Unitized Bidirectional Vanadium-Air

Redox Flow Battery

Kombinierte bidirektionale Vanadium-Luft-Redox-Fluss-Batterie

von der Fakultät für Mathematik und Naturwissenschaften der Carl von Ossietzky

Universität Oldenburg zur Erlangung des Grades und Titels eines Doktors der

Naturwissenschaften (Dr. rer. nat.) angenommene Dissertation

von

Jan Bernhard grosse Austing

geboren am 12.06.1983 in Lohne (Oldb.)

angefertigt am

NEXT ENERGY EWE-Forschungszentrum für Energietechnologie e.V. in Oldenburg

Erstgutachter
Prof. Dr. Gunther Wittstock (Universität Oldenburg)

Zweitgutachter
Prof. Dr. Carsten Agert (Universität Oldenburg)

Tag der Disputation

05.04.2016
“...provides multi-megawatt energy storage solutions using –
and I have no idea what this is – vanadium redox fuel cells.
That’s one of the coolest things I’ve ever said out loud!”
(Barack Obama)
Abstract
The all-vanadium redox flow battery (VRFB) is a promising battery technology
for the compensation of the intermittent electricity production of renewable energy
generators due to its high cycle life, reasonable round-trip efficiency and fast response
time. However, the VRFB suffers from low energy density which can be a hindrance
in decentralized, residential storage installations. In contrast, the vanadium-air redox
flow battery (VARFB) promises a significant higher energy density (roughly doubled)
due to virtually eliminating one electrolyte tank.
In this thesis a new concept for a unitized (i.e. combined) bidirectional VARFB
was developed, investigated and optimized. The proposed setup comprised a two-
layered positive electrode to support both the oxygen evolution reaction (OER)
as the charging reaction as well as the oxygen reduction reaction (ORR) as the
discharging reaction of the positive half-cell. Side reactions were suspected to reduce
the efficiency. Thus, transfer processes through the membrane were investigated by
means of UV/Vis spectroscopy and inductively coupled plasma mass spectroscopy
(ICP-MS) of the electrolytes. The major part of the coulombic efficiency losses was
attributed to oxygen permeation through the membrane. However, a considerable
amount of the vanadium ions of the negative electrolyte underwent crossover during
cycling. As the vanadium amount in the negative electrolyte determines the capacity
of the VARFB, the vanadium crossover is crucial for its life-time. Thus, a layer-by-
layer (LbL) modification routine of the Nafion 117 (N117) membrane with multilayers
of polyethylenimine (PEI) and Nafion ionomer was developed to significantly reduce
vanadium crossover and oxygen permeation. As a result, the energy and coulombic
efficiency were increased substantially as well as the indicated life-time of the VARFB.

v
Zusammenfassung
Die Vanadium-Redox-Fluss-Batterie (VRFB) ist aufgrund der hohen Zyklenlebens-
dauer, der Energieeffizienz und der schnellen Ansprechzeiten ein vielversprechendes
System zur Kompensation der Fluktuation bei der Stromerzeugung durch erneuerba-
re Energien. Die geringe Energiedichte von VRFB ist jedoch ein Hemmnis für die
Anwendung als dezentrales Speichersystem. Die Vanadium-Luft-Redox-Fluss-Batterie
(VARFB) hingegen, bei der nur ein Elektrolyttank benötigt wird, verspricht eine
nahezu verdoppelte Energiedichte.
In dieser Dissertation wurde eine neuartige kombinierte bidirektionale VARFB
konstruiert, untersucht und optimiert. Um sowohl den Anforderungen an die Sauer-
stoffentwicklung beim Laden als auch an die Sauerstoffreduktion beim Entladen in der
positiven Halbzelle gerecht zu werden, wurde eine aus zwei Lagen bestehende positive
Elektrode entwickelt und eingesetzt. Als Ursache für verringerte Effizienzen im Be-
trieb der VARFB wurden Nebenreaktionen vermutet. Daher wurden Transferprozesse
durch die Membran mithilfe von UV/Vis-Spektroskopie sowie Massenspektrometrie
mit induktiv gekoppeltem Plasma der Elektrolytlösungen intensiv untersucht. Ein
Großteil der Ladungseffizienz-Verluste wurde durch Sauerstoff-Permeation verursacht.
Der ebenfalls stattfindende Durchgang von Vanadium-Ionen durch die Membran
verringerte hingegen die Lebensdauer der VARFB, da die Kapazität der Batte-
rie von der Menge an Vanadium-Ionen im negativen Elektrolyten abhängt. Aus
diesem Grund wurde ein Verfahren zur Beschichtung der Membran mit Polyethyle-
nimin und Nafion-Ionomer entwickelt, wodurch sich eine deutliche Reduktion des
Vanadium-Ionen-Durchgangs durch die Membran erreichen ließ. Als Folge zeigte sich
eine Verbesserung der Energieeffizienz und Ladungseffizienz der VARFB sowie eine
verbesserte Prognose bzgl. der Lebensdauer der VARFB.

vii
Contents

Abstract v

Zusammenfassung vii

1 Introduction and Motivation 1

1.1 Global Warming, Renewable Energies and Energy Storage Technologies 1
1.2 Outline of this Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Redox Flow Batteries 7

2.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 General Properties of Redox Flow Batteries . . . . . . . . . . . . . . 9
2.3 Types of Redox Flow Batteries . . . . . . . . . . . . . . . . . . . . . 16
2.4 Concepts for Increased Energy Density . . . . . . . . . . . . . . . . . 20

3 Vanadium-Air Redox Flow Batteries 23

3.1 Introduction and State of the Art . . . . . . . . . . . . . . . . . . . . 23
3.2 Contributions within this Thesis . . . . . . . . . . . . . . . . . . . . . 25
3.3 Author Contributions to Publications I-III . . . . . . . . . . . . . 29

4 Conclusion and Outlook 31

References 35

List of Abbreviations and Symbols 41

Publications 45
Publication I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Publication II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Publication III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

Erklärung I 89

ix
Erklärung II 91

List of Publications 93

CV 97

Danksagung 99

x
1 Introduction and Motivation

1.1 Global Warming, Renewable Energies and Energy

Storage Technologies
The influence of anthropogenic greenhouse gas emissions such as CO2 on terrestrial
global warming is widely consensus in scientific publications [1, 2]. In December
2015, 196 countries agreed in the United Nations conference on climate change in
Paris that the global temperature rise should be well below 2 °C in comparison
to pre-industrial levels to avoid dangerous climate change [3]. A recent report of
the United Kingdom’s national weather service indicates that the average surface
temperature of the year 2015 will be presumably 1.02 °C above the temperature
average of the pre-industrial era from 1850-1900 [4] (Fig. 1). That is, half of the
maximum acceptable temperature increase has already been reached in 2015.
Additionally, the global energy use will continue to rise: a growth of +33 % (based
on 2015) till the year 2040 is expected according to the World Energy Outlook 2015
of the International Energy Agency (IEA) [7]. The rising energy demand is linked to
increasing CO2 emissions due to combustion of fossil fuels. In 2040, the energy-related
global CO2 emissions will be 16 % higher compared to 2013 as predicted by the IEA
[7].
A promising approach to mitigate CO2 emissions related to electricity production
is to increase the share of renewable energies in electricity generation. However, the
production of renewable electricity generators such as photovoltaics (PV) and wind
turbines is fluctuating depending on the presence of insolation or wind. The matching
of electricity supply and demand within a grid can be increased via communication
between electricity appliances and sources (the so-called “smart grids”). Another
important element for the harmonization of electricity generation and consumption
are batteries. They are suitable to shift an excess of produced electricity to periods
with an intermittent shortage of electricity generation within storage time spans of
up to a few days. The battery is charged in periods with a surplus of electricity
production (Fig. 2). In turn, it is discharged in times when power generation of

1
Chapter 1 Introduction and Motivation

1 .2
g lo b a l a v e r a g e te m p e r a tu r e d if f e r e n c e
1 .0 9 5 % c o n f id e n c e in te r v a l

to 1 8 5 0 -1 9 0 0 a v e ra g e [° C ]
te m p e r a tu r e d iffe r e n c e
0 .8

0 .6

0 .4

0 .2

0 .0

-0 .2

-0 .4

1 8 6 0 1 8 8 0 1 9 0 0 1 9 2 0 1 9 4 0 1 9 6 0 1 9 8 0 2 0 0 0
y e a r
Figure 1: Global mean temperature difference from the 1850-1900 average [°C]
(data from [5], data set described in [6]).

photovoltaic devices and wind turbines does not meet the demand. Therefore,
electricity storage devices such as batteries are regarded as an essential element for
an increased renewable electricity production without risking the reliability of the
electricity supply [8–12].
Amongst the existing battery technologies, redox flow batteries (RFB) have a high
potential for balancing electricity generation and consumption in grids with a high
share of renewable energies. RFB properties are the scalability of power and energy
independent from each other, a high cycle life [14], fast response times and a good
round-trip efficiency [15]. The most developed redox flow battery is the all-vanadium
redox flow battery (VRFB). However, like all redox flow battery technologies the
VRFB suffers from low specific energy and low energy density (Fig. 3), i.e. the
amount of energy stored in the battery per unit volume and/or mass of the battery is
low. In most redox flow systems this is caused by limited solubility of active species
and therefore a high ratio of inactive materials (e.g. water, H2 SO4 ) to the active
materials (e.g. vanadium ions). For instance, the concentration of the electroactive
vanadium ions in common electrolytes employed in VRFB does not exceed 2 mol·L−1 .
While a low energy density and low specific energy is generally not a major concern
in stationary storage applications, especially the volumetric energy density can be
an obstacle for residential installations with limited space. Various approaches to

2
1.1 Global Warming, Renewable Energies and Energy Storage Technologies

+
cumulative power
load

photovoltaics

wind

06:00 12:00 18:00 00:00

time
Figure 2: PV and wind power vs. load on a typical summer day in Germany (based
on data from [13]).

increase the energy density of redox flow batteries were proposed in literature; e.g.
utilization of additives to increase the solubility of active species [16–20], application
of organic solvents or ionic liquids (IL) to increase the available potential window
[21] and concepts involving solid species such as metals [14].
Another promising approach to increase the energy density is the vanadium-air
redox flow battery (VARFB). The positive1 redox couple (VO+ 2 /VO ) is replaced
2+

by the couple O2 /H2 O which leads to an increased thermodynamic cell potential.

The oxygen is provided from the ambient air and the size of the H2 O container is
negligible small [22]. Therefore, the theoretical energy density and specific energy of
the VARFB are more than doubled compared to the VRFB.
At the beginning of the research for this thesis, there were only publications in
which a VARFB consisting of two different reaction units had been demonstrated
[23, 24]. Although this setup allows the individual tailoring of the reaction unit
1
In this thesis the half-cells of electrochemical cells (such as batteries) are denoted with the terms
“positive” and “negative” according to the potential of the electrodes versus each other. This
convention is independent of whether the half-cell is being charged or discharged. Likewise,
the liquid electrolyte in the positive half-cell is denoted as “positive electrolyte” and the term
“negative electrolyte” is used for the electrolyte in the negative half-cell compartment. Another
common naming in the field of batteries are the terms “anode” and “cathode” which are assigned
according to the discharging process and this naming is maintained for the charging process as
well. Despite of the fact that the terms “anode” and “cathode” are commonly used when dealing
with batteries, this nomenclature can lead to confusion. Thus, these terms are not used in this
thesis despite of publication I after which the potential ambiguousness was realized.

3
Chapter 1 Introduction and Motivation

500

energy density [Wh·L-1]

400
Li-ion
300

200 Na-S
VRFB Pb-acid
100

50 100 150 200 250

specific energy [Wh·kg-1]

Figure 3: Energy density [Wh·L−1 ] and mass-specific energy [Wh·kg−1 ] of different

battery technologies (based on data from [12]). Pb-acid = lead acid
batteries; Na-S = sodium sulfur batteries, Li-ion = lithium ion batteries.

concerning the opposite requirements of the reactions, the utilization of two reaction
units is at the expense of the weight, volume and costs of the system. Other reports
[25, 26] present and discuss only the discharge process of the VARFB which can also be
regarded as a vanadium-air fuel cell (VOFC). However, these systems represent only
an incomplete battery as the vanadium electrolyte has to be regenerated externally.

1.2 Outline of this Thesis

The scope of this work was to develop a new VARFB cell with a combined reaction unit
(i.e. a single reaction unit in which both charging and discharging can be conducted).
Furthermore, the performance of this system was investigated in order to identify
possible loss mechanisms. This analysis served as a starting point for the rational
optimization of the system. The results of this cumulative thesis are presented in
three articles (denoted as publications I-III) published in peer-reviewed journals.
The structure of the thesis is described briefly in the following. First, RFB are
introduced and discussed in general to provide an overview about the properties and
state of the art of this energy storage technology. Thereupon, the VARFB is reviewed
in detail and its properties are discussed. The own publications are described in
this context and their synopsis and interconnection are given. Additionally, the
contributions of the author to each publication are itemized. The thesis closes with
a Conclusion and Outlook chapter which summarizes and evaluates the outcome of

4
1.2 Outline of this Thesis

this work and discusses outstanding questions related to the VARFB and those that
came up during the course of the thesis. The own publications I-III are attached
at the end of this thesis.
In publication I (“Study of an unitised bidirectional vanadium-air redox flow
battery comprising a two-layered cathode”) a novel VARFB utilizing a combined reac-
tion unit with a two-layered positive electrode is described. The performance of this
system is investigated at different operation conditions. Transfer processes through
the membrane were hypothesized to cause a performance decay during operation and
were thus investigated in publication II (“Investigation of crossover processes in a
unitized bidirectional vanadium/air redox flow battery”). A combination of UV/Vis
spectroscopic and inductively coupled plasma mass spectroscopy (ICP-MS) analysis
of the electrolytes was applied to identify and quantify the crossover of vanadium
species through the membrane as well as the oxygen permeation. An approach to
improve the performance of the VARFB by the modification of the membrane is
presented in publication III (“Layer-by-layer modification of Nafion membranes
for increased life-time and efficiency of vanadium/air redox flow batteries”). Via
layer-by-layer (LbL) deposition of polyelectrolytes on the membrane, the transfer of
vanadium and oxygen through the membrane was reduced significantly which lead to
an improved efficiency of the VARFB and an indicated enhancement of the life-time.

5
2 Redox Flow Batteries

The history of RFB goes back to the 1950s. A patent was issued to Kangro in 1949
in which the storage of electrical energy is claimed in liquids with dissolved reducible
and oxidable species [27]. However, substantial analysis on RFB was not undertaken
before the 1970s. During that time, extensive research such as screening of suitable
redox couples was conducted at the National Aeronautic and Space Administration
(NASA) [15, 28].

In 1983, the polysulfide/bromine redox technology was invented. It was developed

and commercialized owing to the promising costs and abundance of the redox species
[15]. Another relevant hybrid RFB available on the market is the zinc/bromine RFB
which was developed in the 1980s [15].

A promising approach to mitigate capacity fading problems induced by ion crossover

is the employment of two redox couples which are based on the same element. The
most prominent example is the VRFB comprising the redox couples V2+ /V3+ and
VO+ 2 /VO . The VRFB was invented and developed in the 1980s by Skyllas-Kazacos
2+

et al. at the University of New South Wales [21]. This RFB system received most
attention in scientific research [19, 21, 29] and is the most successfully commercialized
RFB (commercialized from 1993 on) [14, 30]. However, the major drawback of RFB
in general is the low energy density which is also inherent in VRFB. Thus, limited
available space can obstruct their application.

Beyond that, numerous different redox flow concepts have been proposed and
studied [8, 14, 15, 19, 21, 28, 30, 31]. New RFB approaches include the utilization
of water-soluble organic redox active species such as anthraquinones [32, 33], redox-
active polymers [34] or other organic redox active species [35] and systems with new
metal-based redox couples such as all-iron [36] or all-copper RFB [37]. However, in
the following the focus will be laid on those types of redox flow batteries which have
overcome the status of laboratory research and have been tested in demonstration
projects.

7
Chapter 2 Redox Flow Batteries

2.1 Setup
In conventional battery systems the storage of electroactive material and its conversion
is in the identical place. In contrast, this is spatially separated in RFB. The conversion
between chemical and electrical energy occurs in the reaction unit while the (dissolved)
electroactive species are stored in external electrolyte tanks (Fig. 4). For operation
of the battery, the electrolytes are circulated through the reaction unit using pumps.
The setup of RFB is similar to that of fuel cells (FC), however, the latter are open
electrochemical systems (exchange of matter with the ambient surrounding) and
they are generally not electrochemically reversible [19, 31].

current collector
porous electrode
membrane

electrolyte – + electrolyte
tank tank

> <
pump reaction unit pump

Figure 4: Schematic of a RFB and the components of one cell of the reaction unit
(magnification).

The reaction unit (also referred to as “stack”) consists of several single cells
connected in series to obtain a higher output voltage. The basic parts of a single
cell are the current collectors or bipolar plates (BPP), the (porous) electrodes and
the electrically insulating ion exchange membrane which separates the two half-cells

8
2.2 General Properties of Redox Flow Batteries

from each other (Fig. 4). In case of RFB operation, the two electrolytes are pumped
through the porous electrodes, electrons can be exchanged between the active species
and the electrode surface and are conducted via the BPP to the outer circuit. Ions
such as H+ or SO2−
4 are transferred through the membrane (cation or anion exchange
membrane) to close the circuit.

2.2 General Properties of Redox Flow Batteries

2.2.1 Independence of Power and Capacity
One unique property of RFB is entailed by the modular setup: the decoupling of
power and capacity from each other. The numbers of cells in the stack and their
active area determine the power of the RFB whereas the volume of electrolyte and
the concentration of active species therein influence its capacity. Therefore, by scaling
the reaction unit and the electrolyte volumes in an appropriate manner, the RFB
can be designed to meet exactly the individual power and capacity demands. This is
especially of interest for those storage installations where power and capacity need to
be tailorable independent from each other to fulfill efficiently any given requirements.

2.2.2 Efficiency and Loss Mechanisms

2.2.2.1 Energy Efficiency

A common measure for the performance of an electrochemical storage device is

the energy efficiency ηE . It describes the ratio of the electrical energy that can
be extracted from the battery during discharging Wdischarge relative to the energy
brought into the system during charging Wcharge (Eq. 1). The electrical energy can
be obtained by integration of the product of the current I (t) and the voltage U (t)
over the time of the discharging or charging process.
´
Wdischarge I (t) U (t) dt
ηE = = ´discharge (1)
Wcharge charge
I (t) U (t) dt

The energy efficiency of common RFB systems is in the range of approx. 75-85 %
[38]. Additionally to the general internal losses caused by crossover, side reactions or
overpotentials of the reactions, the energy efficiency of RFB is lowered due to energy
consumption of external devices such as temperature control units and pumps which
are necessary for electrolyte circulation. In comparison to other common storage

9
Chapter 2 Redox Flow Batteries

techniques, the energy efficiency of RFB is competitive (Fig. 5).

1 0 0

L i-io n
[% ] 9 0
N a -S
E
 
e n e r g y e ffic ie n c y

8 0 P b -a c id R F B
L i- io n P H S

7 0
C A E S

6 0

5 0

Figure 5: Energy efficiencies of electricity storage technologies suitable for stationary

applications (data from [38, 39]).

2.2.2.2 Coulombic Efficiency

The coulombic efficiency ηC is the ratio of the charge during discharging Qdischarge
relative to the charge during charging Qcharge (Eq. 2).

´
Qdischarge I (t) dt
ηC = = ´discharge (2)
Qcharge charge
I (t) dt

A low value indicates that electrical energy consuming side reactions occur. For
instance, these can be gassing reactions such as hydrogen evolution (HER) at the
negative electrode or oxygen evolution reaction (OER) at the positive electrode.
Another well known loss mechanism is the crossover of active species through the
membrane which leads to self-discharging of the battery. Ions of the active species
pass the ion-conduction membrane and react with the active species of the other
half-cell. This leads to lowered coulombic efficiency. Additionally, this leads to a loss
of capacity in those RFB in which the two redox couples are not based on the same
element/molecule. The species that underwent crossover are in the “wrong” half-cell
of the battery and therefore not accessible for further charging/discharging.
Another parasitic phenomenon observed in RFB are the so-called shunt currents.
The intended electric and ionic current path can be described as follows. Ions are

10
2.2 General Properties of Redox Flow Batteries

exchanged between two half-cells of a cell which are ionically connected through the
membrane. Electric current is transmitted between the cells which are connected
electrically via BPP (Fig. 6). However, two half-cells with the same polarity are also
connected by the highly conductive electrolyte which is in the manifolds and channels
(Fig. 6). Thus, parallel to the intended current path, parasitic currents can flow
between two half-cells of the same polarity. Indeed, these shunt currents can severely
reduce the coulombic and therefore the energy efficiency [40–42]. For example, Yin
et al. [42] reported for a non-optimized 5-cell stack a decrease in coulombic efficiency
of 22.1 % compared to a single cell due to shunt currents (ηC,single cell = 95.3 %;
ηC,5 cells = 73.1 %). Nevertheless, with an optimized electrolyte flow channel design
with high electrical resistance (i.e. low cross section area) these currents can be
limited to make up only 2-5 % of the overall energy efficiency losses [38, 42]. In this
thesis, single cell setups were used. Therefore, it was not necessary to consider shunt
currents in the publications I-III.

_ + _ +
iionic ielectric iionic
reaction … …
unit

electrolyte … ishunt
inlet ishunt …

Figure 6: Schematic of the intended ionic current path iionic and electric current
path ielectric and the undesired path for the shunt currents ishunt via the
electrolyte inlet system. The example depicts a RFB with a cation ex-
change membrane (CEM). Shunt currents can be transmitted analogously
through the electrolyte outlet (not shown).

An advantage of RFB is that no self-discharge of the entire electrolyte should occur

when the electrolyte remains in the tanks and the valves to the tanks are closed [8].
Conventional batteries are subject to self-discharge and therefore not suitable for
storing energy for a long time period.

11
Chapter 2 Redox Flow Batteries

2.2.2.3 Voltaic Efficiency

The voltaic efficiency ηV obtained as the ratio of ηE to ηC (Eq. 3) describes the

average discharge voltage in relation to the average charge voltage. It reflects
resistances of the whole battery process such as ohmic cell resistances, charge transfer
resistances of the reactions or concentration polarization resistances. Thus, it is
strongly dependent on operational parameters such as temperature, current density
as well as pumping rates.

ηE
ηV = (3)
ηC

2.2.3 Energy Density and Specific Energy

The energy density is the electrical energy stored in the battery per unit volume.
It depends on the concentration of the active species, the number of transferred
electrons and the cell voltage [31]. Similarly, the (mass) specific energy is the energy
per unit mass of the system. For mobile applications, both the energy density and
specific energy need to be high to achieve lightweight devices (high specific energy)
whose interior is not completely possessed by battery cells (high energy density).
For stationary applications, in many cases the (mass) specific energy is not relevant.
However, the energy density is of importance especially in applications with limited
space.
The low energy density being inherent to most RFB types is an important field of
research activities. In most “true” (i.e. only dissolved redox species) RFB systems it
is less than 25 Wh·L−1 [21]. The low energy density and specific energy are caused by
two limitations in an aqueous electrolyte. First, the solubility limitations of the redox
species lead to a high mass and volume of inactive material in a RFB. For instance, a
positive electrolyte typically employed in a VRFB with the concentration of 2 mol·L−1
VOSO4 in 2 mol·L−1 H2 SO4 has a density of 1.283 kg·L−1 [43]. This means that 2 mol
or 101.8 g vanadium (as metal) are dissolved in 1.283 kg electrolyte, i.e. the active
metal accounts for only 7.9 mass-% of the overall electrolyte mass. Second, the water
splitting reaction limits the usable potential window. The thermodynamic half-cell
potential under standard conditions is E 0 = 0 V vs. standard hydrogen electrode
(SHE) for the HER and E 0 = 1.23 V vs. SHE for the OER. Due to the overpotentials
of HER and OER on several electrode materials, the practical potential window can
be expanded to a certain extent. For example, the HER is negligible on graphite

12
2.2 General Properties of Redox Flow Batteries

felt electrodes which are commonly used in VRFB up to approx. -0.5 V vs. SHE
[44]. The feasible potential window of aqueous electrolytes using graphite electrodes
(green area) and standard potentials of redox couples relevant for RFB are depicted
in Fig. 7.
Nevertheless, the maximal voltage achievable in aqueous solutions is approx. 2.0 V
[45], the highest discharge voltage in aqueous electrolytes was reported for a zinc-
cerium RFB as 2.1 V [46]. Substantial increase of the potential window is only
possible when using alternative solvents (Sec. 2.4). The low energy density of VRFB
is the major motivation for investigating the concept of a VARFB which is the topic
of this thesis.

H 2
e v o lu tio n O 2
e v o lu tio n
e
/F
2+
2+

2+
F e
/C
3+

2-

e
4+

3+
r

/F
3+
2
C
2-
/S

e
/V
r -
n

5+
2+

F e
/C
2+

/C
/Z

4+
2+

4
S

2 /B
V
/V

u
3+

e
Z n

B r
C

C
V

-1 .0 -0 .5 0 .0 0 .5 1 .0 1 .5 2 .0
0
E [V ] v s . S H E
Figure 7: Standard potentials of redox couples in aqueous solutions as well as
HER and OER on graphite electrodes (data from [47]; except S2− 4 /S2 :
2−

[29]). The green area represents the viable potential window on graphite
electrodes under consideration of overpotentials.

2.2.4 Cycle Life

The cycle life of a battery is the number of charge-discharge cycles that can be
conducted until the battery reaches a defined minimum performance. This end-of-life

13
Chapter 2 Redox Flow Batteries

criterion is in most battery applications a minimal capacity that is retained (e.g.

60 % [39]). The cycle life depends on the cycling conditions, e.g. the charge/discharge
current densities, the depth of discharge and the operation temperature [10]. In
contrast to other types of batteries, most RFB exhibit a very high cycle life. Li-ion
batteries have a rated cycle life of 4000-8000 cycles, Pb-acid batteries can be operated
up to 3000 cycles and RFB are reported to perform for more than 13000 cycles
[39]. In conventional batteries morphological changes occur during battery operation.
This can cause the loss of electrical contact between active material and current
collector which in turn results in a reduction of capacity. The origin is mechanical
stress that occurs in the electrode during charging and discharging. In a “true” (i.e.
only dissolved redox species) RFB the active species are present in dissolved state
and hence are not prone to capacity loss due to morphological changes of the active
material [38]. For the same reason, deep discharging of a RFB is unproblematic
whereas in most other battery types this causes instantly a capacity decay [8].

2.2.5 Operation Temperature

The RFB should not exceed certain temperature limits as the solubility of active
species in the electrolyte is temperature dependent. Therefore, some RFB require
an active temperature management system which negatively influences the energy
efficiency [21]. The operation temperature of a VRFB should not be lower than
10 °C (precipitation of V2+ and/or V3+ as oxides) and should not exceed 40 °C
(precipitation of VO+ 2 as V2 O5 ) [30]. An advantage of RFB is that the operation
does not require high temperatures as this is the case with Na-S batteries [15].

2.2.6 Safety
An important criterion for storage technologies is its safety. High-temperature
batteries such as Na-S batteries suffer from the ability to catch fire due to the
presence of highly reactive molten sodium and molten sulfur. In 2011, a Na-S battery
installed in Japan caught fire [48]. Similarly, several incidents with Li-ion batteries
have been reported [49] and highlighted the need of safety considerations. High
energy densities, thin separators and volatile, flammable solvents bear potential risks.
A single defective cell can create excessive heat (e.g. due to a short-circuit) and
provoke a so-called “thermal runaway” (self-accelerating process due to excessive
heat released by exothermic reactions) of surrounding cells which can lead to fire or
explosion of the battery pack [49].

14
2.2 General Properties of Redox Flow Batteries

RFB are intrinsically safe energy storage devices concerning fire or explosion events
which is originated from their setup and components. Aqueous solutions and the
high heat capacity thereof can act as cooling agent in case of heat release due to
unwanted reactions. The external storage of the active material makes a dangerous
side-reaction involving the total active material unlikely to happen [21].
However, many RFB employ acidic heavy metal containing electrolytes (e.g.
vanadium ion containing electrolytes in VRFB) and thus leaking of the battery or
spilling of the electrolyte could occur [38]. Due to environmental and health concerns,
heavy metals should not escape from the systems. Likewise, in RFB employing the
Br− /Br− 3 redox couple it needs to be ensured that no toxic Br2 is released. These
safety hazards can be tackled by sophisticated sealing strategies to avoid leakages
and an extra surrounding dike to prevent electrolyte from leaving the battery.

2.2.7 Costs
Besides technological features, the costs of a storage technology for stationary
application is generally the main criterion for market penetration. Common types of
cost are the investment costs per power [k$·kW−1 ], investment costs per capacity
[$·kWh−1 ] and the costs per capacity and per cycle number [$·(MWh·(total number
of cycles))−1 ]. The latter number also takes into account the cycle life of the specific
system, therefore it represents the costs per capacity over the lifetime of the battery.
Tab. 1 compares the costs of several energy storage devices.

Table 1: Costs of different electricity storage devices [39]. URFC = unitized regen-
erative fuel cell; PHS = pumped hydro storage; CAES = compressed air
energy storage.

system power costs capacity costs cycle costs

[k$·kW−1 ] [$·kWh−1 ] [$·(MWh·cycle)−1 ]
PHS 0.4-5.6 10-350 0.5-3
CAES 1.7-2.2 130-550 4-18
VRFB 3.2 900 70
Li-ion 3-4 600 150-200
Na-S 3.5 550 90-130
Pb-acid 4.6 130 150
URFC 17 >10000 200

The modular setup of RFB allows easy assembling especially in comparison to

other battery types. A cell-making process like in Li-ion batteries which can account

15
Chapter 2 Redox Flow Batteries

for a substantial amount of the final battery cell costs (e.g. the active material makes
up less than 50 % of the cell costs in Li-ion batteries [21]) is not necessary.
RFB are interesting especially for large capacities but moderate power because
costs per kWh do not increase linearly unlike in other battery technologies. The
comparison in Tab. 1 reveals that in terms of power costs and cycle costs RFB are
more expensive than PHS and CAES but cheaper than other storage technologies such
as Li-ion or Pb-acid batteries. However, PHS and CAES are subject to geological and
geographical restrictions. Thus, RFB can compete with existing storage technologies,
especially when distinguishing between initial investment costs and “life-time” costs.
Additionally, recent costs analysis of VRFB predict even lower investment costs as
the values given in Tab. 1. Crawford et al. [50] estimate the costs for a VRFB
with a capacity of 4 times the rated power to be less than 350 $·kWh−1 and predict
that optimization and economies of scale can reduce the investment costs down to
160 $·kWh−1 .

2.3 Types of Redox Flow Batteries

2.3.1 Iron-Chromium
The iron-chromium RFB was one of the first intensively investigated RFB. Judging
by cost and availability of the redox species as criteria, the Fe2+ /Fe3+ //Cr2+ /Cr3+
system was identified by researchers of the NASA as the most promising one [15].

charge
(−) : Cr3+ + e− −
←−
−−→
−− Cr2+ E 0 = −0.42 V
discharge

charge
(+) : Fe2+ −
←−
−−→
−− Fe3+ + e− E 0 = +0.77 V
discharge (4)

charge
cell : Cr3+ + Fe2+ −
←−
−−→
−− Cr2+ + Fe3+ ∆E 0 = 1.19 V
discharge

The iron-chromium RFB suffers from ion crossover through the membrane and
thereby induced capacity decay [21] as well as poor electrochemical reversibility of
the chromium reaction [30]. Additionally, the very negative standard potential of

16
2.3 Types of Redox Flow Batteries

Cr2+ /Cr3+ leads to a high rate of HER as a parasitic side reaction [21].
The problem of capacity decay due to crossover can be tackled by intentionally
premixing the chromium and iron containing electrolytes, however, this approach is
accompanied by the need of additional amounts of iron and chromium not usable for
energy storage [30]. For practical applications, the electrochemical reaction kinetics
of the chromium redox couple need to be enhanced by application of catalysts. Lead
and alloys of lead with noble metals are well suited for this purpose and these
materials additionally increase the overpotential for hydrogen evolution, thus leading
to reduced side reactions [19]. Remaining challenges are the really low specific energy
of the iron-chromium RFB of less than 10 Wh·kg−1 and the need of noble metals for
the catalyst alloys [30]. Nevertheless, the US company EnerVault recently installed
a 250 kW/1 MWh iron-chromium RFB in the USA [31].

2.3.2 Polysulfide-Bromine
The polysulfide-bromine RFB does not utilize any metals as redox active species but
abundant sulfur and bromine are the basis for the redox couples. This latter fact
and the high solubility of the redox couples make this RFB promising concerning
the capacity-related costs [29]. However, the reversibility of the reactions on cheap
carbon electrodes is quite low. Therefore, cobalt or nickel electrodes are used as
the negative electrode for enhancing the reaction rates of the redox couple S2−4 /S2
2−

[19]. The polysulfide-bromine RFB also suffers from capacity fading induced by
crossmixing of the electrolytes through the membrane [30]. Beyond that, formation
of poisonous Br2 and H2 S needs to be prevented [15].

charge
(−) : S2−
4 + 2e
−
−
←−
−−→
−− 2 S2−
2 E 0 = −0.27 V
discharge

charge
(+) : 3 Br− −
←−
−−→
−− Br−
3 + 2e
−
E 0 = +1.09 V
discharge (5)

charge
cell : S2−
4 + 3 Br
−
−
←−
−−→
−− 2 S2−
2 + Br3
−
∆E 0 = 1.38 V
discharge

The polysulfide-bromine RFB was developed for commercialization by the company

Regenesys Technologies Ltd. between 1991 and 2004 and a 1 MW test facility was

17
Chapter 2 Redox Flow Batteries

operated [30, 51]. The reported energy density for the system was 20-30 Wh·L−1 [30].
However, Regenesys stopped its activities in 2003 [31]. In a technological-economic
analysis of a polysulfide-bromide RFB the authors concluded that a significant
enhancement of the electrochemical rate constants is necessary to be able to operate
this storage technology economically [52]. Up to date, polysulfide-bromide RFB are
not commercially available.

2.3.3 Zinc-Bromine
The zinc-bromine RFB is a so-called hybrid redox flow system, i.e. not all species of
the redox couples are present in dissolved state [15]. During charging, the zinc cation
is reduced to elemental zinc metal in the negative electrode compartment. Due to
the involvement of a solid species, there is no limitation concerning solubility and
higher energy densities are possible compared to conventional redox flow batteries.
The theoretical specific energy is 440 Wh·kg−1 while in practical applications 65-
75 Wh·kg−1 have been demonstrated [30].

charge
(−) : Zn2+ + 2e− −
←−
−−→
−− Zn E 0 = −0.76 V
discharge

charge
(+) : 3 Br− −
←−
−−→
−− Br−
3 + 2e
−
E 0 = +1.09 V
discharge (6)

charge
cell : Zn2+ + 3 Br− −
←−
−−→
−− Zn + Br−
3 ∆E 0 = 1.85 V
discharge

Similar to the polysulfide-bromine RFB, precautions are required against the

release of toxic Br2 . In technical application, this issue is tackled by using complexing
agents and/or an additional organic solvent which stabilizes the bromide complexes
[31]. Another obstacle is the self-discharging due to bromine crossover and its
subsequent reaction with the zinc [30]. Nevertheless, an energy efficiency of 80 %
was reported for a zinc-bromine RFB [14]. The most important issue is the dendrite
growth of zinc during deposition which can lead to internal electrical shorting [21].
Despite of these challenges, the zinc-bromine RFB represents a relatively mature
technology and is produced and sold by companies, e.g. RedFlow Ltd. in Australia
and Primus Power in the USA [31].

18
2.3 Types of Redox Flow Batteries

2.3.4 All-Vanadium
Inter-crossover of species between the half-cells is a common issue in RFB because
it leads generally to a capacity decay and hence limited lifetime of the battery.
Employing redox couples based on the same element in different oxidation states,
ion crossover only leads to lowered cycle efficiency but not to capacity fading. As
such, the VRFB in which all active species are based on the element vanadium is a
very promising technology (Fig. 8).

V2+/ – + VO2+/
V3+ VO2+

electrolyte reaction unit electrolyte

tank tank
Figure 8: Schematic of the VRFB.

charge
(−) : V3+ + e− −
←−
−−→
−− V2+ E 0 = −0.26 V
discharge

charge
(+) : VO2+ + H2 O −
←−
−−→
−− VO+
2 + e + 2H
− +
E 0 = +1.00 V
discharge (7)

charge
cell : V3+ + VO2+ + H2 O −
←−
−−→
−− V2+ + VO+
2 + 2H
+
∆E 0 = 1.26 V
discharge

Unlike other RFB, the VRFB does not need any catalysts or special electrode
materials as both half-cell reactions proceed with acceptable rates on activated
carbon felts [30]. The absence of catalysts reduces costs and therefore enhances the
commercial viability of the system. Another advantage, particular in comparison to

19
Chapter 2 Redox Flow Batteries

zinc-bromine RFB, is the low gas evolution rate [30]. Gas evolution in one electrode
compartment leads to inequality of the capacities of the two electrolytes. As the
electrolyte with the lowest capacity is limiting the overall battery capacity, gas
evolution induces a capacity decay due to imbalance of electrolytes.
A disadvantage of the VRFB is the presence of vanadium in the oxidation state
+5. This vanadium species is corrosive which needs to be considered for the material
selection for VRFB [30]. The solubility of the vanadium species in the sulfuric acid
containing electrolyte is limited and temperature-dependent. As already described
above, the upper temperature limit is caused by precipitation of VO+ 2 as V2 O5 . One
of the main drawbacks of the VRFB is the low specific energy of 25-30 Wh·kg−1
[30] and energy density of 25-30 Wh·L−1 [8]. The VARFB which is the focus of
this thesis (publications I-III) aims to overcome the two main disadvantages of
the VRFB: The temperature window will be enlarged due to elimination of the
VO2+ /VO+ 2 couple and the theoretical energy density and specific energy is increased
substantially.

2.4 Concepts for Increased Energy Density

2.4.1 Non-Aqueous Electrolytes
Recent research trends in the field of RFB are the utilization of non-aqueous elec-
trolytes for enhancing the possible potential window [14, 21]. Ionic liquids (IL) for
example have the potential for electrolytes with increased energy density. However,
up to date the reported solubilities of redox couples in non-aqueous electrolytes are
quite low (<1.0 mol·L−1 ) [14, 53]. Additionally, non-aqueous RFB systems have low
coulombic efficiencies (< 90 %) and high electrolyte and membrane resistances [31].
Therefore, non-aqueous RFB systems do not yet compete with conventional, aqueous
systems in terms of energy density [21, 53]. In addition to the described limitations,
the use of non-aqueous solvents would substantially increase the costs of RFB [31]
and involve new safety issues (e.g. flammability of the solvents).

2.4.2 Additives
Common additives in many aqueous RFB electrolytes are acids to improve the
(ionic) conductivity and in some cases to prevent the dissolved active species from
precipitation. For instance, in VRFB sulfuric acid or mixed sulfuric acid and
hydrochloric acid electrolytes are employed [16–20] and in iron-chromium RFB

20
2.4 Concepts for Increased Energy Density

electrolytes hydrochloric acid is used [30]. However, concentrations of solutes in

aqueous electrolytes investigated for relevant RFB systems generally do not exceed
2.0 mol·L−1 [54]. Various additives were investigated for enhancing the solubility of
different redox couples with a focus on the VO2+ /VO+ 2 redox couples employed in the
VRFB [20, 55–59]. The stability of highly concentrated vanadium electrolytes can be
enhanced by additives such as methyl orange or polyvinyl alcohol [17, 19]. However,
the solubility of vanadium salts is limited to 2.5-3.0 mol·L−1 even with additives
[19, 31] Thus, the possible energy density using additives is not substantially high.
Additionally, additives can lead to new problems, e.g. possible evolution of poisonous
chlorine when using chloride additives or new side reactions when using organic
additives which can be oxidized by vanadium(V) [19].

2.4.3 Hybrid Redox Flow Batteries

Hybrid redox flow batteries are distinguished from conventional flow batteries in the
way that at least one of the involved redox active species is not present in dissolved
or liquid form but is a gas or a metal [21, 31, 38].
Hybrid RFB involving metal deposition such as the zinc-bromine RFB promise
higher energy densities compared to conventional RFB due to the occurrence of a
solid metal (high density) as active species. The most developed and commercially
available hybrid RFB is the zinc-bromine RFB [29]. However, the metal deposition
and dissolution which occurs in most hybrid RFB is challenging because of dendrite
growth [21] which can lead to failure due to short-circuiting. Additionally, the zinc-
bromine RFB suffers from high self-discharge and low energy efficiency as already
mentioned above [14]. Other examples for hybrid RFB involving metal deposition
are the flowing lead acid RFB [38], the all-copper RFB [31] or the all-iron RFB [19].
Alternative concepts for hybrid redox flow batteries are proposed with lithium as an
active species [19, 21]. Energy densities of up to 100 Wh·L−1 and specific energies of
up to 100 Wh·kg−1 are predicted for RFB containing solid/semi solid slurries [14].
However, these systems are all in the state of early fundamental research and far
away from commercial viability.
An example for a hybrid flow battery involving a gaseous redox active species
is the hydrogen-bromine RFB which has a cell voltage of ∆E 0 =1.09 V [19] and a
high theoretical specific energy (353 Wh·kg−1 [31]). However, the energy density (in
Wh·L−1 ) of the system strongly depends on the way of hydrogen storage (e.g. in
pressurized form). Other issues occurring in hydrogen-bromine RFB are the crossover
of bromine species [60], the oxidation of carbon electrodes and the dissolution of the

21
Chapter 2 Redox Flow Batteries

platinum catalyst [61].

In terms of energy density and specific energy, the VARFB represents the most
promising hybrid RFB using gaseous compounds. It is described and discussed in
detail in chapter 3 and is the subject of publications I-III.

22
3 Vanadium-Air Redox Flow
Batteries

3.1 Introduction and State of the Art

The VARFB can be regarded as a combination of the negative half-cell of a VRFB
(redox couple: V2+ /V3+ ) and the positive half-cell of a URFC (redox couple: O2 /H2 O).
It can also be considered as a URFC with vanadium species as “fuel”. From the
latter perspective, the VARFB has the advantage that no storage of hydrogen is
necessary as it is the case in a conventional (i.e. H2 /O2 ) URFC. Such a URFC has
a very low energy density if the hydrogen is stored at ambient pressure, whereas
compression of the hydrogen leads to energy losses. In the VARFB there is no need
for storing gas as oxygen is provided by ambient air. The reactions occurring in the
VARFB and their standard potentials are given in Eq. 8. The general setup of the
VARFB is depicted in Fig. 9.

O2

V2+/ – +
V3+
H2O

electrolyte reaction unit

tank
Figure 9: Schematic of the VARFB.

23
Chapter 3 Vanadium-Air Redox Flow Batteries

charge
(−) : 4 V3+ + 4 e− −
←−
−−→
−− 4 V2+ E 0 = −0.26 V
discharge

charge
(+) : 2 H2 O −
←−
−−→
−− O2 + 4 e− + 4 H+ E 0 = +1.23 V
discharge (8)

charge
cell : 4 V3+ + 2 H2 O −
←−
−−→
−− 4 V2+ + O2 + 4 H+ ∆E 0 = 1.49 V
discharge

The VARFB system has a significantly increased theoretical (i.e. thermodynamic)

energy density and specific energy in comparison to VRFB. The standard cell
potential of the VARFB is 18 % higher than that of the VRFB (∆EVRFB 0
=1.26 V
vs. ∆EVARFB = 1.49 V). Thus, the theoretical energy density and specific energy
0

are increased by 18 % assuming that mass and volume of the two systems are equal.
Furthermore, in the VARFB one electrolyte tank is virtually eliminated. The oxygen
consumed at the positive electrode during discharging is provided by the ambient air
and thus no storage is needed. The water split during charging occupies a negligible
volume. In relation to the volume of the negative electrolyte with a vanadium
concentration of cV = 1.5 mol·L−1 , the volume of water needed for OER is only
approx. 1.3 %1 . Assuming a VARFB with high capacity and low power and hence
small and light reaction unit in comparison to the volume and mass of the electrolytes,
the saving of one electrolyte tank yields almost a doubling of energy density and
specific energy. Hence, the overall theoretical energy density and specific energy of a
VARFB is about 2.39 times the one of a VRFB (∆EVARFB 0
= 1.18 · ∆EVRFB
0
; factor
two due to elimination of one electrolyte tank).
However, it should be mentioned that the practical energy density and specific
energy is lower than the theoretical values discussed above. The major reasons are
the high overpotentials of OER and oxygen reduction reaction (ORR). Therefore,
the practical cell potential will deviate from the thermodynamic value of 1.49 V.
Nevertheless, the practical energy density and specific energy will increase significantly
due to the massive reduction of electrolyte.
1

reaction ratio: nV3+ : nH2 O = 2 mol : 1 mol (see Eq. 8)

concentration ratio: cV3+ : cH2 O = 1.5 mol·L−1 : 55.6 mol·L−1
⇒ volume ratio: V V3+ : V H2 O = 1 L : 0.0134 L

24
3.2 Contributions within this Thesis

The increase of energy density and specific energy is the main motivation for
the development of a VARFB. Additionally, the VARFB bears further benefits in
comparison to VRFB. In the positive electrolyte of a VRFB vanadium occurs as
VO+ 2 (oxidation state +5). The avoidance of this vanadium species in the VARFB
is advantageous as on the one hand vanadium(V) is suspected to be carcinogenic
[62, 63] and on the other hand the precipitation of VO+2 as V2 O5 is the reason for
the upper temperature limit of the VRFB [38]. Thus, the VARFB has no upper
temperature limit due to the absence of VO+2.

There are not many publications that deal with the VARFB. It was first described
in a patent by Kaneko et al. in 1992 [24]. Operation tests were performed using two
different reaction units for charging and discharging. An experimental study of a
VARFB employing two different hot-pressed membrane electrode assemblies (MEA)
with titanium electrodes was published by Hosseiny et al. [23]. Menictas et al. [26]
and Noack et al. [25] reported the performance data of a VOFC.
In this monodirectional device only the discharge process of a VARFB is conducted.
Furthermore, the VOFC was modeled by Wandschneider et al. [64]. Other publication
only mentioned the VARFB but did not provide any results or data on the system
[65–67]. Until today, there is only a scarce number of publications about VARFB
and the results are insufficient. For instance, Hosseiny et al. [23] achieved an energy
efficiency of 26.67 % at 40 °C operation temperature and at the low current density
of 0.24 mA·cm−2 . Likewise, Kaneko et al. [24] reported an energy efficiency of only
ηE =27.4 % (50 mA·cm−2 charging current density; 10 mA·cm−2 discharge current
density; 25 °C). Thus, the aim of this thesis is to contribute to the investigation and
improvement of the VARFB performance and to develop a more compact VARFB
employing a unitized reaction unit.

3.2 Contributions within this Thesis

3.2.1 Unitized Vanadium-Air Redox Flow Battery Setup

In the aforementioned publications, only monodirectional (i.e. only charging or
discharging) systems such as the VOFC [25, 26, 64] or bidirectional systems utilizing
separated charge-discharge units [23, 24] were presented. These systems have the
disadvantage that either a fully reversible battery operation is not possible (monodi-
rectional systems) or two different reaction units are necessary which increases the
costs, volume and mass of the system. Due to these reasons, a combined or uni-

25
Chapter 3 Vanadium-Air Redox Flow Batteries

tized bidirectional VARFB is desirable. In publication I (“Study of an unitised

bidirectional vanadium-air redox flow battery comprising a two-layered cathode”)
a novel VARFB with a combined reaction unit is described and the performance
is investigated at different operation conditions. An arrangement of the positive
electrode was presented in which both OER and ORR can be conducted. The
challenge for this setup are the opposite requirements for ORR and OER: While the
ORR is catalyzed by Pt and the electrode should be hydrophobic to repel produced
water and avoid clogging of the electrode, the OER is catalyzed best by IrO2 and
the electrode material should be hydrophilic. These contradictory demands were
overcome by using a two-layered electrode setup. One layer consisted of a hydrophilic,
IrO2 -modified graphite felt for the OER, the other layer was a hydrophobic gas
diffusion layer decorated with platinum supported on carbon black (Pt/C) and Nafion
ionomer similar as it is used in proton exchange membrane fuel cells (PEMFC).

These two electrode layers were assembled in the test cell and compressed be-
tween membrane and current collector, i.e. no hot-pressing was necessary. This
is advantageous as both Hosseiny et al. [23] and Menictas et al. [26] used MEAs
obtained via hot-pressing of the gas diffusion electrodes (GDE) with the membrane
and Menictas et al. [26] reported about serious issues with the detachment of the
GDE due to swelling of the membrane. A detachment of the catalyst layer of a
catalyst-coated membrane during VOFC operation was also reported by Noack et
al. [25]. The hot-pressing technique of MEAs is well-known and established in the
area of PEMFC. However, in contrast to PEMFC, the MEA used in a VARFB is in
contact with liquid aqueous solutions and thus swelling of the membrane is more
pronounced. Generally, membranes become dry during hot-pressing and therefore
the swelling due to water uptake from the aqueous electrolyte can be significant. An
additional concern is that a MEA for a VARFB does not contain a GDE on both
sides but only on one side. The other electrode is in general a porous, compressible
graphite felt. Thus, equally distributed and sufficient pressure on both sides of
the MEA is more difficult to achieve in a VARFB than in PEMFC. This promotes
delamination of the catalyst layer and/or GDL in VARFB. Due to the problems
arising from utilizing hot-pressed MEAs, a new approach was sought in this thesis
that avoids the use of an unsymmetrical hot-pressed MEA. A two-layered positive
electrode obtained without the need of hot-pressing was presented in publication I.
Moreover, the straightforward preparation of the IrO2 -modified graphite felt and its
characterization was shown in publication I followed by the performance analysis
of the VARFB. Half-cell potentials were recorded during operation with an integrated

26
3.2 Contributions within this Thesis

dynamic hydrogen electrode and the behaviour at different current densities and over
several cycles was evaluated. With the presented setup a maximum energy efficiency
of 42 % was achieved (j = 20 mA cm−2 , room temperature) which is the highest
value reported for a VARFB operated at room temperature. A performance decrease
in terms of coulombic and voltaic efficiency was observed during cycling and it was
hypothesized that crossover of vanadium species and corrosion of carbon materials
are the reasons for performance decay.

3.2.2 Investigation of Membrane Transfer Processes

The identification and quantification of unwanted membrane transfer processes as
possible reasons for the losses in coulombic efficiency and capacity during VARFB
cycling are presented in publication II (“Investigation of crossover processes in a
unitized bidirectional vanadium/air redox flow battery”). These investigations are
the basis for optimization strategies of the VARFB system. Up to date, there were
only suppositions and hypotheses about possible loss mechanisms in a VARFB but
no experimental investigation and quantification. For instance, Hosseiny et al. [23]
suspected permeation of dissolved oxygen which was produced during charging to be
the major cause of efficiency losses. Therefore, MEAs with different morphology of
the titanium electrode (mesh and sintered material, respectively) were used to avoid
trapping of oxygenated water in the electrode and its subsequent permeation through
the membrane. Menictas et al. [26] concluded that oxygen permeation would only
play a minor role in a VOFC as they observed a high utilization of active material
(i.e. V2+ electrolyte). This is not in contradiction to the observation of Hosseiny et al.
[23] as they presumed the oxygen produced during charging to undergo permeation
while Menictas et al. [26] conducted only the discharging reaction of a VARFB in
the investigated VOFC. Permeation of O2 was not mentioned in other publications
(e.g. Noack et al. [25], Wandschneider et al. [64] and Kaneko et al. [24]). However,
due to initial experiments conducted in the course of this thesis, oxygen permeation
was suspected to be a relevant phenomenon and this hypothesis was investigated in
publication II.
Besides oxygen permeation, other processes and side reactions such as hydrogen
evolution, crossover of V2+ and/or V3+ from the negative to the positive electrode
can be imagined to occur. Therefore, a combination of analysing methods was
employed in publication II. An experimental setup was designed in order to
monitor the concentrations of V2+ and V3+ in the negative electrolyte in situ during
the VARFB operation using UV/Vis spectroscopy with a flow-through cuvette.

27
Chapter 3 Vanadium-Air Redox Flow Batteries

Additionally, the vanadium concentration of the initially vanadium-free positive

electrolyte was determined after charging and discharging with ICP-MS to quantify
the vanadium crossover from the negative to the positive electrode. These data
combined with ex situ determinations of diffusion coefficients of V2+ and V3+ through
Nafion 117 (N117) membranes revealed the relation of the different loss mechanisms
to each other and the differentiation between diffusional vanadium crossover and
migrational/electroosmotic crossover. It was found that oxygen permeation rather
than vanadium ion crossover is the major cause for lowered coulombic efficiency.
During operation with j = 20 mA·cm−2 the vanadium crossover causes at most 20 %
of the coulombic efficiency losses. The share of 58 % of the vanadium crossover is
due to diffusion and consequently 42 % is related to migration/electroosmosis.
However, the most severe challenge of VARFB is the crossover of vanadium. It
is described as an issue or possible issue in several of the publications that deal
with VARFB or related systems [26, 64–69]. The considerable amount of 6 % of the
vanadium species initially present in the negative anolyte underwent crossover during
a cycle as presented in publication II. In contrast to VRFB where vanadium
crossover induces only a lowered coulombic efficiency, this process leads to a capacity
fade and therefore limited lifetime of the VARFB. Hence, it is crucial to minimize
the crossover of vanadium species for the long-term operationality of the VARFB.

3.2.3 Membrane Modification for Enhanced VARFB

Performance
In publication III (“Layer-by-layer modification of Nafion membranes for increased
life-time and efficiency of vanadium/air redox flow batteries”) a conventional proton
exchange membrane (N117) is modified for improving the membrane performance
concerning vanadium crossover and oxygen permeation and hence enabling a per-
formance enhancement of the VARFB. This modification was carried out with a
LbL deposition of polyelectrolytes. The LbL techniques allows the convenient ap-
plication of very thin multiple layers of oppositely charged polyelectrolytes on a
substrate. Several properties of a substrate such as gas permeability and multivalent
ion transfer rate can be tuned with the deposition of polyelectrolyte multilayer films.
A modification of Nafion membranes with polyelectrolyte multilayers for reduced
transfer rates of both multivalent ions and gases but maintained sufficient proton
conductivity has not been reported before. The thin multilayers investigated in
publication III consisted of polyethylenimine (PEI) and Nafion ionomer. PEI

28
3.3 Author Contributions to Publications I-III

was chosen because it was used in LbL films to successfully reduce multivalent ion
crossover and increase the selectivity towards H+ [70–72]. Nafion ionomer was em-
ployed as polyanion due to its advantageous properties such as chemical stability and
high proton conductivity and because it was also used as substrate material. However,
properties of this pair of polyelectrolytes have not been described yet in literature.
The modification lead to reduced vanadium crossover and oxygen permeation but
also to a decreased proton conductivity. The influence of the number of bilayers on
the aforementioned properties was studied and the optimum was identified. The
application of a N117 membrane modified with 10 bilayers of PEI/Nafion (denoted
as “N117-(PEI/Nafion)10 ”) in a VARFB resulted in 13 % less oxygen permeation
in comparison to unmodified membranes. Even more important, the vanadium
crossover was reduced by 70 % during a cycle in comparison to an unmodified N117
membrane. This reduction is important for increased life-time of the VARFB as
described above. Additionally, the coulombic efficiency during VARFB operation
was increased by 12 % (N117: ηC = 81 %, N117-(PEI/Nafion)10 : ηC = 93 %) and the
energy efficiency by 3.7 % (N117: ηE = 41.5 %, N117-(PEI/Nafion)10 : ηE = 45.2 %)
while the voltaic efficiency was reduced by 2.5 % due to lowered proton conductivity.
Thus, the modified membranes are promising concerning prolonged lifetime of the
battery and improved energy and coulombic efficiency.

3.3 Author Contributions to Publications I-III

Publications in peer-reviewed journals are typically collaborative efforts. This applies
to the publications I-III of this thesis as well. The individual contributions of the
author of this thesis are described below.

3.3.1 Publication I
The idea for the investigation of the presented system was a collaborative work of
Eva-Maria Hammer, Lidiya Komsiyska and the first author (author of this thesis).
Most of the conception/design of the experiments was done by the first author
while the co-authors gave input and advice for specific experimental questions. The
two-layered cathode setup was the idea of the first author. The conduction of the
experiments was mainly done by the first author. Exceptions are the construction
of the test cell (in cooperation with Hayo Seeba and Dietmar Piehler), the preparation
of some of the IrO2 -modified graphite felts used in the study (Andrea Ballarin),
protonation of some of the membranes (Benedikt Berger), design and construction

29
Chapter 3 Vanadium-Air Redox Flow Batteries

of the dynamic hydrogen electrode (Frank Bättermann), X-ray diffraction (XRD)

measurements (Martin Knipper) and ICP-MS sample preparation and analysis
(Dana Schonvogel). All data analysis was done by the first author. The initial
composition of the manuscript was performed together with Lidiya Komsiyska
and Carolina Nunes Kirchner. The manuscript writing was done by the first
author with revisions, suggestions and corrections from all co-authors.

3.3.2 Publication II
The idea for this publication was from Carolina Nunes Kirchner, Lidiya Komsiyska
and the first author. The conception/design of the experiments was done by
the author, except for the suggestion to incorporate the determination of diffusion
coefficients (Gunther Wittstock). The experiments were conducted by the first
author apart from a pre-study on the calibration of UV/Vis spectroscopy (Timo
di Nardo) and ICP-MS analysis of prepared samples (Dana Schonvogel). All data
analysis was performed by the first author. The initial composition of the
manuscript was done by the first author. The manuscript was written by
the first author with revisions, suggestions and corrections incorporated from all
co-authors.

3.3.3 Publication III

The idea for this publication was from the first author. The conception/design
of the experiments was planned by the first author apart from the suggestion
to include TGA measurements in the study (Lidiya Komsiyska). Experiments
were performed by the first author, exceptions are initial experiments related to
the polyelectrolyte deposition routine (Khrystyna Yezerska), the thermogravimetric
analysis (TGA) measurements (Pratik Das), parts of the scanning electron microscopy
(SEM) measurements and parts of the determinations of the diffusion coefficients
(Benedikt Berger) and ICP-MS analysis of prepared samples (Dana Schonvogel). All
data analysis, the initial composition of the manuscript and the manuscript
writing were done by the first author. All co-authors contributed with suggestions
and corrections to the (internal) revision of the manuscript.

30
4 Conclusion and Outlook

The aim of this thesis was to construct, characterize and optimize a unitized VARFB
employing a single reaction unit for charging and discharging which had not been
described in literature before. For this purpose, a new VARFB test cell was engineered
comprising a two-layered cathode. The performance was assessed at different current
densities (publication I). With the presented setup an energy efficiency of 42 % was
achieved at a current density of 20 mA·cm−2 and at room temperature which is higher
than the efficiency reported for non-unitized VARFB by Hosseiny et al. [23] and
Kaneko et al. [24]. Consecutive charging-discharging cycles were conducted which
revealed a performance decay due to capacity losses. Transfer processes through
the membrane were suspected to diminish the efficiency and lifetime of the battery.
The transfer reactions were studied in publication II applying in situ UV/Vis
spectroscopy of the negative electrolyte to monitor the concentrations of V2+ and
V3+ during VARFB operation. To allow the identification and quantification of side
reactions, these results were combined with the determination of the total vanadium
crossover by ICP-MS analysis of the positive electrolyte. Oxygen permeation and
vanadium crossover were ascertained as the major reason for the coulombic efficiency
losses. Moreover, vanadium crossover leads to a limited life-time of the VARFB due
to a continuous loss of capacity. During the studied cycle, a considerable amount of
6 % of the vanadium in the negative electrolyte underwent crossover. To achieve a
reduction of the vanadium crossover, a novel modification routine of the membrane
with thin layers of polyelectrolytes was studied in publication III. The modification
of N117 consisted of a layer-by-layer film of 10 bilayers of PEI and Nafion ionomer. It
resulted in significantly reduced vanadium crossover (-70 %) and oxygen permeation
(-13 %) in comparison to unmodified N117 which is advantageous for enhanced
efficiency and life-time of the VARFB. In summary, a new unitized bidirectional
VARFB system was established and analyzed in the course of this thesis. Based on
this analysis, significant improvements of the performance could be demonstrated in
comparison to the state of the art.
The obtained results might also be relevant for and applicable to other electro-

31
Chapter 4 Conclusion and Outlook

chemical systems. For instance, the issue of crossover is also known for VRFB,
microbial fuel cells and direct methanol fuel cells. The membrane modification
routine presented in this thesis can be applied to improve the performance of the
aforementioned systems. Due to the identical reactions in the positive half-cell of
a VARFB and unitized regenerative fuel cell (URFC), the positive electrode setup
presented in this thesis might also be interesting for a URFC setup. However, the
compatibility of the used polyelectrolytes with the aforementioned systems would
need to be evaluated.
Not all open research questions were addressed in this thesis and some of them
are described below.
The investigation of the corrosion stability of components employed in the reaction
unit needs to be further evaluated. Half-cell potentials of more than 1.5 V vs.
SHE are present at the positive electrode during charging, thus corrosion of carbon
materials or of the Pt/C catalyst might occur. These degradation processes are
suspected to have an impact on the long-term stability of the performance of the
VARFB. Thus, further investigations in this direction are needed for the evaluation
of the system.
As mentioned in this thesis, the costs of electrical storage system for stationary
purposes are crucial. As such, expensive materials need to be avoided. In the
presented VARFB setup precious metal catalysts such as IrO2 and Pt were employed.
For the viability of the VARFB as a stationary storage device, these catalysts need
to be replaced by cheaper, non-noble metal catalysts. There are reports in literature
about bidirectional oxygen catalysts that enhance both oxygen reduction reaction as
well as oxygen evolution reaction (e.g. N-doped graphene [73]). A VARFB employing
alternative catalysts which are not based on noble metals would be desirable.
Another influence that affects especially the voltaic efficiency of the VARFB
requires further investigation. Both OER and ORR are subject to slow kinetics which
is reflected in the overpotentials that are present during charging and discharging. The
operation of the VARFB at higher temperatures enhances the kinetics of ORR and
OER and thus lowers overpotentials. This would result in higher voltaic efficiencies
and hence higher energy efficiency. However, a contrary effect might be an increased
rate of vanadium crossover and/or oxygen permeation at elevated temperatures
which would negatively influence the coulombic efficiency and subsequently lower
the energy efficiency. The optimum temperature for the VARFB operation need to
be evaluated in the future to achieve the best compromise between improved ORR
and OER kinetics and minimized transfer processes through the membrane.

32
Conclusion and Outlook

The remarkable achievement in optimizing the membrane properties presented

in publication III should not hide that the capacity fading is still to high for
commercial applications. The crossover of vanadium species through the membrane
which is responsible for the capacity loss needs to be close to zero or the capacity loss
needs to be recoverable. In hydrogen-bromine RFB the crossover of bromine-species
is tackled by simply returning back the bromine/bromide containing solutions that
were transferred to the hydrogen electrode [60]. This is easily possible because in
the gaseous half-cell (hydrogen electrode) there is no liquid phase present. This
type of capacity recovery could also be pursued in VARFB. However, the procedure
would need to be modified as in VARFB a liquid phase is intentionally present
during charging. Another approach to face the capacity decay in a VARFB could
be followed by merging a VRFB and a VARFB system. The charging reaction in
such a system would be identical to VRFB until all VO2+ is converted to VO+ 2,
subsequently OER would occur at the positive electrode as it is the case in VARFB.
During discharging the situation is vice versa, i.e. reduction of VO+2 to VO
2+
followed
by ORR. This system would have two charging steps and two discharging steps.
The main advantage would be that vanadium species are intentionally present in
the positive electrolyte and crossover of vanadium species merely results in lowered
coulombic efficiency (as it is the case in VRFB) but not in capacity loss. However,
the drawbacks of a VRFB (e.g. a temperature limit due to V2 O5 precipitation) will
also be inherent in such a system and therefore it has to be evaluated whether such
a combination is reasonable.
In conclusion, with the present thesis a copious basis is laid in the understanding
of the novel unitized VARFB system and the relevant processes that occur in this
battery. However, it is desirable to carry on the investigation of this VARFB system
by further studies related to the unanswered questions that are described above.

33
References
[1] S. R. L. Stephen J. Farnsworth, S. R. L. Stephen J. Farnsworth, Int. J. Public
Opin. Res. 24 (2012) 93–103.

[2] J. Cook, D. Nuccitelli, S. A. Green, M. Richardson, B. Winkler, R. Painting,

R. Way, P. Jacobs, A. Skuce, Environ. Res. Lett. 8 (2013) 024024.

[3] UN Framework on Climate Change, Paris Agreement, 2015. http://www.cop21.

gouv.fr/en/more-details-about-the-agreement/ (Accessed: 28/12/2015).

[4] Met Office, "Global temperatures set to reach 1 °C marker for first
time", 2015. http://www.metoffice.gov.uk/news/release/archive/2015/
one-degree (Accessed: 28/11/2015).

[5] Met Office, HadCRUT4 data set, 2015. http://hadobs.metoffice.com/

hadcrut4/data/current/time_series/HadCRUT.4.4.0.0.annual_ns_avg.
txt (Accessed: 28/11/2015).

[6] C. P. Morice, J. J. Kennedy, N. A. Rayner, P. D. Jones, J. Geophys. Res. Atmos.

117 (2012) 1–22.

[7] International Energy Agency, World Energy Outlook 2015, 2015. http://www.
worldenergyoutlook.org/ (Accessed: 28/11/2015).

[8] P. Alotto, M. Guarnieri, F. Moro, Renew. Sustain. Energy Rev. 29 (2014)

325–335.

[9] M. Beaudin, H. Zareipour, A. Schellenberglabe, W. Rosehart, Energy Sustain.

Dev. 14 (2010) 302–314.

[10] K. Divya, J. Østergaard, Electr. Power Syst. Res. 79 (2009) 511–520.

[11] B. Dunn, H. Kamath, J.-M. Tarascon, Science 334 (2011) 928–935.

[12] J. Leadbetter, L. G. Swan, J. Power Sources 216 (2012) 376–386.

35
 References

[13] J. Weniger, J. Bergner, T. Tjaden, V. Quaschning, Dezentrale Solarstromspeicher

für die Energiewende, Berliner Wissenschafts-Verlag, ISBN: 978-3-8305-3548-5,
2015.

[14] Q. Huang, Q. Wang, ChemPlusChem 80 (2015) 312–322.

[15] M. Skyllas-Kazacos, M. H. Chakrabarti, S. A. Hajimolana, F. S. Mjalli,

M. Saleem, J. Electrochem. Soc. 158 (2011) R55.

[16] M. Vijayakumar, W. Wang, Z. Nie, V. Sprenkle, J. Hu, J. Power Sources 241

(2013) 173–177.

[17] G. Wang, J. Chen, X. Wang, J. Tian, H. Kang, X. Zhu, Y. Zhang, X. Liu,

R. Wang, J. Energy Chem. 23 (2014) 73–81.

[18] S. Roe, C. Menictas, M. Skyllas-Kazacos, J. Electrochem. Soc. 163 (2016)

A5023–A5028.

[19] J. Noack, N. Roznyatovskaya, T. Herr, P. Fischer, Angew. Chemie 127 (2015)

9912–9947.

[20] L. Li, S. Kim, W. Wang, M. Vijayakumar, Z. Nie, B. Chen, J. Zhang, G. Xia,

J. Hu, G. Graff, J. Liu, Z. Yang, Adv. Energy Mater. 1 (2011) 394–400.

[21] W. Wang, Q. Luo, B. Li, X. Wei, L. Li, Z. Yang, Adv. Funct. Mater. 23 (2013)
970–986.

[22] J. grosse Austing, C. Nunes Kirchner, E.-M. Hammer, L. Komsiyska, G. Witt-

stock, J. Power Sources 273 (2015) 1163–1170.

[23] S. Hosseiny, M. Saakes, M. Wessling, Electrochem. Comm. 13 (2011) 751–754.

[24] H. Kaneko, A. Negishi, K. Nozaki, K. Sato, M. Nakajima, EU Patent 0517217A1,

1992.

[25] J. Noack, C. Cremers, D. Bayer, J. Tübke, K. Pinkwart, J. Power Sources 253

(2014) 397–403.

[26] C. Menictas, M. Skyllas-Kazacos, J. Appl. Electrochem. 41 (2011) 1223–1232.

[27] W. Kangro, DE Patent 914264C, 1954.

[28] M. Bartolozzi, J. Power Sources 27 (1989) 219–234.

36
References

[29] A. Z. Weber, M. M. Mench, J. P. Meyers, P. N. Ross, J. T. Gostick, Q. Liu, J.

Appl. Electrochem. 41 (2011) 1137–1164.

[30] P. Leung, X. Li, C. Ponce de León, L. Berlouis, C. T. J. Low, F. C. Walsh, RSC

Adv. 2 (2012) 10125.

[31] G. L. Soloveichik, Chem. Rev. 115 (2015) 11533–11558.

[32] K. Lin, Q. Chen, M. R. Gerhardt, L. Tong, S. B. Kim, L. Eisenach, A. W.

Valle, D. Hardee, R. G. Gordon, M. J. Aziz, M. P. Marshak, Science 349 (2015)
1529–1532.

[33] B. Huskinson, M. P. Marshak, C. Suh, S. Er, M. R. Gerhardt, C. J. Galvin,

X. Chen, A. Aspuru-Guzik, R. G. Gordon, M. J. Aziz, Nature 505 (2014)
195–198.

[34] T. Janoschka, N. Martin, U. Martin, C. Friebe, S. Morgenstern, H. Hiller, M. D.

Hager, U. S. Schubert, Nature 527 (2015) 78–81.

[35] T. Liu, X. Wei, Z. Nie, V. Sprenkle, W. Wang, Adv. Energy Mater. (2015).

[36] M. C. Tucker, A. Phillips, A. Z. Weber, ChemSusChem 8 (2015) 3996–4004.

[37] D. Lloyd, E. Magdalena, L. Sanz, L. Murtomäki, K. Kontturi, J. Power Sources

292 (2015) 87–94.

[38] M. Skyllas-Kazacos, M. H. Chakrabarti, S. A. Hajimolana, F. S. Mjalli,

M. Saleem, J. Electrochem. Soc. 158 (2011) R55.

[39] P. Alotto, M. Guarnieri, F. Moro, Renew. Sustain. Energy Rev. 29 (2014)

325–335.

[40] F. T. Wandschneider, S. Röhm, P. Fischer, K. Pinkwart, J. Tübke, H. Nirschl,

J. Power Sources 261 (2014) 64–74.

[41] A. Tang, J. McCann, J. Bao, M. Skyllas-Kazacos, J. Power Sources 242 (2013)

349–356.

[42] C. Yin, S. Guo, H. Fang, J. Liu, Y. Li, H. Tang, Appl. Energy 151 (2015)
237–248.

[43] J. S. Lawton, a. Jones, T. Zawodzinski, J. Electrochem. Soc. 160 (2013) A697–

A702.

37
 References

[44] Z. Yang, J. Zhang, M. C. W. Kintner-Meyer, X. Lu, D. Choi, J. P. Lemmon,

J. Liu, Chem. Rev. 111 (2011) 3577–3613.

[45] M. H. Chakrabarti, F. S. Mjalli, I. M. AlNashef, M. A. Hashim, M. A. Hussain,

L. Bahadori, C. T. J. Low, Renew. Sustain. Energy Rev. 30 (2014) 254–270.

[46] Z. Xie, Q. Liu, Z. Chang, X. Zhang, Electrochim. Acta 90 (2013) 695–704.

[47] D. R. Lide (Ed.), CRC Handbook of Chemistry and Physics, Internet Version,
CRC Press, Boca Raton, FL, 2005.

[48] B. L. Ellis, L. F. Nazar, Curr. Opin. Solid State Mater. Sci. 16 (2012) 168–177.

[49] N. S. Choi, Z. Chen, S. a. Freunberger, X. Ji, Y. K. Sun, K. Amine, G. Yushin,

L. F. Nazar, J. Cho, P. G. Bruce, Angew. Chemie - Int. Ed. 51 (2012) 9994–10024.

[50] A. Crawford, V. Viswanathan, D. Stephenson, W. Wang, E. Thomsen, D. Reed,

B. Li, P. Balducci, M. Kintner-Meyer, V. Sprenkle, J. Power Sources 293 (2015)
388–399.

[51] C. Ponce de León, a. Frías-Ferrer, J. González-García, D. Szánto, F. Walsh, J.

Power Sources 160 (2006) 716–732.

[52] D. P. Scamman, G. W. Reade, E. P. Roberts, J. Power Sources 189 (2009)

1231–1239.

[53] S.-H. Shin, S.-H. Yun, S.-H. Moon, RSC Adv. 069 (2012) 12686–12689.

[54] F. Pan, Q. Wang, Molecules 20 (2015) 20499–20517.

[55] J. Zhang, L. Li, Z. Nie, B. Chen, M. Vijayakumar, S. Kim, W. Wang, B. Schwen-

zer, J. Liu, Z. Yang, J. Appl. Electrochem. 41 (2011) 1215–1221.

[56] X. Wu, S. Liu, N. Wang, S. Peng, Z. He, Electrochim. Acta 78 (2012) 475–482.

[57] F. Chang, C. Hu, X. Liu, L. Liu, J. Zhang, Electrochim. Acta 60 (2012) 334–338.

[58] Y. Chen, K. Santhanam, A. Bard, J. Electrochem. Soc. 128 (1981) 1460–1467.

[59] Y. Wen, H. Zhang, P. Qian, H. Zhou, P. Zhao, B. Yi, Y. Yang, Electrochim.

Acta 51 (2006) 3769–3775.

38
References

[60] G. Lin, P. Chong, V. Yarlagadda, T. Nguyen, R. J. Wycisk, P. N. Pintauro,

M. Bates, S. Mukerjee, M. C. Tucker, a. Z. Weber, J. Electrochem. Soc. 163
(2016) A5049–A5056.

[61] K. T. Cho, M. C. Tucker, M. Ding, P. Ridgway, V. S. Battaglia, V. Srinivasan,

A. Z. Weber, ChemPlusChem 80 (2015) 402–411.

[62] A. Léonard, R. Lauwerys, Mutat. Res. 239 (1990) 17–27.

[63] L. Passantino, A. B. Muñoz, M. Costa, Metallomics 5 (2013) 1357–67.

[64] F. Wandschneider, M. Küttinger, J. Noack, P. Fischer, K. Pinkwart, J. Tübke,

H. Nirschl, J. Power Sources 259 (2014) 125–137.

[65] O. David, K. Percin, T. Luo, Y. Gendel, M. Wessling, J. Energy Storage 1 (2015)

65–71.

[66] C. Gutsche, C. J. Moeller, M. Knipper, H. Borchert, J. Parisi, T. Plaggenborg,

Electrocatalysis 6 (2015) 455–464.

[67] G. Merle, F. C. Ioana, D. E. Demco, M. Saakes, S. S. Hosseiny, Membranes

(Basel). 4 (2013) 1–19.

[68] Y. Wen, J. Cheng, P. Ma, Y. Yang, Y. Xun, Electrochim. Acta 53 (2008)

6018–6023.

[69] Y. H. Wen, J. Cheng, Y. Xun, P. H. Ma, Y. S. Yang, Electrochim. Acta 53

(2008) 6018–6023.

[70] S. Abdu, M. C. Martí-Calatayud, J. E. Wong, M. García-Gabaldón, M. Wessling,

ACS Appl. Mater. Interfaces 6 (2014) 1843–1854.

[71] F. Guesmi, C. Hannachi, B. Hamrouni, Ionics (Kiel). 18 (2012) 711–717.

[72] T. Sata, T. Sata, W. Yang, J. Memb. Sci. 206 (2002) 31–60.

[73] M. Li, L. Zhang, Q. Xu, J. Niu, Z. Xia, J. Catal. 314 (2014) 66–72.

39
List of Abbreviations and Symbols
Abbreviations
BPP Bipolar plate

CAES Compressed air energy storage

CEM Cation exchange membrane

FC Fuel cell

GDE Gas diffusion electrode

HER Hydrogen evolution reaction

ICP-MS Inductively coupled plasma mass spectroscopy

IEA International Energy Agency

IL Ionic liquids

LbL Layer-by-layer

Li-ion Lithium-ion battery

MEA Membrane electrode assembly

N117 Nafion 117

Na-S Sodium-sulfur battery

NASA National Aeronautic and Space Administration

OER Oxygen evolution reaction

ORR Oxygen reduction reaction

Pb-acid Lead-acid battery

41
 List of Abbreviations and Symbols

PEI Polyethyleneimine

PEMFC Proton exchange membrane fuel cell

PHS Pumped hydro storage

PV Photovoltaic

RFB Redox flow battery

SEM Scanning electron microscopy

SHE Standard hydrogen electrode

TGA Thermogravimetric analysis

URFC Unitized regenerative fuel cell

VARFB Vanadium/air redox flow battery

VOFC Vanadium/oxygen fuel cell

VRFB All-vanadium redox flow battery

XRD X-ray diffraction

Symbols
c Concentration

E0 Standard electrode potential

∆E 0 Standard cell potential

I Current

j Current density

n Amount of substance

Q Charge

U Voltage

V Volume

W Electrical energy

42
List of Abbreviations and Symbols

ηC Coulombic efficiency

ηE Energy efficiency

ηV Voltaic efficiency

⊕ / (+) Positive electrode

/ (-) Negative electrode

43
Publications

45
Publication I

47
Study of an unitised bidirectional vanadium/air redox
flow battery comprising a two-layered cathode
Authors: Jan grosse Austing, Carolina Nunes Kirchner, Eva-Maria Hammer, Lidiya
Komsiyska and Gunther Wittstock
Journal of Power Sources, 273, 2015, 1163-1170, DOI: 10.1016/j.jpowsour.2014.09.177

membrane IrO2-mod. graphite felt

graphite felt gas diffusion electrode
current collector flow field

O2
– +
V2+/ V3+
H2O

electrolyte
tank

reaction unit

49
 Journal of Power Sources 273 (2015) 1163e1170

Contents lists available at ScienceDirect

Journal of Power Sources

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

Study of an unitised bidirectional vanadium/air redox flow battery

comprising a two-layered cathode
Jan grosse Austing a, Carolina Nunes Kirchner a, Eva-Maria Hammer a, 1,
Lidiya Komsiyska a, *, Gunther Wittstock b
a
NEXT ENERGY e EWE Research Centre for Energy Technology, Carl-von-Ossietzky-Str. 15, 26129 Oldenburg, Germany
b
Carl von Ossietzky University of Oldenburg, School of Mathematics and Natural Sciences, Centre of Interface Science, Department of Chemistry,
26111 Oldenburg, Germany

h i g h l i g h t s

We introduce a new concept for an unitised vanadium/air redox flow battery.

A novel bidirectional two-layered cathode setup was utilised in the battery.

An energy efficiency of 42% was achieved at 21  C and 20 mA cm2 current density.

Overvoltage increase and species crossover are observed during consecutive cycling.

a r t i c l e i n f o a b s t r a c t

Article history: The performance of a unitised bidirectional vanadium/air redox flow battery (VARFB) is described. It
Received 3 July 2014 contains a two-layered cathode consisting of a gas diffusion electrode (GDE) with Pt/C catalyst for dis-
Received in revised form charging and of an IrO2 modified graphite felt for charging. A simple routine is shown for the modifi-
8 September 2014
cation of a graphite felt with IrO2. A maximum energy efficiency of 41.7% at a current density of
Accepted 29 September 2014
Available online 6 October 2014
20 mA cm2 as well as an average discharge power density of 34.6 mW cm2 at 40 mA cm2 were
obtained for VARFB operation at room temperature with the novel cathode setup. A dynamic hydrogen
electrode was used to monitor half cell potentials during operation allowing to quantify the contribution
Keywords:
Vanadium air redox flow battery
of the cathode to the overall performance of the VARFB. Four consecutive cycles revealed that crossover
Vanadium oxygen fuel cell of vanadium ions took place and irreversible degradation processes within the reaction unit lead to a
Bidirectional oxygen/air electrode performance decrease.
Air battery © 2014 Elsevier B.V. All rights reserved.
Unitized regenerative fuel cell

1. Introduction energy storage device, e.g. for the compensation of the fluctuating
electricity production of renewables [1e4]. Promising properties of
Redox flow batteries (RFB) are electrochemical storage devices redox flow batteries are the independence of power and capacity,
comprising a reaction unit and two electrolyte tanks in which the high cycle life (>12,000 [2]) and competitive round-trip efficiencies
fluidic electrochemical active material is stored externally. By (z80% [4]). The VRFB utilises two vanadium-containing aqueous
circulating the electrolytes through the reaction unit and applying a electrolytes and therefore issues due to the crossover of species from
DC source or load, conversion of chemical to electrical energy and one half cell to the other are reduced. However, one major drawback
vice versa can take place. of VRFB is the low energy density and specific energy of
In the past years all-vanadium redox flow batteries (VRFB) gained 25e30 Wh kg1 [5] being similar to lead-acid batteries
increasing interest due to their potential application as stationary (30e50 Wh kg1 [2]). The low energy density has its origin in the
solubility limitations of the vanadium species and the low standard
* Corresponding author. Tel.: þ49 441 99906 418; fax: þ49 441 99906 109. 0
cell potential of DEVRFB ¼ 1:25 V. The low energy density can be an
E-mail addresses: [email protected] (J. grosse Austing), carolina.
[email protected] (C. Nunes Kirchner), [email protected]
obstruction in electrical storage applications with limited space.
(E.-M. Hammer), [email protected] (L. Komsiyska), gunther. There are different attempts proposed in literature for
[email protected] (G. Wittstock). increasing the vanadium solubility and thus the energy density by
1
Present address: Forschungszentrum Jülich GmbH, Institute of Energy and using various electrolyte additives [6e8]. However, these attempts
Climate Research (IEK-1), Materials Synthesis and Processing, Wilhelm-Johnen-Str.,
again show limits determined by the maximal soluble amount of
52428 Jülich, Germany.

http://dx.doi.org/10.1016/j.jpowsour.2014.09.177
0378-7753/© 2014 Elsevier B.V. All rights reserved.
1164 J. grosse Austing et al. / Journal of Power Sources 273 (2015) 1163e1170

the electroactive species. Another approach for increasing the en- of the GDE to the membrane, they were able to operate the stack for
ergy density is the replacement of the positive (cathode) redox over 120 h.
couple, VOþ 2þ
2 =VO , by O2/H2O (Eq. (1)). This vanadium/air redox Recently, Noack et al. [18] published results of a 280 cm2 VOFC.
flow battery (VARFB) was first proposed and disclosed in a patent The setup consisted of two membranes to avoid crossover of V2þ to
by Kaneko et al., in 1992 [9]. the Pt catalyst on the cathode side. They investigated the influence
of air flow rates on the discharging potentials and on power den-
sities. For a better understanding of the factors causing losses, they
anode : 4V2þ %4V3þ þ 4e E0 ¼ 0:26V
conducted several electrochemical impedance measurements. The
cathode : O2 þ 4e þ 4Hþ %2H2 O E0 ¼ þ1:23V (1)
maximum average power density was 19.6 mW cm2. However, the
cell : 4V2þ þ O2 þ 4Hþ %4V3þ þ 2H2 O DE0 ¼ 1:49V
VOFC can only be used for discharging the electrolyte, charging has
The anodic reactions are the same as in a conventional VRFB and to be done in a different device.
occur readily on activated carbon materials. Carbon-based elec- In this work we report a novel unitised bidirectional vanadium/
trodes are well suited for redox flow batteries as they offer a wide air redox flow battery comprising a two-layered cathode and its
operation potential windows, chemical stability as well as reason- behaviour. This system is, in contrast to fuel cells, rechargeable.
able costs [10]. Different carbon materials (e.g. graphite felts or Driven by the motivation to decrease the system weight and vol-
carbon paper) have been investigated concerning their properties ume, we created a unitised bidirectional system, i.e. charging and
in redox flow battery application [10,11], including several routines discharging process can be conducted in the same reaction unit.
to improve redox kinetics by surface treatments such as thermal The cathode of this system is comparable to the one used in uni-
treatment [12], catalyst deposition [13] or plasma activation [14]. tised regenerative fuel cells (URFC). In URFC a PEM water electro-
The oxygen source for the cathodic discharge reaction can be the lyser is combined with a proton-exchange membrane (PEM) fuel
ambient air while the water for the charging process needs to be cell resulting in a closed-loop device.
stored in a tank. Although the system still needs a cathodic For a better understanding of the processes, the VARFB is
container for the water, a significant increase in the energy density investigated with a dynamic hydrogen reference electrode (DHE) to
is expected due to the fact that the mass and the volume of the monitor separately half cell potentials. Finally, we discuss remain-
cathodic container is decreased. Based on a vanadium concentra- ing drawbacks and possible approaches to improve the perfor-
tion of 1.5 M and assuming that all water in the container can be mance of the system.
consumed, the volume of the water tank is approx. 1% of the volume
of the original VOþ 2 =VO
2þ
containing tank. Therefore, a significant 2. Experimental
decrease in system weight and volume is achieved, resulting in an
increased energy density. Depending on the ratio of reaction unit to 2.1. IrO2-modified graphite felt
tank sizes, the energy density of the VARFB is roughly doubled [1].
Likewise, the standard potential of the VARFB is approx. 20% higher The IrO2-modified graphite felt was prepared by the thermal
than that of the VRFB (DEVRFB0 ¼ 1:25 V) which leads to an addi- decomposition of an iridium compound similar to procedures
tional enhancement of energy density. described elsewhere [19,20] First, a graphite felt of 2 cm  2 cm
A big challenge for such a system is the design of the electrode (GFD5, SGL Carbon GmbH, Germany) was activated at 400  C for
setup on the cathode side. During charging when water is split 18 h in air atmosphere. This hydrophilic felt was immersed in a
(oxygen evolution reaction; OER), a hydrophilic electrode is pref- solution of 18.6 mg (NH4)2IrCl6 (99.994%; Alfa Aesar, GmbH & Co
erable to allow the reactant water to wet the electrode surface and KG, Germany) in 5.5 mL ultrapure water (>18 MU cm at 25  C).
to repel produced oxygen out of the electrode. Additionally, this Subsequently, the felt was dried in a vacuum oven at 60  C for
substrate should contain a suitable OER catalyst (e.g. IrO2). For the 30 min and then calcinated for 15 min at 450  C in air atmosphere.
discharging reaction (oxygen reduction reaction; ORR) a hydro- The soaking/drying/calcinating sequence was repeated two times,
phobic electrode is required to remove the produced water and with the only modification that the third calcination step was
avoid flooding of the electrode, as accumulated water hinders mass conducted for 1 h. The complete volume of the (NH4)2IrCl6 solution
transport of the reactant oxygen to the catalytic sites. Again a was consumed.
suitable catalyst to support the ORR is needed (e.g. Pt). As the re- The modified felt was investigated using XRD (scan step 0.05 ;
quirements for OER and ORR are contrary concerning hydropho- X-Pert Pro MPD diffractometer with copper tube, PANalytical B.V.,
bicity and adequate catalyst, the design of such a bidirectional air Netherlands) and SEM/EDX (NEON 40, Zeiss, Germany).
electrode generally implies compromises. There are different ap-
proaches how to engineer these electrode (e.g. single electrode/ 2.2. GDE preparation and membrane pretreatment
single catalytic layer, separate electrodes/single catalytic layer or
single electrode/multiple catalytic layers [15]). The GDE was prepared by airbrushing a suspension of an ORR
Hosseiny et al. [16] published results about a modular VARFB catalyst and Nafion® ionomer on a gas diffusion layer (GDL). The
using two different membrane electrode assemblies (MEA) for GDL was a carbon cloth material (ELAT HT 1400-W, BASF fuel cell
charging and discharging, respectively. Between charging and dis- GmbH, Germany). A suspension of Pt/C (40 mass % Pt on carbon
charging they exchanged the MEA used for charging by a different black; Alfa Aesar GmbH & Co KG, Germany) in 1:100 (w/w) 2-
MEA for the discharging process. They reported energy efficiencies propanol (99.9%; VWR International GmbH, Germany) was ultra-
of 26.67% at 40  C operating temperature and at a current density of sonicated for 20 min, then 1:20 (w/w, based on Pt/C) of a Nafion®
2.4 mA cm2 for both charging and discharging. perfluorinated resin solution (5 mass % in lower aliphatic alcohols
Menictas et al. [17] investigated a vanadium-oxygen fuel cell and water; SigmaeAldrich, Germany) was added and again ultra-
(VOFC). This system permits the discharging process only, i.e. oxi- sonicated for 45 min. This suspension was airbrushed on the GDL.
dising V2þ to V3þ while reducing oxygen. They examined the in- By differential weighing of the GDL substrate before and after the
fluence of different materials and operation conditions on the airbrushing process a loading of approx. 0.8 mg cm2 Pt was
performance of a 5-cell VOFC stack. One major challenge discussed determined.
in their study is the detachment of the GDE from the membrane Prior to use in the VARFB a Nafion® 117 membrane (Ion Power
due to swelling caused by water uptake. By optimising the bonding GmbH, Germany) was pretreated. The membrane was cleaned in 5%
 J. grosse Austing et al. / Journal of Power Sources 273 (2015) 1163e1170 1165

H2O2 (30%; Carl Roth GmbH & Co. KG, Germany) in deionised water
for 30 min at 80  C. After rinsing it with deionised water it was
protonated in 1 M H2SO4 (95e97%; Fischer Scientific GmbH, Ger-
many) for 30 min at 80  C and finally it was boiled in deionised
water for 10 min.

2.3. V3þ electrolyte production

The V3þ electrolyte was produced electrolytically in a VRFB

setup similar to a routine described elsewhere [21]. Briefly, a so-
lution of 1.13 M VOSO4,xH2O (97%; Sigma Aldrich, Germany) and
2.0 M H2SO4 (95e97%; Fischer Scientific GmbH, Germany) in
deionised water was prepared. For preparing a defined concen-
tration, the amount of crystal water in VOSO4,xH2O was deter-
mined as x z 3 with ICP-MS (XSERIES 2 ICP-MS, Thermo Scientific,
USA). This solution was electrolysed up to 1.8 V with two consec-
Fig. 2. Setup for a unitised bidirectional VARFB comprising a two-layered cathode
utive current densities (80 mA cm2 and 40 mA cm2, respectively) (Hþ(c) ¼ Hþ transport during charge; Hþ(d) ¼ Hþ transport during discharge).
in a VRFB setup, using twice the volume on the cathode as on the
anode. After reaching 1.8 V, half of the catholyte was disposed and
membrane pumps (KNF Neuberger dosing pumps SIMDOS 10,
the electrolytes were discharged up to 0.8 V (again 80 mA cm2 and
Germany).
40 mA cm2, respectively). The anolyte was used as V3þ electrolyte
12 mL of the V3þ anolyte was used (renewed for each experi-
for the VARFB operation experiments. The concentration of vana-
ment, except for those with consecutive cycles) and purged
dium in this electrolyte was determined as 1.2 mol L1 by ICP-MS.
permanently with N2 to prevent contact of the electrolyte with air.
The small volume of anolyte was chosen to be able to perform
2.4. Operating VARFB charging and discharging in reasonable time. To avoid operational
limitations, an excess catholyte volume of 50 mL was employed.
For the investigation of the VARFB performance, a test cell with 2 M H2SO4 was used as catholyte instead of pure water to achieve a
4 cm2 geometric area was constructed. The cell consisted of two higher conductivity and reduce osmotic pressure with respect to
half cells, separated by a membrane. The sealing between the two the anolyte. Mass transport limitations should be avoided by
half cells was achieved with Viton® O-rings (Dichtelemente arcus pumping the electrolytes through the cell with flow rates of
GmbH, Germany). The body of each half cell consisted of a poly- 100 mL min1 while charging the VARFB. During discharging the
carbonate block with manifolds for the electrolyte and cavities 2 M H2SO4 was not pumped but remained in the IrO2-modified
(2 cm  2 cm) for the current collector and the electrodes. The graphite felt, whereas air with 267 mL min1 was pumped through
anodic current collector was made of a graphite-based bipolar plate the flow field to the GDE (anolyte circulated as in charging mode
material (PPG 86, Eisenhuth GmbH Co. KG, Germany) and the anode with 100 mL min1). All experiments were carried out at room
was a thermally activated graphite felt (GFD5, SGL Carbon GmbH, temperature (21  C ± 1  C). The operation of the VARFB was con-
Germany; activation: 400  C/18 h/air atmosphere). The cathodic ducted with a potentiostat/galvanostat (Solartron Analytical Mod-
current collector was made of titanium (grade 2) with an integrated ulab Pstat potentiostat/galvanostat, UK).
serpentine flow field (Keil Feinwerktechnik GmbH, Germany). The In parallel to charging and discharging, half cell potentials were
GDE was placed on top of the flow field, followed by the IrO2- recorded versus a dynamic hydrogen reference electrode (DHE). The
modified graphite felt (Fig. 2). The GDE, the IrO2-modified graphite DHE was integrated into the cell as described by Li and Pickup [22]
felt and the anodic graphite felt were replaced by new ones for each for fuel cell application and similar as Sun et al. [23] reported for
experiment, except for those where the subsequent cycles were redox flow batteries. The DHE consisted of two thin Pt wires
recorded. The hydraulic connections were made from Tygon® (Ø ¼ 0.175 mm) placed between membrane and cathodic half cell
tubings. The electrolytes were pumped through the reaction unit by body close to the edge of the cathode. A constant current source
(5e30 mA) was adjusted to produce small quantities of H2 until a
potential of 0.197 V vs. Ag/AgCl/KCl(sat.), i.e. 0 V vs. standard
hydrogen electrode (SHE), was measured. In order to achieve this,
the Pt wires were placed on the cathodic half cell of the VARFB test
cell, covered with a Nafion® 117 membrane and the latter was
pressed onto the half cell with a cover sheet. This sheet had an
opening in the region where the DHE was placed. Via this hole the
Ag/AgCl reference electrode was coupled ionically to the membrane.
After the current was adjusted, the cover sheet was removed and the
cell was assembled. The current of the DHE remained constant
during the following experiment. A scheme of the overall setup for
the operation experiments of the VARFB is depicted in Fig. 1.

3. Results & discussion

3.1. Concept of the unitised bidirectional VARFB

Fig. 1. Scheme of the VARFB test cell operation system; DHE ¼ dynamic hydrogen The VARFB can be regarded as a combination of the negative half
electrode, DHEeCE ¼ counter electrode for DHE. cell of VRFB and with a bidirectional air electrode as the positive
1166 J. grosse Austing et al. / Journal of Power Sources 273 (2015) 1163e1170

half cell. The anode of the VARFB is similar as in a VRFB, consisting porosity and surface area as well as adequate chemical stability. The
of an activated graphite felt that is electrically connected to a pristine graphite felts are hydrophobic due to a lack of surface
graphite-based current collector, and the electrolyte being circu- functionalities. These functional groups on the surface can be
lated through the felt. As already mentioned, the cathode (¼ pos- created by a mild thermal oxidation [12]. After this treatment the
itive half cell) implies challenges due to the different requirements felts were hydrophilic allowing wetting them with the iridium-
for the charging reaction (ORR) and discharging reaction (OER). containing aqueous solution. The thermal decomposition re-
In order to the contradictory requirements, two different elec- actions of (NH4)2IrCl6 as described in Eq. (2) are proposed to take
trode layers were used as cathode, one layer was optimised for the place [19].
ORR and the other for the OER. The ORR electrode was a hydro-
phobised, Pt/C and Nafion® ionomer loaded GDE similar to PEM fuel D 1
ðNH4 Þ2 IrCl6 /IrCl3 þ 2NH3 þ 2HCl þ Cl
cells, the OER electrode consisted of an IrO2-modified graphite felt. 2 2
The two layers were assembled in direct contact to each other, the (2)
D 3
IrO2-modified graphite felt pressed against the membrane while IrCl3 þ O2 /IrO2 þ Cl
2 2
the GDE was in direct contact with the current collector (Fig. 2).
For charging, H2O/Hþ as the reactant for the OER was circulated Differential weighing of the felt before and after the treatment
through the IrO2-modified graphite felt in parallel to the mem- to determine mass differences as proposed in Ref. [19] was not
brane. The produced Hþ can be transported to the anode to achieve successful. This was related to the harsh annealing conditions of
charge balance. The IrO2-modified graphite felt was electrically 450  C under which the graphite fibres themselves can react with
contacted to the current collector via the GDE. In discharging mode, oxygen from the air to form CO2. Therefore, the mass of the
the GDE was supplied with oxygen from air by the flow field that is modified felt was often less then the mass of the pristine felt. The
integrated in the current collector. The protons that are necessary maximum iridium loading per geometric area can be calculated as
for the ORR were conducted from the anode through the mem- 1.98 mg cm2 (calculated as IrO2) from the iridium content of the
brane and the OER layer which was flooded with H2O/Hþ during solution that was completely consumed during the modification
discharging. On the surface of the GDE, oxygen and protons reacted process.
on the Pt/C catalyst particles to form water. In addition to the in- The IrO2-modified graphite felt was investigated with SEM to
dividual optimisation of the two electrode layers, the advantage of provide information about the coverage with the deposit (Fig. 3).
this cathode design is that hot-pressing of a GDE onto a fully wetted Thermally treated non-modified graphite felt showed a plain sur-
membrane is avoided together with the related detachment prob- face (Fig. 3A, B). EDX analysis (not shown here) confirmed the
lems described in Ref. [17]. presence of Ir in the deposits shown in Fig. 3C, D. These iridium-
containing deposits were well distributed over the sample. The
higher magnification image (Fig. 3D) shows agglomerates of de-
3.2. IrO2-modified graphite felt posits with size in the micrometer range.
Wang et al. [19] proposed a further reduction of the IrO2 to Ir by
The thermal decomposition of (NH4)2IrCl6 on graphite felt graphite (IrO2 þ C / Ir þ CO2). We did not observe metallic Ir but
yielded IrO2 particles on the surface of the fibres. Graphite felts IrO2. This was shown by XRD measurements (Fig. 4). XRD dif-
were chosen as substrate for the IrO2 catalyst as this material is also fractogramms were recorded of the IrO2-modified graphite felt and
successfully used as positive electrode material in VRFB due to high of a graphite felt treated under the same thermal conditions but

Fig. 3. SEM images of non-modified (but thermally treated under same conditions as the IrO2-modified fibres) graphite fibres (A,B) and IrO2-modified graphite fibres (C,D).
 J. grosse Austing et al. / Journal of Power Sources 273 (2015) 1163e1170 1167

the theoretical coulombic capacity of the anolyte. Based on the

measurements shown in Fig. 5, cell efficiencies were calculated
(Table 1).
The general shapes of charging and discharging curves were in
good agreement with the curves shown in literature [16,18]. All
curves exhibited significant overpotentials both in charging and
discharging process which became more pronounced at higher
current densities. This can be explained with increasing over-
potentials of electrochemical reactions at higher current densities.
Notably, the highest discharge capacity was achieved at the highest
current density. This observation was in contrast to the experience
with other battery technologies, as normally the discharge capacity
increases with lowering the current density because of generally
lower overpotentials. This phenomenon observed for the VARFB
test cell indicated that probably crossover processes lead to a self-
discharge of the electrolyte. These losses were more pronounced
when operating with lower current density and therefore longer
operation time during which crossover processes may proceed. The
crossover is important for the performance of the VARFB and will
Fig. 4. XRD pattern of IrO2-modified graphite felt (solid line), a blank but thermally be investigated in detail in future work. However, the charging and
treated graphite felt (dashed line) and calculated IrO2 reflections (vertical lines; [24]). discharging curves show that a stable operation of the VARFB at
different current densities was possible.
The obtained average discharge power density at 40 mA cm2
without immersion in (NH4)2IrCl6 solution. Besides the graphite- was 33.8 mW cm2. The maximum average discharge power den-
related reflections (i.e. reflections at 2Q z 26 ; 43 ; 53.5 ; 79 sity reported in literature for a 280 cm2 VOFC was 19.6 mW cm2
and 82 ), the pattern of the IrO2-modified graphite felt shows [18]. However, the larger geometric area of the VOFC presented in
additional reflections (e.g. 2Q z 28 ; 35 or 40 ). These reflections Ref. [18] may involve additional operational issues leading to the
are related to IrO2 as calculated in Ref. [24]. lower power density.
The modification routine did not allow influencing the The maximum energy efficiency hE of the VARFB operated at
morphology of IrO2. This would be desirable for the formation of room temperature was 41.7% at a current density of 20 mA cm2. It
higher surface area to volume ratio but is beyond the scope of this was much higher than the efficiency of 26.6% reported by Hosseiny
paper. The described method is a simple and straightforward et al. [16] at 2.4 mA cm2 and 40  C and comparable to the one they
routine to load a graphite felt with IrO2 which was used here to test obtained at 2.4 mA cm2 and 80  C (hE ¼ 45.7%). Higher tempera-
the proposed two-layered cathode design for the VARFB. ture enhances the kinetics of ORR and OER leading to lower
charging and higher discharging cell potentials, respectively. Thus
3.3. Operation of the VARFB further improvement of the efficiency is expected at elevated
operation temperatures.
The performance of the constructed unitised bidirectional Table 1 shows clearly that both, low hC and hV, contribute to the
VARFB was examined with current densities of 15 mA cm2, limited energy efficiency hE. As known from conventional VRFB
20 mA cm2 and 40 mA cm2 at room temperature (Fig. 5). As the ([25]), a trend concerning the coulombic efficiencies hC and voltaic
system is new and therefore no information about an adequate efficiencies hV was observed: With increasing current densities, hC
end-of-charge voltage for the different current densities was increased whereas hV decreases. Due to crossover processes the hC
available, the VARFB was charged in all experiments up to 93% of may be lowered. At higher current densities, side reactions (such as
hydrogen evolution on the anode side or oxidative corrosion on the
cathode side) can be promoted due to diffusion limitations or the
higher potentials, thus leading to lower coulombic efficiencies. At
moderate current densities, the described negative effect on hC due
to side-reactions was not so pronounced compared to the
crossover-related effect.
Taking into account the ButlereVolmer equation it is clear that
hV increased with lowering the current density. In order to obtain
more information about the underlying processes, we investigated
the VARFB operation with an integrated DHE reference electrode to
resolve the single contributions of the two half cells to the overall
cell performance. A charge and discharge curve of the VARFB at
40 mA cm2 together with the anode and cathode half cell po-
tentials vs. DHE are shown in Fig. 6. As common for battery
charging curves, the cell voltage increased during the charging
process. This was caused by the change in the V2þ:V3þ ratio during
charging that resulted in an increase of the Nernst potential and
therefore lead to a rising anode half cell potential. The concentra-
tions ratio of the redox pair H2O/O2 in the catholyte did not change
significantly during charging, therefore the cathode half cell po-
Fig. 5. Charge and discharge performance of the VARFB test cell at current densities of tential remained nearly constant. A similar behaviour was recog-
15 mA cm2 (dotted line), 20 mA cm2 (dashed line) and 40 mA cm2 (solid line). nized during discharging.
1168 J. grosse Austing et al. / Journal of Power Sources 273 (2015) 1163e1170

Table 1
Efficiencies of VARFB at 15 mA cm2, 20 mA cm2 and 40 mA cm2; hC ¼ coulombic
efficiency, hV ¼ voltaic efficiency, hE ¼ energy efficiency.

i [mA cm2] hC hV hE
15 55.0% 51.6% 28.4%
20 84.2% 49.5% 41.7%
40 87.6% 44.5% 39.0%

The major reason for the observed low hV were the high
cathodic overpotentials (Fig. 6) The difference in average cathodic
voltage between charging and discharging was DEcathode ¼ 0.7 V
while the corresponding difference of the average anode potentials
was DEanode ¼ 0.1 V. The two cathode (¼ positive half cell) reactions,
ORR and OER, are well known to be kinetically slow and therefore
accompanied with relatively high overpotentials leading to reduced
hV. This observation is in agreement with experiences gained with
unitised regenerative fuel cells (URFC). For URFC, in which the ORR
and OER are also the half cell reactions on the positive electrode, Fig. 7. Four subsequent cycles of VARFB operation at 20 mA cm2. The inset shows a
energy efficiencies of 49% [26] or 53% [27] were reported. magnification of the beginning of the second charging sequence. The fourth cycle was
recorded with fresh anolyte and catholyte.
Enhancing ORR and OER kinetics, e.g. by operating the battery at
higher temperatures (as shown by Hosseiny et al. [16]) or using
other catalysts, could improve the voltaic efficiency. Moreover,
exponential potential increase at the end of the step is missing.
optimisation of the cell design may minimise internal resistances
These observations can be explained with crossover of V2þ/V3þ
and therefore increasing the voltaic efficiency as well. Another
from the anode to the cathode in combination with the fact that the
factor that influences the voltaic efficiency is the degradation of the
VARFB was charged in all cycles with the constant charge of 93% of
electrodes, the catalyst support and the catalyst itself. To gain in-
the initial anolyte capacity. The loss of vanadium on the anode side
formation about the contribution of these processes and of cross-
leads to a loss of the anolyte capacity. Different V2þ:V3þ ratio results
over reactions on the performance of the VARFB, the test cell was
in different electrode potentials and higher charging potential.
operated for consecutive cycles.
Additionally, with decreasing total amount of vanadium in the
Four consecutive cycles of the VARFB at a current density of
anolyte the share of side reactions may increase. After all V3þ is
20 mA cm2 are shown in Fig. 7. During the first three cycles
converted to V2þ, the reaction on the anode will be the evolution of
(charging up to 93% of the theoretical coulombic capacity of the
hydrogen (2Hþ þ 2e / H2) [28]. This reaction is known to proceed
anolyte) the VARFB cell components including peripherals (e.g.
with relatively high overpotentials on a graphite felt electrode [19],
electrolytes) were not exchanged, for the fourth cycle only the
therefore the charging potential increased at the end of the
anolyte (1.2 M V3þ þ 2 M H2SO4) and the catholyte (2 M H2SO4)
charging step. In the fourth cycle the original anolyte capacity was
were exchanged by new, fresh solutions. Table 2 summarises per-
recovered due to an exchange of the electrolyte. The potential did
formance data of the four consecutive cycles operation of the
not increase at the end of the charging step as it was the case in the
VARFB.
second and third cycle. This explanation is also reflected in the
The average charging potential and especially the potential at
coulombic efficiencies hC and the discharge capacity Qdischa for the
the end of the charging step increased from cycle one to three. In
four cycles (Table 2). A decrease of hC and Qdischa with each of the
the fourth cycle (conducted after exchange of electrolytes) the
first three cycles was observed, in the fourth cycle both values were
average charging potential is less than in the third cycle and the
approximately the same as in the first cycle. The produced
hydrogen lead to a loss of charge and therefore a reduction of the
coulombic efficiency. In the third cycle the anolyte capacity was the
lowest, the side reaction (hydrogen evolution) became more sig-
2.0
nificant and therefore the coulombic efficiency was the lowest of all
cycles.
Despite of that, the continuous crossover of vanadium to the
1.5 cathode side can also be noticed by comparing the beginning of the
charging curves. The second and the third cycle showed a small
potential step in the beginning of the charging reaction, while the
E [V]

1.0
first and fourth cycle did not show this feature (see inset in Fig. 7).

0.5
cell potential Table 2
cathode potential vs. DHE Performance data of subsequent cycles of VARFB operation at 20 mA cm2
0.0 anode potential vs. DHE (Q ¼ charge; W ¼ electrical energy, hC ¼ coulombic efficiency, hV ¼ voltaic efficiency,
hE ¼ energy efficiency).
Cycle Qdischa [mAh] hC Wcha [mWh] Wdischa [mWh] hE hV
-0.5
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1st 302.4 83.9% 654.9 272.6 41.6% 49.6%
2nd 277.0 76.7% 661.5 240.0 36.3% 47.3%
time [h]
3rd 269.8 74.7% 674.5 216.2 32.1% 43.0%
4tha 307.9 85.4% 680.3 239.7 35.2% 41.2%
Fig. 6. Charge and discharge performance of VARFB at 40 mA cm2 with additional
a
anode and cathode potentials vs. DHE. The fourth cycle was recorded with fresh anolyte and catholyte.
 J. grosse Austing et al. / Journal of Power Sources 273 (2015) 1163e1170 1169

This potential step was most probably related to the oxidation of investigated. We showed a straightforward and simple routine to
VO2þ to VOþ 2 on the cathode side, since the oxidation of VO
2þ
takes modify a graphite felt with IrO2 which performed well as electrode
place at potentials lower than those necessary for OER. VO2þ was for the reaction in the positive half cell during charging (OER). With
present on the cathode side due to crossover of V2þ/V3þ which the proposed layer arrangement and materials which were used for
were immediately oxidised to VO2þ under the existing potential the positive half cell reactions (ORR in discharging and OER in
conditions. charging), a better performance in terms of efficiency, discharge
Although the increase of charging potential caused by the ca- power density and employable current density could be obtained in
pacity loss was suppressed by exchanging the electrolytes, an comparison to existing reports on VARFB [16] and VOFC [17,18]. At a
additional irreversible potential increase can be noticed. The sharp current density of 20 mA cm2 and room temperature, the VARFB
increase at the end of the charging step vanished in the fourth cycle was operated with a coulombic efficiency of hC ¼ 84.2%, a voltaic
after the electrolytes were exchanged but the average charging efficiency of hV ¼ 49.5% and an energy efficiency of hE ¼ 41.7%. An
potential was higher than in the first cycle. This was also repre- average discharge power density of 34.6 mW cm2 at 40 mA cm2
sented by the increasing energy consumption in the charging steps was achieved.
(Wcha in Table 2) for all four cycles. Even after the electrolyte ex- Half cell potentials showed that energy and voltaic efficiency
change higher charging potentials were still observed. Therefore, it were mainly limited by slow OER and ORR kinetics. Higher opera-
can be concluded that the system additionally underwent degra- tion temperatures are promising to enhance the cathodic reactions
dation processes that were not related to crossover reactions. The and increase efficiency.
same, even more pronounced, was observed for the discharging The operation of the VARFB for four subsequent cycles revealed
curves. The discharge overpotentials increased with each cycle and a performance decrease with each cycle which was not only related
therefore the average discharge potential decreased. The general to loss of active material due to crossover, but also due to irre-
increase of overpotentials for both charging and discharging with versible degradation processes taking place during operation.
each of the four cycles was also reflected by the continuous Therefore, crossover-minimising membranes are needed to reduce
decrease of the voltaic efficiency hV shown in Table 2. coulombic losses and alternative corrosion-resistant GDL as well as
A general increase of overpotentials that can not be suppressed other catalyst support strategies would be promising to achieve a
by electrolyte exchange indicates that the cell resistances (ohmic higher stability.
resistances, but also charge-transfer resistances) increased during
operation. There are several factors that can contribute to this Acknowledgements
irreversibly increased resistances. Firstly, corrosion of conducting
materials in the cell such as the electrodes or current collectors may The authors would like to thank Andrea Ballarin, Benedikt
raise the internal resistance of the cell and/or hinder the diffusion Berger and Frank Ba€ttermann for assisting with the experimental
of reactants leading to increased mass transport resistances. Sec- work. Likewise we appreciate the contributions of Martin Knipper
ondly, the degradation and/or agglomeration of the catalyst parti- (XRD measurements), Dana Schonvogel (ICP-MS measurements),
cles may lead to a loss of catalytically active surface area resulting in Hayo Seeba and Dietmar Piehler (construction of the test cell). Jan
an increase of charge-transfer resistances. The average discharging grosse Austing thanks the Reiner Lemoine-Stiftung, Berlin, Ger-
potential decreased with each cycle much more than the average many for a Ph.D. scholarship.
charging potential increased. Accordingly, it can be assumed that
the degradation of the GDE (the active electrode during discharg- References
ing) was more pronounced than that of the IrO2-modified graphite
felt. This could be attributed to the less graphitic nature of the GDL [1] P. Alotto, M. Guarnieri, F. Moro, Renewable Sustainable Energy Rev. 29 (2014)
compared to the graphite felt. From PEM fuel cells it is known that 325e335.
[2] H. Chen, T.N. Cong, W. Yang, C. Tan, Y. Li, Y. Ding, Prog. Nat. Sci. 19 (2009)
electrochemical oxidation of the GDL leads to a loss of hydropho-
291e312.
bicity [29]. This in turn promotes the flooding of the GDL with [3] C. Ponce de Leo n, a. Frías-Ferrer, J. Gonza lez-García, D. Sza nto, F. Walsh,
water hindering the oxygen transport to the catalytic sites and thus J. Power Sources 160 (2006) 716e732.
[4] M. Skyllas-Kazacos, M.H. Chakrabarti, S.a. Hajimolana, F.S. Mjalli, M. Saleem,
leading to performance loss. Additionally, it is known that the
J. Electrochem. Soc. 158 (2011) R55.
carbon support of the catalyst might undergo corrosion and plat- [5] P. Leung, X. Li, C. Ponce de Leo  n, L. Berlouis, C.T.J. Low, F.C. Walsh, RSC Adv. 2
inum particles might agglomerate in PEM fuel cells when operated (2012) 10125.
at high potentials [30e32]. The GDE seems to be prone to corrosion [6] L. Li, S. Kim, W. Wang, M. Vijayakumar, Z. Nie, B. Chen, J. Zhang, G. Xia, J. Hu,
G. Graff, J. Liu, Z. Yang, Adv. Energy Mater. 1 (2011) 394e400.
under the severe potential conditions. Further investigations of [7] M. Skyllas-Kazacos, J. Power Sources 124 (2003) 299e302.
alternative GDL and different supports for the platinum catalyst [8] S. Peng, N. Wang, C. Gao, Y. Lei, X. Liang, S. Liu, Y. Liu, Int. J. Electrochem. Sci. 7
which could reduce corrosion effects have to be performed. (2012) 4388e4396.
[9] H. Kaneko, A. Negishi, K. Nozaki, K. Sato, M. Nakajima, EU Patent 0517217A1,
The degradation of the IrO2-modified graphite felt electrode 1992.
seems to be less pronounced. On one hand, this can be attributed to [10] M. Chakrabarti, N. Brandon, S. Hajimolana, F. Tariq, V. Yufit, M. Hashim,
the stability of IrO2 even under severe oxidative conditions [33]. On M. Hussain, C. Low, P. Aravind, J. Power Sources 253 (2014) 150e166.
[11] H.-S. Kim, Bull. Korean Chem. Soc. 32 (2011) 571e575.
the other hand, the graphite felt support was shown to undergo [12] B. Sun, M. Skyllas-Kazacos, Electrochim. Acta 37 (1992) 1253e1260.
some degradation in VRFB under certain operating conditions [13] B. Sun, M. Skyllas-Kazacos, Electrochim. Acta 36 (1991) 513e517.
[34e36] but it was also reported that the presence of IrO2 on carbon [14] E.-M. Hammer, B. Berger, L. Komsiyska, Int. J. Renewable Energy Dev. 3 (2014)
7e12.
surface reduces carbon corrosion under oxidative conditions [37]. €rissen, J. Power Sources 155 (2006) 23e32.
[15] L. Jo
Further investigations on the stability of the IrO2-modified graphite [16] S. Hosseiny, M. Saakes, M. Wessling, Electrochem. Commun. 13 (2011)
felt under VARFB operation conditions are necessary to detail 751e754.
[17] C. Menictas, M. Skyllas-Kazacos, J. Appl. Electrochem. 41 (2011) 1223e1232.
specific degradation mechanisms.
[18] J. Noack, C. Cremers, D. Bayer, J. Tübke, K. Pinkwart, J. Power Sources 253
(2014) 397e403.
4. Conclusion [19] W. Wang, X. Wang, Electrochim. Acta 52 (2007) 6755e6762.
[20] L. Ouattara, S. Fierro, O. Frey, M. Koudelka, C. Comninellis, J. Appl. Electro-
chem. 39 (2009) 1361e1367.
In this paper the design of a VARFB comprising a novel two- [21] Q.H. Liu, G.M. Grim, A.B. Papandrew, a. Turhan, T.A. Zawodzinski, M.M. Mench,
layered cathode was presented and its performance was J. Electrochem. Soc. 159 (2012) A1246eA1252.
1170 J. grosse Austing et al. / Journal of Power Sources 273 (2015) 1163e1170

[22] G. Li, P.G. Pickup, Electrochim. Acta 49 (2004) 4119e4126. [30] N. Linse, L. Gubler, G.G. Scherer, A. Wokaun, Electrochim. Acta 56 (2011)
[23] C.-N. Sun, F.M. Delnick, D.S. Aaron, A.B. Papandrew, M.M. Mench, 7541e7549.
T.A. Zawodzinski, ECS Electrochem. Lett. 2 (2013) A43eA45. [31] R.L. Borup, J.R. Davey, F.H. Garzon, D.L. Wood, M.a. Inbody, J. Power Sources
[24] ICDD, PDF-2 Database Release 2009, Code: 03-065-2822, International Centre 163 (2006) 76e81.
for Diffraction Data, Newtown Square, PA, USA, 2009. [32] S. Maass, F. Finsterwalder, G. Frank, R. Hartmann, C. Merten, J. Power Sources
[25] J. Xi, Z. Wu, X. Teng, Y. Zhao, L. Chen, X. Qiu, J. Mater. Chem. 18 (2008) 176 (2008) 444e451.
1232. [33] S. Trasatti, Electrochim. Acta 45 (2000) 2377e2385.
[26] H.-Y. Jung, P. Ganesan, B. Popov, ECS Trans. 25 (2009) 1261e1269. [34] A. Parasuraman, T.M. Lim, C. Menictas, M. Skyllas-Kazacos, Electrochim. Acta
[27] S.-D. Yim, W.-Y. Lee, Y.-G. Yoon, Y.-J. Sohn, G.-G. Park, T.-H. Yang, C.-S. Kim, 101 (2013) 27e40.
Electrochim. Acta 50 (2004) 713e718. [35] M. Rychcik, M. Skyllas-Kazacos, J. Power Sources 19 (1987) 45e54.
[28] M. Skyllas-Kazacos, M. Kazacos, J. Power Sources 196 (2011) 8822e8827. [36] H. Liu, Q. Xu, C. Yan, Electrochem. Commun. 28 (2013) 58e62.
[29] S. Yu, X. Li, S. Liu, J. Hao, Z. Shao, B. Yi, RSC Adv. 4 (2014) 3852. [37] S.-E. Jang, H. Kim, J. Am. Chem. Soc. 132 (2010) 14700e14701.
Publication II

59
Investigation of crossover processes in a unitized
bidirectional vanadium/air redox flow battery
Authors: Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska and
Gunther Wittstock
Journal of Power Sources, 306, 2016, 692-701, DOI: 10.1016/j.jpowsour.2015.12.052

Qdischarge

Qcharge

other losses
(e.g. O2 permeation)
vanadium crossover

diffusion
migration & EOC

61
 Journal of Power Sources 306 (2016) 692e701

Contents lists available at ScienceDirect

Journal of Power Sources

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

Investigation of crossover processes in a unitized bidirectional

vanadium/air redox flow battery
Jan grosse Austing a, Carolina Nunes Kirchner a, Lidiya Komsiyska a, Gunther Wittstock b, *
a
NEXT ENERGY EWE Research Centre for Energy Technology, 26129 Oldenburg, Germany
b
Carl von Ossietzky University of Oldenburg, Faculty of Mathematics and Natural Sciences, Center of Interface Science, Institute of Chemistry, 26111
Oldenburg, Germany

h i g h l i g h t s

Crossover in a VARFB is investigated using ICP-MS and in situ UV/Vis spectroscopy.

Coulombic efficiency of a VARFB is lowered by crossover (mainly O2 permeation).

Vanadium crossover which limits the cycle life of VARFB is quantified.

Diffusion coefficients of V2þ and V3þ through Nafion® 117 are determined.

a r t i c l e i n f o a b s t r a c t

Article history: In this paper the losses in coulombic efficiency are investigated for a vanadium/air redox flow battery
Received 9 September 2015 (VARFB) comprising a two-layered positive electrode. Ultraviolet/visible (UV/Vis) spectroscopy is used to
Received in revised form monitor the concentrations cV2þ and cV3þ during operation. The most likely cause for the largest part of
27 November 2015
the coulombic losses is the permeation of oxygen from the positive to the negative electrode followed by
Accepted 15 December 2015
an oxidation of V2þ to V3þ. The total vanadium crossover is followed by inductively coupled plasma mass
Available online xxx
spectroscopy (ICP-MS) analysis of the positive electrolyte after one VARFB cycle. During one cycle 6% of
the vanadium species initially present in the negative electrolyte are transferred to the positive
Keywords:
Vanadium-air redox flow battery
electrolyte, which can account at most for 20% of the coulombic losses. The diffusion coefficients of V2þ
Crossover redox flow battery and V3þ through Nafion® 117 are determined as DV2þ ; N117 ¼ 9:05$106 cm2 min1 and
Vanadium oxygen fuel cell DV3þ ; N117 ¼ 4:35$106 cm2 min1 and are used to calculate vanadium crossover due to diffusion which
Bidirectional oxygen/air electrode allows differentiation between vanadium crossover due to diffusion and migration/electroosmotic con-
Unitized regenerative fuel cell vection. In order to optimize coulombic efficiency of VARFB, membranes need to be designed with
Diffusion coefficient vanadium Nafion reduced oxygen permeation and vanadium crossover.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction 25e30 Wh kg1 for the all-vanadium redox flow battery (VRFB)
[3]).
In the past years, redox flow batteries (RFB) have gained Different approaches have been proposed to enhance the energy
increasing attention as promising candidates for stationary elec- density of RFB including increasing the solubility of the electrolytes
tricity storage applications. They show advantageous characteris- by using additives [4,5] or following alternative concepts such as
tics such as high cycle life (>12,000 [1]) and good round-trip solid/liquid hybrid redox flow batteries [6,7]. The substitution of
efficiencies (hE z 80% [2]). However, as redox flow batteries are the positive half cell (i.e. the VO2þ/VOþ2 redox couple) by a bidi-
generally based on electroactive materials (e.g. metal cations) dis- rectional air electrode was disclosed in a patent in 1992 [8] and
solved in a solvent (e.g. water) with a limited solubility, the energy holds the potential for roughly doubling the energy density in
density of RFB is low compared to other battery systems (e.g. relation to a VRFB. A modular VARFB system (i.e. two separate re-
action units for charging and discharging) was described by Hos-
seiny et al. [9] and a system for discharging only was reported by
Noack et al. [10] and Menictas et al. [11].
* Corresponding author.
E-mail address: [email protected] (G. Wittstock). In our previous paper [12] we introduced a concept for a

http://dx.doi.org/10.1016/j.jpowsour.2015.12.052
0378-7753/© 2015 Elsevier B.V. All rights reserved.
 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701 693

unitised bidirectional VARFB utilising a two-layered positive elec-

trode. Figure 1a depicts the setup of the VARFB. The negative
electrode of the battery comprises a current collector (➁) and a
graphite felt electrode (➂) through which the V2þ/V3þ containing
electrolyte (➀) is circulated. For brevity we denote the electrolyte in
the negative electrode as ‘‘negative electrolyte’’ (analogous for the
‘‘positive electrolyte’’). The positive electrode is composed of two
electrodes that are pressed together: An IrO2-modified graphite felt
(➄) serves as electrode for the charging reaction (oxygen evolution
reaction, OER; 2H2O / O2 þ 4e þ 4Hþ), while a platinum-
containing gas diffusion layer (➅) which is fed with air through
the flow-field/current collector (➆) supports the discharging reac-
tion (oxygen reduction reaction, ORR; O2 þ 4e þ 4Hþ / 2H2O).
Negative and positive electrode are separated from each other by a
Nafion® 117 membrane (➃). The VARFB was operated for several
cycles and an energy efficiency hE ¼ 39% at 21  C and 40 mA cm2
was achieved [12].
However, comparably low coulombic efficiencies (hC ¼ 55.0% at
15 mA cm2 and hC ¼ 87.6% at 40 mA cm2) were obtained which Fig. 2. Charge/discharge curve of the VARFB test cell (j ¼ 15 mA cm2, 21  C ± 1  C).
also decreased with each consecutive cycle. The coulombic effi-
ciency hC is the ratio of the amount of electric charge delivered
reactions are shown as black solid line, side reactions reducing hC
during discharging (Qdischarge) to the charge consumed during
are numbered and highlighted by red dashed lines. Processes 1 and
charging (Qcharge):
2 illustrate the crossover of V2þ and V3þ from the negative to the
Qdischarge positive electrode. After crossing the membrane, V2þ and V3þ can
hC ¼ (1) be oxidized at the positive electrode to VO2þ (‘‘V4þ’’) and/or VOþ 2
Qcharge
(‘‘V5þ’’) due to the potential of the positive electrode or due to
The capacity of the VARFB depends exclusively on the capacity oxygen present in the positive electrode chamber. This reaction
of the negative electrolyte because the reagents of the positive (gray dotted arrow) does not influence hC because the capacity is
electrode are O2 and H2O which are not limiting in our setup determined by the capacity of the negative electrolyte. Process 3
(Fig. 1b). Therefore, a decrease of hC must only consider all side depicts the oxygen crossover and the subsequent homogeneous
reactions that influence the capacity of the negative electrolyte. oxidation of V2þ to V3þ by O2 which reduces hC because V2þ is
Fig. 2b shows the cell reactions occurring in a VARFB. The main consumed. Hydrogen evolution reaction (HER; process 4) would
also lower hC. The chemical follow-up reaction has no influence on
has no influence on hC and thus depicted as gray dotted line.
For conventional VRFB the crossover of vanadium species
through the membrane has been reported to contribute to hC losses
and was investigated in several studies [13e18]. This crossover (i.e.
the transport of ions through the membrane) is caused by one or
more of the three processes: migration (in the electric field during
battery operation), electroosmotic convection (EOC; caused by
electroosmotic flow of water through the membrane) and/or
diffusion [19]. Recently, these processes were studied in detail for
VRFB by Darling et al. [20] and Yang et al. [19] by modeling ap-
proaches. Yang et al. [19] found that V3þ net transfer from the
negative to the positive electrode is much more pronounced during
discharging than during charging due to superposition of the three
processes. During charging the direction of the diffusive part
(negative to positive) is opposite to the direction of migration and
EOC (positive to negative).
The transport of vanadium cations through the cation exchange
membrane is influenced by the interactions of the cations with the
negatively charged sulfonate groups of the membrane [21,22]. The
transport mechanisms of vanadium ions through a Nafion® mem-
brane on a molecular level have been investigated and discussed in
several publications [22e25]. Vijayakumar et al. [22] reported
about two fouling mechanisms of Nafion® membranes due to va-
nadium cations that blocked the sulfonic acid groups. Recently, the
interaction between triflic acid (F3CeSO3H) as a reference system
for Nafion® and vanadium cations was investigated with hybrid
Fig. 1. a) The VARFB test cell comprising a two-layered positive electrode.
density functional theory [21]. The influence of absorbed V2þ and
➀ ¼ negative electrolyte tank; ➁ ¼ current collector; ➂ ¼ graphite felt; ➃ ¼ membrane; V3þ on the local structure of Nafion® was examined by molecular
➄ ¼ IrO2-mod. graphite felt; ➅ ¼ gas diffusion electrode; ➆ ¼ current collector with dynamics [26].
flow field. b) Possible chemical reactions in the VARFB. Main reactions (), hC reducing Diffusion coefficients for vanadium species through diverse
crossover and side reactions ( ) and subsequent chemical reaction ( ). 1,
(pristine and modified) membranes were investigated in several
2 ¼ vanadium crossover; 3 ¼ oxygen crossover; 4 ¼ hydrogen evolution.
694 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701

studies [13e17,27e29] and used in detailed modeling and simula- runs and the average value was used. The relative standard de-
tion studies of VRFB [30e33]. For Nafion® 115 membranes diffusion viations (RSD) between the runs was less than 1.2%.
coefficients for V2þ and V3þ were reported [13,14], for Nafion® 117
membranes values for V3þ were published [15,16]. In a VARFB, the 2.2. Components for the VARFB test cell and operation
crossover of V2þ and V3þ from the negative to the positive electrode
leads to a decreased amount of vanadium ions in the negative The components for the VARFB test cell were fabricated and the
electrolyte and therefore to an irreversible decrease in battery ca- VARFB operated as described in our previous paper [12]. The gen-
pacity because the negative electrolyte is the capacity limiting eral setup of the VARFB test cell with the different layers of the
factor. No ‘‘back-crossover’’ from the (diluted) positive electrolyte reaction unit is shown in Fig. 1a.
to the (concentrated) negative electrolyte is assumed to occur. Carbon felts (GFD5, SGL Carbon GmbH, Germany) were ther-
Hosseiny et al. [9] suspected crossover of oxygen as a possible mally activated at 400  C for 18 h in air. The IrO2-modified graphite
cause for additional hC reduction. Oxygen could permeate from the felt was prepared by thermal decomposition of (NH4)IrCl6
positive to the negative electrode followed by oxidation of V2þ to (99.994%; Alfa Aesar, GmbH & Co KG, Germany). The gas diffusion
V3þ. However, this assumption has not been investigated experi- electrode (GDE) was loaded with Nafion® and Pt/C (approx.
mentally yet. 0.96 mg cm2 Pt) via airbrushing. Nafion® 117 (Ion Power GmbH,
Furthermore, the HER which might occur on the negative elec- Germany) was used as membrane after protonation [(1) 5% H2O2/
trode would lower hC as well. Unwanted side reactions on the 80  C/30 min; (2) 1.0 M H2SO4/80  C/30 min; (3) deionized water/
positive electrode (e.g. carbon corrosion) have no influence on the 100  C/10 min]. Nafion membranes with the equivalent weight of
loss of hC as the cathodic reactants (H2O and O2, respectively) are 1100 are a benchmark material used in VRFB research. Although
supplied in excess. having a higher ohmic resistance, the thicker Nafion® 117 in com-
In the current contribution, losses of hC in a VARFB were parison to the thinner Nafion® 115 was chosen because lower va-
investigated with the aim to identify and quantify the different loss nadium crossover and oxygen permeation fluxes can be expected.
processes. UV/Vis spectroscopy was applied to investigate the The negative current collector was made from a graphite bipolar
concentrations of the vanadium species in VRFB [34e41]. In the plate material (PPG 86, Eisenhuth GmbH Co. KG, Germany), the
negative electrolyte exists a linear relationship between absorption positive one was manufactured from titanium (grade 2).
and cV2þ and cV3þ at least up to 1.6 M when using a cuvette with a A previously [12] described test cell with 4 cm2 geometric
maximum path length of d ¼ 1 mm [36,39]. Due to dimer formation electrode area was used for VARFB operation. The negative elec-
between VO2 þ and VO2þ species, the monitoring of the positive trolyte was 12 mL of a V3þ electrolyte which was permanently
electrolyte would require a non-linear calibration [36,37,39,42]. purged with humidified N2 to avoid contact with air. 50 mL of
Using a flow-through cuvette with d  1 mm avoids on the one 2 M H2SO4 were employed as positive electrolyte. Both electrolytes
hand the necessity of diluting solutions before UV/Vis analysis were circulated with 100 mL min1. During discharging the positive
[34,38] and avoids on the other hand the contact of the negative electrolyte was not circulated but the GDE on the positive electrode
electrolyte with air (O2 easily oxidizes V2þ to V3þ) allowing UV/Vis was fed with air (267 mL min1). The experiments were carried out
spectroscopy to be applied in situ. The monitoring of cV2þ and cV3þ at room temperature (21  C ± 1  C). After charging and discharging,
during VARFB operation allowed conclusion on occurring loss a sample of the positive electrolyte was taken for ICP-MS
mechanisms. measurements.
Inductively coupled plasma mass spectroscopy (ICP-MS) mea-
surements was used to determine the overall vanadium crossover 2.3. UV/Vis calibration and in situ UV/Vis spectroscopy of the
occurring during charging and discharging by measuring the va- negative electrolyte during VARFB operation
nadium concentration in the positive electrolyte. Diffusion co-
efficients of V2þ and V3þ through Nafion® 117 were determined to A calibration curve was obtained by measuring the UV/Vis ab-
estimate the contribution of fickian diffusion on the vanadium sorption (LAMBDA XLS þ UV/Vis Spectrophotometer, PerkinElmer,
crossover. This information allows relating the crossover processes USA) of a dilution series of V2þ and V3þ in 2 M H2SO4. Diluted V2þ
1, 2 and 3 in Fig. 1b to each other. Finally, recommendations are and V3þ electrolytes were derived from electrolytically prepared
discussed for future membrane development to improve hC. V2þ and V3þ electrolytes (section 2.1). The concentrations as
determined by ICP-MS were cV2þ ¼ 1:06 mol L1 and
2. Experimental cV3þ ¼ 1:09 mol L1, respectively. The UV/Vis measurements were
conducted by circulating the electrolyte through a flow-through
2.1. V2þ and V3þ electrolyte preparation cuvette with 1 mm path length (137-QS, Quartz SUPRASIL®,
Hellma GmbH & Co. KG, Germany). The electrolyte was blanketed
The V2þ and V3þ electrolytes used in this work were prepared with humidified nitrogen to avoid contact with air.
electrolytically in a conventional VRFB setup similar as described During VARFB operation the flow-through cuvette was inte-
elsewhere [43] and as reported in our previous publication [12]. grated in the hydraulic circuit of the negative electrolyte which was
Briefly, a solution of 1.2 M VOSO4 $ x H2O (x z 3; 97%; Sigma pumped through the cuvette before entering the VARFB test cell.
Aldrich, Germany) and 2.0 M H2 SO4 (95e97%; Fischer Scientific Approx. every 15 min a UV/Vis spectrum was recorded and the V2þ
GmbH, Germany) in deionized water was prepared and appropri- and V3þ concentrations were calculated based on the calibration
ately electrolyzed in a VRFB setup using a potentiostat/galvanostat data.
(Solartron Analytical Modulab Pstat potentiostat/galvanostat, UK).
The V2þ and V3þ electrolytes are analysed using ICP-MS 2.4. Determination of diffusion coefficients of V2þ and V3þ through
(XSERIES 2 ICP-MS, Thermo Scientific, USA) to determine the total Nafion® 117
vanadium concentration, as it can deviate from the initial vanadium
concentration of 1.2 mol L1 due to crossover processes during the For the determination of the permeability of V2þ and V3þ
electrolyte preparation [13,44]. Scandium was used as an internal through a Nafion® 117 membrane, a similar routine is used as
standard. For these and all following ICP-MS-analysed samples the described by other authors [13,14,16,45]. The membrane was
concentration of vanadium was determined in three subsequent sandwiched between the compartments of a separable H-cell with
 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701 695

an exposed membrane area of 5.391 cm2. The thickness of the

pretreated, wetted Nafion® 117 membrane was measured as
Aðl1 Þ ¼ εV2þ ðl1 Þ$cV2þ $d þ εV3þ ðl1 Þ$cV3þ $d
x ¼ 207 mm. One side of the H-cell was filled with 130 mL of a V2þ or (2)
Aðl2 Þ ¼ εV2þ ðl2 Þ$cV2þ $d þ εV3þ ðl2 Þ$cV3þ $d
V3þ electrolyte (denoted as excess side). The initial concentration of
vanadium in the V2þ electrolyte was cV2þ ¼ 1:08 mol L1, the con- The spectra of the electrolytes are shown in Fig. 3 for either
centration in the V3þ electrolyte was cV3þ ¼ 1:12 mol L1 as 1.06 M V2þ or 1.09 M V3þ in 2 M H2SO4. V2þ has absorption maxima
determined by ICP-MS. at approximately 370 nm, 570 nm and 850 nm and V3þ at 400 nm
The other side (denoted as deficiency side) was filled with and 605 nm. The absorptions at 605 nm and 850 nm were used for
130 mL of a 1.2 M MgSO4 and 2 M H2 SO4 solution. MgSO4 was used the in situ UV/Vis spectroscopic monitoring of cV2þ and cV3þ in the
to reduce the difference in ionic strengths between the two solu- negative electrolyte during VARFB operation. Calibration curves,
tions and therefore to reduce the osmotic pressure [45]. Both so- obtained by measuring UV/Vis absorption of appropriately diluted
lutions were permanently kept under a blanket of humidified V2þ and V3þ electrolytes, are shown in Fig. 3b with good linearity in
nitrogen to avoid the oxidation of the V2þ or V3þ species by oxygen the entire investigated concentration range allowing the simulta-
from the ambient air. The deficiency side was stirred for 1 min with neous determination of cV2þ and cV3þ using Eq. (2).
a magnetic stir bar (to achieve a homogeneous vanadium concen- Following BeereLambert law (A(l) ¼ ε(l)$c$d), the slope of the
tration in the solution) before taking a sample. Approx. every fitted lines in Fig. 3b is used to calculate the molar absorptivities:
60 min samples of the solution in the deficiency side were taken for ε(l) ¼ slope/d (d ¼ 1 mm). The molar absorptivities ε(l) of V2þ an
the determination of the vanadium concentration by ICP-MS. V3þ at 375 nm, 405 nm, 570 nm, 605 nm and 850 nm are compiled
in Table 1.
The voltage, cV2þ and cV3þ obtained from UV/Vis spectroscopy are
plotted as a function of time during the constant current charging
3. Results & discussion
(Fig. 4a) and discharging (Fig. 4b) at j ¼ 15 mA cm2. Assuming the
absence of any loss mechanism and any transfer process through
A typical charge/discharge cycle at room temperature with a
the membrane except Hþ permeation, cV2þ should increase linearly
current density of j ¼ 15 mA cm2 is shown in Fig. 2. A substantial
and cV3þ should decrease linearly (as schematically shown in
amount of the charge from the charging process cannot be
recovered during discharging (Qcharge ¼ 361 mAh;
Qdischarge ¼ 156 mAh). Possible causes for this observation are the
processes 1 (V2þ crossover), 3 (oxygen permeation) and 4 (HER)
depicted in Fig. 1b. All these processes have the consequence that hC
of the battery is reduced. The processes 1, 3 and 4 have a direct or an
indirect influence on the vanadium concentration in the negative
electrolyte: Process 1 decreases cV2þ , process 3 decreases cV2þ and
increases cV3þ and process 4 consumes charge that would be used to
oxidize V2þ to V3þ (therefore indirectly decreases cV2þ and cV3þ in
comparison to the stoichiometricaly calculated value). It should be
noted that hC of the presented cycles in this contribution
(hC; 15 mA cm2 ¼ 43%, hC; 20 mA cm2 ¼ 65%) differ from the values
obtained in our previous publication (hC; 15 mA cm2 ¼ 55%,
hC; 20 mA cm2 ¼ 84%) [12]. As all other experimental parameters
were the same except the vanadium concentration in the negative
electrolyte was 20% higher in the former publication (1.2 M vs.
0.96 M), it is assumed that the low vanadium concentration in the
negative electrolyte could lead to faster depletion of convertible
species and therefore more pronounced side reactions such as HER.
This relationship between vanadium concentration and coulombic
efficiency was also reported for VRFB [46]. As a possible explanation
for this correlation the authors suggested gas evolution such as HER
and OER which can be promoted due to depletion of species caused
by low vanadium concentrations or low flow rates [46,47].

3.1. Investigation of vanadium concentration in the negative

electrolyte during VARFB operation with in situ UV/Vis spectroscopy

In order to gain insight into the loss mechanisms during VARFB

which affect hC, UV/Vis spectroscopy with a flow-through cuvette
was used to determine cV2þ and cV3þ in the negative electrolyte
during operation. As the vanadium species all have different visible
spectrum, UV/Vis spectroscopy can be applied for their determi-
nation. The additivity of absorption A(l) allows the determination
of cV2þ and cV3þ by measuring the absorption A(l1) and A(l2) of the
mixture at two different wavelengths l1 and l2 and using the Fig. 3. Calibration of UV/Vis absorption of V2þ and V3þ electrolytes in 2 M H2SO4. a)
BeereLambert law. The concentrations can be calculated by solving UV/Vis spectra of concentrated V2þ and V3þ electrolytes. b) UV/Vis absorptions of V2þ
the following system of linear equations: and V3þ solutions with different concentrations at 605 nm and 850 nm and linear fits.
696 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701

Table 1 charging. In general, at higher SOC the electrolyte is depleted of

Molar absorptivities ε(l) [L mol1 cm1] of V2þ and V3þ in 2 M H2SO4 at specific convertible species and the side reactions become more dominant.
wavelengths l.
This effect is expected to be more noticeable at high current den-
ε(375 nm) ε(405 nm) ε(570 nm) ε(605 nm) ε(850 nm) sities. However, the decrease of the total vanadium concentration
V2þ 2.51 1.72 4.34 3.56 2.99 in Fig. 4a contributes to the depletion of the active species in the
V3þ 10.09 11.73 6.04 7.35 0.29 electrolyte, which might be the reason for the asymptotic behavior

Fig. 4. In-situ UV/Vis monitoring of vanadium concentrations in the negative electrolyte during VARFB operation at j ¼ 15 mA cm2. a) Charging; b) Discharging.

Fig. 5. Development of calculated concentrations cV2þ ; calc and cV3þ ; calc based on Eq. (3) in comparison to cV2þ ; meas and cV3þ ; meas during VARFB charging with no side reactions (a),
influenced by water transfer and/or vanadium crossover (b) or by hydrogen evolution and/or oxygen permeation (c).

Fig. 5a). During the charging process (Fig. 4a) is clearly recognizable also at moderate SOC values. The diffusional vanadium crossover
an asymptotic increase of cV2þ above 60% state of charge (SOC), through the membrane due to concentration gradient is expected
whereas the concentration of cV3þ decreases in similar manner. The to be more pronounced at low current densities due to the longer
total vanadium concentration cV2þ þcV3þ is continuously decreasing charging time, since the amount of vanadium ions transferred by
during the charging. This could be caused by water transfer and/or diffusion is independent of current density but increases with time.
vanadium crossover (Fig. 5b). During the discharge process in Therefore, the following studies were performed at a moderate
Fig. 4b, cV2þ decreases linearly and cV3þ increases. The total vana- current density of j ¼ 20 mA cm2.
dium concentration remains constant during discharging. Since the To monitor the deviation between measured and calculated
vanadium concentrations change linearly during the discharge concentration, the volume V() of the negative electrolyte needs to
process, the asymptotic change of the concentrations during be known. However, the used experimental setup does not allow to
charging is most likely due to the HER taking place only during monitor the negative electrolyte volume in situ. Therefore, V() was
 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701 697

determined after charging and after discharging (Table 2). This 1. Hydrogen evolution (hydrogen evolution reaction; HER) at the
approach is justified since the volume changes of negative elec- negative electrode (process 4 in Fig. 1b).
trolyte were not substantial. The theoretical cV2þ ; calc and cV3þ ; calc 2. Contact of oxygen with the negative electrolyte (and subsequent
are obtained from Eq. (3) which is based on Faraday's law of elec- oxidation of V2þ to V3þ).
trolysis (I is the current, nVxþ is the amount of Vxþ and F is the 3. Oxygen permeation from the positive to the negative electrode
Faraday constant) under the assumption that no side reactions take (process 3 in Fig. 1b).
place and, therefore, concentration changes of cV2þ and cV3þ are
exclusively due to their mutual interconversion by electron transfer The negative electrolyte was permanently purged with N2 to
at the negative electrode. The initial concentrations cVxþ ðt ¼ 0Þ prevent contact of negative electrolyte with ambient air and
before charging and discharging were determined with UV/Vis therefore the explanation 2 can be neglected. Although relatively
spectroscopy. The ‘‘þ’’ and ‘‘’’ are used in appropriate manner (e.g. low potentials are present at the negative electrode during
during charging the ‘‘þ’’ is used for cV2þ as this concentration charging and HER could take place as a side reaction [50,51], the
increases). overpotential for HER is very high on graphite electrodes
(hHER; graphite ¼ 0:977 V at 100 mA cm2 and 25  C [52]) and no
DnVxþ ðtÞ I$t bubble formation on the negative electrode was observed during
cVxþ ; calc ðtÞ ¼ cVxþ ðt ¼ 0Þ± ¼ cVxþ ðt ¼ 0Þ± (3)
VðÞ F$VðÞ charging. Therefore, we suppose that HER is not the dominant re-
action. The permeation of oxygen through the membrane and
Taking into account the possible side reactions shown in Fig. 1b, subsequent oxidation of V2þ to V3þ was also suspected by Hosseiny
the time evolution of cV2þ ; meas and cV3þ ; meas is shown schematically et al. [9] to take place in a VARFB and is assumed to contribute most
in Fig. 5a during charging in comparison to the expected evolution significantly to the deviation of cV2þ ; meas and cV3þ ; meas from cV2þ ; calc
in the absence of any side reaction. If water is transferred from the and cV3þ ; calc .
positive electrode to the negative electrode and/or both vanadium All possible explanations 1e3 would have the consequence that
species undergo crossover from the negative to the positive elec- cV2þ ; meas would be lower than cV2þ ; calc and cV3þ ; meas would be
trode, cV2þ ; meas and cV3þ ; meas should be lower than cV2þ ; calc and higher than cV3þ ; calc (Fig. 5c). Taking into account that in parallel to
cV3þ ; calc (Fig. 5b). In case of parallel hydrogen evolution and/or oxygen permeation (or HER) the crossover of vanadium species
oxygen permeation with subsequent chemical oxidation of V2þ to from the negative to the positive electrode takes place which
V3þ, the SOC of the battery is always lower than theoretically ex- lowers cV2þ ; meas and cV3þ ; meas , measured cV2þ ; meas should fall below
pected from the charge measured in the external circuit. The the values predicted by Eq. (3). The measured cV3þ ; meas value can be
occurrence of these side reactions is indicated if cV2þ ; meas is lower higher or lower than cV3þ ; calc , depending on whether oxygen
than cV2þ ; calc and cV3þ ; meas is higher than cV3þ ; calc (Fig. 5c). How- permeation or vanadium crossover is more dominant. The obser-
ever, during real operation of the VARFB a combination of the in- vation concerning cV2þ ; meas (decreasing) and cV3þ ; meas (more or less
dividual loss mechanisms is expected. Therefore, real values for constant) is reflected by Fig. 6a. One possible interpretation for this
cV2þ ; meas and cV3þ ; meas will most probably result from a super- is that V2þ and V3þ crossover take place as well as O2 permeation
position of the effects shown in Fig. 5aec. and subsequent oxidation of V2þ to V3þ. As cV3þ ; meas remains nearly
From the values in Table 2 it is evident that water is transferred constant, the amount of ‘‘lost’’ V3þ due to crossover needs to be
through the membrane from the positive electrode to the negative approximately equal to the amount of V3þ ‘‘gained’’ due to oxida-
electrode during charging and vice versa during discharging. The tion of V2þ by permeated O2. Nevertheless, the observed deviation
transfer of water is also known from VRFB [13,48,49] and is caused of the cV2þ ; meas at the end of the charging process is likely to be
by osmotic pressure between the electrolytes and by water transfer caused by simultaneous hydrogen evolution since at high SOC the
concerted with proton transport due to charge balancing. amount of V2þ depletes and therefore side reactions (such as HER)
The potential during charging and discharging and cV2þ ; meas and are more pronounced. The deviation of cV2þ at the end of the
cV3þ ; meas in the negative electrolyte at j ¼ 20 mA cm2 are shown in charging step is similar to the observation in the VARFB charging
Fig. 6. A significant deviation between the measured and calculated with j ¼ 15 mA cm2 (Fig. 4).
concentrations during charging can be seen at the end of the In the discharging process (Fig. 6b) a more pronounced devia-
charging process for cV2þ (Fig. 6a). As also discussed in section 3.2 tion between measured and calculated concentrations is observed.
and indicated by the amount of vanadium determined in the pos- cV2þ ; meas is much lower than calculated, whereas cV3þ ; meas is much
itive electrolyte by ICP-MS nVxþ ; ðþÞ; ICPMS in Table 3 diffusion of higher than calculated. This observation cannot be explained by
vanadium species occurs from the negative to positive electrode vanadium crossover only (Fig. 5b). HER is not likely to dominate
during the whole cycle. Thus, cV2þ ; meas and cV3þ ; meas lower than galvanic discharging. Thus, the observation concerning cV2þ ; meas
cV2þ ; calc and cV3þ ; calc are expected as schematically depicted in and cV3þ ; meas during discharging indicates indirectly the occur-
Fig. 5b. However, in Fig. 6a only cV2þ ; meas is lower than cV2þ ; calc , rence of O2 permeation. In the discharging process significant loss
while cV3þ ; meas is close to cV3þ ; calc . This observation corresponds to mechanisms like oxygen permeation, which lower cV2þ ; meas and
a combination of the two scenarios shown in Fig. 5b and c. Three raise cV3þ ; meas in comparison to cV2þ ; calc and cV3þ ; calc , occur in
possible reactions that cause the deviation described in Fig. 5c are: parallel to vanadium crossover (lowering both cV2þ ; meas and
cV3þ ; meas ). In contrast to the charging process, the GDE is supplied
Table 2 with compressed air for the oxygen reduction reaction. This air is in
Volume V(), total vanadium concentration measured by UV/Vis spectroscopy excess (to avoid VARFB performance limitation due to insufficient
ðcV2þ þ cV3þ ÞðÞ; UV=Vis and total amount of vanadium ðnV2þ þ nV3þ ÞðÞ; UV=Vis of the O2 supply and to facilitate the removal of product water out of the
negative electrolyte before and after charging and discharging, respectively, during GDE). The increased partial pressure of oxygen facilitates perme-
the VARFB cycle at j ¼ 20 mA cm2.
ation to the negative half cell resulting in more pronounced devi-
V() ðcV2þ þ cV3þ ÞðÞ; UV=Vis ðnV2þ þ nV3þ ÞðÞ; UV=Vis ation of vanadium concentrations from the calculated values during
[mL] [mol L1] [mmol] the discharge process (Fig. 6b).
Before charging 12.0 0.96 12 Overall, significant differences between charging and dis-
After charging 13.6 0.85 12 charging are observed in the VARFB cycle at j ¼ 20 mA cm2. First,
After discharging 13.0 0.83 11
the deviations between cV2þ ; meas and cV3þ ; meas and cV2þ ; calc and
698 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701

Fig. 6. In-situ UV/Vis monitoring of vanadium concentrations in the negative electrolyte during VARFB operation at j ¼ 20 mA cm2 in comparison to calculated concentrations
based on Faraday's law. a) Charging; b) Discharging.

Table 3 to membrane-containing systems working under real conditions

Volume V(þ), total vanadium concentration cVxþ ; ðþÞ; ICPMS and total amount of va- [54]. Additionally, in VARFB operation an electric field is present
nadium nVxþ ; ðþÞ; ICPMS of the positive electrolyte before and after charging and
which could cause EOC flux of water and dissolved oxygen trans-
discharging, respectively, during the VARFB cycle at j ¼ 20 mA cm2.
port through the membrane [55].
V(þ) cVxþ ; ðþÞ; ICPMS nVxþ ; ðþÞ; ICPMS The in situ monitoring of the negative electrolyte with UV/Vis
[mL] [mol L1] [mmol]
spectroscopy during VARFB operation shows indication for the
Before charging 50.0 e e overlapping effects of all loss mechanisms (V2þ/V3þ crossover, ox-
After charging 49.0 0.0079 0.395
ygen permeation, HER) that prevented a quantitative disentangle-
After discharging 47.0 0.0165 0.776
ment of the individual processes. Therefore, the total vanadium
crossover was quantified with ICP-MS analysis of the positive
electrolyte after charging and discharging.
cV3þ ; calc are much more pronounced during discharging. In the
charging process cV2þ ; meas deviates from cV2þ ; calc only towards the
end of the charging and start to show asymptotic behavior at SOC 3.2. Vanadium crossover during VARFB operation determined with
above 80%. cV3þ ; meas is always close to cV3þ ; calc . Supported also by ICP-MS
the values for the total amount of vanadium measured in the
anolyte via UV/Vis spectroscopy ðnV2þ þ nV3þ ÞðÞ; UV=Vis before and The total vanadium crossover after charging and after dis-
after charging (Table 2), it is concluded that the vanadium crossover charging the VARFB at j ¼ 20 mA cm2 was determined by
during charging is less pronounced than during discharging. This measuring the vanadium concentration in the positive electrolyte
finding can be explained by the electric-field dependent transfer using ICP-MS and the volumes of the positive electrolyte (Table 3).
processes through the membrane: migration and EOC. While The volume of the positive electrolyte is decreasing during the cycle
diffusion takes place from negative to the positive electrode due to due to water transfer from the positive to the negative electrode
concentration gradient during charging and discharging, transfer of (charging), water consumption due to oxygen evolution reaction
vanadium ions from negative electrode to the positive electrode (charging) and evaporation due to the stream of compressed air
due to migration and EOC can occur only during discharging. (discharging).
Therefore, the total vanadium crossover is expected to be higher The differences in nVxþ ; ðþÞ; ICPMS are similar after charging and
during discharging than during charging. Indeed, this is observed after discharging (Table 3). This is in contrast to the findings from
here and was also reported for VRFB [19]. UV/Vis measurements. However, the positive electrolyte was not
The crossover of oxygen through Nafion® membranes and pumped during discharging. Consequently, V2þ and V3þ accumu-
possible transport mechanisms have been studied intensively, a lated in the small volume inside the cell and reduced the concen-
comparison of diffusion coefficients for oxygen through Nafion tration difference across the membrane. This leads to an overall
determined with different methods is given in Ref. [53]. However, it decreased diffusional flux of V2þ and V3þ from the negative to the
does not seem to be appropriate to use these values for oxygen positive electrolyte during discharging.
permeation calculations due to several reasons. First, the context of The vanadium concentration during the whole cycle decreases
the studies are fuel cells in which gaseous reactants are in direct in the negative electrolyte and increases in the positive electrolyte.
contact to the membrane, whereas the membrane is in contact with The amount of vanadium found in the positive electrolyte by ICP-
liquid water/acid in VARFB. Second, it was stated that all conven- MS after the complete cycle (nðþÞ; ICPMS ¼ 0:776 mmol; Table 3)
tional techniques to determine the oxygen permeation are ex situ is comparable to the loss of ðnV2þ þ nV3þ ÞðÞ; UV=Vis in the negative
techniques and therefore their results have limited transferability electrolyte DðnV2þ þ nV3þ ÞðÞ; UV=Vis ¼ 1 mmol; Table 2). The
 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701 699

difference is explained with the accuracy limits of the UV/Vis (approx. every hour) by ICP-MS analysis.
measurements and of the volume determination. The calculation of the diffusion coefficients DVxþ follows a deri-
Albeit not determined in this study, the ‘‘backward’’ vanadium vation of Fick's first law in a one-dimensional form (Eq. (5)) re-
crossover is assumed to be negligible in comparison to the cross- ported elsewhere [13,14]:
over from the negative electrolyte to the positive electrolyte. This
was estimated to be adequate as the maximal concentration of 1 dnðtÞ dcðtÞ
¼ D (5)
vanadium species in the positive electrolyte is 0.0165 mol L1 A dt dx
which is less than 2% of the total vanadium concentration in the With Ve and Vd being the electrolyte volume of the excess and
negative electrolyte (0.96 mol L1) even though the concentrations the deficiency side, transitioning from gradient dc/dx to difference
of vanadium species in the oxidation state 4 þ and 5 þ is zero in the Dc/Dx, assuming that the concentration of the excess side ce re-
negative electrolyte due to the potentials occurring on the negative mains nearly constant over time, cd being the concentration in the
electrode. Despite of this assumption, the differentiation between deficiency side and Dx being the thickness of swollen membrane,
the simultaneously occurring ‘‘backward’’ and ‘‘forward’’ crossover Eq. (5) gives Eq. (6) and rearranging leads to the differential
is difficult as each diffusing species is immediately converted to a equation Eq. (7):
different oxidation state when reaching the other electrode.
The total amount of vanadium that underwent crossover Vd dcd ðtÞ ce  cd ðtÞ
(0.776 mmol) represents 6% of the initial vanadium amount in the ¼ DA (6)
dt x
negative electrolyte (12 mmol). hC losses due to vanadium cross-
over are related to V2þ crossover only. Even if the amount of
dcd ðtÞ DA
0.776 mmol of vanadium was V2þ only, it would represent a charge ¼ dt (7)
ce  cd ðtÞ xVd
of 74.9 C (via Faraday's law). However,
DQ ¼ QchargeQdischarge ¼ 370.6 C, thus the vanadium crossover Assuming that D, A, x and Vd are independent of t, integration of
cannot be the only cause for hC losses and can account at most for Eq. (7) leads to:
20% of DQ. It has to be concluded that the other loss mechanisms  
such as oxygen permeation as described above play an important ce DA
ln ¼ t (8)
role. The ICP-MS measurements of the positive electrolyte give ce  cd ðtÞ xVd
information only about the total amount of vanadium that was
By plotting ln(ce/(cecd(t))) vs. t, the slope of the linear fit equals
transferred through the membrane, but the data do not allow dis-
DA/xVd which was used to obtain the diffusion coefficient of V2þ
tinguishing between V2þ and V3þ. In general, V2þ and V3þ cannot
and V3þ through Nafion® 117 membranes (Fig. 7) as
be distinguished in the positive electrolyte at the operation po-
DV2þ ; N117 ¼ 9:05$106 cm2 min1 (¼ 1.51$107 cm2 s1) and
tentials of the positive electrode because both species are imme-
DV3þ ; N117 ¼ 4:35$106 cm2 min1 (¼ 7.25$108 cm2 s1). There are
diately oxidized to the oxidation states þ4 or þ5 when reaching the
no previously reported values for DV2þ ; N117. The value for V3þ fits
positive electrode. Since the crossover of V2þ can only be given as
well to reported values (DV3þ ; N117 ¼ 3:56$106 cm2 min1 [15] or
an upper boundary (0.776 mmol), the fraction of hC losses caused
DV3þ ; N117 ¼ 3:8$106 cm2 min1 [16]). The value for DV2þ ; N117 is
by vanadium crossover can only be given by an upper limit of 20%.
approximately double the value of DV3þ ; N117 . A similar ratio for
For the differentiation of vanadium crossover due to diffusion in
DV2þ /DV3þ was reported for a Nafion® 115 membrane
contrast to electric-field induced migration and EOC, diffusion co-
(DV2þ ; N115 ¼ 2:6$106 cm2 min1 and DV3þ ; N115 ¼ 1:1$106
efficients of V2þ and V3þ through Nafion® 117 were determined.
cm2 min1 [14]; DV2þ ; N115 ¼ 5:261$106 cm2 min1 and
DV3þ ; N115 ¼ 1:933$106 cm2 min1 [13]). This membrane consists
3.3. V2þ and V3þ diffusion coefficients through Nafion® 117 of the same polymer as Nafion® 117 but has a different thickness.
However, for a true comparability, the swelling behavior of both
Vanadium cation crossover from negative to positive electrolyte membranes needs to be the same. Slade et al. [57] reported an
occurs during VARFB operation driven by several processes. The unequal thickness increase of Nafion® 117 and Nafion® 115 in
total flux Ji of vanadium species i through a membrane in an electric relation to their nominal thickness (þ17% for Nafion® 117 and þ 27%
field (as it is the case in VARFB operation) in one direction (x) is
described by the Nernst-Planck-Equation (Eq. (4)) [56]. The first
term on the right side describes the diffusion, the second one the
migration and the third describes the contribution of EOC. A is the
diffusion area, ni is the amount of substance and ci the concentra-
tion of species i, respectively, t is the time, zi is the charge of species
i, F is Faraday constant, R is the gas constant, T is the temperature, Di
is the diffusion coefficient of species i, f is the potential and neo is
the solute velocity due to EOC:

dni dc ðxÞ zi F df
Ji ¼ A ¼ Di i  D c ðxÞ þ ci ðxÞneo ðxÞ (4)
dt dx RT i i dx
The diffusion coefficients of V2þ and V3þ through a Nafion® 117
membrane (denoted as N117) were needed to be able to quantify
the amount of vanadium crossover caused by diffusion. There was
no value reported yet for the diffusion coefficient of V2þ through a
Nafion® 117 membrane.
The diffusion coefficients DV2þ and DV3þ were determined using
the H-cell setup as described in sec 2.4. The vanadium concentra- Fig. 7. Plot of ln(ce/(ce e cd)) vs. t and linear fits for determination of diffusion co-
tion in the deficiency side was probed in regular time intervals efficients of V2þ and V3þ through a Nafion® 117 membrane.
700 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701

for Nafion® 115) due to a similar protonation routine as we used. 3.4. Combined crossover analysis
Therefore, it seems to be advisable to determine the diffusion co-
efficients for each membrane although data might be available for A Sankey diagram (Fig. 8) summarizes the findings concerning
the same polymer material processed to a membrane with different the coulombic losses in the VARFB in one cycle at j ¼ 20 mA cm2.
thickness. Only 65% of Qcharge could be recovered during discharging (Qdi-
Logette et al. proposed that the diffusion coefficient of cations scharge). In Fig. 8 the vanadium crossover determined with ICP-MS is
through a cation exchange membrane is affected by the affinity of given by its maximum share of 20% of the coulombic losses which is
the ion for functional groups of the membrane and the mobility 7% of Qcharge. Based on the calculations in section 3.3, the vanadium
within the membrane [58]. Higher valency of the cation results in crossover is subdivided into diffusion (4% of Qcharge) and the
higher affinity for the membrane and in lower mobility due to remaining quantity that relates to migration and/or EOC (3% of
stronger electrostatic interaction of the cation with the negative Qcharge). The remaining part is related to HER and/or oxygen
groups of the membrane. Depending on which of these opposed permeation and is depicted as ‘‘other losses (e.g. oxygen perme-
effects is more dominant, a higher valent cation can have a higher ation)’’ as its minimal fraction. It accounts for 28% of Qcharge.
or lower diffusion coefficient than a lower valent cation. In the case
of V2þ and V3þ with DV3þ ; N117 z0:5$DV2þ ; N117 , V3þ is either inter-
acting stronger with the sulfonate groups than V2þ so that it be- 4. Conclusion
comes less mobile within the membrane or the interactions are so
poor that V3þ only reaches a low concentration within the mem- As the vanadium/air redox flow battery could theoretically
brane. In aqueous solution, V2þ and V3þ both exist as hexaaqua- double the energy density versus an all-vanadium redox flow
ions [22,24,59,60]. It was shown that in sulfuric acid media SO2 battery, it is a promising technology especially for applications
4
can substitute a water molecule in the inner hydration shell of V3þ where energy density does matter (e.g. stationary electricity stor-
yielding [V(SO4)(H2O)5]þ [21,59], whereas this is not the case for age application with limited space). However, hC need to be
V2þ which exists as [V(H2O)6]2þ [21,24]. The V3þ complex with the improved for practical applications.
sulfate ion carries a lower positive charge than the V2þ complex. In this paper the causes for hC losses of a vanadium/air redox
Another effect that was reported is that V3þ carries a more tightly flow battery were investigated. We integrated successfully a UV/Vis
bound hydration shell which screens effectively the electrostatic spectrometer in the hydraulic circuit of the negative electrolyte for
charge of the V3þ in comparison to V2þ [26]. Both effects would the in situ monitoring of cV2þ and cV3þ during VARFB operation at
decrease the interaction of V3þ with the sulfonate groups in com- j ¼ 15 mA cm2 and at j ¼ 20 mA cm2. This analysis revealed that
parison to V2þ. Therefore, it could be deduced that V3þ probably has hC losses are caused by side reactions such as oxygen permeation
a low affinity to the membrane (weak interaction with membrane) through the membrane and subsequent oxidation of V2þ to V3þ. In
and therefore reaches only a lower concentration within the general, the application of UV/Vis spectroscopy for analysing
membrane. The lower affinity of V3þ to the sulfonate groups of a negative electrolyte concentrations is a powerful and convenient
Nafion® membrane in comparison to V2þ was also reported by Cui technique to monitor the battery operation in situ if appropriate
et al. by means of molecular dynamics [26]. The different affinity of cuvettes are used. The method is not limited to the VARFB.
V2þ and V3þ to the membrane could also be investigated experi- Furthermore, the vanadium crossover during VARFB operation
mentally by determining the concentration of vanadium cations in at j ¼ 20 mA cm2 was quantified by analysing the positive elec-
a membrane equilibrated in V2þ and V3þ solutions, respectively. trolyte by ICP-MS. The vanadium crossover causes at most 20% of hC
However, a detailed study of the interaction of vanadium species losses. Although the vanadium crossover is not the main reason for
with the sulfonate groups in the channels of the Nafion® membrane low hC, it is strongly undesired as it reduces the cycle life of the
is needed for a correlation of diffusivity of the vanadium species VARFB. The amount of vanadium in the negative electrolyte is the
through the membrane to the interaction mechanisms of the capacity-determining part of the VARFB, therefore the continuous
diffusing species inside of the membrane. However, this was crossover of vanadium species from the negative to the positive
beyond the scope of this work. electrode leads to an ongoing capacity decay of the VARFB. Hence a
For the estimation of the diffusional flux during VARFB opera- strongly reduced vanadium crossover is essential for an increased
tion at j ¼ 20 mA cm2 it was assumed that the initial vanadium cycle life of the VARFB which is necessary for practical applications.
concentration in the negative electrolyte (0.96 mol L1) remains
constant and an (arithmetic) average diffusion coefficient for V2þ
and V3þ was used (DV2=3þ ¼ 6:7$106 cm2 min1). With these as-
sumptions 0.451 mmol of V2þ/3þ could have crossed the membrane
during one VARFB cycle at j ¼ 20 mA cm2 by diffusion only. The
total amount of vanadium that crossed over during the VARFB cycle
at j ¼ 20 mA cm2 determined with ICP-MS was 0.776 mmol
(Table 3), therefore the amount of vanadium crossover through the
membrane by migration and/or EOC would be 0.325 mmol. That
means that approximately 58% of the vanadium crossover is
diffusion-related and therefore 42% result from migrational and/or
EOC. The discussion above assumes independent permeation of
oxygen and vanadium. However, this is an approximation as any
oxygen can oxidize diffusing V2þ to V3þ (or even VO2þ (‘‘V4þ’’))
inside the membrane. This not only would influence the overall
transport properties of vanadium species but VO2þ can lead to
fouling of the membrane [22]. A similar fouling process could also
be imagined for V2þ and/or V3þ species. More theoretical and
experimental investigations are necessary to fully unravel these Fig. 8. Sankey diagram quantifying loss mechanisms during the VARFB operation cycle
coupled processes. at j ¼ 20 mA cm2. See text for explanation.
 J. grosse Austing et al. / Journal of Power Sources 306 (2016) 692e701 701

The diffusion coefficients of V2þ and V3þ through Nafion® 117 were 218 (2012) 15e20.
[18] Q. Luo, L. Li, W. Wang, Z. Nie, X. Wei, B. Li, B. Chen, Z. Yang, V. Sprenkle,
determined to differentiate vanadium crossover caused by diffu-
ChemSusChem 6 (2013) 268e274.
sion and migration/electroosmotic convection (EOC). The diffusion [19] X.-G. Yang, Q. Ye, P. Cheng, T.S. Zhao, Appl. Energy 145 (2015) 306e319.
coefficients can also be valuable for detailed simulation and [20] R.M. Darling, a. Z. Weber, M.C. Tucker, M.L. Perry, J. Electrochem. Soc. 163
modeling of all-vanadium redox flow batteries. (2016) A5014eA5022.
[21] F. Sepehr, S.J. Paddison, J. Phys. Chem. A (2015), 150518095410002.
However, as the vanadium species that underwent crossover are [22] M. Vijayakumar, M.S. Bhuvaneswari, P. Nachimuthu, B. Schwenzer, S. Kim,
immediately oxidized at the positive electrode to vanadium in the Z. Yang, J. Liu, G.L. Graff, S. Thevuthasan, J. Hu, J. Memb. Sci. 366 (2011)
oxidation state 4 þ and 5 þ and cannot be distinguished after the 325e334.
[23] S. Quezado, J.C.T. Kwak, M. Falk, Can. J. Chem. 62 (1984) 958e966.
crossover, the share of vanadium crossover on hC losses can only be [24] M. Benmelouka, S. Messaoudi, E. Furet, R. Gautier, E. Le Fur, J.Y. Pivan, J. Phys.
determined as upper limit (20%). Likewise, a more detailed un- Chem. A 107 (2003) 4122e4129.
derstanding of the transport processes in the membrane requires [25] K.M. Cable, K. a. Mauritz, R.B. Moore, J. Polym. Sci. Part B Polym. Phys. 33
(1995) 1065e1072.
an examination of the interactions between the diffusing species [26] S. Cui, S.J. Paddison, J. Phys. Chem. C (2015), 150521065237001.
and the membrane on a molecular level unraveling the influence of [27] X. Teng, Y. Zhao, J. Xi, Z. Wu, X. Qiu, L. Chen, J. Power Sources 189 (2009)
the mutual interaction between the transferred species (V2þ, V3þ 1240e1246.
[28] J. Qiu, L. Zhao, M. Zhai, J. Ni, H. Zhou, J. Peng, J. Li, G. Wei, J. Power Sources 177
and O2). (2008) 617e623.
For future studies in an optimized system it would be interesting [29] J. Xi, Z. Wu, X. Teng, Y. Zhao, L. Chen, X. Qiu, J. Mater. Chem. 18 (2008) 1232.
to investigate the influences of electrolyte flow rates and current [30] K.W. Knehr, E. Agar, C.R. Dennison, a. R. Kalidindi, E.C. Kumbur, J. Electrochem.
densities on the overall hC and the fractions of the individual loss Soc. 159 (2012) A1446eA1459.
[31] D. You, H. Zhang, J. Chen, Electrochim. Acta 54 (2009) 6827e6836.
mechanisms on the global losses. [32] M. Skyllas-Kazacos, L. Goh, J. Memb. Sci. 399e400 (2012) 43e48.
For a more efficient VARFB operation, an optimized membrane [33] R. Badrinarayanan, J. Zhao, K. Tseng, M. Skyllas-Kazacos, J. Power Sources 270
is needed. Our study reveals that oxygen permeation is severely (2014) 576e586.
[34] N.H. Choi, S.-k. Kwon, H. Kim, J. Electrochem. Soc. 160 (2013) A973eA979.
reducing hC. Therefore, a membrane with low oxygen permeation is [35] M. Skyllas-Kazacos, M. Kazacos, J. Power Sources 196 (2011) 8822e8827.
recommended for the VARFB. On the other hand, a membrane with [36] Z. Tang, D. Aaron, A.B. Papandrew, T.A. Zawodzinski, ECS Trans. 41 (2012) 1e9.
low vanadium crossover is needed to enhance hC and cycle life of [37] L. Liu, J. Xi, Z. Wu, W. Zhang, H. Zhou, W. Li, X. Qiu, J. Appl. Electrochem. 42
(2012) 1025e1031.
the VARFB. A membrane combining both requirements seems to be [38] H.J. Lee, N.H. Choi, H. Kim, J. Electrochem. Soc. 161 (2014) A1291eA1296.
most appropriate for an efficient and durable VARFB. Besides other [39] D.N. Buckley, X. Gao, R.P. Lynch, N. Quill, M.J. Leahy, J. Electrochem. Soc. 161
possible approaches, existing membranes could be modified to (2014) A524eA534.
[40] D.C. Sing, J.P. Meyers, ECS Trans. (2013).
reduce both vanadium crossover and oxygen permeation. A [41] R.P. Brooker, C.J. Bell, L.J. Bonville, H.R. Kunz, J.M. Fenton, J. Electrochem. Soc.
promising routine is the layer-by-layer deposition of poly- 162 (2015) A608eA613.
electrolytes thin films which could reduce vanadium crossover [42] N. Quill, C. Petchsingh, R.P. Lynch, X. Gao, D. Oboroceanu, D.N. Eidhin,
M.O. Mahony, C. Lenihan, D.N. Buckley, ECS Trans. 64 (2015) 23e39.
[61e63] and oxygen permeation [64e68] while maintaining proton [43] Q.H. Liu, G.M. Grim, a. B. Papandrew, A. Turhan, T. a. Zawodzinski,
conductivity of the membrane. M.M. Mench, J. Electrochem. Soc. 159 (2012) A1246eA1252.
[44] J. Sun, X. Li, X. Xi, Q. Lai, T. Liu, H. Zhang, J. Power Sources 271 (2014) 1e7.
[45] T. Sukkar, M. Skyllas-Kazacos, J. Appl. Electrochem 34 (2004) 137e145.
Acknowledgments
[46] A.A. Shah, M. Watt-Smith, F. Walsh, Electrochim. Acta 53 (2008) 8087e8100.
[47] M.J. Watt-Smith, P. Ridley, R. G. a. Wills, a. a. Shah, F.C. Walsh, J. Chem.
The authors would like to thank Timo di Nardo for assisting in Technol. Biotechnol. 88 (2013) 126e138.
the experimental work. Likewise the contributions of Dana [48] T. Sukkar, M. Skyllas-Kazacos, J. Memb. Sci. 222 (2003) 235e247.
[49] T. Mohammadi, J. Memb. Sci. 133 (1997) 151e159.
Schonvogel (ICP-MS measurements) are appreciated. The work was [50] C.-N. Sun, F.M. Delnick, L. Baggetto, G.M. Veith, T. a. Zawodzinski, J. Power
supported by the Reiner Lemoine-Stiftung, Berlin, Germany by Sources 248 (2014) 560e564.
supplying a PhD scholarship to J. grosse Austing. [51] a.a. Shah, H. Al-Fetlawi, F. Walsh, Electrochim. Acta 55 (2010) 1125e1139.
[52] E. W. Washburn, N. R. C. (U.S.), C. West, International Council of Scientific
Unions National Academy of Sciences (U.S.), International Critical Tables of
References Numerical Data, Physics, Chemistry and Technology, number Vol. 6 in Inter-
national Critical Tables of Numerical Data, Physics, Chemistry and Technology,
[1] H. Chen, T.N. Cong, W. Yang, C. Tan, Y. Li, Y. Ding, Prog. Nat. Sci. 19 (2009) National Research Council, 1929.
291e312. [53] V. a. Sethuraman, S. Khan, J.S. Jur, A.T. Haug, J.W. Weidner, Electrochim. Acta
[2] M. Skyllas-Kazacos, M.H. Chakrabarti, S. a. Hajimolana, F.S. Mjalli, M. Saleem, 54 (2009) 6850e6860.
J. Electrochem. Soc. 158 (2011) R55. [54] P. Gode, G. Lindbergh, G. Sundholm, J. Electroanal. Chem. 518 (2002)
[3] P. Leung, X. Li, C. Ponce de Leo  n, L. Berlouis, C.T.J. Low, F.C. Walsh, RSC Adv. 2 115e122.
(2012) 10125. [55] P. Millet, R. Ngameni, S. a. Grigoriev, N. Mbemba, F. Brisset, a. Ranjbari,
[4] M. Skyllas-Kazacos, J. Power Sources 124 (2003) 299e302. C. Etievant, Int. J. Hydrogen Energy 35 (2010) 5043e5052.
[5] L. Li, S. Kim, W. Wang, M. Vijayakumar, Z. Nie, B. Chen, J. Zhang, G. Xia, J. Hu, [56] B. Bath, H. White, E. Scott, Anal. Chem. 72 (2000) 433e442.
G. Graff, J. Liu, Z. Yang, Adv. Energy Mater. 1 (2011) 394e400. [57] S. Slade, S. Campbell, T. Ralph, F. Walsh (2002) 1556e1564.
[6] Y. Lu, J.B. Goodenough, Y. Kim, J. Am. Chem. Soc. 133 (2011) 5756e5759. [58] S. Logette, C. Eysseric, G. Pourcelly, A. Lindheimer, C. Gavach, J. Memb. Sci. 144
[7] Q. Huang, Q. Wang, Chempluschem 80 (2015) 312e322. (1998) 259e274.
[8] H. Kaneko, A. Negishi, K. Nozaki, K. Sato, M. Nakajima, (1992) EU Patent [59] M. Vijayakumar, L. Li, Z. Nie, Z. Yang, J. Hu, Phys. Chem. Chem. Phys. 14 (2012)
0517217A1. 10233.
[9] S. Hosseiny, M. Saakes, M. Wessling, Electrochem. Commun. 13 (2011) [60] J. Krakowiak, D. Lundberg, I. Persson, J. Inorg. Chem. (2012) 9598e9609.
751e754. [61] J. Xi, Z. Wu, X. Teng, Y. Zhao, L. Chen, X. Qiu, J. Mater. Chem. 18 (2008) 1232.
[10] J. Noack, C. Cremers, D. Bayer, J. Tübke, K. Pinkwart, J. Power Sources 253 [62] W. Xu, X. Li, J. Cao, H. Zhang, H. Zhang, Sci. Rep. 4 (2014) 4016.
(2014) 397e403. [63] S. Lu, C. Wu, D. Liang, Q. Tan, Y. Xiang, RSC Adv. 4 (2014) 24831.
[11] C. Menictas, M. Skyllas-Kazacos, J. Appl. Electrochem. 41 (2011) 1223e1232. [64] Y.H. Yang, M. Haile, Y.T. Park, F. a. Malek, J.C. Grunlan, Macromolecules 44
[12] J. grosse Austing, C. Nunes Kirchner, E.-M. Hammer, L. Komsiyska, (2011) 1450e1459.
G. Wittstock, J. Power Sources 273 (2015) 1163e1170. [65] Y.-H. Yang, L. Bolling, M. Haile, J.C. Grunlan, RSC Adv. (2012) 12355e12363.
[13] C. Sun, J. Chen, H. Zhang, X. Han, Q. Luo, J. Power Sources 195 (2010) 890e897. [66] D. a. Hagen, C. Box, S. Greenlee, F. Xiang, O. Regev, J.C. Grunlan, RSC Adv. 4
[14] P. Leung, Q. Xu, T. Zhao, L. Zeng, C. Zhang, Electrochim. Acta 105 (2013) (2014) 18354.
584e592. [67] D. a. Hagen, B. Foster, B. Stevens, J.C. Grunlan, ACS Macro Lett. 3 (2014)
[15] X. Luo, Z. Lu, J. Xi, Z. Wu, W. Zhu, J. Phys. Chem. B 109 (2005) 20310e20314. 663e666.
[16] J. Xi, Z. Wu, X. Qiu, L. Chen, J. Power Sources 166 (2007) 531e536. [68] B.E. Stevens, P.K. Odenborg, M.A. Priolo, J.C. Grunlan, J. Polym. Sci. Part B
[17] Q. Luo, L. Li, Z. Nie, W. Wang, X. Wei, B. Li, B. Chen, Z. Yang, J. Power Sources Polym. Phys. 52 (2014) 1153e1156.
Publication III

73
Layer-by-layer modification of Nafion membranes for
increased life-time and efficiency of vanadium/air
redox flow batteries
Authors: Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska and
Gunther Wittstock
Journal of Membrane Science, 510, 2016, 259-269, DOI: 10.1016/j.memsci.2016.03.005

V2+ O2 + 4 H++ 4 e-
⇄ H+ ⇄
_ V3+ + e- H2O
⃝ ⃝
+
Vx+
O2(g/dissolved)

2.0

1.5
E [V]

1.0

0.5 N117
N117-(PEI/Nafion)10

0.0
0 50 100 150 200 250 300
capacity [mAh]

75
 Journal of Membrane Science 510 (2016) 259–269

Contents lists available at ScienceDirect

Journal of Membrane Science

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

Layer-by-layer modification of Nafion membranes for increased life-

time and efficiency of vanadium/air redox flow batteries
Jan grosse Austing a, Carolina Nunes Kirchner a, Lidiya Komsiyska a, Gunther Wittstock b,n
a
NEXT ENERGY EWE Research Centre for Energy Technology, 26129 Oldenburg, Germany
b
Carl von Ossietzky University of Oldenburg, Faculty of Mathematics and Natural Sciences, Center of Interface Science, Institute of Chemistry, 26111 Old-
enburg, Germany

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

Article history: Vanadium/air redox flow batteries (VARFB) promise higher energy densities compared to all-vanadium
Received 4 January 2016 redox flow batteries (VRFB). However, VARFB suffer from crossover processes through the membrane, i.e.
Received in revised form vanadium crossover and oxygen permeation. The vanadium crossover causes ongoing capacity losses and
26 February 2016
therefore reduces the lifetime of the battery. Additionally, the coulombic efficiency is reduced due to
Accepted 1 March 2016
vanadium crossover and oxygen permeation. In this contribution we propose a straightforward routine
Available online 8 March 2016
for Nafion 117 (N117) membrane modification to reduce both vanadium crossover and oxygen per-
Keywords: meation. Layer-by-layer (LbL) deposited films of polyethylenimine (PEI) and Nafion ionomer are build up
Vanadium–air redox flow battery on the membrane by dipping the membrane alternatingly in solutions of the polyelectrolytes. The
Crossover redox flow battery
modification of the membranes is characterized with infrared (IR) spectroscopy, scanning electron mi-
Vanadium oxygen fuel cell
croscopy (SEM), thermogravimetric analysis (TGA) and film thickness measurements. The properties of
Layer-by-layer deposition
the modified membranes are investigated by determining the proton conductivity, vanadium crossover
and oxygen permeation. By the application of a LbL film of PEI/Nafion obtained after 10 LbL deposition
repetitions, the selectivity ( σ H+/P V2 + ) of the membrane towards protons is increased by factor 21. Using
this membrane in a VARFB reveals a strongly reduced vanadium crossover (approx.
70%) during a cycle
as determined with inductively coupled plasma mass spectroscopy (ICP-MS) analysis of the positive
electrolyte. The coulombic efficiency increases from 81% to 93% and the energy efficiency from 41.5% to
45.2%. TGA and IR measurements of the membrane after VARFB operation indicated a vanadium ion
uptake into the membrane and the stability of the LbL film under conditions of VARFB operation.
& 2016 Elsevier B.V. All rights reserved.

1. Introduction reduced to approx. 1% of the initial volume [6]. The O2 for the
discharging process is provided by the ambient air. Recently, we
With increasing share of renewable electricity production, reported the successful operation of a unitized bidirectional VARFB
stationary energy storage is needed to harmonize the fluctuating using a two-layered positive electrode [6]. However, we observed
electricity generation and the demand. Redox flow batteries are a rapid capacity fading and reduced coulombic efficiency [6,8].
suitable candidates for this task due to their interesting properties In a VARFB, vanadium crossover from the negative to the po-
such as independence of power and capacity, long cycle life and sitive electrode (Fig. 1, process 1) can lead to reduced coulombic
the good efficiency [1]. However, redox flow batteries generally efficiency and lowers the cycle life due to irreversible loss of ca-
exhibit a low specific energy in terms of mass and volume [1,2]. pacity. Additionally, the coulombic efficiency is decreased by
Although for most stationary installations irrelevant, low volu- oxygen permeation (Fig. 1, process 2) and the subsequent oxida-
tion of V2 þ to V3 þ in the negative electrode. Vanadium crossover
metric energy density can be an obstacle in applications with
also occurs in conventional redox flow batteries, therefore several
limited space (e.g. residential installation). A promising approach
approaches were published to increase the coulombic efficiency of
to increase the energy density of the all-vanadium redox flow
redox flow batteries by optimizing the membrane [9–20]. They
battery (VRFB) is the vanadium/air redox flow battery (VARFB;
include the use of new membranes based on polymer blends
Fig. 1) [3–7]. By substituting the positive redox couple (VO2 þ /
[13,14], organic/inorganic composite membranes (e.g. Nafion/SiO2
VO2 þ ) by the H2O/O2 couple, the positive tank volume can be or Nafion/(SiO2 modified TiO2) [17,21]), alternative proton con-
ducting polymers (e.g. sulfonated polyether ether ketone (SPEEK)
n
Corresponding author. [22]) or anion exchange polymers [15]. Despite these “bulk”
E-mail address: [email protected] (G. Wittstock). modifications, it was shown that the surface modification of

http://dx.doi.org/10.1016/j.memsci.2016.03.005
0376-7388/& 2016 Elsevier B.V. All rights reserved.
260 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269

polyelectrolyte solutions), the growth behavior of the LbL film can

be adjusted [28–30]. It was shown that a LbL film on the substrate
membrane can increase the performance of a VRFB (e.g. poly
(diallyldimethylammonium chloride) (PDADMAC)/PAA [31],
PDADMAC/poly(sodium styrene sulfonate) (PSS) [24] or chitosan/
phosphotungstic acid [20]).
Besides vanadium crossover, oxygen permeation is playing a
significant role in reducing the performance of a VARFB (Fig. 1)
[7,8]. During charging, oxygen is produced on the porous positive
electrode which is in direct contact with the membrane. The
Fig. 1. Schematic of the vanadium/air redox flow battery, the main reactions (black
solid lines) and undesired transfer processes through the membrane (numbered, produced oxygen can permeate in gaseous or dissolved form from
red dashed lines). the positive to the negative electrode. In the discharging process,
the gas diffusion layer in the positive electrode is supplied with
conventional membranes is a viable approach to improve the compressed air. To avoid mass transfer limitations, the supply is in
performance of the redox flow battery. Covalently bonding and surplus to the stoichiometric requirements. However, parts of this
cross-linking polymers on the membrane surface [12], attaching additional oxygen can also undergo permeation to the negative
polypyrrole [23] or assembling polyelectrolyte layers on the electrode.
membrane surface using layer-by-layer (LbL) deposition are ben- In various applications (e.g. water electrolysis, food packaging
eficial for reduced vanadium crossover [20,24]. or gas separation processes) the gas permeation properties of
The LbL deposition is a convenient and efficient surface mod- substrates were reduced via LbL deposition of polyelectrolytes
ification routine for charged substrates [25–27]. By dipping the [32]. For instance, oxygen permeability of PET substrates is re-
charged substrate alternatingly into oppositely charged polyelec- duced significantly by application of polyethylenimine (PEI) as
trolyte solutions, a layered structure is built up on the substrate polycation in multilayers (e.g. trilayers of PEI/montmorillonite
(Fig. 2a). By adjusting the deposition parameters (e.g. polyelec- clay/PAA [33], quadlayers of PEI/PAA/PEI/clay [34] or bilayer of PEI/
trolyte combination, deposition time, pH or ionic strength of PAA after cross-linking [35,32]). LbL films containing branched PEI

Fig. 2. (a) Principle of LbL deposition and the used nomenclature and (b) structure of the polyelectrolytes employed in this study.
 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269 261

were also proven to decrease multivalent ion crossover and in- Nafion solution for 5 min followed by rinsing with ultrapure water.
crease the selectivity towards monovalent cations such as H þ [36– The polyelectrolyte layers are denoted as follows: substrate-
38]. (polycation/polyanion)k, where k denotes the number of deposi-
The reduced vanadium crossover and reduced oxygen perme- tion cycles (Fig. 2a). The modified membranes were stored in ul-
ability are desired for VARFB operation for increased cycle life of trapure water.
the battery and better coulombic efficiency. In this contribution we
applied LbL films containing the polycation PEI and the polyanion 2.3. Membrane characterization
Nafion on Nafion 117 (N117) as a model substrate. PEI was chosen
because of its barrier properties both to multivalent cations and 2.3.1. Thickness
oxygen. Nafion ionomer was chosen as polyanion because it was The thickness of the membranes was determined at five dif-
shown to be suitable in LbL deposition [39–41], due to its good ferent positions of the membrane sample using a digital thickness
proton conductivity and chemical stability and because it was also dial gauge (Digital Foil Thickness Gauge FD 1000/30-3, Käfer
used as substrate material. Messuhrenfabrik GmbH & Co. KG, Germany). The membrane
To our knowledge, there are no studies dealing with the im- samples were carefully wiped dry with low-lint paper right before
provement of membranes for VARFB application. In the current measuring the thickness.
contribution, a membrane modification approach is presented
which reduced both vanadium crossover and oxygen permeation, 2.3.2. Infrared spectroscopy
and hence enables the improvement of the VARFB efficiency. There The IR absorption spectra of the samples were recorded using
are no reports about the simultaneous reduction of both multi- an attenuated total reflection (ATR) equipment with an FTIR
valent ion crossover and oxygen permeation by application of a spectrometer (Spectrum 100, Perkin Elmer, USA). The membranes
LbL film. Whereas the application of covalently bonded PEI on were investigated in a wetted condition, water on the surface was
Nafion [12] as well as LbL deposition of polyelectrolytes [20,24,31] removed with low-lint paper.
for improving VRFB membranes was described before, there is no
report where LbL films of branched PEI/Nafion were applied on a 2.3.3. Scanning electron microscopy
membrane. Cross-sections of the membrane samples were prepared via a
After an entire characterization of LbL deposited films of PEI freeze-fracture routine, i.e. the samples were freezed in liquid ni-
and Nafion ionomer on N117 substrates, the modified membrane trogen followed by snapping the sample into two parts using
was mounted in a VARFB test cell to evaluate the battery perfor- tweezers [43]. The samples were dried overnight at 60 °C in va-
mance. Furthermore, the stability of the modified membrane un- cuum before imaging the cross-sections with SEM (NEON 40, Zeiss,
der VARFB operation conditions was investigated. Germany). The acceleration voltage was 5 kV and the maximal
magnification was 600  .

2. Experimental 2.3.4. Thermogravimetric analysis

Vacuum-dried samples (60 °C, overnight) were investigated in
2.1. Electrolyte preparation a thermogravimetric analyzer (TGA 4000, Perkin Elmer, USA) be-
tween 30 °C and 600 °C (10 °C min
1) in N2 atmosphere
The V2 þ and V3 þ electrolytes were prepared electrolytically as (40 mL min
1).
described elsewhere [6,8,42]. A solution of 1.2 M VOSO4  x H2O
(x E3; 97%; Sigma Aldrich, Germany) and 2.0 M H2SO4 (95–97%; 2.4. Membrane performance
Fischer Scientific GmbH, Germany) was electrolyzed in a VRFB
setup using a potentiostat/galvanostat (Solartron Analytical Mod- 2.4.1. Vanadium permeability
ulab Pstat potentiostat/galvanostat, UK). The vanadium con- The permeability of V2 þ through the membranes was de-
centration can change during the electrolyte preparation and was termined using a setup described elsewhere [8,17,44]. Briefly, the
therefore determined with ICP-MS (XSERIES 2 ICP-MS, Thermo samples were pressed between the flanges of an H cell (active
Scientific, USA). The prepared V2 þ electrolyte had a vanadium area: 2.545 cm2), one compartment (excess side) was filled with
concentration of 1.04 M, the V3 þ electrolyte had a concentration of 20 mL of V2 þ electrolyte (preparation described above) and the
1.10 M. Both electrolytes were kept under a blanket of humidified other compartment (deficiency side) was filled with 1.2 M MgSO4/
N2 to avoid oxidation by air. 2 M H2SO4 solution and stirred (magnetic stir bar) during the ex-
periment. Both solutions were maintained under a blanket of
2.2. Membrane modification humidified N2 to prevent oxidation of V2 þ to V3 þ by oxygen from
the ambient air. Five aliquotes of the “deficiency side” were taken
Initially, all Nafion 117 membranes were cleaned and proto- within 24 h and the vanadium concentration was determined by
nated by subsequent treatment in 5% H2O2 (30%; Carl Roth GmbH ICP-MS. The same procedure was used to determine the perme-
& Co. KG, Germany) at 80 °C for 30 min, followed by rinsing with ability of V3 þ . The permeability of VO2 þ was determined analo-
deionized water, treatment in 1 M H2SO4 for 30 min at 80 °C and gously using 1.2 M VOSO4/2 M H2SO4 in the “excess side” and the
finally boiling in deionized water for 10 min. The protonated vanadium concentration in the “deficiency side” was determined
membranes were modified with polyelectrolytes via a LbL de- during the measurement by calibrated UV/Vis absorption at
position as depicted in Fig. 2a. The polyelectrolyte solutions con- 765 nm.
sisted of 0.1 mass% PEI (branched, average molecular weight
∼25 000 (by LS), Sigma Aldrich, Germany) and 0.1 mass% Nafion 2.4.2. Conductivity measurements
ionomer (Nafion perfluorinated resin solution, 5 mass% in lower The H cell setup with a reduced active area of 0.785 cm2 and
aliphatic alcohols and water, 1100 EW; Sigma-Aldrich, Germany) in with a 4 electrode arrangement was used for through-plane con-
ultrapure water (Fig. 2b). The LbL film was build up on the ductivity measurements similar to the description in literature
membranes similar to the routine described by Yang et al. [32]. [45]. The membrane was sandwiched between the flanges and
The membranes were repetitively immersed in the PEI solution for both compartments were filled with 2 M H2SO4. In each com-
5 min, rinsed thoroughly with ultrapure water, immersed in the partment a piece of Pt wire served as voltage sense which was
262 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269

placed in a fixed position close to the membrane (∼1 mm dis- 3. Results and discussion
tance). A disk of activated graphite felt (GFD5, SGL Carbon, Ger-
many; activation: 400 °C/18 h/air atmosphere) connected via a Pt 3.1. Membrane characterization
wire was the working electrode in each compartment. As con-
ductivity is strongly depending on temperature, the complete cell To verify the modification of the N117 substrate with the
was maintained at (20 70.1) °C during the measurements with a polyelectrolytes, ATR-FTIR spectra were recorded of the pure
thermostat (F12, Julabo, Germany). The resistance was determined polyelectrolyte PEI, of an unmodified N117 and of N117-(PEI/
using linear sweep voltammetry between
0.1 V and þ0.1 V with Nafion)k with k ¼5, 10, 20 (Fig. 3). The spectrum of PEI shows ty-
10 mV s
1. To obtain the membrane resistance only, the resistance pical features of the C–H bond (νa(C–H)∼2930 cm
1; νs(C–
of the H cell without membrane was subtracted from each mea- H)∼2815 cm
1; δ(C–H)∼1460 cm
1) and the N–H bond (δ(N–
surement. The membrane resistance measurements were repeated H)∼1585 cm
1) [49,50]. The N117 shows absorption related to C–F
three times and the average value was used. bonds (νa(C–F)∼1200 cm
1; νs(C–F)∼1150 cm
1), the C–O–C
group (νa/s(C–O–C)∼970–985 cm
1) and S–O bonds (νs(S–
2.4.3. Oxygen permeation O)∼1060 cm
1) as well as bending vibration of absorbed water (δ
The permeation of oxygen was determined with the so-called (O–H)∼1635 cm
1) in good agreement with the literature [51–54].
electrochemical monitoring technique described in the literature N117 does not show any C–H bond features because it is fully
[46–48]. In this method the permeation of oxygen through mem- fluorinated. It is clearly visible in Fig. 3 that with increasing k, the
branes is determined by reducing the permeated oxygen on a Pt PEI-related features of the C–H and N–H bond also increase which
electrode under mass transfer conditions. Specifically, a thin Pt mesh indicates an increased amount of PEI. Other features of the PEI
was placed on one side of the membrane sample and the assembly than the C–H stretching/deformation and the N–H deformation
was pressed between the flanges of the H cell (active area: cannot be used for identifying PEI on the substrate because these
2.545 cm2), the Pt mesh facing the “liquid” compartment that was absorptions overlap with substrate features. The performed FTIR
filled with 2 M H2SO4. The Pt mesh served as working electrode measurements show that the applied modification routine is sui-
while a Ag/AgCl reference electrode and a Pt grid as counter elec-
table to modify N117 membranes with polyelectrolytes. However,
trode were also placed into the “liquid” compartment. The 2 M H2SO4
the FTIR measurements only confirm qualitatively the deposition
was purged with humidified N2 for 15 min to remove any oxygen
of PEI. Therefore, TGA measurements of the unmodified and
and afterwards kept under a N2 blanket while the “gas” compart-
modified samples were conducted to gain quantitative informa-
ment was also flushed with N2. After purging, a potential of þ0.1 V
tion on the modifications.
vs. Ag/AgCl was applied to the Pt mesh which makes the oxygen
The TGA and negative derivatives of mass loss are shown in
reduction on the Pt mesh mass-transfer limited [46,48]. While still
Fig. 4 for unmodified membranes and membranes modified with
supplying N2 to the “gas” compartment, the background current was
polyelectrolytes. The TGA of Nafion can be divided into three re-
measured. After reaching a steady state current (approx. 10 min), the
gions. Up to 200 °C the Nafion membrane loses water [55–57], in
gas supply was changed from N2 to O2 while continuing the po-
the range of 300–400 °C C–S bonds are broken [56], the –SO3H
tentiostatic measurement and this was defined as t¼ 0. The cathodic
groups decompose [58] and/or the ether groups of the Nafion
current increased as oxygen permeated from the “gas” compartment
break down [55]. At temperatures above 400 °C, the polytetra-
through the membrane reaching the Pt mesh till a steady-state was
obtained and the experiment was stopped. The experiments were fluorethylene (PTFE) backbone of the Nafion is destructed [51,53].
conducted at 2171 °C. The TGA of samples of differently modified membranes appear
similar in the temperature range up to 200 °C. It can be seen in
2.5. Vanadium/air redox flow battery operation Fig. 4 that the TGA of the investigated membranes differs espe-
cially in the region between 300 °C and 400 °C. PEI decomposes at
The VARFB test cell described before [6] consisted of poly- temperatures up to 350/400 °C [59,60]. The difference of TGA be-
carbonate bodies with cavities in which the current collectors and tween 300 °C and 400 °C is most likely due to the different amount
electrodes were placed. The negative electrode consisted of a of PEI and Nafion on the membrane samples which are decom-
graphite current collector and an activated graphite felt, the po- posed in that temperature range. The inset in Fig. 4 shows the
sitive electrode comprised a titanium current collector with in-
tegrated flow field, a gas diffusion electrode decorated with Pt/C
and Nafion ionomer and an IrO2-modified graphite felt. The
membrane was placed between the two polycarbonate bodies,
separating the negative and positive electrodes. 12 mL of the V3 þ
electrolyte were employed as negative electrolyte which was
permanently purged with humidified N2 during the experiment.
The positive electrolyte was 50 mL of 2 M H2SO4. The electrolytes
were pumped through the electrodes with 100 mL min
1 during
charging, during discharging the positive electrolyte was not
pumped but the GDE was fed with 267 mL min
1 air through the
flow field and the anolyte was pumped as in charging mode. As
δ
the reactants of the positive electrode are supplied in excess, the
ν ν
battery capacity is limited by the capacity of the anolyte. To δ
minimize hydrogen evolution and to define comparable experi-
mental conditions, the VARFB was always charged up to 83% of the
anolyte capacity. The applied current density was 20 mA cm
2. An
ICP-MS sample was taken from the positive electrolyte after dis-
charging to determine the vanadium crossover during one cycle.
ν
Operation of the VARFB was conducted at room temperature
(21 71) °C. Fig. 3. ATR-FTIR spectra of PEI, N117 and N117-(PEI/Nafion)k with k ¼5, 10, 20.
 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269 263

modified with k¼ 20 has a spongy and porous structure. As dis-

cussed in several publications [25,62,63], surface roughness of the
LbL film and exponential growth are linked to each other. McA-
loney et al. [64] investigated the growth of PDADMAC/PSS LbL
films in dependence of the ionic strength of the polyelectrolyte
solutions. They found via AFM measurements that in the regime
with the strongest growth per bilayer, the surface roughness in-
creases with k. Likewise, Choi et al. [63] observed for exponential
buildup of poly(2-(dimethylamino)ethyl methacrylate) (PDMAE-
MA)/PAA multilayers a non-uniform surface with a higher surface
roughness which is increasing with k as it is also observed here. It
is assumed that exponential growth occurs due to the “in” and
“out” diffusion of at least one of the polyelectrolytes within the
whole film during each deposition step. This leads to exponential
growth of the LbL film and to a rougher surface in comparison to a
LbL film built up with “true” bilayer consisting of closed, homo-
geneous and equally thick layers on top of oppositely charged with
the same morphological properties. Although exponential growth
is desired for the fast buildup of thick LbL films [25], for well-
Fig. 4. (a) TGA and (b) negative derivatives of N117 and N117-(PEI/Nafion)k with
defined film properties a homogeneous layer deposition is needed.
k ¼ 5, 10, 20. The inset in (a) shows the weight loss of membranes with k ¼5, 10, 20 In the case of the chosen polyelectrolyte system and deposition
in % between 300 °C and 400 °C subtracted by the weight loss of N117 between parameters, (PEI/Nafion)10 can be deposited on N117 as compara-
300 °C and 400 °C. tively compact, smooth and homogeneous film.
The TGA and SEM results suggested a non-linear LbL film
weight loss of the modified membranes between 300 °C and growth. This hypothesis is verified by thickness measurements
400 °C subtracted by the weight loss of the unmodified membrane (Fig. 6). The film thickness increases non-linearly with k which is
between 300 °C and 400 °C, i.e. the weight loss due to the LbL film. in agreement with the aforementioned results. A non-linear re-
With increasing k, the weight loss increases as well. This increase lationship between k and thickness was also found by Yang et al.
is not linear but seems to be exponential which can be a hint that for a (PEI/PAA)k system [35].
the growth of LbL film under the given deposition conditions is not
linear. Exponential growth was found for PEI/PAA films assembled 3.2. Membrane performance
from polyelectrolytes with pH 9 for the PEI and pH 2.85 for the
PAA solution [61]. After preparation without any pH adjustment, For the application of the modified membranes in VARFB the
the PEI solution used in this work had a pH 10–11 and the Nafion proton conductivity, vanadium permeation and oxygen permea-
solution pH 3. tion are important and thus discussed in the following.
To monitor the morphology and roughness of the LbL film, The relevant conductivity in membranes is in general the
cross-sections were examined with SEM (Fig. 5). Both the cross- through-plane conductivity as this is the direction along which
sections and the surface/main planes of the membranes which protons move. However, most conductivity values given in the
were modified with the PEI/Nafion LbL film are imaged. The sur- literature are in-plane conductivities. This is probably due to its
face of the non-modified N117 is smooth and even. The samples easier, reproducible determination [65]. However, there is no
N117-(PEI/Nafion)k with k¼ 5 and k ¼10 show features and have a consensus in the literature whether Nafion has anisotropic con-
slightly heterogeneous surface. The surface roughness of the ductivity [66,67] or not [65,68]. The conductivity of N117-(PEI/
modification increases with k. The difference between the sample Nafion)k must be determined as through-plane conductivity as
k ¼10 and k ¼20 is the most distinct one; the k ¼10 sample has a obviously the in-plane and through-plane conductivity cannot be
quite homogeneous surface structure while the membrane equivalent due to the layered structure. Reported values for

Fig. 5. SEM images of cross-sections of N117 and N117-(PEI/Nafion)k with k ¼5, 10, 20.
264 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269

k
0 5 10 20
100
a)
90

80

70
9.1
b)
9.0

0.8

0.6
Fig. 6. Mean values of the thickness of N117 (k ¼0) and N117-(PEI/Nafion)k with
k ¼1, 5, 10, 20 and standard deviations of the five measurements.
0.4

0.2
Table 1
c)
20
Literature values and value determined in this work for Nafion conductivity in
sulfuric acid media.

Conductivity Temperature [°C] Medium Method [reference]

[S cm
1] 10

0.088 20 1M H2SO4 Through-plane [69]

0.085 20 3M H2SO4 Through-plane [69]
0.11 22 2M H2SO4 n/a [70] 0
0.076 25 1M H2SO4 In-plane [71] 0 5 10 20
0.071 25 1M H2SO4 Through-plane, cell 1
[72] k
0.084 30 1 M H2SO4 Through-plane, cell 2
þ 2þ
[72] Fig. 7. (a) H conductivity σ H+ , (b) V permeability P V2 + and (c) selectivity
0.0899 20 2 M H2SO4 Through-plane (this σ H+/P V2 + of N117 (k ¼ 0) and N117-(PEI/Nafion)k with k ¼5, 10, 20.
study)
along with an unwanted decrease in proton conductivity as the LbL
film reduces the overall cation permeation. For the determination
through-plane conductivity measurements of Nafion show sub-
of the optimal k, the permeation coefficient for vanadium cations
stantial scatter depending on the design of the conductivity cell,
P Vx + through the membrane need to be measured and set in re-
the measurements conditions as well as the pretreatment of the
lation to the H þ conductivity. Following the routine described in
membrane. Table 1 shows selected conductivity values of Nafion
the literature [8,17,44], P Vx + through the membrane is determined
membranes obtained under different measurement conditions in by measuring the transport of vanadium species from a highly
sulfuric acid and the conductivity of N117 determined in this work concentrated vanadium solution (excess side) with concentration
which is in good agreement with most of the reported values gi- ce to an initially vanadium-free solution (deficiency side; vana-
ven in Table 1.
dium concentration cd = 0 at time t¼0) of equal ionic strength.
With increasing k the conductivity decreases (Fig. 7a). The Assuming that ce, the permeation coefficient P, the membrane area
conductivity is decreased by 22% from 0.0899 S cm
1 for N117 to A, the membrane thickness x and the volume of the deficiency side
0.070 S cm
1 for N117-(PEI/Nafion)20. This finding can be ex- Vd are independent of t, the permeation coefficient P can be de-
plained with higher resistance of the modified membrane due to duced from the slope of the line ln(ce/(ce − cd(t ))) vs. t [8,44].
lower cation conduction. The ion conduction and especially proton P V2 + values were determined because V2 þ permeates through a
conduction mechanism in Nafion membranes is well described in
N117 membrane much faster than V3þ ( P V2 + = 9.05·10−6 cm2 min−1,
the literature [73–75]. From a microstructural point of view, hy-
drated Nafion consists of interpenetrating hydrophilic and hy- P V3 + = 4.35·10−6 cm2 min−1; Fig. 8). Consequently, the permeability
drophobic domains [73]. The hydrolyzed sulfonic acid groups of for V2 þ is more critical for the VARFB performance. The dependence
Nafion form water channels. Proton conduction through these of P V2 + on k is shown in Fig. 7b. P V2 + = 3.50 ·10−7 cm2 min−1 of the
channels occurs both via a hopping or Grotthus-like mechanisms N117-(PEI/Nafion)10 is more than one order of magnitude (or approx.
and/or by a diffusion or “vehicular” mechanism [74]. The amine
96%) lower than P V2 + of N117. Interestingly, the N117-(PEI/Nafion)20
groups of PEI interact via hydrogen bonds or protonation with the membrane has a slightly higher P V2 + = 4.03·10−7 cm2 min−1 com-
sulfonic acid groups of the Nafion or with the dopant sulfuric acid, pared to the N117-(PEI/Nafion)10. This observation might be the result
similarly as it was described for polybenzimidazole (PBI) inter- of the porous, spongy structure of the LbL film which was observed
acting with acids [76,77]. Thus, proton conduction through the LbL with SEM. The more opened layer structure enables faster permeation
film occurs via proton hopping between the amine groups and the by cations such as V2 þ . However, the proton conductivity of the N117-
acid (sulfonic acids groups of Nafion or H2SO4). This was also re- (PEI/Nafion)20 membrane is not higher than the conductivity of the
ported for PEI doped with H2SO4 [78]. N117-(PEI/Nafion)10 sample.
A desired reduction of vanadium permeation typically comes Similar to the reduced proton conductivity (Fig. 7a), the lower
 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269 265

Fig. 9. Background-current corrected current transients due to oxygen reduction

Fig. 8. Permeation coefficients of V2 þ , V3 þ and VO2 þ through N117 and N117-(PEI/ after permeation through unmodified N117 and N117-(PEI/Nafion)10.
Nafion)10.
Nafion)10 in comparison to N117. By application of (PEI/Nafion)10
vanadium permeation can be attributed to the repulsive forces due on the N117 membrane the permeability of V3 þ is reduced by
to the same charge of the cationic PEI layers and the vanadium nearly two orders of magnitude from 4.35  10
6 cm2 min
1 to
cations. However, the reduction in vanadium permeation is much 8.55  10
8 cm2 min
1. For the unmodified N117, the permeability
more pronounced than the reduction of proton conductivity (re- of VO2 þ was determined as 39.53  10
7 cm2 min
1 which is in
duction of P V2 + by approx. 92% due to (PEI/Nafion)5 vs. reduction of good agreement with [81] (36.55  10
7 cm2 min
1). The perme-
σ H+ by approx. 2.5% due to (PEI/Nafion)5). The reason for this is ability of VO2 þ is reduced to 6.62  10
7 cm2 min
1 by application
most probably that the proton conductivity in the LbL film can be of (PEI/Nafion)10 on N117 whereas the modification of a N117 with
maintained by proton hopping mechanism while other cations a covalently bonded PEI layer proposed in [81] resulted in a
than protons are transferred only via the vehicular mechanism slightly lower permeability of 5.23  10
7 cm2 min
1.
[79]. Additionally, due to the higher valency of the vanadium ions The oxygen permeation of N117 and N117-(PEI/Nafion)10 were
(þ 2 and þ3, respectively) in comparison to the monovalent determined by electrochemical reduction of oxygen after the
protons, the electrostatic repulsion and thus donnan exclusion of permeation through the membrane. The steady-state oxygen re-
the positively charged V2 þ and V3 þ by the positively charged PEI duction current j∞ (corrected by the background current) re-
is higher than to H þ , leading to a more pronounced decrease of presents the steady-state oxygen permeation through the mem-
Vx þ conductivity compared to H þ [36,80]. brane (Fig. 9). This current is j∞ = 82.1 μA cm−2 for the unmodified
For improving the overall performance of a membrane by a
N117 and j∞ = 71.57 μA cm−2 for the N117-(PEI/Nafion)10 mem-
modification routine, the reduced vanadium permeation should
brane, i.e. the permeation of oxygen is reduced by approx. 13% by
prevail over the reduced proton conductivity. A common measure
the LbL film. However, the N117-(PEI/Nafion)10 membrane is also
is the selectivity which is the ratio of proton conductivity to V2 þ
thicker and therefore a comparison of the permeability which also
ion permeability S = σ H+/P V2 + [13,14]. The higher the selectivity,
the more efficient is the modification. In Fig. 7c the selectivity is accounts for the thickness of the membrane is advisable to draw a
shown for N117 (k ¼0) and N117-(PEI/Nafion)k with k ¼5, 10, 20. S conclusion about the permeation resistance of the LbL film.
of N117 is approx. 1·104 S min cm−3 (Fig. 7). It can be seen that S The permeability P is calculated as P = (j∞ ·x ) /(z·F ) with x being
has a maximum at N117-(PEI/Nafion)10 with approx. the membrane thickness, z ¼4 the transferred electrons per spe-
21·104 S min cm−3, i.e. the modification with (PEI/Nafion)10 is the cies and F is the Faraday constant [47,82]. The permeability is
most efficient one due to the best compromise between reduced 3.94  10
12 mol s
1 cm
1 for the unmodified N117. This value is in
V2 þ permeation and reduced H þ conductivity. The achieved se- good agreement with literature values (Table 2), however, the
lectivity improvement by a factor of 21 is remarkable also in measurement technique, the experimental conditions and the
comparison to literature. Xi et al. [24] reported a maximum se- water content of the membrane influence the permeability mea-
lectivity (concerning VO2 þ ; σ H+/P VO2 + ) increase factor of almost surement significantly [83]. The permeability of N117-(PEI/
7 due to (PDDA/PSS)5 on a Nafion membrane and Lu et al. [20] Nafion)10 was 3.66  10
12 mol s
1 cm
1, i.e. the reduction of P
observed a maximum selectivity ( σ H+/P VO2 + ) for (chitosan/phos- due to (PEI/Nafion)10 is approx. 7%.
photungstic acid)3 on a Nafion membrane which was about The reduction in oxygen permeability due to the LbL film can be
2 times higher than that of the pristine membrane. The reduced explained with the solution–diffusion model which is widely used
vanadium permeation is crucial for the lifetime of the VARFB. to describe gas transport through polymeric materials [86]. Per-
Therefore, the following experiments, evaluation of V3 þ and VO2 þ meation of gases starts with adsorption of the gas on the surface.
permeability, investigation of oxygen permeation and perfor- Subsequently, the gas is dissolved in the membrane material and
mance in VARFB, are conducted with N117-(PEI/Nafion)10 which then it diffuses through the membrane. Finally, the gas exits the
was the membrane with the highest selectivity σ H+/P V2 + . membrane phase by getting dissolved in another phase (e.g. water).
The permeabilities of V2 þ , V3 þ and VO2 þ through unmodified Solution and diffusion depend on the interactions of the permeate
N117 and N117-(PEI/Nafion)10 are shown in Fig. 8. Although not and the membrane material. In a membrane consisting of only one
relevant for VARFB, the permeability of VO2 þ was determined in polymer (such as the unmodified N117), ideally the permeating gas
order to allow an estimation of the performance of N117-(PEI/ only needs to be dissolved once before it can diffuse through the
Nafion)10 in VRFB application. Overall, a significant reduction of membrane and leave it by dissolution in another phase. However, in
permeability of all vanadium species is observed for the N117-(PEI/ the case of LbL polyelectrolyte film on the membrane the situation
266 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269

Table 2
Selected literature values for permeability P of O2 through Nafion membranes determined with different measurement techniques and under different conditions.

P [mol s
1 cm
1] Membrane Temperature [°C] Liquid medium Method, conditions [reference]

11.59  10
12 Nafion 117 25 1 N H2SO4 Electrochemical monitoring technique, gas not humidified, no membrane pretreatment [84]
1.73  10
12 Nafion 120 20 0.5 M K2SO4 Electrochemical monitoring technique, gas humidified, boiling pretreatment [47]
4.2  10
12 Nafion 117 25 – Electrochemical in situ method in a fuel cell using a microelectrode, 82% relative humidity (RH) [85]
3.94  10
12 Nafion 117 21 2 M H2SO4 Electrochemical monitoring technique, membrane pretreated, gas humidified (this study)

is different [35]. The LbL film is not believed to be homogeneous

Table 3
and equally thick but more like an interpenetrating network having Efficiencies and vanadium crossover (determined by ICP-MS of the positive
a scrambled-egg like structure [35]. This explanation is in good electrolyte) during VARFB operation at j¼20 mA cm
2 of N117 and N117-(PEI/
agreement with the increasing roughness of the films with in- Nafion)10.
creasing LbL steps. Due to this irregular structure bearing a lot of
ηC [%] ηE [%] ηV [%] V crossover [mmol]
interfaces between the two polyelectrolytes, gases need to be dis-
solved, diffuse and get redissolved several times when permeating N117 81.32 41.54 51.08 0.763
through the membrane. This slows down the overall mobility of the N117-(PEI/Nafion)10 93.18 45.2 48.51 0.246
gas through the membrane and leads to barrier properties of the
(PEI/Nafion)10 [35].
efficiency was increased from ηE = 41.5% to ηE = 45% .
ICP-MS measurements were conducted to quantify the irre-
3.3. VARFB operation with modified membrane
versible capacity loss due to vanadium crossover during the cycle.
The vanadium crossover during VARFB operation is reduced by
Fig. 10 shows the charge and the discharge curve of the VARFB
0.517 mmol (approx.
70%) if N117 is modified with (PEI/Nafion)10
operation using a N117 (dashed lines) and using a N117-(PEI/ (Table 3). In contrast to the determination of permeation coeffi-
Nafion)10 membrane (solid lines) at j¼ 20 mA cm
2. Interestingly, cients described above, the vanadium crossover during VARFB
the overpotentials during charging are nearly identical, bearing in operation is caused by diffusion, migration and electroosmotic
mind the reduced proton conductivity of the membrane. This convection (EOC) [87].
observation reveals that the membrane resistance is only a small Another important influence on the coulombic efficiency and
part of the total cell resistance and the charge-transfer resistance capacity fade during VARFB operation is the water transfer be-
of the water splitting reaction is the dominant contribution. tween the electrolytes of the two half-cells. The possible water
However, the average discharging potential is lower when using transfer mechanisms occurring during VARFB operation are the
the N117-(PEI/Nafion)10 membrane compared to the unmodified transport of water as hydration shells of diffusing vanadium ions
N117 membrane. This is attributed to a low charge-transfer re- [88], osmosis and EOC/migration (e.g. hydration shell of dragged
sistance of the oxygen reduction reaction in comparison to other protons) as depicted in Fig. 11. Table 4 shows the volumes of the
resistances such as the membrane resistance. The lower dischar- negative electrolyte before charging, after charging and after dis-
ging potential affects the voltaic efficiency of the VARFB which is charging of the VARFB using a N117 membrane and a N117-(PEI/
2.5% lower than in the cycle with the unmodified N117 (Table 3). Nafion)10 membrane, respectively.
During discharging of the VARFB with the N117-(PEI/Nafion)10 Water is transferred from the negative to the positive electrode
membrane more capacity can be retrieved from the battery. The by the hydration shells of V2 þ and V3 þ which undergo crossover
coulombic efficiency of the system with N117 in comparison to the due to diffusion and in the opposite direction due to osmosis
N117-(PEI/Nafion)10 is increased by more than 11% from driven by the ion concentration differences between negative and
ηC = 81.32% to ηC = 93.18% (Table 3). The reason for the increase of positive electrolyte. Water transfer due to EOC/migration depends
coulombic efficiency is the reduction of both loss mechanisms; i.e. on the direction of the electric field and occurs from the positive
vanadium crossover and oxygen permeability. Overall, the energy electrode to the negative electrode during charging and vice versa
during discharging. The net water transfer was observed to take
place from the positive to the negative electrode during charging
whereas it is directed from the negative electrode towards the
positive electrode during discharging for both membranes (Ta-
ble 4). Additionally, it was observed for both membranes that the
net water transfer from the positive to the negative electrode

Fig. 10. Charge/discharge curves of VARFB operated at (217 1) °C applying an Fig. 11. Water transfer during charging and discharging of a VARFB due to hydra-
unmodified N117 and N117-(PEI/Nafion)10. tion shells of diffusing Vx + , osmosis, EOC/migration and the net water transfer.
 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269 267

Table 4
Volumes of the negative electrolyte before charging, after charging and after dis-
charging of the VARFB using a N117 or a N117-(PEI/Nafion)10 membrane.

Volume [mL] Before charging After charging After discharging

N117 12.0 13.9 13.3

N117-(PEI/Nafion)10 12.0 13.8 13.6

during charging is higher than the net water transfer from the
negative to the positive electrode during discharging (Table 4). As
it can be assumed that the water transfer rates due to osmosis and
by hydration shells of diffusing vanadium ions are independent of
the direction of the electric field, the different net water transfer is
caused by the opposite direction of EOC/migration-driven water
transfer during charging and discharging.
Comparing the two membranes, the net water transfer is nearly
the same for both membranes during charging. However, less
water is transferred from the negative electrode to the positive
electrode during discharging when a N117-(PEI/Nafion)10 mem-
Fig. 12. (a) TGA, (b) negative derivative of TGA of the membrane N117-(PEI/
brane is used. These observations can be explained under con-
Nafion)10 before and after VARFB operation.
sideration of migration processes accompanied by water transport
due to hydration shells. During charging (ideally) only H þ can
migrate in the electric field through the cation exchange mem-
brane from the positive electrode to the negative electrode. During
discharging, the electric field is reversed and migration occurs
from the negative to the positive electrode. Amongst proton mi-
gration, vanadium cation migration can also occur. Each trans-
ferred H þ is accompanied by 2.5 H2O molecules [44,89] while 6–
7.5 H2O molecules as inner hydration shell are carried with each
V2 þ ion and 5–6 H2O molecules with each V3 þ ion [90]. Thus, the
migration of vanadium ions during discharging is connected with
a significant water transfer from the negative to the positive
electrode. As shown above, the LbL film represents a barrier to
multivalent ions such as V2 þ and V3 þ . Hence, it can be assumed
that the migration of V2 þ and V3 þ will be reduced when using the
N117-(PEI/Nafion)10 membrane. Consequently, the overall water
transfer from the negative electrode to the positive electrode
during discharging should be less pronounced when using the
N117-(PEI/Nafion)10 membrane which is in agreement with the
observation of the volume changes during discharging (Table 4).
The reduction of oxygen crossover during VARFB operation by
Fig. 13. IR spectra of the membrane N117-(PEI/Nafion)10 before and after VARFB
using N117-(PEI/Nafion)10 can be deduced from the increased operation and after immersing the membrane in 2M H2SO4 for one week.
discharge capacity which is raised by 34.8 mAh ( þ15%) in com-
parison to N117 (Fig. 10). Bearing in mind that only V2 þ crossover operation. It was shown that cations within a Nafion membrane can
reduces the coulombic efficiency, the reduced vanadium crossover reduce the thermal stability since Lewis acids catalyze the break-
which corresponds to a charge of 13.9 mAh ( Q = n·z·F with z ¼1) down of ether bonds of the side chains [91,92].
can account at most for 40% of the discharge capacity increase. In Fig. 13 the IR spectrum of a non-used N117-(PEI/Nafion)10
Consequently, the additional capacity gain when using the N117- membrane is shown in comparison to such a membrane used in a
(PEI/Nafion)10 membrane which cannot be related to reduced va-
VARFB and another one which was immersed in 2 M H2SO4 for one
nadium crossover, 20.9 mAh, is an indirect evidence for lower
week. The features related to C–H bonds (νa(C–H)∼2930 cm
1;
oxygen permeability during VARFB operation.
νs(C–H)∼2815 cm
1; δ(C–H)∼1460 cm
1); the N–H bond (δ(N–
The stability of the membrane under VARFB cycling conditions
H)∼1585 cm
1) can be seen in all three samples. Thus, it can be
was investigated by TGA and IR measurements of the N117-(PEI/
concluded that the PEI and therefore the LbL film is still present on
Nafion)10 membrane before and after the VARFB cycle (Figs. 12 and
13). During TGA measurements, the sample after VARFB operation the membrane even after use in VARFB and after immersion in
shows a weight loss of 14.7% in the temperature range of 300– 2 M H2SO4. Additionally, the N117-(PEI/Nafion)10 membrane was
400 °C where PEI decomposition occurs (as described above) slightly opaque (milky color) due to the LBL film modification. This
whereas the sample before VARFB operation lost 9.2% in that appearance was still present even after VARFB operation.
temperature region (Fig. 12). This higher weight loss after VARFB However, further experimental investigations of the stability of
operation does not allow any conclusion about the stability of the the LbL are necessary. The investigation of the performance of the
LbL film but only indicates that after VARFB operation additional membrane in a VARFB during several cycles should be conducted
substances are present in the membrane which are influencing the to evaluate the stability of the LbL membrane under VARFB cycling
thermal degradation. Indeed, a new degradation process beginning conditions as it was done for example by Xi et al. [93] and Lu et al.
at approx. 340 °C can be observed. This process might be caused by [20] to prove the stability of LbL modified membranes in VRFB.
vanadium cations incorporated into the membrane during VARFB However, for this purpose the optimization of all VARFB cell
268 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269

components with regards to longer life-time is necessary which References

was beyond the scope of this paper.
[1] M. Skyllas-Kazacos, M.H. Chakrabarti, S.A. Hajimolana, F.S. Mjalli, M. Saleem, J.
Electrochem. Soc. 158 (2011) R55, http://dx.doi.org/10.1149/1.3599565, URL
〈http://jes.ecsdl.org/cgi/doi/10.1149/1.3599565〉.
4. Conclusion [2] P. Leung, X. Li, C. Ponce de León, L. Berlouis, C.T.J. Low, F.C. Walsh, RSC Adv. 2
(2012) 10125, http://dx.doi.org/10.1039/c2ra21342g, URL 〈http://xlink.rsc.org/
?DOI¼ c2ra21342g〉.
In this contribution we showed the successful modification of [3] H. Kaneko, A. Negishi, K. Nozaki, K. Sato, M. Nakajima, EU Patent 0517217A1,
N117 membranes with LbL films of PEI and Nafion ionomer and 1992.
[4] C. Menictas, M. Skyllas-Kazacos, J. Appl. Electrochem. 41 (2011) 1223–1232,
the beneficial effect of the application of a N117-(PEI/Nafion)10 http://dx.doi.org/10.1007/s10800-011-0342-8, URL 〈http://link.springer.com/
membrane in VARFB in comparison to a plain N117 membrane. 10.1007/s10800-011-0342-8〉.
[5] J. Noack, C. Cremers, D. Bayer, J. Tübke, K. Pinkwart, J. Power Sources 253
N117-(PEI/Nafion)10 increased the selectivity S = σ H+/P V2 + by a
(2014) 397–403, http://dx.doi.org/10.1016/j.jpowsour.2013.12.070, URL 〈http://
factor of 21 and decreased oxygen permeation by
13% (oxygen linkinghub.elsevier.com/retrieve/pii/S0378775313020491〉.
permeability:
7%) compared to bare N117. The N117-(PEI/ [6] J. Grosse Austing, C. Nunes Kirchner, E.-M. Hammer, L. Komsiyska,
G. Wittstock, J. Power Sour. 273 (2015) 1163–1170, http://dx.doi.org/10.1016/j.
Nafion)10 exhibit a better performance in a VARFB in terms of
jpowsour.2014.09.177, URL 〈http://linkinghub.elsevier.com/retrieve/pii/
coulombic efficiency, energy efficiency and reduced vanadium S0378775314016000〉.
crossover which is crucial for the lifetime of the VARFB. [7] S. Hosseiny, M. Saakes, M. Wessling, Electrochem. Commun. 13 (2011)
751–754, http://dx.doi.org/10.1016/j.elecom.2010.11.025, URL 〈http://linkin
The vanadium crossover reduction is already very efficient with
ghub.elsevier.com/retrieve/pii/S1388248110004996〉.
(PEI/Nafion)10 (
70% in VARFB operation compared to unmodified [8] J. Grosse Austing, C. Nunes Kirchner, L. Komsiyska, G. Wittstock, J. Power Sour.
N117). In this study the focus was the reduction of vanadium 306 (2016) 692–701, http://dx.doi.org/10.1016/j.jpowsour.2015.12.052.
[9] H. Prifti, A. Parasuraman, S. Winardi, T.M. Lim, M. Skyllas-Kazacos, Membranes
crossover which is limiting the lifetime of the battery. However, 2 (2012) 275–306, http://dx.doi.org/10.3390/membranes2020275, URL 〈http://
reduced oxygen permeability is needed for increased coulombic www.pubmedcentral.nih.gov/articlerender.fcgi?ar
efficiency. The permeability of oxygen is reduced by 7% due to tid¼ 4021890&tool¼ pmcentrez&rendertype ¼abstract, http://www.mdpi.
com/2077-0375/2/2/275/〉.
(PEI/Nafion)10. Although not the focus in this study, the oxygen [10] B. Schwenzer, J. Zhang, S. Kim, L. Li, J. Liu, Z. Yang, ChemSusChem 4 (2011)
permeation might be decreased further using an alternative LbL 1388–1406, http://dx.doi.org/10.1002/cssc.201100068, URL 〈http://doi.wiley.
com/10.1002/cssc.201100068〉.
modification. For example, 70% less oxygen permeability was re-
[11] X. Li, H. Zhang, Z. Mai, H. Zhang, I. Vankelecom, Energy Environ. Sci. 4 (2011)
ported for (polyallylamine hydrochloride/PSS)20 [94] or reduction 1147, http://dx.doi.org/10.1039/c0ee00770f, URL 〈http://xlink.rsc.org/?
by factor 100 by applying 4 quadlayers of PEI, PAA, PEI, and sodium DOI¼ c0ee00770f〉.
[12] Q. Luo, H. Zhang, J. Chen, P. Qian, Y. Zhai, J. Membr. Sci. 311 (2008) 98–103,
montmorillonite clay [34]. However, these modifications were http://dx.doi.org/10.1016/j.memsci.2007.11.055, URL 〈http://linkinghub.else
made in different context (fuel cell and food packaging) and might vier.com/retrieve/pii/S0376738807008484〉.
not be as efficient in VARFB application. Additionally, the mod- [13] D. Chen, S. Kim, V. Sprenkle, M.a. Hickner, J. Power Sour. 231 (2013) 301–306,
http://dx.doi.org/10.1016/j.jpowsour.2013.01.007.
ifications might not be efficient in terms of the crucial vanadium [14] S. Liu, L. Wang, D. Li, B. Liu, J. Wang, Y. Song, J. Mater. Chem. A 3 (2015)
crossover reduction. 17590–17597, http://dx.doi.org/10.1039/C5TA04351D, URL 〈http://xlink.rsc.
The modified membranes might also be suitable for application org/?DOI ¼ C5TA04351D〉.
[15] M.-s.J. Jung, J. Parrondo, C.G. Arges, V. Ramani, J. Mater. Chem. A 1 (2013)
in VRFB for enhanced coulombic efficiency due to reduced cross- 10458–10464, http://dx.doi.org/10.1039/C3TA11459G.
over. Likewise, they could be beneficial for applications where [16] T.N.L. Doan, T.K.A. Hoang, P. Chen, RSC Adv. 5 (2015) 72805–72815, http://dx.
proton conductivity is needed and at the same time multivalent doi.org/10.1039/C5RA05914C.
[17] J. Xi, Z. Wu, X. Qiu, L. Chen, J. Power Sour. 166 (2007) 531–536, http://dx.doi.
ion and/or uncharged molecule transport must be reduced. This org/10.1016/j.jpowsour.2007.01.069, URL 〈http://linkinghub.elsevier.com/re
applies for example to microbial fuel cells (MFC) where the trieve/pii/S037877530700239X〉.
[18] X. Teng, Y. Zhao, J. Xi, Z. Wu, X. Qiu, L. Chen, J. Power Sour. 189 (2009)
crossmixing of oxygen and fuel (biological material) through the
1240–1246, http://dx.doi.org/10.1016/j.jpowsour.2008.12.040, URL 〈http://lin
membrane needs to be limited while maintaining proton con- kinghub.elsevier.com/retrieve/pii/S0378775308023689〉.
ductivity of the membrane [95]. The performance of direct me- [19] H.H. Zhang, H.H. Zhang, X. Li, Z. Mai, W. Wei, Energy Environ. Sci. 5 (2012)
6299, http://dx.doi.org/10.1039/c1ee02571f, URL 〈http://xlink.rsc.org/?
thanol fuel cells (DMFC) also suffers from crossover processes such DOI¼ c1ee02571f〉.
as methanol crossover [96]. Albeit not investigated in this con- [20] S. Lu, C. Wu, D. Liang, Q. Tan, Y. Xiang, RSC Adv. 4 (2014) 24831, http://dx.doi.
tribution, the presented modified membranes might also reduce org/10.1039/c4ra01775g, URL 〈http://xlink.rsc.org/?DOI ¼ c4ra01775g〉.
[21] X. Teng, Y. Zhao, J. Xi, Z. Wu, X. Qiu, L. Chen, J. Membr. Sci. 341 (2009) 149–154,
methanol crossover due to the layered structure and could in- http://dx.doi.org/10.1016/j.memsci.2009.05.051.
crease the performance of DMFC. [22] G. Merle, F.C. Ioana, D.E. Demco, M. Saakes, S.S. Hosseiny, Membranes 4 (2013)
For long-term operation the stability of the LbL films needs to be 1–19, http://dx.doi.org/10.3390/membranes4010001.
[23] J. Zeng, C. Jiang, Y. Wang, J. Chen, S. Zhu, B. Zhao, R. Wang, Electrochem.
evaluated. There are not many studies in the literature in which the Commun. 10 (2008) 372–375, http://dx.doi.org/10.1016/j.elecom.2007.12.025.
stability of LbL films is the key aspect [26]. The electrostatic forces [24] J. Xi, Z. Wu, X. Teng, Y. Zhao, L. Chen, X. Qiu, J. Mater. Chem. 18 (2008) 1232,
http://dx.doi.org/10.1039/b718526j, URL 〈http://xlink.rsc.org/?
between the oppositely charged polyelectrolytes are quite strong
DOI¼ b718526j〉.
and the film formation is entropy-driven. Therefore, a high stability [25] Y. Li, X. Wang, J. Sun, Chem. Soc. Rev. 41 (2012) 5998, http://dx.doi.org/
against dissolution can be expected [97] but requires experimental 10.1039/c2cs35107b.
[26] P. Bertrand, A. Jonas, A. Laschewsky, R. Legras, Macromol. Rapid Commun. 21
verification. Likewise, the performance of the modified membranes (2000) 319–348, http://dx.doi.org/10.1002/(SICI)1521-3927(20000401)
at different current densities needs to be evaluated. 21:73.0.CO;2-7, URL 〈http://apps.webofknowledge.com/full_record.do?pro
duct ¼UA&search_mode¼ GeneralSearch&qid ¼ 1&SID¼ V1fSIHRgc2MX8e
V26pd&page¼1&doc¼ 1〉.
[27] Y. Xiang, S. Lu, S.P. Jiang, Chem. Soc. Rev. 41 (2012) 7291, http://dx.doi.org/
Acknowledgments 10.1039/c2cs35048c.
[28] L. Xu, D. Pristinski, A. Zhuk, C. Stoddart, J.F. Ankner, S.A. Sukhishvili, Macro-
molecules 45 (2012) 3892–3901, http://dx.doi.org/10.1021/ma300157p, URL
The authors would like to thank Dana Schonvogel for ICP-MS 〈http://pubs.acs.org/doi/abs/10.1021/ma300157p〉.
measurements and Pratik Das, Benedikt Berger and Khrysthyna [29] P. Bieker, M. Schönhoff, Macromolecules 43 (2010) 5052–5059, http://dx.doi.
Yezerska for assisting with experimental work. The Reiner-Le- org/10.1021/ma1007489, URL 〈http://pubs.acs.org/doi/abs/10.1021/
ma1007489〉.
moine-Stiftung is acknowledged for providing a Ph.D. scholarship [30] D.A. Hagen, B. Foster, B. Stevens, J.C. Grunlan, ACS Macro Lett. 3 (2014)
to Jan grosse Austing. 663–666, http://dx.doi.org/10.1021/mz500276r.
 J. grosse Austing et al. / Journal of Membrane Science 510 (2016) 259–269 269

[31] W. Xu, X. Li, J. Cao, H. Zhang, H. Zhang, Sci. Rep. 4 (2014) 4016, http://dx.doi. [64] R.A. Mcaloney, R.A. Mcaloney, M. Sinyor, M. Sinyor, V. Dudnik, V. Dudnik, M.
org/10.1038/srep04016, URL 〈http://www.pubmedcentral.nih.gov/articler C. Goh, M.C. Goh, Langmuir 17 (2001) 6655–6663, http://dx.doi.org/10.1021/
ender.fcgi?artid ¼3915323&tool¼ pmcentrez&rendertype ¼abstract〉. la010136q.
[32] Y.H. Yang, M. Haile, Y.T. Park, F.A. Malek, J.C. Grunlan, Macromolecules 44 [65] K.R. Cooper, J. Electrochem. Soc. 157 (2010) B1731, http://dx.doi.org/10.1149/
(2011) 1450–1459, http://dx.doi.org/10.1021/ma1026127. 1.3481561.
[33] D.A. Hagen, C. Box, S. Greenlee, F. Xiang, O. Regev, J.C. Grunlan, RSC Adv. 4 [66] C. Gardner, A.V. Anantaraman, J. Electroanal. Chem. 449 (1998) 209–214, http:
(2014) 18354, http://dx.doi.org/10.1039/c4ra01621a, URL 〈http://xlink.rsc.org/ //dx.doi.org/10.1016/S0022-0728(97)00408-7.
?DOI ¼c4ra01621a〉. [67] S. Ma, Z. Siroma, H. Tanaka, J. Electrochem. Soc. 153 (2006) A2274–A2281, URL
[34] B.E. Stevens, P.K. Odenborg, M.A. Priolo, J.C. Grunlan, J. Polym. Sci. Part B 〈http://jes.ecsdl.org/content/153/12/A2274.abstract〉.
Polym. Phys. 52 (2014) 1153–1156, http://dx.doi.org/10.1002/polb.23543, URL [68] R.F. Silva, M. De Francesco, A. Pozio, J. Power Sour. 134 (2004) 18–26, http://dx.
〈http://doi.wiley.com/10.1002/polb.23543〉. doi.org/10.1016/j.jpowsour.2004.03.028.
[35] Y.-H. Yang, L. Bolling, M. Haile, J.C. Grunlan, RSC Adv. (2012) 12355–12363. [69] M.W. Verbrugge, J. Electrochem. Soc. 137 (1990) 3770, http://dx.doi.org/
http://dx.doi.org/10.1039/c2ra21845c. 10.1149/1.2086299.
[36] S. Abdu, M.C. Martí-Calatayud, J.E. Wong, M. García-Gabaldón, M. Wessling, [70] Z. Tang, R. Svoboda, J.S. Lawton, D.S. Aaron, A.B. Papandrew, T.A. Zawodzinski,
ACS Appl. Mater. Interfaces 6 (2014) 1843–1854, http://dx.doi.org/10.1021/ J. Electrochem. Soc. 160 (2013) F1040–F1047, http://dx.doi.org/10.1149/
am4048317. 2.083309jes, URL 〈http://jes.ecsdl.org/content/160/9/F1040.abstract〉.
[37] T. Sata, T. Sata, W. Yang, J. Membr. Sci. 206 (2002) 31–60, http://dx.doi.org/ [71] N. Yoshida, T. Ishisaki, A. Watakabe, M. Yoshitake, Electrochim. Acta 43 (1998)
10.1016/S0376-7388(01)00491-4. 3749–3754, http://dx.doi.org/10.1016/S0013-4686(98)00133-9.
[38] F. Guesmi, C. Hannachi, B. Hamrouni, Ionics 18 (2012) 711–717, http://dx.doi. [72] R. Mohr, V. Kudela, J. Schauer, K. Richau, Desalination 147 (2002) 191–196,
org/10.1007/s11581-011-0627-2. http://dx.doi.org/10.1016/S0011-9164(02)00533-7.
[39] J. Ping, J. Wu, X. Luo, Y. Ying, Ionics 17 (2011) 443–449, http://dx.doi.org/ [73] K.D. Kreuer, S.J. Paddison, E. Spohr, M. Schuster, Chem. Rev. 104 (2004)
10.1007/s11581-011-0530-x. 4637–4678, http://dx.doi.org/10.1021/cr020715f.
[40] D.M. Delongchamp, P.T. Hammond, Chem. Mater. 15 (2003) 1165–1173. [74] S. Peighambardoust, S. Rowshanzamir, M. Amjadi, Int. J. Hydrog. Energy 35
[41] Y. Daiko, K. Katagiri, A. Matsuda, Chem. Mater. (2008) 6405–6409. URL 〈http:// (2010) 9349–9384, http://dx.doi.org/10.1016/j.ijhydene.2010.05.017, URL
pubs.acs.org/doi/abs/10.1021/cm8007705〉. 〈http://linkinghub.elsevier.com/retrieve/pii/S0360319910009523〉.
[42] Q.H. Liu, G.M. Grim, A.B. Papandrew, A. Turhan, T.A. Zawodzinski, M.M. Mench, [75] T.D. Gierke, G.E. Munn, F.C. Wilson, J. Polym. Sci. Polym. Phys. Ed. 19 (1981)
J. Electrochem. Soc. 159 (2012) A1246–A1252, http://dx.doi.org/10.1149/ 1687–1704, http://dx.doi.org/10.1002/pol.1981.180191103.
2.051208jes, URL 〈http://jes.ecsdl.org/cgi/doi/10.1149/2.051208jes〉. [76] Y. Zhai, H. Zhang, Y. Zhang, D. Xing, J. Power Sour. 169 (2007) 259–264, http:
[43] D.-J. Guo, S.-J. Fu, W. Tan, Z.-D. Dai, J. Mater. Chem. 20 (2010) 10159, http://dx. //dx.doi.org/10.1016/j.jpowsour.2007.03.004.
doi.org/10.1039/c0jm01161d. [77] R. Bouchet, Solid State Ionics 118 (1999) 287–299, http://dx.doi.org/10.1016/
[44] C. Sun, J. Chen, H. Zhang, X. Han, Q. Luo, J. Power Sour. 195 (2010) 890–897, S0167-2738(98)00466-4.
http://dx.doi.org/10.1016/j.jpowsour.2009.08.041, URL 〈http://linkinghub.else [78] R. Tanaka, H. Yamamoto, A. Shono, K. Kubo, M. Sakurai, Electrochim. Acta 45
vier.com/retrieve/pii/S0378775309014153〉. (2000) 1385–1389, http://dx.doi.org/10.1016/S0013-4686(99)00347-3.
[45] S. Slade, S. Campbell, T. Ralph, F. Walsh, J. Electrochem. Soc. 149 (2002) [79] M. Saito, N. Arimura, K. Hayamizu, T. Okada, J. Phys. Chem. B 108 (2004)
1556–1564, http://dx.doi.org/10.1149/1.1517281. 16064–16070.
[46] V.A. Sethuraman, S. Khan, J.S. Jur, A.T. Haug, J.W. Weidner, Electrochim. Acta [80] L. Krasemann, B. Tieke, Langmuir 16 (2000) 287–290, http://dx.doi.org/
54 (2009) 6850–6860, http://dx.doi.org/10.1016/j.electacta.2009.06.068, URL 10.1021/la991240z.
〈http://linkinghub.elsevier.com/retrieve/pii/S0013468609008950〉. [81] H. Zhang, ECS Trans. 28 (2010) 1–5, http://dx.doi.org/10.1149/1.3492325, URL
[47] Z. Ogumi, Z. Takehara, S. Yoshizawa, J. Electrochem. Soc. 131 (1984) 769–773. 〈http://ecst.ecsdl.org/content/28/22/1.abstract〉.
[48] T. Lehtinen, G. Sundholm, S. Holmberg, F. Sundholm, P. Bjo, Electrochim. Acta [82] J. Zhang, H.A. Gasteiger, W. Gu, J. Electrochem. Soc. 160 (2013) F616–F622,
43 (1998) 1881–1890. http://dx.doi.org/10.1149/2.081306jes, URL 〈http://jes.ecsdl.org/cgi/doi/10.
[49] J.B. Lindén, M. Larsson, B.R. Coad, W.M. Skinner, M. Nydén, RSC Adv. 4 (2014) 1149/2.081306jes〉.
25063, http://dx.doi.org/10.1039/c4ra02223h, URL 〈http://xlink.rsc.org/? [83] H. Ito, T. Maeda, A. Nakano, H. Takenaka, Int. J. Hydrog. Energy 36 (2011)
DOI¼c4ra02223h〉. 10527–10540, http://dx.doi.org/10.1016/j.ijhydene.2011.05.127, URL 〈http://lin
[50] T. Yang, A. Hussain, S. Bai, I.A. Khalil, H. Harashima, F. Ahsan, J. Control. Release kinghub.elsevier.com/retrieve/pii/S0360319911013760〉.
115 (2006) 289–297, http://dx.doi.org/10.1016/j.jconrel.2006.08.015, URL [84] A.T. Haug, R.E. White, J. Electrochem. Soc. 147 (2000) 980–983, http://dx.doi.
〈http://www.sciencedirect.com/science/article/pii/S0168365906004238〉. org/10.1149/1.1393300.
[51] V.D. Noto, R. Gliubizzi, E. Negro, G. Pace, J. Phys. Chem. B 110 (2006) [85] P. Gode, G. Lindbergh, G. Sundholm, J. Electroanal. Chem. 518 (2002) 115–122,
24972–24986. URL 〈http://www.sciencedirect.com/science/article/pii/S0022072801006982〉.
[52] Z. Liang, W. Chen, J. Liu, S. Wang, Z. Zhou, W. Li, G. Sun, Q. Xin, J. Membr. Sci. [86] J.G. Wijmans, R.W. Baker, J. Membr. Sci. 107 (1995) 1–21, http://dx.doi.org/
233 (2004) 39–44, http://dx.doi.org/10.1016/j.memsci.2003.12.008. 10.1016/0376-7388(95)00102-I.
[53] K. Feng, L. Hou, B. Tang, P. Wu, Phys. Chem. Chem. Phys. 17 (2015) 9106–9115, [87] X.-G. Yang, Q. Ye, P. Cheng, T.S. Zhao, Appl. Energy 145 (2015) 306–319, http:
http://dx.doi.org/10.1039/C5CP00203F, URL 〈http://xlink.rsc.org/? //dx.doi.org/10.1016/j.apenergy.2015.02.038, URL 〈http://linkinghub.elsevier.
DOI ¼C5CP00203F〉. com/retrieve/pii/S0306261915002159〉.
[54] L. Grosmaire, S. Castagnoni, P. Huguet, P. Sistat, M. Boucher, P. Bouchard, [88] T. Mohammadi, J. Membr. Sci. 133 (1997) 151–159, URL 〈http://www.science
P. Bébin, S. Deabate, Phys. Chem. Chem. Phys. 10 (2008) 1577–1583, http://dx. direct.com/science/article/pii/S0376738897000926〉.
doi.org/10.1039/b714870d. [89] T.A. Zawodzinski, J. Electrochem. Soc. 140 (1993) 1981, http://dx.doi.org/
[55] F. Zhang, Z. Zhang, Y. Liu, H. Lu, J. Leng, Smart Mater. Struct. 22 (2013) 085020, 10.1149/1.2220749.
http://dx.doi.org/10.1088/0964-1726/22/8/085020, URL 〈http://stacks.iop.org/ [90] S. Cui, S.J. Paddison, J. Phys. Chem. C 119 (23) (2015) 12848–12855. URL 〈http://
0964-1726/22/i ¼8/a ¼ 085020?key¼ crossref. pubs.acs.org/doi/abs/10.1021/acs.jpcc.5b02876〉, http://dx.doi.org/10.1021/acs.
0f9179f29c114794b3e39b1e46fdf74c〉. jpcc.5b02876.
[56] M. Feng, R. Qu, Z. Wei, L. Wang, P. Sun, Z. Wang, Sci. Rep. 5 (2015) 9859, http: [91] D.L. Feldheim, D.R. Lawson, C.R. Martin, J. Polym. Sci. Part B Polym. Phys. 31
//dx.doi.org/10.1038/srep09859, URL 〈http://www.nature.com/doifinder/10. (1993) 953–957, http://dx.doi.org/10.1002/polb.1993.090310805.
1038/srep09859〉. [92] L. Sun, J.S. Thrasher, Polym. Degrad. Stab. 89 (2005) 43–49, http://dx.doi.org/
[57] M. Rikukawa, K. Sanui, Prog. Polym. Sci. 25 (2000) 1463–1502, http://dx.doi. 10.1016/j.polymdegradstab.2005.01.001.
org/10.1016/S0079-6700(00)00032-0. [93] J. Xi, Z. Wu, X. Teng, Y. Zhao, L. Chen, X. Qiu, J. Mater. Chem. 18 (2008) 1232,
[58] S.K. Tiwari, S.K. Nema, Y.K. Agarwal, Thermochim. Acta 317 (1998) 175–182. http://dx.doi.org/10.1039/b718526j, URL 〈http://xlink.rsc.org/?
[59] L. Zhang, Y. Li, J.C. Yu, Y.Y. Chen, K.M. Chan, J. Mater. Chem. B 2 (2014) DOI¼ b718526j〉.
7936–7944, http://dx.doi.org/10.1039/C4TB01577K, URL 〈http://xlink.rsc.org/? [94] S. Yilmaztürk, N. Ercan, H. Deligöz, Appl. Surf. Sci. 258 (2012) 3139–3146, http:
DOI ¼C4TB01577K〉. //dx.doi.org/10.1016/j.apsusc.2011.11.051, URL 〈http://linkinghub.elsevier.com/
[60] G. Lawson, F. Gonzaga, J. Huang, G. de Silveira, M.A. Brook, A. Adronov, J. retrieve/pii/S0169433211018022〉.
Mater. Chem. 18 (2008) 1694, http://dx.doi.org/10.1039/b715277a. [95] F. Harnisch, U. Schröder, ChemSusChem 2 (2009) 921–926, http://dx.doi.org/
[61] J. Fu, J. Ji, L. Shen, A. Ku, A. Rosenhahn, J. Shen, M. Grunze, Langmuir 25 (2009) 10.1002/cssc.200900111, URL 〈http://doi.wiley.com/10.1002/cssc.200900111〉.
672–675, http://dx.doi.org/10.1021/la803692v. [96] A. Yamauchi, T. Ito, T. Yamaguchi, J. Power Sour. 174 (2007) 170–175, http://dx.
[62] C. Porcel, P. Lavalle, V. Ball, G. Decher, B. Senger, J.C. Voegel, P. Schaaf, Lang- doi.org/10.1016/j.jpowsour.2007.08.081, URL 〈http://linkinghub.elsevier.com/
muir 22 (2006) 4376–4383, http://dx.doi.org/10.1021/la053218d. retrieve/pii/S0378775307017533〉.
[63] I. Choi, R. Suntivich, F.A. Plamper, C.V. Synatschke, H.E.M. Axel, V.V. Tsukruk, J. [97] K. Ariga, J.P. Hill, Q. Ji, Phys. Chem. Chem. Phys. 9 (2007) 2319, http://dx.doi.
Am. Chem. Soc. 133 (2011) 9592–9606. org/10.1039/b700410a, URL 〈http://xlink.rsc.org/?DOI ¼ b700410a〉.
Erklärung I
Die Richtigkeit der Darlegung des Eigenanteils an den Publikationen (“Author
Contributions”) wird hiermit vom Betreuer der Arbeit bestätigt.

Oldenburg, den ___________ ___________________________

(Prof. Dr. Gunther Wittstock)

89
Erklärung II
Hiermit erkläre ich, Jan Bernhard grosse Austing, geb. 12.06.1983 in Lohne (Oldb.),
dass ich die Dissertation selbständig verfasst und die benutzten Hilfsmittel vollständig
angegeben habe. Die vorliegende Arbeit wurde in Teilen bereits veröffentlicht. Diese
Teile sind die im Kapitel “Publikationen” aufgeführten Veröffentlichungen. Eine
Publikationsliste ist der Dissertation beigefügt. Ich erkläre, dass die Dissertation
weder in ihrer Gesamtheit noch in Teilen einer anderen wissenschaftlichen Hoch-
schule zur Begutachtung in einem Promotionsverfahren vorliegt bzw. vorlag. Ich
erkläre weiterhin, dass die Leitlinien guter wissenschaftlicher Praxis an der Carl von
Ossietzky Universität Oldenburg befolgt wurden sind und im Zusammenhang mit
dem Promotionsvorhaben keine kommerziellen Vermittlungs- oder Beratungsdienste
(Promotionsberatung) in Anspruch genommen wurden.

Oldenburg, den _________ ____________________________

(Jan grosse Austing)

91
List of Publications

Printed Publications in Peer-Reviewed Journals

1. Jan grosse Austing, Carolina Nunes Kirchner, Eva-Maria Hammer, Lidiya
Komsiyska, Gunther Wittstock,
“Study of an unitised bidirectional vanadium/air redox flow battery comprising
a two-layered cathode”,
Journal of Power Sources, Volume 273, Pages 1163–1170,
DOI: 10.1016/j.jpowsour.2014.09.177.

2. Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Gunther

Wittstock,
“Investigation of crossover processes in a unitized bidirectional vanadium/air
redox flow battery”,
Journal of Power Sources, Volume 306, Pages 692-701,
DOI: 10.1016/j.jpowsour.2015.12.052.

3. Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Gunther

Wittstock,
“Layer-by-layer modification of Nafion membranes for increased life-time and
efficiency of vanadium/air redox flow batteries”,
Journal of Membrane Science, Volume 510, Pages 259-269,
DOI: 10.1016/j.memsci.2016.03.005.

Conference Proceedings
1. Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Gunther
Wittstock,
"Performance of a vanadium/air redox flow battery (VARFB) comprising a
two-layered cathode",
Printed Conference Mini Papers of the International Flow Battery Forum
(IFBF) 2014, ISBN: 978-0-9571055-4-6, Pages 10/11.

93
 List of Publications

2. Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Oliver Osters,
Gunther Wittstock,
"Investigation of crossover processes in a bidirectional vanadium/air redox flow
battery",
Printed Conference Mini Papers of the International Flow Battery Forum
(IFBF) 2015, ISBN: 978-0-9571055-5-3, Pages 18/19.

Oral Conference Contributions

• Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Gunther
Wittstock,
"Performance of a vanadium/air redox flow battery (VARFB) comprising a
two-layered cathode",
International Flow Battery Forum (IFBF) 2014, Hamburg, July 2014.

• Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Gunther

Wittstock,
"Performance of a vanadium/air redox flow battery (VARFB) comprising a
two-layered cathode",
Electrochemistry 2014, Mainz, September 2014.

• Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Oliver Osters,
Gunther Wittstock,
"Investigation of crossover processes in a bidirectional vanadium/air redox flow
battery",
International Flow Battery Forum (IFBF) 2015, Glasgow, June 2015.

• Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Gunther

Wittstock,
"Investigation of crossover processes in a bidirectional vanadium/air redox flow
battery",
ECHEMS - Electrochemistry in renewable energy based on molecular mecha-
nisms, Bad Zwischenahn, June 2015.

Poster Conference Contributions

• Jan grosse Austing, Udo Martin, Andreas Blömer, Lidiya Komsiyska, Eva-
Maria Hammer,
“Influence of the membrane electrode assembly preparation procedure on the

94
List of Publications

discharge performance of Vanadium-Air-Redox-Flow-Batteries (VARFB)”,

International Flow Battery Forum (IFBF) 2012, Munich, June 2012.

• Jan grosse Austing, Eva-Maria Hammer, Lidiya Komsiyska, Carolina Nunes-

Kirchner, Gunther Wittstock,
“Preparation and Characterization of Cathodes for Vanadium/Air Redox Flow
Batteries (VARFB) by Metal Deposition on Carbon-based 3D Electrodes”,
International Flow Battery Forum (IFBF) 2013, Dublin, June 2013.

• Jan grosse Austing, Carolina Nunes Kirchner, Lidiya Komsiyska, Gunther

Wittstock,
"Performance of a vanadium/air redox flow battery (VARFB) comprising a
two-layered cathode",
Kraftwerk Batterie, Münster, March 2014.

• Timo di Nardo, Carolina Nunes Kirchner, Jan grosse Austing, Oliver Osters;
Lidiya Komsiyska,
"Electrochemical deposition of Ir on graphite felt electrode for vanadium air
redox flow battery cathodes",
CIMTEC, Montecatini Terme, June 2014.

• Timo di Nardo, Carolina Nunes Kirchner, Jan grosse Austing, Oliver Osters;
Lidiya Komsiyska,
"Electrochemical deposition of iridium on graphite felt electrode for vanadium
air redox flow battery cathodes",
13th International Fischer Symposium, Lübeck, June 2015.

95
Jan Bernhard
grosse Austing
born on: 12/06/1983
Curriculum Vitae in: Lohne (Oldb.)

Education & Experiences

07/1996 - 06/2003 Abitur, Gymnasium Damme, Germany, grade: 1.3.
10/2003 - 04/2009 Mathematics and Chemistry, 1. Staatsexamen,
Philipps University of Marburg, Germany, grade: 1.29.
09/2009 - 09/2010 Voluntary Service „weltwärts”, Lwandai Secondary
School, Mlalo, Tanzania.
10/2010 - 06/2012 Renewable Energies and Energy Efficiency, M.Sc.,
University of Kassel, Germany, grade: 1.42.
10/2011 - 04/2012 Master Thesis, NEXT ENERGY - EWE Research
Centre for Energy Technology, Oldenburg, Germany.
11/2012 - 09/2015 PhD Studies, NEXT ENERGY - EWE Research Cen-
tre for Energy Technology, Oldenburg, Germany.
08/2013 - 09/2013 Research Stay, University of Victoria, British
Columbia, Canada.
since 10/2015 Project Manager, EXIST Forschungstransfer “Resi-
Flow”, NEXT ENERGY - EWE Research Centre for
Energy Technology, Oldenburg, Germany.

Awards & Grants

10/2012 Financial Support for PhD Project (5000 €),
Nord-West-Metall Stiftung.
11/2012 PhD Scholarship (3 years), Reiner-Lemoine Stiftung.
04/2013 Award for Environmental Related Master Thesis
(1500 €), VDI-Gesellschaft für Energie und Umwelt
(VDI-GEU).
Danksagung
Im Folgenden möchte ich mich bei all denjenigen bedanken, die direkt oder indi-
rekt zum Gelingen dieser Dissertation beitrugen. Beginnen möchte ich mit Prof.
Dr. Gunther Wittstock, der die wissenschaftliche Betreuung dieser Arbeit und die
Begutachtung als Erstgutachter übernommen hat. Er nahm sich immer Zeit für
fachliche und organisatorische Fragen, auch kurzfristig. Ebenso danke ich Prof. Dr.
Carsten Agert für die Übernahme der Zweitbegutachtung.
Für den intensiven, anregenden fachlichen Austausch bei der täglichen Arbeit gilt
mein Dank besonders Dr. Lidiya Komsiyska, Dr. Carolina Nunes Kirchner und Dr.
Eva-Maria Hammer. Sie haben sich immer Zeit genommen für meine Anliegen und
ich habe ihre fachlichen Fähigkeiten aber auch persönlichen Eigenschaften sehr zu
schätzen gelernt. Ich danke auch allen anderen Kollegen bei NEXT ENERGY für
die hilfsbereite, kollegiale Zusammenarbeit bei vielerlei Fragestellungen. Namentlich
nennen möchte ich Hayo Seeba, Frank Bättermann, Dietmar Piehler, Timo di Nar-
do, Pratik Das, Barbara Satola, Khrystyna Yezerska, Benedikt Berger und Dana
Schonvogel.
Ich danke meiner Frau Karina dafür, dass Sie mich immer in meinen Vorhaben
immer so aktiv unterstützt. Bei meiner Tochter Thea möchte ich mich dafür bedanken,
dass sie insbesondere während der Schreibphase auf meine Anwesenheit als Vater
ab und zu verzichten konnte. Meinem Bruder Arne gilt mein Dank für Hinweise
und Korrekturvorschläge zur Arbeit. Meinen Eltern danke ich für die gesamte
Unterstützung während meines bisherigen Lebens.
Der Reiner-Lemoine Stiftung möchte ich für das Promotionsstipendium danken,
durch welches das Arbeiten an diesem Thema im Rahmen einer Doktorarbeit ermög-
licht wurde.

99