Journal of Power Sources 380 (2018) 105–114

Contents lists available at ScienceDirect

Journal of Power Sources

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

Performance evaluation of thermally treated graphite felt electrodes for T

vanadium redox flow battery and their four-point single cell
characterization
P. Mazúra,∗, J. Mrlíkb, J. Beneša, J. Pocediča, J. Vránab, J. Dundálekb, J. Koseka,b
a
University of West Bohemia, Research Centre - New Technologies, Univerzitni 8, 306 14 Pilsen, Czech Republic
b
University of Chemistry and Technology Prague, Dept. of Chemical Engineering, Technicka 5, 166 28 Prague, Czech Republic

H I G H L I G H T S

• Effect of thermal treatment on graphite felt electrode performance was investigated.

• Results of cyclovoltammetry do not correlate with single-cell testing for cathode.
• Complex 4-point method for single-cell characterization was developed.
• Distribution of ohmic and faradaic losses in single-cell was evaluated.
• Kinetics of cathode reaction was found unaffected by oxygen functionalization.

A R T I C L E I N F O A B S T R A C T

Keywords: In our contribution we study the electrocatalytic effect of oxygen functionalization of thermally treated graphite
Vanadium redox flow battery felt on kinetics of electrode reactions of vanadium redox flow battery. Chemical and morphological changes of
Graphite felt the felts are analysed by standard physico-chemical characterization techniques. A complex method four-point
Oxygen functionalities method is developed and employed for characterization of the felts in a laboratory single-cell. The method is
Electrochemical impedance spectroscopy
based on electrochemical impedance spectroscopy and load curves measurements of positive and negative half-
Load curve
cells using platinum wire pseudo-reference electrodes. The distribution of ohmic and faradaic losses within a
Cyclic voltammetry
single-cell is evaluated for both symmetric and asymmetric electrode set-up with respect to the treatment
conditions. Positive effect of oxygen functionalization is observed only for negative electrode, whereas kinetics
of positive electrode reaction is almost unaffected by the treatment. This is in a contradiction to the results of
typically employed cyclovoltammetric characterization which indicate that both electrodes are enhanced by the
treatment to a similar extent. The developed four-point characterization method can be further used e.g., for the
component screening and in-situ durability studies on single-cell scale redox flow batteries of various chemistries.

1. Introduction concept is schematically described in Supplementary data 1. The flow

The actual growth in the electricity production from the renewable
energy sources operating in the intermittent regime (such as wind
turbines and photovoltaic cells) emphasizes the need for flexible, cheap
and reliable technology for large scale energy storage. Among various
energy storage options the vanadium redox flow battery (VRFB) re-
presents an interesting alternative. Comparing to classical static accu-
mulators, the energy is stored in external tanks in the form of aqueous
vanadium-based electrolytes which are circulated through the battery
stack, where the energy conversion is associated with the electro-
chemical reactions of vanadium ions on inert electrodes. The VRFB
concept provides the battery with several advantageous features such
as: i) decoupled power (kW) and capacity (kWh); ii) easy removal of
waste heat; and iii) reduced self-discharge. Moreover, superior cy-
clability (over 10 000 cycles) is expected due to the absence of phase
changes on the electrodes (e.g., metal deposition, gas evolution or ion
intercalation) within the battery operation and a single electro-
chemically active element used in both, positive and negative, elec-
trolytes. These features, together with reasonable efficiency of energy
conversion (above 80% DC-DC) and fast response time (in tens of ms),
predetermine the technology for stationary energy storage applications
[1,2].

∗
Corresponding author.
E-mail address: mazurp@ntc.zcu.cz (P. Mazúr).

https://doi.org/10.1016/j.jpowsour.2018.01.079
Received 30 November 2017; Received in revised form 24 January 2018; Accepted 28 January 2018
0378-7753/ © 2018 Elsevier B.V. All rights reserved.
P. Mazúr et al.
The VRFB electrode reactions, together with the corresponding
standard reduction potentials, are presented in the direction of reduc-
tion in eqs. (1) and (2).
VO2+ + 2H+ + e− ⇆ VO 2 + + H2 O ∘
Epos = 1.000 V (1)
V 3 + + e− ⇆ V 2 + ∘
Eneg = −0.255 V
(2)
The sluggish reaction kinetic of typically used graphite felt elec-
trodes towards the VRFB electrode reactions is usually addressed by the
heat treatment of the felts in the presence of air. The partial oxidation of
the surface of graphite fibres by the oxygen contained in air increases
the wettability of the originally rather hydrophobic graphitic material,
which, in turn, increases the electrode-electrolyte interface area. At the
same time, oxygen functional groups created on the fibre surface are
believed to catalyse the VRFB electrochemical reactions [3]. The three-
step reaction mechanism is usually used to describe the positive (eqs.
(3)–(5)) and the negative (eqs. (6)–(8)) charging electrode reaction at
hydroxylated carbon electrode [3]:
(3)
R − OH + VO 2 + → R − O − V + = O + H+
R−O− V+ = O + H2 O → R − O − V (=O )2 + 2H+ + e− (4)
R − O − V (=O)2 + H+ → R − OH + VO2+ (5)
(6)
R − OH + V 3+ →R−O− V 2+ + H+
R − O − V 2 + + e− → R − O − V + (7)
R − O − V + + H+ → R − OH + V 2 + (8)
The specific optimal conditions of thermal activation are known to
be strongly dependent on the overall history of the carbon electrode
material including: choice of fibre precursor (rayon or poly-
acrylonitrile) [4], texture properties of the porous matrix (carbon felt or
paper) [5], conditions of carbonization and graphitization [6] as well as
the content of impurities [7]. Besides the apparent differences between
various porous carbon materials, Rabbow et al. recently draw attention
to the poor reproducibility of the thermal activation of the PAN-based
felts [8,9]. For different charges of the same felt material significantly
different effect of the same thermal activation procedure was observed
with respect to mass losses, specific capacitance and content of oxygen
functional groups. The oxidative effect of the thermal treatment can be
also enhanced by the increase of oxygen concentration in the gas
mixture [10], resulting in a significant improvement of performance of
graphite paper electrodes as a consequence of the increased roughness
of the fibres surface.
The electrochemical activity of carbon materials is known to be
strongly related to the microstructure of the surface layer, more spe-
cifically to the content of the carbon edge sites and defects. According
to Chu et al. [7] the gas-phase treatment of graphitic material increases
the content of edge sites due to etching of the basal plane skin layer.
According to Ehrburger et al. [6] the sensitivity of carbon materials
towards the oxidation can be related to the so-called active surface area
(ASA), which express the content of the catalytically active edge sites
and defects of graphene sheets. This can be experimentally quantified
by oxygen chemisorption followed by the measurements of carbon
oxides evolved during the subsequent degassing at elevated tempera-
tures.
Despite two decades which have passed since Sun et al. firstly re-
ported on the thermal treatment of graphite felts for VRFB [11], the role
of the oxygen functionalities on the electrochemical performance of the
felt electrodes still seems to be rather controversial. The positive effect
of the oxygen functionalities on the kinetics of the battery electrode
reactions has been repeatedly reported in the past [4], [12–14]. How-
ever, contrary to that, the inhibiting effect of electrode surface oxida-
tion on positive VRFB electrode reaction has been recently indicated by
several papers [15–17]. Bourke et al. reported the negative effect of
anodic polarization on the various amorphous and crystalline carbon
106
Journal of Power Sources 380 (2018) 105–114
materials (including graphite fibres) on the kinetics of positive VRFB
electrode reaction. Similar trends for the VO2+/VO2+ redox couples
were observed by Friedl et al. who tested the effect of chemical func-
tionalization of multi-walled carbon nanotubes [16]. Also Melke et al.
reported negative effect of thermal treatment on electrode reaction
kinetics for less graphitized carbons due to decreased content of sp2
carbon atoms in the surface layer and related increased ohmic losses
[17,18].
The discrepancy between these reports may originate in in-
appropriate experimental approaches which are typically employed for
the evaluation of electrocatalytic activity of carbon-based electrodes
towards the battery electrode reactions. Most of the publications in the
field employ cyclic voltammetry (CV) measured in diluted vanadium
electrolytes (as it is comprehensively reviewed in Ref. [19]), with the
evaluation based on to the separation of oxidation and reduction peak
potential. This is typically followed by a single-cell characterization in
symmetric electrode set-up, i.e., with both positive and negative elec-
trode being treated equally.
In our contribution, the relevancy of such characterization approach
is critically assessed on a series of thermally treated graphite felt elec-
trodes where the discrepancy between the results of CV and symmetric
single-cell characterization is observed. Based on the results we propose
a complex single-cell characterization method combining electro-
chemical impedance spectroscopy (EIS) and load curve (LC) measure-
ments in 4-electrode set-up. The distribution of ohmic and faradaic
losses of individual cell components within the single-cell is directly
evaluated from these measurements. This enables us to study the effect
of electrode functionalization of positive and negative electrode in-
dependently.
2. Experimental part
2.1. Thermal modification of graphite felts
Rayon-based graphite felt with the thickness of 5.0 mm and fibre
diameter of approx. 10 μm was used within the study. The thermal
treatment was conducted in electrical furnace under various tempera-
tures (400, 450, 500, 550 and 600 °C) for 9 h. The temperature was
ramped with the 10 °C min−1 rate to a given temperature and after 9 h
time period, it cooled down to room temperature spontaneously. The
weight of felt before and after the treatment was measured using ana-
lytical laboratory scales.
2.2. Physical characterization of the felt
X-ray photoelectron spectroscope (XPS) ESCAProbeP (Omicron
Nanotechnology) and confocal DXR Raman microscope (Thermo-
Fischer Scientific) were used to observe the changes of surface com-
position of the felt due to thermal treatment. Scanning electron mi-
croscope (SEM) VEGA3 (Tescan) was used to observe the morphological
changes of the graphite fibres. Thermogravimetric analysis (TGA) using
thermogravimetre Q 500 (TA Instruments) was used to characterize
thermal stability of the felt under nitrogen, air and oxygen atmosphere.
The specific surface area (SSA) of felts was measured by nitrogen ad-
sorption at −196.15 °C using 3Flex physisorption instrument
(Micromeritics) and calculated from BET isotherms. Prior to the mea-
surement, the sample was desorbed for 1000 min in vacuum at 105 °C.
The wettability of the felts was characterized by dropping of demi-
neralized water on to the felt sample and classified in terms of wett-
ability degrees, according to Rabbow et al. [8].
Through-plane area specific resistance (ASR) of graphite felt was
measured by ohmmeter BS407 (Aim-TTi Instruments) in 4-probe setup
with a measured felt sample surrounded by two pieces of commercial
composite plate PPG86 (Eisenhuth). The felt resistance was measured at
25% compression of its initial thickness, which corresponds to the
compression in our laboratory single-cell.
P. Mazúr et al.
2.3. Cyclovoltammetric characterization in diluted electrolyte
The cyclovoltammetric characterization was performed in 3-elec-
trode arrangement using the in-house designed and constructed cell.
The working electrode assembly consisted of the felt sample (1.0 cm2
active area) supported by a conductive composite plate PPG86
(Eisenhuth). A mercurous sulphate reference electrode (MSRE) with
saturated K2SO4 inner electrolyte with the potential 0.65 V vs. standard
hydrogen electrode was used within the study (all potentials are re-
ferred to MSRE, unless stated otherwise). The platinized titanium sheet
(active area of 6 cm2) was used as a counter electrode. The working and
counter electrode compartments were mutually separated by per-
fluorinated cation-exchange membrane (Fumatech).
The specific electric double layer capacitance (EDLC) of the felts
was obtained from CV in the 1 mol dm−3 H2SO4 in potential range from
−0.5 V to 0.5 V and 20 mV s−1 scan rate. The EDLC of the electrode
was calculated from charge ΔQ integrated in potential window ΔE
(–0.4 V–0.4 V) and weigh of the sample m using equation (9).
E2
ΔQ ∫E1 I dt
EDLC = =
ΔE m (E2 − E1) m (9)
The kinetics of electrodes towards the reactions of vanadium redox
couples was studied by CV in 0.05 mol dm−3 solution of corresponding
vanadium ions (V4+ for positive and V3+ for negative electrode reac-
tion, respectively) in 1 mol dm−3 H2SO4 supporting electrolyte. The
electrolytes were prepared from VOSO4 (97% purity by Sigma-Aldrich),
sulphuric acid (p.a. grade by Penta) and demineralised water. The V3+
electrolyte was prepared by electrochemical reduction in a flow battery
cell and the UV-VIS spectroscopy was used to determine its full con-
version to V3+. Prior to the experiments, the negative electrolyte was
deaerated by nitrogen purging for 30 min. Potential was swept at
2 mV s−1 scan rate and the potential limits were adjusted ad hoc to
minimise the contribution of the parasitic water splitting reactions. The
electrochemical measurements were carried out using potentiostat/
galvanostat/FRA EC-LAB SP-300 (Biologic).
2.4. Single-cell characterization – 4-point measurements
The felts were subsequently characterized in a laboratory VRFB
single-cell of own design and construction. The battery body was
manufactured from polyvinylchloride and sealed by elastomeric gas-
kets. The graphite felt electrodes (dimensions of 40 × 50 mm and 25%
compression of the initial thickness of 5.0 mm) were supported by
carbon composite plates PPG86 (Eisenhuth). The electrode compart-
ments were mutually separated by PFSA type cation-exchange mem-
brane (Fumatech). Four high purity platinum wires were positioned in
the cell according to scheme depicted on Fig. 1. SM+ and SM- were
placed in the electrolyte inlet channels from both sides of membrane in
the proximity of graphite electrodes (without touching them) and they
served as pseudo-reference electrodes. SP+ and SP- were connected to
the inner side of carbon composite plates. Such arrangement enabled
the direct measurement of ohmic and faradaic losses of the individual
components of the cell, according to the equivalent circuit presented in
Fig. 1. The method is into more details described in our related Meth-
odsX article titled “A complex four-point method for the evaluation of
ohmic and faradaic losses within a redox flow battery single-cell”.
VRFB single-cell was characterized by EIS, LC measurements and
galvanostatic charge-discharge cycles. All the measurements were
conducted under electrolyte flow rate of 40 cm3 min−1, which corre-
sponds to the linear flow velocity of approx. 27 cm min−1. Commercial
electrolyte comprising 1.6 mol dm−3 of vanadium ions (1:1 M ratio of
V3+:V4+), 2 mol dm−3 of H2SO4 and 0.3% of H3PO4 (purchased from
GfE, Gesellschaft für Elektrometallurgie mbH) was used as the initial
catholyte and anolyte (50 ml for each electrolyte circuit which corre-
spond to the battery capacity of 2144 mAh). Nitrogen gas was bubbled
107
Journal of Power Sources 380 (2018) 105–114
through the negative electrolyte to prevent its self-discharge due to the
oxidation of V(II) ions by oxygen from air. Electrolytes were charged by
constant current density of 50 mA cm−2 to +50% state-of-charge (SOC)
(defined by equimolar ratios of V4+:V5+ and V2+:V3+ in catholyte and
anolyte, respectively) and the EIS and LC measurements were per-
formed. EIS with 5 mV potential amplitude was used in the frequency
range from 200 kHz to 50 mHz. The measured EIS spectra were fitted to
appropriate equivalent circuit using EC-Lab software tool. SM+ and SM-
sense electrodes were used to measure the spectra of individual half-
cells in three-electrode arrangements. LC measurements were realised
by linearly increasing current scans from 0 to +4 A (charging) and from
0 to −4 A (discharging) with 20 mA s−1 scan rate. The slopes of charge
and discharge U-j curves were taken as the value of cell resistance under
load. The half-cell LCs were obtained in three-electrode set-up using
platinum wire pseudo-reference electrodes to evaluate the resistances of
individual cell components under current load.
Finally, the battery was characterized by constant current charge-
discharge cycles at various constant current densities. Voltage limits
1.65 V for charging and 0.8 V for discharging were used to prevent the
presence of parasitic electrode reactions and battery components de-
gradation. The efficiencies of charge-discharge cycle were evaluated
using eqs. (10)–(12). Coulombic efficiency ηC is defined as a ratio of
discharge capacity Qd to charge capacity Qc of the cycle
Qd
ηC =
Qc (10)
Voltage efficiency ηV is given as ratio of average battery voltage
during discharge Ud to average voltage during charge Uc
Ud
ηV =
Uc (11)
Energetic efficiency ηE is a ratio of energy released during discharge
Ed to the energy consumed during charge Ec
Ed
ηE = = ηV ηC
Ec (12)
Battery testing was carried out with Potentiostat/galvanostat/FRA
EC-LAB VSP-126 (Biologic) with 4 A current booster. Whole apparatus
was placed into the thermally insulated box tempered at 20 °C.
3. Results and discussion
3.1. Physicochemical characterization of graphite felts
The surface of graphite felt fibres was functionalised by thermal
treatment, which is a standard modification method used in VRFB in-
dustrial production [8]. Dry pieces of the graphite felt were treated at
temperatures of 400, 450, 500, 550 and 600 °C for 9 h. The temperature
range was chosen based on the results of preliminary TGA character-
ization of the untreated dry felt sample, see Supplementary data 2,
where the significant weight loss of the felt we observed at tempera-
tures around 600 °C in air and, more intensively, in pure oxygen at-
mosphere. As such behaviour was not observed in nitrogen atmosphere,
the mass loss can be clearly attributed to the oxidation of the felt.
Thermal treatment resulted in optically visible changes of the felt, as
shown on photographs (see Supplementary data 3), colour of the felt
became darker with rising temperature – and at 550 °C several small
pits in felt were observed that developed at 600 °C to massive de-
struction of the felt. This correspond well with noticeable weight losses
observed at these temperatures: approx. 20 wt.% for 550 °C and 60 wt.
% for 600 °C, see Fig. 2a. Besides the macroscopic changes of the felt the
surface morphology of graphite fibres was also significantly influenced
by treatment conditions; see SEM images on Supplementary data 3.
When compared to originally smooth and defect-less untreated fibre,
the fibre surface becomes rough at 450 °C and several micron-size pits
appears at 500 °C. At 600 °C, apparent degradation of the fibre is
P. Mazúr et al.
Fig. 1. Scheme of VRFB single-cell with indication of positions of platinum wires. 1 – current collectors, 2 – composite plates, 3 ± – positive/negative felt electrode, 4 – membrane, SP ± –
platinum wires attached to composite plates, SM ± – platinum wires placed in electrolyte inlets.
Fig. 2. a) Dependence of electrode relative weight, specific surface area and electrical
double-layer capacitance on thermal treatment temperature; b) The ID/IG ratio obtained
from Raman spectra and oxygen surface coverage evaluated from XPS spectra as a
function of the thermal treatment temperature.
observed together with presence of larger holes.
The changes of wettability of the felts were observed by applying
drops of demineralized water onto the surface of the felts by dropper. In
accordance to our assumptions, thermal treatment increased the
108
Journal of Power Sources 380 (2018) 105–114
wettability, however not monotonically: the felts treated under 400 °C
and 600 °C showed no affinity to absorb water droplets (W0), the felts
treated at 450 °C and 550 °C absorbed the droplets in about a minute
(W1), whereas highly wettable 500 °C felt absorbed the droplet im-
mediately (W2). Such a behaviour was not observed in the related
publication of Rabbow et al. [8], however they studied different felt
(PAN-based) and used shorter treatment time. More precise quantifi-
cation of the wettability, e.g., by measurement of droplet-felt contact
angles, was disabled due to non-homogeneity of the felt surface.
The changes in surface morphology and wettability of graphite fibre
resulted in significant increase of electrode capacitance (EDLC), which
was evaluated from CV curves taken in 1 M sulphuric acid solution, see
Fig. 2a. The original curves are presented in Supplementary data 4.
Assuming purely capacitive behaviour of electrode-electrolyte inter-
face, the double-layer charging current is proportional to the size of the
interface [20]. The deviation of the anticipated rectangular shape of the
CV curve was caused by surface redox processes of quinone functional
groups present on the fibre surface. These are represented on CV curve
as a couple of broad oxidation and reduction peaks in the potential
range from −0.4 to 0.2 V, which is in good agreement with the pre-
vious observations [9,21]. The EDLC of the felt electrodes increases
with the treatment temperature up to 500 °C where the highest capa-
citive current was observed indicating on the largest electrode-elec-
trolyte interface. As illustrated in Fig. 2a, the trend of EDLC correlates
well with that of SSA obtained by BET analysis of nitrogen physisorp-
tion data; maximum of both was found for 500 °C. The increase of
electrode-electrolyte interface of the thermally activated felt thus seems
to be tightly related to the increase of micro-roughness on the fibre
surface induced by the treatment, which was also confirmed by our
SEM observations. The decrease of EDLC and SSA at 600 °C is most
probably due to the partial destruction of the graphite fibres by the
oxidation. The EDLC observed for treated felts were significantly higher
than reported by Rabbow [8] which indicates higher electrocatalytic
activity of our rayon-based felt.
The changes of the surface composition caused by the treatment
were studied by confocal Raman microscopy. The obtained Raman
spectra, which are for all the felts presented in Supplementary data 5,
contain a pair of characteristic D- and G-bands in the region around
1330 and 1580 cm−1, respectively. The G-band is known to be ex-
clusively presented in the first-order spectrum of perfect graphite and
corresponds to sp2 hybridized carbons, whereas D-band is related to sp3
hybridized carbon atoms [22–25]. The shoulder peak on G-band with
the position of approx. 1620 cm−1, typically referred to as D′-band, is
P. Mazúr et al. Journal of Power Sources 380 (2018) 105–114

typically presented, together with D-band, in the spectra of carbon
materials with lower structural order [26,27]. The dependence of ID/ IG
ratio on temperature of thermal treatment is graphically presented on
Fig. 2b. The ratio of sp3/sp2 hybridized carbon atoms on the fibre
surface increases due to the treatment most likely due to the oxidation
of basal-plane sp2-hybridized carbon atoms by atmospheric oxygen. At
treatment temperatures higher than 500 °C, however, the ID/ IG ratio
decreased which can be attributed to the oxidative removal/gasification
of the surface layer with higher representation of defects and edge-site
carbons.
To investigate the effect of thermal treatment on chemical compo-
sition of the surface layer of graphite fibres, the felts were characterised
by XPS. The C1s details of the studied felts are presented on
Supplementary data 6. Each detail was deconvoluted into two carbon
peaks in sp2 (284.8 eV) and sp3 (285.6 eV) hybridization forms and a
single peak corresponding to hydroxyl group (286.5 eV). Other types of
oxygen functionalities with higher binding energies, such as carbonyl
and carboxyl groups, were not present in evaluable quantities. This is
slightly surprising as the presence of C=O groups on thermally treated
carbon felts has been previously reported [4,28]. The detailed results of
XPS characterization are summarized in Table 1. The oxygen content
increased due to thermal treatment, see Fig. 2b, however, similarly to
the results of Raman and wettability characterization, the observed
trend is not monotonous with the maximum oxygen coverage at 500 °C.
Also here, the decrease of the oxygen on the surface of the fibres at
higher temperatures can be related to the oxidative removal of the
surface layer. The significantly increased surface content of hydroxyl
groups, which is observed between 450 and 500 °C, is accompanied by
the decrease of sp3-hybridized carbon atoms. This corresponds to the
presumption that edge-plane carbon sites react more readily with at-
mospheric oxygen during the thermal treatment when compared to
basal-planes [6]. The slight decrease of sp2-hybridized carbon atoms
with increased treatment temperature is most probably caused by
partial destruction of basal-plane surface layers, which had been re-
ported in the literature previously [9,18].
The results of through-plane ASR measurements of thermally
treated graphite felt are summarized in Table 1. The untreated felt
provided the ASR of 0.33 Ω cm2 and it only slightly increased to
0.36 Ω cm2 at 500 °C, whereas the significant resistivity increase was
observed at 550 °C (0.46 Ω cm2) and 600 °C (1.00 Ω cm2) due to the
oxidative destruction of the felt. This is consistent with our visual ob-
servations and weight loss measurements.
The untreated graphite felt provided us with the highest peak potential
separation ΔEp of more than 600 mV indicating the electrochemical
irreversibility of the electrode reaction [20]. Thermal treatment of the
felt resulted in significant enhancement of the reaction kinetics, ΔEp
dropped below 0.35 V for felt treated at 500 °C, which exhibited the
lowest ΔEp value within the series, see Fig. 3b. The highest improve-
ment of the electrochemical reversibility was observed between 450
and 500 °C, which correlates well with the previous results of physi-
cochemical characterizations. Thus, the enhancement of electrode
performance seems to be tightly connected with the increase of the
electrode-electrolyte interface and presence of hydroxyl functional
groups on the surface of graphite fibre. The performance of graphite felt
electrode deteriorated when treated at temperatures higher than
500 °C, most probably due to the degradation of the felt and related
increased resistance of electrode, as discussed above.
Analogically, performance of the felts towards the negative elec-
trode reaction was examined in 0.05 mol dm−3 V3+ solution in
1 mol dm−3 H2SO4 supporting electrolyte. Measured CVs are on Fig. 3b.
Also in this case, the ΔEp decreased significantly due to the treatment
and the lowest peak potential separation was found also for 500 °C.
However, the differences between the curves are less pronounced which
indicates on the lower sensitivity of the reaction kinetics to the presence
of oxygen functional groups.
Based on the CV results, both positive and negative VRFB electrode
reactions appear to be enhanced by the thermal treatment to a similar
extent with the optimal treatment temperature of 500 °C. The sig-
nificant improvement of electro-catalytic activity can be attributed to
the combination of increased electrode-electrolyte interface (as in-
dicated by results of BET analysis and EDLC measurements) and in-
creased surface coverage by oxygen functionalities. At the same time,
the treatment at 500 °C does not deteriorate the electric conductivity of
the felt, the felt degradation takes place at higher temperatures. The
observed results are in accordance with previous reports
[10,11,30–32].
3.3. Symmetric single-cell characterization
Thermally treated felt electrodes were subsequently tested in a la-
boratory single-cell using standard full-cell measurements in symmetric
set-up with the respect to electrode treatment conditions. The felt
treated at 600 °C was excluded from these measurements due to sever
degradation of conductivity and mechanical properties.

3.2. Cyclovoltammetric characterization 3.3.1. Characterization in initial electrolyte

Characterization of the felt towards the positive electrode reaction
was examined by CV in 0.05 mol dm−3 V4+ solution in
1 mol dm−3 H2SO4 supporting electrolyte with scan rate of 2 mV s−1.
Measured CVs are presented in Fig. 3a. For all tested samples a pair of
oxidation-reduction peaks was observed at potentials around 0.35 V vs.
MSRE, which is in accordance with thermodynamic assumptions [29].
Firstly, EIS characterization was performed with the initial elec-
trolyte (equimolar ration of V3+:V4+ ions). The resulted spectra of full-
cell measurements of symmetric cell (both electrodes treated equally)
are presented in the form of Nyquist diagram on Fig. 4a. The equivalent
circuit shown on Fig. 1 was used to fit the spectra. According to the
theory [20, 33, 34], the intercept of the curve with horizontal axis re-
presents the ohmic resistance of the cell (RIN ) and the diameter of the

Table 1
Results of physicochemical characterization of thermally treated graphite felts.

Treatment temperature Wettability Resistance Capacitance Surface area Relative weight Raman Surface coverage (XPS)

W ASR EDLC SSA m/m0 ID/ IG C O C (sp2) C (sp3) -OH

−1 −1
°C - Ω cm 2
Fg 2
m g - - at.% at.% at.% at.% at.%

25 W0 0.33 0.11 1.0 1.00 0.95 97.3 2.7 90.6 5.4 4.0
400 W0 0.33 0.19 1.0 1.00 1.36 97.9 2.1 85.2 12.0 2.8
450 W1 0.36 1.65 1.0 1.00 1.80 96.9 3.1 83.8 12.5 3.7
500 W2 0.36 9.21 30.2 1.00 1.98 90.8 9.2 81.0 8.9 10.1
550 W1 0.46 9.11 26.5 0.80 1.68 94.9 5.1 80.4 14.2 5.5
600 W0 1.00 5.97 16.3 0.40 1.50 92.3 7.7 79.4 12.2 8.4

109
P. Mazúr et al.
Fig. 3. Cyclovoltammograms of thermally treated graphite felt electrodes for positive (a)
and negative (b) VRFB electrode reaction and corresponding dependences of ΔEp on
thermal treatment temperature (c).
flattened semi-circle in low frequency region represents the charge
transfer resistance (RCT ) of electrode reactions. The flattening of the
semi-circle is caused by non-uniform potential distribution within
porous felt electrode and thus the constant phase element (CPE) was
used in equivalent circuit instead of capacitor to account for unideal
behaviour of electrode-electrolyte double layer capacitance [33]. For
the symmetric cell set-up, EIS spectra of the full-cell contain two
identical overlapped semi-circles of positive and negative electrode
110
Journal of Power Sources 380 (2018) 105–114
Fig. 4. a) EIS spectra of VRFB single-cell with graphite felt treated at different tem-
peratures; b) Dependence of resistances evaluated form EIS spectra (Fig 4a) on treatment
temperature. State of charge −50%.
reactions. On Fig. 4b the values of RIN and RCT measured in −50% SOC
electrolytes are plotted as a function of the treatment temperature. The
untreated felt exhibited extremely high value of RCT , i.e., slow kinetics
of reaction (13)
VO 2 + + 2H+ + e− ⇆ V 3 + + H2 O E ° = 0.340 V (13)
The RCT decreased significantly due to the treatment and the lowest
value was found for the felt treated at 500 °C, i.e., with the highest
content of hydroxyl groups. The dependence of RIN on felt treatment
temperature does not show any clear trend; however, also here the
lowest value was found at 500 °C. This is slightly surprising, as the
electric resistivity of dry felt slightly increased due to the treatment, see
section 3.1. A minor increase of electric resistivity of the felt is probably
compensated by substantial increase of electrode-electrolyte interface
due to increased surface roughness and wettability, as indicated by
EDLC measurements.
Although the redox reaction eq. (13), which takes place in both
directions in −50% SOC electrolyte, does not proceed in operating
VRFB, we present the data to demonstrate that the results of the
characterization correlates well with the physicochemical properties.
Characterization in −50% SOC can provide us with valuable indirect
information regarding the content of hydroxyl electrocatalytic sites on
the fibre surface, which can be used e.g., for in-situ durability char-
acterization of the battery stack.
3.3.2. Characterization in +50% SOC
Subsequently, the battery was charged to +50% SOC by applying
50 mA cm−2 constant current density for corresponding time and EIS
characterization was repeated. The resulting Nyquist diagrams are
presented in Fig. 5a and b. Due to the different composition of positive
and negative electrolyte in +50% SOC of the battery, each spectrum
P. Mazúr et al. Journal of Power Sources 380 (2018) 105–114

electrode reactions was significantly reduced by the thermal treatment

of the felt. Identically as in −50% SOC, the lowest RIN and RCT were
found for the felt treated at 500 °C, which is consistent with the pre-
vious results of cyclovoltammetric characterization.
The same trend was found in results of charge and discharge load
curve measurements, which are presented in Fig. 5b. Except for the
untreated felt, strictly linear U-j dependence was measured for all the
felts in the measured load range, which is an indication of fast electrode
reaction kinetics. The mass transfer polarization was prevented by fast
circulation of electrolytes. The slopes of charging and discharging load
curves represent the total resistance of the cell under charging R charge
and discharging Rdischarge current load. For all the LC measurements,
slightly higher resistances were found for discharging, as it can be seen
in Fig. 5c. In accordance to the EIS results, the significant decrease of
cell resistance under load was achieved by the treatment and the lowest
cell polarization was found for the felt treated at 500 °C because of the
lowest ohmic and faradaic efficiency losses.
Finally, the cell was characterized by charge-discharge cycling at
various constant current densities in the voltage range of 0.8–1.65 V.
The 3rd complete cycles of the cycling at 50 mA cm−2 are for individual
treatments presented in Fig. 6a. The individual efficiencies are plotted
as a function of treatment temperature on Fig. 6b. In accordance to the
previous results, the mutual separation of charging and discharging U-Q
curves was significantly reduced by the thermal treatment, due to re-
duced RCT of electrode reactions, which resulted in substantial im-
provement of voltage efficiency (ηV ). The negligible decrease of cou-
lombic efficiency (ηC ) with treatment temperature is due to the

Fig. 5. Characterization of single-cell with felt electrodes thermally treated at various

temperatures: a) EIS spectra; b) Charge and discharge load curves; c) Dependence of
resistances evaluated form EIS spectra and load curves on the treatment temperature.
State of charge +50%.

consists of two partially overlapped semi-circles which correspond to

the processes on positive and negative electrode-electrolyte interfaces,
followed by low frequency region of finite diffusion [35]. The depen-
dence of cell resistances evaluated from EIS measurements on treatment Fig. 6. a) Charge and discharge curves of VRFB single-cell with graphite felt treated at
different temperatures at current density of 50 mA cm−2, b) Dependence of efficiencies of
temperature is graphically presented in Fig. 5. The kinetic barrier of
charge-discharge cycle at 50 mA cm−2 on the treatment temperature.

111
P. Mazúr et al.
Fig. 7. Dependence of voltage efficiency on current density of charge-discharge cycles.
increased cycle length and related capacity losses caused by vanadium
ion permeation [36]. The highest energy efficiency (ηE ) was found for
electrodes treated at 500 °C, i.e., the electrode with the highest surface
content of oxygen functional groups, which agrees well with the pre-
vious results.
3.4. Asymmetric single-cell characterization – 4-point measurements
To examine the role of positive and negative electrode separately,
we assembled two asymmetric cells containing: i) untreated positive
electrode and negative electrode treated at 500 °C (labelled as “negative
treated”) and ii) untreated negative electrode and positive electrode
treated at 500 °C (labelled as “positive treated”). Both cells were
characterised as in the previous cases. Opposite to our expectations,
with “negative treated” set-up, we achieved almost the same cell
parameters in +50% SOC as in case of symmetric cell with both elec-
trodes treated at 500 °C (labelled as “both treated”). This asymmetric
cell also exhibited only slightly lower values of voltage efficiency of
charge-discharge cycle in wide range of current densities from 50 to
150 mA cm−2, as shown in Fig. 7. Contrary to that, the performance of
“positive treated” cell was significantly worse and comparable with
symmetric untreated cell (labelled as “both untreated”) in terms of
resistances and efficiencies. These results clearly show that, at least for
the felt used, surface oxidation is significantly more effective for ne-
gative electrode performance in real battery cell.
Similar findings were previously reported by Agar et al. [31] who
also observed lower sensitivity of positive electrode reaction on thermal
treatment by asymmetric electrode single-cell measurements. However,
they explained the deficient performance of untreated negative gra-
phite felt electrode to be caused by blocking of electrode active sites by
evolved hydrogen bubbles. This explanation, however, is not supported
by our observations as, for untreated negative electrode felt, we did not
observe lower ηC of battery cycle due to hydrogen evolution nor sig-
nificant difference between R charge and Rdischarge at +50% SOC, see
Fig. 5b (assuming that hydrogen evolution, as a reduction reaction
takes place preferably during charging).
To clarify the role of electrode functionalization on VRFB single-cell
performance, we did a closer analysis of the half-cell EIS and LC mea-
surements at +50% SOC. Four platinum wires were positioned in the
cell as described in Fig. 1 which allowed us to measure EIS and LC of
individual battery half-cells. From the obtained data, we directly
evaluated the ohmic and faradaic contribution of individual cell com-
ponents on the overall cell resistance. The specification of experimental
set-up and method of evaluation is into more details described in the
112
Journal of Power Sources 380 (2018) 105–114
Fig. 8. Distribution of ohmic and faradaic resistances within the cell evaluated from EIS
and LC measurements at +50% SOC.
related MethodX article titled: “A complex four-point method for the
evaluation of ohmic and faradaic losses within a redox flow battery
single-cell”. The resistances are for all four cell set-ups presented on
Fig. 8. For each battery set-up, the left bar on the graph represents
internal resistances (RIN ) evaluated form high frequency region of
corresponding full-cell or half-cell EIS spectra. Analogically, the right
bar represents the cell resistance under current load (R charge ) evaluated
from corresponding charging half-cells and full-cell load curves. The
R charge thus includes both ohmic and faradaic contributions.
For all four set-ups, the distribution of ohmic losses was almost
unaffected by the treatment and membrane represented the main
source of resistance. The lowest ohmic losses were observed for “both
treated” cell, which is consistent with the results of symmetric cell
characterizations (slight decrease of RIN due to the treatment). The si-
tuation is different under current load where, for “both untreated” and
“positive treated”, i.e., the set-ups with untreated negative electrode,
we observed high faradaic resistance of negative electrode resulting in
high R charge of the full-cell (over 2 Ω cm2). In contrast, the faradaic re-
sistance of untreated positive electrode is substantially lower and is
almost unaffected by the treatment. The faradaic resistance of negative
electrode is considerably reduced by the treatment resulting in low
R charge of “both treated” and “negative treated” set-ups (around
1 Ω cm2). These originate mainly from ohmic losses of individual cell
components thanks to the high catalytic activity of both electrode
materials. The battery parameters evaluated from full-cell and half-cell
EIS and LC characterizations are for symmetric and asymmetric cell set-
ups summarized in Table 2.
Let us look more closely to the results of EIS half-cell measurements.
In initial electrolyte (equimolar V3+:V4+ ratio), we observed massive
RCT + for “both untreated” and “negative treated”, i.e., for both set-ups
with untreated positive electrode (approx. 1500 Ω cm2) and compar-
ably high RCT − for “both untreated” and “positive treated”, i.e., the set-
ups with untreated negative electrode. This is consistent with previous
observation in symmetric cell set-up. At +50% SOC, however, the ef-
fect of thermal treatment on RCT + is significantly lower than on RCT −,
which is in accordance with asymmetric full-cell characterization re-
sults. Surprisingly, slightly higher values of RCT + are observed for “both
untreated” and “positive treated” (i.e., both set-ups with untreated
negative electrode), when compared to “both treated” and “negative
treated”. This is most probably caused by the kinetics limitation of
negative untreated electrode, which serves as counter electrode for
positive half-cell measurements.
Based on these results we can conclude that the kinetics of positive
electrode reaction is reasonably fast even on untreated graphite felt
electrode and it is only slightly affected by the thermal treatment.
P. Mazúr et al. Journal of Power Sources 380 (2018) 105–114

Table 2
Results of single-cell characterization of thermally treated graphite felt in symmetric and asymmetric configurations.

Treatment temp./°C Initial electrolyte +50% SOC Efficienciesa

RIN RCT RCT + RCT − RIN RCT RCT + RCT − R charge ηC ηU ηE

Positive Negative Ω cm2 Ω cm2 Ω cm2 Ω cm2 Ω cm2 Ω cm2 Ω cm2 Ω cm2 Ω cm2 % % %

25 25 0.92 2349 1398 969.3 0.90 1.22 0.18 1.08 2.00 97.4 54.8 53.4
500 500 0.80 11.3 5.7 5.5 0.82 0.14 0.09 0.06 1.06 96.9 75.9 73.5
25 500 0.88 1627 1599 6.3 0.86 0.11 0.06 0.05 1.06 96.8 75.2 72.8
500 25 0.88 1523 3.9 1446 0.85 1.28 0.19 1.10 2.08 99.1 49.9 49.5

a
Efficiencies evaluated from charging discharging cycles at 150 mA cm−2.

Despite the substantial differences in the relevant electrode properties Acknowledgement

such as surface area (SSA and EDLC measurements) and content of
hydroxyl groups (XPS results) and edge-plane carbon atoms (Raman) Financial supports from Grant Agency of the Czech Republic (post-
described in section 3.1, the kinetics of positive electrode reaction is doctoral project no. 14-33400P) and from specific university research
almost identical. Similar findings were recently published by Langner fund (MSMT no. 20-SVV/2017) are gratefully acknowledged.
et al. [37] who employed Ag/AgCl reference electrodes on the elec-
trolyte inlet and outlet of the cell to determine overpotentials of posi- Appendix A. Supplementary data
tive and negative felt electrode. However, as their approach was based
only on polarization curve measurements, it did not allow to differ- Supplementary data related to this article can be found at http://dx.
entiate between individual ohmic and faradaic contributions of the doi.org/10.1016/j.jpowsour.2018.01.079.
half-cell polarization.
It is important to mention that, for positive electrode reaction, our List of symbols
results of single-cell characterization are inconsistent with our results of
cyclovoltammetric characterization in diluted vanadium electrolyte, T temperature °C
where both electrode reactions were found to be enhanced by the m weight g
thermal treatment to a similar extent. This discrepancy might be related U voltage V
to the differences in concentration of vanadium ions (0.05 vs. E potential V
1.6 mol dm−3) and/or sulphuric acid (1.0 vs. 2.0 mol dm−3) in elec- Q charge C
trolytes used. This is suggested by previous findings of Kaneko et al. j current density mA cm−2
−
[38] who reported the effect of sulphuric acid concentration on reaction Ip peak current of reduction A
mechanism and kinetics parameters of positive electrode reaction on I p+ peak current of oxidation A
graphite reinforcement electrode due to presence of vanadium com- ΔEp peak potential separation V
plexes. However, closer analysis of this phenomenon is beyond the ID intensity of D-band in Raman spectrum c.p.s
scope of this contribution. It is evident, that cyclovoltammetric char- IG intensity of G-band in Raman spectrum c.p.s
acterization, which represents a standard experimental research ap- ρ electric resistivity Ω cm
proach within the VRFB community, provides irrelevant information RIN internal resistance Ω cm2
regarding the electrode performance under real battery conditions. At RCT charge transfer resistance Ω cm2
the same time, our findings support the objections against the generally RCT + charge transfer resistance of positive electrode Ω cm2
accepted hypothesis [3] of catalytic effect of oxygen functional groups RCT − charge transfer resistance of negative electrode Ω cm2
on positive VRFB reaction which have been reported recently by several R charge resistance of charging Ω cm2
authors [15–17]. Rdischarge resistance of discharging Ω cm2
ηC coulombic efficiency of cycle %
ηV voltage efficiency of cycle %
4. Conclusions ηE energy efficiency of cycle %
List of acronyms
We studied the effect of oxygen functionalization of graphite felt
electrode on kinetics of VRFB electrode reactions. For that purpose, we VRFB Vanadium Redox Flow Battery
developed a complex four-point characterization method based on CV Cyclic voltammetry
impedance and load curve measurements in full-cell and half-cell ar- EIS Electrochemical impedance spectroscopy
rangements enabling direct evaluation of ohmic and faradaic re- CPE Constant phase element
sistances of individual components of single-cell. For the negative OCV Open circuit voltage
electrode reaction, we observed significant decrease of faradaic re- SEM Scanning electron microscope
sistance by the treatment which is in accordance with standard cyclo- BET Brunauer–Emmett–Teller
voltammetric characterization in diluted electrolyte. The enhancement XPS X-ray photoelectron spectroscopy
of negative electrode performance can be clearly attributed to the EDCL Electric double layer capacitance
functionalization of graphite fibres by oxygen containing groups. For SSA Specific surface area
positive electrode reaction, however, we observed almost no effect of ASR Area specific resistance
the treatment on electrode performance, which is in contradiction to
the results of cyclovoltammetric characterization in diluted electro- References
lytes. The developed method can be further employed for the compo-
nents characterization and in-situ durability studies on single-cell scale. [1] M.-J. Li, W. Zhao, X. Chen, W.-Q. Tao, Appl. Therm. Eng. 114 (2017) 802–814.
[2] P. Alotto, M. Guarnieri, F. Moro, Renew. Sustain. Energy Rev. 29 (2014) 325–335.

113
P. Mazúr et al.
[3] K.J. Kim, M.-S. Park, Y.-J. Kim, J.H. Kim, S.X. Dou, M. Skyllas-Kazacos, J. Mater.
Chem. A 3 (2015) 16913–16933.
[4] S. Zhong, C. Padeste, M. Kazacos, M. Skyllas-Kazacos, J. Power Sources 45 (1993)
29–41.
[5] D.S. Aaron, Q. Liu, Z. Tang, G.M. Grim, A.B. Papandrew, A. Turhan,
T.A. Zawodzinski, M.M. Mench, J. Power Sources 206 (2012) 450–453.
[6] P. Ehrburger, F. Louys, J. Lahaye, Carbon 27 (1989) 389–393.
[7] X. Chu, K. Kinoshita, Mater. Sci. Eng. B 49 (1997) 53–60.
[8] T.J. Rabbow, M. Trampert, P. Pokorny, P. Binder, A.H. Whitehead, Electrochim.
Acta 173 (2015) 17–23.
[9] T.J. Rabbow, M. Trampert, P. Pokorny, P. Binder, A.H. Whitehead, Electrochim.
Acta 173 (2015) 24–30.
[10] A.M. Pezeshki, J.T. Clement, G.M. Veith, T.A. Zawodzinski, M.M. Mench, J. Power
Sources 294 (2015) 333–338.
[11] B. Sun, M. Skyllas-Kazacos, Electrochim. Acta 37 (1992) 1253–1260.
[12] X.-g. Li, K.-l. Huang, S.-q. Liu, N. Tan, L.-q. Chen, Trans. Nonferrous Metals Soc.
China 17 (2007) 195–199.
[13] C. Flox, M. Skoumal, J. Rubio-Garcia, T. Andreu, J.R. Morante, Appl. Energy.
[14] C. Gao, N. Wang, S. Peng, S. Liu, Y. Lei, X. Liang, S. Zeng, H. Zi, Electrochim. Acta
88 (2013) 193–202.
[15] A. Bourke, M.A. Miller, R.P. Lynch, J.S. Wainright, R.F. Savinell, D.N. Buckley, J.
Electrochem. Soc. 162 (2015) A1547–A1555.
[16] J. Friedl, C.M. Bauer, A. Rinaldi, U. Stimming, Carbon 63 (2013) 228–239.
[17] J. Melke, P. Jakes, J. Langner, L. Riekehr, U. Kunz, Z. Zhao-Karger, A. Nefedov,
H. Sezen, C. Wöll, H. Ehrenberg, C. Roth, Carbon 78 (2014) 220–230.
[18] J. Langner, M. Bruns, D. Dixon, A. Nefedov, C. Wöll, F. Scheiba, H. Ehrenberg,
C. Roth, J. Melke, J. Power Sources 321 (2016) 210–218.
[19] H. Kabir, I.O. Gyan, I. Francis Cheng, J. Power Sources 342 (2017) 31–37.
[20] A.J. Bard, L.R. Faulkner, Electrochemical Methods: Fundamentals and Applications,
114
Journal of Power Sources 380 (2018) 105–114
second ed., Wiley, 2001.
[21] F. Regisser, M.-A. Lavoie, G.Y. Champagne, D. Bélanger, J. Electroanal. Chem. 415
(1996) 47–54.
[22] D.R. Dreyer, S. Park, C.W. Bielawski, R.S. Ruoff, Chem. Soc. Rev. 39 (2010)
228–240.
[23] W. Li, J. Liu, C. Yan, Carbon 55 (2013) 313–320.
[24] C.T.J. Low, F.C. Walsh, M.H. Chakrabarti, M.A. Hashim, M.A. Hussain, Carbon 54
(2013) 1–21.
[25] S.Y. Toh, K.S. Loh, S.K. Kamarudin, W.R.W. Daud, Chem. Eng. J. 251 (2014)
422–434.
[26] K. Angoni, J. Mater. Sci. 33 (1998) 3693–3698.
[27] M.A. Montes-Morán, R.J. Young, Carbon 40 (2002) 845–855.
[28] K.J. Kim, Y.-J. Kim, J.-H. Kim, M.-S. Park, Mater. Chem. Phys. 131 (2011) 547–553.
[29] M. Pavelka, F. Wandschneider, P. Mazur, J. Power Sources 293 (2015) 400–408.
[30] L. Yue, W. Li, F. Sun, L. Zhao, L. Xing, Carbon 48 (2010) 3079–3090.
[31] E. Agar, C.R. Dennison, K.W. Knehr, E.C. Kumbur, J. Power Sources 225 (2013)
89–94.
[32] D.M. Kabtamu, J.-Y. Chen, Y.-C. Chang, C.-H. Wang, J. Power Sources 341 (2017)
270–279.
[33] M.E. Orazem, B. Tribollet, Electrochemical Impedance Spectroscopy, Wiley, 2008.
[34] C.-N. Sun, F.M. Delnick, D.S. Aaron, A.B. Papandrew, M.M. Mench,
T.A. Zawodzinski, ECS Electrochem. Lett. 2 (2013) A43–A45.
[35] C.-N. Sun, F.M. Delnick, D.S. Aaron, A.B. Papandrew, M.M. Mench,
T.A. Zawodzinski, J. Electrochem. Soc. 161 (2014) A981–A988.
[36] A. Di Blasi, N. Briguglio, O. Di Blasi, V. Antonucci, Appl. Energy 125 (2014)
114–122.
[37] J. Langner, J. Melke, H. Ehrenberg, C. Roth, ECS Trans. 58 (2014) 1–7.
[38] H. Kaneko, K. Nozaki, Y. Wada, T. Aoki, A. Negishi, M. Kamimoto, Electrochim.
Acta 36 (1991) 1191–1196.