Full Paper

www.afm-journal.de

Dual-Ion-Diffusion Induced Degradation in Lead-Free

Cs2AgBiBr6 Double Perovskite Solar Cells
Mehri Ghasemi, Lei Zhang, Jung-Ho Yun, Mengmeng Hao, Dongxu He, Peng Chen,
Yang Bai, Tongen Lin, Mu Xiao, Aijun Du, Miaoqiang Lyu,* and Lianzhou Wang*

challenges, lead-free halide perovskites

Lead-free double perovskite Cs2AgBiBr6 has attracted increasing research with improved stability are a forefront
interest in addressing the toxicity and stability challenges confronted by lead research focus. Structural and electronic
halide perovskites. While most of the studies on this Cs2AgBiBr6 material considerations have inspired research
have been focusing on photovoltaic performance and potential applications, efforts to replace lead with tin. The ther-
modynamically favored high valence (i.e.,
its long-term stability and degradation mechanism are well under-explored. Sn4+) in tin-based perovskites, however,
Herein, high-quality Cs2AgBiBr6 thin-films are developed for lead-free double causes even worse stability over their lead
perovskite solar cells with a decent efficiency of 1.91%. By exploring the counterparts.[7,8] Another approach is the
ambient stability of these photovoltaic devices, it is found that the Cs2AgBiBr6 substitution of lead (Pb2+) with isoelec-
exhibits a unique dual-ion-migration phenomenon, where Ag and Br ions tronic bismuth (Bi3+).[9–11] Nevertheless,
replacing Pb solely with Bi tends to form
gradually diffuse through the hole-transporting layer in the long-term
0D perovskite derivatives with isolated bis-
operation. This phenomenon leads to the degradation of the Cs2AgBiBr6 muth halide dimers of face-sharing octa-
perovskite and subsequent device failure. Theoretical calculations indicate hedra, leading to confinement of excitons
that low formation energies of the Ag and Br vacancies, and low diffusive and low photovoltaic efficiencies.[12,13]
energy barriers contribute to the dual-ion-migration effect. A possible Lead-free halide double perovskites
mechanism involving a vacancy-mediated ion-migration is proposed to have been recently emerging as promising
candidates among the bismuth halide
explain this phenomenon. These key findings are essential for halide perovskites with a typical 3D structure.
double perovskites not only in providing a new knowledge base for further These double perovskites share a formula
addressing the challenge of double perovskite stability, but also in extending of A2B(I)B(III)X6, where A is typically an
their optoelectronic/electronic applications where mixed electronic, ionic and alkali metal cation (Na, K, Rb, and Cs),
photonic properties may be desired. while B is a monovalent (Cu (I), Ag (I),
and Au (I), etc.) and a trivalent metal
cation (Bi (III), Sb (III) and In (III), etc.)
1. Introduction and X is a halide anion (F, Cl, Br, and I).[14,15] Therefore,
double perovskites have rich combinational chemistry in com-
Lead halide perovskites have witnessed significant progress in positions, leading to a large family of potential candidates.
low-cost and high-efficiency photovoltaics, with a rapid increase Among various double perovskites, Cs2AgBiBr6 has recently
in photovoltaic efficiencies from 3.8% to a certified record of received much attention due to its potential applications in
25.2% in the past decade.[1–4] However, the viability and practical solar cells,[16–18] radiation detectors,[19,20] photodetectors,[21] sen-
scale-up implementation are limited by the stability and tox- sors,[22] and photocatalysis.[23,24] Although Cs2AgBiBr6 features
icity of the lead halide perovskites.[5,6] To circumvent these two an indirect bandgap, it exhibits long radiative photolumines-
cence lifetime, long charge diffusion length, high defect toler-
M. Ghasemi, Dr. J.-H. Yun, M. Hao, D. He, Dr. P. Chen, Dr. Y. Bai, ance and superior stability in ambient conditions compared to
T. Lin, Dr. M. Xiao, Dr. M. Lyu, Prof. L. Wang traditional lead halide perovskites.[18,25,26]
Nanomaterials Centre Despite its superior ambient stability over its lead coun-
School of Chemical Engineering terparts, the intrinsic ionic nature of this Cs2AgBiBr6 double
Australian Institute for Bioengineering and Nanotechnology
The University of Queensland
perovskite still raises similar ion migration concerns in its
St Lucia, Brisbane, QLD 4072, Australia long-term stability. Ion migration in the lead halide perov-
E-mail: [email protected]; [email protected] skites has been intensively studied previously, which is
L. Zhang, Prof. A. Du revealed to be contributing factors for many anomalous phe-
Centre for Materials Science and School of Chemistry and Physics nomena, including current–voltage hysteresis, ferroelectricity,
Queensland University of Technology and poor operational stability.[27–29] In lead halide perovskites,
Gardens Point Campus, Brisbane, QLD 4000, Australia
halide anions are found to be the predominant migrating spe-
The ORCID identification number(s) for the author(s) of this article
can be found under https://doi.org/10.1002/adfm.202002342.
cies owing to its much lower activation energy, while the metal
cations tend to be immobile.[29–31] In terms of the lead-free
DOI: 10.1002/adfm.202002342 double perovskites, there are very limited reports on the ion

Adv. Funct. Mater. 2020, 2002342 2002342 (1 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

migration with majority only on theoretical predictions.[32–35] for optoelectronic device applications to suppress device short-
Recently, Yang et al. demonstrated a heteroepitaxial growth of circuit due to the presence of pin-holes.[16,18]
BiOBr for suppressing ion-migration in the Cs2AgBiBr6 wafer, An optical bandgap of ≈2.39 eV was derived from the
which can effectively improve the stability of its application ultraviolet–visible (UV–vis) spectroscopy for the Cs2AgBiBr6
in X-ray imaging.[36] However, the experimental evidence for considering the indirect bandgap transition (Figure 1c,d).
the ion-migration of this double perovskite still remains rare, Interestingly, a sharp absorption peak presents at around
such as the exact “mobile species,” their diffusive behaviors 425 nm. This intriguing feature exists both in the polycrys-
and influence on the material/device stability. Herein, we talline thin-film and nanocrystals of Cs2AgBiBr6, which was
report a unique ion-migration behavior of high-performance initially attributed to the quantum confinement of excitons.
Cs2AgBiBr6 perovskite solar cells. Unlike the predominant However, a recent study suggests this sharp absorption peak
role of halide in the ion-migration within the heavily studied may originate from the 6s-6p transitions of the isolated octa-
lead-halide perovskites, we found that the Cs2AgBiBr6 based hedral BiBr63− cluster.[39] The photoluminescence (PL) emis-
devices adopt an interesting dual-ion-migration induced deg- sion shows a broad peak at around 513 nm with a large Stokes
radation, where Ag and Br diffusion occurs in the long-term shift (Figure 1e) in comparison to its UV–vis spectrum, which
operation of devices. The dual-ion-migration is intrinsically suggests a strong electron–phonon coupling effect within the
resulted from the low formation energies of Ag and Br vacan- Cs2AgBiBr6 double perovskite.[40] The time-resolved PL life-
cies and low diffusive energy barriers in Cs2AgBiBr6, which time measurement reveals a lifetime of 14.44 ns (Figure 1f).
can be mitigated by using thick hole-transporting material in We further determined the energy band alignments of the
the solar cells. Cs2AgBiBr6 double perovskite thin-film using ultraviolet
photoelectron spectroscopy (UPS), indicating a valence band
and conduction band levels of 6.39 and 4.00 eV (Figure 1g),
2. Results and Discussion respectively. The Fermi level was determined to be 5.15 eV,
confirming the p-type semiconducting of this material. An
2.1. Characterizations of Cs2AgBiBr6 energy-level diagram was summarized based on the different
photovoltaic components (Figure 1h).
High-quality perovskite thin-film is the prerequisite of high-
performance solar cells. By optimizing the deposition para­
meters and annealing temperatures, a compact and uniform 2.2. Photovoltaic Performance
Cs2AgBiBr6 thin-film was developed by using a facile solution
process based on a 0.6 m Cs2AgBiBr6 precursor solution in The double perovskite shows suitable band alignment in
dimethyl sulfoxide (DMSO). It was observed that the film has the photovoltaic devices. Previous work on the Cs2AgBiBr6
better coverage on the substrate and much lower roughness double perovskite solar cells were mainly focusing on the
factors at higher spin-coating rates, as shown in the scanning spiro-OMeTAD as the hole-transporting layer and Au as the
electron images (SEM) and atomic force microscope (AFM), metal contact.[16,18] Spiro-OMeTAD is an expensive small-
respectively (Figures S1 and S2, Supporting Information). The molecular organic material that requires finely controlled
higher spin-coating rate can facilitate solvent spreading and hygroscopic dopants to achieve optimized performance.[41] In
evaporation during deposition processes, leading to a com- this work, low-cost and dopant-free P3HT was selected as the
pact and homogenous thin-film with grain sizes in the range hole-transporting material (HTM). In addition, low-cost Cu was
of 300–500 nm. Cs2AgBiBr6 double perovskites require an used to replace Au as the metal contact, leading to a device
annealing temperature of over 250 °C in the previous report.[37] configuration of fluorine-doped tin oxide (FTO)/compact
The optimized temperature in this work is 285 °C which gives TiO2 (C-TiO2)/Cs2AgBiBr6/poly 3-hexylthiophene (P3HT)/Cu
a well-crystallized Cs2AgBiBr6 thin-film and the best photo- (Figure 2a). A typical cross-sectional device structure is pre-
voltaic performance. Interestingly, it was found that fresh sented in Figure 2b. It was found that the thickness of
precursor solution is critically important for achieving high- the P3HT layer has a significant influence on the photo-
quality Cs2AgBiBr6 thin-films, while aged precursor solutions voltaic performance. Photovoltaic results of the best device
lead to Ag-rich nanoparticles on the film surface, which may using four different thicknesses of 80 ± 6, 65± 5, 45 ± 4,
be due to the degradation of the light-sensitive AgBr in the and 35 ± 5 nm for P3HT are shown in Table S1 (Supporting
precursor solution. According to the X-ray diffraction (XRD) Information). Devices based on the 65 nm P3HT show the best
result, the as-prepared Cs2AgBiBr6 thin-film crystallizes into power-conversion-efficiency (PCE) of around 2% (Figure 2c).
a cubic structure with Fm3̅m space group (Figure 1a), which It’s noteworthy to mention that the PCE achieved is among
is consistent with previous reports.[38,39] In the Cs2AgBiBr6 the best in the same category of Cs2AgBiBr6 solar cells, where
double perovskite, the A, B(I), B(III) and X sites are occupied typical efficiencies were around 1.32% and 1.37% (Table S2,
by Cs, Ag (I), Bi (III), and Br, respectively. Double perovskites Supporting Information).[42] Incident photon-to-current con-
belong to elpasolite structure, where Ag(I) and Bi(III) cen- version efficiency (IPCE) measurement was conducted for the
tered octahedra are alternately connected with each other by best-performing Cs2AgBiBr6 solar cell (Figure S3, Supporting
corner-sharing, leading to a rock-salt configuration in a 3D net- Information). An integrated photocurrent density of 2.53 mA cm−2
work. A compact film has been observed in the SEM surface can be calculated from the IPCE curve, which is compa-
morphology (Figure 1b) with a grain size of a few hundreds rable to the value from the current–voltage measurement
of nanometers. The fully covered thin-film is highly desirable (2.58 mA cm−2).

Adv. Funct. Mater. 2020, 2002342 2002342 (2 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

Figure 1. a) X-ray diffraction (XRD) pattern of Cs2AgBiBr6 thin-film deposited from fresh solution on FTO substrate at 285 °C. b) Top view scanning
electron microscope (SEM) images of Cs2AgBiBr6 on FTO at 4000 rpm spin-coating rate. c) Absorption spectrum of the Cs2AgBiBr6 film on FTO using
the transmission-mode of ultraviolet-visible (UV–vis) measurement. d) Indirect energy bandgap of the Cs2AgBiBr6 film on FTO. e) PL emission of
Cs2AgBiBr6 on the glass substrate. f) PL decays of the Cs2AgBiBr6 on the glass substrate. g) Ultraviolet photoemission spectroscopy (UPS) spectrum
of Cs2AgBiBr6 film. h) Energy band levels of Cs2AgBiBr6 relative to other hole/electron transporting layers.

Devices with configurations of FTO/C-TiO2/Cs2AgBiBr6/ Cs2AgBiBr6 double perovskite thin-film due to the protection
Spiro-OMeTAD/Au and FTO/C-TiO2/Cs2AgBiBr6/P3HT/Au of the polymer layer (Figure S6a,b, Supporting Information).[44]
were also fabricated, exhibiting PCE of 1.38% and 1.77% respec- However, thicker P3HT also increases the series resistance, and
tively (Figure S4, Supporting Information). Although better thus an optimization of the layer thickness on the device per-
device efficiencies can be anticipated for the Au and Spiro- formance is required. When the P3HT thickness increases to
OMeTAD-based double perovskite solar cells by further opti- 80 nm, the device efficiency drops obviously, which may be due
mization, the straightforward comparisons presented here to the increased series resistance and has been observed in the
highlight the viability and potential of using low-cost Cu and previous report.[44] Therefore, increasing the thickness of P3HT
P3HT for these double perovskite solar cells. Spiro-OMeTAD is beneficial for device long-term stability but may comprise the
as the HTL was reported in the Cs2AgBiBr6 solar cells with device efficiency if the HTL is too thick. The improvement of
efficiencies ranging from 1.11–2.5%, where the device optimi- the device stability by increasing the thickness of P3HT was
zations and use of a MoO3 interfacial layer were demonstrated evident from the long-term stability tests (Figure 2d–f). The
to be effective.[17,37,43] Alternative candidates of inorganic hole devices based on the 65 nm P3HT show significantly enhanced
transporting layer (HTL) were also examined, including CuI long-term stability compared to the thinner ones with an effi-
and CuSCN (Figure S5a,b, Supporting Information). Despite ciency retention rate of 73% even after 68 days’ storage in the
their suitable band-alignment relative to the other photovoltaic ambient conditions with a humidity of 68 ± 7%. P3HT as a
layers, inferior device efficiencies were observed. low-cost alternative hole-transporting material was also demon-
strated excellent performance in improving the stability of lead
halide perovskite solar cells.[45]
2.3. Ambient Stability To compare the stability of the devices with varied HTLs,
double perovskite solar cells with configurations of FTO/C-TiO2/
Higher P3HT thickness also improves the reproducibility of Cs2AgBiBr6/Spiro-OMeTAD/Au and FTO/C-TiO2/Cs2AgBiBr6/
device performance (Figure 2d–f), which can be attributed P3HT/Au were assembled and tested. Time-dependent device
to the reduced shunting paths and improved stability of the efficiencies were recorded (Figure S7a,b, Supporting Information),

Adv. Funct. Mater. 2020, 2002342 2002342 (3 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

Figure 2. a) Device configuration of the Cs2AgBiBr6 solar cell. b) Cross-sectional SEM image of the device with a planar structure of FTO/C-TiO2/
Cs2AgBiBr6/P3HT/Cu. c) Current–voltage curve of devices based on the planar structure of FTO/C-TiO2/Cs2AgBiBr6/P3HT/Cu with different P3HT
thicknesses. Statistics of the device efficiency stored in the ambient conditions with different P3HT thicknesses: d) 35 nm, e) 45 nm, f) 65 nm.

suggesting superior ambient stability of the devices using P3HT transmission electron microscope (STEM). Three different
over Spiro-OMeTAD. The comparison between the devices conditions were compared depending on the varied thickness
using Au and Cu (Figure 2f and Figure S7b, Supporting Infor- of the P3HT layer (Figure 3a–d and Figure S8, Supporting
mation) shows that both Au and Cu based devices exhibit Information). It is noted that some bright dots present on the
considerable long-term stability and Au based devices are surface of the Cs2AgBiBr6 thin-film, which can be attributed to
slightly more stable than that of Cu due to the inertness of the Ag-rich phase of the aggregates (Figure 3a,b). The forma-
noble-metal contact. tion of this Ag-rich phase was also observed in the Cs2AgBiBr6
In previous reports, the Cs2AgBiBr6 double perovskite thin- nanocrystals due to the low standard reduction potential of
film shows excellent ambient stability compared to its lead silver.[39] The devices with thin P3HT layers (i.e., 35 and 45 nm)
counterparts and no decomposition was observed after 30 days’ show serious degradation after 68 days’ storage and the inter-
storage at a relative humidity of ≈50% without any encapsula- faces of different device layers are not readily distinguishable.
tion.[26] Wu et al. reported over four months’ ambient stability In particular, there are many cavities in the Cs2AgBiBr6 layer
for nonencapsulated Cs2AgBiBr6 double perovskite solar and almost no trace of the bright-colored layer that belongs
cells.[46] These reports on the superior ambient stability of the to Cu metal contact (Figure 3c). Therefore, the degradation of
Cs2AgBiBr6 double perovskite and its photovoltaic devices the double perovskite layer can explain the observation of the
are encouraging. However, to our knowledge, no degradation faster decay in the long-term operational photovoltaic perfor-
mechanism study on this Cs2AgBiBr6 double perovskite has mance in the devices with 35 and 45 nm P3HT. However, the
been performed to date, which is critically important to prop- device with 65 nm P3HT still retains well-defined Cs2AgBiBr6
erly evaluate its potential for various practical applications. The thin-film even after 68 days’ storage, indicating the beneficial
significantly faster degradation in the devices based on the thin protection effect of the thick P3HT on improving the stability of
P3HT (Figure 2d,e) inspired further study on the underlying the perovskite and the device (Figure 3d). Elemental mapping
reasons for this phenomenon. of the cross-sections of the fresh and aged devices reveals an
interesting observation that Br and Ag tend to migrate through
the HTM layer and accumulate under the Cu metal contact. As
2.4. Ion Diffusion Phenomenon it is illustrated in Figure 4a and Figure S8 (Supporting Infor-
mation), the fresh device shows homogeneous distribution of
To reveal the origins, the lamellae of cross-sectional devices Bi, Ag, Cs, and Br across the Cs2AgBiBr6 thin-film, while the
were prepared by the assistance of the focused ion beam (FIB) S and Cu are originated from the P3HT and Cu metal elec-
technique. High angle annular dark-field (HAADF) images trode, respectively. However, aged device shows completely dif-
were recorded for the cross-sectional views of the fresh and aged ferent morphologies depending on the thickness of the P3HT
Cs2AgBiBr6 double perovskite solar cells by using a scanning layer. The devices based on the 35 and 45 nm P3HT have no

Adv. Funct. Mater. 2020, 2002342 2002342 (4 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

Figure 3. Cross-sectional STEM-HAADF images of the fresh (upper) and aged (lower) devices in ambient conditions with a relative humidity of
68 ± 7% for 68 days in dark with different P3HT thicknesses of a,c) 35 nm and b,d) 65 nm.

well-defined interfaces between the adjacent layers, indicating cell. Despite the formation of the AgBr, it should be noted that
the damage of the layer interface due to the device degradation the decomposition enthalpy (ΔHd) of the Cs2AgBiBr6 is larger
(Figure S8, Supporting Information). Regarding to the device than that of the typical lead halide perovskite, which accords
based on the 65 nm P3HT, it was observed that accumulation with better ambient and thermal stability of the former in
of Ag and Br elements occurs close to the Cu metal contact, the previous reports.[32] Exposure of the Cs2AgBiBr6 to light-
while the other two elements (i.e., Cs and Bi) are remarkably illumination and moisture may lead to accelerated material
stable in the perovskite structure (Figure 4b). degradation.[25] It’s of interest to investigate whether the light-
Line scans of the energy-dispersive X-ray spectroscopy (EDS) illumination may contribute to the ion-migration, which will be
can give a better visual understanding of the compositional dif- further studied in the future research.
ference of each layer in the fresh and aged Cs2AgBiBr6 devices. Depth profiling X-ray photoelectron spectroscopy (XPS) on
Figure 5a reveals that in the fresh sample all the four elements the aged device reveals considerable amount of Br at the upper
contributing to the Cs2AgBiBr6 layer show distinct interfaces side of Cu layer and also near the P3HT/Cu interface in com-
one another in the different photovoltaic layers. Whereas, for parison to the middle position of the Cs2AgBiBr6 layer in the
the aged sample after 68 days’ storage in ambient conditions, device (Figure S9a–c, Supporting Information). The survey
Ag and Br show an obvious diffusion into the P3HT layer and scan unveils the existence of 3.5% atomic ratio of Br near
accumulate under the metal electrode (Figure 5b). In addition, P3HT/Cu interface (Figure S9b, Supporting Information), in
the EDS line scan spectrum of Cu in the aged device shows good agreement with the result from Figure 4b. Another point
a tail that is overlapping with the migrated Br and Ag in the observed here is the presence of very small amount of Ag at
P3HT layer, suggesting the possibility of degradation of Cu the upper side of the Cu layer and P3HT/Cu interface, while
due to reaction between Cu and Br. In contrast, the Cs and Bi no signal corresponding to Bi and Cs was detected in these
show no obvious tendency of ion-migration in both the fresh interfaces (Figure S9a,b, Supporting Information). These XPS
and aged devices. The Ag and Br dual-ion-migration phenom- results further highlight the higher tendency of Br and Ag
enon was further supported by monitoring the XRD spectra of migration than that of Bi and Cs in the Cs2AgBiBr6 thin-film. To
the bare Cs2AgBiBr6 thin-film stored in the ambient air with examine the ion-migration behavior of elements in the absence
a relative humidity of 68 ± 7% (Figure 5c). Additional XRD of Cu metal contact, we further performed a stability test on a
peaks for AgBr can be identified with enhanced peak inten- device with the structure of FTO / C-TiO2/Cs2AgBiBr6/P3HT
sity after 68 days’ storage. AgBr was suggested as one of the without any Cu metal contact on top of the P3HT layer. The
possible resultants upon decomposing of the Cs2AgBiBr6. The device was stored in the relative humidity of 50% for 38 days
observation of the AgBr from the XRD is consistent with the in the dark. It is known that the majority of the XPS signals
results from the EDS line scan of the aged Cs2AgBiBr6 solar are collected on the sample surface within 10 nm penetration

Adv. Funct. Mater. 2020, 2002342 2002342 (5 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

Figure 4. Cross-sectional STEM-HAADF images of a) fresh and b) aged devices in ambient conditions with a humidity of 68 ± 7% for 68 days in dark
with P3HT thickness of 65 nm. The figures of the right are corresponding EDS mappings of the cross-sectional STEM-HAADF images in the fresh and
aged device.

length. Considering the 65 nm P3HT layer on the top, it’s Br < Ag < Cs < Bi. Therefore, the Br and Ag ions are more
anticipated that the fresh sample shows little elemental signal likely to migrate compared to that of Cs and Bi when intrinsic
from the double perovskite layer underneath, while the ion dif- vacancies exist. According to the theoretical calculation results,
fusion into the P3HT after long-term storage can contribute the internal Br and Ag ions in the Cs2AgBiBr6 crystal network
to the XPS results. Ag and Br signals were detected after the can migrate through the vacancies and finally diffused into the
long-term storage in the humid ambient conditions, as shown P3HT layer, while the Bi and Cs ions hardly move due to much
in Figure S10 (Supporting Information). This further confirms larger energy barriers and ionic radii. The Br and Ag vacan-
that Ag and Br ions diffuse through the HTL even without the cies have the lowest formation energies thus these vacancies
Cu metal contact. Note that the elemental ratio of the migrated can abound in the Cs2AgBiBr6, which could potentially pro-
Ag and Br in the XPS results are underestimated in compar- vide more paths for the diffusion of Ag and Br ions.[32] This
ison to that of EDS mapping in Figure 4, which may be attrib- phenomenon may be extended to other halide double perov-
uted to the differences in the storage conditions of the samples skites, especially those contain Ag and/or Br in their struc-
and the mechanism of the characterization techniques. tures (e.g., Cs2AgBiCl6 and Cs2AgInBr6). Further theoretical
To further understand this dual-ion-diffusion phenom- and experimental studies are required to understand the ion-
enon, computational simulations were carried out to explore migration mechanism in other double-perovskite compounds.
the energy barriers for Cs, Bi, Ag, and Br ion-migration to The energy bandgap of Cs2AgBiBr6 is calculated using dif-
their adjacent vacancy position within the 3D Cs2AgBiBr6 net- ferent methods (Table S3 and Figure S11, Supporting Infor-
work (Figure 6a,b). The energy barriers for Cs, Bi, Ag, and mation). The optimized model for the calculations provides
Br ions migration are 1.413, 3.363, 0.895, and 0.438/0.427 eV simulated lattice parameters and bond lengths that are reason-
(Br1/Br2), respectively (Figure 6c), which follow the order of ably close to the experimental values, indicating its reliability

Adv. Funct. Mater. 2020, 2002342 2002342 (6 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

Figure 5. Cross-sectional STEM-HAADF images and line scan EDS of the device based on 65 nm P3HT a) Fresh device and b) Aged device in a relative
humidity of 68 ± 7% for 68 days. c) XRD spectra of the Cs2AgBiBr6 thin-film stored in ambient conditions with a humidity of 68 ± 7% for 68 days in the
dark. The peaks with “*” indicate FTO peaks while the peaks with “@” can be assigned to AgBr.

(Table S4, Supporting Information). The vacancy formation Bi has less possibility to migrate due to its large ionic radii
energy (ΔEf ) was calculated by ΔEf = Evac + Eele − Eperfect, where and high ion-migration energy barrier (3.363 eV). Therefore,
Evac is the energy of the model containing the vacancy, Eele is Ag and Br tend to be the dominant ions that prone to migrate
the energy per atom of the vacant species in the elemental in the Cs2AgBiBr6. Thus it is reasonable to propose a vacancy-
phase, and Eperfect is the energy of the perfect Cs2AgBiBr6 bulk mediated ion migration in the Cs2AgBiBr6. After the forma-
structure. The calculated Ef for Cs, Ag, Bi, and Br vacancy is tion of the intrinsic vacancies in the Cs2AgBiBr6, the Br and
4.74, 0.59, 3.55, and 2.56 eV respectively. It is clear that the Ag near the surface of the thin-film are more readily to leave
Ag and Br vacancies have the lowest formation energy thus can be the perovskite lattice and migrate through the P3HT/perovskite
the main vacancy defects in Cs2AgBiBr6, which is consistent with interface, leaving much higher concentrations of Br and Ag
the previous report.[32] We also calculated the formation energy vacancies near the Cs2AgBiBr6 surface. The bulk elements may
of Cs2AgBiBr6 (ΔEfbulk) by ΔEfbulk = Ebulk − 2ECs – EAg – EBi – 6EBr diffuse toward the surface due to the concentration gradient to
where Ebulk is the energy of Cs2AgBiBr6 bulk phase per chem- achieve a thermodynamic equilibrium, leading to a continuous
ical formula and ECs, EAg, EBi, and EBr are the energy per atom loss of Ag and Br in the bulk perovskite and consequently an
of elemental Cs, Ag, Bi, and Br respectively. The calculated accumulation of the Br and Ag elements in the P3HT layer
ΔEfbulk is −12.30 eV, which is close to the reported value and the P3HT/Cu interface. It is interesting to observe that two
of −11.36 eV.[32] oppositely charged ions (i.e., Ag+ and Br−) can diffuse to the
same direction and accumulate between the metal contact and
the P3HT layer. We speculate that the volatility of the halide
2.5. Degradation Mechanism and reactivity of the halide with metal elements (Ag and Cu)
may partially contribute to this phenomenon. In addition to the
A possible mechanism for dual-ion-migration in the Cs2AgBiBr6 formation of AgBr in the Cu and P3HT interface, the migrated
was proposed based on the experimental and theoretical studies Br may react with the Cu metal contact to form copper/
(Figure 6d). In principle, various intrinsic point defects in the cuprous bromide. This can be supported by the experimental
Cs2AgBiBr6 can potentially form, including elemental vacan- observation of accelerated degradation of both Cs2AgBiBr6 and
cies, interstitials, cation-on-cation antisites, cation-on-anion Cu metal electrode in the devices based on thin P3HT layer,
antisites, and anion-on-cation antisites.[32] Most of the forma- where the inhomogeneous P3HT increases the chance of direct
tion enthalpies of these defects are relatively too high to facili- contact between the Cs2AgBiBr6 thin-film and Cu, leading to
tate their formation. However, Ag vacancies (VAg) can readily much faster deterioration of both the perovskite layer and the
form, leading to the p-type conductivity in the Cs2AgBiBr6.[32,47] device. However, this degradation can be effectively mitigated
Bi vacancies and Br vacancies/interstitials can also be the domi- by using a thicker P3HT layer. Suppressing the ion-migration
nant defects depending on the growth conditions.[32] However, has been proven effective in improving the long-term stability

Adv. Funct. Mater. 2020, 2002342 2002342 (7 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

Figure 6. a,b) Atomic layers of the (0.25 0 0) and (0.5 0 0) surfaces of Cs2AgBiBr6 respectively. The ion migration paths are shown by arrows with
different colors and the vacancies are marked by the dotted circle. c) Energy profiles of ion migrations shown in (a,b) achieved by the climbing-image
nudged elastic band (NEB) method. d) Schematic of degradation mechanism of Cs2AgBiBr6 solar cell.

of lead halide perovskite solar cells, including interfacial engi- to the vacancy-mediated dual-ion-migration of the Ag and Br,
neering on the surface of perovskite layer/grain boundaries which can be relieved at least by using a thick hole-transporting
and chemical bonding enhanced cations/anions immobiliza- material in the solar cells. The new mechanism reported herein
tion.[27,28] Developing viable strategies for enhancing the long- emphasizes the importance of fundamental mechanism under-
term stability is also critically imperative for any future practical standing on the stability of halide double perovskites for better
implementations of the Cs2AgBiBr6 double perovskite consid- solar cell design. On the other hand, this unique property that
ering the low toxicity, interesting optoelectronic properties and involves a mixed ionic, photonic and electronic feature in the
various demonstrated applications. The new understandings Cs2AgBiBr6 may stimulate potential applications of halide
of the ion-migration in the lead-free halide double perovskites double perovskites in other electronic/optoelectronic devices
can promote designing strategies for further materials and such as artificial synapses and photonic memristors.
device stability improvement. Despite the adverse influence on
the stability, it’s noted that the ion-migration phenomenon in
the metal halide perovskites is a “double-edged sword,” which
has been found to facilitate the fast response in photodetectors 4. Experimental Section
by forming dynamic electronic junctions[48,49] and also demon- All steps of materials deposition and device fabrication were performed
strate important applications in resistance-switching memory in glovebox except spray pyrolysis of compact TiO2 layers.
devices[50–52] and artificial synapses.[53] Cs2AgBiBr6 Thin-Film Preparation: For Cs2AgBiBr6 thin-film deposition,
first 0.6 mmol of BiBr3 (bismuth(III) bromide 99.999%, Alfa Aesar) and
0.6 mmol of AgBr (silver bromide 99%, Sigma Aldrich) and 1.2 mmol of
CsBr (caesium bromide, 99.999%, Sigma Aldrich) were dissolved in DMSO
3. Conclusion (dimethyl sulfoxide ≥99%, Sigma Aldrich) to prepare a 0.6 m solution.
The solution was stirred overnight at a temperature of 80 °C to
In summary, this work reported efficient planar double perov- completely dissolve all the powders and prepare a transparent solution.
skite solar cells with an efficiency of 1.91% by fabricating high- The final solution was filtered with polytetrafluoroethylene (PTFE) syringe
quality Cs2AgBiBr6 thin-films. The long-term stability study filters (0.45 µm). Then this transparent solution was used to deposit the
Cs2AgBiBr6 film. In this regard, 45 µL of the Cs2AgBiBr6 solution was
reveals that an unusual dual-ion-migration of Br and Ag is an dispersed onto the fluorine doped tin oxide (FTO) substrate (FTO22-7,
intrinsic property in the Cs2AgBiBr6 thin-film and can lower its Yingkou OPV Tech New Energy Co. Ltd), and then the solution was spin-
long-term operational stability. The low formation energies of the coated at different deposition rate of 2000 rpm (revolutions per minute),
Ag and Br vacancies and low diffusive energy barriers contribute 3000, 4000, and 5000 rpm for 45 s with a ramping rate of half of the

Adv. Funct. Mater. 2020, 2002342 2002342 (8 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

deposition rate (1000, 1500, 2000, and 2500 rpms−1 respectively). The Supporting Information
spin-coated samples were annealed at different annealing temperatures
for 20 min with the best condition at 285 °C. Supporting Information is available from the Wiley Online Library or
Device Fabrication: For device fabrication, FTO glass slides were from the author.
cleaned sequentially by ultra-sonicating in acetone, ethanol, and
isopropanol for 10 min. Then the surface was further cleaned with
UV–Ozone for 20 min. Afterward, a 0.42 m precursor solution of
titanium diisopropoxide bis(acetylacetonate) (75 wt% in isopropanol) in Acknowledgements
anhydrous 2-propanol was used to deposit a thin compact-TiO2 (C-TiO2) M.G. and L.Z. contributed equally to this work. The research leading
layer by spray pyrolysis. The light-absorbing layer of Cs2AgBiBr6 was to these results was financially supported through Discovery, Laureate
deposited according to the aforementioned steps. For planar-structured Fellowship and Linkage programs by the Australian Research Council.
devices, the Cs2AgBiBr6 layer was directly deposited onto the C-TiO2 layer. This work was performed in part at the Queensland node of the
The hole transporting layer was prepared by spin-coating pristine P3HT Australian National Fabrication Facility, a company established under the
(Sigma-Aldrich, Mw 54 000–75 000, 15 mg mL−1 in 1,2-dichlorobenzene) National Collaborative Research Infrastructure Strategy to provide nano-
solution at 2000 rpm for 30 s at 1000 rpms−1. Annealing of the P3HT and micro-fabrication facilities for Australia’s researchers. The support
layer at 100 °C for 15 min was conducted to dry the layer and remove from the Australian Microscopy & Microanalysis Research Facility at the
the additional solvent. For Spiro-OMeTAD layer deposition, 72.3 mg of Centre for Microscopy and Microanalysis, The University of Queensland
2,2′,7,7′-tetrakis(N,N-di-p-methoxyphenylamine)-9,9-spirobifluorene (UQ) and UQI scholarship are also acknowledged.
(spiro-OMeTAD, Merck) was dissolved in 1 mL anhydrous chlorobenzene
(CB, 99.8%, Sigma–Aldrich) containing 28.8 µL of 4-tert-butylpyridine
and 17.5 µL of lithium bis(trifluoromethanesulfonyl)imide (Li-TFSI)
solution (520 mg of Li-TFSI in 1 mL of acetonitrile, Sigma–Aldrich). The Conflict of Interest
fresh solution was deposited at 5000 rpm for 30 s at 2500 rpms−1. Finally,
The authors declare no conflict of interest.
≈60 nm Cu or Au was thermally deposited as the contact electrode by a
thermal evaporator (DD ONG high technology). The aperture area of the
solar cells was 0.08 cm2.
Characterizations and Measurements: For characterizations, powder Keywords
X-ray diffraction measurement was performed on a Bruker Advanced X-ray
Diffractometer with Cu Kα radiation from 10° to 50° (2θ). A Field-Emission halide double perovskites, ion migration, lead-free, solar cell, stability
Scanning Electron Microscope (FE-SEM, JEOL 7100) recorded scanning
electron microscope images of the sample morphologies. Absorption Received: March 12, 2020
spectra of the films were performed on a UV–visible light spectrometer Revised: June 22, 2020
(Shimadzu UV2450). Kratos Axis Supra Photoelectron Spectroscopy Published online:
was used for ultraviolet photoelectron spectroscopy measurement.
A fluorescence spectrophotometer (FLSP-920, Edinburgh Instruments)
was used to measure the steady-state and time-resolved PL spectra.
[1] A. Kojima, K. Teshima, Y. Shirai, T. Miyasaka, J. Am. Chem. Soc.
A solar simulator (AM 1.5, 100 mW cm−2, Oriel) equipped with a Keithley
2420 digital source meter was used to record current–voltage (I–V) plots. 2009, 131, 6050.
A Newport 1918-c power meter equipped with a 300 W Oriel xenon light [2] M. M. Lee, J. Teuscher, T. Miyasaka, T. N. Murakami, H. J. Snaith,
source and an Oriel Cornerstone 260¼ m monochromator in a DC mode Science 2012, 338, 643.
was used to measure the incident photon-to-current conversion efficiency. [3] J. Burschka, N. Pellet, S.-J. Moon, R. Humphry-Baker, P. Gao,
A standard Si reference was used before the IPCE measurement. For M. K. Nazeeruddin, M. Grätzel, Nature 2013, 499, 316.
transmission electron microscope (TEM) lamella sample preparation [4] NREL, Best Research Cell Efficiencies, http://www.nrel.gov/ncpv/
a Scios DualBeam focused ion beam-scanning electron microscope imaGes/efficiency_chart.jpg (accessed: January 2020).
(FIB-SEM, Oregon, USA) system was used to prepare lamellae on TEM [5] M. Gratzel, Nat. Mater. 2014, 13, 838.
lift-out grids. XPS with Al Kα (hν = 1253.6 eV) radiation source was used to [6] A. Babayigit, A. Ethirajan, M. Muller, B. Conings, Nat. Mater. 2016,
characterize the chemical composition of the samples and C 1s (284.8 eV) 15, 247.
was used to calibrate binding energies. For the stability test, devices were [7] N. K. Noel, S. D. Stranks, A. Abate, C. Wehrenfennig, S. Guarnera,
stored in ambient conditions without any control on humidity for 68 days A.-A. Haghighirad, A. Sadhanala, G. E. Eperon, S. K. Pathak,
in the dark. The average humidity was calculated to be 68 ± 7%. M. B. Johnston, Energy Environ. Sci. 2014, 7, 3061.
Theoretical Calculation: The band structure of bulk Cs2AgBiBr6 is [8] F. Hao, C. C. Stoumpos, D. H. Cao, R. P. Chang, M. G. Kanatzidis,
calculated by Heyd–Scuseria–Ernzerhot (HSE06) method with the spin- Nat. Photonics 2014, 8, 489.
orbital coupling (SOC) effect.[54] The density functional theory (DFT) [9] A. J. Lehner, D. H. Fabini, H. A. Evans, C.-A. Hébert, S. R. Smock,
calculations were carried out using the Vienna ab-initio simulation J. Hu, H. Wang, J. W. Zwanziger, M. L. Chabinyc, R. Seshadri, Chem.
package (VASP).[55–57] Generalized gradient approximation (GGA) of the
Mater. 2015, 27, 7137.
Perdew–Burke–Ernzerhof (PBE) type[58,59] and the projector augmented
[10] M. Lyu, J.-H. Yun, M. Cai, Y. Jiao, P. V. Bernhardt, M. Zhang,
wave (PAW)[60] method were applied for the exchange-correlation
Q. Wang, A. Du, H. Wang, G. Liu, Nano Res. 2016, 9, 692.
potential and the electron-core interaction respectively. The van der
[11] R. L. Z. Hoye, R. E. Brandt, A. Osherov, V. Stevanović, S. D. Stranks,
Waals (vdW) interactions were taken into consideration by applying
Grimme method.[61] The plane-wave energy cutoff was set to 300 eV for M. W. B. Wilson, H. Kim, A. J. Akey, J. D. Perkins, R. C. Kurchin,
all calculations and a k-point mesh of 3 × 3 × 3 was applied to sample J. R. Poindexter, E. N. Wang, M. G. Bawendi, V. Bulović,
the first Brillouin zone. The bulk structure of Cs2AgBiBr6 was fully relaxed T. Buonassisi, Chem. - Eur. J. 2016, 22, 2605.
(both the lattice and the atomic positions) until the residual force per [12] M. Ghasemi, M. Lyu, M. Roknuzzaman, J.-H. Yun, M. Hao, D. He,
atom was less than 0.03 eV Å−1. The energy barriers for ion migrations Y. Bai, P. Chen, P. V. Bernhardt, K. K. Ostrikov, J. Mater. Chem. A
in Cs2AgBiBr6 were studied by the climbing-image nudged elastic band 2019, 7, 20733.
(cNEB) method.[62,63] Nine images were interpolated between the initial [13] C. Ni, G. Hedley, J. Payne, V. Svrcek, C. McDonald,
and final states linearly and the convergence criteria for the force were L. K. Jagadamma, P. Edwards, R. Martin, G. Jain, D. Carolan,
set to 0.05 eV Å−1. Nat. Commun. 2017, 8, 170.

Adv. Funct. Mater. 2020, 2002342 2002342 (9 of 10) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.afm-journal.de

[14] L. R. Morss, M. Siegal, L. Stenger, N. Edelstein, Inorg. Chem. 1970, [37] E. Greul, M. L. Petrus, A. Binek, P. Docampo, T. Bein, J. Mater.
9, 1771. Chem. A 2017, 5, 19972.
[15] I. Flerov, M. Gorev, K. Aleksandrov, A. Tressaud, J. Grannec, [38] M. Pantaler, K. T. Cho, V. I. E. Queloz, I. G. Benito, C. Fettkenhauer,
M. Couzi, Mater. Sci. Eng., R 1998, 24, 81. I. Anusca, M. K. Nazeeruddin, D. C. Lupascu, G. Grancini,
[16] M. Pantaler, K. T. Cho, V. I. Queloz, I. s. García Benito, ACS Energy Lett. 2018, 3, 1781.
C. Fettkenhauer, I. Anusca, M. K. Nazeeruddin, D. C. Lupascu, [39] Y. Bekenstein, J. C. Dahl, J. M. Huang, W. T. Osowiecki, J. K. Swabeck,
G. Grancini, ACS Energy Lett. 2018, 3, 1781. E. M. Chan, P. D. Yang, A. P. Alivisatos, Nano Lett. 2018, 18, 3502.
[17] F. Igbari, R. Wang, Z.-K. Wang, X.-J. Ma, Q. Wang, K.-L. Wang, [40] J. A. Steele, P. Puech, M. Keshavarz, R. X. Yang, S. Banerjee,
Y. Zhang, L.-S. Liao, Y. Yang, Nano Lett. 2019, 19, 2066. E. Debroye, C. W. Kim, H. F. Yuan, N. H. Heo, J. Vanacken,
[18] W. Ning, F. Wang, B. Wu, J. Lu, Z. Yan, X. Liu, Y. Tao, J. M. Liu, A. Walsh, J. Hofkens, M. B. J. Roeffaers, ACS Nano 2018, 12, 8081.
W. Huang, M. Fahlman, Adv. Mater. 2018, 30, 1706246. [41] M. Zhang, M. Lyu, H. Yu, J. H. Yun, Q. Wang, L. Wang, Chem. - Eur.
[19] W. C. Pan, H. D. Wu, J. J. Luo, Z. Z. Deng, C. Ge, C. Chen, J. 2015, 21, 434.
X. W. Jiang, W. J. Yin, G. D. Niu, L. J. Zhu, L. X. Yin, Y. Zhou, [42] C. C. Wu, Q. H. Zhang, Y. Liu, W. Luo, X. Guo, Z. R. Huang, H. Ting,
Q. G. Xie, X. X. Ke, M. L. Sui, J. Tang, Nat. Photonics 2017, 11, 726. W. H. Sun, X. R. Zhong, S. Y. Wei, S. F. Wang, Z. J. Chen, L. X. Xiao,
[20] W. N. Yuan, G. D. Niu, Y. M. Xian, H. D. Wu, H. M. Wang, H. Yin, Adv. Sci. 2018, 5, 1700759.
P. Liu, W. Z. Li, J. D. Fan, Adv. Funct. Mater. 2019, 29, 1900234. [43] W. H. Ning, F. Wang, B. Wu, J. Lu, Z. B. Yan, X. J. Liu, Y. T. Tao,
[21] J. Yang, C. Bao, W. Ning, B. Wu, F. Ji, Z. Yan, Y. Tao, J.-M. Liu, J. M. Liu, W. Huang, M. Fahlman, L. Hultman, T. C. Sum, F. Gao,
T. C. Sum, S. Bai, J. Wang, W. Huang, W. Zhang, F. Gao, Adv. Opt. Adv. Mater. 2018, 30, 1706246.
Mater. 2019, 7, 1801732. [44] P. Zhou, T. Bu, S. Shi, L. Li, Y. Zhang, Z. Ku, Y. Peng, J. Zhong,
[22] Z. Weng, J. Qin, A. A. Umar, J. Wang, X. Zhang, H. Wang, X. Cui, Y.-B. Cheng, F. Huang, J. Mater. Chem. C 2018, 6, 5733.
X. Li, L. Zheng, Y. Zhan, Adv. Funct. Mater. 2019, 29, 1902234. [45] E. H. Jung, N. J. Jeon, E. Y. Park, C. S. Moon, T. J. Shin, T.-Y. Yang,
[23] Z. Z. Zhang, Y. Q. Liang, H. L. Huang, X. Y. Liu, Q. Li, L. X. Chen, J. H. Noh, J. Seo, Nature 2019, 567, 511.
D. S. Xu, Angew. Chem., Int. Ed. 2019, 58, 7263. [46] C. Wu, Q. Zhang, Y. Liu, W. Luo, X. Guo, Z. Huang, H. Ting, W. Sun,
[24] H. L. Wu, X. B. Li, C. H. Tung, L. Z. Wu, Adv. Mater. 2019, 31, X. Zhong, S. Wei, Adv. Sci. 2018, 5, 1700759.
1900709. [47] T. S. Li, X. G. Zhao, D. W. Yang, M. H. Du, L. J. Zhang, Phys. Rev.
[25] E. T. McClure, M. R. Ball, W. Windl, P. M. Woodward, Chem. Mater. Appl. 2018, 10, 041001.
2016, 28, 1348. [48] H. Wang, M. Zhou, H. Luo, ACS Omega 2018, 3, 1445.
[26] A. H. Slavney, T. Hu, A. M. Lindenberg, H. I. Karunadasa, [49] T. Zhang, C. Hu, S. Yang, Small Methods 2019, 4, 1900552.
J. Am. Chem. Soc. 2016, 138, 2138. [50] X. Guan, W. Hu, M. A. Haque, N. Wei, Z. Liu, A. Chen, T. Wu,
[27] N. Li, S. Tao, Y. Chen, X. Niu, C. K. Onwudinanti, C. Hu, Z. Qiu, Adv. Funct. Mater. 2018, 28, 1704665.
Z. Xu, G. Zheng, L. Wang, Nat. Energy 2019, 4, 408. [51] E. J. Yoo, M. Lyu, J. H. Yun, C. J. Kang, Y. J. Choi, L. Wang,
[28] E. Bi, H. Chen, F. Xie, Y. Wu, W. Chen, Y. Su, A. Islam, M. Grätzel, Adv. Mater. 2015, 27, 6170.
X. Yang, L. Han, Nat. Commun. 2017, 8, 15330. [52] F. Zhou, Y. Liu, X. Shen, M. Wang, F. Yuan, Y. Chai, Adv. Funct.
[29] G. Y. Kim, A. Senocrate, T.-Y. Yang, G. Gregori, M. Grätzel, J. Maier, Mater. 2018, 28, 1800080.
Nat. Mater. 2018, 17, 445. [53] Y. Sun, L. Qian, D. Xie, Y. Lin, M. Sun, W. Li, L. Ding, T. Ren,
[30] Y. Yuan, J. Huang, Acc. Chem. Res. 2016, 49, 286. T. Palacios, Adv. Funct. Mater. 2019, 29, 1902538.
[31] Y. Luo, P. Khoram, S. Brittman, Z. Zhu, B. Lai, S. P. Ong, [54] A. V. Krukau, O. A. Vydrov, A. F. Izmaylov, G. E. Scuseria, J. Chem.
E. C. Garnett, D. P. Fenning, Adv. Mater. 2017, 29, 1703451. Phys. 2006, 125, 224106.
[32] Z. W. Xiao, W. W. Meng, J. B. Wang, Y. F. Yan, ChemSusChem 2016, [55] G. Kresse, J. Furthmüller, Phys. Rev. B 1996, 54, 11169.
9, 2628. [56] G. Kresse, J. Furthmüller, Comput. Mater. Sci. 1996, 6, 15.
[33] J. Xu, J. B. Liu, B. X. Liu, B. Huang, J. Power Sources 2019, 412, 689. [57] G. Kresse, J. Hafner, Phys. Rev. B 1994, 49, 14251.
[34] C. F. Lan, S. Zhao, J. T. Luo, P. Fan, Phys. Chem. Chem. Phys. 2018, [58] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865.
20, 24339. [59] J. P. Perdew, M. Ernzerhof, K. Burke, J. Chem. Phys. 1996, 105, 9982.
[35] J. Xu, J. B. Liu, B. X. Liu, J. F. Wang, B. Huang, Adv. Funct. Mater. [60] G. Kresse, D. Joubert, Phys. Rev. B 1999, 59, 1758.
2019, 29, 1805870. [61] S. Grimme, J. Comput. Chem. 2006, 27, 1787.
[36] B. Yang, W. C. Pan, H. D. Wu, G. D. Niu, J. H. Yuan, K. H. Xue, [62] G. Henkelman, H. Jónsson, J. Chem. Phys. 2000, 113, 9978.
L. X. Yin, X. Y. Du, X. S. Miao, X. Q. Yang, Q. G. Xie, J. Tang, [63] G. Henkelman, B. P. Uberuaga, H. Jónsson, J. Chem. Phys. 2000,
Nat. Commun. 2019, 10, 1989. 113, 9901.

Adv. Funct. Mater. 2020, 2002342 2002342 (10 of 10) © 2020 Wiley-VCH GmbH