Journal of Membrane Science 478 (2015) 37–48

Contents lists available at ScienceDirect

Journal of Membrane Science

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

Efficiency of electrochemical gas compression, pumping and power

generation in membranes
Jacopo Catalano a, Anders Bentien a,n, David Nicolas Østedgaard-Munck a, Signe Kjelstrup b
a
Department of Engineering, Aarhus University, Hangoevej 2, 8200 Aarhus N, Denmark
b
Department of Chemistry, Norwegian University of Science and Technology, 7491 Trondheim, Norway

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

Article history: The electrokinetic effects in membranes and porous materials can be used for direct conversion of
Received 22 August 2014 pressure into electricity or conversion of electricity into pumping power, and they have potential
Received in revised form applications within actuators, small scale pumping and energy harvesting devices. Still, in the literature
17 December 2014
only electrokinetic effects with liquid reservoirs on each side of the membrane are considered. In the
Accepted 22 December 2014
present paper, isothermal electrokinetic effects are investigated using non-equilibrium thermodynamics
Available online 1 January 2015
in the case with gas phase reservoirs on each side of the membrane. For comparison the case with liquid
Keywords: reservoirs is included. We describe how the figure-of-merit, energy conversion efficiency, power density,
Proton exchange membrane compressor and generator curves depend on observable transport properties. The derived equations are
Electrochemical gas compression
used to analyse two examples with gas phase reservoirs. The first being related to electrochemical
Power generation
hydrogen compression/liquefaction, and the second being the electrochemical gas compression for
Electrokinetic
cooling cycles. In the latter case, experimental transport data predict high efficiency and a power density
which is promising with respect to membrane-based electrochemical cooling applications.
& 2015 Elsevier B.V. All rights reserved.

1. Introduction values of the elecktrokinetic figure-of-merit [1,2,24] have shown

Electrokinetic effects in porous materials and membranes arise
due to the coupling between the movement of ions and solvent. In
e.g. water such a coupling can be explained and modelled by
electrostatic interactions between charged ions and polar water
molecules. Electrokinetic effects can be utilised for direct conver-
sion of potential or kinetic energy (e.g. Pressurised liquids and
gases) into electrochemical energy or vice versa [1]. This is
potentially attractive for many applications [2] such as microfluidic
pumps, gas compressors for cooling cycles, small generators in e.g.
domestic compressed air energy storage [3] or heat engines e.g.
organic Rankine cycles [4].
The Saxèn relations have been known since long, but specific
studies of electrokinetic energy conversion were done first in the
1960s by Osterle and co-workers [1,5] and Burgreen and Nakache
[6,7]. They predicted a maximum first law efficiency (ηEK ) of the
order 1–3% and up to 17%, respectively. During the past decade
renewed interest in this topic has emerged; and most reports
focus on the transport properties of straight nanochannels with
well-defined dimensions [8–23]. Recently [2,24], experimental
n
Corresponding author.
E-mail address: bentien@eng.au.dk (A. Bentien).
that efficiency of about 20% can be obtained in commercial Nafion
membranes and it is increasing with temperature.
Up to now all theoretical studies of electrokinetic energy
conversion have only considered incompressible fluids and in
particular aqueous solutions. However, gas phase electrokinetic
energy conversion is possible too and the potential applications
are electrochemical gas compression or power generation (expan-
der). Some experimental works have been reported on electro-
chemical compression of H2 through a membrane [25–29], with
focus on high pressure liquefaction of H2. Electrochemical gas
compression in membranes for cooling applications is described
in the patent literature, e.g. Refs. [30–32]. None of the patents
disclose technical details or theoretical considerations with res-
pect to e.g. efficiency and power density.
Fig. 1 gives a schematic illustration of electro-osmosis with
liquid (left) and gas reservoirs (right). In the first case, an ion, say
Li þ , migrates through the membrane due to an external electric
potential (ϕ) difference. Each Li þ couples with a number of water
molecules, that are quantified by the water transport number
(t w ). In popular terms one may say that Li þ is dragging or
pumping water through the membrane as it is moving in the
electric field.
Alternatively, a gas phase can supply the conducting ion by gas
oxidation and subsequent reduction on the two membrane sides.

http://dx.doi.org/10.1016/j.memsci.2014.12.042
0376-7388/& 2015 Elsevier B.V. All rights reserved.
38 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48
Fig. 1. Schematic representation of electrokinetic pumping (left) and gas compres-
sion (right) across a membrane (M) having thickness Δx ¼ x2 −x1 . Electrodes
(E) generate an external electrical potential difference (Δϕ ¼ ϕ2  ϕ1 ) across the
membrane that results in an ion current density (j) through the membrane. In
the liquid phase a volume flux (J v ) and in the vapour phase a gas flux (J n ) arises due
the coupling between the ion and the solvent. During operation a pressure
difference (Δp ¼ p2  p1 ) between the two sides is present. In a practical system
the pump could be designed similar to that of a redox flow battery where faradaic
processes takes place in the whole volume of the porous (e.g. carbon felt) electrode.
The compressor could be designed similar to that of a fuel cell with a membrane
electrode assembly. Here the faradaic processes only take place in the catalyst layer
(C) between the membrane surface and the electrode (electrically conductive gas
diffusion layer).
The membrane surface is then covered with an electrode/catalyst.
Examples of such gas reactions are 1=2 H2 ⇆H þ þe  , 1=2 Br2 þ

e  ⇆Br  or 1=2 Cl2 þ e  ⇆Cl . Other gas molecules, e.g. water,
methanol or ammonia, in the gas phase can dissolve in the membrane
and be transported along with the ion. This is quantified by the
gas transport number (t n ) and in cooling cycles, this transported fluid
acts as a refrigerant. Since electrokinetic pumping/compression with
liquid or gas reservoirs are reversible processes an external pressure
difference can be converted into an electrical power.
The basic theoretical description [1,5] of electrokinetic energy
conversion is obtained from non-equilibrium thermodynamics. The
theory has been used to find a general expression for maximum
efficiency that increases with the electrokinetic figure-of-merit
(β ¼ ðν2 σ Þ=κ H ), where ν is the streaming potential coefficient, σ is
the ion conductivity and κ H is the hydraulic permeability of the
membrane [1,5]. One of the aims of the present article is to explore
this theory in more details. In particular, we want to define pump
and generator curves, and find requirements for maximum effi-
ciency, power density, maximum pressure (pump head) etc. These
parameters are very important for assessing electrokinetic energy
conversion with respect to various applications.
Still, the major aim of the current article is to formulate the
transport equations for the gas phase in a practically useful way. The
theory of non-equilibrium thermodynamics is indispensable when it
comes to correctly define the coupling phenomena, which are central
in electrokinetic energy conversion. This includes the derivation of a
general figure-of-merit and maximum efficiency along with other
important parameters e.g. compressor and generator (expander)
curves, requirements for maximum efficiency, power density, max-
imum pressure etc. Despite applications within the field of electro-
kinetic H2 compression and the many patent applications for
electrochemical cooling cycles, a theoretical basis has not yet been
reported in the literature.
The following two theoretical sections cover electrokinetic
energy conversion with liquid and gas phase reservoirs, respec-
tively, and can be read independently. They are followed by
a section that discusses the significance of the results along
with estimates of efficiency of electrochemical gas compression
based on transport data from the literature. The theory does not
include effects from mass transport and catalytic activity in
electrodes and electrode membrane interface. It can however
be included emperically as overpotentials as will be done later in
the discussion.
2. Electrokinetic energy conversion with liquid phase
reservoirs
2.1. Transport equations
Fig. 2 shows a circuit diagram of electrokinetic energy conver-
sion. The central unit is a membrane, characterised by the ion
conductivity (σ ) and hydraulic permeability (κ H ). Electric power
can be generated (Fig. 2 left) when a pressure difference
(Δp ¼ p2  p1 ) is applied across the membrane, resulting in a
volume flux (J v ) through the membrane. Due to the coupling
between the movement of solvent and ions an electric current
density (j) and a potential difference (Δϕ ¼ ϕ2  ϕ1 ) are induced
in the electrical circuit, where electrical power can do work in an
variable external load (RL ). In the pumping operation (Fig. 2 right),
an electric potential difference across the membrane induces a
current density along with a volume flow and a pressure differ-
ence arises across the membrane. The hydraulic work can be
dissipated within a variable hydraulic load/resistance (RL ).
An isothermal system, at a fixed temperature T, exposed to an
electric and a mechanical driving force obeys the following
phenomenological transport Eqs. [1,5,33,34].



1 Δp 1 Δϕ
J v ¼ Lvv  þLvϕ  ð1Þ
T Δx T Δx



1 Δp 1 Δϕ
j ¼ L ϕv  þ Lϕϕ  ð2Þ
T Δx T Δx
The direction and magnitude of the volume flux (J v ) and current
density (j) is determined by the transport coefficients (Lij ) where
the off-diagonal coefficients (written with appropriate units) obey
the Onsager relation Lvϕ ¼ Lϕv .
In the following, the phenomenological transport coefficients
are related to observable transport coefficients. The electric
Fig. 2. Diagram of electrokinetic power generation/expander (left) and compres-
sion (right) across a membrane. The system is charachterised by an ion conductivity
(σ), electrical current density (j), gas permeability (κ n ) and gas flux (J n ). During
power generation (left) the pressure difference (Δp ¼ p2  p1 ) is negative and the
potential difference (Δϕ ¼ ϕ2  ϕ1 ) is positive. For compression the signs are
reversed. With this convention j and J n are always positive. The streaming potential
coefficient (not shown in figure) is negative for cation conductive membranes and
postive for anion conductive membranes. The arrows at the top and bottom
indicate wheter electric power (P el ) or compression/expansion work (P comp ) is
inserted to or extracted from the system. The case for liquids reservoirs on each
side is fully equivalent, here κ n and J n are replaced with the hydraulic permeability
(κ H ) and volume flux (J v ), respectively. While P comp is replaced with the hydraulic
power (P hyd ¼ J v ðΔp=ΔxÞ).
 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48 39

conductivity is defined for uniform pressure, as In the power generation mode, a constant Δp=Δx is used as
  driving force, alternatively this can be a constant J v , and again this
jΔx Lϕϕ
σ¼ ¼ ð3Þ does not change the generality of the analysis. In this case Δϕ=Δx
Δϕ Δp ¼ 0 T
decreases linearly with current density and Δϕ=Δx ¼ 0 at jgen ¼ 1,
The hydraulic permeability (κ H ) is defined for zero current density where the external electric load is zero. With respect to J v , it seen
(j ¼ 0), and is: that it increases linearly with current density and can be explained
  ! by the lower hydraulic resistance due to a change in the electro-
J v Δx Lvv Lvϕ Lϕv
κH ¼  ¼ 1 ð4Þ viscous effect (electrolyte drag effects) when the electrical load is
Δp j ¼ 0 T Lvv Lϕϕ lowered.
The equations for the pump and generator curves along with
The streaming potential is related to the transport coefficients
the maximum and minimum values of the potentials and fluxes
through:
  are collected in Table 1. Electro-viscous effects can be observed
Δϕ L ϕv experimentally as changes in the flow or current whenever the
ν¼ ¼ ð5Þ
Δp j ¼ 0 Lϕϕ external electrical or hydraulic load is changed, respectively. An
   
interesting point is that the ratios Δϕ=Δx J ¼ 0 = Δϕ=Δx Δp ¼ 0 ¼
and can be related to the salt (t s ) and water (t w ) transport v
  J v;Δϕ ¼ 0 = J v;j ¼ 0 ¼ 1 þ β, that define the electro-viscous effects,
numbers by ν ¼  t s V s þ t w V w þ ΔV =F [33–35], where V s and
depend on the elecktrokinetic figure-of-merit only, and can be
V w are the molar volume of the salt and water, respectively, and
used for a direct experimental determination of β . For instance
ΔV is the volume transferred by the electrode reaction. The value measurements of J v;Δϕ ¼ 0 and J v;j ¼ 0 correspond to measurement
of ΔV is often known, and can be added to the sum t s V s þ t w V w .
of κ H with and without electro-viscous effects, respectively. In a
The transport Eqs. (1) and (2) can now be rewritten in terms of
practical setup this corresponds to measurement of κ H where the
observable transport coefficients (Eqs. (3)–(5)):
two sides of the membrane are electrically short- and open-
   
  Δp Δϕ circuited, respectively. Similarly, Δϕ=Δx J ¼ 0 and Δϕ=Δx Δp ¼ 0
Jv ¼  κH 1 þ β þ σν ð6Þ v
Δx Δx correspond to measurements of the ion conductivity of the mem-
brane with infinite and zero hydraulic resistance, respectively. A
Δp Δϕ
j ¼ σν σ ; ð7Þ figure-of-merit above 1 has been found experimentally in Nafion
Δx Δx  
[2,24], and significant differences between e.g. Δϕ=Δx J v ¼ 0 and
where β ¼ ðν2 σ Þ=κ H is the electrokinetic figure-of-merit in the case Δϕ=Δx Δp ¼ 0 are thus expected.
with liquid reservoirs on each side.

2.2. Pump and generator curves

2.3. Efficiency and power density
For evaluating the performances and efficiency of generators or
For both pumping and power generation the energy inserted
pumps, the corresponding generator or pump curves must be known.
into or work extracted from the membrane system is the area
For generators the performance is characterised by the ðΔϕ=ΔxÞ  j
under the pump or generator curve. In terms of volume power
curve, while pumps are characterised by the ðΔp=ΔxÞ  J v curve.
density (P), power per unit volume, it is
These curves are equivalent to e.g. the Δϕ  j curve of a fuel cell
or battery.
2
Δp Δp Δp J 2 J jν
By rearrangement and mutual insertion of Eqs. (6) and (7), one P hyd ¼ J v ¼  κH  jν ¼ v v ð10aÞ
Δx Δx Δx κH κH
can write:
Δp 1 1 β  1 Δϕ 1 ν Δϕ j2 Δp j2   J jν
¼   J þ ¼  Jv  j ð8Þ
Δx 1 þ β κH v 1 þ β ν Δx κH κH P el ¼ j ¼  þ jν ¼  1þβ  v ð11aÞ
Δx σ Δx σ κH
Δϕ Δp j ν  1 where the subscripts refer to hydraulic and electric power,
¼ν  ¼ J  1þβ j ð9Þ
Δx Δx σ κH v σ respectively. A negative value of P corresponds to work that is
This gives the pump (Eq. (8)) and generator (Eq. (9)) curve for done on the membrane system while a positive value corresponds
electrokinetic energy conversion. The middle parts of Eqs. (8) and to work that can be extracted from the membrane system. The
(9) express the curves in terms of potential/pressure difference area power density is found from P Δx.
along with a flux, while the right hand sides are expressed as The lower part of Fig. 3 includes the power density curves, Eqs.
functions of the fluxes only. (10b) and (11b), that explicitly are
The pump and generator curves are shown in Fig. 3, where the
2 
dimensionless volume flow, J pump ¼  ðJ v =jνÞ, and the dimension- j2 ν2  Δp
P hyd ¼ J pump  J 2pump ¼  κ H 1 þ β jgen ð10bÞ
less current density, jgen ¼ j=ðσνΔp=ΔxÞ are introduced. In an κH Δx
experimental setup the fluxes J v or j are varied by adjusting the

2 
external hydraulic or electrical resistances (RL in Fig. 2) in pump or Δp j2 h i
P el ¼ βκ H jgen  j2gen ¼  1 þ βð1  J pump Þ ð11bÞ
generator operation, respectively. Δx σ
In a pumping mode, the electric input power can be either a
constant current density or constant potential difference. Here a in terms of the dimensionless fluxes.
constant current density is used, however, the outcome of the For pumping it is seen that the electric power introduced into
following analysis with respect to efficiency, power density etc. is the system decreases with J pump . The hydraulic power output
independent of this choice. From Fig. 3 it is seen that Δp=Δx shows a maximum output for J pump ¼ 1=2. For power generation,
decreases linearly with J pump up to J pump ¼ 1, where Δp=Δx ¼ 0. similar curves are obtained, here P hyd increases with jgen and the
Since j is constant, Δϕ=Δx varies with the volume flux and reaches a output P el shows a maximum at jgen ¼ 1=2.
minimum value at J pump ¼ 1 whenever the external hydraulic resis- The electrokinetic energy conversion efficiency (η) is given as
tance (Fig. 2) is zero. the ratio between the amount of energy that is extracted from the
40 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48

Fig. 3. (Top) Pump (top left) and generator (top right) curves for electrokinetic energy conversion as function of the dimensionless fluxes J pump ¼  ðJ v =jνÞ and
jgen ¼ j=ðσνΔp=ΔxÞ. For pump operation it is assumed that j is constant whereby Δϕ=Δx is variable (right y-axis of the pump curve). For generator operation a constant
Δp=Δx is assumed, whereby J v is variable (rigth y-axis of the generator curve). Equation numbers refer to equations in text. Specific values of Δp=Δx, Δϕ=Δx and J v have been
included for J pump and jgen at values 0 and 1 have been included and can also be found in Table 1. (Bottom) Power density curves for pump (bottom left) and generator
(bottom right) operation as function of the dimensionless fluxes. The electric power input in pumping and hydraulic power input in power generation is dependent on the
dimensionless flux and is a consequence of the value of the external loads. The graphs also show the efficiency (η) (rigth y-axis). The conditions for maximum power density
and maximum effiencieny are indicated by dotted lines and the conditions for this can be found in Table 1 along with the corresponding expressions for the power and
efficiency. Values of Δp=Δx, Δϕ=Δx and J v at maximum power density and maximum effiency can be derived from the equations for the pump (Eq. (8)) and generator curves
(Eq. (9)).

Table 1
Collection of equations that characterise electrokinetic energy conversion in the liquid phase based on measurements of ion conductivity (σ), hydraulic permeability (κ H ) and
streaming potential coefficient (ν). Left side is with respect to pump operation, while right side is with respect to power generation.

Pump operation with constant j Generator operation with constant Δp.

Δp Δϕ Δp
Pump curve Δx ¼  κ1H J v  κνH j Generator curve j
Δx ¼ νΔx  σ
J pump ¼ 0-J v ¼ 0 J pump ¼ 1-Δp ¼ 0 jgen ¼ 0 -j ¼ 0 jgen ¼ 1 -Δϕ ¼ 0
Δp Δϕ Δp
jν
Δx J v ¼ 0 ¼  κ H (pump head)
J v;Δp ¼ 0 ¼  jν (max flow)
Δx j ¼ 0 ¼ νΔx (open circuited potential) jΔϕ ¼ 0 ¼ σνΔp
Δx (max current)
Δϕ Δϕ
j
Δx J ¼ 0 ¼  σ ð1 þ β Þ Δx Δp ¼ 0 ¼  σ
j
J v;j ¼ 0 ¼  κ H Δp
Δx J v;Δϕ ¼ 0 ¼  κ H ð1 þ βÞΔp
Δx
v

Figure-of-merit β ¼ νκHσ
2

Dimensionless volume flow J pump ¼  Jjνv j

Dimensionless current density jgen ¼ σνΔp=Δx
Maximum power density is obtained at J pump;MaxP ¼ 1=2 Maximum power density is obtained at jgen;MaxP ¼ 1=2
 2
ηpump;MaxP ¼ 2ð2βþ βÞ ηgen;MaxP ¼ 2ð2βþ βÞ
2
P hyd;MaxP ¼ βj
4σ P el;MaxP ¼ βκ4H Δp
Δx
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
1þβ 1þβ1
Maximum efficiency is obtained at J pump; Max η ¼ pffiffiffiffiffiffiffiffi Maximum efficiency is obtained at jgen; Max η ¼ β
1þ 1þβ
p ffiffiffiffiffiffiffi
ffi pffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffi
2 β 1þβ 1þβ1 Δp2  2 þ 2 1 þ β  β þ β2 1þβ1
P hyd; Max η ¼ jσ  pffiffiffiffiffiffiffiffi2 ηpump; Max η ¼ pffiffiffiffiffiffiffiffi P el; Max η ¼ κ H Δx β
ηgen; Max η ¼ pffiffiffiffiffiffiffiffi
1þ 1þβ 1þβþ1 1þβþ1

system and the amount inserted and for pumping it is and the condition for maximum efficiency is

pffiffiffiffiffiffiffiffiffiffiffi
P  J 2pump þ J pump 1þβ
J pump ¼ pffiffiffiffiffiffiffiffiffiffiffi ð14Þ
ηpump ¼ hyd ¼    ð12Þ 1þ 1þβ
 P el J pump þ 1 þ β =β and
pffiffiffiffiffiffiffiffiffiffiffi
while for power generation 1þβ1
jgen ¼ ð15Þ
β
P el  j2gen þ jgen
ηgen ¼ ¼ ð13Þ for pumping and power generation, respectively. In both cases the
 P hyd jgen þ ð1=β Þ
maximum efficiency is given by
The efficiency curves are also plotted in the lower parts of pffiffiffiffiffiffiffiffiffiffiffi
Fig. 3. The maximum efficiency (ηmax ) is found by standard 1þβ1
ηmax ¼ pffiffiffiffiffiffiffiffiffiffiffi ð16Þ
mathematical methods with differentiation of Eqs. (12) and (13), 1þβþ1
 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48 41

These results are in accordance with literature [1,5,36], how- number:

ever, with the difference that the conditions for maximum


Lϕn p tn p
efficiency (Eqs. (14) and (15)) are explicitly shown only here. The Δϕj ¼ 0 ¼  RT ln 2 ¼  RT ln 2 ð18Þ
Lϕϕ p1 F p1
conditions for maximum efficiency, maximum power density  
along with the efficiencies and power densities are also included With t n ¼ 1=2 for H2 this gives Δϕj ¼ 0   13 mV ln p2 =p1 and is
in Table 1. in agreement with experimental literature data [25]. Furthermore,
the streaming potential coefficient (νn ) in the gas phase
" #
Δϕ  L ϕn
3. Electrokinetic energy conversion with gas phase reservoirs νn ¼  ¼ RT ð19Þ
ln p2 =p1 Lϕϕ
j¼0
3.1. Phenomenological transport equations is defined.
The ion conductivity of the membrane is defined for uniform
This section is outlined in the same manner as the previous pressure, as
section where the liquid phase reservoirs were used. The basic  
jΔx Lϕϕ
principles of electrokinetic energy conversion in the presence of σ¼ ¼ ð20Þ
gas phases were shown in Figs. 1 and 2. As an example, we Δϕ Δp ¼ 0 T
consider H2 and methanol (MeOH), in the gas phase on each side The total gas permeability of a gas mixture of non-interacting
of the membrane, at different total and partial pressures. At the gases is normally defined as being a linear combination of the
left hand side of the membrane/electrode surface 12 H2 dissociates single-gas permeabilities and the respective driving forces: i.e.
into H þ ð12 H2 -H þ þ e  Þ. The proton migrates through the mem- P P
J n ¼ J i ¼  κ gas;i ðΔpsurf i =ΔxÞ in which Δpsurf i represents the
brane due to the externally applied electrical field. At the other i
partial pressure i
difference of the species i between the membrane
side of the membrane the opposite reaction occurs surfaces. Considering the case of a binary mixture, no concentra-
(H þ þ e  -12 H2 ) and H2 is formed. Due to electrostatic interac- tion polarisation (psurf ¼ pi ¼ pyi ) and the same gas concentrations
i
tions, MeOH molecules are transported along with the proton ¼
through the membrane.
in both  sides (y
i;1 y  for i ¼ 1; 2), the gas flux simplifies
i;2  as:
J n ¼  y1 κ gas;1 þ 1  y1 κ gas;2 p2 p1 ¼  κ gas p2  p1 in which
Electrokinetic gas compression and power generation in the κ gas is a lumped permeability coefficient. However, for electro-
gas phase has, to the best of the authors' knowledge, not been kinetic energy
described in terms of non-equilibrium thermodynamics. A rigor-  conversion
 a more convenient definition is as
function of ln p2 =p1 . The gas permeation for zero current density
ous derivation of simplified phenomenological transport equations (j ¼ 0) where both sides of the membranes are electrically isolated
is given in Appendix A. The result is from each other, is found from Eqs. (A6) and (A7):



  
R p 1 Δϕ ln p2 =p1
J n ¼ Lnn  ln 2 þ Lnϕ  ðA6Þ Jn ¼  κn ð21Þ
Δx p1 T Δx
Δx



 where the gas permeability coefficient is defined as
R p 1 Eϕ
j ¼ Lϕn  ln 2 þ Lϕϕ  ðA7Þ " # !
Δx p1 T Δx J n Δx Lnϕ Lϕn
κn ¼    ¼ Lnn 1  R ð22Þ
where Lnϕ ¼ Lϕn through the Onsager relation and J n is the total ln p2 =p1 Lnn Lϕϕ
j¼0
molar gas flux. In the example shown in Fig. 1 J n ¼ J H2 þ J g , where J g
is the molar flux of e.g. MeOH. It is assumed that the gases are If the ratio p2 =p1 is sufficiently small Eq. (21) can be Taylor
ideal and that the compositions on both sides of the membrane expanded to give J n ¼  κ n ð1=p1 ÞðΔp=ΔxÞ. For practical purposes
are the same. In a practical application, the latter may not always an experimentally determined/literature value of the gas perme-
be the case; as the concentrations of the transported ion and gas ability obtained from J n ¼  κ gas ðΔp=ΔxÞ can be translated into κ n
may change over time. This accompanying change in the chemical from the expansion, κ n ¼ p1 κ gas , where the units of κ gas and κ n are
potential must then be included into the equations, but they are [mol m  1 Pa  1 s  1] and [mol m  1 s  1], respectively. Alterna-
left out here for simplicity. tively, κ n can be estimated from Eq. (21) provided that J n , p2 and
In the following the phenomenological transport coefficients p1 are given, which often is the case in the literature.
(Lij ) in Eqs. (A6) and (A7) will be related to observable transport According to non-equilibrium thermodynamics theory, the
coefficients using the definitions of the observable transport phenomenological coefficients in the flux-force relations do not
coefficients. depend on the fluxes or forces themselves. They may depend on
Electroosmosis is defined as the number of molecules trans- state variables, such as temperature  or pressure. Any experimental
ported across the membrane along with the electric current ðJ n =jÞ deviation from a linear J n vs ln p2 =p1 relationship can be caused
in the case where there is no trans-membrane pressure difference by such a dependence, which will appear as a consequence of the
ðΔp ¼ p2  p1 ¼ 0Þ. From Eqs. (A6) and (A7) this gives integration across the membrane (Eq. (A5)). We have assumed that

 the gasses are ideal in the integration, while say, a temperature
Jn Lnϕ t n dependence, may appear as a non-linearity after integration. This
¼ ¼ ð17Þ
j Δp ¼ 0 Lϕϕ F will also mean that compression and power generation efficiency
becomes pressure dependent.
where F is the Faraday constant and t n is the gas transport number,
By introducing the definitions of the observable transport
which is the total number of gas molecules transported across the
coefficients the transport Eqs. (A6) and (A7) can be rewritten as
membrane per unit of charge. In comparison to the case with  
liquid reservoirs, t n also includes the H2 while t w is only the  ln p2 =p1 t n σ Δϕ
Jn ¼  κn 1 þ β  ð23Þ
number of water/solvent molecules that are dragged through the Δx F Δx
membrane. If only H2 is present and it is assumed that H2 does not  
diffuse through the membrane, t n ¼ 1=2. t n σ RT ln p2 =p1 Δϕ
j¼  σ ð24Þ
The open-circuited potential (j ¼ 0), the streaming potential F Δx Δx
that arises whenever a pressure difference is present, is also found where β ¼ ðRT=F 2 Þðt 2n σ =κ n Þ is defined as the figure-of-merit for
from Eqs. (A6) and (A7) and the definition of gas transport electrokinetic energy conversion with gas phase reservoirs.
42 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48

3.2. Compressor and generator (expander) curves external electrical

 or
 compression
  load is changed, respectively.
The ratios Δϕ=Δx J ¼ 0 = Δϕ=Δx Δp ¼ 0 ¼ J n;Δϕ ¼ 0 = J n;j ¼ 0 ¼ 1 þ β,
n
The performance and efficiency of the electrokinetic generator/ that define the electro-viscous effects depend on the electrokinetic
gas expander can be evaluated from the Δϕ=Δx j characteristics. figure-of-merit only and can be used for a direct experimental
For electrokinetic gas compression
 theperformance and efficiency measurement of β . I.e. J n;Δϕ ¼ 0 and J n;j ¼ 0 correspond to measure-
is evaluated from the ðRT=ΔxÞln p2 =p1  J n curve. The area under ments of κ n where the two sides of the membrane  are electrically

this curve gives the (reversible) work per unit time and volume  open-circuited, respectively. Similarly, Δϕ=Δx J n ¼0 and
short- and
that is done on the gas. Δϕ=Δx Δp¼0 correspond to the measurements of the ion conduc-
From rearrangement and introduction of Eqs. (23) and (24), the tivity of the membrane with infinite and zero external compression
compressor and generator curves are found: load, respectively.
 
ln p2 =p1 1 1 F β 1 Δϕ 1 tn
¼   J    ¼  Jn þ j ð25Þ
Δx 1 þ β κ n n RT 1 þ β t n Δx κn F κn
3.3. Efficiency and power density
 
Δϕ t n RT ln p2 =p1 j t n RT ð1 þ β Þ
¼  ¼ J  j ð26Þ The volume power density with respect to the amount of work
Δx F Δx σ κn F n σ
that is done on gas for compression and generation is the area
 
The middle part of Eqs. (25) and (26) gives the pressure and under the ðRT=ΔxÞln p2 =p1  J n and ðΔϕ=ΔxÞ  j curves, respec-
potential difference as function  of the fluxes and the potentials tively. The area power density can be calculated from P Δx.
Δϕ=Δx and ðRT=ΔxÞln p2 =p1 , while the right hand side of the Explicitly, the equations found are
equations gives the curves as function of the fluxes J n and j only.
   
  2
With Eqs. (25) and (26) the compressor and generator curves for ln p2 =p1 tn ln p2 =p1 ln p2 =p1 t n RT RT
P comp ¼ J n RT ¼ jRT  κ n RT ¼ jJ  J 2
electrokinetic energy conversion in the gas phase can be con- Δx F Δx Δx F κn n κn n
structed and they have the same functional behaviour as the ones ð27aÞ
for liquid reservoirs (Eqs. (8) and (9)) and are shown in Fig. 4,
   
where they are plotted as function of the dimensionless gas Δϕ j2 t n RT ln p2 =p1 t n RT 1þβ 2
flow (J comp ¼ J n F=ðjt n Þ) and the dimensionless current density P el ¼ j ¼  j ¼ J j j ð28aÞ
Δx σ F Δx κnF n σ
(jgen ¼ −ðjFΔxÞ=ðσt n RT lnðp2 =p1 ÞÞ). The explicit dependence of Eqs.
(25) and (26) on J comp and jgen can be found by inserting the As for the case of liquid reservoirs (Eqs. (10a), (10b), (11a) and
expressions for into the equations. (11b)), a negative value corresponds to external work that is
As for liquid reservoirs, and without loss of generality, it is inserted into the membrane while a positive value corresponds
assumed that the constant driving force for compression is a to work that can be extracted from the membrane.
constant j while Δϕ=Δx is variable. For generation
 it is assumed Eqs. (27a) and (28a) are also shown in the lower part of Fig. 4
that the constant driving force is ln p2 =p1 while J n is variable. and the explicit dependence on J comp and jgen can be found:
The equations for the pump and generator curves are collected in

 2 
Table 2. From an experimental point-of-view electro-viscous effects β ln p2 =p1
P comp ¼ J J 2 ¼  κ n RT jgen þ 1 ð27bÞ
are observed as changes in the flow or current whenever the σ comp comp Δx

Fig. 4. (Top) Compressor (top left) and generator (top right) curves for electrokinetic energy conversion as function of the dimensionless fluxes J comp ¼ J n F=ðjt n Þ
and jgen ¼ −ðjFΔxÞ=ðσt n RT lnðp2 =p1 ÞÞ. For pump operation it is assumed that j is constant whereby Δϕ=Δx is variable (right y-axis of the pump curve). For generator operation
   
a constant ln p2 =p1 is assumed, whereby J n is variable (rigth y-axis of the generator curve). Equation numbers refer to equations in text. Specific values of ln p2 =p1 , Δϕ=Δx
and J n have been included for J comp and jgen at values 0 and 1 have been included and can also be found in Table 2. (Bottom) Power density curves for compression (bottom
left) and generator (bottom right) operation as function of the dimensionless fluxes. The electric power input in compression and expansion power input in power
generation is dependent on the dimensionless flux and is a consequence of the value of the external loads. The graphs also show the efficiency (η) (rigth y-axis). The
condition for maximum power density and maximum effiencieny are indicated by dotted lines and the conditions for this can be found in Table 2 along with the
 
corresponding expressions for the power and efficiency. Values of ln p2 =p1 , Δϕ=Δx and J n at maximum power density and maximum effiency can be derived from the
equations for the compressor (eq. 25) and generator curves (eq. 26).
 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48 43

Table 2
Collection of equations that characterise electrokinetic energy conversion in the gas phase based on measurements of ion conductivity (σ), gas permeability (κ n ) and gas
transport number (t n ). Left side is with respect to compressor operation, while right side is with respect to power generation.

Compressor operation with constant j Generator operation with constant Δp.

lnðp2 =p1 Þ Δϕ lnðp2 =p1 Þ

Compressor curve Δx ¼ tn
 κ1n J n þ Fκ E
j Generator curve Δx ¼  tnFRT Δx  σj
J comp ¼ 0 -J n ¼ 0 J comp ¼ 1 -Δp ¼ 0 jgen ¼ 0 -j ¼ 0 jgen ¼ 1 -Δϕ ¼ 0
 Δϕ
lnðp2 =p1 Þ J n;Δp ¼ 0 ¼ jtFn (max flow) t n RT lnðp2 =p1 Þ lnðp2 =p1 Þ
Δx ¼ jFκtnn (max compression ratio) Δx j ¼ 0 ¼  F Δx (open circuited potential) jΔϕ ¼ 0 ¼  σ tnFRT Δx (max current)
Jn ¼ 0
Δϕ j
Δϕ lnðp2 =p1 Þ lnðp2 =p1 Þ
Δx J ¼ 0 ¼  σ ð1þ β Þ Δx Δp ¼ 0
¼  σj J n;j ¼ 0 ¼  κ n Δx J n;Δϕ ¼ 0 ¼  κ n ð1 þ βÞ Δx
n

RT t n σ
2
Figure-of-merit β ¼ F 2 κn

Dimensionless gas flow J comp ¼ JjtnnF Dimensionless current density jgen ¼  σt jFΔx
lnðp2 =p1 Þ
n RT

Maximum power density is obtained at J comp;MaxP ¼ 1=2 Maximum power density is obtained at jgen;MaxP ¼ 1=2
 2
ηcomp;MaxP ¼ 2ð2βþ βÞ ηgen;MaxP ¼ 2ð2βþ βÞ
2
P comp;MaxP ¼ βj lnðp2 =p1 Þ
4σ P el;MaxP ¼ βκn4RT Δx
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
1þβ 1þβ1
Maximum efficiency is obtained at J comp; Max η ¼ pffiffiffiffiffiffiffiffi Maximum efficiency is obtained at jgen; Max η ¼ β
1þ 1þβ
p ffiffiffiffiffiffiffi
ffi pffiffiffiffiffiffiffiffi  2   2 þ 2pffiffiffiffiffiffiffi
ffi  pffiffiffiffiffiffiffiffi
β 1þβ 1þβ1 κ n RT lnðp2 =p1 Þ 1þβ1
2
1 þ β  β þ β2
P comp; Max η ¼ jσ  pffiffiffiffiffiffiffiffi2 ηcomp; Max η ¼ pffiffiffiffiffiffiffiffi P el; Max η ¼ Δx β
ηgen; Max η ¼ pffiffiffiffiffiffiffiffi
1þ 1þβ 1þβþ1 4 1þβþ1

  2 
ln p2 =p1 j2 h  i power density are still valid, but the effective potential rather than the
P el ¼ βκ n RT jgen  j2gen ¼  1 þ β 1  J comp
Δx σ potential difference at reversible conditions will enter the equations.
ð28bÞ Alternatively, the transport properties should be measured in a way
that includes the effect of overpotentials, whereby the equations for
The power density that can be extracted from the membrane the efficiency and power density are fully valid.
shows a maximum whenever J comp ¼ 1=2 or jgen ¼ 1=2 for com-
pression and generation, respectively. Due to the electro-viscous 4.1. Condition for maximum for efficiency in power generation
effect, the amount of work/power (density) inserted into the
system depends on gas flux and current density. For instance with From an experimental point of view it is often an advantage to
compression, P el decreases with J comp and is a consequence of a have a theoretical prediction of the conditions for maximum
decreased external compression load (see Fig. 2) that goes to zero efficiency. As an example an electrokinetic generator with liquid
at J comp ¼ 1. For power generation P comp increases with jgen and is a reservoirs (Fig. 2, left) is chosen. The generator operates with
consequence of decreasing external electrical load that goes to constant pressure difference and the maximum efficiency is
zero at jgen ¼ 1 obtained by varying the magnitude of the external electrical load
The efficiency for compression is (RL) inducing the question of the magnitude of RL which max-
P comp  J 2comp þJ comp imises the efficiency. From Fig. 2 it is seen that the current density
ηcomp ¼ ¼    ð29Þ in the generator is j ¼ σ sys ðΔϕ=ΔxÞ, where σ sys 1
¼ σ internal
1
þ σ L 1 is
 P el J comp þ 1 þ β =β
the combined reciprocal electrical conductivity from the mem-
For generation the efficiency is: brane and the external load. The product between the streaming
potential coefficient and the pressure difference is  the internal
P el j2gen þ jgen
ηgen ¼ ¼ ð30Þ voltage of the generator Δϕ=Δx ¼ νðΔp=ΔxÞ and hence
 P comp jgen þ ð1=β Þ j ¼ σ sys νðΔp=ΔxÞ. From Table1pitffiffiffiffiffiffiffiffiffiffiffi
is seen that maximum efficiency
 
These curves coincide with the ones for energy conversion in the is obtained at jgen; Max η ¼ 1 þ β  1 =β ¼ j= σ internal νΔp=Δx ,
liquid state (Eqs. (12) and (13)). Maximum efficiency can be found where the subscript internal has been added to the conductivity
frompstandard in order to distinquish between the two conductivities. By com-
ffiffiffiffiffiffiffiffiffiffiffi mathematical
pffiffiffiffiffiffiffiffiffiffiffi methods and is found to be
η¼ 1þβ1 = 1 þ β þ 1 where β ¼ ðRT=F 2 Þðt 2n σ =κ n Þ is the bining these two expression
pffiffiffiffiffiffiffiffiffiffi it is found that
figure-of-merit for electrokinetic compression or power genera- σ internal =σ L ¼ RL =Rinternal ¼ β=ð 1 þ β−1Þ−1, and in the case with e.
tion in p the gasphase. The condition for maximum g. β ¼ 1 the external resistance has to be about 1.4 times the
ffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi p ffiffiffiffiffiffiffiffiffiffiffi efficiency is
J comp ¼ 1 þ β= 1 þ 1 þ β and jgen ¼ 1 þ β  1 =β for com- internal resistance in order to obtain maximum efficiency. On the
pression and power generation, respectively. other hand it is well-known that maximum power output is
Table 2 summarises the results of the compressor, generator obtained from RL ¼ Rinternal , which is stated as jgen;MaxP ¼ 1=2 in
curves, power density and efficiency. Some of the results that are Table 1. The corresponding power densities and efficiencies can be
of particular relevance with respect to dimensioning and applica- found from the relevant equations in the table.
tions are the equations for the maximum compression ratio,
power density and the condition for maximum efficiency. 4.2. Energy conversion with liquid reservoirs

As an example we consider the electrokinetic transport proper-

4. Discussion
The equations for efficiency and power density follow from
standard non-equilibrium thermodynamics and are generally valid
for reversible electrodes. In real electrokinetic energy conversion
systems, electrode overpotentials and polarisation phenomena at the
electrode and membrane surfaces may decrease the conversion
efficiency and power density. The equations for the efficiency and
ties (σ , ν, κ H and β) of Nafions 117 (in the following referred to as
Nafion), which has been thoroughly investigated [2,24]. A rela-
tively high conversion efficiency of the order 20% was found in
0.03 M LiCl solutions. The transport properties of Nafion can be
used to estimate the pump characteristics of electrokinetic pump-
ing. Here it is desirable to have a large current density in order to
maximise the power density. At a high electrolyte concentration
(40.3 M) a current density of 100 mA cm  2 is reasonable without
44 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48
very large concentration polarisation effects. From the equations
for electrokinetic energy conversion with liquids in Table 1 some
of the important pump characteristics like the area power density
and pressures are calculated and given in Table 3 for a 180 μm
Nafion 117 membrane.
A current density (100 mA cm  2) can generate a maximum
pressure difference of 226 bar, whereas the pressure differences at
maximum power density and efficiency are 113 bar and 92 bar
respectively. Experimentally, values up to about 25 bar with
j 30 mA cm  2 to 180 mA cm  2 have been observed [37,38].
However, in these experiments the pressure difference was far
from being fully developed, presumably due to mechanical limita-
tions of the test cell. A maximum pressure difference of the order
200 bar is not unrealistic and is in agreement with predictions
from the transport properties.
In many applications, high power densities and efficiencies are
key parameters. In the case of a 180 μm Nafion membrane high
performance can only be obtained if the system is operated at high
pressure differences (4100 bar). If a more moderate operating
pressure difference is needed (e.g. 10 bar), it can be obtained by
(i) shifting to a more extreme point at the pump curve at the
expense of a an efficiency that drops below 5%, (ii) reduction of the
current density to 10 mA cm  2, which is at the expense of a power
density that is reduced by a factor of 100 or (iii) reduction of the
membrane thickness, which maintains the efficiency and increases
the power density.
Due to their relatively high pressure difference at maximum
efficiency and power density, Nafion membranes are potential
targets for applications in HPLC (high pressure liquid chromato-
graphy). The pumps in HPLC operate at very high pressure and
with relatively small volume flow.
4.3. Electrokinetic hydrogen gas compression
One of the main results in the present article is the derivation of
conditions for maximum efficiency and power density for electro-
kinetic energy conversion with gas phase reservoirs. It can be used for
dimensioning of and search for efficient membrane materials for
electrochemical gas compression. The figure-of-merit for gas phase
electrokinetic energy conversion, β ¼ ðRT=F 2 Þðt 2n σ =κ n Þ, resembles that
for conversion with liquid phases, with the squared streaming
potential term (strength of the electrokinetic coupling) and the
conductivity and permeability term in the nominator and denomi-
nator, respectively. The figure-of-merit increases with the gas trans-
port number (t n ) squared and this can be used for screening
membrane materials with high efficiency. t n can be determined from
t n ¼  νn ðF=RTÞ (Eqs. (17) and (19)) by using the experimental values
of the streaming potential coefficient (νn ). In the case of pure H2 on
both sides of the membrane t n ¼ 1=2.
The efficiency and power density for energy conversion in
Nafion membranes is discussed as practical examples for the case
with pure H2 and mixed gas phases on both sides of the
membrane. Nafion is well characterised with respect to both H2
permeation and proton conductivity for a range of different
humidity levels. Table 4 gives a compilation of literature transport
Table 3
Calculation of pump characteristics for Nafion 117 membrane with a thickness of 180 μm based on literature data of the transport data [2,24] in the left column. The power
density is calculated from the relevant equations in Table 1 and assuming a current density of 100 mA cm  2.
Transport coefficient Value Current density: j¼100 mA cm  2
κ H (m2 s  1 Pa-1) 3.1  10  17 Pump head (zero flow)
ν (μV bar  1) 390 Maximum power density
σ (S m  1) 2.36 Maximum efficiency
β 1.2
η (%) 20
data for Nafion surrounded by H2 in the gas phase [39]. Estimates
of κ n are made using the definition of κ n (Eq. (22)) and with values
of J n from literature permeation rates [39] assuming p2 ¼ 10 bar
and p1 ¼ 1 bar. This most likely leads to overestimated values of κ n .
The interesting result here is that for Nafion in pure H2 β ¼ 74
corresponding to η ¼ 79%. Even if κ n is a factor 10 larger, which
results in β ¼ 7:4, the maximum efficiency still has a high value of
49% and experimental data unambiguously show that there is a
large coupling in the case of Nafion and pure H2. This fact is not
surprising, in view of the electrode reaction, which fully couples to
proton transport to the electric current density. It is only the
ohmic loss of ion transport and the impact of H2 diffusion on the
reservoir pressures, which can decrease the efficiency.
Despite the very high efficiency, there are two major draw-
backs, the first being a very low power density and the second is
that overpotentials in the dissociation/recombination of H2 at the
electrodes have not been included in the calculation of the
efficiency. With respect to the power density two different
examples of gas compression with varying membrane thickness
(180 mm and 50 mm corresponding to commercial Nafion 117 and
212 respectively) and with a compression ratio of 210 between the
low and high pressure side have been included. It is seen that
highest obtainable power density is about 11 mW cm  2, for a
50 mm membrane with a compression ratio of 210. Although the
current density is relatively high (180 mA cm  2) the low power
density comes from the relatively low potential drop of 78 mV
across the membrane. If this technology is to be used for large
scale H2 compression/liquefaction, large membrane areas are
needed, adding significantly to the total system costs. In the
literature [29], experimental value of a steady state compression
ratio of 150 was found for j ¼ 200 mA cm  2, in quite good
agreement with the estimated values in Table 4. Furthermore
Grigoriev et al. also made an experimental plot of the cell voltage
as function of compression (Fig. 6 in ref. [29]) and found an
approximate
   logarithmic relationship. This is expected from
ln p2 =p1 =Δx J ¼ 0 ¼ jt n =ðFκ n Þ (Table 2) if the ratio κn F=ðt n ΔxÞ is
n
independent of pressure indicating that κ n too is somewhat
independent of pressure.
Still, in the efficiency calculation, overpotentials in the elec-
trode materials have not been included. Depending on the defini-
tion, the overpotential accounts for both activation losses and
reaction kinetic effects at the electrode. In the present study a
crude model for the overpotential is applied. A constant over-
potential of 50 mV per electrode that is independent of current
density is assumed, this is most likely a low and optimistic value
[25,29]. Nonetheless, the main point is that the overpotential has a
large impact and the maximum efficiency decreases from 79% to
35% (last column Table 4) because the overpotential is of the same
magnitude as the potential difference across the membrane.
4.4. Electrokinetic gas compression – cooling applications
Alternatively, electrokinetic effects with gas phase reservoirs
on each side of the membrane can be used in cooling cycles or
power generation from heat sources. In the present discussion
Power density (mW cm  2) Pressure (bar)
– 226
2.2 113
2.1 92
 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48 45

only gas compression for cooling applications is considered and it
is implicitly assumed that it is a condensation cycle type similar to
that in mechanical refrigeration systems.
The use of pure H2 in cooling cycles is not realistic since it
would require extremely high compression ratios in order to
liquefy H2 and the power density is at the same time quite low.
Alternatively, a system containing H2 along with a refrigerant will
be analysed and it will be shown that higher system efficiencies
and power densities can be obtained and with compression ratios
that are much more adequate for cooling applications.
From the literature it is well established that the hydration shell
increases with K þ , Na þ , Li þ , (H þ ) as observed by the largest water
transport number in e.g Nafion [40]. Water transport numbers
t w  2–3 that are found experimentally for H þ in Nafion [41] are
significantly lower than those for any of the other monovalent ions
[40]. This discrepancy is explained by the extra transport tunnelling
mechanism of protons in water, that reduces the drag of the water
molecules, thereby also t w . In electrokinetic energy conversion the
tunnelling mechanism decreases the efficiency and is thus not a
wanted feature. Alternatively, in a solvent such as methanol (MeOH)
protons cannot tunnel, and MeOH may therefore be a suitable
choice. High values of the coupling between Li þ and MeOH have
been observed experimentally with liquid reservoirs [42], corre-
sponding to t MeOH  21. To the best of our knowledge there are no
data for H þ and MeOH in Nafion. However, the transport number is
expected to be of the same magnitude. This is supported by the
large decrease of σ H þ ;water  10 S m  1 /σ H þ ;MeOH  1 S m  1 com-
pared to σ Li þ ;water  1:3 S m  1 /σ Li þ ;MeOH  0:6 S m  1 in Nafion
going from 100% water to 100% MeOH [43,44]. The large magnitude
of σ H þ ;water =σ H þ ;MeOH indicates that tunnelling does not contribute
in MeOH while the similar magnitude of σ H þ ;MeOH and σ Li þ ;MeOH
suggests that the dragging effect of the solvent is similar for
both ions.
There are no experimental data available for t n in a system with
Nafion and gas phase MeOH and H2. However, assuming that the
coupling is similar to that with liquid reservoirs a value of t n from
5–20 is reasonable and is used in the following.
The gas permeability of H2O and MeOH is several orders of
magnitude larger than for H2 [45] and κ n is estimated to be
5  10  6 mol m  1 s  1. At the same time σ will decrease in MeOH
and with increasing tn/solvent drag. Experimental values of
σ H þ ;MeOH are found to be of the order 1 S m-1 [44]. However, in
the calculations the values are chosen to be in the range
0.3–4.8 S m  1 under the constraint that β ¼ 32, in order to make
the different cases comparable. With these transport properties
the power density and current density were calculated for 50 mm
membranes where the compression ratio is 10. The results are
shown in Table 4.
The main result is that lower current densities are needed
whenever t n increases in order to maintain a compression ratio of
10. At the same time the potential difference across the membrane
increases because the amount of work that is done per unit charge
increases with t n and the combined effect is a constant power
density. Despite the higher compression ratio for pure H2 it can be
seen from Table 4 that the power density is 1–2 orders of
magnitude larger in the example with MeOH. In fact, it can be
shown from the equations in Table 2 that P comp p κ n for a constant
β and constant compression ratio. For practical applications this is
very important since this shows that high power densities could
be found in membrane/solvents/gas systems with relatively large
gas permeability.
An equally important point is that the total system efficiency is
larger whenever a refrigerant is used along with the gas e.g. MeOH
together with H2. As already noted, the potential difference across
the membranes is larger in the case where refrigerants are used
and it increases with t n , as can be seen from Table 4. When
electrode overpotentials are included in the calculation of the
efficiency it is seen that the effect is relatively small and that the
system efficiency still is of the order 60%, much larger than that
with pure H2. This is explained by the larger electrical potential
difference across the membrane that is a consequence of the larger
amount of work that is done per unit charge whenever proton
couple to MeOH.
Domestic refrigerators, about 100 W cooling effect, typically have a
coefficient of performance (COP) of about 1.5, depending on the
cooling temperature. If we assume that most losses are related to the
mechanical compressor, the compressor efficiency is below 30%. As
seen from Table 4, compressor efficiencies with Nafion above 50% are
realistic. Furthermore, a cooling capacity of 100 W can be obtained by
a membrane area of approximately 20 cm  20 cm. This is a relatively
small area and it further underlines the feasibility of this technology
for domestic cooling applications. Furthermore, the cooling capacity is
simply regulated by the magnitude of the DC current in an electro-
kinetic gas compressor and can be kept at the maximum efficiency of
the compressor curve at all times. Mechanical compressors need
frequency regulation and permanent magnet motors to regulate the
capacity, which adds significantly to the costs.
The weaknesses of the above calculations are of course a lack of
comprehensive experimental data. Nevertheless, the main result is
that very high efficiencies and power densities can be predicted

Table 4
Calculation of efficiency, current density, potential difference and power density for different membrane thicknesses and for different compression ratios. Calculations are
based on the equations in Table 2. Gas transport properties of Nafion with pure H2 have been estimated from Ref. [39]. In the case of Nafion with H2 and MeOH the transport
properties have been estimated based on a fixed gas permeability that is approximately 550 times larger [45] than in the case for pure H2 and under the constraint that the
figure-of-merit is 32.

Gas Gas Ion Figure- Efficiency Current Potential Membrane Area power Compression Efficiency
permeability transport conductivity of- (η) density difference thickness density at ratio at including 100 mV
(κ n ) number (σ) merit (j) across (Δx) maximum maximum electrode
(t n ) (β) membrane (Δϕ) efficiency efficiency overpotential
mol m  1 s  1 S m1 % mA cm  2 mV μm mW cm  2 %

Nafion 9  10  9 1/2 10 74 79 50 78 180 3.1 210 35

with 180 78 50 11.2 210 35
pure
H2

Nafion 5  10  6 5 4.8 32 70 600 360 50 152 10 55

with 10 1.2 300 720 152 62
H2 20 0.3 150 1440 152 66
and
MeOH
46 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48

for gas compression in cooling cycles if a refrigerant is included in η first law efficiency, dimensionless
the system. This does in particular hold when overpotentials are κH , κn Hydraulic permeability and gas permeability coefficient,
included in the model. The calculations should be regarded as a m2 s  1 Pa  1 and mol s  1 m  1
guide to obtain high efficiencies and power densities. μi chemical potential of component i, J mol  1
In order to identify compressor systems that have practical ν; νn Streaming potential with liquid and gas phase reservoirs,
relevance, aimed research in membrane properties along with a V Pa  1 and V
suitable refrigerant must be addressed. In the above examples MeOH σ electrical/ionic conductance, S m  1
and Nafion has been chosen only because it is well characterised in the σe volumetric entropy production rate, J m  3 s  1 K  1
literature and this data can be used for estimating the performance ϕ; Δϕ electric potential (difference), V
based on the theory developed in the present article. Nonetheless,
enhanced electrokinetic properties is most likely not found in Nafion
membranes and similar. These membranes are so dense that the
solvent (water) uptake per unit charge is about 20–40 water molecules Acknowledgements
per SO3 group depending on the solvent. This ultimately limits the
gas/water transport number and thereby the streaming potential AB wishes to thank the Villum Foundation (grant number:
coefficient. Larger streaming potential coefficients can be found for VKR022356) and The Aarhus University Research Foundation for
membranes with larger solvent uptake and larger pore diameters [2]. funding. SK is grateful to the RENERGI project no. 164466, and
With respect to the refrigerant it must be a polar solvent, otherwise the FRIENERGI project no. 197598/V30, both of the Norwegian
the coupling to the ions will be diminished; it must otherwise be Research Council.
chemically compatible with the membrane and it should have a
boiling point in the range around room temperature. Again MeOH has
Appendix A
been chosen in the example because of the available literature data.
Alternatives may be organic polar solvent/gases and an even more
Consider a membrane surrounded by two gas reservoirs (Fig. 1).
obvious choice may be ammonia that is a known refrigerant with
The reservoirs contain a mixture of two gases. One of them is H2;
excellent properties.
the other may be water or methanol. The membrane is saturated
with solvent (e.g. H2O or MeOH), and electrodes (Pt) are deposited
on to its two external surfaces. The catalyst is connected via an
5. Summary and conclusions
external electrical circuit. The arrangement allows H2 to dissociate
on the membrane surface (left side). Hydrogen is transported
The main goal of the present study was to formulate a theoretical
through the membrane as H þ , and it recombines at the right hand
framework based on non-equilibrium thermodynamics for describ-
side to form gas molecules again. The other gas in the reservoir can
ing the electrokinetic effects in membranes with both gas phase and
be transported through the membrane alone and along with the
liquid reservoirs. In particular for gas phase reservoirs this has not
H þ . This scenario is likely if the gas is well dissolved in the
yet be done in the literature. The theoretical description can have
membrane and is polarisable, like MeOH is. It is assumed that the
large impact on practical applications of membrane electrokinetic
system is isothermal.
effects, e.g. electrochemical gas compression/cooling, micro pumps
The entropy production (σ e ) in such a system, showing various
and power generation. Based on experimental transport data for
electrokinetic effects has in the outset three independent terms:
dense ion conducting membranes (Nafion) we predict high effi-




ciency for electrochemical gas compression/cooling in practical 1 ∂μH2 ;T 1 ∂μg;T 1 ∂ϕ
σ e ¼ J H2  þ Jg  þj  ðA1Þ
systems if mixtures of a transport gas (e.g. H2) and a refrigerant T ∂x T ∂x T ∂x
(e.g. MeOH) are used.
Where J H2 and J g are molar gas fluxes, and μH2 ;T and μg;T are the
chemical potentials at constant temperature of the H2 and gas,
respectively, and ϕ is the electrical potential. Effects from elec-
Nomenclature
trode activation and reaction kinetics at the electrodes are not
specifically included here, however, in the discussion section the
F Faraday constant, C mol-1
effects are included empirically in some examples.
Lij phenomenological transport coefficients in appropri-
The chemical potential depends on the concentration and the
ate units
pressure through the relation
j electrical current density, A m-2

Jv volume flux, m s  1 1 ∂μi;T 1 ∂ μ i ðc Þ ∂p
 ¼ þV i ðA2Þ
Jw molar water flux, mol s  1 m  2 T ∂x T ∂x ∂x
J n ; J g ; J H2 Molar gas flux, mol s  1 m  2 where V i is the molar volume of species i. In the following it is
J pump dimensionless volume flow, dimensionless assumed that there are no electrolyte concentration gradients, and
jgen dimensionless current density, dimensionless only pressure gradients are considered and eq. (A1) can be
J comp dimensionless gas flow, dimensionless rewritten as:
p; Δp pressure (difference), Pa



P hyd ; P comp ; P el Hydraulic, compressor and electric power density, 1 ∂p 1 ∂p 1 ∂ϕ
σ e ¼ J H2  V H2 þ Jg  V g þj  ðA3Þ
W m3 T ∂x T ∂x T ∂x
R gas constant, J K  1 mol  1 If ideal gasses are considered, the molar volumes are the same
RL external load (electric or hydraulic), Ω or Pa s m  3 V g ¼ V H2 ¼ RT
p and
t s , t w , t n transference coefficient of salt, water and gas, dimensi-


1 RT ∂p 1 ∂ϕ
onless σe ¼ Jn  þj  ; ðA4Þ
T p ∂x T ∂x
T, ΔT temperature (difference), K
Vi molar volume of species i, m3 mol  1 where J n ¼ J H2 þJ g is the total molar gas flux. The entropy produc-
x, Δx distance parallel to fluxes, membrane thickness, m tion in an infinitesimal area section is σ e dx. The total entropy
β electrokinetic figure-of-merit, dimensionless production is obtained by integrating between the states that are
 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48 47

bounding the membrane. This assumes that we can anywhere find and (A11), we see that they have the same form and will result in
a state in equilibrium with the real membrane. The integration is slightly different definitions of the observable transport properties
carried out over the appropriate pressure and potential and the (Eqs. (17)–(22)).
outcome is:
Z Δx Z p2
 Z Δϕ

1 RT 1 11
σ e; tot ¼ σ e dx ¼ Jn  dp þ j  dϕ References
0 p1 T p ∂x 0 T ∂x




1 RT p 1 Δϕ [1] F.A. Morrison, J.F. Osterle, Electrokinetic energy conversion in ultrafine
¼ Jn  ln 2 þj  ðA5Þ
T Δx p1 T Δx capillaries, J. Chem. Phys. 43 (1965) 2111–2115.
[2] B.S. Kilsgaard, S. Haldrup, J. Catalano, A. Bentien, High figure of merit for
Where integration has been along a path of solutions in electrokinetic energy conversion in Nafion membranes, J. Power Sources 247
(2014) 235–242.
equilibrium with the membrane at any location. This equation is
[3] H.S. Chen, T.N. Cong, W. Yang, C.Q. Tan, Y.L. Li, Y.L. Ding, Progress in electrical
valid under isothermal conditions and it assumes that the gas energy storage system: a critical review, Prog. Nat. Sci. 19 (2009) 291–312.
compression/expansion is reversible. In this case the total internal [4] B. Saleh, G. Koglbauer, M. Wendland, J. Fischer, Working fluids for low-
energy is constant and W ¼ -q, where W is the work that the gas temperature organic Rankine cycles, Energy 32 (2007) 1210–1221.
[5] R.J. Gross, J.F. Osterle, Membrane transport charachteristic of ultrafine capil-
performs on the surroundings and q is the heat involved in the laries, J. Chem. Phys. 49 (1968) 228–234.
process. In order to maintain isothermal conditions under the [6] D. Burgreen, F.R. Nakache, Electrokientic flow in ultrafine capillary slits, J. Phys.
compression/expansion process, it must be coupled to a heat Chem. 68 (1964) 1084–1091.
[7] D. Burgreen, F.R. Nakache, Efficiency of pumping and power generation in
reservoir. In the present case it will be the membrane and its ultrafine electrokinetic systems, J. Appl. Mech. 32 (1965) 675–679.
coupling to the electrochemical cell and surroundings. Thus, the [8] J. Yang, F.Z. Lu, L.W. Kostiuk, D.Y. Kwok, Electrokinetic microchannel battery by
fulfilment of the isothermal condition depends strongly on the means of electrokinetic and microfluidic phenomena, J. Micromech. Microeng.
13 (2003) 963–970.
design of the mechanical set-up. [9] H. Daiguji, P.D. Yang, A.J. Szeri, A. Majumdar, Electrochemomechanical energy
The transport equations that follow from Eq. (A5) and describe conversion in nanofluidic channels, Nano Lett. 4 (2004) 2315–2321.
electrokinetic gas compression and expansion are: [10] F.H.J. van der Heyden, D.J. Bonthuis, D. Stein, C. Meyer, C. Dekker, Electro-



 kinetic energy conversion efficiency in nanofluidic channels, Nano Lett. 6
1 RT p 1 Δϕ (2006) 2232–2237.
J n ¼ Lnn  ln 2 þ Lnϕ  ðA6Þ
T Δx p1 T Δx [11] X.C. Xuan, D.Q. Li, Thermodynamic analysis of electrokinetic energy conver-
sion, J. Power Sources 156 (2006) 677–684.



 [12] Y. Ren, D. Stein, Slip-enhanced electrokinetic energy conversion in nanofluidic
1 RT p 1 Δϕ channels, Nanotechnology 19 (2008) 195707.
j ¼ Lϕn  ln 2 þLϕϕ  ðA7Þ
T Δx p1 T Δx [13] C. Davidson, X. Xuan, Effects of stern layer conductance on electrokinetic
energy conversion in nanofluidic channels, Electrophoresis 29 (2008)
where Lij are phenomenological transport coefficients (with app- 1125–1130.
[14] C. Davidson, X. Xuan, Electrokinetic energy conversion in slip nanochannels,
ropriate units) and the cross-term coupling coefficients are linked J. Power Sources 179 (2008) 297–300.
through the Onsager relation (Lnϕ ¼ Lϕn ). The coupling coefficient [15] R. Chein, H. Chen, C. Liao, Analysis of electro-kinetic pumping efficiency
in these equations measures the number of vapour molecules through finite-length nano-scale surface-charged capillaries, J. Electroanal.
Chem. 630 (2009) 1–9.
transported with the protons through the membrane.
[16] R. Chein, C. Liao, H. Chen, Electrokinetic energy conversion efficiency analysis
Eqs. (A6) and (A7) are derived under ideal gas assumption and using nanoscale finite-length surface-charged capillaries, J. Power Sources 187
that the composition of H2 and the second gas is the same in both (2009) 461–470.
[17] R. Chein, K. Tsai, L. Yeh, Analysis of effect of electrolyte types on electrokinetic
reservoirs. If at the same time it is assumed that J H2 is fully coupled
energy conversion in nanoscale capillaries, Electrophoresis 31 (2010) 535–545.
to the electrical current: [18] C.-C. Chang, R.-J. Yang, Electrokinetic energy conversion in micrometer-length
nanofluidic channels, Microfluidics Nanofluidics 9 (2010) 225–241.
1j
J H2 ¼ ðA8Þ [19] R. Chein, C. Liao, H. Chen, Thermodynamic reversibility analysis of electro-
2F kinetic energy conversion in nanofluidic channels, Nanoscale Microscale
Thermophys. Eng. 14 (2010) 75–94.
With the definition of J n , the total entropy production (Eq. (A5)) [20] M. Wang, Q. Kang, Electrochemomechanical energy conversion efficiency in
can be rewritten as: silica nanochannels, Microfluidics Nanofluidics 9 (2010) 181–190.




 [21] C.-C. Chang, R.-J. Yang, Electrokinetic energy conversion efficiency in ion-
1 RT p 1 Δϕ 1 RT p selective nanopores, Appl. Phys. Lett. 99 (2011) 083102.
σ e; tot ¼ J g  ln 2 þj   ln 2 ðA9Þ
T Δx p1 T ∂x T 2F Δx p1 [22] D. Gillespie, High energy conversion efficiency in nanofluidic channels, Nano
Lett. 12 (2012) 1410–1416.
For which the two following transport equations hold: [23] J. Hoffmann, D. Gillespie, Ion correlations in nanofluidic channels: effects of




 ion size, valence, and concentration on voltage- and pressure-driven currents,
1 RT p 1 Δϕ 1 RT p Langmuir 29 (2013) 1303–1317.
J g ¼ Lgg  ln 2 þ Lgϕ   ln 2 ðA10Þ
T Δx p1 T Δx T 2F Δx p1 [24] J. Catalano, A. Bentien, Influence of temperature on the electrokinetic properties
and power generation efficiency of Nafions 117 membranes, J. Power Sources




 262 (2014) 192–200.
1 RT p 1 Δϕ 1 RT p [25] R. Strobel, M. Oszcipok, M. Fasil, B. Rohland, L. Jorissen, J. Garche, The
j ¼ Lϕg  ln 2 þ Lϕϕ   ln 2 ðA11Þ
T Δx p1 T Δx T 2F Δx p1 compression of hydrogen in an electrochemical cell based on a PE fuel cell
design, J. Power Sources 105 (2002) 208–215.
These equations hold under the assumption in Eq. (A8), that is the [26] K. Onda, K. Ichihara, M. Nagahama, Y. Minamoto, T. Araki, Separation and
compression characteristics of hydrogen by use of proton exchange mem-
membrane is impermeable to  H2 and H2 does not diffuse through brane, J. Power Sources 164 (2007) 1–8.
the membrane. The last term  T1 2FRTΔx ln pp2 can be interpreted [27] C. Casati, P. Longhi, L. Zanderighi, F. Bianchi, Some fundamental aspects in
1
as a correction to the electric potential due to volume transfer by electrochemical hydrogen purification/compression, J. Power Sources 180
(2008) 103–113.
the electrode reaction. [28] K.A. Perry, G.A. Eisman, B.C. Benicewicz, Electrochemical hydrogen pumping
In the present article we analyse examples with H2O/MeOH, that using a high-temperature polybenzimidazole (PBI) membrane, J. Power
have gas permeabilities about three orders of magnitude higher than Sources 177 (2008) 478–484.
[29] S.A. Grigoriev, I.G. Shtatniy, P. Millet, V.I. Porembsky, V.N. Fateev, Description
H2 in Nafion membranes [39,45]. In this case H2 permeation is
and characterization of an electrochemical hydrogen compressor/concentrator
negligible and Eqs. (A10) and (A11) should hold. Still in the main part based on solid polymer electrolyte technology, Int. J. Hydrog. Energy 36 (2011)
of the article we use Eqs. (A6) and (A7) since these equations also 4148–4155.
apply for the case with pure H2 in both reservoirs. Eqs. (A10) and [30] B. Bahar, W. Grot, W. Parmelee, Electrochemical compression system.
WO2013106718-A1, 2013 Jul 18.
(A11) leads to an infinite electrokinetic figure-of-merit (100 % con- [31] D.P. Bloomfield, Electrochemically driven heat pump. US4593534-A, 1986 10
version efficiency). By comparison of Eqs. (A6) and (A7) to Eqs. (A10) Jun.
48 J. Catalano et al. / Journal of Membrane Science 478 (2015) 37–48
[32] R.B. Trusch, Joule-Thomson refrigeration cycle employing a reversible drive
electrochemical compressor. S5024060-A; EP463985-A; EP463985-A3;
EP463985-B1; DE69100584-E 463985-B463981, 1991 18 Jun.
[33] A. Bentien, T. Okada, S. Kjelstrup, Evaluation of nanoporous polymer mem-
branes for electrokinetic energy conversion in power applications, J. Phys.
Chem. C 117 (2013) 1582–1588.
[34] K.D. Sandbakk, A. Bentien, S. Kjelstrup, Thermoelectric effects in ion conduct-
ing membranes and perspectives for thermoelectric energy conversion,
J. Membr. Sci. 434 (2013) 10–17.
[35] S. Kjelstrup, D. Bedeaux, Non-equilibrium Thermodynamics of Heterogeneous
Systems, World Scientific, Singapore, 2008.
[36] J.F. Osterle, Unified Treatment of Thermodynamics of Steady-State Energy
Conversion, Appl. Sci. Res. Sect. A—Mechan. Heat Chem. Eng. Math. Methods
12 (1964) 425–434.
[37] C.E. Evans, R.D. Noble, C.A. Koval, A nonmechanical, membrane-based liquid
pressurization system, Ind. Eng. Chem. Res. 45 (2006) 472–475.
[38] K.A. Jeerage, R.D. Noble, C.A. Koval, Investigation of an aqueous lithium iodide/
triiodide electrolyte for dual-chamber electrochemical actuators, Sens. Actua-
tors B—Chem. 125 (2007) 180–188.
[39] T. Sakai, H. Takenaka, N. Wakabayashi, Y. Kawami, E. Torikai, Gas permeation
properties of solid polymer electrolyte (SPE) membranes, J. Electrochem. Soc.
132 (1985) 1328–1332.
[40] T. Okada, G. Xie, O. Gorseth, S. Kjelstrup, N. Nakamura, T. Arimura, Ion and
water transport characteristics of Nafion membranes as electrolytes, Electro-
chim. Acta 43 (1998) 3741–3747.
[41] S.K. Ratkje, M. Ottøy, R. Halseid, M. Strømgard, Thermoeletric power relevant
for the solid-polymer-electrolyte fuel-cell, J. Membr. Sci. 107 (1995) 219–228.
[42] V.M. Barragan, J.P.G. Villaluenga, M.R. Godino, M.A. Izquierdo-Gil, C. Ruiz-
Bauza, B. Seoane, Swelling and electro-osmotic properties of cation-exchange
membranes with different structures in methanol-water media, J. Power
Sources 185 (2008) 822–827.
[43] L. Chaabane, L. Dammak, V.V. Nikonenko, G. Bulvestre, B. Auclair, The
influence of absorbed methanol on the conductivity and on the microstruc-
ture of ion-exchange membranes, J. Membr. Sci. 298 (2007) 126–135.
[44] C.A. Edmondson, P.E. Stallworth, M.C. Wintersgill, J.J. Fontanella, Y. Dai,
S.G. Greenbaum, Electrical conductivity and NMR studies of methanol/water
mixtures in Nafion membranes, Electrochim. Acta 43 (1998) 1295–1299.
[45] R. Struis, S. Stucki, M. Wiedorn, A membrane reactor for methanol synthesis,
J. Membr. Sci. 113 (1996) 93–100.