is the value of the exchange current density
Fig. 3. Flow redox cell with current flow perpendicular to the at a reference composition ci._r. It is further assumed
hid flow. that the transfer coefficients GL,and a, sum to n (in this
ficial velocities v, and v- and average mass-transfer case n = 1) for an electrode reaction of the form
coefficients k, + and km _ . 1 sjMf’ + ne-,
6. Only one electrode reaction occurs at each
electrode.
and that the exponents y, have the dependence
The validity of some ofthe assumptions made above
cannot be tested without
development
experimental
of more sophisticated
data and the
models. However,
yi = qi +ssi
n
the resdts obtained illustrate the important features such that qi = -si = 1 for a cathodic reactant and zero
that would be observed with a more sophisticated otherwise. For a,, = u, = 0.5, (38) becomes
approach. It is hoped that this work will stimulate
more research on the flow-by system which has
potentially significant industrial importance.
Anulysis. Figure 3 shows a schematic of the flow-
by configuration to be modeled. The cell potential V is where c.,._$a= been set equal to ci, (theupstream feed
defined to be the difference between the positive and concentration of the oxidized and reduced species of a
negative current collectors redox couple).
V=@,l+(d+)-@l_(d-), (36) Material balances. Continuity equations for re-
actant and product species can be used to describe the
which may be decomposed:
variation of concentration and current density through
RT co, the porous electrode. In addition, the electrode length
--nln-_, may be calculated for set values of the cell voltage and
CR*
utilization of the reactants. From these results, the
open-circuit potential power densities on charge and discharge can be
obtained.
The above calculations are accomplished by relating
the unknown wall concentration in (37) and (42) to the
bulk concentration of the oxidized species for the
positive electrode. The wall concentrations c,, and the
bulk concentrations ci are then linked through the
limiting current density relationship

civ = ci(l - i/iL), (43)

where
i, = F adk,c, (44)
surface overpotential and ohmic-potential drop is the limiting current density in the direction of
current flow, which follows from Faraday’s law
(Newman and Tiedemann[ 133):
where 0 and R are the oxidized and reduced species of
the redox couple and subscripts + and - denote the
positive and negative electrode.
462 [Link]

or neglect COST ANALYSIS

(46) An objective function must be specified in order to

compare the performance of the flow-by and flow-
with i,.(d) = i,, i,,(o) = 0, and si = - 1. through configuration on a sound economic basis.
The bulk and wall concentrations of thereduced and Such an objective function should relate the overall
oxidized species in each electrode can be related costs (capital, fixed, and variable costs) for an entire
through the pore-wall fluxes, since these fluxes are system to local variables, such as current density, cell
equal and opposite for a redox reaction: voltage, and flow rate. However, for the comparison
considered here, only the costs and system variables
k,,(c, - co,) = - &Z(c, -c&J. (47) related to the net specific power of a particular
Equations (43) and (47) can be combined to yield the configuration are important. One objective function
reactant-wall concentrations in terms of the bulk will be based solely on the variable costs. The termi-
concentrations and the limiting current nology “net specific power” refers to the amount of
power in watts per square meter recovered during the
day, over and above that required to pay for the
(48) nighttime power. In general, the objective function
should be constructed in terms of intensive variables,
The local bulk concentration of the reduced species
that is, independent of the separator or gap area
is given by (membrane area for the flow-by configuration; super-
CR = cof+ cw.- co, (49) ficial-flow area for the flow-through configuration).
The simplified cost analysis presented here allows a
provided that diffusion in the x-direction is neglected
quick evaluation of the two different electrode
(or D, = D,).
configurations. However, it must be emphasized that
Equations (47)-(49) relate the wall concentrations of construction of large scale systems should be based on
the reduced and oxidized species to the bulk con- more detailed cost and performance evaluations.
centration of the oxidized species in each electrode.
The differential material balance for a species i may Capital costs andfixed costs. In the study consid-
be simplified (owing to the assumptions stated ered here, the exact functional form for these costs is
previously): not important because it is assumed that the energy
neglect neglect storage facilities are built and are operational.
However, for the sake of completeness an approximate
“Q+$)2#+sY). (50) formulation will be given (see Newman[lIJ) for the
case of variable, capital, and fixed costs.
An integration over y with i,,(d) = - i and i,,(o) = 0, The capital costs and the fixed costs in $/year-m’
gives for each configuration are assumed to have one value:

de. Flow-by
vd>= -?-i
(51)
dx nF * C :F = c(C,++CB+~+)+(CA_ +C,_d_)]r+C;,
It follows that for each electrode on discharge (56)

Flow-through
(52)
Cc, = [(CA+ +CB+L+)+(C,.- +C,_L_)]r+C;,
or (57)
d+v+ where C,, and CAi are the total costs per unit volume
cR- = cRf- +d-v_@Of+-CO+). (53) ($/m3) and total costs per superficial-electrode area
(!$/m”), respectively. The factor P (year-‘) represents
With the use of (43), (44), (48), (49), and (53), all the the effect of interest, depreciation, and taxes. In
unknown compositions in (37) and (42) can be related addition, r should reflect the average lifetime of the
to the differential material balance for the oxidized components in the system. If the expected lifetime of
species in the positive electrode (51). Either the elec- the components depends significantly on the current
trode length can be calculated for a set superficial density or cell voltage, this simplified treatment of
velocity and reactant utilization or the superficial capital costs would have to be modified.
velocity for a set electrode length and reactant The terms C; in (56) and (57) represent the fixed
utilization costs (ie, labor, materials, and maintenance). The fixed
costs are sometimes taken to be a fraction of the total
r~f+ dcb, capital investment plus the cost of labor. Such an
L=d+v+F - (54) assumption requires that the labor costs can be scaled
s CO+ i
with the superficial+Jectrode area.
or
Variable costs. Thecosts to be emphasized here are
L the power costs on charge and discharge, and the
V += (551 pumping costs. The variable costs also include materi-
d+F als makeup plus waste treatment, which may or may
not be functions of the operating conditions. In any
 A comparison between flow-through and flow-by porous electrodes 463

case, the relationships are not known, and it will optimized separately, and will not be considered
therefore be assumed that these additional costs are further in this analysis.
negligible. Dimensional considerations indicate that only ratios
The variable costs ($/year-m2) are given by of costs are needed to determine the optimum
Flow-by design and operating conditions. For redox energy
storage applications, only the ratio of the cost of
c;s = I,V$,SC,, - I,V@C,D daytime and nighttime power is important.
charge discharge Consequently, the objective function to be maximized
can now be expressed as

NSP(W/m’) = Pd -:CP,, (62)

+ Qg (“l,f,CpN + v2,t,cJ
1
, (58) where C = C,,/C,,
pressure drop
Flow-through OPTIMIZATION PROCEDURES AND
RESULTS
C;S = t,vct,SC,N - ldVctdSCpD
charge discharge To simplify the optimization, only the case where
V, = v, and t, = td will be considered. It is also
convenient to equate the superficial velocities in each
+-4;s y (u:,t,CpN + u:JdCpD)
electrode ucc = u-,. With these assumptions, (62)
P [
becomes
NSP(Wjm’) = P_,- CP,. (63)
The method of optimization is different for each
The only difference between (58) and (59) is that the configuration. For the flow-through configuration, the
length of electrode over which the pressure drop net specific power is maximized over discharge current,
occurs is explicit for the flow-through configuration charging current, and Aow rate. For the flow-by
and implicit for the flow-by configuration. configuration, it was only necessary to optimize the net
In equation (SS), the electrode thicknesses (per- specific power with respect to the cell voltage on
pendicular to the direction of fluid flow) d + and d _ discharge in order to obtain a valid comparison with
account for the fact that the volumetric flow rate may the results obtained for the Bow-through con-
be different for each electrode. figuration.
The factor S is the on-stream factor (either days It was not necessary to optimize the flow-by con-
operating per 365 days or operating cycles per year). figuration with respect to flow rate because the pres-
In equations (58) and (59), the pressure drop cor- sure drop is ignored in the objective function.
relation used is for Re = v/uv < 1 (see Bird et a/.[ 191). Furthermore, the flow-by configuration can operate at
This is typical of the flow rates encountered in any velocity with a corresponding change in electrode
the cell stacks at NASA-Lewis for the flow-by- length. Also, because the efIIuent is isolated and side
configuration[9] and, as will be seen later, the op- reactions are ignored, the number of coulombs passed
timum flow rates for the flow-through configuration on charge and discharge must balance, whereas with
typically give Re < 0.001. the flow-through configuration, extra charge is lost
As can be seen in (56 j(59), there is no optimum according to (1).
value for electrode length. An order of magnitude
analysis of the terms in these equations indicates that Chemical system. The results presented are for the
practical systems can be built so that the pressure drop FeeTi redox system. The Fe-Ti chemical system was
terms are small compared to the remaining terms. In chosen because data on fundamental parameters, such
the case of the flow-by configuration, the electrode as exchange current densities, solution conductivities,
length is determined by the desired conversion and the and viscosities were available (Giner et a/_[E] and
minimum flow rate which can be easily controlled in Millerf20], see Table 3). However, the values of the
practice. Therefore, the variable costs for either eon- exchange current densities were not available on the
figuration can be exressed as electrode materials used in the cells at NASA-Lewis for
C&S = P&K,, - P,fJCpD, (60) which a power density of approximately 237 W/m*[9]
was achieved for the Fe-Ti system.
where P, = V,I, and Pd = VJ,, The primary electrochemical reactions were as-
The total cost of a redox system can now be given by sumed to be

= = [Cc, + P&C,, - PdtrSCpD]A,,, , (61) FeCl z + FeCI: t-e-, (64)

year
and
where A,,, is the total superficial electrode area of the
Ti’* + 2H,O + Ti(OH);+ +2H+ +e-. (65)
system.
For the comparison considered here, only the vari- The species assumed in these reactions are based on
able costs CpNand C,,, and the system variables related stability data (Baes and Mesmer[21]), experimental
to these costs P,, Pd, t,, and t,, are important. The measurements of open-circuit potentials (Reid and
capital and fixed costs per unit superficial area can be Gahn[22]), and secondary reference state quantities
464 JAMES A. TRAINHAM AND JOHN NEWMAN

0.9
(Wagman[23]). For Ti(JV) the hydrolyzed species is
expressed Ti(OH):+, as there is no strong evidence
0.8
supporting the commonly written “titanyl” ion, TiO’+
(Baes and Mesmer[Z 11). 0.7
The flow-through model has provision for side
reactions, and these were assumed to be evofution of 0.6
hydrogen and chlorine
0.5
2H+ +2e- -+ HZ, (66)
and 0.4

2c1- * Cl,(g) + 2.K. (47) 0.3 0

I 5 9 13 17 20
Flow-through As mentioned
configuralion.
previously, this configuration was optimized with I,. mA/cm-
respect to discharge current density, charging current
density, and flow rate, with (63) as a criterion. To
simplify the calculational procedure, and also to Fig. 5. Effect of charging current density on the reciprocal of
reduce computer costs, the charging current density I, the superficial velocity.
was set and the value of C (cost of nighttime to
daytime power) was eventually calculated such that the velocity v if all the reactant in the feed were completely
set value of I, was an optimum. This procedure has the reacted) as shown in Fig. 4. I: is really a dimensionless
advantage, first of all, that the discharge current substitute for v because I,, PI,F, and cRI are set. In
density can be optimized separately--operation is to Figure 4, the point of intersection of the two curves
he at the power maximum independent of the value of represents the value of the cost ratio for which I,, I,,
C. Therefore, the search is one-dimensional rather than and v are optimum. Figure 4 also shows that (68) is
three-dimensional. more dependent on u than is (69).
To be more specific, two equations result for the cost Figure 5 shows that for an increase in optimum flow
ratio C when (63) is differentiated with respect to u at rate by a factor of 8.1, I, increases by a factor of 20. This
constant I, and then with respect to I, at constant u, implies a greater utilization of the reactant species on
setting the results equal to zero charge at higher current densities. It should be noted
that the optimum superficial velocity varied from
(68) (0.961 to 7.78) 10ms cm/s over the charging range
considered. However, these low velocities would prob-
ably be difficult to attain in practice.
(69 Figure 6 shows that the cost ratio decreases with
current density. Therefore, the higher the cost of
The procedure is to set I, and guess a value for II; daytime power relative to nighttime power, the higher
calculate the derivatives in (68) and (69); while simul- the current density at which the system can be
taneously adjusting thecurrent density on discharge so economically charged.
that cell operation is at the power maximum. The As might be expected from the previous graphs, Fig.
procedure is repeated at a set I, adjusting u until a plot 7 shows that the ratio of the optimum current density
can be made of (68) and (69) us 1: = f,/nFvcRf (the recovered on discharge to that of the required charging
current density on charge made dimensionless with the current density, decreases with charging current
limiting current density that would exist for fluid density. Thus, the current efficiency is low. Also, a low
value of C results in a low energy efficiency.
0.47 The decrease in the energy efficiency with current
I I I I
density on charge is clearly shown in Fig. 8. At I,
= 1 mA/cms, PJP, is 0.6, whereas at I, = 20 mA/cm2,
0.46 W,
\
P,JP, drops to 0.27.
\
\
0.45 -
\
\
0.6
‘\
0.44 - t- \ I
C 0.5
‘\\
\
0.43 - \
\ 0.4
\
\ C
\
0.42 - \
0.3
‘\
\
\
0.41 - \ 0.2

0.1 p’;Ti
5 9 13 17 20
0.40
0.58 0.59 0.60 0.61 0.62 0.63
I,, mA/cm’

Fig. 6. Effect of the charging current density on the cast ratio

Fig. 4. Computer-aided graphical optimization procedure. (nighttime to daytime power COSU).
 A comparison between flow-through and flow-by porous electrodes 465

0.65

0.55 -
I 5 9 13 17 20

I,. mA/cm- 0. I 0.2 0.3 0.4 0.5 0.6

C
Fig. 7. Ratio of the optimum discharge current density to the
charging current density versus the charging current density. Fig. 9. Net specific power recovered by the Row-through
configuration US the cost ratio (nighttime to daytime
The NSP (W/m’) is shown in Fig. 9 as a function of power).
C. The NSP is much higher at low values of C.

Flow-by configuration. As mentioned previously, Equation (72) is analogous to (68) and (69) used for
this configuration was optimized only with respect to optimization of the flow-through reactor.
the-cell voltage on discharge. Again, optimization of Figure 10 shows a comparison between the flow-
the flow rate is unnecessary because the pressure drop through and flow-by configurations. Except for very
contribution to the objective function is not included. high reactant utilization on discharge (c 90 per cent),
In order to compare the net specific power of the two the flow-by configuration yields the greater return on
systems, similar results as were obtained in Fig. 9 are investment.
required. The values of V,, V, and coL are set, and the In Fig. 11, the effect of increasing the specific
electrode length is calculated from (37) and (54). Once interfacial area and decreasing the electrode thickness
L is known, a value of V, that will recharge the cell to is shown for the flow-by reactor. An increase in a by a
the initial concentrations can be calculated. The total
current density, and consequently the power, for either
charge or discharge can be obtained from
I, = n+FMcR,--RL)+lL (70)
and
4 = n+F4co,--c,L)+lL- (71)
A value for C is then determined for which V, is an 10

optimum from the equation

C = (aP,/a v,)l(aP,l~ V,), (72)
and finally the net specific power is calculated with (63). “6
5
i
”

z
I

I I I I\
0
I 5 9 I3 17 20
Fig. 10. A comparison between the flow-by configuration and
I,, mA/cm’
the flow-through configuration based on the net specific
power recovered. The curve for the flow-through con-
Fig. 8. Dependence of the power required on charge and figuration is the same as that shown in Fig. 9. flow-by
power recovered on discharge as a function of the charging are the feed compositions shown at
current density. various utilizations of reactant.
466 JAMES A. TRAINHAM AND JOHNNEWMAN

zoo*
Table 3. Values of the parameters wed in generating
I
1 I I the
cFe"',L/%e"',,=0.95 _ results in Figs 4-l 1
Y = 1.7x IO-'..,. -
Parameters used in both electrode systems

a = 25cm-‘, e =0.3Do+,~~ = DO,Ti(OH);+

= 5 x 10-bcm2/s
PI= 1, T= 298.15K, Do FcCI,= 5.5 x 10-6cm2/s[26]
Do,Fd3,; = 5.7 x 10-6cm2/s[26], a, = a, = OS*
FKuFK,~ =0.6513V[22, 231, U&B.~~,~~);. so.1 V[21,273
u@

Flow-through configuration
0.2 -
~~ = 0.25 mhoJcm, CT= 2 x lOI4 n&o/cm, L = km,
0.1b I I 1 L =lcm
0.I 0.3 0.5 0.7 0.Q
on charge: cHg,P=SM
C
%(oH):‘J = %clt, f = 3 M
+a+,, = crea;,l= 3.2942 x lo-4M
Fig. 11. Dependence of the net specific power on the cost 2C1- + C&(g) +2e- : i,,s = lo-‘c,-,[28] Q = 0.38,[28]
ratio for the flow-byconfigurationwith the specificinterfacial
aes = 0.69[28]
area and the electrode thickness as parameters.
ca-,f = 20A4, u&,a- = 1.3595[13]

factor of ten yields an increase in the performance by a H2 --t 2HC + 2e- : id = 6.708 x lo-‘“A/cm2,[28]
factor of three. Similar results can be expected for the =,s = =,s = 0.5[28]
flow-through configuration. 0.5454
WI

DISCUSSION
Flow-by co@iguration
The results of the optimization have shown several K._ = 0.207 mho/cm[E]
K.+ = 0.176 mho/cm[KJ,
important differences between the two configurations. d, = d_ = 3cm, d, = d, = 0.5cmt, L, =O.O47cm[24]
However, these results have raised several pertinent
u=1.7xlO-scm/s, l/K, = 25mho/cm[24]
questions:
cm discharge:
1. Were the parameters used in the calculations
cTia+,, =c~+-,;+,~= 2.7 M
equivalent for both systems, and if so, were values
representative of actual systems? %lCOH):* ./= cFCl,,f'o.27 M
2. Why is the performance of the flow-through
configurations so much poorer? “EL&= 0.91 &?(-$)o.19~$L)“3[19]
3. Why is the optimum flow rate for the flow-through
$ = 0.86, Y = 0.0251 cm2/s
configuration so small?
4. Is there perhaps a better configuration coupled
with a different chemical system for which the flow- l Ail primary reactions.
through electrodes would yield competitive results?
+ Fig. 11 only.
5. Under the conditions of the calculations, is the
assumption of linear kinetics valid for the Bow-by
model? electrodes on mixing after passing through the elec-
6. What determines the flow-by electrode thickness? trodes on discharge
We will endeavour to answer these questions.
Ti3+ + FeCl: + Hz0 * Ti(OH): + + FeCl, + 2H +,
Parameters. In order to obtain a valid comparison (73)
between the two configurations, the parameter values
which has an equilibrium constant
were shared whenever possible (see Table 3). However,
the differences need some clarification. The total iron 'Ti'*'FcCl:
concentration and titanium concentration were each = 4.823 x lo-*’ --& ‘. (74)
3.00 M for the flow-through configuration and 2.97 M cTi(OH):*cFcCl,c~* ( >
for the flow-by configuration. In equation (74), the activity coefficients and the
In the flow-through case, the concentrations of Ti’ + activity of water have been set equal to unity.
and FeCl: in the feed (set equal to 3.2942 x NJ-’ M, The results of the optimization show that the charge
seeTable 3) were determined from the equilibrium for imbalance due to incomplete conversion of FeCl; on
the reaction which occurs in the gap between the discharge is very small. In fact, the percent of charge
 A comparison between flow-through and flow-by porous electrodes 467

(stored on discharge) required to restore the system to copper removal process of Bennion and Newman[25].
the initial state of charge (ie, all FeClg converted to This contributed substantally to the two-orders of
FeCl,), was always less than 0.05 per cent. Therefore, magnitude difference in flow rate.
the side reactions on charge are extremely small. These 3. A comparison of Fig. 8 with Figs 9 and 10 shows
resultsreflect the optimization procedure, which forces that the system cannot be run economically at high
the system to operate at nearly 100 per cent current current densities.
efficiency on charge.
An estimate of the maximum allowable velocity on
The correlation used in predicting the mass transfer
charge can bc obtained from the limiting current
coefficient for the flow-through configuration is the
design equations of Bennion and Newman[25]. The
same on that determined by Trainham and
difference in the standard electrode potentials for the
Newman[l 11. For the flow-by configuration, a mass
system is AU’ = 0.5513 V (see Table 3). If we assume
transfer correlation appropriate for the Reynolds
0.0513 V is lost to the separator, then there is approxi-
numbers encountered was used[19].
mately 0.2 V available for ohmic losses in each
Other variations in the parameters [Link] to the
electrode. If we combine the equation for the mass
inherent differences between the two systems. For
transfer coefficient given in Table 3 for the flow-
example, the superficial velocity for the flow-by system
through configuration[ 111 with the expression for the
(a = 1.7 x 10-j cm/s) was based on some experiments
ohmic drop across an electrode[25]
at NASA-Lewis[9], whereas an optimum value was
determined on charge for the other system. Experi- nFv* ca,
A0’2=--- (76)
mental conductivity and viscosity data were available ak,s,gc ’
for the mixtures FeCIJFeCl~ and T?+/Ti(OH)i+
encountered in typical flow-by systems[8], but were an expression for the velocity results
not available for the Fe/T? mixtures of the flow- SRKAWZa1.4546p.4546.
u’JS46 = 0.07054 (77)
through system. In the latter case, a value of
snFcsf
0.25 mho/cm was assumed (see Table 3). Experimental
data for the membrane separator thickness and re- Using the values of the parameters in Table 3 and Alo;
sistivity were also known for the flow-by system = 0.2 V, we obtain u = 2.1 x lo-’ cm/s. This estimate
(Alexander and Hodgdon[24]), where a 1 cm gap was is the same order of magnitude as the velocities
assumed in the flow-through calculations. obtained by the optimization procedure.
It can be concluded from the above discussion that As mentioned previously, the results for the flow-by
the differences between the parameters used in each configuration operating at a fixed conversion are
system are minor and, therefore, will not account for essentially independent of velocity. For a desired
the discrepancy in performance. conversion, the electrode length is set, and the velocity
As mentioned previously, values for the kinetic can be calculated or vice versa. This operational
parameters and the specific interfacial area were not flexibility makes the flow-by configuration superior.
available for the system at NASA-Lewis for which
a discharge power density of 237 W/m2[9] was Effects of electrode thickness. For the flow-by
achieved. This compares with a calculated maximum configuration, results were obtained for electrodes of
power density of 170 W/m’ (top curve Fig. 11). A value thickness 3 cm and 0.5 cm in the direction of current
of 25 cm- ’ for the specific interfacial area was used in flow. Figure 11 shows that there is a small improve-
most of the calculations. However, Fig. 11 has shown ment in the performance of the 3 cm thick electrode
that for a tenfold increase in the value of a, the NSP over the 0.5 cm thick electrode. The length of the 3 cm
increases by a factor of three. Even a value of a equal to thick electrode varied from approximately 5.2 to 5.5
250 cm- ’ could be conservative by a factor of two or times that of the 0.5 cm thick electrode over the
three, and this would account for most of the discharge range considered. These results imply that
discrepancy. In addition, the values of the exchange for equal capacity systems (ie systems capable of
current densities used in the calculations (see Table 3) processing the same volume of reactants per unit time)
could easily be low by an order of magnitude. At the a thicker electrode is more efficient and the total
present time, there is no way to tell how well the models volume of the electrode material would be less.
predict reality, because there are no steady-state data However, because the flow-by model assumes no
available. concentration variations across the thickness of the
electrode, this claim can only be justified by a more
Superficial velocity. In Fig. 5, the optimum super- sophisticated approach. Thus, the important design
ficial velocity for the flow-through system varied from question which cannot be answered here is: what limits
(0.961 to 7.78) lo-’ cm/s over the charging range of the flow-by electrode thickness?
1-20mA/cm2. Again, these velocities would be
extremely difficult to achieve in practice.
Linear kinetics. In the development of the flow-by
The optimum velocity is low for several reasons:
model, it was assumed that linear polarization was
1. The conversion is moderate on both charge appropriate owing to the high exchange current
(0.11 < c,&E~ < 0.64) and discharge (0.15 -K densities inherent in these systems. The assumption of
c&c,, < 0.6) for the charging current density range of linear kinetics can be tested using the numerical results
1 mA/cm’ to 20 mA/cm*. A low conversion, which generated from the model.
would allow a much higher flow rate, is not possible The criterion for testing the validity of the linear
due to the single pass operation. polarization assumption depends on the electrode
2. The feed composition of the reactants on charge is thickness. For thick electrodes (ie thicker than the
3 M. This is 300 times the feed composition of the effective depth of penetration of the reaction into the
 468 JAMES A. TRAMHAM AND JOHN NEWMAN

electrode) the linear polarization assumption is valid if 2. Lawrence H. Thaller, “Electrically rechargable redox flow
cells.” U.S. Patent 3996,064 (1976).
1 Klaus D. Beccu, Ch4mnip IngenCur Technik 46.95 (1974).
i-,1, (78) :: Walther Kangro, “Verfahren zur Speicherung van elek-
L/v ai
trischer Energie.” Deutsches Patentamt 914.264 (1954).
where i is the average transfer current density and L/v 5. Walther Kangro, “Couples redox liquides.” Space Power
is the effective penetration depth given by [13] Systems Energy Storage. Proc. Sixth ESRO Summer
[Link] Nerherlands (1968): Vol. X, ESRO

LJ~
= &zzzzj (79)
SP-49 (1969).
6. N. P. Yao and J. R. Birk, “Battery energy storage for
utility load leveling and electric vehicles: a review of
If the electrode is thinner than the penetration depth, advanced secondary batteries.” IECEC Record, p. 1107
the actual electrode thickness d should be substituted (1975).
for LJv. 7. Marvin Warshay and Lyle Wright. ‘%ost and size
estimates for a redox bulk energy storage concept.” J
An analysis of the results generated by the model elecrtochem. Ser. 124, 173 (1977).
using (78) and (79) show that for the results presented, 8. 3. Giner, L. Swette and K. Cahill, “Screening of rcdox
the linear polarization assumption is valid. couples and electrode materials.” (Giner Inc.; NAS 3-
19760), NASA CR-134705 (1976).
Other systems. It should be kept in mind that the 9. Gahn Randall and Norman Hagedom. Unpublished
Fe-Z chemical system has been used for a relative results, NASA-Lewis Research Center, Cleveland,
comparison between two electrode configurations. Ohio, May 1978.
Other chemical systems offer larger open-circuit pot- 10. John Newman, ‘Simultaneous reactions at disk and
entials and reaction rates per unit volume (ai,). porous electrodes.” Electrochimica Acta 22, 903 (1977).
11. James A. Trainham and John Newman, “A flow-through
However, the mathematical models and the optimi-
norous electrode model: annhcation to metal-ion re-
zation procedures developed here can be used directly moval from dilute streams:” J. electrochem. Sot. 124,
or modified accordingly, so that other systems can be 1528 (1977b).
compared. 12. James A. Trainham and John Newman, “The effect of
electrode placement and finite matrix conductivity on the
performance of flow-through porous electrodes.” J.
SUMMARY AND CONCLUSIONS elertrot,hem. Ser. 125, 58 (1978).
13. John Newman and William Tiedemann. “Porous elec-
Theoretical models have been developed for two trode theory with battery applications.” em. Inst. them.
Engrs .I. 21, 25 (1975).
electrode configurations for redox battery appli-
14. P. V. Danckwerts, “Continuous fiow systems. Distri-
cations: 1 flow-through porous electrodes and 2 flow- bution of residencetimes.” Chem. Engng Sci. 2, 1 (1953).
by porous electrodes. These models were used to 15. J. F. Wehner and R. H. Wilhelm. “Boundary conditions of
determine which system yields the greater return on a flow reactor.” Chem. Engng Sri. 6, 89 (1956):
investment. To do this, both systems were optimized 16. Richard Alkire and Patrick K. Ng, “Two-dimensional
with respect to an objective function appropriate for current distributionwithin apacked-bed electrochemical
the application. flow reactor.” J. elecrrorhem. Sot. 121, 95 (1974).
The results show that the flow-by configuration is 17. Richard Alkire and Patrick K. Ng, “Studies on flow-by
superior. Not only does the flow-by system yield a porous electrodes having perpendicular directions of
current and electrolyte flow.” J. olectrochem. Sot. 124,
greater return on investment, but it also offers the
1220 (1977).
operational flexibility of variable flow rate and 18. John Newman. “Optimization of potential and hydrogen
conversion. Its limitations compared to the flow- utilization in an acid fuel cell.” Elecrrochimicn Acra 24,
through system will be associated with the mainten- 223 (1979)
ance and reliability of the membrane separator. 19. R. Byron Bird. Warren E. Stewart and Edwin N.
The flow-through configuration in addition to yield- Lightfoot, Transport Phenuneno, pp. 4ll-~412. Wiley,
ing a lower return on investment, offers an impractical New York. (1960).
superficial velocity (the optimum superficial velocity 20. Riley 0. Milier, “Electrochemical Behavior of 0.2 to 3
was of order IO-’ cm/s). Its failure is due to current Molar Ferrous Chloride-Ferric Chloride mixtures on
Pyrolytic Graphite Rated Disk
flow (and ohmic potential drop) in the same direction Electrodes.”
ERDA/NASA-5022/77/2, NASA TM 73716 (1977).
as the fluid flow.
21 C.F. Baes, Jr. and R. E. Mesmer, The Hydrolysis oj
The perspective here has been to use a simplified Cations. John Wiley, New York (1976).
optimization procedure to obtain a valid compa- 22. Margaret A. Reid and Randall F. Gahn, “Factors Affect-
rison between two electrode systems. ing the Open Circuit Voltage and Electrode Kinetics
of Some Iron/Titanium Redox Flow Cells.” ERDA-
Acknow/&rmenr-This work was supported by the Division 1022/77/10, NASA 7MX-73669 (1977).
of Solar, Geothermal, Electric and Storage Systems,Officeof 23. D. D. Wagman, W. H. Evans,V.B. Parker.1. Halow, S. M.
the Assistant Secretary for Energy Technology. U.S. Depart- Bailey and R. H. Schumm, “Selected Values of Chemical
ment of Energy under contract No. W-7405-Eng-48. Thermodynamic Properties. Tables for tee First Fifty-
Three Elements in Standard Order of Arrangement.”
NBS Technical Notes 270-3 and 2704. Washington
D.C., National Bureau of Standards (1968).
REFERENCES 24. S. S. Alexander and R. B. Hodgdon, “Anion Perme-
selective Membranes.” Ion&, Inc. Watertown, Mass.,
1. Lawrence H. Thaller. “Electrically rechargable redox flow CONS/010801, NASAA CR-135316 (1978)
cells.” NASA TMX-7154. Ninrh lnrersoriety Energy 25. Douglas N. Bennion and John Newman, “Electro-
Conversion Engineering Conf: San Francisco, California, chemical removal of copper ions from very dilute
August (1974). solutions.” J. Appl. Elecrrorh~m. 2, 113 (1972).
 A comparison between flow-through and flow-by porous electrodes 469

26. B. G. Ateya and L. G. Austin, “The Kinetics of Eleventh edition, p. 6-14. McGraw-Hill, New York
Fe2*/FeCI:/HCI (aq.) on Pyrolytic Graphite (1973).
Electrodes.” J. olertrochem. Sot. 120, 1216 (1973). 28. Tibor Erdey-Gruz, Kinetics of Electrode Processes,
27. John A. Dean, (editor) Lange’s Handbook of Chemistry, p. 165. Wiley Interscience, New York (1972).