Protonic conductivity and fuel cell tests of nanocomposite membranes based on bacterial cellulose

Accepted Manuscript Title: Protonic conductivity and fuel cell tests of nanocomposite membranes based on bacterial cellulose Author: Tiago D.O. Gadim Francisco J.A. Loureiro Carla Vilela Nataly Rosero-Navarro Armando J.D. Silvestre Carmen S.R. Freire Filipe M.L. Figueiredo PII: DOI: Reference:

S0013-4686(17)30426-7 http://dx.doi.org/doi:10.1016/j.electacta.2017.02.145 EA 29010

To appear in:

Electrochimica Acta

Received date: Revised date: Accepted date:

1-11-2016 2-2-2017 26-2-2017

Please cite this article as: T.D.O. Gadim, F.J.A. Loureiro, C. Vilela, N. Rosero-Navarro, A.J.D. Silvestre, C.S.R. Freire, F.M.L. Figueiredo, Protonic conductivity and fuel cell tests of nanocomposite membranes based on bacterial cellulose, Electrochimica Acta (2017), http://dx.doi.org/10.1016/j.electacta.2017.02.145 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Protonic conductivity and fuel cell tests of nanocomposite membranes based on bacterial cellulose

ip t

Tiago D. O. Gadima, Francisco J. A. Loureiroa,+, Carla Vilelab, Nataly Rosero-

CICECO – Aveiro Institute of Materials, Department of Materials and Ceramic

us

a)

cr

Navarroa,++, Armando J. D. Silvestreb, Carmen S. R. Freireb, Filipe M. L. Figueiredoa,*

b)

an

Engineering, University of Aveiro, 3810-193 Aveiro, Portugal

CICECO – Aveiro Institute of Materials, Department of Chemistry, University of

M

Aveiro, 3810-193 Aveiro, Portugal

te

[email protected]

d

+ Present address

Nanotechnology Research Division, Centre for Mechanical Technology and Automation,

Ac ce p

Department of Mechanical Engineering, University of Aveiro, Aveiro, Portugal

++ Present address

[email protected]

Division of Applied Chemistry, Faculty of Engineering, Hokkaido University, 060-8628 Sapporo, Japan.

* Corresponding author: Filipe M. L. Figueiredo [email protected]

1 Page 1 of 29

Abstract The effect of the preferential orientation of supporting bacterial cellulose (BC) nanofibrils

ip t

on the conductivity of composite proton conducting electrolytes with poly(4-styrene sulfonic acid) (PSSA) is reported. Data obtained by impedance spectroscopy show that

cr

the in-plane conductivity at 40% relative humidity (RH) is more than half order of

magnitude higher than that measured through-plane, indicating significant discontinuity

us

of proton transport at the PSSA/BC interface. The difference becomes less than 20% in

an

nearly saturated conditions (98% RH), demonstrating the key role of water in ensuring proton transport through those interfaces. The negative impact of the conductivity

M

anisotropy in fuel cell performance is mitigated due to operation in wet conditions and fuel cell tests of PSSA/BC-based membrane electrode assemblies under humidified

d

hydrogen/air gradients at room temperature yield 40 mW cm-2 at 125 mA cm-2, which is

te

amongst the highest values reported for a biopolymer -based electrolyte. It also results from the presented investigation that conventional electrode preparation used for

Ac ce p

thermoplastic polymer electrolytes must be modified in order to ensure proper adhesion to BC-based MEAs necessary to lower polarization losses.

Keywords: bacterial cellulose, poly(4-styrene sulfonic acid), nanocomposite membrane, protonic conductivity, fuel cell

2 Page 2 of 29

1. Introduction The research associated with polymer electrolyte fuel cells is extensive, as emphasized by

ip t

a large number of original scientific papers, reviews and books [1–23]. However, only a small number of studies have addressed the use of alternative bio-based materials to

cr

replace the state-of-the-art perfluorinated sulfonic acid membranes (PFSA) [6]. Bacterial cellulose (BC) is a highly pure and crystalline form of cellulose with a tridimensional

us

nanofibrillar structure allying an intrinsic “green” character with excellent thermo-

an

mechanical stability and low hydrogen permeation [24]. Evans and co-workers were the first (in 2003) to recognise the potential of this biopolymer as an alternative to PFSAs

M

such as Nafion [12], and the interest has recently started to grow, either as a mechanical reinforcement for another phase of high protonic conductivity, e.g. polymers such as

d

poly(4-styrene sulfonic acid) (PSSA), Nafion®, poly(methacryloyloxyethyl phosphate)

te

(see [16,19,25,26] and references therein), simple mixtures with phosphotungstic, phosphoric or phyptic acids [11,14], and also as a single phase membrane grafted with 2-

Ac ce p

Acrylamido-2-methyl-1-propanesulfonic acid functional groups [10]. Recently, Bayer et al. reported fuel cells based on membranes of cellulose nanofibers (CNF) and cellulose nanocrystals (CNC), which like BC are highly crystalline forms of cellulose but, instead of the bacterial origin, are obtained by purification of vegetable cellulose fibre pulps in order to remove lignin or pectin impurities (in the case of CNFs), as well as to dissolve the amorphous regions (in CNCs) [27]. The authors report a protonic conductivity of ~10 µS cm-1 for CNF membranes at room temperature and 100 % RH, which is remarkably similar to what we have measured for BC [16], thus confirming the strong resemblance between CNFs and BC. CNCs are more conductive (~500 µS cm-1) due to the presence of sulfuric acid groups introduced during acid hydrolysis. In fact, the excellent mechanical 3 Page 3 of 29

properties of nano-crystalline cellulose are severely affected by the hydrolysis and the CNC membranes are brittle. In addition, they are easily dispersed in water [27], which may also preclude their long-term use in a fuel cell where water is the reaction product.

ip t

The overall performance of fuel cells based on those BC-based membranes is lacking behind other massively studied and developed materials such as Nafion. This is partly

cr

due to the significant electrode polarization, which is apparent in the I-V curves reported

us

in the various aforementioned cases, or may also result from the anisotropic microstructure of these membranes. The latter effect has not been studied so far, and yet

an

the strong alignment of the BC nanofibrilar layers with the plane of the membrane is likely to favour proton transport along that direction, whereas the layer-to-layer contact

M

resistance should lead to a lower conductivity through the plane of the membrane. Moreover, in the case of the BC composites, the distribution of the second conducting

d

phase is likely to occupy the BC interlayer space, thus displaying regions of preferential

te

orientation along the membrane plane with enhanced proton transport. This paper resumes our previous work on PSSA/BC composite membranes [16] to assess

Ac ce p

the effect of the preferential orientation of the host BC nanofibrils and of the PSSA bulk domains along the membrane plane on both the protonic conductivity and in the performance of air/hydrogen fuel cells, thus also demonstrating their technological application.

2. Materials and Methods

2.1. Chemicals and materials Sodium 4-styrene sulfonate (NaSS, ≥90%, Aldrich), potassium persulfate (KPS, 98%, Panreac), poly(ethylene glycol) diacrylate (PEGDA, Mn 258, Aldrich) were used as received without any further purification. Other chemicals and solvents were of

4 Page 4 of 29

laboratory grades. Bacterial cellulose (BC) (tridimensional network of nano and microfibrils with 10–200 nm width) in the form of wet membranes was produced in our laboratory using the Gluconacetobacter sacchari bacterial strain [28] and following

ip t

established culture procedures [29].

cr

2.2. Preparation and characterization of PSSA/BC membranes

Nanocomposite membranes of PSSA/BC were prepared according to the scheme in Fig.

H OH

H

O HO

H

HO O

OH H

H OH

O

HO

H OH

n

d

H

H

M

HO

an

us

1, following the procedure reported in our previous study.[16]

te

PSSA/BC membranes

Ac ce p

Bacterial Cellulose

n

SO3-Na+

Sodium 4-styrene sulfonate

SO3-Na+

Fig. 1. Scheme of fabrication of the nanocomposite membranes.

Briefly, wet BC membranes with 40% water content (~400 mg dry weight) were placed in Erlenmeyers stoppered with rubber septa and purged with nitrogen. Simultaneously, aqueous solutions of NaSS, KPS (1.2%, w/w relative to monomer) and PEGDA were prepared (Table 1) and transferred with a syringe to the Erlenmeyers containing the 5 Page 5 of 29

drained BC membranes. After the complete incorporation of the solutions, the reaction mixtures were placed in an oil bath at 70 ºC during 6 h. The nanocomposite membranes were then repeatedly washed with water and dried at 40 ºC in a ventilated oven for 12 h,

ip t

before being converted into the acid form by ionic exchanging with an aqueous solution of 0.5 M HCl for 24 h at room temperature. The acidic membranes were again washed

cr

with distilled water following the above mentioned protocol, dried at 40 ºC, and kept in desiccators until their use. All experiments were performed in triplicate and analysed in

an

us

the form of thin nanocomposite membranes.

Table 1. List of the prepared membranes with the corresponding nominal composition,

/WBC

WNaSS

BC

-

-

PSSA/BC_2

3.0

PSSA/BC_3

3.0

Nafion

-

[H+]

d

WPEGDA/

te

L / µm

WNaSS

120

Water

λ

mmol g-1 uptake / %

(nH2O/nSO3H)

-

100

-

0.2

45

1.91

154

44.8

0.4

120

2.25

170

42.0

-

100

0.90

54

33

Ac ce p

Membranes

M

thickness (in dry condition), ion exchange capacity, water uptake and hydration level (λ).

The ion exchange capacity (IEC) of the membranes was determined by the usual back titration method after immersion of the membranes in an aqueous 0.1 M NaCl solution for 24 h. The water uptake of the membranes was also assessed by usual procedures and expressed as the fractional increase of weight from a dry condition and after immersion in distilled water for 24 h. Details can be found in [16]. Scanning electron microscopy (SEM) analysis of the surfaces and cross-sections of membranes was carried out on a HR-FESEM SU-70 Hitachi microscope operating at 4 kV, which also allowed elemental 6 Page 6 of 29

analysis with an energy dispersive X-ray spectroscopy (EDS) Bruker QUANTAX 400 detector.

ip t

2.3. Protonic conductivity The protonic conductivity (σ) of the membranes was determined by impedance

cr

spectroscopy using an Agilent E4980A Precision LCR meter. The measurements were

us

carried out under variable temperature (40 to 100 ºC) and RH (20 to 98%) conditions in an ACS Discovery DY110 climatic chamber. The electrode configurations schematized

an

in Fig. 2 were used to assess the resistance across the membrane thickness (through-plane configuration) and along the membrane plane (in-plane configuration). A sample with an

M

area of approximately 1×1 cm2 was used for the through plane measurements, on which two circular silver electrodes (Agar Scientific silver paste) with ∼0.6 cm in diameter were

d

painted on opposite sides of the membrane. A rectangular piece with ca. 1.5×0.5 cm2 was

te

used for the in-plane measurements, here with two silver stripes separated by approximately 1 cm are used as electrodes. These membranes were placed in tubular

Ac ce p

sample holders designed to ensure fully exposure of the membrane surface to the surrounding atmosphere, while providing the necessary electrical contact between the sample and the LCR meter through four platinum wires. The impedance spectra were recorded between 20 Hz and 2×106 Hz with test signal amplitude of 100 mV (in some cases also with 500 mV), and analysed with the ZView software (Version 2.6b, Scribner Associates) in order to assess the ohmic resistance (R) of the membrane. The conductivity was then calculated using σ = L(RA)-1, where L is the gap between electrodes for in-plane measurements and the thickness of the dry membrane for the through-plane measurements, and A is the area of the cross-section of the membrane for in-plane setup and the area of the electrodes in the case of the through-plane setup (see Fig. 2 for

7 Page 7 of 29

details). The temperature dependence of the conductivity was fitted to the Arrhenius

an

us

cr

ip t

equation by simple linear regression.

Fig. 2. Schemes of the setups of the proton conductivity measured via through-plane and

M

in-plane configurations.

d

2.4. Preparation of Membrane Electrode Assemblies (MEAs) and fuel cell tests

te

A commercial Pt:C electro catalyst powder (nominally 40% Pt on carbon black, HiSPEC 4000, Johnson Matthey) was mixed with Nafion® ionomer suspension (1100 EW 20

Ac ce p

wt.%, D2021, Ion Power) and isopropanol in 1:3:4.5 weight proportions, respectively. This suspension was ultra-sonicated for 30 min and sprayed directly onto a gas diffusion layer (GDL, Sigracet 34BC, SGL Group) using a nitrogen spray gun. The Pt loading was adjusted to approximately 1 mg cm-2 (within ±0.05 mg cm-2 for the various cells prepared) by multiple spraying steps with intermediate drying and weighting. After drying overnight in air, a very small amount of the Nafion® dispersion was brushed on the catalysed surface of the GDL to improve the adhesion to the BC membranes, which proven to be difficult even under applied pressure and temperature due to the nonthermoplastic character of both BC and the cross-linked PSSA. Two 2.7×2.7 cm2 pieces of the catalysed GDL, corresponding to an electrode active area of 7.3 cm2, were then

8 Page 8 of 29

placed aligned on both sides of a membrane and compressed in situ (at 0.4 N m-1) between the interconnector plates (interdigitated flow field design, Pragma Industries) of the fuel cell setup. The system was then heated up to 80 ºC for 1 h to consolidate the

ip t

assembly before the measurements. Pure Nafion membranes were also prepared for comparison purposes. The procedure

cr

stars by drying the commercial Nafion dispersion (EW 1100) in order to remove the

solvent, and re-dissolving it in N,Ndimethylacetamide. This dispersion was then casted

us

onto a Petri dish with 5 cm in diameter, where the membrane was formed after

an

evaporating the solvent under ambient conditions. The membranes were subsequently dried at 100 °C for 5 h, and left at 60 °C overnight before being hot-pressed at 120 °C

M

under 10 MPa for 10 min, to improve their mechanical robustness. Lastly, the membranes were activated and cleaned with a series of sequential treatments in 3 vol.% H2O2 (1 h), 1

d

M H2SO4 (up to 3 h) and de-ionized water (1 h), always boiling. The results reported here

te

were obtained with a membrane with a thickness of 100 µm. Prior to test the cell performance, a run-in stage was carried out during 2 h at 0.6 V at

Ac ce p

room temperature by feeding H2 (25 sccm) humidified at room temperature through a bubbler containing distilled water, and atmospheric air at 50% RH (125 sccm). The performance of the single cell was evaluated, using the same gas fluxes and humidification conditions, through current density-potential curves measured using a PAR VersStat 4 potentiostat. The cell resistance was investigated during the polarization by electrochemical impedance spectroscopy (EIS) between 10 mHz and 1 MHz with an applied AC amplitude of 10–50 mV, using the EIS module of the potentiostat. Due to the current upper limit of the potentiostat, data above 1 A for the Nafion membrane (roughly corresponding to j ≥ 140 mA cm-2) were obtained using external loads switched

9 Page 9 of 29

manually. The polarization and resistance data presented were collected after 10 min of stabilization time, which was previously confirmed as sufficient to attain steady-state.

3.1. Composition and microstructure of PSSA/BC nanocomposites

ip t

3. Results and Discussion

cr

Mechanically and thermally stable nanocomposite membranes consisting of cross-linked PSSA supported on a BC tridimensional nanofibrilar network were easily obtained as

us

pointed out in our previous study [16]. The three membranes have IEC values (Table 1)

an

well above of the reference Nafion® (0.9 mmol [H+] g-1 [30]), thus confirming that a significant amount of the PSSA polyelectrolyte was effectively incorporated into the BC

M

matrix. The increasing IEC indicates an increasing fraction of PSSA in the membrane due to an increasing fraction of the PEGDA cross-linker. The high acid load of the materials

d

implies higher water absorption, and indeed the water uptake more than doubles the dry

te

weight of the membranes, again increasing with increasing PSSA fraction (Table 1). Despite the homogeneous distribution of the PSSA within the BC matrix confirmed by

Ac ce p

the EDS map of sulphur depicted in Figs. 3B and 3C, the microstructure of these membranes is highly anisotropic resulting from the alternating BC nanofibril layers and the PSSA phase (Fig. 3A). Since BC is a poor ionic conductor,[16] this particular phase distribution may have a negative impact on the ionic transport across the membrane.

10 Page 10 of 29

ip t cr us

an

Fig. 2. (A and B) SEM micrographs of the cross-section of membrane PSSA/BC_2 depicting the oriented PSSA layers due to the BC network, with (C) showing the EDS

M

map of sulphur demonstrating the homogeneous distribution of PSSA.

d

3.2. In-plane vs. through-plane protonic conductivity

te

The typical impedance spectra of membranes PSSA/BC_1, PSSA/BC_2 and PSSA/BC_3

Ac ce p

collected under variable RH with the through-plane configuration are shown in Fig. 4. The Nyquist plot of the membrane with lower PSSA content (PSSA/BC_1) at low RH conditions is dominated by one semicircle with amplitude corresponding to the membrane ohmic resistance (30% RH in Fig. 4A and 4D), and a capacitive tale at low frequency which is ascribed to the electrode impedance. The semicircle tends to disappear with increasing RH, with the spectrum at 98% RH showing only the electrode tale (Fig. 4C and 4F) and where the ohmic resistance of the sample corresponds to the interception with the real axis obtained by extrapolating to high frequency the electrode tale. There is an obvious and systematic decrease of the sample impedance with increasing RH, which confirms that it is mainly associated with protonic conduction. The trend in the shape of the spectra is analogous when observing the effect of increasing the 11 Page 11 of 29

PSSA content, with the semicircle progressively disappearing and a significant decrease

Ac ce p

te

d

M

an

us

cr

ip t

of the overall impedance until only the electrode tale is observed.

Fig. 4. Nyquist plots of the PSSA/BC membranes collected at 60 ºC and different RH%, with the through-plane configuration: A) 30%, B) 60% and C) 98% (D, E and F are the corresponding magnifications).

12 Page 12 of 29

The evolution of the spectra for the in-plane assembly is similar, as illustrated by Fig. 5 for the most conductive membrane (PSSA/BC_3) on a broader humidity range (from 5% to 98% RH). The spectra show an almost perfect semicircle at 5% RH (partly observed at

ip t

40% RH), whereas above 60% RH the membrane exhibits an impedance response dominated by the electrode response. It should be noticed, however, that the capacitive

cr

response observed for low RH is not due to the ionic relaxation (very low for the in-plane membrane geometry) but rather the result of the stray capacitance generated by the two

us

platinum current collector wires of the sample holder. Nevertheless, the membrane ohmic

an

resistance can be easily determined for a vast set of temperature and RH conditions enabling direct comparison of the through-plane vs. in-plane conductivity. Figure 6 is an

Ac ce p

te

d

M

Arrhenius plot showing such comparison for membrane PSSA/BC_3.

Fig. 5. Nyquist plots of membrane PSSA/BC_3 collected at 60 ºC under variable RH with the in-plane configuration (B is a zoomed view of A). 13 Page 13 of 29

0.0 -0.5 80% RH

98% RH

ip t

] -1 -1.0 m c-1.5 S / -2.0 [ g-2.5 o l

60% RH

40%

cr

RH

-3.5 2.7

2.8

2.9

3

3.1

3.2

3.3

an

2.6

us

-3.0

1000 T-1 / K-1

M

Fig. 6. Arrhenius plot of the through-plane (○) and in-plane (●) conductivities of membrane PSSA/BC_3 at RH of 40, 60, 80 and 98%. The straight lines are linear fits to

te

d

the Arrhenius model, whereas the dashed curves are fits to the VTF equation.

Ac ce p

The temperature dependence of both the through-plane and in-plane conductivities follows the typical Arrhenius behaviour for high RH, whereas for low RH the presence of a curvature recalls the Vogel-Tamman-Fulcher (VTF) equation, in agreement with our previous work [16]. The conductivity of these composites under humid conditions is comparable or even higher than PFSA membranes such as Nafion (Table 2), which is to some extent expected since PSSA is a polyelectrolyte with very high ion exchange capacity and thus potentially very high protonic conductivity. The estimated Ea values obtained for the through- and in-plane configurations at high RH conditions (80 and 98%) decrease with increasing RH in the range between 19 and 26 kJ mol-1. Values of this order indicate that the structural diffusion of protons is the predominant transport mechanism. At low RH, the VTF-type behaviour suggests that proton transport is likely 14 Page 14 of 29

to be assisted by migration of the segmental motion of cross-linked PSSA chains, in agreement with dynamic mechanical analysis data [16]. Following the afore-mentioned description of the impedance spectra, it can be seen that

ip t

the conductivity increases with increasing RH for both through-plane and in-plane measurements. Furthermore, the data for both electrode configurations are approximately

cr

parallel but with the in-plane arrangement yielding distinctly higher conductivity values

at low humidity, thus confirming the underlying hypothesis of this work. For instance, at

us

40% RH the in-plane conductivity is more than half order of magnitude higher than the

an

conductivity measured through the plane of the membrane.

In nearly saturated conditions with 98% RH, both sets of data converge (Fig. 6),

M

indicating that the resistive contribution of the BC layers becomes negligible when the membranes are almost fully hydrated. This suggests the formation of an extensive

d

network of highly conductive, hydrophilic domains across the BC/PSSA interface. The

te

though-plane conductivity at 40 °C in this humidity condition is nearly 0.2 S cm-1. Such high value is not surprising considering the high concentration of functional acid groups

Ac ce p

in the membranes containing PSSA.

A fuel cell in operando conditions below 90 ºC generates liquid water and the membrane is actually exposed to potentially 100% RH, which is higher than the conditions used in the impedance measurements. This suggests that the effect of the conductivity anisotropy should be negligible, as observed for the ex-situ conductivity measurements shown in Fig. 6. The conductivity in low humidity (∼50% RH) is considerably lower, but the attained few mS cm-1 (close to 10 mS cm-1) is still compatible with fuel cell operation, especially at temperatures close to 120 °C and 50% RH where the membrane benefits from the excellent viscoelastic stability provided by BC [16], even in comparison with short side chain PFSAs with higher glass transition temperature (e.g. Aquivion®).

15 Page 15 of 29

3.3. Fuel cell tests Figure 7 shows the MEAs based on pure BC and the PSSA/BC_2 nanocomposite membranes after the fuel cell tests. Worth noting is the conservation of transparency and

ip t

mechanical integrity of the membranes, which is an indication of stability, at least at room temperature. This may not be the case at higher temperature, where pure PSSA is

cr

well known to be prone to degradation on the long term under fuel cell operation

conditions. While one expects a potentially higher stability of BC and cross-linked PSSA,

us

the issue of the thermo-chemical stability of these materials at high temperature must be

an

studied in greater detail.

A

Ac ce p

te

d

M

B

Fig. 7. Snapshot of the (A) pure BC and (B) PSSA/BC_2 MEAs obtained after the fuel cell tests in H2-air showing no signs of degradation, but the (arrowed) groove caused by the from the rubber seal (the electroactive area is 2.7×2.7=7.3 cm2).

The typical polarization and power curves obtained at room temperature under humidified air-hydrogen gradients with a pure BC membrane and two of the most conductive PSSA/BC membranes (PSSA/BC_2 and PSSA/BC_3) are displayed in Fig. 8.

16 Page 16 of 29

The open circuit voltage (OCV) for the BC cell is in excess of 1 V, which is higher than what is generally reported for Nafion®-based MEAs. This might indeed be associated with the lower H2 permeability of BC when compared to Nafion®, as initially reported by

ip t

Evans et al. [12], and recently confirmed [27]. However, it is immediately apparent from Fig. 8 that the OCV decreases with increasing PSSA content in the membrane, very little

cr

for PSSA/BC_3, but with PSSA/BC_3 already reaching a value low enough (0.83 V) to have a negative impact on the fuel cell efficiency. This may indicate an increasing fuel

us

cross-over with increasing PSSA content, which is unlikely to be due to the intrinsic

an

permeation of cross-linked PSSA, according to reported data on MEAs based on PSSA [3,31]. The contribution of any major defect or crack is unlikely as these typically would

M

lead to even lower and unstable OCV readings, and indeed such defects were not detected by SEM or optical microscopy. A possible explanation can be related to the large

d

hydrated domains located at the PSSA/BC interface that were previously assumed to

te

explain the enhanced through-plane conductivity in wet conditions (Fig. 5). These localized regions represent domains of bulk-like water that could provide the path for the

Ac ce p

gas molecules to diffuse faster due to the higher hydrogen permeability of water. Electrical short-circuits between the anode and the cathode due to membrane thinning or existing pinholes could also explain the lowering of the OCV, although the highly developed nanofibrilar BC network and the relatively thick membranes contribute to decrease the likelihood of such defects. Either way, improvements should be possible upon further optimization of the PSSA/BC interface through the membrane composition and fabrication process. As expected, the low protonic conductivity of BC [16] severely limits the power output, attaining a maximum of 0.06 mW cm-2 at circa 0.1 mA cm-2 (Fig. 8a). This is confirmed by the very low conductivity of this membrane measured by impedance spectroscopy

17 Page 17 of 29

both ex-situ on the climatic chamber and in-situ under polarization, and also estimated from the slope of the linear part of the I-V curve, in all cases being of the order of few µS cm-1 (Table 2). It is interesting to notice that the performance of a MEA based on a

ip t

vegetable CNF membrane is about 10 times higher (0.79 mW cm-2 at 1.8 mA cm-2) [27]. However, this CNF membrane is 4 times thinner than our BC membrane (thus 4 times

cr

less resistive), and the respective cell was operated at 80 °C, when the membrane

conductivity is 4 to 5 times higher than at room temperature, both factors yielding higher

us

power output. The relative agreement between both sets of data is indeed remarkable,

an

considering the potential variability of membranes with biological origin. 0.07

a)

1.0 0.8

1.0

0.06

40

0.8

2 -

2 -

m c V 0.6 W/ m V / 0.4 P

30

0.2

10

M

0.05

50

b)

0.04

V 0.6 / V 0.4

0.03 0.02

d

0.2

0.01 0

te

0.0 0.05 0.1 j / mA cm-2

Ac ce p

0

1.0

20

0.0

0 0

0.15

m c W m / P

25

50

75 100 125 150 175 j / mA cm-2

40

d)

c)

0.8

30

V 0.6 / V 0.4

20

-2

m c W m / P

10

0.2 0.0

0

25

0

50 75 100 125 150 j / mA cm-2

Fig. 8. Typical polarization and power curves (no IR correction) obtained at 25 °C under humidified H2-air for membranes a) BC, b) PSSA/BC_2, c) PSSA/BC_3, and d) Nafion.

18 Page 18 of 29

Table 2. Comparison of the room temperature through-plane AC protonic conductivity measured ex-situ by impedance spectroscopy at 98% RH (σAC), the AC conductivity

ip t

(σAC,pol) and the membrane ohmic resistance (RAC,pol) estimated from the impedance spectra collected under polarization at V in range 0.6 V – 0.5 V, and the DC conductivity

cr

estimated from the linear part of the polarization curves in Fig. 8 (electrode area is

σAC / S cm-1

σAC,pol / S cm-1

BC

8.3×10-6

5.5×10-6

PSSA/BC_2

5.4×10-3

2.6×10-3

PSSA/BC_3

8.8×10-2

1.4×10-2

Nafion®

7.1×10-2

2.9×10-2

RAC,pol / Ω

1.7×10-3

2.7×10-1

5.9×10-3

1.3×10-1

1.1×10-2

4.6×10-2

σDC / S cm-1 3.5×10-6

3.40×102

d

M

an

Membrane

us

2.7×2.7=7.3 cm2 for all membranes).

te

The much higher conductivity of the PSSA/BC nanocomposites than pure BC enables much higher power outputs, with membrane PSSA/BC_2 yielding 41 mW cm-2 at 123

Ac ce p

mA cm-1 (Fig. 8b), and PSSA/BC_3 generating 32 mW cm-2 at 93 mA cm-1 (Fig. 8c). It is noteworthy the better performance of PSSA/BC_2 in comparison to PSSA/BC_3, despite the lower RAC,pol of the former (Table 2). This means that either the higher fuel cross-over or electrical short circuit in PSSA/BC_3 suggested by the lower OCV is also the main responsible by the lower performance of the MEA based on this membrane. One common feature to both PSSA/BC MEAs is the drop of around 0.25 V in the cell potential for low current, which indicates significant activation losses. As a consequence of the sluggish electrode kinetics, the AC conductivity of the BC-based membranes obtained ex-situ by impedance spectroscopy is consistently higher than the estimates

19 Page 19 of 29

obtained in-situ also by impedance spectroscopy under applied polarization, which in turn are higher than the DC estimates obtained from the I-V curves (Table 2). Nevertheless, the performance of these PSSA/BC cells is also higher than most of the

ip t

other cellulose-based membranes reported in the literature (often obtained in more favourable conditions – at higher temperature and feeding pure oxygen), including BC

cr

composites with phosphoric or phytic acids [14], 2-acrylamido-2-methyl-1-

propanesulfonic acid-grafted BC [10], and pure CNF or CNC [27]. MEAs based on

us

Nafion®/BC (1:1) composite membrane prepared using the same electrodes and using the same reactor configuration in our laboratory also have a lower performance (16 mW cm-2

an

at 40 mA cm-2) [25]. The exception is a Nafion®/BC (on a ratio 1:7, roughly

M

corresponding to a BC content od 12.5%) composite membrane with reported I-V curves denoting lower activation losses and depicting a maximum output of 100 mW cm-2 at 225

d

mA cm-2 [19]. One should note that these MEAs were assembled with similar cathode

te

composition (0.95 mg cm-2 Pt) but higher catalyst load on the anode (1.72 mg cm-2 30% Pt / 15% Ru) than in our case, which may lead to somewhat improved performance.

Ac ce p

Another factor contributing to the higher power output of the BC:Nafion®=1:7 composite from Jiang et al. is the higher conductivity of this membrane (∼50 mS cm-1 vs. 14 mS cm1

for PSSA/BC_3), which combined with an apparent lower thickness (≤75 µm according

to SEM images - the authors do not quote a number), leads to a potentially lower ohmic polarization. However, the poor performance of the pure Nafion membrane used as reference by Jiang and co-workers (less than 100 mW cm-2 at 200 mA cm-2) may suggest that above average polarization losses may be affecting their cells. In order to gain insight on the possible limitations imposed by the electrode kinetics on the performance of the PSSA/BC MEAs – in comparison to the intrinsic properties of the membrane, a reference MEA was prepared based on a Nafion® membrane casted from a 20 Page 20 of 29

commercial dispersion to a thickness of 100 µm and using the same electrodes. This MEA tested in the same laboratory set-up and in the same conditions can achieve more than 120 mW cm-2 at 300 mA cm-2 (Fig. 8d), which is clearly better than any of

ip t

BC:PSSA membranes, and also better than Jiang’s Nafion®-based composites and Nafion® reference. The latter confirms that the polarization losses in the data from Ref.

cr

[19] may indeed be significant. In comparison to the PSSA/BC membranes, the higher

us

current density achieved with Nafion is to some extent explained by the lower ohmic resistance of this membrane, as shown by the RAC,pol values in Table 2. The correction of

an

the resistance values by the geometric factor of the membranes indicates that the conductivity of Nafion measured under polarization cell is also higher by a factor of 2

M

than that of composite PSSA/BC_3. All these values measured in-situ are lower than those measured ex-situ in the climatic chamber, but the differences are distinctly greater

d

for the PSSA/BC membranes than for Nafion. This may be an indication of imperfect

te

adhesion and/or coverage of the BC-based membranes to the catalysed GDL, which is easier in the case of the silver electrodes used in the ex-situ studies.

Ac ce p

Besides the higher electrical resistance, poor electrodes may also lead to increased activation losses due to lower amounts of platinum in contact with the PSSA, which is the conducting phase. The IR-corrected cell voltage plotted as a function of the log (j) in Fig. 9 yields an apparent Tafel slope of 135 mV per decade for our Nafion reference. This value is likely to be slightly overestimated due the small number of data points in the low current density range, but it still in good agreement with the 120 mV per decade usually observed,[32] and it thus serves well the purpose of comparison with the BC-based MEAs. The latter have clearly higher slope, starting at 200 mV per decade for PSSA/BC_3, and further increasing to 280 mV per decade for PSSA/BC_2. It is

21 Page 21 of 29

impossible to confirm the Tafel linearity pure BC (Fig. 9a), but a regression on the first

te

d

M

an

us

cr

ip t

two points at low current confirms the trend, yielding ∼550 mV per decade.

Ac ce p

Fig. 9. Tafel plots for membranes a) BC, b) PSSA/BC_2, c) PSSA/BC_3, and d) Nafion, obtained after correcting the ohmic drop of the I-V curves shown in Fig. 8.

These results confirm that the comparatively poor performance of the PSSA/BC membranes is also attributable to the activation losses. One possible explanation for the higher Tafel slope may be that the Platinum surface that is active for the oxygen reduction reaction may have been covered by the Nafion® applied on the last step of the electrode preparation, thus slowing down the reaction. This layer was the solution to improve the adhesion of the catalysed GDL to the membrane, which was found quite difficult to achieve given the fact that nor BC nor the cross-linked PSSA are 22 Page 22 of 29

thermoplastic. This characteristic coupled to the very high Young modulus of the BCbased membranes severely limits the efficiency of the usual process used to assemble thermoplastic membranes to the GDL, which consists on applying pressure at a

ip t

temperature close or above the glass transition temperature of the material. Further and detailed studies of the electrodes are thus mandatory to pursue the development of this

cr

new type of bio-based membranes.

us

4. Conclusions

an

Membranes of cross-linked PSSA supported on BC display morphological anisotropy that originates protonic conductivity higher along the plane than through the plane of the

M

membrane, particularly at low humidity. The differences get much smaller near saturated conditions corresponding to 98% RH, probably due to the formation of an extensive

d

network of percolating hydrated domains along the PSSA/BC interface. Since the effect

te

of anisotropy decreases with increasing humidity, its negative impact in fuel cell performance is mitigated.

Ac ce p

Fuel cell tests under humidified hydrogen/air gradients at room temperature demonstrate the application of such membranes as proton separators for PEMFCs, despite the poor electrode kinetics limiting the overall power output to a maximum of ca. 40 mW cm-2 at 125 mA cm-2. This is still one of the highest values reported for a bio-based polymer electrolyte, but additional work is clearly needed to clarify the underlying reasons of the sluggish electrode performance - which seems to affect most published data on similar materials - in order to profit from the high levels of protonic conductivity of these biobased membranes.

Acknowledgments

23 Page 23 of 29

This work was developed within the scope of projects CICECO-Aveiro Institute of Materials, POCI-01-0145-FEDER-007679 (FCT Ref. UID/CTM/50011/2013) and UniRCell (Ref. SAICTPAC/0032/2015, POCI-01-0145-FEDER-016422), financed by

ip t

national funds through the FCT/MEC and when appropriate co-financed by FEDER under the PT2020 Partnership Agreement, and CelFuelCel (EXPL/CTM-

cr

ENE/0548/2012). The Portuguese Foundation for Science and Technology (FCT) is also acknowledged for a post-doctoral grant to C. Vilela (SFRH/BPD/84168/2012), and

us

research contracts under Investigador FCT to C.S.R. Freire (IF/01407/2012) and F.M.L.

an

Figueiredo (IF/01174/2013).

[1]

M

References

K.D. Kreuer, S.J. Paddison, E. Spohr, M. Schuster, Transport in proton

d

conductors for fuel-cell applications: Simulations, elementary reactions, and

[2]

te

phenomenology, Chem. Rev. 104 (2004) 4637–4678. doi:10.1021/cr020715f. N. Cele, S.S. Ray, Recent Progress on Nafion-Based Nanocomposite Membranes

Ac ce p

for Fuel Cell Applications, Macromol. Mater. Eng. 294 (2009) 719–738. doi:10.1002/mame.200900143. [3]

A. Kraytsberg, Y. Ein-Eli, Review of Advanced Materials for Proton Exchange

Membrane Fuel Cells, Energy & Fuels. 28 (2014) 7303–7330. doi:10.1021/ef501977k. [4]

H. Zhang, P.K. Shen, Advances in the high performance polymer electrolyte

membranes for fuel cells, Chem. Soc. Rev. 41 (2012) 2382–2394. doi:10.1039/c2cs15269j. [5]

M. Díaz, A. Ortiz, I. Ortiz, Progress in the use of ionic liquids as electrolyte

membranes in fuel cells, J. Memb. Sci. 469 (2014) 379–396. doi:10.1016/j.memsci.2014.06.033.

24 Page 24 of 29

[6]

M.E. Scofield, H. Liu, S.S. Wong, A concise guide to sustainable PEMFCs: recent

advances in improving both oxygen reduction catalysts and proton exchange membranes, Chem. Soc. Rev. 44 (2015) 5836–5860. doi:10.1039/c5cs00302d. P. Mustarelli, A. Carollo, S. Grandi, E. Quartarone, C. Tomasi, S. Leonardi, et al.,

ip t

[7]

Composite Proton-Conducting Membranes for PEMFCs, Fuel Cells. 7 (2007) 441–446.

[8]

cr

doi:10.1002/fuce.200700010.

E. Guilminot, R. Gavillon, M. Chatenet, S. Berthon-Fabry, A. Rigacci, T.

us

Budtova, New nanostructured carbons based on porous cellulose: Elaboration, pyrolysis

an

and use as platinum nanoparticles substrate for oxygen reduction electrocatalysis, J. Power Sources. 185 (2008) 717–726. doi:10.1016/j.jpowsour.2008.08.030. T. Soboleva, Z. Xie, Z. Shi, E. Tsang, T. Navessin, S. Holdcroft, Investigation of

M

[9]

the through-plane impedance technique for evaluation of anisotropy of proton conducting

d

polymer membranes, J. Electroanal. Chem. 622 (2008) 145–152.

[10]

te

doi:10.1016/j.jelechem.2008.05.017.

C.W. Lin, S.S. Liang, S.W. Chen, J.T. Lai, Sorption and transport properties of 2-

Ac ce p

acrylamido-2-methyl-1-propanesulfonic acid-grafted bacterial cellulose membranes for fuel cell application, J. Power Sources. 232 (2013) 297–305. doi: 10.1016/j.jpowsour.2013.01.047. [11]

J. Yang, D. Sun, J. Li, X. Yang, J. Yu, Q. Hao, et al., In situ deposition of

platinum nanoparticles on bacterial cellulose membranes and evaluation of PEM fuel cell performance, Electrochim. Acta. 54 (2009) 6300–6305. doi:10.1016/j.electacta.2009.05.073. [12]

B.R. Evans, H.M. O’Neill, V.P. Malyvanh, I. Lee, J. Woodward, Palladium-

bacterial cellulose membranes for fuel cells., Biosens. Bioelectron. 18 (2003) 917–923. doi:10.1016/S0956-5663(02)00212-9.

25 Page 25 of 29

[13]

J. Rooke, C. de Matos Passos, M. Chatenet, R. Sescousse, T. Budtova, S. Berthon-

Fabry, et al., Synthesis and Properties of Platinum Nanocatalyst Supported on CelluloseBased Carbon Aerogel for Applications in PEMFCs, J. Electrochem. Soc. 158 (2011)

[14]

ip t

B779–B789. doi:10.1149/1.3585744. G. Jiang, J. Qiao, F. Hong, Application of phosphoric acid and phytic acid-doped

cr

bacterial cellulose as novel proton-conducting membranes to PEMFC, Int. J. Hydrogen Energy. 37 (2012) 9182–9192. doi:10.1016/j.ijhydene.2012.02.195.

J. Song, F. Mighri, A. Ajji, C. Lu, Polyvinylidene Fluoride/Poly (Ethylene

us

[15]

an

Terephthalate) Conductive Composites for Proton Exchange Membrane Fuel Cell Bipolar Plates : Crystallization , Structure , and Through-Plane Electrical Resistivity, Polym.

[16]

M

Eng. Sci. 52 (2012) 2552–2558. doi:10.1002/pen.

T.D.O. Gadim, A.G.P.R. Figueiredo, N.C. Rosero-Navarro, C. Vilela, J.A.F.

d

Gamelas, A. Barros-Timmons, et al., Nanostructured Bacterial Cellulose-Poly(4-styrene

te

sulfonic acid) Composite Membranes with High Storage Modulus and Protonic Conductivity, ACS Appl. Mater. Interfaces. 6 (2014) 7864–7875.

Ac ce p

doi:10.1021/am501191t. [17]

F. Müller, C.A. Ferreira, D.S. Azambuja, C. Alemán, E. Armelin, Measuring the

proton conductivity of ion-exchange membranes using electrochemical impedance spectroscopy and through-plane cell, J. Phys. Chem. B. 118 (2014) 1102–1112. doi:10.1021/jp409675z. [18]

N.C. Rosero-Navarro, E.M. Domingues, N. Sousa, P. Ferreira, F.M. Figueiredo,

Protonic conductivity and viscoelastic behaviour of Nafion® membranes with periodic mesoporous organosilica fillers, Int. J. Hydrogen Energy. 39 (2014) 5338–5349. doi:10.1016/j.ijhydene.2013.12.197.

26 Page 26 of 29

[19]

G. Jiang, J. Zhang, J. Qiao, Y. Jiang, H. Zarrin, Z. Chen, et al., Bacterial

nanocellulose/Nafion composite membranes for low temperature polymer electrolyte fuel cells, J. Power Sources. 273 (2015) 697–706. doi:10.1016/j.jpowsour.2014.09.145. K.R. Cooper, Characterizing through-plane and in-plane ionic conductivity of

polymer electrolyte membranes, ECS Trans. 41 (2011) 1371–1380.

cr

doi:10.1149/1.3635668.

ip t

[20]

G. Hoogers, Fuel Cell Technology Handbook, CRC Press, 2003.

[22]

J. Zhang, PEM Fuel Cell Electrocatalysts and Catalyst Layers, Springer-Verlag

an

London, 2008. doi:10.1007/978-1-84800-936-3.

us

[21]

F. Barbir, PEM Fuel Cells: Theory and Practice, Elsevier, USA, 2013.

[24]

D. Klemm, F. Kramer, S. Moritz, T. Lindström, M. Ankerfors, D. Gray, et al.,

M

[23]

Nanocelluloses: a new family of nature-based materials, Angew. Chemie Int. Ed. 50

T.D.O. Gadim, C. Vilela, F.J.A. Loureiro, A.J.D. Silvestre, C.S.R. Freire, F.M.L.

te

[25]

d

(2011) 5438–5466. doi:10.1002/anie.201001273.

Figueiredo, Nafion® and nanocellulose: a partnership for greener polymer electrolyte

Ac ce p

membranes, Ind. Crops Prod. (2016). doi:10.1016/j.indcrop.2016.01.028. [26]

C. Vilela, T.D.O. Gadim, A.J.D. Silvestre, C.S.R. Freire, F.M.L. Figueiredo,

Polymer electrolyte composite membranes based on nanocellulose and a phosphorouscontaining polymer, Cellulose. (2016). [27]

T. Bayer, R. Selyanchyn, M. Nishihara, S. Fujikawa, K. Sasaki, S.M. Lyth, High

Temperature Proton Conduction in Nanocellulose Membranes: Paper Fuel Cells, Chem. Mater. (2016). doi:10.1021/acs.chemmater.6b01990. [28]

E. Trovatti, L.S. Serafim, C.S.R. Freire, A.J.D. Silvestre, C.P. Neto,

Gluconacetobacter sacchari: An efficient bacterial cellulose cell-factory, Carbohydr. Polym. 86 (2011) 1417–1420. doi:10.1016/j.carbpol.2011.06.046.

27 Page 27 of 29

[29]

S. Hestrin, M. Schramm, Synthesis of cellulose by Acetobacter xylinum - II.

Preparation of freeze-dried cells capable of polymerizing glucose to cellulose, Biochem. J. 58 (1954) 345–352. DuPontTM Nafion® PFSA Membranes, Prod. Bull. P-12. (2016).

ip t

[30]

https://www.chemours.com/Nafion/en_US/assets/downloads/nafion-extrusion-cast-

[31]

cr

membranes-product-information.pdf.

J. Yu, B. Yi, D. Xing, F. Liu, Z. Shao, Y. Fu, et al., Degradation mechanism of

us

polystyrene sulfonic acid membrane and application of its composite membranes in fuel

[32]

an

cells, Phys. Chem. Chem. Phys. 5 (2003) 611–615. doi:10.1039/b209020a. A. Parthasarathy, C. R. Martin, Investigations of the O2 reduction reaction at the

M

platinum/Nafion interface using a solid-state electrochemical Cell, J. Electrochem. Soc.,

Ac ce p

te

d

138 (1991) 916-921. doi: 10.1149/1.2085747.

Highlights •

Polymer electrolytes based on bacterial cellulose are produced.

•

Preferential protonic conductivity along the membrane plane is observed.

•

The conductivity anisotropy decreases in wet conditions.

•

Effect of anisotropy is mitigated in fuel cells running on humidified air/H2.

•

Amongst highest fuel cell performance for a biopolymer-based polyelectrolyte.

Graphical Abstract 28 Page 28 of 29

29

Page 29 of 29

d

te

Ac ce p us

an

M

cr

ip t