Solute Transport
7.2 Ionic current and transport and transference numbersnumbers
Ionic current and transport and transference numbers 165 important, and a certain change inΞπΎimportant to make whenπΎis low, i.e., a small amount of extra stirring then helps a lot, at some point the incremental effect of stirring is small and the energy costs thereof will surpass the advantage of more stirring. In addition, at some point of increasing stirring intensity, another mass transport limitation in the process will become rate-limiting.
A very accurate approximation to the Airy equation (with the given boundary conditions) for the profileπ(π₯Λ)is
π(π₯Λ) βπβ πββπβ
=1βtanh(πΌ π) (7.17)
whereπΌandπare the same as above.
In conclusion, in this section we presented four models to describe steady-state diffusion and dispersion near a reactive or selective surface for neutral solutes. For the relationship between flux and concentration, these models all end up with the same result that flux is proportional to the concentration difference between bulk and surface. However, for the proportionality factor different expressions are obtained, with a dependence on diffusion coefficient that can be to the power1/2,2/3, or 1.iii
7.2 Ionic current and transport and transference
166 Solute Transport
Eq. (7.2), extended with convection, which describes the molar flux of an ionπby π½π=πππ£
Fβπ·π π ππ
π π₯ +π§πππ
π π
π π₯
where π·π is the diffusion coefficient of the ion in the electrolyte phase, and where the potential π relates to a voltage π by π = π/π
T. In Β§7.8 we explain how this extended NP-equation derives from a balance of forces acting on ions in solution.
A summation over all ions of their fluxπ½πtimes their valencyπ§πresults for current density (unit A/m2) in
πΌ=πΉ
βοΈ
π
π§ππ½π =πΉ π£
F
βοΈ
π
π§πππβπΉ π
βοΈ
π
π·ππ§π
π ππ
π π₯
βπ πβ
ππ
π π₯ (7.18)
where we introduce the ionic conductivity of a solution (unit S/m, where S=A/V and A=C/m2)
πβ= πΉ2/π π
βοΈ
π
π§2
ππ·πππ (7.19)
also called the electrical conductivity of a solution, π . In bulk electrolyte we have local electroneutrality,Γ
ππ§πππ =0, and thus the convective term in Eq. (7.18) can be omitted, but that is not the case in a charged porous medium. We discuss ion conduction in a porous medium in the next box.
Conductivity in porous media, such as membranes. The conductivity of a porous medium where pores are filled with electrolyte, such as a flow channel that contains a mesh of material,π
m, is lower than the conductivity of bulk solution by a factorπ, thus πm=π πβ, see p. 159. The factorπis porosity, π, divided by the tortuosity factor,π, i.e.,π=π/π. For flow inside a nanoporous medium such as an ion-exchange membrane, ions also have friction with the structure, the matrix. In that case, the conductivity, πm is lower again, now by an additional factorπΎ
f,π that we discuss at p. 197. Thus πβ
m=πΎ
f,ππ
m =πΎ
f,ππ πβ.
An important topic is the electrolyte conductivity inside charged membranes, of special importance for electrodialysis (ED), see Ch. 12. Also here Eq. (7.19) can be directly applied, as conclusively proven by J.C. DΓaz and J. Kamcev1. We illustrate this fact by an extensive data set of DΓaz and Kamcev for an AEM with|π|=3.1 M and a CEM with|π|=2.5 M for membranes with several thicknesses, tested with NaCl solutions with salt concentration from 1 mM to 1 M, see Fig. 7.1A for results obtained with an
Ionic current and transport and transference numbers 167
AEM membrane. From the slope of a trendline running through the origin, we can derive a conductivity ofπβ
mβΌ16 mS/cm for salt concentrations 1β100 mM andπβ
mβΌ17 mS/m forπβ=1 M. Thus conductivity is independent of salt concentration, except for a slight increase around 1 M NaCl, see Fig. 7.1B. The constant conductance is in agreement with Eq. (7.19) given that at low enoughπβthe concentration of counterions in the membrane is almost equal to|π|and the concentration of coions is very low. The increase ofπβ
mat 1 M NaCl is due to the increase in co- and counterion concentrations in the membrane.
To describe these data, we extend Eq. (2.11) in Ch. 2 to account for a non-unity partition coefficient,Ξ¦π(which we use to describe non-electrostatic contributions to the chemical potential of an ion), resulting inπ2
T,m=π2+ (2Ξ¦ππβ)2, which relates the total ion concentration in the membrane,πT,m, to salt concentration and membrane charge.
The counterion concentration in the membrane is thenΒ½ π
T,m+ |π|
and the coion concentrationΒ½ π
T,mβ |π|
. These concentrations are then used in Eq. (7.19) which results in the theoretical lines in Fig. 7.1B (Ξ¦π=0.6). To fit to the data we introduce a membrane reduction factor, mrf, which describes by how much the rate of diffusion of ions is reduced in the membrane compared to solution, thus mrf=πβ/πβ
m=1/πΎ
f,ππ. For the AEM considered in Fig. 7.1B, we then have mrfβΌ15 while for the CEM we derive mrfβΌ9.5. Note that measurement ofπβ
mbased on conductivity (current-voltage relationship) is only correct when convection and diffusion can be neglected, see 1stand 2ndterm on the right in Eq. (7.18). For a charged membrane, convection is typically non-zero while diffusion is only zero if we can assume the sameπ·π for all ions.
Interestingly, a similar analysis by DΓaz and Kamcev of data for a commercial CEM membrane (Neosepta CMX) with|π| βΌ5.7 M, leads to a value ofπβ
mβΌ5 mS/cm while other literature reports 7-10 mS/cm.These numbers lead to a reduction factor between mrf=30β60, in line with estimates for the mrf of Neosepta CMX membranes in Tedesco et al.(2018).2
We summarize these data of mrf vs. |π| with the empirical function mrf = exp(πΌ
1|π|πΌ2) where|π| has the unit M, with πΌ
1=1.32 andπΌ
2 =0.6. (For π =0 M, mrf=1, i.e., no reduction in ion mobility. In this analysis, we use mrf=β
30Β·60βΌ42 for the Neosepta membrane.) Data points and the fit line are presented in Fig. 7.2.
A similar relationship between conductivity and membrane charge density is given by Fanet al. (2022)3 where effectively mrf is expressed as an exponential function of the ionic valency squared, and with membrane charge density to the power 2/3, i.e., the empirical function above is modified to mrfβexp π΄Β· |π|3/2
. The prefactor π΄is
168 Solute Transport
10 12 14 16 18
1 10 100 1000
0 0.5 1 1.5 2 2.5
0 100 200 300 400
membrane conductanceο³m(mS/cm)
salt concentration (mM) CEM, X=2.5 M
AEM, X=3.1 M
membrane areal resistance(οοcm2)
membrane thickness (οm) AEM, X=3.1 M
ο³m~17 οS/cm 1 M
1 mM 10 mM 100 mM
ο³m~16 οS/cm
Fig. 7.1:Resistance to ionic transport in an anion and cation exchange membrane (AEM and CEM).
A) Resistance as function of membrane thickness for an AEM. B) Ionic conductivity as function of external salt concentration. Data from DΓaz and Kamcev (2021), and lines based on Eq. (7.19) and Donnan model, as explained in the nearby box.
inversely proportional to the dielectric constant in the membraneπ
rsquared, and the equation in Fanet al. matches that by us ifπ
r=32 is used. Note that mrf as calculated by Fanet al. is extremely sensitive to the assumed value ofπrin the membrane.
In Fanet al. (2022)3 also an expression is provided for the influence of membrane porosity, or water fraction,p, on the diffusion coefficient in the membrane, which is mrf = (2/πβ1)2. This expression implies that when π=0.48, mrf=10, and when π=0.18, mrf=100, which are realistic predictions. However, it is not known whether this formula relates to an influence on diffusion itself (relating to porosity and tortuosity), or also to ion-membrane friction (resulting inπΎ
f,π<1). In Β§7.8 we discuss approaches to include ion-membrane friction in a transport model.
We can now define the transference number of an ion,π‘π, as π‘π =
π§2 ππ·πππ Γ
ππ§2
ππ·πππ
. (7.20)
The transference numbers,π‘π, are positive for all ions, and add up to unity, i.e.,Γ
ππ‘π=1. For a binaryπ§+:π§β salt solution (only one cation, one anion), the transference number of each
Ionic current and transport and transference numbers 169
0 10 20 30 40 50
0 1 2 3 4 5 6
membrane reduction factor
membrane charge density (M) mrf=exp(ο‘1ο|X| )
ο‘1=1.32, ο‘2=0.6
ο‘2
Fig. 7.2:Data for ionic conductivity of an ion-exchange membrane (IEM) recalculated to a membrane reduction factor, mrf, which is the ratio of ion diffusion coefficient in an IEM relative to that in free solution, plotted as function of membrane charge density,|π|. An empirical fit for mrf starts at mrf=1 atπ=0, and increases to above mrf=40 for the Neosepta CMX membrane (|π| βΌ5.7 M).
of the ions is
π‘π = |π§π|π·π
π§+π·++ |π§β|π·β (7.21)
and we haveπ‘++π‘β=1.
In the absence of concentration gradients we only have electromigration as a driving force for ion transport, and then implementing the above equations in Eq. (7.2) results for the ionic flux in
π½π =πππ£
Fβ π‘ππΌ π§ππΉ
. (7.22)
This is a very elegant equation, but it must be noted that it has a very small range of application: it is only valid in the complete absence of any concentration gradient, so in the boundary region near a selective surface it cannot be used.
Transference numbers in membranes. In this chapter we discuss transport in bulk electrolyte, where we have charge neutrality based on the ions. However, in a charged membrane, we also have the fixed charge groups, with concentration|π|. In a description of transference numbers, these groups are mathematically treated as if they are ions, with a diffusion coefficient of zero. Thus, also in a membrane we can use Eq. (7.20) to calculate transference numbers of all (mobile) ions. Transference numbers will then strongly depend on the concentrations of counterions and of coions, leading to values
170 Solute Transport
of π‘π close to zero or unity, even though the diffusion coefficients (mobilities) of the ions are similar. These numbers forπ‘π in the membrane via their dependence on ion concentration in the membrane, will then depend on the concentrations of ions outside the membrane (because of Donnan equilibrium). But transference numbers have no direct relation to the rates of transport of ions, as we will discuss next.
β
Besides the transference number, π‘π, there is the transport number, ππ, and this is a very different parameter. The transference number, π‘π, was defined by Eq. (7.20) and can be calculated when we know all ion valencies and diffusion coefficients. It describes the contribution of ions to the current only if the composition of the solution phase is uniform (all concentration gradients are zero). Instead, the transport number describes the actual or
βrealβ contribution of ion transport to the current, thus also when there are concentration gradients, which always is the case near a selective surface, such as near an electrode or membrane. Thus, the transport numberππis the local contribution of a certain ion flux to the local current, whether or not there are ion concentration gradients. The differences between π‘πandππcan be large, with the transference number,π‘π, in a binary salt solution always at the same value, see Eq. (7.21), but the transport number,ππ, changing in time and with position.
Only with vigorous stirring, and away from selective interfaces, do the two parameters have the same value. But in general they have different values.
Transport numbers can be used to describe a process where bulk electrolyte is in contact with a selective interface, with current running through this interface. This interface can be an electrode, absorbent material, or membrane. As just mentioned, the transport numbers, ππ, describe the contribution of each ion to the current density, and they are often position- dependent, and for a dynamic process also time-dependent. Only in a one-dimensional steady-state transport problem, without convection along the surface, i.e., without dispersion modelled as a βsidewaysβ refreshment, are they constant, i.e., independent of position. In other situations they vary with position. If the selective layer (for instance a membrane) operates in steady-state, with ion fluxes assumed to cross the membrane in one particular directionx, then inside the membrane the transport number of a certain ion will be the same at eachx-position, i.e.,ππ will be invariant withx.iv Note that this is only true for inert ions.
When we have reactive ions, transport numbers of individual ions will change with position, also for a membrane layer in steady state, see Ch. 10.
ivIn a full membrane module, with also a coordinatezdirected along the membrane, i.e., through the flow channel from entrance to exit, transport numbers will depend onz.
Ionic current and transport and transference numbers 171 So how to calculate transport numbers right next to a selective interface, such as a membrane? They are not material properties, or properties of the ionic solution, like the transference number is. They are not membrane properties in the same way that an ion mobility or a membrane charge density is. Instead, they follow from a combination of a transport model for the membrane, with a transport model for the flow channel, and they (i.e., their values) emerge in a calculation that considers the complete membrane-electrolyte system. In some cases we can assume the membrane (or other selective layer) to have a certain transport number for a certain ion, for instance we can assume that a certain membrane is perfectly selective and only allows access to a certain type of ion, so for this ionππ=1 and for all other ionsππ=0. For a reactive electrode where only a certain species is consumed, for that species the transport number is unity, while it is zero for all other species.v
The transport number is defined (at any position) as the fraction of the current density carried by a certain ion, i.e., it is the local ion flux timesπ§πover the local current density
ππ =π§ππ½π/π½
ch (7.23)
without any requirements such as uniformity of concentration profiles. Also, because π½ch=Γ
ππ§ππ½π, it is the case thatΓ
πππ =1, just like for transference numbers. However, what is different is that transport numbers can also be<0 or>1, while transference numbers are always between 0 and 1. For instance, in a process called βosmotic power generationβ, or reverse electrodialysis (RED), both ions have fluxes in the same direction, thus for the coion the transport number is<0, while it is>1 for the counterion.
For solutions with all ions monovalent (a 1:1 salt, but more than one type of cation and anion possible), we can use the concept of the current efficiency,π, which is the ratio of ions flux into a surface,π½
ions, over the current density,π½
ch. This current efficiency,π, can be used to characterize ion adsorption in porous electrodes, see Β§15.3, and also describes ion selectivity in membrane transport, see Β§12.3. Current efficiency is defined as
π= π½
ions
π½ch
= π½++π½β
π½ch (7.24)
and for a 1:1 salt can be related to transport numbers according to
π=π+βπβ (7.25)
whereπ+is calculated as a summation over theππβs of all cations, and the same forπβand all anions. Thus when all anions together contribute 50% to the current density, and likewise
vThis is then the case at the very surface, not necessarily at distancesπ₯ >0, unless there is steady-state, no dispersion, etc.
172 Solute Transport
all cations contribute 50%, thenπ+=πβ=Β½ and thenπ=0, i.e., there is transport of current, with all cations going in one direction and all anions in the opposite direction, but there is no βnetβ transport of ions as a whole.
As mentioned, when both the anions and cations go in the same direction (as is the case for the aforementioned RED-process), then the transport number will be<0 for the coion and>1 for the counterion, and consequentlyπ >1. As a metric to define efficiency, numbers beyond unity are not very intuitive, and that is why in RED and similar processes, instead a different efficiency is used, which is the salt transport efficiency, π. This efficiency is defined asπ=1/πand is a measure of how effectively the salt concentration difference is used to generate electrical current. For a well-designed RED process this ratio is close to unity. However, when membranes are used with large pores (a pore diameter of several nmβs or more), or membranes that are too thin (for instance a thickness less than 1πm), then the ions flux (leakage of ions,π½++π½β) is high relative to the generated current, and thusπwill be low. On p. 338 we discuss in more detail the low value ofπwhen very thin membranes are used, and the effect thereof on osmotic power production.