Articles

https://doi.org/10.1038/s41565-020-0765-7

2D metal–organic framework for stable perovskite

solar cells with minimized lead leakage
Shengfan Wu1, Zhen Li1, Mu-Qing Li   1,2, Yingxue Diao   1,3, Francis Lin   1, Tiantian Liu1, Jie Zhang1,
Peter Tieu   4, Wenpei Gao5, Feng Qi1, Xiaoqing Pan   5,6,7, Zhengtao Xu   1 ✉, Zonglong Zhu   1 ✉ and
Alex K.-Y. Jen   1,3,8 ✉

Despite the notable progress in perovskite solar cells, maintaining long-term operational stability and minimizing potentially
leaked lead (Pb2+) ions are two challenges that are yet to be resolved. Here we address these issues using a thiol-functionalized
2D conjugated metal–organic framework as an electron-extraction layer at the perovskite/cathode interface. The resultant
devices exhibit high power conversion efficiency (22.02%) along with a substantially improved long-term operational stability.
The perovskite solar cell modified with a metal–organic framework could retain more than 90% of its initial efficiency under
accelerated testing conditions, that is continuous light irradiation at maximum power point tracking for 1,000 h at 85 °C. More
importantly, the functionalized metal–organic framework could capture most of the Pb2+ leaked from the degraded perovskite
solar cells by forming water-insoluble solids. Therefore, this method that simultaneously tackles the operational stability and
lead contamination issues in perovskite solar cells could greatly improve the feasibility of large-scale deployment of perovskite
photovoltaic technology.

T
he certified power conversion efficiency (PCE) of organic–
inorganic lead halide perovskite solar cells (PVSCs) has
reached a new high at 25.2%, which can rival those of
market-dominating inorganic photovoltaic technologies, includ-
ing polycrystalline silicon (p-Si), thin film copper indium gallium
selenide and cadmium telluride1–5. The demonstrated high PCE
indicates that it is no longer a limiting factor for commercializing
PVSCs. As such, research efforts in this field have been gradually
shifted to focus on the reliability and stability issues of PVSCs under
ambient conditions, light soaking and thermal stresses6–9. By apply-
ing composition and crystal engineering, encapsulation, electrode
selections, defect passivation and interfacial modifications, the
device lifetime can be prolonged by excluding external environ-
mental tensions10–13. Nevertheless, long-term stability of PVSCs
under operational conditions has yet to be realized, which remains
a critical hurdle to overcome before their large-scale implementa-
tion. Moreover, compared to the thermally and chemically stable
cadmium telluride having low solubility in water (with a solubility
product constant (Ksp) below 10−30) and presenting a lower toxic-
ity threat to the environment, lead (Pb)-containing compounds in
perovskites (or degraded perovskites) have a much higher Ksp of
10−8 in rainwater11,14–16. Therefore, to improve PVSC operational
stability effectively and eliminate/minimize Pb2+ leakage from the
degraded perovskite absorbers, an integrated solution with proper
materials and device configurations is required to enable PVSCs for
practical applications.
Metal–organic frameworks (MOFs) have been recently explored
in the perovskite photovoltaic community due to their potential
for enhancing photovoltaic performance17–21. This stems from the
relative ease in tuning their functionalities as a versatile class of
well-defined hybrid materials. So far, there are only very few cases
that report the use of MOFs in PVSCs, indicating the very embry-
onic stage of this area. Among the limited studies, MOFs have been
applied as scaffolds to template the growth of perovskites or as addi-
tives/surface modifiers to passivate the defects of perovskites for
enhancing the device performance and stability18,21. However, it is
challenging to use a MOF as an efficient charge-transporting layer
(CTL) because most of them are quite insulating with low carrier
mobilities19. Additionally, the MOFs utilized in PVSCs are mostly
unfunctionalized prototypes that lack the diverse chemical func-
tionalities needed for integrating with PVSCs.
To address these problems, we have developed a simple and
effective strategy by applying a 2D conjugated MOF that pos-
sesses n-type electrical behaviour and suitable energy levels as an
electron-extraction layer (EEL) at the perovskite/cathode interface.
The use of Zr(iv) ions to engage selectively the hard carboxyl groups
allows dense free-standing thiol arrays to be built around the Zr(iv)–
oxo cluster to offer steric shielding and stabilization22,23. The dense
array of thiol groups on the MOF is particularly important because of
their capabilities in rigidifying the networks with crosslinked disul-
fide linkages and in trapping heavy metal ions (such as lead, mercury
and so on). In addition, the thiol-functionalized MOF can also form
intimate contact with the silver electrode to extract electrons effi-
ciently from PVSCs by reducing the contact resistance and trapping
mobile Pb2+ ions at the perovskite/electrode interface to mitigate the
potential impact of the device on environmental sustainability.
Characterization of 2D MOF and perovskite/MOF films
As shown in Fig. 1a, a new linker molecule H3L3 was first synthesized
(Supplementary Figs. 1–4). The peak numbers and assignments

Department of Chemistry, City University of Hong Kong, Kowloon, Hong Kong. 2Frontier Institute of Science and Technology (FIST), Xi’an Key Laboratory
1

of Sustainable Energy and Materials Chemistry, Xi’an Jiaotong University, Shanxi, China. 3Department of Materials Science and Engineering, City University
of Hong Kong, Kowloon, Hong Kong. 4Department of Chemistry, University of California-Irvine, Irvine, CA, USA. 5Department of Materials Science and
Engineering, University of California-Irvine, Irvine, CA, USA. 6Department of Physics and Astronomy, University of California-Irvine, Irvine, CA, USA. 7Irvine
Materials Research Institute, University of California-Irvine, Irvine, CA, USA. 8Department of Materials Science and Engineering, University of Washington,
Seattle, WA, USA. ✉e-mail: zhengtao@cityu.edu.hk; zonglzhu@cityu.edu.hk; alexjen@cityu.edu.hk

Nature Nanotechnology | www.nature.com/naturenanotechnology

Articles Nature Nanotechnology

a b c d
O OH

HS SH

HS SH

HO O

O SH SH OH 5 nm 5 nm

H3L3 ZrL3

e f
500
0.6
0.4 r = 0.999994 r = 0.999991
N2 uptake (cm3 g–1 STP)

400 0.4 Boiling water treated ZrL3

0.2
300 0.2
Activited ZrL3
0 0
0 0.02 0.04 0.06 0 0.02 0.04 0.06
200
As-made ZrL3
100
528 m2 g–1 369 m2 g–1 Calculated from the structural model of ZrL3
0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 10 20 30 40
Relative pressure (P/P0) Relative pressure (P/P0) 2θ (°)

g h i
6
Pero Pero Pero/ZrL3
ZrL3-as made 100 Pero/PC61BM Pero/PC61BM/ZrL3
Pero/ZrL3
5 Band gap = 2.53 eV 12 Pero/PC61BM

Normalized intensity (a.u.)

PL intensity (×105 counts)

Pero/PC61BM/ZrL3
4 9 10
0

intensity (a.u.)
10–1
α (S)

Normalized
3 10
–1

6
2 –2
10
3
1
10–2 0 25 50 75 100 125 150
Time (ns)
0 0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 650 700 750 800 850 0 150 300 450 600 750
Energy (eV) Wavelength (nm) Time (ns)

Fig. 1 | Characterization of ZrL3 and the perovskite films with different EELs. a, The thiol-decked carboxyl linker H3L3. b, Structural model of corresponding
2D MOF ZrL3. c, Bright field image. d, Model of ZrL3. e, N2 sorption isotherm at 77 K and Brunauer–Emmett–Teller plot (insets) for the activated ZrL3 before
(left) and after (right) boiling-water-treatment. f, PXRD patterns of ZrL3. g, UV-vis diffuse reflectance spectrum of the as-made ZrL3. h,i, PL (h) and TRPL
(i) spectra of the perovskite films with different EELs.

in the NMR spectra confirmed the molecular structure of the should be noted that their ratios here cannot be pinpointed by the
H3L3 linker and indicated the high purity of the H3L3 sample used current data as their small weight fractions do not impact signifi-
for assembling the Zr(iv)-based MOF product, ZrL3. The corre- cantly the elemental and TGA results.
sponding polycrystalline framework solid, ZrL3 (Fig. 1b), was then The morphology, micro and atomic structures of the 2D MOF
solvothermally prepared, using ZrOCl2∙8H2O as the metal source, were also characterized by scanning electron microscopy (SEM) and
formic acid as the modulator and ethylene dithiol for stabilizing the transmission electron microscopy (TEM). The SEM image of ZrL3
thiol groups on the organic linker. The peak S–H stretch at around features the morphology of thin platelets with dimensions between
2,555 cm−1 can be clearly observed from the FT-IR spectra of both 100 and 500 nm (Supplementary Fig. 10). The low magnification
H3L3 and ZrL3 (Supplementary Figs. 5 and 6), indicating the suc- image, fast Fourier transform and electron diffraction show the dis-
cessfully installed thiol groups in the ZrL3 networks. The powder tinct hexagonal arrays of the 2D layers in ZrL3, as demonstrated
X-ray diffraction (PXRD) pattern of as-made ZrL3 (Supplementary in Supplementary Fig. 11. Moreover, atomic resolution high angle
Fig. 7) can be indexed onto hexagonal unit cells (parameters: annular dark field images from aberration corrected scanning trans-
a = 19.98 Å, c = 7.24 Å (see Supplementary Note 1). The short c axis mission electron microscopy indicate the direct (eclipsed) stacking
suggests a single-layer structure based on Zr6O8 clusters as it com- of the MOF layers (Supplementary Fig. 12). Individual Zr atoms
pares well with the corresponding interlayer distance in a known are seen within the Zr clusters at high magnification. The surface
structure. Elemental analysis and thermogravimetric analysis clusters of a thin region were found to be more diffuse compared
(TGA) results fit the formula Zr6O4(OH)4(L3)2(OH)6(H2O)27(DMF) to the interior. The organic linkers between Zr clusters were seen
(DMF = N,N-dimethylformamide) for ZrL3 (Supplementary Fig. 8 in the bright field image and the surface of ZrL3 was also shown
and Note 2). The linker/Zr6 cluster ratio thus determined suggests to be terminated by the organic linkers as in the model (Fig. 1c,d).
full occupancy for the linker sites in ZrL3. The capping sites can TEM studies also suggest superior stability of the 2D networks as
be occupied by the formate (HCOO−) and HO−/H2O species, but it ZrL3 was stable under prolonged parallel electron beam radiation

Nature Nanotechnology | www.nature.com/naturenanotechnology

Nature Nanotechnology
(~30 e− Å−2 s−1) while imaging. The stability of the ZrL3 lattice is
further verified by gas sorption studies and PXRD. The N2 sorption
isotherm (77 K) features dominant type-I characteristics of micro-
porous solids (Fig. 1e), with additional uptake at high P/P0 regions
that can be attributed to capillary condensation in mesopores (for
example, arising from the inter-particle/layer spaces).
The Brunauer–Emmett–Teller surface area is calculated to be
528 m2 g−1 for ZrL3, which agrees with the simulated accessible
solvent surface of 594 m2 g−1 (Supplementary Fig. 13). To probe
respectively the hydrolytic and thermal stability, ZrL3 was placed
in boiling water for 24 hours and it was also heated at 100 °C in air
for 7 days. The boiling-water-treated sample still showed substan-
tial porosity as per gas sorption studies, even though the surface
area was somewhat decreased (369 m2 g−1, see also Supplementary
Figs. 14 and 15 for the density functional theory analysis of the N2
sorption data). Moreover, the PXRD pattern of ZrL3 after either
boiling-water-treatment or continual heating at 100  °C in air
remained strong, indicating the retention of the crystalline phase
(Fig. 1f and Supplementary Fig. 16). These results demonstrated the
superior moisture and thermal stability of ZrL3. The optical band-
gap of ZrL3 was determined and calculated to be 2.53 eV from its
ultraviolet-visible (UV-vis) diffuse reflectance spectrum, as shown
in Fig. 1g. Additionally, the absorption spectra were also obtained
to study the effect of ZrL3 on light harvesting of the perovskite layer
(Supplementary Fig. 17). No obvious change could be found in the
absorption spectra with varied thicknesses of ZrL3 and the bandgap
of the mixed-cation (Cs/formamidinium (FA)/methylammonium
(MA)) perovskite was calculated to be 1.60 eV from the Tauc plot.
The electron- and hole-only devices were fabricated to study the
electrical properties of ZrL3 by the space-charge-limited-current
method (Supplementary Figs. 18 and 19 and Note 3)24. As a result,
the electron- and hole-mobilities were calculated to be 2.81 × 10−7
m2 V−1 s−1 and 5.36 × 10−9 m2 V−1 s−1, respectively. The relatively
higher electron mobility indicated that ZrL3 is more of a n-type
material, which would be appropriate for extracting electrons
in PVSCs. To investigate the interfacial charge carrier dynam-
ics between perovskite and different EELs, steady-state photolu-
minescence (PL) and time-resolved photoluminescence (TRPL)
spectra were obtained for the perovskite films coated with ZrL3,
phenyl-C61-butyric acid methyl ester (PC61BM) and PC61BM/ZrL3
layers, respectively. As can be seen in Fig. 1h, the pure perovskite
film showed the strongest PL intensity due to the efficient radia-
tive recombination of the photogenerated carriers, while it was
apparently quenched upon contacting with the EELs. We note
that the perovskite film showed substantially quenched PL when
it contacted with ZrL3, suggesting efficient charge extraction from
perovskite to ZrL3 (ref. 25,26). This phenomenon is further con-
firmed by the TRPL spectra, as shown in Fig. 1i and summarized
in Supplementary Table 1. The perovskite/ZrL3 film exhibited biex-
ponential decay with a fast (1.6 ns) and a slow (13.8 ns) component
after being excited by a pulsed laser. This decay is much shorter
than for the perovskite-only sample (137.8 and 632.8 ns). In addi-
tion, the perovskite/PC61BM/ZrL3 film exhibited a slightly faster PL
decay than the perovskite/PC61BM film, which might be attributed
to the cascade energy alignment in perovskite/PC61BM/ZrL3 (see
Supplementary Fig. 20 and Note 4).
Device architecture and photovoltaic performance
The inverted planar PVSCs with the device configuration of indium
tin oxide (ITO) glass/poly(triarylamine) (PTAA)/perovskite/
PC61BM/EEL/silver (Ag) were fabricated and the corresponding
energy alignment is provided in Fig. 2a,b. The device without any
EEL showed a relatively low PCE of 18.80% with a fill factor (FF) of
71.21%, which was attributed to the non-ideal electron extraction
from PC61BM to the electrode and the insufficient hole-blocking
ability. Encouragingly, by introducing ZrL3 as the EEL between
Nature Nanotechnology | www.nature.com/naturenanotechnology
Articles
PC61BM and Ag, a much higher FF of 76.42% along with a higher
PCE of 20.34% was achieved (Supplementary Figs. 21 and 22). These
results suggest the enhanced electron extraction and hole-blocking
ability originated from ZrL3, which is consistent with the TRPL
results. To extract electrons even more efficiently, a bis-C60 surfac-
tant (previously demonstrated to be able to tune the cathode work
function for improving electron extraction in both organic photo-
voltaic devices and PVSCs (ref. 27–30)) was further incorporated with
ZrL3 as a hybrid EEL. Although the mobilities of the ZrL3:bis-C60
hybrid EEL are similar to those of ZrL3, the reduced energy barrier
with the cathode will facilitate electron collection (Supplementary
Fig. 23).
As expected, the PCE of the device with the ZrL3:bis-C60 EEL
(denoted as M-PVSC) could be further increased to 22.02% with
an FF of 81.28%, open-circuit voltage (Voc) of 1.20 V and negligible
hysteresis. This is among the highest values reported for the pla-
nar inverted PVSCs (Fig. 2c, Supplementary Figs. 24 and 25 and
Table 2). The short-circuit current density (Jsc) obtained from the
J–V curve matched well with the integrated value from the external
quantum efficiency (EQE) (Fig. 2d), demonstrating the good reli-
ability of the results. The ‘champion’ (best) cell of M-PVSC was fur-
ther held at the maximum power point (MPP) with a bias voltage
of 1.03 V under continuous 1 sun illumination (AM 1.5G, 100 mW
cm−2) to obtain a stabilized photocurrent of 20.97 mA cm−2, which
corresponds to a stabilized PCE of 21.60% (Fig. 2e). More impor-
tantly, the efficiency of the encapsulated M-PVSC was certified
to be 21.3% by an independent accredited institute (the National
Institute of Metrology, China), which is close to the value measured
in house (Supplementary Fig. 26).
To elucidate the potential roles ZrL3 and bis-C60 played in the
hybrid EEL for device efficiency, the PVSCs using only bis-C60
as the EEL were fabricated as the reference device (denoted as
r-PVSC), which affords a slightly lower PCE of 21.32% than that of
M-PVSC (Supplementary Fig. 27 and Table 3). The device perfor-
mance of the PVSCs with different EELs is summarized in Fig. 2f
and Supplementary Tables 3–5. There are no obvious enhancements
in Voc and Jsc, while the enhanced PCE for M-PVSCs mainly comes
from the FF improvements (Supplementary Fig. 28). To explore the
effect of EEL morphology on the improved FF, atomic force micros-
copy was employed to characterize the morphology of bis-C60,
ZrL3 and ZrL3:bis-C60 prepared on perovskite/PC61BM film. The
roughness of perovskite/PC61BM was slightly reduced from 10.02 to
8.86 nm after coating with bis-C60 (Supplementary Fig. 29).
By contrast, the film roughness of both perovskite/PC61BM/
ZrL3 and perovskite/PC61BM/ZrL3:bis-C60 films were increased
to 36.31 nm and 35.45 nm, respectively and the morphologies of
perovskite/PC61BM/ZrL3 and perovskite/PC61BM/ZrL3:bis-C60
were quite similar. These results implied that the film morphol-
ogy of hybrid EEL was dominated by ZrL3 and it should not be the
reason for the varied FF between the devices with different EELs.
Furthermore, given that the hole-blocking ability of the EEL can sig-
nificantly affect the FF in photovoltaic devices, the energy levels of
ZrL3, bis-C60 and ZrL3:bis-C60 were then calculated and are sum-
marized in Supplementary Table 6 (also Supplementary Figs. 30 and
31 and Note 4). It was found that bis-C60 possesses a larger energy
barrier for the holes than ZrL3. Therefore, it can potentially provide
better hole-blocking ability which is consistent with the higher FF
observed in r-PVSC than the devices with ZrL3 EEL alone.
On the other hand, it is known that the thiol group can inter-
act with Ag by forming S–Ag bonds, which can reduce the contact
resistance at the interface31–34. Thus, it is expected that the S–Ag
bonds would form between the dense thiol and disulfide groups
in ZrL3 and the Ag electrode to enable M-PVSCs to have better
EEL-Ag contact than r-PVSCs. This allows M-PVSCs to have lower
contact resistance and higher FF, which is consistent with our pre-
vious report33. Based on these considerations, the high efficiency
Articles Nature Nanotechnology

a b c
25

Current density (mA cm–2)

20
3.9
4.1

Energy level (–eV)

60 4.3
ZrL3:bis-C 15 Voc = 1.20 V

PTAA
PC 61BM Jsc = 22.58 mA cm–2

Perovskite
Ag
ITO

PC61BM
FF = 81.28%

ZrL3:bis-C60
10
PCE = 22.02%
Perovskite 4.8
5.2 5
PTAA 5.5
te
ITO substra 0
0 0.2 0.4 0.6 0.8 1.0 1.2
Voltage (V)

d e f
100 25 25 12
Average PCE = 20.85%
80 20 10
20
Stablized at 20.97 mA cm–2
8
Integrated Jsc

J (mA cm–2)

60 15
EQE (%)

15

Counts
6
40 10 10
4
20 5 5
EQE 2
Integrated Jsc In N2-filled glovebox
0 0 0
300 400 500 600 700 800 0 10 20 30 40 50 60 19 20 21 22 23
Wavelength (nm) Time (min) PCE (%)

Fig. 2 | Device structure and performance of the PVSCs with ZrL3:bis-C60 EEL. a, Device structure of inverted PVSC. b, Energy level alignment of the used
materials. c, J–V curve of the champion PVSC with ZrL3:bis-C60 EEL. d, EQE spectrum and integrated current density of the champion PVSC with ZrL3:bis-C60
EEL. e, Stabilized power output at the MPP for the PVSC with ZrL3:bis-C60 EEL. f, PCE histogram obtained from 30 PVSCs with ZrL3:bis-C60 EEL.

achieved for M-PVSCs should be mainly attributed to bis-C60, Long-term device stability
while ZrL3 with the dense thiol groups plays a role in reducing the To verify the feasibility of using a ZrL3:bis-C60 EEL in enhancing
contact resistance, thus slightly improving the FF. device stability, a comprehensive study was carried out on the encap-
The resultant PVSCs were also characterized by light intensity (I) sulated PVSCs in both humid air and nitrogen atmosphere. The
dependent photovoltage and photocurrent density. Supplementary M-PVSC showed superior shelf stability in an ambient environment
Fig. 32 shows the linear relationships between Jsc and I in (with a RH of 75%), maintaining 90% of its initial efficiency after
double-logarithmic scale and the fitted α values are close to 1, which 1,100 hours (Fig. 3a). By contrast, the PCE of r-PVSC dropped sig-
demonstrates the efficient carrier transport and the few bimolecular nificantly to <50% of its original value. The significantly enhanced
recombinations for both devices. Additionally, the slopes were cal- stability can be attributed to the strong coordinating ability of Zr(iv)
culated from the semi-natural logarithmic plot of Voc versus I. The ions and the dense thiol and crosslinked disulfide linkages in ZrL3,
M-PVSC exhibited a slightly lower slope (1.41 kBT/q) than r-PVSC which protect perovskites against moisture and oxygen38.
(1.57 kBT/q), where kB, T and q are Boltzmann’s constant, absolute Long-term operational stability of the devices was tested under
temperature and elementary charge, respectively. These results a white LED lamp with the light intensity equivalent to 1 sun in
indicate that the trap-assisted Shockley–Read–Hall recombination a N2-filled glovebox (Fig. 3b). Specifically, the operational stability
is reduced in M-PVSC because of the better contact between ZrL3 was tested under an accelerated ageing condition: (1) no UV filter
and the Ag electrode, which contributes to faster charge extraction for the light source and (2) the devices were monitored at an ele-
in M-PVSC. vated temperature of 85 °C to estimate their potential for long-term
In a real-use scenario, perovskites and CTLs in a complete PVSC stability. The ZrL3-incorporated cell was tracked at the MPP and
are susceptible to hydration and oxidation due to abundant mois- the corresponding PCEs obtained from the J–V curve were con-
ture and an oxygen atmosphere35–37. To highlight the superior sta- stantly monitored during the test (Fig. 3c). The PCE of M-PVSC
bility of ZrL3, perovskite/PC61BM films were prepared initially on only dropped slightly to 92% of its initial value after 1,000 hours of
glass and coated with either bis-C60 or ZrL3:bis-C60 hybrid EEL ageing. Thus, M-PVSC is among the most stable inverted PVSCs
to investigate their ability to protect perovskites against moisture tested under the MPP tracking condition at 85 °C. Moreover, it is
and oxygen. The perovskite/PC61BM film with only bis-C60 EEL observed that the constantly recorded PCEs from the J–V curves of
bleached almost completely after being placed in an ambient envi- M-PVSC during MPP tracking showed a similar decay trend, dem-
ronment with a relative humidity (RH) of 75% for over two weeks onstrating the good reliability of the testing (Fig. 3c). In the litera-
(Supplementary Fig. 33). In sharp contrast, no obvious degrada- ture, small-molecule based CTLs/EELs tend to crystallize and form
tion was observed for the perovskite/PC61BM film protected by the aggregates under thermal stress. This creates pathways for direct
ZrL3:bis-C60 hybrid EEL. Besides, the contact angle with a drop interactions between perovskite and metal electrode and so impairs
of water on top of ZrL3:bis-C60 protected perovskite/PC61BM film the device performance37. The superior stability of ZrL3 under ther-
rose from 68.37 to 81.23° (Supplementary Fig. 34), implying ZrL3 mal stress significantly enhanced the thermal stability of M-PVSC,
can indeed help prevent moisture from penetrating into perovskite which was further verified by the stability profile of the PVSCs with
film, which is consistent with its enhanced film stability. only ZrL3 as the EEL (Supplementary Fig. 35).

Nature Nanotechnology | www.nature.com/naturenanotechnology

Nature Nanotechnology Articles
a d
Ag ETL Perovskite HTL ITO Ag ETL Perovskite HTL ITO
20 ZrL3:bis-C60
105 Bis-C60
105
I I
16 104 104
PCE (%)

Intensity

Intensity
12 103 Ag 103 PbI2
PbI2 Ag
2
8 10 102
M-PVSC
r-PVSC In humid air 101 101 Zr
4
0
10 100
0 150 300 450 600 750 900 1,050
0 500 1,000 1,500 0 500 1,000 1,500
Time (h)
Sputtering time (s) Sputtering time (s)

b e Ag ETL Perovskite HTL ITO

Ag ETL Perovskite HTL ITO
1.0 105 ZrL3:bis-C60
105 Bis-C60
Normalized PCE

I I
0.8 104 104

Intensity
Intensity
0.6 103 Ag
PbI2
103 Ag PbI2

M-PVSC 10 2
102
0.4
r-PVSC MPP in N2-filled glovebox 101 101 Zr
0.2 Lead leakage
10 0
100
0 200 400 600 800 1,000
0 500 1,000 1,500 0 500 1,000 1,500
Time (h)
Sputtering time (s) Sputtering time (s)

c f
H2O

20 H2O

H2O

16
PCE (%)

Lead immobilization
Zr
L3

via S–Pb bond

12
:b C 61
is BM
P

-C
60

8
Pe
ro
vs
kit

Pb Pb
e

Pb
C O S Pb Zr
4
0 200 400 600 800 1,000
Time (h)

Fig. 3 | Long-term stability studies for the PVSCs with bis-C60 (r-PVSC) or ZrL3:bis-C60 (M-PVSC) EEL. a, Shelf stability of the encapsulated PVSCs in
humid air with RH of 75%. b, Long-term operational stability of the encapsulated PVSCs under MPP tracking at 85 °C under continuous light irradiation
with a white LED lamp (100 mW cm−2) in a N2-filled glovebox. c, PCEs constantly recorded from the J–V curves of M-PVSC during MPP tracking. d,e,
ToF-SIMS depth profiles of the pristine (d) and aged (e) PVSCs with bis-C60 (left) or ZrL3:bis-C60 (right) EEL. f, Schematic of degradation process of
PVSCs and the immobilization effect of ZrL3 on leaked Pb2+ ions.

Lead contamination of PVSCs
To probe the possible sources and mechanisms that cause the deg-
radation in our devices, the depth profiles of Pb, Ag and I ions
throughout the un-encapsulated PVSCs were investigated by using
time-of-flight secondary-ion mass spectrometry (ToF-SIMS) (Fig.
3d,e). The Pb2+ signal could be clearly observed to shift towards
the Ag electrode through the electron-transporting layer in the
r-PVSC after it was aged under the accelerated condition in air
(85 °C and 85% RH). This indicates that the perovskite films can
be easily decomposed into Pb-containing by-products and ammo-
nium salts under such an aggressive ageing condition, resulting in
Pb2+ diffusion into other layers35,39. Very encouragingly, almost no
Pb2+ ions could be detected in the Ag layer of M-PVSC and they
were “trapped” at the ZrL3:bis-C60/Ag interface to form Pb(ii)–
ZrL3 complexes. Unlike the physical encapsulation methods used
in reducing lead leakage, this in situ chemical adsorption of Pb2+ by
ZrL3:bis-C60 EEL in the device is a much more effective and sus-
tainable process for long-term practical applications (Fig. 3f).
To quantify the adsorption ability of ZrL3 for Pb2+ ions, sorption
kinetics and sorption isotherm measurements were obtained (Fig.
4a,b and Supplementary Notes 5 and 6). A distribution coefficient
Kd of around 105 ml g−1, a substantial value of 355 mg g−1 for the
Pb adsorption capacity (qmax) and a sorption rate constant (pseudo
first-order) of 0.103 min−1 were achieved. These results strongly
support the idea that ZrL3 EEL can effectively adsorb Pb2+ by rapid
chemical reactions between thiols and Pb2+ to minimize the leak-
age of Pb2+ ions into the environment. To investigate the lead leak-
age that occurred in the degraded PVSCs, they were immersed in
deionized water with a pH value of ~5.6 (simulating acidic rain)
to compare the concentration of water soluble Pb components
between M-PVSCs and r-PVSCs, as shown in Fig. 4c. Pb2+ testing
paper was used to perform a quick check on the Pb2+ concentra-
tion in the contaminated water and it was found that the aqueous
extracts from the r-PVSC showed a much higher Pb2+ concentration
compared with those of the M-PVSC.
An inductively coupled plasma optical emission spectroscopy
(ICP-OES) instrument was further used to quantify accurately
the Pb2+ concentrations (Fig. 4d and Supplementary Fig. 36).
The Pb2+ concentration decreased drastically to an average value
of 7.6 ppm for M-PVSCs compared to 38.4 ppm for r-PVSCs and
38.1 ppm for pure perovskite film, which means over 80% of the
leaked Pb2+ ions from the degraded perovskite could be captured
by thiol-functionalized ZrL3 to form water-insoluble Pb(ii)–ZrL3
complexes. This substantially reduced lead leakage was attributed
to the chemical reaction between the dense array of thiol and
disulfide groups on ZrL3 with leaked Pb2+ ions from the degraded
perovskites. Although this Pb(ii)-trapping process can be further
optimized, our results demonstrate a promising strategy to improve
simultaneously PVSC long-term operational stability and reduce/
prevent Pb2+ leakage into the environment. This will greatly facili-
tate the large-scale deployment of highly efficient PVSCs for clean
energy applications.
Conclusion
We have developed an effective strategy to address simultaneously
the most challenging operational stability and lead contamination

Nature Nanotechnology | www.nature.com/naturenanotechnology

Articles Nature Nanotechnology

a b
10 25
250

Absorbed amount (mg g–1)

20

Absorbed amount (mg g–1)

8 1.2 2.8
Pb concentration (ppm)

200
15

ln(Ct)
0.8 2.4

Ce/qe (g L–1)
6 10
0.4 150 2.0

5 1.6
4 0.0
100
0 2 4 6 8 10 12 14 16 1.2
0 t (min)
0.8
2 50
0 60 120 180 240 300 360 420
Time (min) 150 300 450 600 750
Ce (mg L–1)
0 0
0 60 120 180 240 300 360 420 0 150 300 450 600 750 900

Time (min) Equilibrium concentration (Ce) (mg L–1)

c d
60
r-PVSC
Lead concentration (ppm)
50 M-PVSC
0 ppm

40

M-PVSC M-PVSC
10 ppm 30

20
20 ppm
10

r-PVSC r-PVSC 0
40 ppm
1 2 3 4 5 6 7 8 9 10 11 12
Immersed into deionized water
Devices

Fig. 4 | Lead contamination of the PVSCs with bis-C60 (r-PVSC) or ZrL3:bis-C60 (M-PVSC) EEL. a, Pb(ii) sorption kinetics of an activated ZrL3 sample
(20.0 mg) at an initial Pb(ii) concentration of 10.0 ppm. The two insets are the corresponding absorbed Pb(II) amount and the Lagergren first-order kinetic
plot of the Pb(II) sorption, respectively. b, Pb(ii) Langmuir sorption isotherm of the activated sample of ZrL3. The inset is the linear fit with the Langmuir
adsorption model. c, Degraded PVSCs with bis-C60 (r-PVSC) or ZrL3:bis-C60 (M-PVSC) EEL are immersed into the deionized water with a pH value of
~5.6 and Pb ion testing paper is used to quickly detect the Pb concentration in the contaminated water. d, Pb concentration in the contaminated water was
accurately detected by ICP-OES measurements.

issues in PVSCs by incorporating a functionalized 2D MOF as 4. Yoshikawa, K. et al. Silicon heterojunction solar cell with interdigitated back
an EEL into a PVSC. The resultant inverted PVSC achieved a contacts for a photoconversion efficiency over 26%. Nat. Energy 2, 17032 (2017).
high PCE of >22% (a certified value of 21.3%) and substantially 5. Kim, M. et al. Methylammonium chloride induces intermediate phase
stabilization for efficient perovskite solar cells. Joule 3, 2179–2192 (2019).
improved long-term operational stability. More importantly, the 6. Bai, S. et al. Planar perovskite solar cells with long-term stability using ionic
thiol-containing MOF can capture most of the leaked Pb2+ (over liquid additives. Nature 571, 245–250 (2019).
80%) from degraded PVSCs by forming a water-insoluble com- 7. Christians, J. A. et al. Tailored interfaces of unencapsulated perovskite solar
plex. These combined results showed the promise of simultane- cells for >1,000 hour operational stability. Nat. Energy 3, 68–74 (2018).
ously realizing stable and environmental-friendly PVSCs through 8. Li, N. et al. Cation and anion immobilization through chemical bonding
enhancement with fluorides for stable halide perovskite solar cells. Nat.
multi-functional MOFs. Energy 4, 408–415 (2019).
9. Wang, L. et al. A Eu3+-Eu2+ ion redox shuttle imparts operational durability to
Online content Pb-I perovskite solar cells. Science 363, 265–270 (2019).
Any methods, additional references, Nature Research report- 10. Niu, T. et al. High performance ambient-air-stable FAPbI3 perovskite solar
cells with molecule-passivated Ruddlesden–Popper/3D heterostructured film.
ing summaries, source data, extended data, supplementary infor-
Energy Environ. Sci. 11, 3358–3366 (2018).
mation, acknowledgements, peer review information; details of 11. Jiang, Y. et al. Reduction of lead leakage from damaged lead halide perovskite
author contributions and competing interests; and statements of solar modules using self-healing polymer-based encapsulation. Nat. Energy 4,
data and code availability are available at https://doi.org/10.1038/ 585–593 (2019).
s41565-020-0765-7. 12. Alharbi, E. A. et al. Atomic-level passivation mechanism of ammonium salts
enabling highly efficient perovskite solar cells. Nat. Commun. 10, 3008 (2019).
13. Chen, J., Zhao, X., Kim, S.-G. & Park, N.-G. Multifunctional chemical linker
Received: 9 December 2019; Accepted: 13 August 2020; imidazoleacetic acid hydrochloride for 21% efficient and stable planar
Published: xx xx xxxx perovskite solar cells. Adv. Mater. 31, 1902902 (2019).
14. Zayed, J. & Philippe, S. Acute oral and inhalation toxicities in rats with
References cadmium telluride. Int J. Toxicol. 28, 259–265 (2009).
1. Jiang, Q. et al. Surface passivation of perovskite film for efficient solar cells. 15. Yu, Y., Hong, Y., Gao, P. & Nazeeruddin, M. K. Glutathione modified gold
Nat. Photon. 13, 460–466 (2019). nanoparticles for sensitive colorimetric detection of Pb2+ ions in rainwater
2. Yoo, J. J. et al. An interface stabilized perovskite solar cell with high stabilized polluted by leaking perovskite solar cells. Anal. Chem. 88, 12316–12322 (2016).
efficiency and low voltage loss. Energy Environ. Sci. 12, 2192–2199 (2019). 16. Hailegnaw, B., Kirmayer, S., Edri, E., Hodes, G. & Cahen, D. Rain on
3. Jung, E. H. et al. Efficient, stable and scalable perovskite solar cells using methylammonium lead iodide based perovskites: Possible environmental
poly(3-hexylthiophene). Nature 567, 511–515 (2019). effects of perovskite solar cells. J. Phys. Chem. Lett. 6, 1543–1547 (2015).

Nature Nanotechnology | www.nature.com/naturenanotechnology

Nature Nanotechnology
17. Wu, H. B. & Lou, X. W. Metal-organic frameworks and their derived
materials for electrochemical energy storage and conversion: Promises and
challenges. Sci. Adv. 3, eaap9252 (2017).
18. Li, M. et al. Doping of [In2(phen)3Cl6]·CH3CN·2H2O indium-based
metal-organic framework into hole transport layer for enhancing perovskite
solar cell efficiencies. Adv. Energy Mater. 8, 1702052 (2018).
19. Ryu, U. et al. Nanocrystalline titanium metal–organic frameworks for highly
efficient and flexible perovskite solar cells. ACS Nano 12, 4968–4975 (2018).
20. Lee, C. C., Chen, C. I., Liao, Y. T., Wu, K. C. & Chueh, C. C. Enhancing
efficiency and stability of photovoltaic cells by using perovskite/Zr-MOF
heterojunction including bilayer and hybrid structures. Adv. Sci. 6,
1801715 (2019).
21. Chang, T. H. et al. Planar heterojunction perovskite solar cells incorporating
metal–organic framework nanocrystals. Adv. Mater. 27, 7229–7235 (2015).
22. Yee, K.-K. et al. Effective mercury sorption by thiol-laced metal–organic
frameworks: in strong acid and the vapor phase. J. Am. Chem. Soc. 135,
7795–7798 (2013).
23. Wong, Y. L., Diao, Y., He, J., Zeller, M. & Xu, Z. A thiol-functionalized
UiO-67-type porous single crystal: Filling in the synthetic gap. Inorg. Chem.
58, 1462–1468 (2019).
24. Ren, H. et al. Efficient and stable Ruddlesden–Popper perovskite solar
cell with tailored interlayer molecular interaction. Nat. Photon. 14,
154–163 (2020).
25. Lu, J. et al. Interfacial benzenethiol modification facilitates charge transfer
and improves stability of cm-sized metal halide perovskite solar cells with up
to 20% efficiency. Energy Environ. Sci. 11, 1880–1889 (2018).
26. Tan, F. et al. In situ back-contact passivation improves photovoltage and fill
factor in perovskite solar cells. Adv. Mater. 31, e1807435 (2019).
27. Zhu, Z., Chueh, C.-C., Li, N., Mao, C. & Jen, A. K. Y. Realizing efficient
lead-free formamidinium tin triiodide perovskite solar cells via a sequential
deposition route. Adv. Mater. 30, 1703800 (2018).
28. Zhu, Z. et al. Highly efficient and stable perovskite solar cells enabled by
all-crosslinked charge-transporting layers. Joule 2, 168–183 (2018).
Nature Nanotechnology | www.nature.com/naturenanotechnology
Articles
29. Li, C.-Z. et al. Effective interfacial layer to enhance efficiency of polymer solar
cells via solution-processed fullerene-surfactants. J. Mater. Chem. 22,
8574–8578 (2012).
30. Zhu, Z., Chueh, C.-C., Lin, F. & Jen, A. K. Y. Enhanced ambient stability of
efficient perovskite solar cells by employing a modified fullerene cathode
interlayer. Adv. Sci. 3, 1600027 (2016).
31. Gan, W., Xu, B. & Dai, H.-L. Activation of thiols at a silver nanoparticle
surface. Angew. Chem. Int. Ed. 50, 6622–6625 (2011).
32. Fenter, P. et al. Structure of octadecyl thiol self-assembled on the silver (111)
surface: an incommensurate monolayer. Langmuir 7, 2013–2016 (1991).
33. Laibinis, P. E. et al. Comparison of the structures and wetting properties of
self-assembled monolayers of n-alkanethiols on the coinage metal surfaces,
copper, silver, and gold. J. Am. Chem. Soc. 113, 7152–7167 (1991).
34. Yip, H.-L., Hau, S. K., Baek, N. S. & Jen, A. K. Y. Self-assembled monolayer
modified ZnO/metal bilayer cathodes for polymer/fullerene
bulk-heterojunction solar cells. Appl. Phys. Lett. 92, 193313 (2008).
35. Tiep, N. H., Ku, Z. & Fan, H. J. Recent advances in improving the stability of
perovskite solar cells. Adv. Energy Mater. 6, 1501420 (2016).
36. Wang, S. et al. Unveiling the role of tBP–LiTFSI complexes in perovskite solar
cells. J. Am. Chem. Soc. 140, 16720–16730 (2018).
37. Li, M. et al. Interface modification by ionic liquid: A promising candidate for
indoor light harvesting and stability improvement of planar perovskite solar
cells. Adv. Energy Mater. 8, 1801509 (2018).
38. Chen, W., Xu, L., Feng, X., Jie, J. & He, Z. Metal acetylacetonate series in
interface engineering for full low-temperature-processed, high-performance,
and stable planar perovskite solar cells with conversion efficiency over 16%
on 1 cm2 scale. Adv. Mater. 29, 1603923 (2017).
39. Yang, J. & Kelly, T. L. Decomposition and cell failure mechanisms in lead
halide perovskite solar cells. Inorg. Chem. 56, 92–101 (2017).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2020
Articles
Methods
Materials. The starting materials, reagents and solvents for MOF synthesis were
purchased from commercial sources, including J&K, Sigma-Aldrich and Anaqua
and were used without further purification. Caesium iodide (CsI), formamidinium
iodide (FAI) and methylammonium bromide (MABr) were purchased from
Dysol. Lead iodide (PbI2) and lead bromide (PbBr2) were purchased from TCI.
Phenyl-C61-butyric acid methyl ester (PC61BM) was purchased from ADS.
Poly[bis(4-phenyl)(2,4,6-trimethylphenyl) amine] (PTAA) and bathocuproine
(BCP, purity of 99.9%) were purchased from Xi’an Polymer Light Technology
Corporation. The solvents, including DMF, dimethyl sulfoxide (DMSO),
isopropanol (IPA) and chlorobenzene (CB) were purchased from J&K and used as
received. Toluene was purchased from Tian Hang Technology Co., Ltd. Silver with
high purity was purchased from commercial sources.
Synthesis of methyl 2,6-bis(benzylthio)-4-bromobenzoate (S1). 4-Bromo-
2,6-difluorobenzoic acid (4.74 g, 0.02 mol) and potassium carbonate (27.6 g,
0.20 mol) were loaded in a 250 ml two neck round-bottom flask. N-Methyl-
2-pyrrolidone (100 ml) was added to the flask and the mixture was stirred and
bubbled with nitrogen at room temperature for 30 min. Benzyl mercaptan (6.0 ml,
0.05 mol) was then added to the mixture under nitrogen. The mixture was stirred
at 80 °C for 4 days. After cooling to room temperature, methyl iodide (3.0 ml,
48.19 mmol) was added to the mixture and the mixture was stirred at room
temperature for 30 min. The resulting mixture was poured into distilled water
(800 ml) and extracted with dichloromethane (DCM) (3 × 200 ml). The combined
organic layer was dried over anhydrous MgSO4, evaporated on a rotary evaporator
and purified by column chromatography (eluent: DCM/hexane, 1:5) to afford S1
as a: white solid (6.96 g, 76% yield based on 4-bromo-2,6-difluorobenzoic acid); 1H
NMR (300 MHz, CDCl3): δ = 7.34–7.18 (m, 12H), 4.05 (s, 4H), 3.91 (s, 3H) ppm;
13
C NMR (75 MHz, CDCl3): δ = 167.35, 138.72, 136.35, 135.67, 133.41, 129.23,
128.69, 127.70, 123.22, 52.70, 40.48 ppm.
Synthesis of dimethyl 1,3,5-tris[3′,5′-bis(benzylthio)-4′-(methoxycarbonyl)
phenyl] benzene (M3). Molecule S1 (2.00 g, 4.36 mmol), bis(pinacolato)diboron
(1.107 g, 4.36 mmol), PdCl2(PPh3)2 (77 mg, 0.11 mmol) and anhydrous potassium
acetate (941 mg, 9.59 mmol) were loaded in a 250 ml two neck round-bottom
flask and dried under vacuum for 30 min. The flask was then connected to a
nitrogen atmosphere. Anhydrous 1,4-dioxane (30 ml, bubbled with nitrogen for
10 min beforehand) was added to the flask. The mixture was stirred at 90 °C for
12 h. After cooling to room temperature, K3PO4 aqueous solution (2.3 M, 5.0 ml
bubbled with nitrogen for 10 min beforehand) was added, followed by the addition
of 1,3,5-tribromobenzene (400 mg, 1.27 mmol). The mixture was stirred at 90 °C
for 12 h. After cooling to room temperature, the mixture was poured into water
(400 ml) and extracted with DCM (3 × 100 ml). The combined organic phases
were washed with distilled water (3 × 200 ml). The organic phase was then dried
over anhydrous MgSO4 and evaporated with a rotary evaporator. The solid residue
obtained was purified by flash column chromatography (eluent: DCM/n-hexane/
ethyl acetate, 1:1:0.01) to obtain M3 as a: white solid (987 mg, 64% yield based on
1,3,5-tribromobenzene); 1H NMR (300 MHz, CDCl3): δ = 7.27 (s, 6H), 7.16–7.26
(m, 30H), 6.88 (s, 3H), 4.11 (s, 12H), 4.03 (s, 9H) ppm; 13C NMR (75 MHz, CDCl3):
δ = 168.01, 142.17, 141.46, 140.49, 137.12, 133.02, 132.58, 129.26, 128.58, 127.58,
125.36, 52.70, 41.11 ppm.
Synthesis of 1,3,5-tris[3′,5′-bis(benzylthio)-4′-carboxyphenyl]benzene (MA3).
Molecule M3 (485 mg, 0.40 mmol) was dissolved in tetrahydrofuran (7.0 ml) in a
100 ml two neck round-bottom flask, followed by adding KOH solution (8.9 M,
7.0 ml in MeOH/H2O, v-v = 1:1). The mixture was stirred at 70 °C for 24 h. After
cooling to room temperature, 10% HCl(aq.) was added to the resulting mixture to
attain pH < 2. A yellow precipitate formed was filtered, washed extensively with
distilled water and suction-dried on a filter paper to afford MA3 as a: brown solid
(455 mg, 97% yield based on M1); 1H NMR (300 MHz, DMSO-d6): δ = 7.55 (s, 6H),
7.52 (s, 3H), 7.16–7.26 (m, 18H), 7.32–7.35 (m, 12H), 4.35 (s, 12H) ppm; 13C NMR
(75 MHz, DMSO-d6): δ = 168.25, 140.59, 140.33, 138.13, 137.19, 133.64, 129.09,
128.45, 127.23, 125.52, 37.95 ppm.
Synthesis of 1,3,5-tris(3′,5′-bimercapto-4′-carboxyphenyl)benzene (H3L3).
Molecule MA3 (234 mg, 0.20 mmol), anhydrous AlCl3 (640 mg, 4.80 mmol), dry
DCM (17.0 ml) and dry toluene (2.0 ml) were mixed in a reaction tube in a N2-filled
glove box. After the reaction tube was taken out, the tube was connected to a N2
manifold and stirred at room temperature for 2 h. Afterwards, 10% HCl (aq., 10 ml,
bubbled with N2 for five min beforehand) was added to the mixture, followed by
stirring at room temperature for 2 h. The yellow precipitate formed was collected
by suction filtration and washed with 10% HCl (aq.), distilled water, then DCM
and dried under vacuum to afford H3L3 as a: brown solid (103 mg, 82% yield based
on MA3); 1H NMR (400 MHz, acetone-d6): δ = 8.01 (s, 3H), 7.78 (s, 6H) ppm; 1H
NMR (400 MHz, DMF-d7): δ = 8.11 (s, 3H), 7.96 (s, 6H) ppm; 13C NMR (100 MHz,
DMF-d7): δ = 169.14, 142.78, 141.27, 137.32, 129.61, 127.05, 126.85 ppm.
Polycrystalline framework solid of ZrL3. H3L3 (30 mg, 0.048 mmol) and 10 drops
of 1,2-ethanedithiol (170 mg, 1.81 mmol) were dissolved in DMF (3.0 ml), while
Nature Nanotechnology
ZrOCl2∙8H2O (92 mg, 0.29 mmol) and formic acid (≥ 88%, 490 mg, 9.37 mmol)
were ultrasonically dissolved in another 3.0 ml DMF. The two mixtures were added
to a 25 ml Pyrex glass ampoule. The ampoule was flame-sealed and heated in an
oven at 120 °C for 48 h, followed by programmed cooling to room temperature for
12 h. The pale brown powder formed was collected by filtration, then washed with
DMF (3 ml × 10) and acetone (3 ml × 10) and dried under nitrogen flow for weight
measurement (90 mg). The above powder was denoted as as-made ZrL3.
Preparation of activated ZrL3. In a nitrogen-filled glovebox, an as-made ZrL3
sample (~80 mg) and DMF (~8 ml) were added to a 15 ml glass vial and heated at
60 °C for 12 h. After cooling to room temperature, the supernatant was replaced
with fresh DMF (~8 ml) and heated at 60 °C for 12 h. The above solvent exchange
and heating procedures were repeated three times. After that, the above activation
procedure was repeated using CH3CN. The vacuum-dried powder was denoted as
activated ZrL3.
Elemental analysis found (C (26.70%), H (4.286%), N (0.47%)) and a
fitting formula can be determined to be Zr6O4(OH)4(L3)2(OH)6(H2O)27(DMF)
(molecular mass 2,596.45), which gives a calculated profile as (C (26.37%), H
(3.92%), N (0.54%), S (14.82%), Zr (21.08%)). The activated ZrL3 sample was
also characterized by TGA in both N2 and air flows (Supplementary Fig. 8, see
also Supplementary Fig. 9 for an informative comparison of TGA data). The TGA
measurement conducted in an air flow showed 25.1% ZrO2 residue, equivalent
to a Zr content of 18.6%. Both TGA and elemental analysis results agree with the
evacuated, dehydrated formula Zr6O8(C27H15O6S6)2O (molecular mass 1947).
Boiling-water-treatment of ZrL3. To a 100 ml round-bottom flask charged with a
stirring bar, an activated ZrL3 sample (120 mg) and deionized water (30 ml) were
added and the mixture was refluxed (using an oil bath at 120 °C) for 24 h. After
cooling to room temperature, the resultant MOF solid was collected by filtration
and dried under vacuum at 105 °C for 12 h. The above powder was denoted as
boiling-water-treated ZrL3.
Perovskite solar cell fabrication. The pre-patterned ITO (15 Ω sq−1) glass
substrates were sequentially cleaned by sonication with detergent (Decon 90/
deionized water, v/v = 1:1), deionized water, acetone and IPA for 15 min,
respectively. The cleaned ITO glass substrates were then transferred into an
oven at a temperature of 100 °C and treated with O2 plasma for 10 min before
use. Then, the hole-transporting layer, that is, 10 mg ml−1 PTAA in toluene
(doped with 1 wt% of 2,3,5,6-tetrafluoro-7,7,8,8-tetracyanoquinodimethane
(F4-TCNQ)), was spin-coated onto the ITO glass substrate at 4,000 rpm for 30 s
and this was subsequently annealed at 150 °C for 10 min. After cooling for 5 min,
the mixed-cation perovskite film was then deposited on PTAA by a one-step
spin-coating method. The perovskite precursor solution was prepared by mixing
CsI, FAI, MABr, PbI2 and PbBr2 in DMF/DMSO (v/v = 5:1) to a chemical formula
of Cs0.05(FA0.85MA0.15)0.95Pb(I0.85Br0.15)3.
Considering the poor wetting ability of the PTAA surface, 50 μl DMF was first
spun onto the PTAA layer to improve the wetting ability before depositing the
perovskite layer. Then, 60 μl of perovskite precursor solution was spin-coated on
PTAA at 1,000 rpm for 10 s and 5,000 rpm for 30 s. During the second step, 110 μl
CB as anti-solvent was quickly dripped onto the centre of the perovskite film 15 s
before the end of the spin-coating process and then annealed at 60 °C for 5 min and
100 °C for 30 min. The electron-transporting layer, that is PC61BM (20 mg ml−1 in
CB), was spun onto the perovskite layer with a speed of 2,000 rpm for 30 s, followed
by annealing at 90 °C for 10 min. As for the devices with a MOF EEL, ZrL3 solution
was first prepared by dispersing ZrL3 into IPA at different concentrations followed
by ultrasonication and stirring. Then, the ZrL3 (or ZrL3 doped with 10 wt% of
bis-C60) solution was spun onto the PC61BM layer at a speed of 2,000 rpm for 60 s.
For the reference devices, bis-C60 (2 mg ml−1 in IPA) was spun onto the PC61BM
with a speed of 2,000 rpm for 30 s. All the spin-coating processes were conducted in a
N2-filled glovebox with contents of O2 and H2O < 1 ppm at a controlled temperature
of 20–25 °C. Finally, 120 nm Ag was thermally evaporated in a high vacuum chamber
(<5 × 10−6 Torr) through a metal shadow mask (aperture area of 0.13 cm2 is defined
by the overlap between the pre-patterned ITO and the back metal electrode).
Characterizations and measurements. Elemental analysis was performed
with a Vario EL III CHN elemental analyser. Infrared spectra in the range
400–4,000 cm−1 were recorded on a Nicolet Avatar 360 FT-IR spectrophotometer.
Solution 1H NMR and 13C NMR spectra were recorded at room temperature on
Bruker superconducting-magnet high-field NMR spectrometers with working
frequencies of 300 and 400 MHz, using tetramethylsilane as the internal standard.
Chemical shifts (δ) are expressed in ppm relative to the residual solvent reference
(for example chloroform-d 1H: 7.26 ppm, 13C: 77.0 ppm). Coupling constants are
expressed in hertz. TGA was carried out in a nitrogen or air stream using Perkin
Elmer thermal analysis equipment (STA 6000) with a heating rate of 2 °C min−1,
with an empty Al2O3 crucible being used as the reference. The porosity and
surface area analyses were performed using a Quantachrome Autosorb iQ gas
sorption analyser. Samples were outgassed at 0.03 Torr with a 2 °C min−1 ramp to
120 °C and held at 120 °C for 12 h. The samples were then held in vacuum until
the analyses started to run. Pore analysis was performed using N2 at 77.35 K (P/P0
Nature Nanotechnology | www.nature.com/naturenanotechnology
Nature Nanotechnology
range of 2 × 10−7–0.995). PXRD data were collected in the reflection mode at room
temperature on a Philips X’Pert diffractometer equipped with a CPS 180 detector
using monochromated Cu-Kα (λ = 1.5418 Å) radiation. The X-ray tube operated
at a voltage of 30 kV and a current of 30 mA. The amounts of the metal ions were
determined using a PerkinElmer Optima™ 8000 ICP-OES instrument.
Scanning transmission electron microscopy images were obtained on a
double-aberration-corrected JEOL JEM-300CF instrument at 300 kV. UV-vis
absorption spectra were recorded on a Perkin Elmer UV-vis spectrometer
(model Lambda 2S). PL and TRPL spectra were recorded on an FLS980
spectrofluorometer (Edinburgh Instruments) with a Xe lamp and a pulsed
excitation laser, respectively. Ultraviolet photoelectron spectrum studies were
carried out on a VG ESCALAB 220i-XL surface analysis system equipped with a
He discharge lamp (hv = 21.22 eV). The morphology of the samples was monitored
by SEM (QUATTRO S, Thermal Fisher Scientific). Atomic force microscopy was
carried out using a Nanoscope IIIa instrument (Veeco). Cyclic voltammetry was
measured on a CH Instruments CHI760E electrochemical analyser/workstation
using a conventional three-electrode cell with an ITO strip (5 mm × 35 mm) coated
with ZrL3 as the working electrode, a Pt wire counter electrode and a Ag/AgCl
reference electrode at a scan rate of 50 mV s−1 using 0.1 M tetrabutylammonium
hexafluorophosphate in dry DCM as the electrolyte and referenced to the FeCp2/
FeCp2+ (FeCp2 = Fe(C5H5)2) redox couple (−5.10 eV below vacuum level)40.
The working electrode was prepared using a similar method as reported in the
literature41 by casting a dispersion of ZrL3 in isopropanol onto the ITO glass.
ToF-SIMS measurement was perfomed using a TOF-SIMS V instrument (ION‐
TOF GmbH, Cameca IMS 4F), where a 3 keV Cs+ ion beam was used for erosion
and a 25 keV Bi+ pulsed primary ion beam was used for the analysis. The area of
analysis was 48 × 48 µm2 while the sputtering area was 300 × 300 µm2.
The current density–voltage (J–V) characteristics of the photovoltaic devices
were measured in a N2-filled glovebox at room temperature using a Keithley 2400
SourceMeter under simulated sunlight from a solar simulator (EnliTech, SS-F5). A
National Renewable Energy Laboratory calibrated silicon solar cell (with a KG-2
filter) was used to obtain the AM 1.5G (100 mW cm−2) light intensity from the solar
simulator. The PVSCs were covered with a shading mask with an aperture area of
0.104 cm2 to ensure the accuracy of the current density obtained from the J–V curves.
The J–V measurements were carried out in sweep mode with reverse (from 1.23 V to
−0.02 V) and forward (from −0.02 V to 1.23 V) scans with a scan rate of 10 mV s−1.
EQE was carried out using an EQE measurement system (EnliTech, QE-R).
Reporting Summary. Further information on research design is available in the
Nature Research Reporting Summary linked to this article.
Data availability
The data that support the plots within this paper and other findings of this study
are available from the corresponding author upon reasonable request.
Nature Nanotechnology | www.nature.com/naturenanotechnology
Articles
References
40. Cardona, C. M., Li, W., Kaifer, A. E., Stockdale, D. & Bazan, G. C.
Electrochemical considerations for determining absolute frontier orbital
energy levels of conjugated polymers for solar cell applications. Adv. Mater.
23, 2367–2371 (2011).
41. Wang, T. C. et al. Rendering high surface area, mesoporous metal-organic
frameworks electronically conductive. ACS Appl. Mater. Interfaces 9,
12584–12591 (2017).
Acknowledgements
The work was supported by the APRC Grant of the City University of Hong Kong
(9380086, 9610421), an ECS grant from the Hong Kong Research Grants Council
(21301319) and Innovation and Technology Support Programme (ITS/497/18FP,
GHP/021/18SZ), the Guangdong Major Project of Basic and Applied Basic Research
(no. 2019B030302007) and Guangdong-Hong Kong-Macao Joint Laboratory of
Optoelectronic and Magnetic Functional Materials (no. 2019B121205002). TEM work
was conducted using the facilities in the Irvine Materials Research Institute (IMRI) at the
University of California-Irvine and supported by the NSF under grants (CBET-1159240
and DMR-1506535). This work was also supported by an ARG grant (Project 9667168)
from the City University of Hong Kong.
Author contributions
Z.Z. and A.K.-Y.J. conceived the ideas and designed the project. Z.Z., A.K.-Y.J. and Z.X.
directed and supervised the research. S.W. fabricated and characterized devices. Z.L.,
T.L. and J.Z. also contributed to device fabrication, characterizations and helped S.W.
analyse the data. Z.X., M.-Q.L. and Y.D. designed and synthesized the materials. P.T. and
W.G. performed the TEM measurement and P.T., W.G. and X.P. analysed the TEM data.
F.L. conducted cyclic voltammetry measurements and analysed the data. F.L. and F.Q.
designed the NMR characterization of the digested MOF sample and analysed the data.
S.W., Z. Z. and A.K-Y.J. drafted and finalized the manuscript. All the authors revised the
manuscript.
Competing interests
Authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/
s41565-020-0765-7.
Correspondence and requests for materials should be addressed to Z.X., Z.Z. or
A.K.-Y.J.
Reprints and permissions information is available at www.nature.com/reprints.
 nature research | solar cells reporting summary
Corresponding author(s): Zhengtao Xu, Zonglong Zhu and Alex K.-Y. Jen

Solar Cells Reporting Summary

Nature Research wishes to improve the reproducibility of the work that we publish. This form is intended for publication with all accepted papers
reporting the characterization of photovoltaic devices and provides structure for consistency and transparency in reporting. Some list items might
not apply to an individual manuscript, but all fields must be completed for clarity.
For further information on Nature Research policies, including our data availability policy, see Authors & Referees.

` Experimental design
Please check: are the following details reported in the manuscript?
1. Dimensions
Yes The area of the tested solar cells is 0.13 cm2 (see "Methods" section in main text).
Area of the tested solar cells
No
Yes The device area is defined by the overlap between the pre-patterned ITO and back
Method used to determine the device area metal electrode, where the metal electrode is thermally evaporated.
No
2. Current-voltage characterization
Current density-voltage (J-V) plots in both forward Yes Supplementary Figs. 22, 25 and 27.
and backward direction No

Voltage scan conditions Yes The J-V measurements were conducted with sweep mode with reverse (from 1.23 V
For instance: scan direction, speed, dwell times No to -0.02 V) and forward (from -0.02 V to 1.23 V) scan with a scan rate of 10 mV s-1
(see "Methods" section in main text).

Test environment Yes All J-V measurements were conducted in N2-filled glovebox at room temperature
For instance: characterization temperature, in air or in glove box No (see "Methods" section in main text).

Protocol for preconditioning of the device before its Yes No preconditioning protocol was used.
characterization No

Stability of the J-V characteristic Yes The steady-outputs were measured with time evolution of the maximum power point
Verified with time evolution of the maximum power point or with No in Figs. 2E and 3B.
the photocurrent at maximum power point; see ref. 7 for details.

3. Hysteresis or any other unusual behaviour

Description of the unusual behaviour observed during Yes No unusual behaviour was observed during the characterization.
the characterization No
Yes Supplementary Figs. 22, 25 and 27.
Related experimental data
No

4. Efficiency
External quantum efficiency (EQE) or incident Yes Fig. 2D.
photons to current efficiency (IPCE) No

A comparison between the integrated response under Yes The Jsc obtained from J-V measurement is consistent with the Jsc integrated from
the standard reference spectrum and the response No EQE spectrum. (Fig. 2C, 2D).
measure under the simulator

For tandem solar cells, the bias illumination and bias Yes Not relevant.
November 2017

voltage used for each subcell No

5. Calibration
Light source and reference cell or sensor used for the Yes Simulated AM 1.5 G irradiation was from solar simulator (EnliTech, SS-F5, Taiwan). A
characterization No silicon solar cell (with a KG-2 filter) was used as reference cell (see "Methods" section
in main text).

1
 Confirmation that the reference cell was calibrated Yes The reference cell was calibrated by National Renewable Energy Laboratory (more
and certified No see "Methods" section in main text).

nature research | solar cells reporting summary

Calculation of spectral mismatch between the Yes The test light source for the device measurement matches well with the one-sun light
reference cell and the devices under test No spectrum and reference cell. The spectral mismatch factor is 1 for the devices

6. Mask/aperture
Yes The size of the mask used for J-V measurements is 0.104 cm2 (See "Methods" section
Size of the mask/aperture used during testing in main text).
No

Variation of the measured short-circuit current Yes All devices has been measured with identical masks
density with the mask/aperture area No

7. Performance certification
Identity of the independent certification laboratory Yes Certified by the National Institute of Metrology (NIM, China).
that confirmed the photovoltaic performance No

A copy of any certificate(s) Yes Certificate is provided in Supplementary Fig. 26.

Provide in Supplementary Information No

8. Statistics
Yes 30 solar cells for each device configurations were tested and statistic for efficiency
Number of solar cells tested distribution demonstration (Fig. 2F and Supplementary Fig. 28).
No
Yes Fig. 2F and Supplementary Fig. 28.
Statistical analysis of the device performance
No

9. Long-term stability analysis

Type of analysis, bias conditions and environmental Yes Long-term stability for the cells was measured and analyzed at different conditions
conditions No including shelf stability in humid air and MPP tracking at elevated temperature in
For instance: illumination type, temperature, atmosphere N2-filled glovebox (detailed conditions are described in main text).
humidity, encapsulation method, preconditioning temperature

November 2017