Nano Energy (2016) 22, 396–405

Available online at www.sciencedirect.com

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

FULL PAPER

Unbiased solar energy storage:

Photoelectrochemical redox flow battery
João Azevedoa,b, Thorsten Seippc, Jens Burfeindc, Célia Sousab,
Anders Bentiend, João Pedro Araújob, Adélio Mendesa,n

a
LEPABE-Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal
b
IFIMUP and IN – Institute of Nanoscience and Nanotechnology, Departamento de Física e Astronomia,
Faculdade de Ciências da Universidade do Porto, Rua do Campo Alegre, 678, 4169-007 Porto, Portugal
c
Fraunhofer Institute for Environmental, Safety, and Energy Technology UMSICHT, Osterfelder Street 3,
46047 Oberhausen, Germany
d
Department of Engineering, Aarhus University, Hangøvej 2, 8200 Aarhus, Denmark

Received 28 October 2015; received in revised form 20 January 2016; accepted 15 February 2016
Available online 22 February 2016

KEYWORDS Abstract
Redox flow battery; Electricity from renewable energy sources is craving for efficient storage technologies, in particular
Solar energy storage; solar industry, to enable practical small-scale solutions for residential and offices use. The best
Tandem device; stationary technology is probably the redox flow batteries. This article proposes the direct
Dye-sensitized solar conversion of sunlight into electrochemical energy stored in a redox flow battery. A photoelec-
cell;
trochemical cell is used to charge a vanadium redox flow cell (CdS(s)|V3 + , VO2 + ||V3 + , V2 + |Carbon
Cadmium sulfide
Felt(s), E0 =0.6 VNHE). A CdS thin film photoelectrode is prepared to directly charge the cell, pairs
V3 + /VO2 + . CdS photoanode exhibits competitive photocurrents, when compared to other photo-
electrochemical devices, and yields enough photovoltage to charge the vanadium battery up to 75%
with no external bias. An overlayer of CdSe improves the performance of CdS with current densities
up to ca. 1.4 mA cm
2. Finally, a tandem configuration is tested using a dye-sensitized solar cell and
a CdS photoanoade creating  1.3 V of photovoltage. This tandem arrangement proves to charge
conventional all vanadium redox cell (DSC/CdS(s)|VO2 + , VO2+ ||V3 + , V2 + |Carbon Felt(s), E0 =1.2 VNHE)
without external bias.
& 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction
n
Corresponding author. The efficient exploitation of renewable energy sources is
E-mail address: [email protected] (A. Mendes). the only route for long-term energy supply. However,

http://dx.doi.org/10.1016/j.nanoen.2016.02.029
2211-2855/& 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Unbiased solar energy storage. 397

renewable energy is often intermittent, local in nature [1] overpotentials needed to drive the hydrogen and oxygen
and there is an inherent mismatch between production and evolution reactions, the energy required for photoelectro-
consumption. If future goals of fully renewable energy lysis is frequently reported to have values around 2 eV per
sources are to be met it is paramount that efficient storage electron-hole pair generated [2,11]. Considering the sig-
technologies are developed. Compared to other storage nificant amount of research in the field for the past two
systems, such as conventional batteries, gravity or decades, practical semiconductors solutions that enable
mechanical-based and supercapacitors, that can be used unbiased and efficient single photoelectrode solar water
for the applications requiring moderate energies and high splitting appear unlikely.
power densities, chemical fuels combine the advantages The use of photoeletrodes for converting solar into
of ease of transportation and high-energy storage densities electrochemical energy in a redox flow battery (RFB)
[2–4]. arrangement is a disruptive approach that allows an effi-
For the past two decades, storage in hydrogen through cient storage of solar energy. Contrary to water splitting,
electrolysis of PV electricity or direct photoelectrochemical where oxidation and reduction potentials are unique, in the
water splitting has been some of the most investigated case of direct solar charging redox flow batteries it is
technologies for storage of solar energy [5–8]. Much effort potentially possible to find two redox pairs for a selected
was devoted to fabricate and optimize photoelectrochem- photoelectrode, fitting the redox potentials with the semi-
ical (PEC) cells for water splitting offering the prospect of conductor band edges. Moreover, the round trip energy
clean, dispatchable, abundant, and affordable energy. efficiency using redox pairs for storing electricity is very
However, hydrogen is a very low-density gas requiring high, ca. 85%, compared to hydrogen with round trip energy
high-end technology for storage and transport. Also, its efficiencies of 20–50% [12]. The direct solar charging of
production via PEC cells route still presents many unsolved redox flow batteries is then viewed as a premium opportu-
challenges, in particular low conversion efficiency [9,10]. nity for the research on photoelectrochemical cells.
PEC cells require an abundant semiconductor material with Preliminary experiments using a photoelectrode for char-
a band gap energy (Eg) large enough to split water and a ging redox pairs have been reported in the 80 s but limited
conduction band-edge energy (Ecb) and valence band-edge choice of redox pairs and membrane limitations made the
energy (Evb) that straddle the electrochemical potentials for combined PEC and RFB device unable to efficiently convert
generation of hydrogen (E 0H þ =H2 ) and oxygen (E 0O2 =H2 O ) [9]. solar into electrochemical storable energy and research in
This requirement limits the possibilities to a handful of this topic came to a halt for decades [13,14]. Compared to
semiconductor materials. Since the positions of both the PEC water splitting, for solar-RFB there are multiple choices
band edges in many semiconductors and E 0H þ =H2 /E 0O2 =H2 O shift of the redox couples given a great flexibility for choosing
with
59 mV per pH unit [2], pH-tuning strategies to match suitable photoelectrodes. Semiconductors that are unsuita-
energy levels are not effective. ble for water splitting due to thermodynamic potential
In a PEC cell, the photo induced free charge carriers mismatch or corrosion may now be used for solar-RFB since
migrate to a semiconductor/liquid junction reacting with the choice of redox pairs is quasi-unlimited with redox
electroactive species directly at the semiconductor inter- potentials varying almost continuously [15–17]. Further-
face. Given the energy efficiency losses at semiconductor/ more, most of the charge transfer processes involving redox
liquid junctions due to the concentration and kinetic couples normally used in RFB have faster kinetics since they

Figure 1 Schematic diagram of (a) all vanadium solar redox flow battery charged with a CdS photoanode and (b) energy diagram of
the system, including the standard redox reactions.
398 J. Azevedo et al.

rely on single or double electron transfer processes a good match between selected electrolyte and semicon-
[15,16,18,19]. Thus, the overall system of solar-RFB is very ductor. In this work it is intended to prove the possibility to
likely to be more efficient and the direct conversion of achieve easily unbiased charging of a RFB. In the future
sunlight into electrochemical energy may potentially be a other electrolyte/semiconductor combinations can be used
cost effective alternative to photovoltaic (PV) charged RFB. to achieve higher efficiencies and energy densities. With the
The RFB can receive simultaneously charged electrolytes current developments in RFB technology [16,19,26,27] an
from the PEC panels or electricity from the grid or PV panels even larger range of electrolytes could be available for
and convert it into the chemical energy stored in the energy fine-tuning between photo absorber and redox pairs.
charged electrolytes. Sunlight can be used to heat up the
electrolytes (thermal solar collector) in a cogeneration
arrangement. The heated electrolytes can then heat
Materials and methods
exchange with sanitary waters and for thermal comfort,
where for cooler climates a heat pump may be needed – Substrate preparation
Figure S1.
While this work was being developed, two different Conducting fluorine doped tin oxide (F:SnO2) glass was used
research groups published work on the solar battery con- as substrate (Solaronix TCO22-7). FTO-glass was cut into
cept. Wu et al. combined a dye sensitized TiO2 photoelec- 1  3 cm2 pieces and then cleaned by sonication sequen-
trode and a lithium based battery describing the advantages tially in soapy water (15 min), acetone (15 min), ethanol
of having a photoassisted battery charging [20,21]. Despite (15 min), and finally in distilled water (15 min).
the reported advantages, single photoelectrode unassisted To obtain reproducible high quality samples, the FTO
photocharge is highly unlike for the reported system and a substrate was delimited with an epoxy resin resistant to high
tandem arrangement is most probably not beneficial due to temperatures (Araldites Super Strength). The active area was
the photoanode broad absorption spectrum. Moreover, the measured from photographs of the samples with image
energy storage capacity of a lithium based battery is limited analysis software [28], and found to be 0.94 cm270.37 cm2.
by the size of the battery, making this technology unprac-
tical for very large applications. Liu et al. integrated an all CdS thin film fabrication
vanadium redox flow battery (RFB) with a TiO2 photoelec-
trode with high Faradaic efficiencies [22,23]. RFBs are easily
CdS thin films can be prepared by different methods; here
scalable, display high round trip efficiencies, long lifetimes
electrodeposition [29] has been chosen because the mor-
and acceptable costs for large-scale electrical energy
phology and nanoparticle size of the films can be easily
storage, when compared with lithium-based batteries
controlled by e.g. pH, temperature and deposition potential
[15,16,24]. The charging measurements of the all vanadium
[30–32].
RFB, by Liu et al. were conducted with discharged batteries
The chosen electrodeposition conditions of the CdS thin
and a voltage gain of less than 0.05 V is reported [23].
films were based on previous reports [33,34] and optimized
Contrarily to solar water splitting, where the redox poten-
for the present system. Briefly, CdS thin films were depos-
tials of the chemical species in solution are constant during
ited in a plating bath containing 2 mM CdSO4 (99%, Sigma-
electrochemical reactions (not accounting small pH
Aldrich) and 100 mM Na2S2O3 (98%, Sigma-Aldrich), with pH
changes), in RFB an intrinsic electrochemical potential
adjusted to 2.5 using 2 M H2SO4 (95–97% Fluka). The bath
variation is observed governed by the Nernst equation
was kept at 40 1C with stirring throughout the deposition. A
[25]. The battery was charged up to 5% state of charge
three-electrode configuration cell with a ZENNIUM (Zahner
(SOC); beyond this point more potential is required from the
Elektrik, Germany) electrochemical workstation was used at
photoanode to charge the battery, which the authors have
a constant potential of
0.8 V for 2 h for most of the
not demonstrated to be possible with the proposed photo-
samples. The counter-electrode was a 99.9% pure platinum
electrode. Moreover, a TiO2 photoelectrode, which has a
wire (Alfa Aesar, Germany) and the reference electrode was
wide band gap, can only absorb a small part of the solar
Ag/AgCl/Sat.KCl (Metrohm, Switzerland). After electrode-
spectrum, making TiO2 based devices energy inefficient.
position the samples were annealed at 450 1C for 30 min, in
Figure 1a is a schematic representation of the proposed
ambient atmosphere, under forced air flux. The films had a
solar RFB that consists in a cadmium sulfide (CdS) semi-
thickness of ca. 200 nm.
conductor electrode placed at the negative (photoanode)
side of a RFB. The photogenerated electron/hole pairs
conduct the necessary redox reactions to charge the battery Under and overlayers deposition
whereas the charged electrolytes are stored in two external
tanks. Given that the photoelectrode provides enough Different configurations were tested in order to improve the
voltage to charge the battery, direct charge can be con- performance of the CdS photoelectrodes. Layers of n-type
ducted in a stand-alone device (with no external bias semiconductors were deposited on top of CdS or directly on
needed). The vanadium electrolytes were selected not only FTO such as TiO2 and CdSe. Epoxy resin or Teflons tape was
for their good kinetics or reversibility, but also for energy fit used to cover any exposed FTO substrate and delimit the
with the CdS band gap positioning – Figure 1b. In this work, active area.
for the first time, it is demonstrated that the solar RFB, CdS Cadmium selenide (CdSe) was electrodeposited from a
(s)|V3 + , VO2 + ||V3 + , V2 + |Carbon Felt(s), E0 = 0.6 VNHE, can be 2 mM aqueous solution of SeO2 (99.8%, Sigma-Aldrich) in
charged up to 75% SOC without any external bias with 20 mM CdSO4, at pH= 2. Galvanostatic deposition was done
photocurrents of ca. 1.4 mA cm
2. This was possible due to with 1 mA cm
2 at 30 1C using the same three electrode
Unbiased solar energy storage. 399

configuration as described above. The deposition time was Flow cell construction and cycling analysis
adjusted to 400 s to give a 200 nm thick film.
Titanium dioxide (TiO2) overlayer films were deposited The flow cell used had 25 cm2 active membrane area and a
using atomic layer deposition (Beneq TFS 200) at a substrate thickness of 3 mm. The RFB was assembled in a flow-through
temperature of 225 1C using titanium isopropoxide (99.99%, configuration. A Nafion™ NM 117 (Quintech) membrane served
Aldrich), at a precursor temperature of 45 1C, and water at as the separator. High surface area carbon felt electrodes were
room temperature. TiO2 was deposited in pulse mode under used (Sigratherms GFA5 from SGL Technologies GmbH) with
a nitrogen flow of 0.250 LSTP min
1, with a pulse length of an uncompressed thickness of 6 mm and a specific weight of
1 s and a purge time of 5 s for TiO2 and a pulse length of 1 s 500 g m
2. The carbon felt was heat-treated at 350 1C for 7 h,
and a purge time of 3 s for water. The growth per cycle of in oxygen atmosphere, prior testing for activation. The current
TiO2 corresponded to 0.58 Å cycle
1 at 225 1C. The samples collectors were graphite bipolar plates (Sigracets TF6 from
were annealed at 500 1C for 15 min. SGL). Contacts to the current collectors were made with
Underlayer of porous TiO2 (denoted as TiO2 paste) was nickel-coated copper plates. Battery cycling studies were
deposited by doctor blade of a TiO2 paste (Ti-Nanoxide T/SP carried out with a ZENNIUM workstation in a two-electrode
from Solaronix) followed by drying at 80 1C for 20 min. The configuration. Charge–discharge cycles were recorded galva-
samples were annealed at 500 1C for 15 min. The thickness nostatically, with battery at 0% SOC. All cycling measurements
was approximately 5 μm. were conducted at current densities of 1 mA cm
2. Further, all
Crystalline structure of underlayers and overlayers was experiments were performed at room temperature with no
checked by X-ray diffraction and energy-dispersive X-ray active temperature control. Watson Marlow 323 pumps were
spectroscopy as described in Section 2.4. used to maintain an electrolyte flow rate of 20 mL min
1
during cycling measurements. Both reservoirs were argon
Thin Film Characterization purged before testing to minimize oxidation of the vanadium
species. Charge/discharge cycles of Carbon Felt(s)|V3 + , VO2 + ||
The morphology and surface topography of the films was V3 + , V2 + |Carbon Felt(s), Eo =0.6 VNHE are represented – Figure
characterized using a high-resolution scanning electron S2.
microscope (FEI Quanta 400 FEG ESEM) with a through-
the-lens detector for secondary electrons. The acceleration Solar battery characterization
voltage was 15 keV while an in-lens detector was employed
with a working distance of about 10 mm. Cross-sectional
The photoelectrochemical cell used in the present study is
images were acquired from freshly cleaved samples. The
based on previous design [36], and shown in Figure S3. The
compositions of as-obtained thin films were studied by
distance between membrane and each photoelectrode was
energy-dispersive X-ray spectrometry (EDAX Genesis X4M).
1 cm and a Nafion™ NM 117 membrane served as separator.
The crystalline structure of the samples was analyzed by
The photoelectrodes were placed as close as possible
X-ray diffraction (XRD) in a Rigaku MiniFlex 600 diffract-
( 1 mm) from the transparent window to minimize the
ometer in the θ
2θ geometry using the Cu-Kα line with
light absorption from the electrolytes. The carbon felt
wavelength λ= 1.540593 Å. The spectra were acquired by
counter-electrode was bonded to the graphite current
moving the detector between 2θ = 201 and 2θ = 601 at a scan
collector plate with a carbonized resorcinol resin and no
rate of 0.6 deg min
1 with a step width of 0.021. The
compression is thus needed for obtaining a good electrical
spectra peaks were fitted with Lorentzian functions to
contact. The carbonization/pyrolysis process is described
determine the average grain size according to Scherer
elsewhere [37].
equation, as described elsewhere [33]. Reflection patterns
For measurements at different SOC, the electrolytes were
were matched literature values [29].
charged using an electrochemical cell equipped with two carbon
UV–vis spectra were obtained on a Shimatzu UV-3600 UV–
felt electrodes. For example, for a standard all vanadium
vis-NIR spectrophotometer, equipped with a 150 mm inte-
battery, comprising of initial redox species V3 + and VO2 +
grating sphere and using BaSO4 as 100% reflectance stan-
(Carbon Felt(s)|VO2 + , VO2+ ||V3 + , V2+ |Carbon Felt(s)), the bat-
dard. A quartz cuvette with path length of 1 cm was used
tery was charged up to 1.2 V until a small current was observed
for electrolyte absorption measurements.
(corresponding to 20% SOC according to Nernst equation-Figure
S4) and the electrolyte absorption spectra were measured.
Preparation of vanadium solutions Using Beer–Lambert law [38], the concentration of the vana-
dium species with different oxidation states was estimated and
Vanadium electrolyte solution (AMG-GfE GmbH) containing the experimental SOC computed. Given the high concentration
1.6 M vanadium in 2 M H2SO4 was used. The vanadium of vanadium ions in solution, the electrolyte samples were
species were separated in V3 + and VO2 + using a flow cell diluted before determining the absorption.
at constant potential 1 V (see Eq. (1)), until a clear blue The current-potential characteristic curves were
solution was observed on the cathode side and a dark green recorded by linearly sweeping the potential using a ZEN-
solution on the anode side, which indicates a good separa- NIUM workstation. The measurements were performed in
tion of both species. Prior to any test, the initial solutions the dark and under simulated sunlight (Newport solar
were screened in the UV–vis-NIR spectrophotometer to simulator, 150 W Xe lamp, AM 1.5 G, 1000 W m
2) at a scan
check the complete separation of V3 + and VO2 + . The rate of 10 mV s
1 in a two-electrode configuration. The
obtained solutions were then diluted to 0.4 M using 2 M light beam was calibrated with a c-Si photodiode. For
H2SO4. stability measurements, no external bias was applied and
400 J. Azevedo et al.

the photocurrent response was recorded over time. No for 15 min. Both electrodes were then sealed using a laser
hydrogen or oxygen bubbles were observed during the assisted glass frit method. Dye (1 mM N719 dye solution) was
experiments. recirculated for 10 h, electrolyte was inserted in the cell
(Iodolyte Z-150 from Solaronix) and the holes sealed using a
Tandem system glass lamella glued to the substrate using a thermo-plastic
sealant (Surlyns, DuPont). In order to insert the DSC inside
the electrolyte, the entire cell edge was protected with
To test a tandem configuration of a CdS photoanode with a
epoxy resin resistant to high temperatures (Pattex Nural
dye-sensitized solar cell (DSC), a DSC was fabricated as
22). The DSC was placed in the back of the photoanode,
described elsewhere [39]. Briefly, cleaned FTO glasses were
with which was connected in series using an electrical wire.
immersed in a 40 mM TiCl4 (99.9% Sigma-Aldrich) aqueous
The characteristic curves were obtained for both stand-
solution at 70 1C, for 20 min. Then, the samples were
alone DSC and tandem system.
coated with porous TiO2 layer by screen printing a commer-
cial TiO2 paste (Tiscreen printi from Solaronix), followed by
drying at 80 1C for 20 min. Finally, the samples were Results and discussion
annealed at 500 1C for 15 min in an infrared electrical oven
(Nabertherm Gmbh model GF75). After, samples were again Electrodeposited CdS films were 200 nm thick, compact and
treated with 40 mM TiCl4 aqueous solution and annealed at homogeneous (Figure 2a). The surface morphology of the
500 1C for 15 min. The counter electrodes were made from samples is nanostructured with cauliflower-like features as
the same cleaned FTO glass, after being drilled with two small as 50 nm (Figure 2b). Depending on the deposition
holes of 1 mm diameter. A commercial platinum based paste potential, different morphologies were obtained (Figure S5)
(Platisol T/SP from Solaronixs) was applied on the glass and the cauliflower-like nanostructured surface was
substrate by screen-printing followed by annealing at 400 1C obtained with a deposition potential of
0.8 V. The

0.5 µm
CdS ~ 200 nm

FTO ~ 500 nm

glass
1 µm 5 µm
(002)

Observed Transmission
Fitted Reflection
Difference Absorption
Intensity / (arb. units)
Intensity (arb. units)

(101)
(100)

(110)

(004)
(103)
(102)

20 30 40 50 60 200 300 400 500 600 700 800

(º) / nm

Figure 2 Cross-sectional (a) and surface (b) SEM micrograph of a CdS photoanode. Inset shows an amplified view of the surface;
(c) X-ray diffraction patterns of CdS photoanode. Different crystalline phases are marked with the corresponding miller indexes.
Orange circles mark the TCO diffraction peaks; (d) The UV–VIS spectra of CdS thin films.
Unbiased solar energy storage. 401

increased roughness is associated with a better photoactiv- 0.7

ity of the photoanode [40] and in tests this cauliflower-like
nanostructure yielded the best photocurrent performance. 0.6
The films revealed good adhesion to the FTO glass, impera-
0.5
tive for a good electrical performance.
The CdS layer was characterized by X-ray diffraction-

J / mA cm-2
0.4
Figure 2c; literature values are shown as reference – Table in
S1 [29]. Each peak was fitted with a Lorentzian function to 0.3
assess the grain size using Scherer’s equation [35]. The fit
0.2
and difference to measured curves are plotted to verify the
quality of the fitting. XRD spectra of CdS films indicate a 0.1
nanocrystalline size of about 10 nm. Given the small thick-
ness of the samples, the underlying FTO peaks are also 0.0
visible and are identified in the spectra.
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2
UV–vis spectrophotometry, shown in Figure 2d, was
performed to investigate the optical absorption of the films E/V
in the range of 200–800 nm. It can be observed that these Figure 3 (a) Photocurrent–photovoltage characteristics under
films exhibit very good light absorption below 500 nm and it 1.5 AM light and dark conditions (in 0.4 M V3/V3 / 2 M H2SO4
transmits most of the light in the 600–800 nm range. electrolyte) of the best performing CdS photoanode. Inset
The selected electrolyte for the solar RFB were vanadium shows stability.
aqueous solutions:
V3 þ þH2 O-VO2 þ þ e
þ2H þ E 0ox ¼
0:337VNHE ð1Þ
to charge the battery further. For this reason, an initial low
3þ
2þ onset is required to charge the battery to high SOC.
V þ e -V E 0red ¼
0:255VNHE ð2Þ
The charging state of the battery was assessed by spectro-
This configuration is named V3/V3 from now on. These photometry. A fully charged and discharged electrolyte sample
vanadium species exhibit a standard cell potential of was analyzed and the absorption spectra were according to
E0 = 0.6 VNHE (Figure S4). Also, V3 + /V2 + redox pair is pH literature: V3 + showed absorption peaks at 400 nm and 600 nm
independent (Figure S6) offering the possibility to adjust and VO2 + solution at 760 nm, Figure 4b. Starting from a
the open circuit voltage (OCV) changing the electrolyte pH complete discharged state, the battery was charged during
up to 2 for better band energy tuning; above this pH some equal time intervals at 0 V bias, under continuous 1.5 AM
vanadium species may start precipitating. In this work, pH illumination, using several CdS samples. Absorption spectra of
was set to 0 to avoid precipitation and maximize energy the anolyte were taken at each interval and a clear change
density, at the expense of system stability. The OCV was from V3 + to VO2 + is observed over time.
experimentally measured using a carbon-felt electrode on The charging was confirmed from the absorption data
cathode and anode sides, at different SOC values; it was assuming a linear relationship between concentration and
done by applying a constant potential until a small residual the 760 nm absorption peak of VO2 + [38]. From Lambert–
current was observed. Beer law, using 18.46 as VO2 + molar absorptivity at 760 nm,
Under AM 1.5 illumination, the as-deposited bare CdS the produced amount of VO2 + was estimated. After passing
produced an unbiased photocurrent as high as 0.47 mA 0.6 C through the photoanode under continuous illumina-
cm
2 in a nitrogen-purged 0.4 M V3 + electrolyte in 2 M tion, an amount of 0.13 mM of VO2 + was computed (total
H2SO4, for completely discharged V3/V3 cell-Figure 3; data electrolyte volume of 14 ml). A conversion rate of0.1 mM
shown are from top performing batches. The dark current h
1 for VO2 + was estimated.
was negligible up to 0.2 V. Preliminary experiments with a To improve performance it was attempted to coat the
1.6 M vanadium solution presented smaller photocurrent ( photoelectrode with an overlayer. Here it is important that
0.3 mA cm
2) attributed to the electrolyte darker color. it has a staggered type-II band offset when interfaced with
The photocurrent onset is at
0.73 V-sufficient to charge CdS. This ensures that charge carriers are injected in the
the battery up to 90% SOC without external bias, according correct direction for the desired reactions i.e. holes to the
to Nernst equation (Figure S4). electrolyte and electrons to the counter electrode. This is
To confirm the possibility of unbiased charging battery, the case of some n-type semiconductors with sufficiently
photocurrent characteristic curves were obtained at differ- low conduction band edge. Secondly, the layer should have
ent SOCs (Figure 4a). Given the high concentration of a catalytic activity that promotes the charging reaction (Eq.
vanadium ions in solution, the redox reactions kinetics (1)). Finally, no corrosive side reactions should occur at
remained mostly the same across the different SOC, and potentials within the bandgap potential range or at least
small differences in fill factors and plateau photocurrents have sufficiently low corrosion kinetics. Among n-type
are attributed to sample-to-sample variation. Only after semiconductors CdSe and TiO2 meets these criteria and
75% SOC a decrease in photocurrent is observed which was were selected to be tested.
attributed to a depletion of V3 + ions at the surface of the Table 1 summarizes the PEC response produced by the
photoelectrode and counter-electrode. A shift in the onset two protected photoelectrodes. The maximum photocurrent
is observed as the battery charges. This shift fits perfectly corresponds to the photocurrent density at 0 V during the
with the Nernst equation: as the battery charges a built in linear sweep voltammetry, corrected by the dark current.
chemical potential is created which requires more voltage The stability was assessed over 5 min with
402 J. Azevedo et al.

a attempt to reduce the resistive behavior, very thin layers of

TiO2 (5 nm) were tested, however, without improved per-
formance (results not shown).
CdSe, on the other hand, creates a favorable potential
step when in contact with CdS [44]. In fact, CdS/CdSe
structure is widely studied [29,45–47]. With a band gap of
1.7 eV, good conductivity and light absorption, CdSe has
been studied for improving both CdS onset potential and
photocurrent. Also, anodic decomposition potential [48] lies
below the vanadium oxidation potential implying that CdSe
should be more stable than CdS in vanadium electrolytes
[29,44,45].
The optimized CdS/CdSe photoelectrode showed an onset
at
0.67 V and a photocurrent of 1.4 mA cm
2 at 0 V bias
(Figure 5). CdSe protected samples displayed better stabi-
lities (up to 5 min); but given the extreme pH conditions this
still needs to be improved. A stable device can be achieved
using sacrificial reagents [49] in the electrolyte but it is not
the scope of this work, which solely intends to demonstrate
the charge of the vanadium battery with no bias.
The open circuit potential of a RFB is directly related to
the energy density it can provide. Open circuit potentials
depend on different parameters according to Nernst equa-
tion but the standard redox potential is the predominant
factor. Most of the commercial redox flow batteries have
open circuit potentials over 1 V. The tested photoelectrodes
provide insufficient photopotential to charge such a RFB
b without external bias. The necessary extra voltage can be
obtained with a tandem system, placing a PV cell aligned
Discharged
Charged with the photoelectrode [50–52]. The PV cell in front of the
After 80 min charge photoanode generates an extra bias that is added to the
After 160 min charge photovoltage of the photoelectrode, provided that enough
Absorbance / (arb. units)

After 240 min charge

After 320 min charge light reaches the photoanode through the PV cell. A
schematic representation of the proposed system is given
in Figure 6a. The selected PV technology was a dye
sensitized solar cell since it can be made semi-transparent
and it is fabricated from low cost and environmentally safe
components. The semiconductor samples with lower onset
potential were selected, in this case CdS.
As a higher energy density battery the following electro-
lytes were selected (named V3/V4 from now on):

VO2 þ þ H2 O-VO2þ þe
þ2H þ E 0ox ¼
0:991V NHE

300 400 500 600 700 800 ð3Þ
/ nm
V3 þ þe
-V2 þ E 0red ¼
0:255VNHE ð4Þ
Figure 4 (a) Photocurrent–photovoltage characteristics under
chopped 1.5 AM light CdS photoanode at different SOC values in 2
The DSC cell used had 4 cm active area with an open
V3/V3; (b) Absorption spectra of V3 + –VO2 + mixtures at differ- circuit potential of 0.66 V, a short circuit current of E 2.7
ent SOC and for the parent V3 + and VO2 + solutions. mA cm
2 and a fill factor of 41% – see Figure S8a; a CdS
photoanode matching the useful area of the DSC, 4 cm2, was
chronoamperometric measurements at 0 V and chopped light. prepared. The low current density of the DSC cell was
The final photocurrent density (J) was compared with the attributed to the thin, semitransparent, photoelectrode
initial photocurrent density (J0). while the low fill factor was assigned to the large active
TiO2 is a well-tested material as a protective layer for area that promotes recombination [39]. The selected dye
other semiconductors [5,41,42], as well as for CdS [43]. It is was N719 (Solaronix), which gives a reddish color to the DSC
stable in a wide range of pH values in aqueous environments and high absorption across the visible range.
and it can be deposited in a conformal and dense layer by Given that the DSC has the widest absorption range –
atomic layer deposition. Despite previous reports using TiO2 Figures 2d and S8b – The DSC was placed on the backside
as protective layer of CdS, the low Evb of TiO2 (Figure S7) of the photoanode. This configuration originated higher
imposes an undesired potential step for holes diffusion to photocurrents, in the range of 1 mA cm
2 and photopoten-
the electrolyte, thus a resistive behavior is observed. In an tials ca. 1.3 V (Figure 6b); a SOC of 70% could be achieved.
Unbiased solar energy storage. 403

Table 1 Summary of PEC performance for the different ALD layers investigated.

Sample ID Anode/Cathode Configuration J at 0 V (mA cm
2) J/J0 after 5 min at 0 V (%)

1 Anode 200 nm CdS 0.47 46

2 Anode 200 nm CdS/200 nm CdSe 1.40 86
3 Anode 200 nm CdS/200 nm CdSea 0.66 42
4 Anode TiO2b/ 200 nm CdS/200 nm CdSe 0.12 63
5 Cathode 200 nm CdS/200 nm CdSe/100 nm TiO2c
1.9 47
6 Anode 200 nm CdS/200 nm CdSe/100 nm TiO2c 0.4 26
a
Back illumination.
b
Deposited from TiO2 paste.
c
Deposited using atomic layer deposition.

2.0

1.8

1.6

1.4
J / mA cm-2

1.2

1.0 t / min

0.8

0.6

0.4

0.2

0.0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2
E/V

Figure 5 Photocurrent–photovoltage characteristics of the

best CdS/CdSe sample. Inset shows stability.

Conclusions

In this work we demonstrated the possibility of storing solar

energy in a simple way using a vanadium redox flow battery
(RFB) and a CdS photoanode; unbiased charged of a solar-
RFB (CdS(s)|V3 + , VO2 + ||V3 + , V2 + |Carbon Felt(s), E0 = 0.6
VNHE) was achieved. A good semiconductor/electrolyte
match was achieved allowing the unbiased solar charge of
a RFB up to high state of charge (SOC). CdSe protective over
layer had a twofold increase with photocurrents up to
1.4 mA cm
2 with no external bias. A tandem system
battery using a dye-sensitized solar cell was also studied Figure 6 (a) Schematic representation of a tandem device
to provide necessary photovoltage to charge a standard all comprising a DSC and a CdS photoanode to charge a vanadium
vanadium redox flow (DSC/CdS(s)|VO2 + , VO2+ ||V3 + , V2 + | RFB; (b) Photocurrent–photovoltage characteristics of a tandem
Carbon Felt(s), E0 = 1.2 VNHE); in this configuration, the system composed by CdS(s)/DSC|VO2 + , VO2+ ||V3 + , V2 +
observed photocurrent was 1 mA cm
2. |Carbon Felt(s).
Given the acidic nature of the electrolytes (pHE 0), the
photoanodes are still limited by their long-term stability. To useful in minimizing CdS photocorrosion [53]. It is antici-
improve further the performance of the proposed solar-RFB pated that new approaches will be proposed and photo-
system, the CdS/CdSe junction should be further optimized electrodes able to delivery higher current densities and
to produce more reproducible photoelectrodes displaying photopotentials will be soon available. For example, organic
higher photopotentials, currents and stability. Another alkaline electrolytes are currently being tested that allow
option is reduced graphene oxide, that have been proven the use of stable semiconductors such as hematite.
404 J. Azevedo et al.

Acknowledgments [20] M. Yu, X. Ren, L. Ma, Y. Wu, Nat. Commun. 5 (2014) 1–17.
[21] M. Yu, W.D. McCulloch, D.R. Beauchamp, Z. Huang, X. Ren,
J. Azevedo and C. Sousa are grateful to the FCT SFRH/BD/ Y. Wu, J. Am. Chem. Soc. 137 (2015) 8332–8335.
79207/2011 PhD grant and postdoctoral grant SFRH/BPD/ [22] D. Liu, Z. Wei, C.-J. Hsu, Y. Shen, F. Liu, Electrochim. Acta 136
(2014) 435–441.
82010/2011, respectively. A. Bentien acknowledges the
[23] Z. Wei, D. Liu, C. Hsu, F. Liu, Electrochem. Commun. 45 (2014)
Danish Council for Independent Research (Technology and
79–82.
Production Sciences, Grant # 4005-00517B) for funding. This [24] P. Alotto, M. Guarnieri, F. Moro, Renew. Sust. Energ. Rev 29
work was supported by research project BI-DSC (grant (2014) 325–335.
agreement no. 321315) funded by European Research Coun- [25] K.W. Knehr, E.C. Kumbur, Electrochem. Commun. 13 (2011)
cil. We also thank José Nogueira (LEPABE) for DSC fabrica- 342–345.
tion, Dr. Verena Stockhausen (LEPABE) and João Barros for [26] N. Arroyo-Curras, J.W. Hall, J.E. Dick, R.A. Jones, A.J. Bard, J.
graphical design. Electrochem. Soc. 162 (2014) A378–A383.
[27] 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,
Appendix A. Supplementary material Science 349 (2015) 1529–1532.
[28] Rasband, W.S., ImageJ, U. S. National Institutes of Health,
Supplementary data associated with this article can be Bethesda, Maryland, USA, 〈〈http://imagej.nih.gov/ij/〉〉, 1997–
found in the online version at http://dx.doi.org/10.1016/ 2014.
j.nanoen.2016.02.029. [29] C. Lu, L. Zhang, Y. Zhang, S. Liu, G. Liu, Appl. Surface Sci. 319
(2014) 278–284.
[30] M. Takahashi, S. Hasegawa, M. Watanabe, T. Miyuki, S. Ikeda,
K. Iida, J. Appl. Electrochem. 32 (2002) 359–367.
References [31] G. Sasikala, R. Dhanasekaran, C. Subramanian, Thin Solid Films
302 (1997) 71–76.
[1] H. Lund, Energy 32 (2007) 912–919. [32] J. Azevedo, C.T. Sousa, J. Ventura, A. Apolinário, A.
[2] R. van de Krol, M. Gratzel, Photoelectrochemical Hydrogen M. Mendes, J.P. Araujo, Mater. Res. Express 1 (2014) 1–16.
Production, Springer Science & Business Media, New York, [33] F. Goto, K. Shirai, M. Ichimura, Sol. Energ. Mater. Solar Cells 50
2011. (1998) 147–153.
[3] T.N. Murakami, N. Kawashima, T. Miyasaka, Chem. Commun. [34] E. Fatás, R. Duo, P. Herrasti, F. Arjona, E. Garcia-Camarero, J.
26 (2005) 3346. Electrochem. Soc. 131 (1984) 2243–2246.
[4] M. Skunik-Nuckowska, K. Grzejszczyk, P.J. Kulesza, L. Yang, [35] J. Azevedo, C.T. Sousa, A.M. Mendes, J.P. Araujo, J. Nanosci.,
N. Vlachopoulos, L. Häggman, E. Johansson, A. Hagfeldt, J. Nanotech 12 (2012) 9112–9117.
Power Sources 234 (2013) 91–99. [36] T. Lopes, P. Dias, L. Andrade, A.M. Mendes, Sol. Energy Mater.
[5] J. Azevedo, L. Steier, P. Dias, M. Stefik, C.T. Sousa, J.P. Araujo, Solar Cells 128 (2014) 399–410.
A.M. Mendes, M. Gratzel, S.D. Tilley, Energy Environ. Sci. 7 [37] S.C. Rodrigues, R. Whitley, A.M. Mendes, J. Membr. Sci. 459
(2014) 4044–4052. (2014) 207–216.
[6] T. Zhu, M.N. Chong, Nano Energy 12 (2015) 347–373. [38] D.N. Buckley, X. Gao, R.P. Lynch, N. Quill, M.J. Leahy, J.
[7] J.Y. Kim, G. Magesh, D.H. Youn, J.-W. Jang, J. Kubota, Electrochem. Soc. 161 (2014) A524–A534.
K. Domen, J.S. Lee, Sci. Rep. 3 (2013) 2681. [39] J. Maçaira, L. Andrade, A.M. Mendes, RSC Adv. 4 (2014)
[8] M. Zhong, T. Hisatomi, Y. Kuang, J. Zhao, M. Liu, A. Iwase, 2830–2844.
Q. Jia, H. Nishiyama, T. Minegishi, M. Nakabayashi, N. Shibata, [40] S.D. Tilley, M. Cornuz, K. Sivula, M. Gratzel, Angew. Chem.
R. Niishiro, C. Katayama, H. Shibano, M. Katayama, A. Kudo, Int. Ed. 49 (2010) 6405.
T. Yamada, K. Domen, J. Am. Chem. Soc. 137 (2015) [41] S.D. Tilley, M. Schreier, J. Azevedo, M. Stefik, M. Graetzel,
5053–5060. Adv. Funct. Mater. 24 (2014) 303–311.
[9] O.V. Marchenko, S.V. Solomin, Int. J. Hydrog. Energy 40 (2015) [42] X. Li, P.S. Bassi, P. Boix, Y. Fang, L.H. Wong, ACS Appl. Mater.
3801–3805. Interfaces 7 (2015) 16960–16966.
[10] J.R. McKone, N.S. Lewis, H.B. Gray, Chem. Mater. 26 (2014) [43] A.S. Aliev, M.N. Mamedov, M.T. Abbasov, Inorg. Mater. 45
407–414. (2009) 965–967.
[11] M.G. Walter, E.L. Warren, J.R. McKone, S.W. Boettcher, Q. Mi, [44] Y.-L. Lee, C.-F. Chi, S.-Y. Liau, Chem. Mater. 22 (2010)
E.A. Santori, N.S. Lewis, Chem. Rev. 110 (2010) 6446–6473. 922–927.
[12] H. Chen, T.N. Cong, W. Yang, C. Tan, Y. Li, Y. Ding, Prog. Nat. [45] S. Cheng, W. Fu, H. Yang, L. Zhang, J. Ma, H. Zhao, M. Sun,
Sci. 19 (2009) 291–312. L. Yang, J. Phys. Chem. C 116 (2012) 2615–2621.
[13] M. Sharon, P. Veluchamy, C. Natarajan, D. Kumar, Electro- [46] H.-S. Rao, W.-Q. Wu, Y. Liu, Y.-F. Xu, B.-X. Chen, H.-Y. Chen,
chim. Acta 36 (1991) 1107–1126. D. Kuang, C.-Y. Su, Nano Energy 8 (2014) 1–8.
[14] S. Licht, G. Hodes, R. Tenne, J. Manassen, Nature 326 (1987) [47] Y.-F. Xu, W.-Q. Wu, H.-S. Rao, H.-Y. Chen, D. Kuang, C.-Y. Su,
863–864. Nano Energy 11 (2015) 621–630.
[15] A.Z. Weber, M.M. Mench, J.P. Meyers, P.N. Ross, J.T. Gostick, [48] S. Chen, L.-W. Wang, Chem. Mater. 24 (2012) 3659–3666.
Q. Liu, J. Appl. Electrochem. 41 (2011) 1137–1164. [49] X. Wang, G. Liu, G.Q. Lu, H.-M. Cheng, Int. J. Hydrog. Energy
[16] W. Wang, Q. Luo, B. Li, X. Wei, L. Li, Z. Yang, Adv. Funct. 35 (2010) 8199–8205.
Mater. 23 (2012) 970–986. [50] X. Shi, K. Zhang, K. Shin, M. Ma, J. Kwon, I.T. Choi, J.K. Kim, H.
[17] S. Er, C. Suh, M.P. Marshak, A. Aspuru-Guzik, Chem. Sci. 6 K. Kim, D.H. Wang, J.H. Park, Nano Energy 13 (2015) 182–191.
(2015) 885–893. [51] D. Gurudayal, M.H. Sabba, L.H. Kumar, J. Wong, M. Barber,
[18] B. Yang, L. Hoober-Burkhardt, F. Wang, G.K. Surya Prakash, S. N. Gratzel, Mathews, Nano Lett. 15 (2015) 3822–3839.
R. Narayanan, J. Electrochem. Soc. 161 (2014) A1371–A1380. [52] F.F. Abdi, L. Han, A.H.M. Smets, M. Zeman, B. Dam, R. van de
[19] B. Huskinson, M.P. Marshak, C. Suh, S. Er, M.R. Gerhardt, C. Krol, Nat. Commun. 4 (2013) 1–18.
J. Galvin, X. Chen, A. Aspuru-Guzik, R.G. Gordon, M.J. Aziz, [53] Y. Tang, X. Hu, C. Liu, Phys. Chem. Chem. Phys. 16 (2014)
Nature 505 (2015) 195–198. 25321–25329.
Unbiased solar energy storage. 405

João Azevedo obtained a master’s degree Anders Bentien is a permanent staff

in Experimental Physics at Porto University, member at the Department of Engineering
Portugal, in 2011. Currently he is a Physics of AU section for Bio- & Chemical Engineer-
PhD candidate at Porto University and ing. He obtained his Ph.D. in Physics in 2004
researching on solar energy storage both at Max Planck Institute for Chemical Physics
at LEPABE, Laboratory for Process Engineer- of Solids of Dresden Technical University.
ing, Environment, Biotechnology and Energy Current research topics within membranes
and IFIMUP-IN, Institute of Nanoscience and and energy conversion include synthesis/
Nanotechnology. His research centers on the characterization of ion conducting polymer
production of solar fuels, with a new photo- membranes for electrochemical energy
eletrode development to improve the efficiency of water splitting conversion applications, electrochemical compressors, redox flow
processes and solar batteries. batteries and solar charging of redox flow battery electrolytes with
photoelectrochemical cells.
Thorsten Seipp (born 1986) studied Chemi-
cal Engineering in Dortmund. His Diploma J. P. Araújo obtained his PhD in Physics at the
Thesis dealed with scale-up of Redox Flow University of Porto in 2002 and became
Batteries. Since 2011 he is scientific Assistant Professor at the Physics and Astron-
researcher at Fraunhofer Institute UMSICHT omy Department at the Faculty of Sciences of
in Oberhausen with focus on redox-flow- University of Porto in the same year. He is now
batteries. Since 2015 he is furthermore Vice-President of IFIMUP-IN, Institute of
managing director of Volterion GmbH, a Nanoscience and Nanotechnology and counts
spin-off company of Fraunhofer, which pro- with over 200 papers in international journals.
duces and sells compact redox-flow-stacks He has relevant contributions in the field
and batteries for decentralized energy storage. of magnetic intermetallics and oxides, and
recently, in new phenomena and new technologies based on the
Jens Burfeind graduated in Chemistry. He magnetism of nanostructured materials.
exhibits over 13 years experience in elec-
trochemistry. From the year 2000 he devel- Adélio Mendes is full professor at FEUP and
oped fuel cells and electrolyzer at h2- researcher at LEPABE, where he leads the
interpower. The last 6 years Dr. Burfeind Energy, Processes and Products area. His
was head of high temperature PEM working publications include over 250 papers in
Group at center for fuel cell technology peer-reviewed scientific journals and
(ZBT). From March 2011 to February 2012 he 5 reviews or invited book chapters. He is
worked at ReVolt GmbH in the field of zinc the inventor or co-inventor of over 20
air batteries. Since March 2012 Dr. Burfeind patents and authored one book. Prof.
works at Fraunhofer Institute UMSICHT as group manager for Mendes is presently the coordinator of
Electrochemical Storage and Processes. He is responsible for the project BeingEnergy (Grant agreement no:
Vanadium-Redox-Flow Battery development and other electroche- 303476), project BI-DSC (advanced research grant from ERC) and
mical processes especially in unconventional solvents. GOTSolar (Grant agreement no: 687008). He was awarded with Air
Products Faculty Excellence Award (2011), ACP Diogo Vasconcelos
Célia Sousa is specialized in nanoscience (2011) and Solvay & Hovione Innovation Challenge 2011 (SHIC'11)
and nanotechnology and her interdisciplin- prizes.
ary research has been mainly focused on the
fabrication of nanostructures using top-
down methods for biomedical and fuel cells
technological applications. She completed
her PhD degree in Physics–Nanotechnology
in 2011 and has an h-factor of 13, published
48 papers in international peer review
journals. She is also author of an invited
review article, 2 book chapters and 9 proceedings. Her research led
to 3 invited talks, 23 oral presentations and 30 posters in interna-
tional conferences.