Electrochimica Acta 51 (2006) 2743–2755

The effect of membrane thickness on the conductivity of Nafion

M.N. Tsampas, A. Pikos, S. Brosda, A. Katsaounis, C.G. Vayenas ∗
LCEP, Department of Chemical Engineering, Caratheodory 1, St., University of Patras, GR-26504 Patras, Greece
Received 8 April 2005; received in revised form 28 July 2005; accepted 10 August 2005
Available online 21 September 2005

Abstract
The conductivity of fully hydrated Nafion 112, 1135, 115 and 117 membranes was measured via ac impedance spectroscopy and steady-state
current–potential measurements both in symmetric H2 , Pt|Nafion|Pt, H2 and D2 , Pt|Nafion|Pt, D2 PEM cells and in H2 , Pt|Nafion|Pt, air and D2 ,
Pt|Nafion|Pt, air PEM fuel cells. In agreement with recent studies, it was found that the conductivity, σ, increases almost linearly with membrane
1/2
thickness L and also depends exponentially on potential and almost linearly on PH2 . These and other observations, including the strong isotope
effect obtained upon switching between H2 and D2 at the anode, show that the conductivity of Nafion contains two components, one due to proton
migration in the aqueous phase, the other due to proton tunneling between adjacent sulfonate groups in narrow pores. The observed near-linear
increase of σ with L is consistent with the proton tunneling mechanism but can also be explained by the existence of skin layers with lower
conductivity at the ionomer interfaces with the anode and cathode.
© 2005 Elsevier Ltd. All rights reserved.

Keywords: Nafion conductivity; Effect of thickness; Proton tunneling; Classical jump model; Isotope effect of conductivity

1. Introduction the experimental U–I curves of PEMFCs, including the appear-

ance of steady-state multiplicity.
The conductivity of fully hydrated Nafion membranes plays The purpose of the present work was to study and to rational-
a key role in the performance of PEM fuel cells [1–24]. Recent ize the experimentally observed increase of Nafion conductivity
work has shown that the conductivity of Nafion 117, which with membrane thickness, which has also been reported recently
is an increasing function of PH2 [24,25], also increases expo- by other workers [26–28].
nentially with cell potential [24,25] and almost linearly with
membrane thickness [24–28]. These striking features of the 2. Experimental
fully hydrated Nafion conductivity together with its low acti-
vation energy are difficult to interpret by classical models and A state-of-the-art PEM fuel cell was used (NuVant, Fig. 1)
require a thorough investigation of the proton transfer mecha- with state-of-the-art Pt and Pt–Ru electrodes (4 mg Pt/cm2
nism. In a recent paper [24], we have shown that the Nafion 117 (unsupported Pt black) on E-TEK carbon cloth and Pt (30%) Ru
conductivity contains two components, a potential-independent (15%) supported on Vulcan XC-72 carbon deposited by NuVant
component corresponding to proton conduction in the aqueous [20,21] on E-TEK carbon cloth (4 mg/cm2 , thus 1.8 mg/cm2
phase and a second component which increases exponentially metal basis)) in order to minimize any electrode overpotential
with potential and decreases by a factor of at least two when D2 losses [20,21,24]. In all experiments, the active Nafion elec-
replaces H2 as the fuel in diluted anodic mixtures [24]. In a more trolyte area was 23 mm × 23 mm.
recent paper [25], we have developed a first-principles macro- The Nafion 112, 1135, 115 and 117 samples conductivity was
scopic mathematical model, based on Kirckhoff’s law and on measured using both steady state current–overpotential mea-
the experimentally observed exponential dependence of Nafion surements as well as ac impedance spectroscopy (Solartron SI
conductivity on potential, which describes semiquantitatively 1255) combined with a Solartron 1286 electrochemical inter-
face. The anode and cathode feeds were continuously humidified
using thermostated gas saturators. The gases were air liquid cer-
∗ Corresponding author. Tel.: +30 2610 997576; fax: +30 2610 997269. tified (N4.5) H2 or 1% H2 in He mixtures or air liquide (N4.5)
E-mail address: [email protected] (C.G. Vayenas). D2 . The H2 or D2 feed could be diluted with ultrapure He to

0013-4686/$ – see front matter © 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.electacta.2005.08.021
2744 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755

Fig. 2. Nyquist plots of the symmetic H2 , Pt/Nafion 117/Pt, H2 cell with PH2 =
1 kPa (triangles) and PH2 = 100 kPa (circles). Frequency at semicircle top was
0.24 and 20 Hz, respectively; signal amplitude 5 mV.

itance (from 0.67 to 0.77 F), which is surprisingly high. The

origin of this very high capacitance, of the order of 0.5–0.8 F, is
addressed in Section 4.

3.2. The effect of membrane thickness and potential

As shown in Fig. 3a, at constant potential and PH2 , the con-

ductivity σ increases almost linearly with membrane thickness
L. Fig. 3b shows that the conductivity depends exponentially
Fig. 1. Schematic of the PEMFC geometry [20,21,24,25] and basic electrical on cell potential U. This leads to the appearance of steady-state
circuit showing the membrane electrode assembly (MEA) and the gas diffusion multiplicity with a maximum current for low PH2 values both
layers (GDL) at the anode and cathode.
in PEM fuel cells [24,25] (Fig. 4a), as well as in symmetric H2 ,
Pt/Nafion/Pt, H2 cells, as shown in Fig. 4b. The curves in Fig. 4
control the H2 or D2 partial pressure, PH2 or PD2 , at desired lev- come from the macroscopic model, termed γ-model [25], pre-
els. Constant current or potentials were applied using either the sented in references [24,25] and briefly presented in Section 4. It
Solartron Interface or an AMEL 553 Galvanostat–Potentiostat can be seen that the fit of the data to the γ-model is quantitative
in conjunction with a J. Fluke 893 A differential voltmeter. The in the case of fuel cell operation (Fig. 4a) while some deviations
conductivity of the Nafion 112, 1135, 115 and 117 membranes for low overpotentials are observed in the case of the symmetric
was measured at 25 ◦ C with both anode and cathode exposed cells (Fig. 4b).
to H2 /He and D2 /He mixtures with partial pressure PH2 or PD2 The surprising near-linear effect of membrane thickness on
varied between 0.3 and 100 kPa. In different sets of experiments, Nafion conductivity shown in Fig. 3a has been also reported in
humidified air was used at the cathode and the above humidified other recent experimental works [26–28] (Fig. 5) and has been
H2 /He and D2 /He mixtures were used only at the anode. Three predicted by the quantum mechanical proton tunneling model
parameters were investigated, i.e. PH2 , applied potential or cell (Eq. (5) in ref. [24]), as further discussed below.
potential and film thickness. As shown in Fig. 3a, upon plotting σ versus L one obtains a
straight line for every fixed potential and the σ-axis intersects
3. Results corresponding to L → 0, increase with potential (Fig. 6). The
increase can be described by:
3.1. H2 , Pt/Nafion/Pt, H2 cells σ(L = 0) ≡ σs (L = 0) = σs0 exp(U/γ) (1)
Fig. 2 shows two typical Nyquist plots obtained with no where the parameter γ seems to take two different values, a small
applied potential (U = 0) and with both sides of the fully hydrated one (47 mV) at small overpotentials and a larger one (1200 mV)
Nafion 117 membrane exposed to PH2 = 1 kPa and to PH2 = at higher overpotentials (Fig. 6).
100 kPa. Increasing PH2 causes a significant decrease in the total On the other hand, upon replotting both the literature [26–28]
resistance (from 0.67 to 0.1 ) and a small increase in the capac- and the present data as σ −1 versus L−1 , (Figs. 7 and 8) one again
 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755 2745

Fig. 3. (a) Effect of membrane thickness on the conductivity of fully hydrated

Nafion membranes exposed to 1.1 kPa H2 on both sides at 25 ◦ C at various
imposed potentials. (b) Corresponding effect of potential across the membrane Fig. 4. Typical I–U curves obtained with PH2 = 1 kPa at the anode and air
on the conductivity of fully hydrated Nafion 112, 1135, 115 and 117 membranes. (with Nafion 117 membrane) (top) or H2 –He (PH2 = 1 kPa) (bottom) at the
cathode showing the appearance of steady-state multiplicity and a maximum
current (except for the N112 membrane) at Urev − γ ≈ Urev − 0.43 V curves
obtains straight lines for large L values (low L−1 , solid lines from γ-model (Eqs. (17) and (18)) (top) and from modified γ-model (Eqs.
in Figs. 7 and 8), which all practically converge to the same (21) and (22)) (bottom) (γ 1 , γ 2 and Uc1 values for N112, N1135, N115 and
N117: γ 1 = −0.42, −0.225, −0.205 and −0.200 V; γ 2 = −1.9, −0.62, −0.61
conductivity value for L−1 → 0 (Figs. 7 and 8).
and −0.60 V; Uc1 = −0.42, −0.32, −0.2 and −0.2 V).
Consequently, the data can be described mathematically by
the following expression:
The parameter γ(V), which appears in Eq. (3) is an increas-
1 1 1 ing function of PH2 ([24], Fig. 9a) and, in fact γ −1 is a nearly
= + (L, U) (2)
σ σb σs −1/2
linear function of PH2 ([24], Fig. 9b). As shown in Fig. 9c, the
where σ b is thickness- and potential-independent and σ s (L, U) parameter γ can also be approximated very well by the expres-
is a linear function of L and an exponential function of potential: sion:

3/16
σs = σs0 L exp(U/γ) (3) PH2
γ = 0.572 V (4)
1 kPa
which is of the type reported recently [24,25] regarding the U
dependence and predicted recently [24] regarding the L depen- where PH2 is expressed in kPa. Eq. (4) can be derived theoreti-
dence. cally using the Debye–Huckel theory [29].
2746 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755

Fig. 6. Effect of potential on the σ intersects of Fig. 3a.

Fig. 11 compares the σ values obtained with H2 , Pt/Nafion

117/Pt, H2 and H2 , Pt/Nafion 117/Pt, air as a function of potential
and PH2 . One observes that the cathode gas has a rather small
effect on σ and on the slopes γ. The transition, however, from
σ s to σ b appears much more clearly in the curves obtained with
air at the cathode.

3.3. Effect of deuterium

Figs. 12–14 have been obtained by running air at the cathodic

side of the PEM fuel cell and H2 or D2 at the anode side. It
has been found recently [24,25] that PEM fuel cells operating
at low PH2 or PD2 values exhibit a strong isotope effect. The

Fig. 5. Effect of Nafion membrane thickness on membrane conductivity at dif-

ferent temperatures. (a) From refs. [28] and (b) [26].

As analyzed recently [24,25], one can assign the potential and

thickness-independent parameter σ b to the bulk proton conduc-
tivity in the aqueous phase of the Nafion membrane and the L-
and U-dependent conductivity σ s to proton migration between
adjacent sulfonate groups on the membrane walls (Fig. 10).
Upon combining Eqs. (2) and (3) one obtains:
σb σs σb σs0 L exp(U/γ)
σ= = (5)
σb + σ s σb + σs0 L exp(U/γ)
and
1 1 1
= + 0 (6)
σ σb σs L exp(U/γ)
It is worth noting that the change in the slope of the 1/σ ver-
sus 1/L lines at large 1/L values and very negative U values
(Figs. 7 and 8) is due to the previously discussed significant
change in the parameter γ observed between low and high over-
potentials (Figs. 3 and 6). Fig. 7. Effect of L−1 on σ −1 . Conditions as in Fig. 3.
 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755 2747

One can use these transients to compute the proton content,

CH
0 , of the membrane. Thus, a simple proton–deuterium mass
+
balance gives:
dI
CH
0
+V = −I/F (7)
dt
where V is the active membrane volume. Integration of Eq. (7)
gives:
 
−t I
I = Iin + (If − Iin ) exp VF (8)
CH
0
+

where Iin and If are the initial and final current values during the
transient.
Thus, the time constant, τ, of the system is CH 0 VF/Ī, where
+
Ī is the average current value ((Iin + If )/2 = 0.45 A) and from the
observed time constant of the system (60 s) and the active mem-
brane volume (2.3 cm × 2.3 cm × 185 ␮m = 9.8 × 10−8 m3 ) one
computes CH 0 ≡ τ Ī/VF = 2850 mol/m3 . This value is higher
+

than that of CH 0 = 900 mol/m3 corresponding to the proton

+
concentration resulting only from the protons produced by the
dissociation of the sulfonate groups of the Nafion membrane
[24,25], i.e. without accounting for additional protons produced
in the membrane by the dissociation of H2 at the anode. The fact
that the Nafion conductivity depends so strongly on PH2 (e.g.
refs. [24,25] and Figs. 2 and 11) implies that increasing PH2
increases the average proton concentration in the membrane, as
also supported by the transient experiments of Fig. 12 and the
above computed CH 0 value. Very recently [29], we have var-
+
ied systematically PH2 and confirmed the pronounced increase
in proton concentration with PH2 using the transient technique
of Fig. 12. We have thus found CH 0 values up to 4800 mol/m3
+
for PH2 = 100 kPa [29], i.e. increasing PH2 causes a significant
increase in the average proton concentration CH 0 . This, however,
+
in view of the overall electroneutrality requirement, raises the
question of where the compensating negative charge is stored.
Although the development of some delocalized negative charge
Fig. 8. Effect of L−1 on σ −1 . Conditions as in Fig. 5; from refs. [28] (a) and on the polymeric Nafion chains cannot be excluded, it is much
[26] (b).
more likely that the additional protons and charge-compensating
electrons are generated at the anode and stored at the very large
effect becomes smaller as PH2 and PD2 approach 1 atm. This is capacitor (supercapacitor [30]) in the three-dimensional anode-
because at low PH2 (or PD2 ) values the conductivity of Nafion, Nafion interfacial region (Fig. 13) between the highly porous
σ, is dominated by the surface tunneling proton conductivity carbon/Pt anode and the equally porous ionomer phase, via the
σ s [24,25]. Typical examples of switching the fuel cell feed reaction:
between H2 and D2 are shown in Fig. 12 for PEMFC operation at H2 → 2H+ (dl) + 2e− (dl) (9)
fixed external resistance Rex = 0.1 . The three sets of transients
shown in Fig. 12a–c, respectively, were obtained, as indicated where proton and electrons are stored on the opposing sides
in their time axis, consecutively and after having turned on the of that interface forming a double layer (dl). This three-
saturators at t = 0. This was done because it was anticipated, dimensional anode is prepared from hot-pressing of solutions
and partly verified by these experiments, that the isotope effect of suspended Pt/C catalyst and solubilized Nafion ionomer
would be more pronounced, in a partly dry membrane where and has a very high interfacial area between the Pt/C cat-
σ b would be smaller. Thus, indeed, one can observe in Fig. 12 alyst and the hydrated ionomer phase. The very large mea-
that the isotope effect becomes gradually less pronounced from sured capacitance values (∼0.7 F) allow for an estimation of
Fig. 12a–c as the membrane gradually becomes fully hydrated. the surface area of that interface and of the amount of charge
In general, as Fig. 12 shows, deuterium introduction at the anode stored there. Typical values of the area specific capacitance
causes typically a two-fold decrease in current and in potential formed between porous C or Pt and aqueous solutions is
and thus a four-fold decrease in power output. 30 ␮F/m2 [30]. Thus, one estimates a real surface area for the
2748 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755

−1/2
Fig. 9. Effect of PH2 on γ potential: (a) linear plot, (b) γ −1 vs. PH2 and (c) semilogarithmic plot.

anode of 1.7 × 104 cm2 per superficial cm2 of the PEM mem- It is also worth noting that protons generated by reaction
brane. For a surface atom density of 1015 atoms/cm2 (Pt) or (9) cannot in general be expected to remain localized only
4 × 1015 atoms/cm2 (C) this corresponds to 2.73 × 10−5 mol in the ionomer phase of the three-dimensional anode, i.e. at
Pt or 1.1 × 10−4 mol C. The metal loading at the anode the compact double layer between the anode and the hydrated
is 1.8 mg/cm2 , i.e. 9 × 10−6 mol metal/cm2 and that of C is Nafion membrane. This is due to the purely protonic con-
2.2 mg/cm2 , i.e. 1.8 × 10−4 mol C/cm2 . Thus, if a fraction 80% ductivity of Nafion membranes and the resulting spatial con-
(=1.8 × 10−4 mol/2.2 × 10−4 ) of the C atoms are surface atoms stancy of the electrochemical potential, µ̄H+ , of the protons
one can account for the very large measured capacitance, throughout the membrane which is valid both in symmet-
even neglecting the protons present at the Pt metal/solution ric H2 , Pt|Nafion|Pt, H2 cells but also in H2 , Pt|Nafion|Pt,
interface. O2 fuel cells as long as the purely ohmic overpotential is
On the other hand, at CH 0 = 900 mol/m3 the protons
+ small. The spatial constancy of µ̄H+ implies that an increase
resulting from the dissociation of the sulfonate groups are in proton concentration in the ionomer phase will cause a (not
1.67 × 10−5 mol H+ /cm2 of the 185 ␮m thick Nafion 117 mem- generally equal) increase in proton concentration throughout
brane, i.e. a factor of 7 smaller than those which can be stored the membrane. As already noted, this extension of the dif-
at the anode/membrane interface. Thus, it is entirely possible to fuse part of the anodic double layer in the membrane itself is
anodically generate and store at the anode/membrane interface not necessary for explaining the above high measured average
several (up to roughly seven) times more protons than those CH0 values.
+
resulting from the dissociation of the sulfonic groups of the In view of the above discussion one can explain why com-
Nafion membrane. This can explain why the estimated CH 0
+ mercial PEM units, such as the one used here, always utilize a
3
value (2850 mol/m ) is a factor of three higher than the nomi- porous C-supported anode, which in addition to efficient anodic
nal value of 900 mol/m3 corresponding to protons resulting only electrocatalysis, ensures a very high capacitance value, while a
from the, almost complete, sulfonic group dissociation. Pt-black electrode is usually sufficient at the cathode.
 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755 2749

Fig. 10. Physical model used to rationalize Eqs. (2) and (6) (top) and schematic of the Nafion molecular structure showing the ether bond bending and resulting
optimal conditions for proton tunneling [24].

Steady-state current–potential PEMFC curves are presented As already shown in Figs. 7, 8 and 15 and as also recently
in Fig. 14. They were obtained in the two time intervals between reported in the literature [24,25] the conductivity σ is described
the transients shown in Fig. 12a–c. The D2 -fueled PEMFC gives by
significantly lower currents at low potentials and exhibits steady-
state multiplicity, i.e. the current exhibits a local maximum, as σ = min(σb , σs ) (10)
described by the γ-model [24,25]. For current densities between or
600 and 950 A/m2 the ratio of the power outputs of the D2 - and σb σs
H2 -fueled PEMFCs is zero. σ= (11)
σb + σ s
Fig. 15 shows the Nafion conductivity values computed from
the I–U curves of Fig. 14, where activation and gas-phase mass- (both equations are compared with the data in Fig. 15) and the
transfer overpotentials are clearly absent. As expected [24,25], extracted σ max (=σ b ) and σ min (=σ s for U = 0) parameters are
D2 has a minor (15%) effect on the bulk conductivity, σ b , but also shown in Fig. 15. It is clear from Fig. 15 that the observed
causes a significant, near two-fold, decrease in σ s and in the strong isotope effect in Figs. 12 and 14 is primarily (∼85%) due
slope γ (Fig. 15). to the pronounced decrease in σ s upon replacing H2 with D2 .
2750 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755

Fig. 13. Schematic of PEM fuel cell showing the three-dimensional superca-
pacitive anode and cathode regions. See text for discussion.

Fig. 11. Effect potential or overpotential on the conductivity of Nafion 117.

(Open symbols) symmetric H2 , Pt/Nafion/Pt, H2 cell; (filled symbols) H2 ,
Pt/Nafion/Pt, air; T = 25 ◦ C.

Fig. 12. Transient response of cell potential and current to step changes between H2 and D2 at the anode (PH2 = PD2 = 0.7 kPa). Anode flowrate = 2.4 l/min; cathode
air flowrate = 0.2 l/min; (a), (b) and (c) were obtained consecutively as discussed in the text.
 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755 2751

4. Discussion

4.1. Summary of experimental observations

The main experimental observations regarding the very pecu-

liar conductivity behaviour of fully hydrated Nafion membranes
can be summarized as follows:

1. The conductivity increases almost linearly with membrane

thickness L (Fig. 3). This is valid both for conductivity mea-
surements performed in H2 –H2 cells as well as in H2 –O2 fuel
cells.
2. The conductivity, σ, contains two components. One, denoted
σ b , is potential-independent, the other, denoted σ s , increases
exponentially with potential U across the membrane:

σ = min(σb , σs ) (10)


U − Uc
σs = σb exp , U ≤ Uc ≤ Urev (12)
γ
σ = σb , Uc ≤ U ≤ Urev (13)

where Uc is the potential where σ = σ b . For any fixed U,

both σ b and σ s increase with PH2 (Figs. 2 and 11). The
non-linear conductivity component σ s dominates at low PH2
values (Fig. 11) and low potentials (Figs. 11 and 15). The
cathode gas has a rather small effect on the membrane con-
Fig. 14. Steady state I–U curves in PH2 = 0.7 kPa and PD2 = 0.7 kPa at ductivity (Fig. 11).
2.4 l/min; cathode air flowrate = 0.2 l/min. 3. The parameter γ is also an increasing function of PH2 and
this can be described mathematically in two ways. The first
way is via (Fig. 9b):
1/2
1/γ = 1/γmax + 1/(γmax kH PH2 ) (14)

with γ max = 1.72 V and kH = 0.64 kPa−1/2 , which shows that

γ can be expressed as:
1/2 1/2
γ = γmax ΘH+ , ΘH+ = kH PH2 /(1 + kH PH2 ) (15)

where ΘH+ can be associated with the fraction of sulfonate

groups with a (partially solvated) proton counterion, i.e. with
a proton localized within the immediate vicinity (Debye
length) of the sulfonate group [24,25,29] according to the
equilibrium

−RSO− + − +
3 + H
−RSO3 H (16)

as also discussed in Section 4.4.Alternatively, γ can be

expressed (Fig. 9c) as:

3/16
PH2
γ = 0.572 V (4)
1 kPa
The theoretical basis for the latter type of γ versus PH2 depen-
dence (Fig. 9c) is given elsewhere [29].
Fig. 15. Effect of potential across the membrane on the conductivity of Nafion 4. As a result of the above strongly non-linear component of
117. Anode PH2 (or PD2 )=0.7 kPa; cathode: air; open symbols: H2 ; filled sym- Nafion conductivity, the I–U curves are described by the γ-
bols: D2 . model described in refs. [24,25], i.e. by (Fig. 4):
2752 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755


σb S U − Uc obey Ohm’s Law, where the transport of the charge carrier (elec-
I= (Urev − U) exp , U ≤ Uc ≤ Urev trons in the case of certain oxides [31], ions in the case of some
L γ
(17) glasses [32]) occurs via thermally activated classical jumps of
length j in the material. The current density, i, is then given by
[31]:
σb S

I= U, Uc ≤ U ≤ Urev (18) e0 j U
L i = 2FCH+ j κd sinh (23)
2kb TL
where S is the active electrolyte surface area and Urev is the
open-circuit potential. where κd is the jump frequency.
These equations (with Urev ≈ 0.9 V for the H2 –O2 fuel cell This leads to Ohm’s Law for:
and 0 V for the H2 –H2 cell, where the current I is taken pos- e0 j U/2kb TL  1 (24)
itive when it leads to H2 oxidation at the anode) predict that
the current I is maximized at U = Urev − γ. and otherwise to:


e0 j U
U(Imax ) = Urev − γ (19) i = FCH+ j κd exp (25)
2kb TL
The maximum current value is given by

 For such a material, where i does not depend linearly on the field
σb S Urev strength U/L, the conductivity, σ, is given by:
Imax = γ exp −1 (20)
L γ

iL L e0 j U
σ= = FCH+ j kd sinh (26)
The above expression shows that Imax is maximized when U U 2kb TL
γ = Urev .
As shown in Fig. 4, agreement of the above model (Eqs. The conductivity is predicted to decrease monotonically with
(17) and (18)) is quantitative for the case of H2 –O2 (i.e. L (Fig. 16) and for large L values, i.e. for small values of the
regular PEMFC) operation (Fig. 4a) but not satisfactory for parameter (e0 j U/kb TL), to approach a constant value corre-
low overpotentials in the case of symmetric H2 –H2 operation sponding to the ohmic law limit (Fig. 16):
(Fig. 4b, continuous curve). This is because of the existence FCH+ 2j kd e0
of two γ values in the case of H2 –H2 operation (Fig. 6). Defin- σ= (27)
2kb T
ing by Uc1 the potential at which the γ-slope changes (Fig. 6)
and noting that in the case of H2 –H2 operation Uc = Urev , one This is clearly (Fig. 16) not the experimentally observed con-
can appropriately modify the above γ-model to obtain: ductivity dependence on L, which is increasing and near linear,

 exactly as Eqs. (3) or (5) describe in agreement with the predic-
σb S U − Urev
I= (Urev − U) exp , Uc1 ≤ U ≤ Urev tions of the proton tunneling model [24,25], i.e. (Fig. 16).
L γ1
(21) σ = C∗ L (28)
where C* is a constant.



σb S Uc1 − Urev U − Uc1
I= (Urev − U) exp exp ,
L γ1 γ2
U ≤ Uc1 (22)
As shown in Fig. 4b (dashed curves continued by the contin-
uous curves), the above modified γ-model (with the Uc1 , γ 1
and γ 2 values given in the Fig. 4 caption) provides a satis-
factory fit to the experimental data.
5. The capacitance of a Pt, C/Nafion/C, Pt cell with a
23 mm × 23 mm 117 Nafion membrane is very high, i.e. of
the order of 0.7 F (Fig. 2). As already noted, this very high
capacitance value is to a significant extent due to the superca-
pacitor properties of the interface between the porous Vulcan
XC-72 C-supported anode and the ionomer phase [30].

4.2. Classical jump model considerations

It is at first tempting to try to interpret the observed expo-

nential potential dependence of the non-linear (σ s ) component
of Nafion conductivity (Eq. (3)) on the basis of the hyperbolic Fig. 16. Comparison of the experimental points (from Fig. 3a) with the predic-
sinus I–U dependence of certain materials [31,32] which do not tions of the classical jump theory and of the proton tunneling model.
 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755 2753

Furthermore, by comparing the exponential terms of Eqs. (3) some internal process in the material, which is not affected to
and (23) one obtains any significant effect by the field U/L.
kb TL
j = (29) 4.4. Proton tunneling mechanism
e0 γ
which for the experimentally observed γ values (0.3–1.5 V) As already noted, the strong isotope effect as well as the
for L = 185 ␮m (Nafion 117) gives j values between 72 and strong effect of potential, PH2 and thickness on the conductiv-
14.5 ␮m, i.e.: ity of Nafion membranes is consistent with a proton tunneling
mechanism for the transport of protons between adjacent sul-
14.5 ␮m < j < 72 ␮m (30) fonate groups on the wall surfaces of Nafion membranes. Each
Such long proton jumps are, of course, not meaningful. In sum- proton, including those in the immediate vicinity of RSO− 3
mary, the classical thermally activated proton jump approach groups, creates around it its own Debye–Huckel negatively
predicts a decrease of σ with L (Fig. 16) and also gives unrea- charged cloud [29,31], described by the Poisson–Boltzmann
sonably high j values and thus totally fails to describe the (PB) formulation [29,31] and its potential is screened by the
experimental observations. potential of the cloud. The proton can migrate via tunneling
to neighboring empty RSO− 3 sites which belong to that cloud
and are at a distance of the order of the Debye length, κ−1 , of
4.3. Implications of the linear σ versus L dependence
Nafion (∼0.4 × 10−10 m, depending strongly on the local H+
concentration). The tunneling distance is of the order of the pro-
The observed near-linear increase in conductivity with mem-
ton wavelength, which is again of the order of 0.2 × 10−10 m
brane thickness can be attributed to an inhomogeneous layered
for proton energies of the order of 1 eV. The proton tunneling
Nafion membrane with a high resistance, low conductivity, skin
probability, which determines the current density depends on
layer at the anode-ionomer and possibly cathode-ionomer inter-
the local environment of the proton in its PB cloud, which is
face [27]. Since the ionomer phase occupies typically 30–40%
practically unaffected by the field strength U/L. In fact, it can
of the volume of the three-dimensional anode-ionomer interfa-
be shown [29] that the tunneling current density is given by:
cial region (Fig. 13), it is indeed possible that the macroscopic
ionomer conductivity is significantly lower in that region than kb T
i= C + F∗T PT ΘH+ (1 − ΘH+ ) (34)
in the bulk of the membrane. h H
Since, however, the proton concentration is very high in this
where ΘH+ is the fraction of sulfonate groups with a proton
supercapacitor region, such a geometric skin effect can be antic-
counterion, i.e. with a proton localized in their DH cloud, ∗T the
ipated only if the proton concentration is uniform throughout
average distance between two adjacent sulfonate groups (con-
the membrane.
nected with the main polymer chain via flexible ether bonds
Although such a skin effect may play a significant role, there
(Fig. 10)) and the tunneling probability, PT , is given by the
is an alternative, and not mutually exclusive, explanation for the
Gamow equation [29,31]:
observed near-linear increase in σ with L which is valid even for
a homogeneous material. 2 / h)r [2m[E0 −E]]1/2
As already noted, any homogeneous material, where the local PT = e−(π T (35)
current density, i, is not proportional to the field strength U/L Both the tunneling distance rT and the energy barier E0 − E can
(where U is the potential applied across the material and L is it be computed using the Debye–Huckel theory [29]:
thickness in the direction of i) will exhibit thickness-dependent
conductivity. This is because, by the definition of σ, it is: rT = κ−1 v1/2 (36)
I L i E0 − E = e0 [Φv1/2 e−v
1/2
+ Φv] (37)
σ= = L (31)
U S U
where the potential Φ is given by:
where i = I/S and S is the cross-sectional surface area.
Thus, if Ohm’s Law is obeyed and me30
Φ= (38)
U ε 2 h2
i = i0 (32)
L and m is the proton mass. The dimensionless proton potential v
where i0 is a constant, one obtains from Eqs. (31) and (32) that is defined from:
σ = i0 , i.e. constant. But if Eq. (32) is not obeyed, then σ will be −Up
L-dependent. We concentrate now to the specific case where σ v= (39)
Φ
depends linearly on L, as is the present case for σ s :
Since i does not vary linearly with the field U/L it follows that
σs = σs0 L (33) σ will be proportional to L [29]:

Comparison of Eqs. (31) and (33) shows that Eq. (33) will kb T 0 L
σ= C + F∗T PT ΘH+ (1 − ΘH+ ) (40)
be satisfied if i/U is fixed to some constant value, due, e.g. to h H U
2754 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755

where U equals the open-circuit potential Urev for fuel cell oper- j length of classical jumps (Eq. (18)) (m)
ation and the applied potential U for symmetric H2 –H2 cells. L0 thickness of a nanobattery (m)
The derivation of Eqs. (34) and (40) on the basis of the PB for- N number of nanobatteries
mulation will be given elsewhere [29]. These equations enable NAv Avogadro’s number (6.022 × 1023 mol−1 )
one to compute a priori σs0 and σ s . For the purpose, however, of PH2 partial pressure (kPa)
rationalizing Eq. (3), i.e PT tunneling probability
R resistance ()
σs = σs0 L (41)
Rc resistance of a nanobattery ()
for any fixed overpotential U, it is sufficient to note that, as rT Tunneling radius (m)
already discussed, the current density, i, is determined by the S active membrane surface (m2 )
local environment of the proton (proton plus Debye–Huckel T temperature (K)
cloud) and by the concomitant tunneling probability PT which U potential (V)
do not depend on the field strength U/L. As shown elsewhere Urev reversible potential (V)
[29], Eq. (40) provides a semiquantitative fit to experimental V active membrane volume (m3 )
Nafion conductivity values. v dimensionless potential defined in Eq. (39)
V volt
5. Conclusions
Greek symbols
The conductivity of fully hydrated Nafion membranes
U overpotential or applied potential (V)
increases very significantly with potential, PH2 and membrane
Φ tunneling potential (Eq. (38)) (V)
thickness. These observations, together with the strong observed
ΘH + fraction of protonated sulfonate groups
isotope effect, suggest strongly a proton tunneling mechanism
ΨT proton potential at tunneling distance (V)
between adjacent sulfonate groups in narrow membrane pores.
ε (=4πε0 εr ), where ε0 is the permittivity of vacuum
The near linear increase in σ with L may also be due to a skin
(8.85 × 10−12 C2 /(J m)) and εr is the dielectric constant
effect. In this case the potential, PH2 and isotope effects inter-
in Nafion (Eq. (38))
pretation presented here applies to the skin region and not to the
γ characteristic membrane potential (Eq. (1)) (V)
entire membrane. Additional work will be necessary to deter-
κ−1 Debye length (m)
mine whether a skin conductivity layer exists and, if so, how
κd jump frequency (Eq. (18)) (s−1 )
much it extends in the Nafion membrane.
σ conductivity (S/m)
σb bulk conductivity (S/m)
Acknowledgments
σs surface conductivity (S/m)
τ time constant (s)
We thank Professor E. Smotkin and his company, NuVant,
for supplying us with the state-of-the-art PEMFC unit. We also
thank DuPont (Drs. Simone Arizzi and Raj Rajendran) for sup- References
plying us with the various Nafion samples and our reviewers for
helpful comments. [1] W. Vielstich, A. Lamm, H. Gasteiger, Handbook of Fuel Cells, Funda-
mentals Technology and Applications, Wiley, New York, 2003.
[2] H.H. Voss, D.P. Wilkinson, M.C. Johnson, V. Basura, Electrochim. Acta
Appendix A. Nomenclature 40 (1995) 321.
[3] T.V. Nguyen, R.E. White, J. Electrochem. Soc. 140 (1993) 2178.
[4] M.S. Wilson, S. Gottesfeld, J. Electrochem. Soc. 139 (1992) L28.
CH 0
+ proton concentration (mol/m3 ) [5] V.A. Paganin, C.L.F. Oliveira, E.A. Ticianelli, T.E. Springer, E.R. Gon-
C capacitance (F) zalez, Electrochim. Acta 43 (1998) 3761.
[6] M. Cappadonia, J.W. Erning, S.M.S. Niaki, U. Stimming, Solid State
CV volume-specific capacitance (F/m3 )
Ionics 77 (1995) 65.
E energy (J) [7] M. Eikerling, A.A. Kornyshev, J. Electroanal. Chem. 502 (2001) 1.
E0 energy barrier (J) [8] P. Commer, A.G. Cherstvy, E. Spohr, A.A. Kornyshev, Fuel Cells 2
F farad (2002) 127.
F Faraday’s constant (96,485 C/mol) [9] M.M. Mench, C.Y. Wang, J. Electrochem. Soc. 150 (2003) A79.
h Planck’s constant (6.626 × 10−34 J s) [10] J.F. Moxley, S. Tulyani, J.B. Benziger, Chem. Eng. Sci. 58 (2003) 4705.
[11] B. Andreaus, A.J. McEvoy, G.G. Scherer, Electrochim. Acta 47 (2002)
i current density (A/m3 ) 2223.
If final value of current (Eq. (8)) (A) [12] M. Ciureanu, S.D. Mikhailenko, S. Kaliaguine, Catal. Today 82 (2003)
Iin initial value of current (Eq. (8)) (A) 195.
Ī average value of current (Eq. (8)) (A) [13] M. Ciureanu, H. Wang, J. Electrochem. Soc. 146 (1999) 4031.
kb Boltzmann’s constant (1.3806 × 10−23 J/K) [14] Z. Qi, C. He, A. Kaufman, J. Power Sources 111 (2002) 239.
[15] J. Lee, C. Eickes, M. Eiswirth, G. Ertl, Electrochim. Acta 47 (2002)
kH adsorption coefficient (kPa−1/2 ) 2297.
L thickness of Nafion (m) [16] K. Scott, W.M. Taama, S. Kramer, P. Argyropoulos, K. Sundmacher,
∗T tunneling length (m) Electrochim. Acta 45 (1999) 945.
 M.N. Tsampas et al. / Electrochimica Acta 51 (2006) 2743–2755 2755

[17] B. Bahar, C. Cavalca, S. Cleghorn, J. Kolbe, D. Lane, M. Murthy, G. [26] F.N. Büchi, G.G. Scherer, J. Electrochem. Soc. 148 (3) (2001) 183.
Rusch, J. New Mater. Electrochem. Syst. 2 (1999) 179. [27] P. Dimitrova, K.A. Friedrich, B. Vogt, U. Stimming, J. Electroanal.
[18] T. Seiler, E.R. Savinova, K.A. Friedrich, U. Stimming, Electrochim. Acta Chem. 532 (2002) 75.
49 (2004) 3927. [28] R.F. Silva, M. De Fransesco, A. Pozio, J. Power Sources 134 (1) (2004)
[19] J.A. Collins, U. Stimming, in: A. Wieckowski, E. Savinova, C.G. 18.
Vayenas (Eds.), Catalysis and Electrocatalysis at Nanoparticles, Marcel [29] C.G. Vayenas, M. Tsampas, A. Katsaounis, J. Phys. Chem. B, submitted
Dekker Inc., New York, 2003, p. 531. for publication.
[20] R. Liu, E.S. Smotkin, J. Electroanal. Chem. 535 (2002) 49. [30] B.E. Conway, Electrochemical Supercapacitors: Scientific fundamentals
[21] B. Gurau, E.S. Smotkin, J. Power Sources 112 (2002) 339. and Technological Applications, Kluwer Academic/Plenum Publishers,
[22] M. Schell, J. Electroanal. Chem. 457 (1998) 221. New York, 1999.
[23] K.D. Kreuer, Solid State Ionics 136–137 (2000) 149. [31] J.O.M. Bockris, A.K.M. Reddy, M. Gamboa-Aldeco, Modern
[24] A. Katsaounis, S. Balomenou, D. Tsiplakides, S. Brosda, S. Neophytides, Electrochemistry, Kluwer Academic/Plenum Publishers, New York,
C.G. Vayenas, Appl. Catal. B: Environ. 56 (2004) 237. 2000.
[25] A. Katsaounis, S.P. Balomenou, D. Tsiplakides, M. Tsampas, C.G. [32] A. Schirmeisen, A. Taskiran, H. Fuchs, B. Roling, S. Murugavel, H.
Vayenas, Electrochim. Acta 50 (2005) 5132. Bracht, F. Natrup, Appl. Phys. Lett. 85 (11) (2004) 2053.