ARTICLE

Received 29 Jun 2015 | Accepted 13 Nov 2015 | Published 11 Jan 2016 DOI: 10.1038/ncomms10214 OPEN

High-efficiency robust perovskite solar cells on

ultrathin flexible substrates
Yaowen Li1,2,*, Lei Meng1,*, Yang (Michael) Yang1, Guiying Xu2, Ziruo Hong1, Qi Chen1, Jingbi You1, Gang Li1,
Yang Yang1 & Yongfang Li2

Wide applications of personal consumer electronics have triggered tremendous need for
portable power sources featuring light-weight and mechanical flexibility. Perovskite solar
cells offer a compelling combination of low-cost and high device performance. Here we
demonstrate high-performance planar heterojunction perovskite solar cells constructed on
highly flexible and ultrathin silver-mesh/conducting polymer substrates. The device perfor-
mance is comparable to that of their counterparts on rigid glass/indium tin oxide substrates,
reaching a power conversion efficiency of 14.0%, while the specific power (the ratio of power
to device weight) reaches 1.96 kW kg  1, given the fact that the device is constructed on a
57-mm-thick polyethylene terephthalate based substrate. The flexible device also demon-
strates excellent robustness against mechanical deformation, retaining 495% of its original
efficiency after 5,000 times fully bending. Our results confirmed that perovskite thin films are
fully compatible with our flexible substrates, and are thus promising for future applications in
flexible and bendable solar cells.

1 Department of Materials Science and Engineering, University of California, Los Angeles, California 90095, USA. 2 Laboratory of Advanced Optoelectronic

Materials, College of Chemistry, Chemical Engineering and Materials Science, Soochow University, Suzhou 215123, China. * These authors contributed equally
to this work. Correspondence and requests for materials should be addressed to Y. W. L. (email: [email protected]) or to Z.H. (email: [email protected]) or to
Y.Y. (email: [email protected]).

NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214

S
olution processing perovskite-based photovoltaic cells have flexible pero-SCs. Therefore, for the flexible pero-SCs, the flexible
reached a power conversion efficiency (PCE) of 20% (ref. 1). electrodes and substrates should have sufficient thermal stability,
The rapid progress in both device efficiency2–8 and high electrical conductivity and optical transmission, bending
stability9–11 indicates the potential application of perovskite durability, facilitating growth of high-quality perovskite
materials in next generation solar cells. There are two device crystalline films.
architectures currently dominating in the field of perovskite solar Recently, a flexible PET substrate with embedded Ag-mesh
cells (pero-SCs): mesoporous type and planar heterojunction. (FEAM) was successfully fabricated via nano-imprinting
Solar cells using a TiO2 mesoporous layer as a scaffold still lead in technique, and used as electrodes in flexible optoelectronic
terms of efficiency. In addition to using a thick perovskite layer devices27–29. In combination with high-conductivity transparent
for stronger light absorption, obtaining high-quality TiO2 conducting polymer (Clevios PH1000), the hybrid electrode
nanocrystals and forming high-quality interfacial contacts with with high lateral conductivity can be readily fabricated
the perovskite layer greatly improves device performance by for wide application in thin film devices. Our previous work
reducing interface recombination. The TiO2 scaffold layer usually shows that this hybrid electrode substrate with thickness of
requires high temperature annealing at 450 °C that is commonly 200 mm exhibits a sheet resistance as low as 3 ohm sq  1 (ref. 28)
used in pero-SCs12,13. The promising advantages of this hybrid electrode such as
As an alternative to the mesoporous pero-SCs, the planar high flexibility and robustness, low sheet resistance and high
heterojunction design, which features a straightforward fabrica- transmission (78%) strongly suggest its future applications in
tion procedure, is also becoming attractive5,14,15. In this device flexible solar cells.
architecture, organic semiconductor hole- and electron- In this work, we integrate FEAM/PH1000 hybrid electrode
transporting layers, which have also been widely investigated, into perovskite-based photovoltaic cells, and demonstrate an
are proper candidates for high-throughput low temperature ultrathin flexible device delivering a PCE of 14.0%. This value is
manufacturing (o150 °C), opening the possibility of the highest among those pero-SCs based on TCO-free pero-SCs,
constructing flexible photovoltaic cells on various polymer and more importantly the device exhibits superior durability
substrates. Most of the flexible pero-SCs have been reported against mechanical bending over 5,000 times with 5-mm
with PCEs in the range of 6–10% (refs 16–19). Jung and curvature and extremely high bending stability through rational
colleagues have reported a perovskite device on polyethylene device design. To fabricate the FEAM, we adopt low coverage,
naphthalate (PEN)/indium-tin-oxide (ITO) based flexible hexagonal Ag-mesh embedded in the ultrathin (B57 mm)
substrates, giving a 12.2% PCE. 50% loss in initial PCE was ultraviolet-resin-coated PET substrate that ensures high optical
observed after 1,000 fully bending cycles20. More recently, Seok transmission. The bottom of the ultrathin FEAM substrate
and colleagues reported a flexible pero-SCs with a PCE of is laminated on a 100-mm-thick highly hardened PET
14.85% based on the similar substrate21. It is catching up with protection film for thermal processing. It can be readily peeled
the state-of-the-art for flexible thin film solar cells, such as off after completing the device, giving the ultrathin flexible
CIGS, CdTe and so on22,23. pero-SCs.
Still most efficiency values of flexible devices lag behind those
of the references on rigid substrates, and the bending durability of
PEN/ITO is subpar. The major deficiencies arise from the Results
relatively low conductivity and poor mechanical robustness of the The structure and properties of FEAM-based hybrid electrode.
transparent electrode on polymer substrates. Transparent Ultrathin FEAMs substrate can be readily scaled up via roll-
conductive oxide (TCOs) electrodes, such as ITO and Al-doped to-roll nano-imprinting, which makes testing the feasibility of
ZnO usually have sheet resistance B10 ohm sq  1, increasing this type of substrate for applications in pero-SCs the main
series resistance of the photovoltaic cells. Besides, TCOs are motivation of this work. Figure 1a shows a 15-inch FEAM sheet
normally fragile against repeated bending, which has been exhibiting excellent flexibility and high transmission in the
recognized as the origin of device degradation/failure20,24. visible region. The substrate structure is shown in Fig. 1b. The
Going beyond flexible solar cells, wearable electronics require substrate consists of periodically hexagonal Ag-mesh (diagonal
high durability against bending. It imposes further challenges to length of 180 mm, width of 3 mm and height of 2 mm) embedded
build up highly flexible pero-SCs. Until recently, the reported in ultraviolet-resin coated on PET, giving a substrate with total
mechanical bending test of pero-SCs ended up with the failure of thickness of 57 mm. It is noteworthy that the coverage rate of
TCO electrodes, and whether perovskite thin films can endure Ag-mesh is 3.2%, which causes very low optical loss of o4%
multiple-times bending still remains unknown. Consequently, according to Fig. 1d. As the Ag-mesh period is several orders of
TCO-free electrodes or even indium-free electrodes are desirable magnitude larger than the visible wavelength, there is no
owing to the poor mechanical robustness of poly-crystalline metal significant light diffraction and scattering29. The hybrid
oxide films and the scarcity of indium and so on. Thin film electrode substrate PET/Ag-mesh/PH1000 was formed by
conductors based on carbon materials25 and metal nanowires26 spin-coating PH1000 onto PET/Ag-mesh substrate, then
have been successfully demonstrated in flexible pero-SCs. annealing at 120 °C for 20 min (Fig. 1c). The 150-nm-thick
However, the unsatisfying PCEs still limit their further PH1000 layer shows smooth and continuous coverage on the
applications, which is possibly ascribed to the imperfect PET surface with a root mean square (RMS) roughness of
perovskite film or poor contact between the neighbouring layers 2.0 nm (Supplementary Fig. 1a). According to Fig. 1d, the
including the electrode, interfacial layer and perovskite film. The resulting hybrid PET/Ag-mesh/PH1000 electrode shows a
substrate of the flexible electrode is particularly important as well. transmission of 82–86% in the visible region, with an excellent
Although the planar heterojunction pero-SC can effectively lower sheet resistance of B3 ohm sq  1, almost one-third lower than
the annealing temperature below 150 °C, it is still higher than the that of commercial ITO-glass substrates. It should be noted here
glass-transition temperature of typical flexible substrates such as that the groove is not completely filled with Ag nanoparticles
polyethylene terephthalate (PET) and PEN. As a result, it is during the fabrication, resulting in an B500-nm-height
desirable to develop flexible substrates or technology which are difference between the Ag-mesh and the substrate surface.
able to stand an annealing temperature of 150 °C. This is also However, PH1000 still forms sufficient contact with the
critical for the roll-to-roll manufacturing of high-performance Ag-mesh as discussed later.

2 NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214 ARTICLE

a d 100

90 3.9% loss

Transmittance (%)
80 PET
PET/Ag-mesh
70 PET/Ag-mesh/PH1000

60

50

40
300 400 500 600 700 800 900
b Embedded Ag-mesh substrate Wavelength (nm)

c Hybrid electrode (Ag-mesh/PH1000)

m
0μ
18 )
(7 μm
esin
UV-r PH1000
PET
57 μm

2 μm
3 μm
Ag nanoparticle

Figure 1 | Schematic illustration of the FEAMs substrate and hybrid electrode. (a) An image of the large-area FEAMs substrate. (b) The structure of the
FEAMs substrate with detail parameters. (c) The diagram for the hybrid electrode (PET/Ag-mesh/PH1000). (d) Transmission spectra of bare PET, PET/
Ag-mesh, PET/Ag-mesh/PH1000-based substrates.

FEAM substrate-based flexible pero-SCs design. To build (5.30 eV) in MAPbI3 are in good alignment with the LUMO level
up flexible pero-SCs, we adopted a conventional p-i-n of PCBM (3.9 eV) and the work function of PEDOT:PSS
architecture in which the hybrid electrode is used as an (5.15 eV), respectively. Recent reports have alluded that
anode, that is, PET/Ag-mesh/PH1000/poly(3,4-ethylenediox- minimizing the mismatch between the work functions of
ythiophene):poly(styrenesulfonate) (PEDOT:PSS) (35 nm)/ electrode and transporting layer can enhance charge transport
CH3NH3PbI3 (MAPbI3) (B280 nm)/phenyl-C61-butyric acid through the transporting layer to its corresponding electrode35.
methyl ester (PCBM) (B60 nm)/Al (100 nm) as shown in Fig. 2a. Accordingly, we utilized scanning Kelvin probe microscopy to
This device structure offers the benefit of efficient charge transfer determine the surface potential (SP) difference of the PH1000
from the perovskite thin film to the charge transport layers, electrode before and after coating PEDOT:PSS. As shown in
ensuring sufficient charge extraction30–32. The B57-mm-thick Supplementary Fig. 1b and f, in the flat surface, both the scanning
hybrid flexible electrode (PET/Ag-mesh/PH1000) and all organic Kelvin probe microscopy images of the PH1000 electrode before
interlayer materials (PEDOT:PSS and PCBM) ensure the and after coating PEDOT:PSS are rather homogeneous. A slightly
ultrathin and flexible properties of pero-SCs. Notably, PH1000 increased SP of 80 mV was obtained after coating PEDOT:PSS
and Ag-mesh in PET/Ag-mesh/PH1000 are used as electrode and film, giving corresponding work function of 5.09 and 5.17 eV,
transporting channel, respectively. PH1000 (B0.9 S cm  1) can respectively, which are rectified by the highly ordered pyrolytic
efficiently collect charges along both the lateral and vertical graphite as standard sample36. Considering the similar work
direction in less than 100 mm (ref. 33). The charges were then function and chemical compositions in PEDOT:PSS and PH1000,
further gathered by the highly conductive Ag-mesh the contact should be Ohmic contact. These results indicate that
(4104 S cm  1). As the periodic Ag-mesh honeycomb has a there is no barrier for hole conduction from PEDOT:PSS to
diagonal of 180 mm, the charge–transport pathway of PH1000 PH1000, which can efficiently avoid hole accumulation and
along the lateral direction is within 90 mm to ensure efficient facilitate greater transportation from PEDOT:PSS to PH1000.
charge extraction to the Ag mesh, and this is consistent with the Moreover, in the Ag-mesh-covered area, the SP is not disturbed,
low sheet resistance of the hybrid electrode. The perovskite films and this further confirms the formation of continuous and
were fabricated by a standard two-step method (see Methods for homogeneous PEDOT:PSS films throughout the substrate
details)34. surface.
Cross-section scanning electron microscopy (SEM) images of
an optimized device are presented in Fig. 2b. The MAPbI3 layer,
with its distinct feature and contrast, can be well distinguished Performance of flexible pero-SCs. Figure 3a shows the current
from the hole- and electron-transporting layers. The well-defined density–voltage (J–V) characteristics of an inverted, planar,
and pinhole-free perovskite film was observed from both the flexible PET/Ag-mesh/PH1000/PEDOT:PSS/perovskite/PCBM/
Ag-mesh area and the flat PET area, effectively depressing the Al solar cell. The champion flexible device achieved a power
shunting pathways. It is noteworthy to mention that the conversion efficiency of 13.7% with a short-circuit current (JSC) of
perovskite poly-crystalline structure can be grown in the 19.3 mA cm  2, open-circuit voltage (VOC) of 0.90 V, fill factor
Ag-mesh groove. Even though the depth of the groove is (FF) of 0.79 in the forward scan direction, and a PCE of 14.2%
B400 nm after filling PH1000, the film quality remains. As with a JSC of 19.5 mA cm  2, VOC of 0.91 V and FF of 0.80 in the
shown in Fig. 2c–e, the top-view SEM images of the perovskite reverse scan direction under 100 mW cm  2 AM 1.5G illumina-
film further identified homogeneous MAPbI3 crystals in both the tion. The flexible cell exhibits competitive efficiency as the control
Ag-mesh and the flat PET area, confirming a high-quality film. devices on rigid glass/ITO substrate as shown in Supplementary
Figure 2f shows the energy band diagram of the pero-SC. The Fig. 2, indicating the enormous potential of our hybrid electrode
conduction band (CB) (3.75 eV) and the valence band (VB) in pero-SC. The slightly lower JSC should be attributed to the

NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214

a b f
3.75 eV
3.9 eV
– Flat PET area
AI AI PCBM 4.3 eV 4.3 eV
e e P C BM e MAPbl3
e e e MAPbl3 Ag AI
e
h h MAPbl3 h 5.08 eV 5.15 eV PCBM
h
h PH1000
Ag PEDOT:PSS
0 5.3 eV
+ P H 10 0 Ag
U V - r es i n
Ag-mesh area
PET 5.9 eV
c d e
n fi
Peelable
P r o t e ctio lm

Figure 2 | Flexible pero-SCs architecture and morphology. (a) Device architecture of the hybrid electrode/PEDOT:PSS (35 nm)/MAPbI3 (B280 nm)/
PCBM (B60 nm)/Al (100 nm) cells tested in this study. (b) Cross-section SEM images of complete perovskite devices showed both Ag-mesh area (scale
bar, 1 mm) and Flat PET area (scale bar, 500 nm). SEM top-view images of perovskite films; (d) : low-resolution image of perovskite film coated on both flat
PET and Ag-mesh (scale bar, 2 mm); (c): film surface on bare PET (scale bar, 500 nm); (e): film surface on Ag-mesh (scale bar, 500 nm). (f) The
corresponding energy-level diagram of each layer.

a b –22
Current density

Current density (mA cm–2),

0 –20 PCE
Dark
Jsc (mA cm–2)

–5 Reverse scan PCE (%) –18 17.3 mA cm–2

Forward scan

–10 –16
14.0%
–14
–15
–12
–20
–10
0.0 0.2 0.4 0.6 0.8 1.0 0 20 40 60 80 100 120
Voltage (V) Time (s)

c 100 d 20

80
15
EQE (%)

60
Counts

10
40

5
20

0 0
300 400 500 600 700 800 11.5 12.0 12.5 13.0 13.5 14.0 14.5
Wavelength (nm) PCE (%)

Figure 3 | Photovoltaic performance characteristics. (a) J–V curves in reverse and forward scan measured under 100 mWcm  2 AM 1.5G illumination and
dark for the champion flexible PET/Ag-mesh/PH1000/PEDOT:PSS/MAPbI3/PCBM/Al solar cell: Inset shows photograph of corresponding ultra-thin
flexible pero-SCs. (b) Steady-state photocurrent output at the maximum power point (0.8 V). (c) External quantum efficiency (EQE) spectrum of champion
flexible pero-SC. (d) Histograms of device PCE measured for 50 devices of PET/Ag-mesh/PH1000/PEDOT:PSS/MAPbI3/PCBM/Al.

lower transmission of the PET/Ag-mesh/PH1000 electrode 14.0 mW cm  2, equal to 14.0% PCE. To the best of our
(82–86% in the visible region) compared with that of glass/ITO knowledge, this is the highest PCE observed in a TCO-free
electrode (82–91% in the visible region, Supplementary Fig. 3). flexible pero-SC. It is also noteworthy to mention that the flexible
The key photovoltaic parameters of the flexible and rigid planar pero-SC gives a high specific power (the ratio of power to device
perovskite solar cells are summarized in Supplementary Table 1. weight) reaching 1.96 kW kg  1 owing to the ultrathin device,
According to previous reports17,18, such p-i-n type devices which also holds promising application in light-weight
usually show weak hysteresis with respect to the scan direction photovoltaic purpose, such as unmanned aerial vehicles38.
and rate owing to their fast response to an incident light signal37. To accurately evaluate the hysteresis behaviours, the J–V curves
In the case of flexible cells, the devices show even less hysteresis of the optimizing devices based on PET/Ag-mesh/PH1000 and
than the rigid ones. To accurately evaluate the efficiency, we glass/ITO electrodes were measured at varied scan rates of
took a steady-state current output at a forward bias of 0.8 V, 2.4 V s  1, 200 mV s  1 and 20 mV s  1 from both scanning
which corresponds to the point of maximum power output, as directions as summarized in Supplementary Fig. 4 and
shown in Fig. 3b. The flexible cell yielded a stabilized current Supplementary Table 2. In the case of flexible pero-SC, we confirm
density of 17.4 mA cm  2, corresponding to a power output of that the flexible pero-SC has less hysteresis than the control device,

4 NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214 ARTICLE

with slight change in JSC in the range of 19.2–19.6 mA cm  2 and B68% in the long wavelength region of 630–700 nm, which
FF between 0.77 and 0.80. In contrast, the rigid pero-SCs showed should be owing to the incomplete absorption of light. The
obvious hysteresis. FF varied from 0.65 to 0.73 when the perovskite layer is relatively thin, and the absorption coefficient is
scanning rate is changed from 2.4 V s  1 to 200 mV s  1 in both also low in this range. As the reproducibility of device efficiency
scanning directions. Meanwhile, JSC changes significantly from has been a concern of the scientific community, we further show a
21.1 mA cm  2 at high scan rate to 20.0 mA cm  2 at low scan rate. histogram of the PCE of the flexible pero-SCs in Fig. 3d. Over 50
It indicates that it takes longer for the glass/ITO-based device to devices were fabricated and characterized, producing a 4.0%
reach steady state. To confirm that, we measured photocurrent relative standard deviation in PCE, and thus the PCEs and
response at the maximum power point using manually controlled reproducibility are also encouraging.
shutter (response time within 0.1 s) to turn on and off the incident To further understand the device operation in the flexible
light. As shown in Supplementary Fig. 4c and d, the photocurrent pero-SCs, light intensity dependence of J–V characteristics of the
of flexible device rose quickly to maximum steady-state photo- device was measured under illumination of a solar simulator with
current. However, photocurrent from the rigid cell needs more a set of neutral density filters. Figure 4a shows the linear
than 1.9 s to fully saturate. Considering the similar structure of relationship of the photocurrent with light intensity in a double
flexible and rigid devices, it is reasonable to attribute the less logarithmic scale for the device with a slope close to unity,
hysteresis of the flexible device to the high charge carrier extraction indicating no substantial space charge build-up and confirming
ability of hybrid electrode. the good electrical contact in the device39. Figure 4b gives VOC as
The external quantum efficiency spectra of the flexible pero-SC a function of light intensity. We can see that VOC increases
is shown in Fig. 3c. The integrated JSC is 17.9 mA cm  2, close to monotonically with light intensity for the flexible pero-SCs. The
the JSC (19.5 mA cm  2) obtained under AM 1.5G solar simulator. dependence of VOC on light intensity implies that trap-assisted
The external quantum efficiency exceeds 70% from 400 to Schockley–Read–Hall recombination plays a minor role even at
610 nm, with a peak value of 78% at 520 nm. However, it drops to low light intensity40.

a b 1.00
Current density (mA cm–2)

10 0.95
Voltage (V)

0.90
1
0.85

0.1 0.80
1 10 100 1 10 100
Light intensity (mW cm–2) Light intensity (mW cm–2)

c 12 d Glass/ITO/PEDOT:PSS/MAPbI3
Glass/ITO/PEDOT:PSS/MAPbI3
PL intensity (counts ×105)

PET/Ag-mesh/PH1000/PEDOT:PSS/MAPbI3
10 PET/Ag-mesh/PH1000/PEDOT:PSS/MAPbI3
1,000
PL intensity (a.u.)

6 100

2 10

0
700 770 840 –200 0 200 400 600 800 1,000
Wavelength (nm) Time (ns)

e f
1.0 1.0
Glass/ITO Glass/ITO
Normalized transient

Normalized transient

0.8 PET/Ag-mesh/PH1000 PET/Ag-mesh/PH1000

0.8

0.6 0.6
voltage

current

=8.9 μs
 =0.65 μs
0.4 0.4
=8.2 μs
0.2 0.2
=0.42 μs
0.0
0.0
0 2 4 6 8 10 12 14 0.0 0.5 1.0 1.5 2.0
Time (μs) Time (μs)

Figure 4 | Characterization of flexible pero-SCs. (a) JSC as a function of light intensity in a double log plot. (b) VOC as a function of light intensity in a
semi-log scale. (c) The steady-state PL spectra of perovskite film on various substrates, and (d) TRPL decay transient spectra of PET/Ag-mesh/PH1000/
PEDOT:PSS/MAPbI3 and glass/ITO/PEDOT:PSS/MAPbI3. Charge transient time and traces for optimized devices based on glass/ITO and PET/Ag-mesh/
PH1000 electrodes by (e) transient photovoltage measurement under 0.5 sun bias light and (f) transient photocurrent measurement.

NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214

We then looked into the charge dynamics by measuring Ag-mesh and perovskite and also balanced the growth of
steady-state photoluminescence (PL) and time-resolved photo- perovskite crystal on the substrate consisting of double
luminescence (TRPL) characterization. Both have been recog- components (flat PET and Ag-mesh groove). On the PET/
nized to be proper tools to understand the charge extraction in Ag-mesh/PH1000 substrates, we found that there was a strong
perovskite-based solar cells. Figure 4c shows the steady-state PL dependence of photovoltaic performance on thermal annealing as
spectra of MAPbI3 films on PET/Ag-mesh/PH1000/PEDOT:PSS shown in Fig. 5a and Supplementary Table 1. When the annealing
and glass/ITO/PEDOT:PSS. The PL intensity on PET/Ag-mesh/ time was extended from 10 s to 2 min at 130 °C, the PCEs
PH1000/PEDOT:PSS was reduced by one order of magnitude improved from 7.0% to 13.1%. With further thermal annealing up
compared with that on glass/ITO/PEDOT:PSS. It strongly to 8 min, the PCEs dramatically dropped to 6.8%. We then
indicates that this PET/Ag-mesh/PH1000 hybrid electrode coated studied the relationship between photovoltaic performance and
with PEDOT:PSS hole-transporting layer can extract charge thermal annealing time, and the composition/morphology
carrier more efficiently, which is consistent with the less evolution of two-step solution-processed MAPbI3 crystals on
hysteresis of the flexible pero-SCs. The TRPL was measured by PET/Ag-mesh/PH1000 substrate was further investigated by
monitoring the peak emission at 768 nm as shown in Fig. 4d. using SEM and X-ray diffraction. Films with different annealing
Fitting the data with bi-exponential decay yields a fast decay times (10 s, 30 s, 1 min, 2 min, 4 min and 8 min) were chosen for
(t1) component and a slow decay (t2) component, and the the X-ray diffraction patterns and SEM images shown in Fig. 5.
detailed parameters are summarized in Supplementary Table 3. As shown in Fig. 5b, the perovskite film on the PET/Ag-mesh/
The fast decay would be originated from the quenching of charge PH1000 substrate, after quick annealing for 10 s, exhibited clear
carriers by the anode contact, and the slow decay component diffraction peaks at 14.1°, 28.5° and 31.9°, and were assigned to
could be attributed to the radiative recombination of free the (100), (200) and (210) planes, respectively, characteristic of a
charge carriers before the charge collection41. In the case of cubic perovskite structure42. With increasing annealing time, the
glass/ITO/PEDOT:PSS/MAPbI3, the fast decay lifetime was diffraction peaks remain, which indicates that the major part of
43.67 ns with a 61.3% ratio and the slow decay lifetime was the precursors was converted to MAPbI3 crystal even after short-
165.95 ns with a 38.7% ratio, suggesting that the depopulation time annealing at 130 °C. Nevertheless, residual PbI2 is present
of photogenerated charges was dominated by charge according to the X-ray diffraction pattern, which is favourable for
collection through the PEDOT:PSS/perovskite interface and pero-SC via surface passivation43. Moreover, as illustrated in
appropriate radiative recombination. For PET/Ag-mesh/ Supplementary Fig. 6a, the full width at half maximum of the
PH1000/PEDOT:PSS/MAPbI3, both t1 and t2 were shortened (100) peak reduce gradually from 0.203° to 0.178° when the
to 22.34 and 129.40 ns, respectively, and t1 dominated the PL thermal annealing time increased from 30 s to 8 min. It should be
decay, which can be interpreted that the hybrid PET/Ag-mesh/ noted that both increased crystallinity and a larger grain size with
PH1000 electrode induces a faster charge transfer from MAPbI3. less grain boundaries are evidenced by narrowed X-ray diffraction
Therefore, it is apparent that the majority of photogenerated peak, which is strongly related to the photovoltaic performance44.
carriers were collected by the electrode with minimal However, films annealed for 4 and 8 min exhibit higher
recombination loss. The fast charge transfer and efficient crystallinity, but dramatically decreased PCEs, compared with
collection shown in the devices based on PET/Ag-mesh/ optimized condition. It is probably caused by decomposition of
PH1000/PEDOT:PSS/MAPbI3 may be closely related to the less perovskite film and reaction between perovskite and Ag-mesh as
interfacial barrier and a more efficient charge carrier extraction as discussed below.
discussed above, thus contributing to an enhancement of SEM images (Fig. 5c–h) further reveal the morphology
photovoltaic properties, such as less hysteresis and high FF. evolution of the perovskite films upon annealing. The perovskite
We then further study the effects of the flexible hybrid film (Fig. 5c) was compact and conformal on top of both the PET
electrode on the charge carrier recombination and transport and the Ag-mesh after thermal annealing for 10 s. On the edge of
characteristics in complete devices by measuring the charge the Ag-mesh, we observed relatively high brightness contrast
carrier lifetime and charge carrier extraction time via transient compared with the adjacent area, possibly owing to less
photovoltage (TPV) and transient photocurrent (TPC) methods, conductivity and more accumulation of charges. Precursor
respectively. As shown in Fig. 4e, the carrier lifetime of flexible conversion is not complete, likely owing to the non-uniform
device was slightly increased to 8.9 ms, compared with 8.2 ms in thermal conductivity of the hybrid electrode, considering the
glass/ITO-based device, under 0.5 sun bias light at open-circuit higher thermal conductivity of Ag-mesh than those of the
condition, suggesting the similar perovskite film quality on both polymer and conducting polymer components. With the anneal-
flexible and rigid substrates. In Fig. 4f, the photocurrent decay ing time increased to 30 s, the appearance of pinholes was
time at short circuit condition decreased to 0.42 ms for the flexible observed. Again the change in film morphology, along with
device, compared with 0.65 ms for the rigid device. The high improved photovoltaic efficiency, (that is, dramatic JSC enhance-
charge carrier extraction capability of PET/Ag-mesh/PH1000 ment and reduction in series resistance) indicates that the
electrode reduced charge carrier extraction time by 1.5 times, thus conversion of precursors into perovskite is still favoured. As for
benefiting for the efficient charge carrier separation and the pinholes, we coated a PCBM layer to fully cover the
extraction. perovskite film surface and thus prevent shunting owing to
One concern regarding the Ag-mesh electrode is that Ag may direct contact between Al cathode and PEDOT:PSS layer41,45.
react with the halogen in the perovskite film, which is likely owing Increasing the annealing time to 1 and 2 min, the film
to the formation of a silver halide, especially when annealed in morphology was further improved, giving a continuous and
ambient air. When we deposited MAPbI3  xClx film via the one- homogeneous surface with larger grain size and a compact
step method, which requires thermal annealing for B1.5 h, it was perovskite film. The optimal morphology of the perovskite film
found that the long thermal annealing process can dramatically can facilitate the charge carrier transportation and collection, and
decrease the conductivity of the embedded Ag-mesh and the can further contribute to an improved PCE value. Further
corresponding device performance (Supplementary Fig. 5). We increasing the annealing time to 4 and 8 min, however, results in a
attribute this to the formation of non-conducting silver halide. large number of bright contrast disordered grains, especially
Therefore, we used the two-step deposition method to avoid long- around the Ag-mesh, strongly indicating decomposition of the
time thermal annealing, which results in the reaction between perovskite films owing to overheating. The photovoltaic cells

6 NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214 ARTICLE

a 4
Annealing 10 s
b * Perovskite
+ Substrate
Annealing 30 s Δ Pbl2 8 min
0
Annealing 1 min
*

Intensity (a.u.)
Annealing 2 min 4 min

Jsc (mA cm–2)

–4
Annealing 4 min *
Annealing 8 min Δ
* + + + 2 min
–8 *
1 min
–12
30 s
–16
10 s
–20
0.0 0.2 0.4 0.6 0.8 1.0 10 30 32 34 36 38 40 42 44
Voltage (V) 2 (degree)

Ag-mesh area Ag-mesh area Ag-mesh area Ag-mesh area Ag-mesh area Ag-mesh area

c d e f g h

Flat PET area Flat PET area Flat PET area Flat PET area Flat PET area Flat PET area

Figure 5 | Annealing-time evolution of the perovskite films. Annealing MAPbI3 grown on PET/Ag-mesh/PH1000/PEDOT:PSS substrate at 130 °C for
10 s, 30 s, 1 min, 2 min, 4 min and 8 min, respectively: (a) J–V curves of the corresponding solar cells. (b) X-ray diffraction patterns of the perovskite films.
(c–h) Top-view SEM images of the perovskite films coated on both flat PET and Ag-mesh area (scale bar, 2 mm): (c) 10 s; (d) 30 s; (e) 1 min; (f) 2 min;
(g) 4 min; (h) 8 min. Ag-mesh area row: the SEM images of corresponding perovskite films coated on Ag-mesh (scale bar, 500 nm); Flat PET area row: SEM
images of corresponding perovskite films coated on flat PET (scale bar, 500 nm).

based on these films yielded inferior PCEs below 10%. Compared great importance concerning practical applications, such as por-
with the optimized devices, all the three parameters determining table and wearable electronics. Bending tests, taking into account
PCE, that is, VOC, JSC and FF, decreased. It was further found that the bending radii and cycles, were carried out to evaluate the
long-time annealing at 130 °C will result in the decomposition of device stability against mechanical bending. The flexible pero-SCs
the perovskite films. were bent with five different radii of curvature (r ¼ N, 7, 5, 3.5
To understand the correlation between charge carrier extrac- and 2 mm) in one bending cycle as shown in Fig. 6a inset. The
tion dynamics and crystallographic properties of perovskite films, PCE showed almost no decrease in efficiency even with 2 mm of
we performed TPV and TPC measurement as summarized in bending radius, which is closing to the limit bending radius,
Supplementary Fig. 6 and Supplementary Table 1. According to retaining 98.1% of the original PCE value (Fig. 6a). This result
the TPV data, extending the annealing time from 1 to 2 min indicates that both the PET/Ag-mesh/PH1000 electrode and
increased carrier lifetime from 5.3 to 8.2 ms, which can be ascribed perovskite film possess high flexibility. We further evaluated the
to the fewer defects in the perovskite film consistent with the effects of mechanical bending on PET/Ag-mesh/PH1000
increased crystallinity of perovskite films and larger grain size. electrode via multiple-cycles bending test. A reference pero-SC on
With further annealing up to 4 and 8 min, the carrier lifetime PET/ITO electrode was also fabricated for comparison. A total of
dramatically decreased to 3.4 and 1.7 ms, respectively, which, as 5,000 consecutive bending cycles at radius of 5 mm were
we discussed in previous section, may result from perovskite conducted. As shown in Fig. 6b, the flexible pero-SCs based on
decomposition. From TPC measurement, the 2 min annealed PET/Ag-mesh/PH1000 electrode display very promising
device has the fastest charge extraction, which is consistent with mechanical bending stability. The PCE sustained even after
continuous and homogeneous film surface, large grain size, high 1,500-cycle bending. We then further went up to 5,000 cycles, the
crystallinity and beneficial to the device performance, especially device still retaining 95.4% of its initial PCE, accompanying the
FF and JSC. However, after annealing up to 4 and 8 min, charge well-maintained key photovoltaic parameters such as JSC, VOC
carrier extraction times are significantly increased. It could be and FF (Supplementary Fig. 7). Such high bending durability is
caused by more defects in the perovskite film, resulting from the very competitive to the state-of-the-art in flexible solar cell field,
decomposition of the perovskite films and reaction between while still holding the advantage of high efficiency46. The
Ag-mesh and perovskite because of overheating. Therefore, the reference on PET/ITO substrate showed obviously deteriorated
high device performance is likely owing to the high perovskite trend of device performance after 100 bending cycles, consistent
crystallinity with less defects and well protected Ag-mesh. with previous report20. PCE decreased to 30% of its initial value
after 1,000 bending cycles. As reported by Jung and colleagues20,
the sheet resistance of PET/ITO electrode abruptly increased
Bending and thermal stability test of flexible pero-SCs. From owing to the fracture in ITO polycrystalline structure, which
the viewpoint of the utilization of the flexible pero-SCs, the results in the dramatic deterioration in device performance.
mechanical flexibility and durability under bending stress are of Therefore, our results demonstrated that the high flexibility and

NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214

a 16 structure with the same full width at half maximum after

14
bending 2,000 and 5,000 cycles, indicating that the crystallinity
does not degrade upon bending. Meanwhile, 2,000 bending
12 cycles does not cause any observable changes in film
10 morphology (Supplementary Fig. 8b). Similarly, the perovskite
PCE (%)

8 film after 5,000 bending cycles shows no cracks from both top
6 r =∞ r =7 mm r =5 mm
and cross-section view as shown in Supplementary Fig. 8c and d.
However, after 5,000 bending cycles, as shown in higher-
4
resolution SEM image (Supplementary Fig. 8c), some pinholes
2
r =3.5 mm r =2 mm with diameter of 20–50 nm were observed close to the grain
0 boundaries. The developed pinholes could be responsible for the
∞ 7 5 3.5 2 efficiency loss. Steady-state PL and TRPL measurement of the
Curvature radius (mm) bent-state perovskite film were used to further investigate the
b 16 issue, and the data are shown in Supplementary Fig. 9 and
14
Supplementary Table 4. The PL intensities of perovskite film
before and after bending are almost identical, indicating that the
12
electronic coupling between layers in device is not affected by
10 the bending process, and the flexible electrode coated with
PCE (%)

8
PET/Ag-mesh/PH1000
PEDOT:PSS hole-transporting layer can still extract charge
6 PET/ITO carrier efficiently even after 5,000 bending cycles. The fast
component of PL decay, t1, were also similar according to TRPL
4
data. However, with increasing bending cycles to 5,000, the slow
2 component, t2, slightly decreases, and the ratio of t2 in the
0 overall decay process increases. As t2 comes from the delayed
0.1 1 10 100 1,000 10,000 recombination of trapped charges in perovskite bulk film, the
Bending cycles
pinholes originating from the mechanical bending could be the
c 16 reason of slightly increased trap density and thus the decrease in
14 photovoltaic efficiency.
Device thermal stability is always another major concern with
12
regard to the potential of hybrid electrodes. In particular, solar
10 cells normally operate at elevated temperatures. Therefore we
PCE (%)

8 PET/Ag-mesh/PH1000 recorded the device efficiency as a function of time at various

Glass/ITO
6 temperatures. The flexible pero-SC stored in inert atmosphere is
relatively stable at room temperature as illustrated in Fig. 6c. PCE
4
of 12.2% retains 91.6% of initial value after 500 h storage, which is
2 slightly better than that of the rigid control device (90.5% of
0 initial PCE). At 45 and 70 °C, a similar trend holds as shown in
0 100 200 300 400 500 Supplementary Fig. 10. The efficiency of flexible device decreased
Time (h) by B25% (PCE of 9.60%) after 500 h at 45 °C, while the rigid
Figure 6 | Bending and stability test of flexible pero-SCs. (a) PCEs device lost B31% (PCE of 9.45%). At 70 °C, both flexible and
measured after bending PET/Ag-mesh/PH1000 electrode-based flexible rigid pero-SCs deteriorated quickly and showed photovoltaic
pero-SCs within a specified radius of N, 7, 5, 3.5 and 2 mm. The inset characteristics with B23 and 15% of their initial values,
shows the real images of the corresponding bending radii, respectively. respectively, after 97 h. It could be attributed to the intrinsic
(b) PCEs of flexible pero-SCs based on both PET/Ag-mesh/PH1000 and issues of this device structure, such as the acidic PEDOT:PSS,
PET/ITO electrodes as a function of bending cycles at a radius of 5 mm. crystallization of PCBM, as well as the thermal stability of
(c) Stability of pero-SCs based on both PET/Ag-mesh/PH1000 and perovskite film47. According to our results, the flexible electrode
glass/ITO substrates under room temperature in N2-filled glove box in a does not induce noticeable device degradation at the initial stage
timescale of a few hundred hours. PCE values are obtained from statistical of 500 h of stability test. Therefore, the PET/Ag-mesh/PH1000
distribution of six devices for each condition. The ball symbols represent the hybrid electrode is a promising candidate for large-scale
mean, while the line across the ball represents the distribution. manufacturing of pero-SCs.

durability of pero-SCs on flexible hybrid substrates are ascribed to

the combination of embedded Ag-mesh charge transport channel Discussion
in a PET substrate and high conductivity polymer electrode. The In summary, ultrathin flexible pero-SCs with planar heterojunc-
former reserved the characteristic of high mechanical flexibility tion architecture have been successfully constructed on
and durability of PET substrate owing to the low coverage and 50-mm-thick PET substrate-based Ag-mesh/PH1000 hybrid
embedded Ag-mesh. The latter utilized the high flexibility nature electrode. A currently highest PCE (14.0%) of TCO-free flexible
of the polymer. More importantly, we proved that the inorganic– pero-SCs has been reported. The keys to achieve highly efficient
organic halide perovskites possessed the high flexibility and flexible pero-SCs are the high transmission, low sheet resistance,
promising durability, which can further extend its state-of-the-art excellent thermal stability of this hybrid electrode, as well as the
flexible electronics expectation. well-designed device architecture with aligned work functions of
The mechanical deformation caused B5% loss in PCE PH1000/PEDOT:PSS efficiently extracting charge carriers, and a
after 5,000 bending cycles, stimulating us toward more insight perfect perovskite film. The ultrathin flexible pero-SCs exhibit
into bending durability of pervoskite film. As illustrated in excellent bending durability retaining over 95.4% of its initial
Supplementary Fig. 8e, the perovskite film retains the cubic PCE value even after 5,000 fully bending cycles. This study opens

8 NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214 ARTICLE

an effective protocol for fabricating efficient, high specific power 13. Wang, J. T.-W. et al. Low-temperature processed electron collection layers of
and bending durable flexible pero-SCs. graphene/TiO2 nanocomposites in thin film perovskite solar cells. Nano Lett.
14, 724–730 (2014).
14. Liu, D. et al. Perovskite solar cells with a planar heterojunction structure
Methods prepared using room-temperature solution processing techniques. Nat. Photon.
FEAM substrate fabrication. The fabrication process of FEAM is described as 8, 133–138 (2014).
follows. First, the bottom of PET (thickness of 50 mm) substrate was covered by a 15. Chen, Q. et al. Planar heterojunction perovskite solar cells via vapor-assisted
highly hardening treatment PET protection film (thickness of 100 mm), which can solution process. J. Am. Chem. Soc. 136, 622–625 (2014).
well prevent distortion of the PET substrate during device fabrication process and
16. Kumar, M. H. et al. Flexible, low-temperature, solution processed
be easily peeled from the FEAMs-based device. Second, we design and fabricate a
ZnO-based perovskite solid state solar cells. Chem. Commun. 49, 11089–11091
protruding hexagon-patterned nickel (Ni) mould (diagonal length of 90 mm, width
(2013).
of 3 mm and height of 2 mm) through electroforming. Then the Ni mould with an
area 500 mm  630 mm was wrapped around a steel roller with the diameter of 17. Roldan-Carmona, C. et al. Flexible high efficiency perovskite solar cells. Energy
200 mm, forming Ni roller mould. The ultraviolet-resin (thickness of 7 mm)-coated Environ. Sci. 7, 994–997 (2014).
PET (thickness of 50 mm) substrate was transferred at a speed of 10 m min  1 by 18. Docampo, P. et al. Efficient organometal trihalide perovskite planar-
roll-to-roll nanoimprinting machine, and imprinted by the Ni roller mould with a heterojunction solar cells on flexible polymer substrates. Nat. Commun. 4, 2761
pressure of 0.25 g under ultraviolet irradiation. The roll-to-roll ultraviolet (2013).
nanoimprinting lithography can significantly improve the substrate imprinting 19. Das, S. et al. High-performance flexible perovskite solar cells by using a
efficiency. Third, the imprinted groove was filled with Ag ink containing combination of ultrasonic spray-coating and low thermal budget photonic
conductive submicron Ag nanoparticles (viscosity of 500 c.p.s.) and the excess Ag curing. ACS Photon. 2, 680–686 (2015).
ink was removed by means of scraping. The same process was repeated at least 20. Kim, B. J. et al. Highly efficient and bending durable perovskite solar cells:
three times for complete filling. Meanwhile, to make the surface clear, the wiping toward a wearable power source. Energy Environ. Sci. 8, 916–921 (2015).
process was performed with ethanol. Subsequently, the substrate underwent 21. Shin, S. S. et al. High-performance flexible perovskite solar cells exploiting
sintering at 80 °C for 10 min. Finally, an antistatic protective film was covered on Zn2SnO4 prepared in solution below 100 °C. Nat. Commun. 6, 7410 (2015).
the completed substrate. 22. Chirilă, A. et al. Highly efficient Cu(In,Ga)Se2 solar cells grown on flexible
polymer films. Nat. Mater. 10, 857–861 (2011).
23. Kranz, L. et al. Doping of polycrystalline CdTe for high-efficiency solar cells on
Flexible pero-SCs fabrication. A 15-inch FEAM substrate containing 100 pieces
flexible metal foil. Nat. Commun. 4, 2306 (2013).
of modules with a size of 15 mm  15 mm were cut off and then peeled off the
24. You, J. et al. Low-temperature solution-processed perovskite solar cells with
antistatic protective film, on which we fabricated perovskite solar cells, directly. The
processing and device geometry are exactly the same as the reference cells made on high efficiency and flexibility. ACS Nano 8, 1674–1680 (2014).
the ITO-coated glass. The Supplementary Fig. 11 showed the fabricated process of 25. Li, Z. et al. Laminated carbon nanotube networks for metal electrode-free
flexible pero-SC as the following steps: Step 1, the high conductivity polymer efficient perovskite solar cells. ACS Nano 8, 6797–6804 (2014).
PH1000 (Heraeus, Clevios PH1000, Germany) with the thickness of 150 nm was 26. Han, J. et al. Fully indium-free flexible Ag nanowires/ZnO:F composite
spin-coated on the surface of FEAM under 800 r.p.m., and then baked at 120 °C for transparent conductive electrodes with high haze. J. Mater. Chem. A 3,
15 min in ambient air. Step 2, an B40-nm-thick buffer layer of PEDOT-4083 5375–5384 (2015).
(Heraeus, Clevios PVPAI 4083, Germany) was subsequently spin-coated at 27. Ahn, B. Y. et al. Omnidirectional printing of flexible, stretchable, and spanning
2,000 r.p.m. and baked at 120 °C for 20 min in ambient air. Step 3, the substrates silver microelectrodes. Science 323, 1590–1593 (2009).
were transferred into nitrogen glove box for coating of the perovskite layer. In this 28. Mao, L. et al. Flexible silver grid/PEDOT:PSS hybrid electrodes for large area
study, for perovskite layer, we adopted two-step spin-coating method to make a inverted polymer solar cells. Nano Energy 10, 259–267 (2014).
dense and highly uniform film. The 1 M PbI2 solution was stirred in N,N-Dimethyl 29. Zhou, L. et al. High-performance flexible organic light-emitting diodes using
formamide solvent at 70 °C for 4 h. The PbI2 solution was spin-coated on substrates embedded silver network transparent electrodes. ACS Nano 8, 12796–12805
and then dried at 70 °C for 10 min. Step 4 and 5, 50 mg ml  1 of CH3NH3I (MAI) (2014).
in 2-propanol solution was coated on PbI2 layer and the film was taken out for 30. Sun, S. et al. The origin of high efficiency in low-temperature solution-
annealing in ambient air at 130 °C for 1 min 40 s for the champion device. Step 6, processable bilayer organometal halide hybrid solar cells. Energy Environ. Sci. 7,
for PCBM coating (B60 nm), a 2 wt% PCBM in chlorobenzene solution was spin- 399–407 (2014).
coated onto the perovskite layer. Step 7, the devices were transferred into a vacuum 31. Jeng, J.-Y. et al. CH3NH3PbI3 perovskite/fullerene planar-heterojunction hybrid
chamber for 100 nm Al electrode deposition by thermal evaporation at 4  10  6 solar cells. Adv. Mater. 25, 3727–3732 (2013).
bar with a shadow mask. The active area was 0.1 cm2. Finally, step 8, the highly 32. Xing, G. et al. Long-range balanced electron- and hole-transport lengths in
hardening treatment PET protection film coated on the bottom of FEAM substrate organic-inorganic CH3NH3PbI3. Science 342, 344–347 (2013).
was peeled off. The ultra-thin flexible pero-SCs were obtained. The detailed 33. Xia, Y. et al. Solution-processed metallic conducting polymer films as
device and perovskite film for characterization are described in Supplementary transparent electrode of optoelectronic devices. Adv. Mater. 24, 2436–2440
Methods.
(2012).
34. Xiao, Z. et al. Efficient, high yield perovskite photovoltaic devices grown by
References interdiffusion of solution-processed precursor stacking layers. Energy Environ.
1. Yang, W. S. et al. High-performance photovoltaic perovskite layers fabricated Sci. 7, 2619–2623 (2014).
through intramolecular exchange. Science 347, 522–525 (2015). 35. Shi, J. et al. Interfaces in perovskite solar cells. Small 11, 2472–2486 (2015).
2. Kojima, A. et al. Organometal halide perovskites as visible-light sensitizers for 36. Shimizu, H. et al. Synthesis, structures, and photophysical properties of
photovoltaic cells. J. Am. Chem. Soc. 131, 6050–6051 (2009). p-expanded oligothiophene 8-mers and their saturn-like C60 complexes. J. Am.
3. Im, J. H. et al. 6.5% efficient perovskite quantum-dot-sensitized solar cell. Chem. Soc. 137, 3877–3885 (2015).
Nanoscale 3, 4088–4093 (2011). 37. Dou, L. et al. Solution-processed hybrid perovskite photodetectors with high
4. Kim, H. S. et al. Lead iodide perovskite sensitized all-solid-state submicron thin detectivity. Nat. Commun. 5, 5404 (2014).
film mesoscopic solar cell with efficiency exceeding 9%. Sci. Rep. 2, 591 (2012). 38. Shiu, K.-T. et al. Ultrathin film, high specific power InP solar cells on flexible
5. Liu, M. et al. Efficient planar heterojunction perovskite solar cells by vapour plastic substrates. Appl. Phys. Lett. 95, 223503 (2009).
deposition. Nature 501, 395–398 (2013). 39. Zhao, D. et al. High-efficiency solution-processed planar perovskite solar cells
6. Burschka, J. et al. Sequential deposition as a route to high-performance with a polymer hole transport layer. Adv. Energy Mater. 5, 1401855 (2015).
perovskite-sensitized solar cells. Nature 499, 316–319 (2013). 40. Mandoc, M. M. et al. Effect of traps on the performance of bulk heterojunction
7. Zhou, H. et al. Interface engineering of highly efficient perovskite solar cells. organic solar cells. Appl. Phys. Lett. 91, 263505 (2007).
Science 345, 542–546 (2014). 41. Liang, P.-W. et al. Additive enhanced crystallization of solution-processed
8. Jeon, N. J. et al. Compositional engineering of perovskite materials for high- perovskite for highly efficient planar-heterojunction solar cells. Adv. Mater. 26,
performance solar cells. Nature 517, 476–480 (2015). 3748–3754 (2014).
9. Mei, A. et al. A hole-conductor–free, fully printable mesoscopic perovskite solar 42. Baikie, T. et al. Synthesis and crystal chemistry of the hybrid perovskite
cell with high stability. Science 345, 295–298 (2014). (CH3NH3)PbI3 for solid-state sensitised solar cell applications. J. Mater. Chem.
10. Guarnera, S. et al. Improving the long-term stability of perovskite solar cells A 1, 5628–5641 (2013).
with a porous Al2O3 buffer layer. J. Phys. Chem. Lett. 6, 432–437 (2015). 43. Chen, Q. et al. Controllable self-induced passivation of hybrid lead iodide
11. Kim, J. H. et al. Enhanced environmental stability of planar heterojunction perovskites toward high performance solar cells. Nano Lett. 14, 4158–4163
perovskite solar cells based on blade-coating. Adv. Energy Mater. 5, 1401229 (2014).
(2015). 44. Bi, C. et al. Understanding the formation and evolution of interdiffusion grown
12. Lee, M. M. et al. Efficient hybrid solar cells based on meso-superstructured organolead halide perovskite thin films by thermal annealing. J. Mater. Chem. A
organometal halide perovskites. Science 338, 643–647 (2012). 2, 18508–18514 (2014).

NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10214

45. Seo, J. et al. Benefits of very thin PCBM and LiF layers for solution-processed the other characterization experiments. Y.W.L., Z.H., Y.Y. and Y.F.L. wrote the manu-
p-i-n perovskite solar cells. Energy Environ. Sci. 7, 2642–2646 (2014). script. All the authors commented on the manuscript.
46. Lin, Q. et al. Flexible photovoltaic technologies. J. Mater. Chem. C 2, 1233–1247
(2014).
47. Stranks, S. D. et al. Metal-halide perovskites for photovoltaic and light-emitting Additional information
devices. Nat. Nanotechnol. 10, 391–402 (2015). Supplementary Information accompanies this paper at http://www.nature.com/
naturecommunications

Acknowledgements Competing financial interests: The authors declare no competing financial interests.
Y.W.L. acknowledged the financial support from Jiangsu Government Scholarship for Reprints and permission information is available online at http://npg.nature.com/
Overseas Studies of China. Zachary St Louis and Sam Sarhaddi are acknowledged for reprintsandpermissions/
extensive manuscript editing. Dr. Yulong Gao from Shine Optoelectronics (Kunshan)
Co., Ltd. is acknowledged for collaboration of flexible substrates. This work was finan- How to cite this article: Li, Y. et al. High-efficiency robust perovskite solar cells
cially supported by a grant from National Natural Science Foundation of China (Grant on ultrathin flexible substrates. Nat. Commun. 7:10214 doi: 10.1038/ncomms10214
No. 21204057), National Science Foundation (grant number ECCS-1202231, Program (2016).
Director Dr Paul Werbos), Air Force Office of Scientific Research (grant number
FA9550-15-1-0333, Program Manager Dr Charles Lee).
This work is licensed under a Creative Commons Attribution 4.0
International License. The images or other third party material in this
Author contributions article are included in the article’s Creative Commons license, unless indicated otherwise
Y.W.L., Z.H., G.L., Y.Y. and Y.F.L. developed the concept and designed the in the credit line; if the material is not included under the Creative Commons license,
experiments. Y.W.L. and L.M. performed the device and SEM experiments. Y.W.L. and users will need to obtain permission from the license holder to reproduce the material.
G.X. designed the flexible hybrid electrode. Y.W.L., J.Y. and Y.(Michael) Y. performed To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

10 NATURE COMMUNICATIONS | 7:10214 | DOI: 10.1038/ncomms10214 | www.nature.com/naturecommunications