Article [Link]

1038/s41467-023-41187-9

Monolithic FAPbBr3 photoanode for

photoelectrochemical water oxidation with
low onset-potential and enhanced stability
Received: 3 June 2023 Hao Yang1,5, Yawen Liu2,5, Yunxuan Ding3, Fusheng Li 4, Linqin Wang3, Bin Cai2,
Fuguo Zhang1, Tianqi Liu 1, Gerrit Boschloo2, Erik M. J. Johansson 2 &
Accepted: 25 August 2023
Licheng Sun 1,3,4

Check for updates Despite considerable research efforts on photoelectrochemical water splitting
1234567890():,;
1234567890():,;

over the past decades, practical application faces challenges by the absence of
efficient, stable, and scalable photoelectrodes. Herein, we report a metal-
halide perovskite-based photoanode for photoelectrochemical water oxida-
tion. With a planar structure using mesoporous carbon as a hole-conducting
layer, the precious metal-free FAPbBr3 photovoltaic device achieves 9.2% solar-
to-electrical power conversion efficiency and 1.4 V open-circuit voltage. The
photovoltaic architecture successfully applies to build a monolithic photo-
anode with the FAPbBr3 absorber, carbon/graphite conductive protection
layers, and NiFe catalyst layers for water oxidation. The photoanode delivers
ultralow onset potential below 0 V versus the reversible hydrogen electrode
and high applied bias photon-to-current efficiency of 8.5%. Stable operation
exceeding 100 h under solar illumination by applying ultraviolet-filter pro-
tection. The photothermal investigation verifies the performance boost in
perovskite photoanode by photothermal effect. This study is significant in
guiding the development of photovoltaic material-based photoelectrodes for
solar fuel applications.

Artificial photosynthesis is a system that uses artificial materials to electrochemical (EC) reaction to produce hydrogen; the other is the
mimic the conversion of solar energy to chemical energy in nature. As direct approach which takes inspiration from the working principles of
solar energy is the most abundant energy source on Earth, using water natural photosynthesis, and converts solar energy into hydrogen and
as a raw material to produce hydrogen with solar energy is an ideal oxygen through the photoelectrochemical (PEC) water splitting with-
approach to obtaining renewable fuels, known as the solar fuel out external power4.
pathway1–3. Developing efficient, stable, and low-cost water-splitting Figure 1c lists the band structure of typical materials for PEC water
devices is one of the key scientific challenges to achieving large-scale splitting. Although visible (vis.) light-absorbing materials, including
electrolysis and photolysis of water for hydrogen production. Cur- WO3, BiVO4, and Fe2O3, have been extensively studied due to favorable
rently, there are two main artificial photosynthesis approaches for band edge positions, their limited charge separation/transport prop-
converting solar energy into fuels: one is the indirect approach, which erties remain major obstacles to achieving commercially viable
uses the electricity generated by a photovoltaic (PV) device to drive an performance5. Most PEC cells with single-junction semiconductor

1
Department of Chemistry, School of Engineering Sciences in Chemistry, Biotechnology and Health, KTH Royal Institute of Technology, 10044
Stockholm, Sweden. 2Department of Chemistry-Ångström, Physical Chemistry, Uppsala University, 75120 Uppsala, Sweden. 3Center of Artificial Photo-
synthesis for Solar Fuels and Department of Chemistry, School of Science, Westlake University, 310024 Hangzhou, China. 4State Key Laboratory of Fine
Chemicals, Institute of Artificial Photosynthesis, DUT-KTH Joint Education and Research Centre on Molecular Devices, Dalian University of Technology, 116024
Dalian, China. 5These authors contributed equally: Hao Yang, Yawen Liu. e-mail: [Link]@[Link]; lichengs@[Link]

Nature Communications | (2023)14:5486 1

Article [Link]

a 4e- b 10 )
work
FAPbBr 3 (this

Current density (mA cm-2)

020)
Ta 3N 5 (2
8

)
023
6 O 4 (2
BiV
17 )
( 20
O3
Fe 2
hv 2H2O +
4H + O2
4
020)
WO 3 (2

2 TiO 2 (2018)

Epoxy

0
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
ITO glass
SnO2
FAPbBr3
Carbon
Graphite
NiFe alloy
NiFe LDH
E (V vs. RHE)

c -1 d -1

E vs. NHE (pH=13.6)

E vs. Vacuum

E vs. Vacuum
-3 -3
-2
-2 -2

-2.9
-3.0
-3 -2

-3.2
-3.8

-4.0
-4.0

-1
1.1 -4.1

-3.4

-3.4
-3.4
-4.2
-4.2

-3

-3.6
-4.5

-4

-3.9
H2O/H2
-4.7

-4.0
-4.8

-1
-4.9

1.4

0
-5 -4 H2O/H2
1.7
2.1

2.4
O2/H2O

2.9
1 0
-5.2

1.8
2.2
-5.2

2.4
2.9

2.2
3.0
3.0

1.5
2.4

1.4
2.7

2.2

-6 -5
2.8

-5.9

2 O2/H2O
-6.1

E vs. NHE (pH=13.6)

-6.4

-7 1

-5.4

-5.4
-5.4
-6.9

-5.6
-5.6
3
-7.1

-6
-7.2

MAPbCl 3 -5.8

-5.8
-7.4

FAPbCl 3 -6.0
-8
-7.6

CsPbCl 3 -6.1
2
4

MAPbBr 3
-7

FAPbBr 3
CsPbBr 3

MAPbI3
-9

FAPbI3
CsPbI 3
BiVO4

Fe2O3

Ta3N5

CdSe
Bi 2O3

CdTe

3
WO3

5
TiO2

CdS
Si

-10 -8

Fig. 1 | Comparing band structures and photoanode performance. a Schematic from the corresponding literature. Band positions of (c) typical PEC and (d) lead
illustration of FAPbBr3 photoanode for oxygen evolution. b Comparison of the halide perovskite semiconductors in the pH 13.6 aqueous electrolyte compared
linear sweep voltammetry (LSV) curves between the FAPbBr3 photoanode and the with the energy potential for water splitting reaction. red=valence band edge;
best-reported photoanodes made from a single junction absorber layer, including blue=conduction band edge. Tabulated values are retrieved from references7,18,61.
TiO2, WO3, Fe2O3, BiVO4, and Ta3N5, with catalysts layer. The LSV data was extracted

photoelectrodes, even those made from wide band gap (Eg) semi- when used as a single junction photoanode, their onset potential is
conductor materials such as TiO2 and WO3, show high current onset typically greater than 0.5 V vs. reversible hydrogen electrode (RHE)
potential due to the high thermodynamic requirements and the multi- (Supplementary Fig. 2). Meanwhile, the performance of chloride per-
proton and electron transfer processes involved in water oxidation ovskites is limited by their wide band gap, which leads to extremely low
reactions. As such, the development of a perfect semiconductor that is visible light absorption. Therefore, bromide perovskites with suitable
stable, low-cost, has suitable band structures, high charge carrier energy band structures are ideal components for the preparation of
mobility, and high absorption coefficient remains a considerable low-onset potential photoanodes. Out of the three mono-cation bro-
challenge. At the same time, better water oxidation catalysts, as well as mide perovskites, cesium lead bromide (CsPbBr3) has the larger band
reliable, stable, and economical catalyst loading techniques, are also gap at 2.4 eV, i.e., the lowest theoretical saturation current. Addition-
highly desired to prepare high-efficiency water oxidation ally, its inorganic counterparts require high-temperature annealing
photoanodes. over 300 °C for stabilization, which makes it difficult to obtain high-
Alternatively, the PV material (Si, GaAs, GaInP) with excellent crystallinity films11. On the other hand, methylammonium lead bro-
optical and electronic properties can be arranged as an absorber layer mide (MAPbBr3) components have poor thermal stability and are not
in an integrated solar hydrogen device, analog as the semiconductor suitable for long-term high-temperature operations. Therefore, after a
layer in traditional PEC configuration. Among PV absorbers, lead halide comprehensive consideration of the preparation difficulty, band
perovskite (ABX3) materials have emerged as a promising alternative structure, and material stability, formamidinium lead bromide
to overcome the limitations of metal-oxide-based photoelectrodes (FAPbBr3) is selected in this study to be the top choice for producing
due to their exceptional optoelectronic properties6. These materials low-onset potential photoanodes. The photovoltaic materials-based
offer several advantages over traditional metal-oxide-based photo- photoelectrochemical (PVM-PEC) cell can immerse in an electrolyte
electrodes, including higher light absorption coefficients, facile exci- after coating with waterproof layers12. Among the strategies reported,
ton dissociation ability, and longer carrier diffusion lengths7. Poli et al. utilize a mesoporous carbon/graphite sheet structure as
Additionally, their unique crystal structure enables facile tuning of protective layers on a CsPbBr3-based photoanode13. This simple, cost-
their electronic properties through compositional variations, making effective encapsulation method provides stable protection of the
them an attractive platform for the development of high-performance absorptive layer against moisture, while maintaining high con-
photoelectrochemical systems. As such, perovskite materials are a ductivity. This approach offers promising opportunities for the use of
subject of intense research and development in the field of solar fuel water-sensitive perovskite materials in integrated PEC cells.
production8–10. Lead halide perovskite materials with iodine compo- Herein, we present the first efficient and stable FAPbBr3-based
nents have a high theoretical saturation current due to their narrow photoanode for solar water oxidation. This FAPbBr3 absorber was
band gap (Fig. 1d). However, it is challenging for a single cell to gen- initially employed in mesoporous carbon solar cells, yielding effi-
erate enough photoelectric voltage for water oxidation. Therefore, ciencies of 9.16% and open-circuit voltages of up to 1.4 V. Furthermore,

Nature Communications | (2023)14:5486 2

Article [Link]

the FAPbBr3 absorber, after encapsulated with graphite and resin/ provide statistical data of 20 devices, indicating that the average onset
epoxy, was utilized to produce a photoanode for water oxidation potential is −0.066 V vs. RHE, and the average photocurrent density at
(Fig. 1a). The optimized photoanode with integrated carbon/graphite/ 1.23 V vs. RHE is approximately 8.7 mA cm−2. The photoanode exhibits
alloy interlayers and electrodeposited NiFe layered double hydroxide a hysteresis effect similar to a solar cell, with a decay of about 50 mV on
(LDH) catalyst subsequently demonstrates a remarkable performance the back scanning of the onset potential; additionally, the cyclic vol-
and stability: the device outperforms oxide photoanodes by exhibiting tammetry (CV) curves show almost no scan rate dependency (Sup-
an ultralow onset potential below 0 V vs. RHE and achieving a max- plementary Fig. 18). As shown in Fig. 2d, the incident photon-to-current
imum applied bias photon-to-current efficiency (ABPE) of 8.5%. The efficiencies (IPCEs) of FAPbBr3 photoanode were measured and cal-
performance of FAPbBr3 system far exceeds that of the best-reported culated under different potentials. The IPCE values are approximately
single-junction absorber photoanodes to date (Fig. 1b), including 80 % and 90 % at 0.23 and 1.23 V vs. RHE, respectively. Integrating the
TiO214, WO315, Fe2O316, BiVO417, and Ta3N518. Additionally, it maintains IPCE curves over the AM 1.5 G solar spectrum, the theoretical photo-
95% of its performance for 100 h of operation, surpassing that of all current densities are estimated to be 7.4 mA cm−2 at 0.23 V vs. RHE and
known reported perovskite photoanodes. We further discuss the 8.3 mA cm−2 at 1.23 V vs. RHE. These calculated values are consistent
ultraviolet (UV) degradation of the perovskite interface and emphasize with the measured data in Fig. 2b, which are 7.8 mA cm−2 at 0.23 V and
the feasibility of utilizing the photothermal effect to enhance cell 8.5 mA cm−2 at 1.23 V. The FAPbBr3 photoanode exhibits an ultralow
performance. The as-fabricated FAPbBr3 photoanode is produced overpotential and over 8 mA cm−2 saturation current, making it a per-
without the need for high temperature, inert atmosphere, or high fect match with photocathodes, such as CuO or GaInP2, which have a
vacuum evaporation techniques, providing promising opportunities saturation current of around 10 mA cm−219,20. When subjected to
for the development of large scale high-performance, low-cost per- chopped-light illumination, the photoanode exhibits notable current
ovskite photoanodes for water oxidation. spikes around onset potentials in both LSV and chronoamperometry
current due to the slow water oxidation kinetics induced carrier
Results recombination on the surface (Fig. 2c and Supplementary Fig. 19). The
Photoanode preparation and characterization excess charge may be dissipated through radiative recombination,
As illustrated in Fig. 2a, the wide band gap FAPbBr3 perovskite pho- which can be detrimental to the device’s lifetime. The spikes are sur-
toanode consists of two parts, the light-absorbing part (ITO/SnO2/ passed when the potential increases to 0.2 V vs. RHE, indicating that
FAPbBr3/carbon) and the water oxidation catalyst (WOC) part (gra- only a subtle additional driving force is needed for the FAPbBr3 pho-
phite sheet/NiFe alloy/NiFe LDH); the edge of the device is sealed with toanode to improve charge separation and catalytic efficiency, which
epoxy resin to prevent water degradation of FAPbBr3 material (Sup- greatly decrease the onset potential requirements of the counter
plementary Fig. 1). The light-absorbing part, which is a hole transport photocathodes in Z-scheme PEC cells. The steady-state current density
material (HTM)-free FAPbBr3 solar cell, exhibits outstanding perfor- of the FAPbBr3 photoanode is also consistent with LSV current density;
mance with a power conversion efficiency (PCE) of 9.16%, a high open- under the applied potentials of 0.2 and 1.2 V vs. RHE, steady current
circuit voltage (Voc) of 1.38 V, a short-circuit current density (Jsc) of densities of 7.20 and 8.46 mA cm−2 were achieved, respectively (Sup-
8.69 mA cm−2, and a fill factor (FF) of 76.04% (see Supplementary plementary Fig. 19). The amount of photogenerated O2 in the head-
Note 1 for details). Inspired by the work of Poli et al. 13, a commercially space was quantified using a pressure transducer. The number of
available 160 μm self-adhesive graphite sheet (GS) was placed on top electrons passing through the electrode agreed well with the amount
of the FAPbBr3/carbon layer to prevent water penetration into the of O2 detected, representing Faradaic efficiencies of 94.7% and 96.2 %
light-absorbing part. As demonstrated in Supplementary Fig. 17, for the at the potential of 0.23 and 1.23 V vs. RHE, respectively (Supplemen-
epoxy resin-encapsulated device with solely the mesoporous carbon tary Fig. 20).
layer, notable degradation was observed after just 5 min of water The ABPE was calculated from the corresponding LSV curve in a
immersion. Conversely, the device protected by the GS layer exhibited three-electrode system (Supplementary Fig. 21). A maximum ABPE of
minimal alterations even after 100 h of water immersion. The compact 8.52 % is achieved at 0.082 V vs. RHE. This ABPE value is compared to
and stacked GS separator offers excellent thermal (350 W m−1 K−1) and those reported systems in Fig. 2e. To the best of our knowledge, the
electrical conductivity while also providing more efficient protection value is one of the best of those reported photoanodes, including
for the perovskite from liquid water percolation. Moreover, the self- metal oxide, perovskite, polymer bulk heterojunction (BHJ), and
adhesive and commercially available characteristics of the GS simplify silicon-based photoanodes. Compared to bismuth vanadate-based
the device fabrication process, making it an ideal water separator photoanodes, which have a similar wide Eg, the FAPbBr3 photoanode is
component for the assembly of a perovskite photoanode. By utilizing far superior in terms of maximum ABPE. Moreover, considering the
facile electrodeposition methods, the WOCs could be attached to the ultralow potential at which the maximum ABPE value is achieved, the
conductive GS layer, which also allows individual optimization of the FAPbBr3 photoanode is one of the most promising candidates for
WOC component on the indium tin oxide (ITO)/GS substrate. The realizing efficient photo-driven total water splitting. Fig. 2f summarizes
overpotential requirements of the as-fabricated GS/NiFe alloy/NiFe the best-reported photovoltages for representative PEC and PV devi-
LDH electrode to reach a constant current density of 10 mA cm−2 is ces with varying band gaps. The photovoltage of PV devices is pre-
only 228 mV; the overpotential shows only an increment of 19 mV after sented as the open circuit potential, whereas the photovoltage of
100 h of catalysis (see Supplementary Note 2 for details). photoanodes is calculated as the difference between the photocurrent
onset potential and the Nernstian potential for water oxidation (1.23 V
Photoelectrochemical performance of photoanode vs. RHE)21. Analog to trends of PV devices, a wide band gap absorber
After conducting detailed studies on the performance, stability, mor- (>2.0 eV) should be capable of photovoltages approaching >1.4 V when
phology, and structure of the WOC part, highly active GS/NiFe alloy/ considering a free energy loss of 0.6 eV21, while it is clear that tradi-
NiFe LDH layers were applied to the light-absorbing part to construct a tional oxide or nitride wide band gap absorbers have significant room
photoanode with an architecture of ITO/SnO2/FAPbBr3/carbon/GS/ for photovoltage improvement to reach the goal. To the best of our
NiFe alloy/NiFe LDH (i.e., FAPbBr3 photoanode). The digital images of knowledge, the wide band gap, HTM-free FAPbBr3 photoanode pre-
FAPbBr3 photoanode are illustrated in Fig. 2a. Fig. 2b shows the LSV sented in this work has achieved the lowest onset potential (around
curve of the as-fabricated FAPbBr3 photoanode; the photocurrent −0.12 V) and record photovoltage (around 1.35 V) among single junc-
sharply rose from −0.12 V vs. RHE and reached a saturated photo- tion photoanodes to date, which brings the photovoltage much closer
current density over 8.5 mA cm−2 at 1.23 V vs. RHE. The inset figures to the optimal value (>1.6 V). The ultralow onset potential greatly

Nature Communications | (2023)14:5486 3

Article [Link]

a b 10
Light
Dark

Current density (mA cm-2)

8 @ 1.23 V vs. RHE

Current density (mA cm-2)

10
Glass side
6
O2
hv 8
Onset potential
4
H2O
NiFe side 6
2
ITO Glass

SnO2
FAPbBr3

Carbon
Graphite

NiFe LDH
Alloy

-0.4 -0.2 0.0 0.2 0.4

E (V vs. RHE)
0
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
E (V vs. RHE)
c 10 d 100 10

Integrated current density (mA cm-2)

Scaning current with chopped irradiation
Steady state current
Current density (mA cm-2)

8 80 8
Integrated current density (1.23 V vs. RHE)
6 IPCE (%) Integrated current density (0.23 V vs. RHE)
IPCE (1.23 V vs. RHE)
60 IPCE (0.23 V vs. RHE)
6
4
40 4
2
20 2
0

-2 0 0
-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 350 400 450 500 550 600
E (V vs. RHE) Wavelength (nm)
# Hole transport materials-free device
e 10 Best device 8.5% @ 0.08 V
f 2.0 Band gap
FA0.93MA0.07PbI3/NF.3D [Link]
SQ limit
FAPbBr3/NiFe alloy/NiFe LDH

8
1.5
Photovoltage (V)

GaInP (This work)

Eg ≈ 1.5 - 1.6 eV FAPbBr3 # SQ - 1V
H2O/O2
ABPE (%)

Perovskite
6 Eg ≈ 2.3 eV
FA0.83Cs0.17Pb(I0.8Br0.2)3/Ni GaAs
(narrow band gap)
CdTe PNDITCVT/PBDTTTPD
FA0.80MA0.15Cs0.05PbI2.55Br0.4/FeNi(OH)x E
g ≈ 1.1 eV 1.0 FA0.80MA0.15Cs0.05PbI2.55Br0.4
BiVO4
n+p-Si/NiFe-LDH
FA0.93MA0.07PbI3 FA0.83Cs0.17Pb(I0.8Br0.2)3
Eg ≈ 2.4 eV CsPbBr3 (2023) # WO3
4 MAPbI3/Ni np+-Si/NiO /NiFe
MAPbI3
Fe2O3
BiVO4/Co3O4/FeOOH /NiOOH x Si
CsPbBr3/NiFeOOH Ta3N5 CsPbBr3 (2019) #
BiVO4/NiFeOx n-Si/NiFe MAPbI3
MAPbI3/Ni FA0.93MA0.07PbI3
BaTaO2N
BiVO4/Ni-N4-O/FeOOH
n-Si/Ni/NiOx/NiOOH
0.5 ZnFe2O4
2 BiVO4 photoanodes BiVO /Co@CB[5]
4 PV cell
n-Si photoanodes BiVO4/g-C3N4/CoOOH n-Si/Ni/NiOOH/NiFe # (5-AVA)x(MA)1−xPbI3
Other Perovskite photoanodes CsPbBr3/Ir WOC
PEC photoanode
BiVO4/Co4O4 cubane
BHJ photoanode n-Si/NiFeP Perovskite photoanode
FTO/mZnO/BHJ/PTAA/LIO
FAPbBr3 photoanodes (5-AVA)x(MA)1−xPbI3/Carbon BHJ photoanode
0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.0 1.5 2.0 2.5 3.0
E (V vs. RHE) Band gap (eV)
Fig. 2 | Assessing and comparing FAPbBr3 photoanode performance. a Device benchmarks of as-prepared FAPbBr3 photoanodes and other perovskite, BiVO4, BHJ
structure of FAPbBr3 photoanode in this work. Inset: digital images of as-prepared and silicon-based photoanodes. Tabulated values and literature references for lis-
photoanode. b LSV curve of FAPbBr3 photoanode in 1.0 M KOH (scan rate: ted photoanodes are provided in Supplementary Table 4. f Photovoltage bench-
50 mV s−1). Inset: statistical photoelectrochemical parameters (20 devices) of the marks for PEC and PV materials as a function of the optical band gap. Diagonal lines
onset potential and saturated catalytic current density at 1.23 V vs. RHE. c LSV curve represent the band gap, the Shockley-Queisser (S-Q) photovoltage limit, and the
of FAPbBr3 photoanode under chopped irradiation (red curve) and current density- S-Q limit minus 1 V. Tabulated values of PV cells and PEC anodes were retrieved
voltage response of FAPbBr3 photoanode obtained under steady-state current from reference21; Tabulated values and literature references for other perovskite
conditions (blue dots) in 1.0 M KOH. d IPCEs and corresponding integral current and BHJ anodes are provided in the Supplementary Tables 5 and 6.
densities (AM 1.5 G) of FAPbBr3 photoanode at 1.23 V and 0.23 V vs. RHE. e ABPE

Nature Communications | (2023)14:5486 4

Article [Link]

Fig. 3 | Band structure analysis of FAPbBr3 photoanode. a UPS spectrum of defective FAPbBr3 perovskites. c The band alignment structures of FAPbBr3 pho-
FAPbBr3 film. b Geometric structures and calculated TDOS of pristine and Br- toanode under specific conditions.

reduces the requirements of the counter photocathode, making it to reveal the energy band structure. Specifically, the projected density
more feasible to construct a low-cost Z-scheme water-splitting system of states (PDOS) in Supplementary Fig. 23 revealed that the conduction
with over 5% solar-to-hydrogen efficiency. Supplementary Fig. 22 band minimum (CBM) mainly consists of Pb 6p orbitals, while Br 4p
shows the values of solar cell Voc, photovoltage, and corresponding orbitals contribute to the valence band maximum (VBM); moreover,
voltage loss of various perovskite photoanodes for water oxidation. the trap state energy level of VBr is close to the CBM, as it is composed
Compared to all other reported photoanodes based on solar cell of the Pb 6p orbitals. The Fermi energy level is found to be shifted to
architecture, which exhibit a voltage loss of over 0.2 V, the FAPbBr3 the CBM, indicating the n-type property of Br vacancy due to the
photoanode shows a significantly lower voltage loss of only 0.05 V. presence of donor states within the band gap. The TDOS of other
This low voltage loss suggests good internal contact and well-matched vacancy and anti-site structures were also calculated accordingly. As
band structure in the device. shown in Supplementary Fig. 24, VPb, VFA, and BrPb will reduce the
To gain deep insight into the origin of the ultralow onset poten- Fermi level into the VBM, while PbBr will shift the Fermi level pene-
tial, ultraviolet photoelectron spectroscopy (UPS) was used to explore trating into CBM. The defect formation energies (DFEs) for the afore-
the band alignments in devices (Fig. 3a). The Fermi level of the FAPbBr3 mentioned defect types were calculated according to the procedures
film is measured to be approximately −3.7 eV vs. vacuum, and the in Supplementary Note 3. As detailed in Supplementary Table 7, the
valence band energy of the film was found to be 1.9 eV lower than the DFE for PbBr was found to be 0.78 eV, which is higher than the −0.51 eV
Fermi level (i.e., −5.6 eV vs. vacuum), indicating an n-type semi- for the Br vacancy, which implies that PbBr is less stable in comparison
conductor feature of the FAPbBr3 absorber layer. Density functional to VBr. These theoretical results provide valuable insight into the effect
theory (DFT) calculations were performed to gain a detailed under- of vacancy and anti-site structures on the Fermi level of FAPbBr3
standing of the energy states and n-type features of the as-fabricated materials. Based on the results, it is suggested that the vacancy
FAPbBr3 materials. The lattice parameters and band gap of pristine structure of Br leads to an n-type property of the FAPbBr3 film.
FAPbBr3 were calculated to be 6.02 Å and 2.41 eV, respectively, which The energetic situations at the n-type FAPbBr3/electrolyte inter-
agreed well with the reported experiment values (6.01 Å and face are presented in Fig. 3c. After the contact of the semiconductor
2.3 eV)22–24. As perovskite is made through a two-step process, the surface with the electrolyte, the thermodynamic equilibrium on both
vacancy state was considered to dominate the band structure due to sides of the interface is established with the formation of the depletion
non-stoichiometric components. Calculations were performed for layer. After light illumination, the quasi-Fermi level for electrons is
both the pristine structure and possible vacancy structures, including consistent with the Fermi level at equilibrium; in contrast, the quasi-
A-site vacancy (VFA), B-site vacancy (VPb), X-site vacancy (VBr), B anti- Fermi level of holes shifts downwards to the vicinity of the valence
site X (PbBr), and X anti-site B (BrPb). Fig. 3b presents the calculated band edge. In a PV device, the corresponding quasi-Fermi level split-
total density of states (TDOS) of pristine and X-site vacancy perovskite ting (QFLS) is directly related to the open circuit potential25. In this

Nature Communications | (2023)14:5486 5

Article [Link]

Fig. 4 | Stability assessment of FAPbBr3 photoanode. a Normalized chron- photoanode in oblique view. d Schematic diagram of the photoanode degradation
oamperometric measurement of FAPbBr3 photoanode at 1.23 V vs. RHE in 1.0 M process and digital images of pristine and tested FAPbBr3 photoanode. e SEM
KOH solutions (AM 1.5 G, 100 mW cm−2). Inset: digital images of photoanode at images for the glass part of tested FAPbBr3 photoanode in cross-section view. f LSV
different light-exposed times. b Normalized chronoamperometric measurement of curves of FAPbBr3 photoanode with different filters in 1.0 M KOH (scan rate:
FAPbBr3 photoanode at 1.23 V vs. RHE with different filters in 1.0 M KOH solutions 50 mV s−1). g AM 1.5 G spectrum and corresponding integral current densities for UV
(AM 1.5 G + UV or Vis. filter). Inset: digital images of photoanode at different light- (< 400 nm) and maximum absorption edge (<560 nm) regions. h Schematic dia-
exposed conditions. c SEM images for the carbon part of tested FAPbBr3 gram of the proposed mechanism for UV degradation of FAPbBr3 photoanode.

context, the QFLS of the FAPbBr3 perovskite layer should not be less Stability of photoanode
than 1.4 eV, as the perovskite layer can generate an open circuit Although the NiFe alloy/NiFe LDH-based catalysts as well as HTM-free
potential of about 1.4 V in the carbon-based PV device. In the FAPbBr3/ FAPbBr3 solar cells have already shown good stability under operation
electrolyte system, the flat-band potential is estimated to be the Fermi conditions, achieving long-term stability using integrated photoelec-
level of the FAPbBr3 film, which is −3.7 V vs. vacuum. The onset trodes remains a challenging task due to the harsh electrolyte envir-
potential of photocurrent can be considered to coincide with the flat- onment of water splitting. To investigate the long-term stability of
band potential when the band bending is less pronounced26. In our FAPbBr3 photoanodes, chronoamperometry with a constant applied
case, the theoretical onset potential is Eonset = (3.7 − 4.5) + 0.059 × potential was conducted in 1.0 M KOH solutions. As shown in Fig. 4a
13.6 = 0.02 V vs. RHE. Meanwhile, with an QFLS over 1.4 eV, the oxida- and Supplementary Fig. 27, the initial photocurrent exceeds
tion potential of holes reaches over 1.42 V, which could offer a suffi- 8.8 mA cm−2 under the applied potential of 1.23 vs. RHE. The device
cient overpotential of 0.19 V for apparent water oxidation reaction. exhibits a current half-life of over 55 h, which is better than other
The theoretical value of Eonset coincides with the measured Eonset of reported systems listed in Fig. 4a. Notably, the current shows no decay
−0.2 to 0.1 V vs. RHE in Fig. 2b, which confirms that the ultralow onset in the initial 12 h of illumination and remains at around 95% of its initial
potential in the FAPbBr3 photoanode is derived from the shallower current after 24 h of operation. During the measurement, light-
Fermi level and higher photovoltage (see Supplementary Note 4 for soaking-induced photocurrent enhancement is observed, where the
detailed discussions). The above analysis highlights the crucial role of photocurrent gradually increased with the initial 5 h of illumination13,27.
the Fermi level and band structure alignment of semiconductor The stability of the FAPbBr3 photoanode was also examined under a
materials in constructing high-performance, low-overpotential pho- lower applied potential of 0.3 vs. RHE. The current shows almost no
toanode devices. decay after 18 h of illumination with an initial current density of

Nature Communications | (2023)14:5486 6

Article [Link]

7.7 mA cm−2 (Supplementary Fig. 28). The inset images in Fig. 4a show Table 1 | Stability of the reported perovskite photoanodes for
photographs of the PEC cell after 24 and 55 h of photo-electrolysis. water oxidation
With time increasing, the orange active area of the photoanode gra-
Perovskite composition Year Stability data (Operation Ref.
dually turns white with the decayed photocurrent, indicating the gra- time/residual current)
dual degradation of the perovskite layer inside the sealed PEC cell (not MAPbI3 2016 0.5 h/64% 54
due to seal failure) since the degradation was only observed at the light
MAPbI3 2018 6 h/0% 55
illumination area (Supplementary Fig. S29a). The following investiga-
(5-AVA)x(MA)1−xPbI3a 2019 12 h/69% 56
tion aimed to reveal possibilities for improving the stability of degra-
ded photoanodes. To this end, deactivated photoanodes were FA0.80MA0.15Cs0.05PbI2.55Br0.45 2021 13 h/84% 57
disassembled to study the degradation mechanism. As shown in CsPbBr3a 2019 35 h/7% 13
Fig. 4d, the tested sample could be easily separated into two parts, with CsPbBr3a 2023 24 h/78% 70 h/77% 58
the graphite/carbon layer peeled off from the glass substrate (see FA0.83Cs0.17Pb(I0.8Br0.2)3 2021 47 h/0% 59
Supplementary Fig. 30 for details). Scanning electron microscope FA0.93MA0.07PbI3 2022 12 h/86% 48 h/50% 60
(SEM) was employed to observe the interface of the two peeled parts. FAPbBr3a 2023 12 h/100% 24 h/94% This
As shown in Fig. 4c, the perovskite layer tightly adheres to the carbon 60 h/43% work
layer; in contrast, only a small portion of perovskite crystal is attached FAPbBr3a (UV filter)b 2023 100 h/95% 125 h/57% This
to the ITO glass substrate (Fig. 4e). For the pristine sample, the ITO work
glass substrate should be an orange color with the growth of FAPbBr3 a
HTM-free device.
layer. However, after stability tests, the FAPbBr3 layer detached from b
After 72 h of operation, the device rested in the dark for 24 h before further testing.
the substrate, resulting in a transparent glass substrate. The detach-
ment of the perovskite layer from the ITO substrate under working
conditions was further confirmed in SEM analysis of selectively
exposed photoanode (Supplementary Fig. 29c). The functional groups rapidly under UV light, with a current half-life of only 2.2 h, confirming
of the pristine and degraded FAPbBr3 films were investigated using the sensitivity of the perovskite photoanode system to UV light. The
Fourier transform infrared (FT-IR) spectroscopy. As presented in color of the FAPbBr3 layer also turns white, indicating the rapid
Supplementary Fig. 31, a sharp peak at around 1716 cm−1 is observed in degradation of the ITO-SnO2/FAPbBr3 interface. In contrast, under UV-
the pristine FAPbBr3 film, which is attributed to the antisymmetric C-N free illumination, the photoanode is much more stable under operat-
vibration; as a reference, the antisymmetric C-N peak of pure FABr is ing conditions for 50 h, retaining almost 100% of the initial photo-
broad and centered at 1697 cm−1. The antisymmetric C-N peak of current and gradually decaying to 88% of the initial photocurrent for
degraded FAPbBr3 film is found to be blue-shifted and broader than an additional 24 h. More importantly, the photocurrent can be
that of the pristine sample, indicating the evolution of ITO-SnO2/ restored to the initial level after resting in the dark for 12 h at open
FAPbBr3 interface under light irradiation, with the possibility that the circuit conditions, as commonly observed in regularly structured
perovskite degrades into a corresponding monomer (i.e., FABr). perovskite solar cells with n-i-p architectures31–34. The reversible losses
The stability of lead halide perovskite solar cells is influenced by and dark recovery phenomenon are typically attributed to ion and
several factors, including heat, moisture, voltage, and UV light. Among defect migration within the perovskite layer and at the interfaces
these factors, the stability upon UV light exposure is particularly pro- between perovskite and selective contacts32. The system demonstrates
blematic due to the photocatalytic effect of electron transport mate- remarkable operational stability, with no current decay even after
rials like TiO2 and SnO2, which are considered to be the main reasons cumulative 96 h of electrolysis. Eventually, the photocurrent density
for perovskite degradation28. The theoretical oxidation potential of retains more than 57% of its initial level even after 125 h. As revealed in
bromine anion (Br−) is 1.087 V vs. normal hydrogen electrode, which is Supplementary Fig. 34 and the inset image in Fig. 4b, the current
equivalent to −5.59 eV vs. vacuum level29. The potential is almost the reduction should be due to a sealing failure in the electrolyte, which
same as the VBM of FAPbBr3 perovskite. Thus, under visible light led to moisture penetration on the edges. Nevertheless, the center of
irradiation, oxidation of Br− by excited holes is less likely to occur on the perovskite layer does not turn white, indicating its intrinsic stability
FAPbBr3 because such the reaction requires a certain additional driv- toward water oxidation catalysis under visible light, which also implies
ing force. However, as indicated in Fig. 4h, highly oxidative holes that further improvements in sealing materials and processes can
would accumulate close to the SnO2/perovskite interface when UV further enhance device stability. To the best of our knowledge, the as-
light exposure. These charges could extract electrons from the Br− in fabricated FAPbBr3 photoanode is the first lead halide perovskite-
the SnO2/FAPbBr3 junction, compromising the perovskite interface based photoanode with a stability of more than 100 h to date (Table 1).
structure and producing FABr or PbBr230. To further confirm the Although the use of a UV filter would cause an additional performance
negative impact of UV light, stability tests were conducted using a penalty, the saturation current of 7 to 8 mA cm−2 is still superior to
long-pass UV filter with a cut-off wavelength of >400 nm and a band- oxide-based photoanodes like BiVO4 and WO3, making FAPbBr3 per-
pass visible light filter with a range of 200–400 nm (Supplementary ovskite a competitive candidate for low-cost photoanode materials.
Fig. 32). As shown in Fig. 4f, the photocurrent of the FAPbBr3 photo-
anode decreased by approximately 25% under UV-filtered light, redu- Photothermal effect of photoanode
cing to about 7.0 mA cm−2 at 1.23 V vs. RHE. When the cut-off While it has been demonstrated that increased electrolyte tempera-
absorption edge is set at 560 nm, the energy of UV light in AM 1.5 G tures can enhance the PEC performance of oxide-based
spectrum accounted for 13%, delivering a maximum photocurrent photoanodes35,36, the effect of temperature on perovskite photo-
density of 1.36 mA cm−2 (Fig. 4g). When a visible light filter was applied, anode performance remains uncertain. Introducing the photothermal
the photocurrent of FAPbBr3 photoanode is limited to 1.23 mA cm−2 at effect into the catalytic process has an evident advantage, as increased
1.23 V vs. RHE, and the onset potential dramatically increased under UV temperature can accelerate the kinetics of electrochemical reactions.
light. This observation could be explained by the fact that UV light However, in the view of charge separation, higher temperatures are
cannot penetrate deep into the perovskite film, leading to the insuffi- detrimental to the performance and stability of perovskite solar
cient driving force for the carriers to flow to the HTM side. The stability cells37,38, which presents a challenge for balancing these two competing
measurements with filtered light are presented in Fig. 4b and Supple- factors. Therefore, investigating the photothermal effects on
mentary Fig. 33. It was found that the FAPbBr3 photoanode degraded perovskite-based photoanodes is crucial. The question to be

Nature Communications | (2023)14:5486 7

Article [Link]

Fig. 5 | Photothermal performance of FAPbBr3 photoanode. Thermal images of illumination (AM 1.5 G, 100 mW cm−2). k Structure diagram of GS/carbon-coated
(a) FAPbBr3/carbon solar cell and (b) FAPbBr3 photoanode under continuous light ITO glass. l, m Temperature profiles and (n) thermal images of ITO/carbon/GS film
illumination (AM 1.5 G, 100 mW cm−2). c Corresponding heating curves of FAPbBr3/ under continuous light illumination (AM 1.5 G, 100 mW cm−2). o Temperature-
Carbon solar cell and FAPbBr3 photoanode. d Thermal image and (e) corre- dependent LSV curves of GS/NiFe alloy/NiFe LDH catalyzed OER in 1.0 M KOH (scan
sponding temperature profile of FAPbBr3 photoanode heated by continuous light rate: 5 mV s−1). Inset: Arrhenius plots under different overpotentials (p)
illumination in the electrolyte (AM 1.5 G, 100 mW cm−2). f Heating curves of ITO/ Temperature-dependent density-voltage (J-V) curves of FAPbBr3/Carbon solar cell
carbon and ITO/carbon/GS under continuous light illumination (AM 1.5 G, under light illumination (AM 1.5 G, 100 mW cm−2). q Temperature-dependent LSV
100 mW cm−2). g Structure diagram of carbon-coated ITO glass. h, i Temperature curves of FAPbBr3 photoanode under light illumination (scan rate: 50 mV s−1, AM
profiles and (j) thermal images of ITO/carbon film under continuous light 1.5 G, 100 mW cm−2).

addressed in this section is whether high temperatures have a positive graphite sheet and epoxy resin, as the temperature immediately rises
or negative impact on photoanode performance. upon illumination and the temperature increases at 470 s are found
Prior to examining the catalytic performance, the photothermal to be 30 °C (Fig. 5c). As shown in Fig. 5b, the heat generated by the
characteristics of HTM-free FAPbBr3 solar cell and photoanode were photoanode is primarily concentrated in its central region, while the
evaluated by recording the time-dependent temperature variation in surrounding epoxy resin functions as a competent insulator, inhi-
ambient air (Supplementary Fig. 35). As shown in Fig. 5c, when biting heat dissipation. The photothermal effect of the FAPbBr3
FAPbBr3 solar cell is exposed to simulated AM 1.5 G light photoanode in electrolyte was also assessed using the infrared
(100 mW cm−2), the temperature rises by 14 °C within 60 s and camera. As presented in Fig. 5d, the surface temperature of the
reaches 45 °C within 470 s, indicating a pronounced photothermal photoanode reaches over 30 °C after 300 s continuous illumination
effect. The infrared image presented in Fig. 5a implies that the heated in 1.0 M KOH, implying that light can heat the interior of the photo-
region is primarily concentrated in the central area covered by the anode even in the electrolyte. The temperature profile reveals that
carbon layer, indicating that the carbon conduction layer can serve as the photoanode surface is evenly heated, with a temperature differ-
a photothermal layer. Moreover, the photothermal performance of ential of 10 °C between the photoanode surface and the electro-
the FAPbBr3 photoanode is enhanced after encapsulation with a lyte (Fig. 5e).

Nature Communications | (2023)14:5486 8

Article [Link]

Carbon and carbon/graphite layers were coated on the ITO glass perovskite photoanodes. Fig. 5q shows the photocurrent density-
to investigate the photothermal conversion properties of carbon and overpotential curves of FAPbBr3 photoanode at different tempera-
graphite components separately (the structures are shown in Figs. 5g tures. The results indicate that changes in temperature do not sig-
and 5k). Under AM 1.5 G irradiation, the ITO/carbon film exhibits a nificantly affect the onset potential under the same driving force for
considerable photothermal conversion, with an 18 °C temperature the reaction, which suggests that the decrease in voltage for light-
increase after 300 s of exposure. Conversely, after attaching the gra- absorbing perovskite material and the reduction in overpotential
phite sheet, the temperature rises of the ITO/carbon/GS film decreased requirements for the catalytic process offset each other, resulting in a
by 1.5 °C compared to that of the ITO/carbon film (Fig. 5f). This consistent apparent onset potential for FAPbBr3 perovskite photo-
decrease in temperature could be attributed to the enhanced surface anode. Furthermore, the increased short-circuit current in the light-
heat dissipation in the carbon/GS film, as graphite sheet are widely absorbing portion and the increased catalytic current in the catalytic
employed as thermal pads in diverse industries due to their superior portion leads to a significant improvement in the saturated catalytic
heat dissipation characteristics. Fig. 5j and Fig. 5n show the thermal current of the photoanode. Based on the above analysis, it can be
images of the ITO/carbon and ITO/carbon/GS films, respectively, concluded that the increase in temperature has a positive effect on the
under continuous light illumination. The graphite sheet, which exhi- photoanode performance. Utilizing photothermal effect can be an
bits excellent thermal conductivity, shows a uniform temperature effective strategy to further enhance the performance of perovskite-
distribution. As illustrated in Fig. 5h, i, the heat primarily accumulates based photoanodes, provided that stability is not compromised.
in the center of the ITO/carbon film, producing parabolic temperature Moreover, the FAPbBr3 material exhibits superior thermostability
profile curves. In contrast, ITO/carbon/GS film exhibits uniform heat compared to other candidates, such as MAPbBr3 and MAPbI3, making
distribution on the surface, resulting in trapezoidal-shaped tempera- it an ideal building block for highly efficient photoanodes that can fully
ture profile curves (Fig. 5l, m). Based on this analysis, it can be con- leverage the photothermal effect without suffering from thermal
cluded that the carbon layer primarily generates the photothermal degradation.
effect in the device; moreover, the graphite layer utilized for con- Solar hydrogen can be produced using a variety of methods,
ductive sealing also functions as a thermal pad layer, facilitating the including the use of PV cells, PEC cells, and hybrid systems that com-
rapid and uniform heat conduction toward the catalyst layer to miti- bine both technologies. Our work demonstrates that PVM-PEC con-
gate the accumulated heat damage to the perovskite layer. figuration can fully utilize the photothermal effect, like PEC, to
The use of a GS separator also allows for the individual investi- maximize solar energy utilization (see summary in Supplementary
gation of the temperature effect on electrochemical water oxidation Note 6). As-fabricated FAPbBr3 photoanode can be synthesized with
catalysis of the ITO/GS/NiFe alloy/NiFe LDH electrode. Temperature remarkable accessibility without requiring high temperature, inert
(T)-controlled electrocatalysis was conducted in 1.0 M KOH to deter- atmosphere, or vacuum evaporation techniques. The carbon/GS pro-
mine the temperature dependency for the electrode. In an electro- tection strategy effectively prevents water corrosion from damaging
chemical reaction, the activation overpotential (η) is the potential the sensitive perovskite, and the photothermal properties of the car-
difference from the equilibrium value required to produce a current, bon material allow for extended spectral utilization. Meanwhile, this
which depends on the activation energy of the redox event. Hence, the HTM-free and precious metal-free structure dramatically reduces
catalytic rate, as indicated by the current densities, should be com- fabrication costs. The scalable electrodeposition preparation of cata-
pared at the same overpotential in order to accurately assess the cat- lyst layers enables facile fine-tuning of the composition, structure, and
alytic activity (see Supplementary Note 5 for the calculation of morphology of electrocatalysts using electrochemical technologies.
overpotential). After calibration with driven force, the corresponding We believe that these findings have significant implications for the
Arrhenius plots of the current at various potentials against 1/T are development of cost-effective, high-performance integrated photo-
calculated and shown in the inset figure of Fig. 5o. At an overpotential electrodes for solar energy conversion applications.
of 250 mV, the apparent activation energy is calculated to be
16.2 kJ mol−1. The activation energy at zero overpotential is derived Discussion
from the intercept of apparent activation energy vs. overpotential In summary, this work has demonstrated a FAPbBr3-based photoanode
plots at η = 0 (Supplementary Fig. 38)39. The obtained value featuring a photocurrent density around 8.5 mA cm−2 at 1.23 V vs. RHE,
(65 kJ mol−1) is smaller than that of other nickel-based catalysts, such as an onset potential below 0 V vs. RHE, and half-life stability over 55 h
NiTe (100.2 kJ mol−1) > Ni2Si (98.6 kJ mol−1) > Ni (94.0 kJ mol−1) > NiAs using a FAPbBr3 absorber, a carbon/graphite conductive protection
(86.3 kJ mol−1) > NiP (78.4 kJ mol−1)40, indicating faster intrinsic OER layer and NiFe alloy/NiFe LDH catalysts under alkaline conditions. Due
kinetics for the electrodeposited NiFe alloy/NiFe LDH catalysts. to its fast catalytic kinetics, high photovoltage, and rational band
The temperature effect on the FAPbBr3 solar cell is studied using a structure, a high ABPE of 8.5% was reached at 0.08 V vs. RHE. Further
lab-made setup (Supplementary Fig. 39). As shown in Fig. 5p, an stability studies and post-characterizations revealed that the per-
increase in temperature will cause a decrease in the Eg of the semi- ovskite interface was damaged by UV light, resulting in a detached
conductor material, resulting in a slight increase in the short-circuit absorber layer and performance loss. Accordingly, by applying a UV
current38. However, this increase in Jsc is not enough to compensate for filter, the device achieved record stability of over 100 h in 1.0 M KOH
the loss of FF and Voc values due to higher recombination probabilities, under constant simulated solar illumination, with currents around
which ultimately leads to a reduced PCE at higher temperatures41. On 7.5 mA cm−2 at 1.23 V vs. RHE. More importantly, this work has con-
one hand, for photoanode devices with perovskite-based light- ducted a detailed investigation of the photothermal effect on FAPbBr3
absorbing layer, the decrease in voltage is certainly detrimental to the photoanode, confirming that the photoanode based on photovoltaic
performance of photoanode devices, particularly in terms of onset materials can also take full advantage of the photothermal effect and
potential and maximum ABPE. On the other hand, the reduction in Eg, improve catalytic performance. In the design structure of ITO/SnO2/
as well as the increase in Jsc, can lead to a higher saturated current FAPbBr3/carbon/GS/NiFe alloy/NiFe LDH, the carbon layer serves as
under higher applied potential. Additionally, an increase in tempera- both the hole transport layer and the photothermal layer, while the
ture can significantly decrease the theoretical potential of the water graphite layer functions not only as a waterproof and conductive layer
oxidation reaction while greatly increasing the catalytic current under but also efficiently transfers heat, serving as a thermal homogenization
the same driven force. These competing effects make it uncertain how layer. Meanwhile, the dense electrodeposited alloy layer provides a
temperature affects the performance of photoanode devices, and robust substrate for high-performance LDH catalyst loading, ensuring
thus, it is necessary to study the temperature effect on complete efficient electrical and thermal conductivity. The compact and

Nature Communications | (2023)14:5486 9

Article [Link]

multifunctional design endows the monolithic device with high per- the perovskite films (4000 rpm, 30 s) and 80 nm Au layer was then
formance and excellent stability. These observations, along with the deposited via vacuum evaporation on top of spiro-OMeTAD as a
ultralow onset potential and the facile processability of FAPbBr3 pho- contact electrode. The devices were stored in a dry box overnight
toanode, demonstrate the potential of lead halide perovskite-based before the J-V measurement.
PEC as a promising approach to achieving efficient, cost-effective, and
scalable photoelectrochemical solar fuel production. Fabrication of NiFe catalysts
The structure of catalyst layers is glass/ITO/GS/NiFe alloy/NiFe LDH.
Methods The self-adhesive GS layer was stuck onto the ITO glass. The NiFe
Chemicals alloy layer was electrodeposited on the GS layer (1.0 × 1.0 cm) using a
Colloidal tin oxide solution (SnO2, 15% in H2O) was purchased from Alfa reported method with subtle modification43. In brief, sodium citrate
Aesar. Sodium hydroxide (semiconductor grade, 99.99%), potassium dihydrate (20.0 mmol, 6.00 g), ammonium sulfate (45.0 mmol,
hydroxide (semiconductor grade, 99.99%), tri-Sodium citrate dihy- 6.00 g), and sodium hypophosphite monohydrate (57.0 mmol,
drate (99%), Ammonium sulfate (99.0%), sodium hypophosphite 6.00 g) were dissolved in 100 mL water at ambient temperature.
monohydrate (99.0%), Nickel(II) sulfate hexahydrate (99.99%), Iron(II) Nickel sulfate hexahydrate (10.0 mmol, 2.62 g) and iron(II) sulfate
sulfate heptahydrate (99.0%), Nickel(II) chloride hexahydrate (99.9%), heptahydrate (2.0 mmol, 0.54 g) were added into the solution with
Iron(III) chloride (anhydrous, 99.99%), Nickel(II) nitrate hexahydrate stirring. The pH of the solution was adjusted to 10.0 using 10 M
(99.999%), Iron(III) nitrate nonahydrate (99.95%), dimethylformamide NaOH solution. The deposition process was performed using a
(DMF, anhydrous, 99.8%), dimethyl sulfoxide (DMSO, anhydrous, three-electrode electrochemical cell (platinum mesh as the counter
99.9%), acetonitrile (HPLC), chlorobenzene (anhydrous, 99.8%) electrode and an Ag/AgCl (saturated KCl) as the reference electrode)
methanol (HPLC, 99.9%), bis(trifluoromethanesulfonyl)imide lithium filled with 15 mL electrodeposition solution at an applied potential of
(Li-TFSI) and 4-tert-butylpyridine (TBP) were purchased from Sigma- −1.10 V vs. Ag/AgCl for 1600 s. The alloy film was rinsed with deio-
Aldrich and used as received without further purification. 2,20,7,70- nized water and then dried under an N2 atmosphere before further
tetrakis [N, N-di(4-methoxyphenyl) amino]−9,90-spirobifluorene use. The NiFe LDH layer was electrodeposited on the alloy layer
(SpiroOMeTAD) was purchased from Borun. PbBr2 (99.99%, trace using a modified reported method44. The electrolyte bath contained
metals basis) was purchased from TCI Chemicals. FK209 [tris(2- 5 mM nickel(II) nitrate hexahydrate and 5 mM iron(III) nitrate non-
(1Hpyrazol-1-yl)−4-tert-butylpyridine)-cobalt (III) tris(bis(tri- ahydrate. The electrodeposition was then carried out at −1.0 V vs.
fluoromethylsulfonyl)imide) and FABr were provided by Greatcell Ag/AgCl for 900 s. NiFe LDH was also electrodeposited onto the GS
solar. Carbon pastes (DN-CP01) were purchased from Dyenamo. 160 electrode following the same procedures. The LDH film was rinsed
um graphite thermal sheet (794–3973) was purchased from RS com- with deionized water and dried under an N2 atmosphere before
ponents. High-purity deionized water (18.2 MΩ cm−1), supplied with a further use.
Milli-Q system (Millipore, Advantage A10), was used in all experiments.
Indium tin oxide (ITO) glass substrate and epoxy adhesive (Loctite Fabrication of FAPbBr3 photoanode
9466) were purchased from a local company. All other reagents were Following Poli’s methods13, after the FAPbBr3 solar cell part was pre-
commercially available and used as received. Organic solvents were of pared, the above-mentioned 160 um self-adhesive GS layer was stuck
analytical reagent grade and used without further purification. onto the carbon film (step 2, Supplementary Fig. 1). The bottom edge
part of the attached GS was further covered by carbon paste to ensure
Fabrication of FAPbBr3 solar cell conductivity. Then copper tape and copper wire were stuck on the ITO
The structure of hole transport materials-free FAPbBr3 solar cell is and GS parts, respectively (step 3, Supplementary Fig. 1). Latterly, the
glass/ITO/SnO2/FAPbBr3/carbon. ITO glass was cleaned with deter- photoanode was sealed by chemical-resistant epoxy resin Loctite 9466
gent, acetone, isopropanol, and water for 20 min, respectively. (step 4, Supplementary Fig. 1) and hardened at 60 °C for 30 min. The
UV/ozone (UVO) treatment was then applied for 30 min before use. devices were stored in a dry box overnight before subsequent use. The
SnO2 film was deposited on the ITO substrate by spin-coating method catalyst layers were deposited on GS using the same procedure and
using the commercial SnO2 colloidal solution (diluted with deionized conditions as above. In particular, the deposition process was per-
water, the volume ratio of water to SnO2 is 4:1) at 3000 rpm for 30 s, formed using a three-electrode electrochemical cell with the working
followed by annealing treatment at 150 °C on a hotplate for 30 min. electrode connected to copper wire. After the preparation is complete,
After cooling down, UV/ozone treatment was conducted on SnO2 film the copper wire should be cut short to prevent possible corrosion in
for 20 min before further use. The FAPbBr3 film was prepared using a the electrolyte.
two-step spin coating method. PbBr2 precursor solution was pre-
pared by dissolving 1.3 M of PbBr2 in a mixture of DMF and DMSO Electrochemical measurements
with a volume ratio of 9:1 and stirred at 70 °C overnight. Then the All electrochemical characterizations were carried out in a standard
PbBr2 precursor solution was spin-coated onto SnO2 substrates at three-electrode cell connected to a CHI 660E electrochemical work-
3000 rpm for 30 s, followed by annealing treatment in the ambient station, using the prepared films as the working electrode, a Pt mesh as
conditions at 70 °C for 1 min. Next, 0.065 g FABr was dissolved in the counter electrode, and Hg/HgO (1.0 M KOH) as the reference
1.0 ml methanol and spin-coated on top of PbBr2 precursor film electrode. The temperature was controlled at 25 °C by a thermostatic
(3000 rpm, 30 s), followed by annealing treatment in the ambient water bath unless otherwise noted. All measured potentials were
conditions at 150 °C for 25 min. Finally, A carbon paste (DN-CP01) was converted to values relative to RHE by a commercialized RHE reference
doctor-bladed as the top contact on FAPbBr3 film. Devices were electrode (HydroFlex of Gaskatel) at 25 °C45. All linear sweep voltam-
further annealed at 100 °C for 30 min in the air for the formation of a metry curves were measured in a three-electrode configuration with iR
solid carbon layer. The reference solar cell with structure of glass/ correction (95%) unless otherwise noted. Chronopotentiometry was
ITO/SnO2/FAPbBr3/spiro-MeOTAD/Au was fabricated using same recorded under the same experimental setup with iR-correction.
procedure. The HTM layer was prepared by spin coating method: All the correction was done manually according to the equation
70 mM Spiro-OMeTAD was dissolved in chlorobenzene with addi- Ecorr = Emeas – iRu, where Ecorr is iR-corrected potential, Emeas is the
tives of TBP, Li-TFS (1.8 M in acetonitrile), and FK209 (0.25 M in experimentally measured potential, and Ru is the equivalent series
acetonitrile). The Molar ratio of Spiro-OMeTAD: TBP: Li-TFSI: FK209 resistance extracted from Nyquist plots (~2.3 Ω in 1 M KOH). Tafel
was 1: 3.3: 0.5: 0.0542. The Spiro-OMeTAD solution was spin-coated on slopes were calculated based on polarization curves by fitting to the

Nature Communications | (2023)14:5486 10

Article [Link]

following equation η = b × log(j) + a, where j is current density was reached when the forces on the relaxed atoms became less than
(mA cm−2), η is overpotential (mV), and b is Tafel slope. 0.02 eV/Å. The van der Waal (vdW) interaction was described by the
DFT-D3 method51,52. Note that, due to the presence of heavy atoms,
Photoelectrochemical measurements spin-orbit coupling (SOC) is also included through all calculations53.
All photoelectrochemical measurements were carried out at room
temperature by using a CHI 660E electrochemical workstation. The Data availability
photoelectrochemical performances of photoanodes were measured The complete dataset in the main text is available at [Link]
with a three-electrode configuration, using the prepared photoanodes with the identifier [Link] Other data
as the working electrode, a Pt mesh as the counter electrode, and Hg/ that support the findings of this study are available from the corre-
HgO (1.0 M KOH) as the reference electrode. All measured potentials sponding author upon reasonable request.
were converted to values relative to RHE by a commercialized RHE
reference electrode (HydroFlex of Gaskatel) at 25 °C. The simulated References
solar illumination was obtained by a NEWPORT LCS-100 solar simu- 1. Bard, A. J. & Fox, M. A. Artificial photosynthesis: solar splitting of
lator (type 94011A-ES, a 100 W Xenon arc lamp with an AM 1.5 G filter). water to hydrogen and oxygen. Acc. Chem. Res 28, 141–145 (1995).
The incident light power intensity was calibrated to 100 mW cm−2 by a 2. Queyriaux, N., Kaeffer, N., Morozan, A., Chavarot-Kerlidou, M. &
Thorlabs S401C photometer (190 nm −20 µm). J-V curves were Artero, V. Molecular cathode and photocathode materials for
obtained by LSV without iR-correction. The chopped light was con- hydrogen evolution in photoelectrochemical devices. J. Photo-
trolled using an electronic shutter and corresponding driver (type chem. Photobiol. C: Photochem. Rev. 25, 90–105 (2015).
76995) provided by Newport. The long-term electrolysis was carried 3. Lewis, N. S. & Nocera, D. G. Powering the planet: Chemical chal-
out with a constant potential under the same experimental setup. The lenges in solar energy utilization. Proc. Natl. Acad. Sci. U.S.A. 103,
ABPE was calculated from the J-V curves (in a three-electrode system) 15729–15735 (2006).
under illumination according to the following equation: 4. Styring, S. Artificial photosynthesis for solar fuels. Faraday Discuss.
155, 357–376 (2012).

 
1:23  V RHE × j light  jdark 5. Kim, J. H., Hansora, D., Sharma, P., Jang, J.-W. & Lee, J. S. Toward
ABPEð%Þ = × 100% ð1Þ practical solar hydrogen production – an artificial photosynthetic
P light
leaf-to-farm challenge. Chem. Soc. Rev. 48, 1908–1971 (2019).
6. Chen, J., Dong, C., Idriss, H., Mohammed, O. F. & Bakr, O. M. Metal
where Plight was the light intensity of incident illumination, VRHE indi- Halide Perovskites for Solar-to-Chemical Fuel Conversion. Adv.
cated the applied potential in RHE scale, jlight and jdark were photo- Energy Mater. 10, 1902433 (2020).
currents and dark current, respectively. The IPCE values of 7. Huang, H., Pradhan, B., Hofkens, J., Roeffaers, M. B. J. & Steele, J. A.
photoanodes were calculated using the following equation: Solar-Driven Metal Halide Perovskite Photocatalysis: Design, Stabi-
  lity, and Performance. ACS Energy Lett. 5, 1107–1123 (2020).
1240 × j light  j dark 8. Singh, M. & Sinha, I. Halide perovskite-based photocatalysis
IPCEð%Þ = × 100% ð2Þ systems for solar-driven fuel generation. Sol. Energy 208,
λ × Pλ
296–311 (2020).
The monochromatic light was produced using a monochromator. 9. Ren, K. et al. Metal halide perovskites for photocatalysis applica-
The light intensity (Pλ) at each wavelength (λ) was determined by a tions. J. Mater. Chem. A 10, 407–429 (2022).
photometer; jlight and jdark were photocurrents and dark current, 10. Singh, S. et al. Hybrid Organic–Inorganic Materials and Composites
respectively, measured by an electrochemical workstation. The onset for Photoelectrochemical Water Splitting. ACS Energy Lett. 5,
potential (Eonset) for photoanode was defined as the intersection point 1487–1497 (2020).
of the J-V curve with the x-axis. The corresponding photovoltage (Vph) 11. Tong, G. et al. Phase transition induced recrystallization and low
was calculated according to the following equation: surface potential barrier leading to 10.91%-efficient CsPbBr3 per-
ovskite solar cells. Nano Energy 65, 104015 (2019).
V ph = 1:23  E onset ð25 CÞ ð3Þ 12. Li, Z. et al. Solar Hydrogen. Adv. Energy Mater. 13, 2203019 (2023).
13. Poli, I. et al. Graphite-protected CsPbBr3 perovskite photoanodes
It’s worth noting that this approach neglects catalytic over- functionalised with water oxidation catalyst for oxygen evolution in
potentials, resulting in a potential underestimation of the photo- water. Nat. Commun. 10, 2097 (2019).
voltage for PEC devices. 14. Dong, G., Hu, H., Huang, X., Zhang, Y. & Bi, Y. Rapid activation of
Co3O4 cocatalysts with oxygen vacancies on TiO2 photoanodes for
Computational details efficient water splitting. J. Mater. Chem. A 6, 21003–21009 (2018).
All the DFT calculations were performed using the Vienna Ab-initio 15. Han, H. S. et al. (020)-Textured tungsten trioxide nanostructure with
Simulation Package (VASP)46. The projector-augmented wave (PAW) enhanced photoelectrochemical activity. J. Catal. 389,
pseudopotentials method46,47 was utilized to describe the electron-ion 328–336 (2020).
interactions with the plane-wave basis expansion cut-off energy set to 16. Jeon, T. H., Moon, G.-H., Park, H. & Choi, W. Ultra-efficient and
be 500 eV. The generalized gradient approximation (GGA) was used durable photoelectrochemical water oxidation using elaborately
with Perdew–Burke–Ernzerhof (PBE)48 to perform the exchange- designed hematite nanorod arrays. Nano Energy 39, 211–218 (2017).
correlation potential. The valence states of the potential chose 17. Liu, B. et al. A BiVO4 Photoanode with a VOx Layer Bearing Oxygen
5d106s26p2 for the lead (Pb) atom, 4s24p5 for the bromine (Br) atom, Vacancies Offers Improved Charge Transfer and Oxygen Evolution
2s22p2 for the carbon (C) atom, 2s22p3 for the nitrogen (N) atom, and 1s1 Kinetics in Photoelectrochemical Water Splitting. Angew. Chem.
for the hydrogen (H) atom. The Hubbard-type correction was set on Pb Int. Ed. 62, e202217346 (2023).
6p and Br 4p orbitals with a U-J value of 9 eV and 8 eV, respectively49,50. 18. Xiao, Y. et al. Band structure engineering and defect control of
The p(2 × 2 × 2) supercell of FAPbBr3, consisting of 96 atoms, is con- Ta3N5 for efficient photoelectrochemical water oxidation. Nat.
sidered for the defect study. A 5 × 5× 5 Monkhorst–Pack k-point mesh Catal. 3, 932–940 (2020).
sampling for Brillouin-zone integration was used for the bulk optimi- 19. Gu, J. et al. A graded catalytic–protective layer for an efficient and
zation, while it was 2 × 2 × 2 for all defective structures. The equilibrium stable water-splitting photocathode. Nat. Energy 2, 16192 (2017).

Nature Communications | (2023)14:5486 11

Article [Link]

20. Pan, L. et al. Boosting the performance of Cu2O photocathodes for 43. Yang, H. et al. Improving the performance of water splitting elec-
unassisted solar water splitting devices. Nat. Catal. 1, 412–420 (2018). trodes by composite plating with nano-SiO2. Electrochim. Acta.
21. Mayer, M. T. Photovoltage at semiconductor–electrolyte junctions. 281, 60–68 (2018).
Curr. Opin. Electrochem. 2, 104–110 (2017). 44. Lu, X. & Zhao, C. Electrodeposition of hierarchically structured
22. Hanusch, F. C. et al. Efficient Planar Heterojunction Perovskite Solar three-dimensional nickel–iron electrodes for efficient oxygen evo-
Cells Based on Formamidinium Lead Bromide. J. Phys. Chem. Lett. lution at high current densities. Nat. Commun. 6, 6616 (2015).
5, 2791–2795 (2014). 45. Li, F. et al. A Cobalt@Cucurbit[5]uril Complex as a Highly Efficient
23. Eperon, G. E. et al. Formamidinium lead trihalide: a broadly tunable Supramolecular Catalyst for Electrochemical and Photoelec-
perovskite for efficient planar heterojunction solar cells. Energy trochemical Water Splitting. Angew. Chem. Int. Ed. 60,
Environ. Sci. 7, 982–988 (2014). 1976–1985 (2021).
24. Slimi, B. et al. Synthesis and characterization of perovskite 46. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the
FAPbBr3−xIxthin films for solar cells. Monatsh. Chem. 148, projector augmented-wave method. Phys. Rev. B 59, 1758–1775
835–844 (2017). (1999).
25. Caprioglio, P. et al. On the Relation between the Open-Circuit 47. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50,
Voltage and Quasi-Fermi Level Splitting in Efficient Perovskite Solar 17953–17979 (1994).
Cells. Adv. Energy Mater. 9, 1901631 (2019). 48. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient
26. Beranek, R. (Photo)electrochemical Methods for the Determination approximation made simple. Phys. Rev. Lett. 77, 3865–3868 (1996).
of the Band Edge Positions of TiO2-Based Nanomaterials. Adv. Phys. 49. Peng, C., Chen, J., Wang, H. & Hu, P. First-Principles Insight into the
Chem. 2011, 786759 (2011). Degradation Mechanism of CH3NH3PbI3 Perovskite: Light-Induced
27. Poli, I. et al. Screen printed carbon CsPbBr3 solar cells with high Defect Formation and Water Dissociation. J. Phys. Chem. C. 122,
open-circuit photovoltage. J. Mater. Chem. A 6, 18677–18686 27340–27349 (2018).
(2018). 50. Ding, Y., Shen, Y., Peng, C., Huang, M. & Hu, P. Unraveling the
28. Lee, S.-W. et al. UV Degradation and Recovery of Perovskite Solar Photogenerated Electron Localization on the Defect-Free
Cells. Sci. Rep. 6, 38150 (2016). CH3NH3PbI3(001) Surfaces: Understanding and Implications from a
29. Bard, A. Standard potentials in aqueous solution. Routledge (2017). First-Principles Study. J. Phys. Chem. Lett. 11, 8041–8047 (2020).
30. Ito, S., Tanaka, S., Manabe, K. & Nishino, H. Effects of Surface 51. Grimme, S., Antony, J., Ehrlich, S. & Krieg, H. A consistent and
Blocking Layer of Sb2S3 on Nanocrystalline TiO2 for CH3NH3PbI3 accurate ab initio parametrization of density functional dispersion
Perovskite Solar Cells. J. Phys. Chem. C. 118, 16995–17000 (2014). correction (DFT-D) for the 94 elements H-Pu. J. Chem. Phys. 132,
31. Wang, Q., Phung, N., Di Girolamo, D., Vivo, P. & Abate, A. 154104 (2010).
Enhancement in lifespan of halide perovskite solar cells. Energy 52. Grimme, S., Ehrlich, S. & Goerigk, L. Effect of the damping function
Environ. Sci. 12, 865–886 (2019). in dispersion corrected density functional theory. J. Comput. Chem.
32. Saliba, M., Stolterfoht, M., Wolff, C. M., Neher, D. & Abate, A. Mea- 32, 1456–1465 (2011).
suring Aging Stability of Perovskite Solar Cells. Joule 2, 53. Even, J., Pedesseau, L., Jancu, J.-M. & Katan, C. Importance of
1019–1024 (2018). Spin–Orbit Coupling in Hybrid Organic/Inorganic Perovskites for
33. Domanski, K., Alharbi, E. A., Hagfeldt, A., Grätzel, M. & Tress, W. Photovoltaic Applications. J. Phys. Chem. Lett. 4,
Systematic investigation of the impact of operation conditions on 2999–3005 (2013).
the degradation behaviour of perovskite solar cells. Nat. Energy 3, 54. Hoang, M. T., Pham, N. D., Han, J. H., Gardner, J. M. & Oh, I. Inte-
61–67 (2018). grated Photoelectrolysis of Water Implemented On Organic Metal
34. Tan, H. et al. Efficient and stable solution-processed planar per- Halide Perovskite Photoelectrode. ACS Appl. Mater. Interfaces 8,
ovskite solar cells via contact passivation. Science 355, 722–726 11904–11909 (2016).
(2017). 55. Nam, S., Mai, C. T. K. & Oh, I. Ultrastable Photoelectrodes for Solar
35. Zhou, C., Zhang, L., Tong, X. & Liu, M. Temperature Effect on Pho- Water Splitting Based on Organic Metal Halide Perovskite Fabri-
toelectrochemical Water Splitting: A Model Study Based on BiVO4 cated by Lift-Off Process. ACS Appl. Mater. Interfaces 10,
Photoanodes. ACS Appl. Mater. Interfaces 13, 61227–61236 (2021). 14659–14664 (2018).
36. Dias, P., Lopes, T., Andrade, L. & Mendes, A. Temperature effect on 56. Tao, R., Sun, Z., Li, F., Fang, W. & Xu, L. Achieving Organic Metal
water splitting using a Si-doped hematite photoanode. J. Power Halide Perovskite into a Conventional Photoelectrode: Outstanding
Sources 272, 567–580 (2014). Stability in Aqueous Solution and High-Efficient Photoelec-
37. Tress, W. et al. Performance of perovskite solar cells under simu- trochemical Water Splitting. ACS Appl. Energy Mater. 2,
lated temperature-illumination real-world operating conditions. 1969–1976 (2019).
Nat. Energy 4, 568–574 (2019). 57. Chen, H. et al. Integrating Low-Cost Earth-Abundant Co-Catalysts
38. Lin, L. & Ravindra, N. M. Temperature dependence of CIGS and with Encapsulated Perovskite Solar Cells for Efficient and Stable
perovskite solar cell performance: an overview. SN Appl. Sci. 2, Overall Solar Water Splitting. Adv. Funct. Mater. 31, 2008245
1361 (2020). (2021).
39. Yang, H. et al. Intramolecular hydroxyl nucleophilic attack pathway 58. Daboczi, M., Cui, J., Temerov, F., Eslava, S. Scalable all-inorganic
by a polymeric water oxidation catalyst with single cobalt sites. Nat. halide perovskite photoanodes with 100 h operational stability
Catal. 5, 414–429 (2022). containing Earth-abundant materials. ChemRxiv Cambridge: Cam-
40. Masa, J. et al. Ni-Metalloid (B, Si, P, As, and Te) Alloys as Water bridge Open Engage, This content is a preprint and has not been
Oxidation Electrocatalysts. Adv. Energy Mater. 9, 1900796 (2019). peer-reviewed. (2023).
41. Aharon, S., Dymshits, A., Rotem, A. & Etgar, L. Temperature 59. Wang, M. et al. High-Performance and Stable Perovskite-Based
dependence of hole conductor free formamidinium lead iodide Photoanode Encapsulated by Blanket-Cover Method. ACS Appl.
perovskite based solar cells. J. Mater. Chem. A 3, 9171–9178 (2015). Energy Mater. 4, 7526–7534 (2021).
42. Jesper Jacobsson, T. et al. Exploration of the compositional space 60. Kim, T. G. et al. Monolithic Lead Halide Perovskite Photoelec-
for mixed lead halogen perovskites for high efficiency solar cells. trochemical Cell with 9.16% Applied Bias Photon-to-Current Effi-
Energy Environ. Sci. 9, 1706–1724 (2016). ciency. ACS Energy Lett. 7, 320–327 (2022).

Nature Communications | (2023)14:5486 12

Article [Link]

61. Li, F. Design of Water Splitting Devices via Molecular Engineering. Additional information
PhD diss., KTH Royal Institute of Technology (2016). Supplementary information The online version contains
supplementary material available at
Acknowledgements [Link]
L.S. acknowledges the financial support from the National Key R&D
Program of China (2022YFA0911902), the National Natural Science Correspondence and requests for materials should be addressed to Erik
Foundation of China (Grants No. 22088102) and the starting-up package M. J. Johansson or Licheng Sun.
of Westlake University. Y.L. and E.M.J.J. acknowledge the financial sup-
port obtained from the Swedish Energy Agency, ÅForsk, Swedish Peer review information Nature Communications thanks Jong Hyeok
Research Council (VR), and Olle Engkvist Foundation. G.B. thanks sup- Park and the other, anonymous, reviewer(s) for their contribution to the
port from the STandUP for Energy program. F.L. acknowledges the peer review of this work. A peer review file is available.
financial support from the National Natural Science Foundation of China
(Grants No. 22172011). The authors acknowledge Westlake University Reprints and permissions information is available at
High-Performance Computing Center for computation support. The [Link]
authors thank the Instrumentation and Service Center for Physical Sci-
ences at Westlake University for facility support and technical Publisher’s note Springer Nature remains neutral with regard to jur-
assistance. isdictional claims in published maps and institutional affiliations.

Author contributions Open Access This article is licensed under a Creative Commons
H.Y., Y.L., E.M.J.J., and L.S. conceived the project design and initiated the Attribution 4.0 International License, which permits use, sharing,
project. H.Y. and Y.L. performed synthesis, structural and electro- adaptation, distribution and reproduction in any medium or format, as
chemical characterizations. Y.D. performed DFT calculations. H.Y., Y.L., long as you give appropriate credit to the original author(s) and the
F.L., L.W., B.C., F.Z., T.L., and G.B. contributed to data analysis and source, provide a link to the Creative Commons licence, and indicate if
interpretation. All authors contributed to discussions. H.Y. and Y.L. wrote changes were made. The images or other third party material in this
the original draft with inputs from the other authors, and all authors article are included in the article’s Creative Commons licence, unless
reviewed and revised the manuscript. E.M.J.J. and L.S. supervised the indicated otherwise in a credit line to the material. If material is not
research. included in the article’s Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted
Funding use, you will need to obtain permission directly from the copyright
Open access funding provided by Uppsala University. holder. To view a copy of this licence, visit [Link]
licenses/by/4.0/.
Competing interests
The authors declare no competing interests. © The Author(s) 2023

Nature Communications | (2023)14:5486 13