A polymer scaffold for self-healing perovskite solar cells — TEI-tagged view

Yicheng ZhaopersName, Jing WeipersName, Heng LipersName, Yin YanpersName, Wenke ZhoupersName, Dapeng YupersName, Qing ZhaopersName (2016date). “A polymer scaffold for self-healing perovskite solar cells.” Nature Communicationstitle·j.

Object of study — perovskite solar cellsterm built on a polyethylene glycolterm polymer scaffoldterm·key

Research question — whether a polymer scaffoldterm·key gives uniform, efficient and humidity-resistant perovskite solar cellsterm

Contribution — a mild-temperature polymer-scaffold architecture giving ~16% efficiency, humidityterm·key resistance and self-healingterm·dom

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

A polymer scaffoldterm·key for self-healingterm·dom perovskite solar cellsterm Abstract Advancing of the lead halide perovskite solar cellsterm towards photovoltaic market demands large-scale devices of high-power conversion efficiency, high reproducibility and stability via low-cost fabrication technology, and in particular resistance to humid environment for long-time operation. Here we achieve uniform perovskite film based on a novel polymer-scaffold architecture via a mild-temperature process. These solar cells exhibit efficiency of up to ∼16% with small variation. The unencapsulated devices retain high output for up to 300 h in highly humid environment (70% relative humidityterm·key). Moreover, they show strong humidityterm·key resistant and self-healingterm·dom behaviour, recovering rapidly after removing from water vapour. Not only the film can self-heal in this case, but the corresponding devices can present power conversion efficiencyterm·dom recovery after the water vapour is removed. Our work demonstrates the value of cheap, long chain and hygroscopic polymer scaffoldterm·key in perovskite solar cellsterm towards commercialization.

The rapid rising of the lead halide perovskite solar cellsterm (PSCsterm), typically based on CH 3 NH 3 PbI 3 (MAPbI 3 ) or its analogues, demonstrate great potential for market applications 1 2 3 4 5 6 7 8 9 10 11 12 13 . The worldwide power conversion efficiencyterm·dom (PCEterm·dom) race for the perovskite solar cellsterm has stimulated the highest record of >20% update via solvent engineering 8 , interface engineering 5 and composition engineering 10 , which approaches the best of polycrystalline silicon modules. The excellent performance of PSCsterm is attributed to the high absorption coefficient, weak exciton binding energy, and long diffusion length of the perovskite phase 12 14 15 16 17 . Two dominant device architectures, namely mesoporous scaffoldterm·mth 1 2 3 and planar heterojunctionterm·mth structures 4 5 have been developed. The PCEterm·dom in mesoporous scaffoldterm·mth architecture depends on the pore size, porosity and morphology of the metal oxide nanoparticles, which dominantly determine the coverage, morphology variation of the perovskite layer and the carrier lifetime. Moreover, the fabrication of inorganic metal oxide mesoporous scaffoldterm·mth is complicated and needs sintering at high temperature of >450 °C (refs 2 , 3 ), which makes the fabrication more expensive. The planar heterojunctionterm·mth PSCsterm fabricated layer by layer is either fabricated via high vacuum deposition 4 or solution-based method 5 18 . However, the low-cost solution process is rather challenging to form homogeneous film due to the dewetting process and sensitivity to the atmosphere 18 19 20 . Recently, progress has been reported to control the nucleation rate by changing dissolvent or temperature of the substrate 21 22 . Besides high efficiency and high reproducibility, long-term stability is also a crucial barrier to commercialize the PSCsterm, because the lead halide perovskite materials are very sensitive to ambient humidityterm·key and easy to dissolve and degrade in humid environment 5 23 24 . Therefore, it is very challenging and pressing to develop ultra stable and high PCEterm·dom devices resistant to hostile operation environment.

Here we report a novel perovskite solar cellterm architecture based on an insulating polymer scaffoldterm·key structure. The polymer-scaffold perovskite layer is simply fabricated via a single one-step process under mild temperature of 105 °C. The as-made polymer-scaffold perovskite solar cellterm·mth (PPSCterm·mth) devices demonstrate high PCEterm·dom up to ∼16% under standard AM1.5 illumination and very small efficiency variation compared with that without the insulating polymer. Remarkably, unsealed PPSCsterm·mth exhibit excellent performance in stability test in highly humid environment (70% relative humidityterm·key), which can endure the moisture for over 300 h. Most strikingly, the PPSCsterm·mth show strong self-healingterm·dom or water resistant effect under vapour spray. The perovskite film, as well as the corresponding devices can recover rapidly to its original phase and performance after being vapour sprayed and dried in ambient air, respectively.

Results Design, fabrication and characterization of PPSCsterm·mth The PPSCterm·mth architecture inherits the basic structure of PSCsterm with inorganic mesoporous scaffoldterm·mth 1 2 3 , instead, the inorganic nanocrystal scaffolds are replaced with long- and flexible-polymer chain networks, which can be processed under mild temperature ( Fig. 1a–c ). The fabrication of PPSCterm·mth is similar to that of planar heterojunctionterm·mth PSCsterm, except for the precursor solution preparation ( Fig. 1d ). Long-chain insulating polymers, dissolved in precursor solution, assemble the polymer scaffoldterm·key simultaneously as the precursor solution is spin coated on the substrate. The selected polymer should be chemically inert to ingredients for perovskite synthesis; it should also act as an electrical insulator to guarantee the effective transport of photo-generated carriers. Here polyethylene glycolterm (PEGterm), which meets the requirements, is chosen to be added in the precursor solution of PbCl 2 and CH 3 NH 3 I (MAI; 1:3 molar ratio) in dimethylformamide. The molar ratio of PEGterm monomers (C 2 H 4 O) to ultimate product MAPbI 3 is close to 1:1 (PEGterm in 40 mg ml −1 ). The PEGterm scaffold perovskite absorber layer was prepared on the TiO 2 (∼40 nm thickness)-coated fluorine doped tin oxide (FTO) glass by spin coating followed by annealing at 105 °C for 70 min. 2,29,7,79-tetrakis-(N,N-di-p-methoxyphenylamine)9,99-spirobifluorene (Spiro-OMeTAD) and gold were used as the hole transport layer and electrode, respectively.

Figure 1e,f present top-view scanning electron microscope (SEM) images of the perovskite films without and with polymer scaffoldterm·key. Those without scaffold exhibited large pinholes and bare conducting substrates, increasing the risk of short circuits. Those with scaffold demonstrated continuous and complete coverage. The cross-section SEM image showed the perovskite layer without polymer to have considerable variation in thickness and many voids ( Fig. 1g ). The film with polymer was uniform in thickness over large area ( Fig. 1h ). The PEGterm scaffold can also work in another precursor solution PbI 2 :MAI (1:1) ( Supplementary Fig. 1 ). PEGterm plays an important role in improvement of the film quality for the following reasons. First, the simultaneously formed PEGterm scaffold, which is formed by entangled long-chain molecules, acts as a three-dimensional skeleton to support perovskite crystals, causing it to cover the substrate more uniformly ( Fig. 1c ), demonstrating an advantage over planar heterojunctionterm·mth PSCsterm. Uniformly distributed PEGterm scaffolds function as Al 2 O 3 scaffolds 3 , and they can be produced in a much simpler and less costly manner. Second, the wetting properties of the precursor solution have been improved after introducing PEGterm ( Supplementary Fig. 2 ), which can improve film morphology 19 . Third, PEGterm molecules in precursor solution can slow down the crystallization of perovskite from the X-ray diffraction characterization by tracking the intermediate phase in the crystallization process, which is shown below.

To confirm that PEGterm acts as scaffold in the architecture, detailed energy dispersive spectroscopy (EDS) analysis in a Cs-corrected scanning transmission electron microscope (STEM) was conducted for element mapping investigation ( Fig. 2a–c ). The high-angle annular dark field (HAADF) STEM image ( Fig. 2a ) reveals that the perovskite crystalline grains are surrounded by polymer material. The profiles of perovskite grains are highlighted with dashed lines according to the EDS mapping of Pb component ( Fig. 2b ). The amorphous material surrounding the perovskite grains is PEGterm in accordance with the EDS mapping of O element because only PEGterm contains O. The O mapping demonstrates that PEGterm not only covers the perovskite grains, but also cross-links the neighbouring perovskite grains via self-assembly forming a polymer-scaffold network to support the perovskite grains ( Fig. 2c ). Above conclusion is also evidenced by the fact that O mapping covers larger area than Pb and extends from crystal boundaries, indicating that the perovskite grains are imbedded in the PEGterm networks, corresponding well to the HAADF–STEM image. In addition, the O element mapping on the cross-sectional area of the PPSCterm·mth film confirms a homogeneous distribution of PEGterm molecules in the perovskite film ( Supplementary Fig. 3 ). High-resolution transmission electron microscope image of perovskite grain demonstrates good crystal lattice ( Supplementary Fig. 4 ), suggesting that PEGterm molecules do not damage the perovskite crystal structure.

The phase evolution of PEGterm scaffold perovskite film during annealing was characterized by X-ray diffraction analysis ( Fig. 2d ). Over the course of 30–50 min, the film turned from orange to yellow, the (110) and (220) peaks of perovskite phase 25 26 became more pronounced and some new peaks appeared between 10 and 20°. The intermediate phase disappeared after annealing for 60 min, and the sample became deep black at 105 °C after 70 min of annealing, with stronger (110) and (220) peaks. X-ray diffraction patterns of perovskite film without PEGterm and with PEGterm under the same 100 °C sintering temperature are also shown in Supplementary Fig. 5 . Note that intermediate phase disappears after 70 min in film without PEGterm. While in film with PEGterm, intermediate phase disappears after 90 min ( Supplementary Note 1 ), which strongly suggests a slower crystallization process in PEGterm scaffold perovskite film.

Photoluminescence (PL) spectra were measured to verify whether excess recombination sites and defect states are induced in the PEGterm scaffold perovskite film 27 28 . Compared with the PL spectrum of pristine perovskite film, the intensity is not influenced after being coated with PEGterm ( Fig. 3b ), indicating the insulating nature of PEGterm, as shown in Fig. 3a . Figure 3c shows the time-resolved PL spectra with and without PEGterm polymer scaffoldterm·key, with the similar carrier lifetime by using single exponential fitting (110.8 ns with PEGterm, 120.1 ns without PEGterm). This result indicates ∼1 μm diffusion length if diffusion coefficient is supposed to be 0.1 cm 2 s −1 , which is sufficient to guarantee effective collection efficiency. On the other hand, no visible shift (peak position of 760 nm) and broadening (half width of 40 nm) of peaks associated with perovskite layers with and without PEGterm scaffold is observed, indicating that PEGterm does not induce any excessive recombination ( Supplementary Fig. 6 ).

Photovoltaic performance of PPSCsterm·mth The photovoltaic properties of PPSCsterm·mth are characterized systematically. The PEGterm molecular weight and its concentration in precursor solution were found to play an important role in the performance of PPSCsterm·mth. Figure 4a shows the current density–voltage ( J – V ) curves of samples with PEGterm in different molecule weights (12,000/20,000/100,000 Da) measured under simulated AM1.5 illumination of 100 mW cm −2 . The molecular weight of PEGterm, proportional to the length of polymer, has a significant effect on the performance of PPSCsterm·mth, and the optimal molecular weight was found to be >20,000 Da because longer PEGterm molecules facilitate the formation of three-dimensional molecular network in the perovskite film, and guarantee uniform morphology. PEG concentrationterm·var is also critical to devices because it is related to the series resistanceterm·dom. As demonstrated in Fig. 4b , the device shows much better performance with 20 and 40 mg ml −1 PEG concentrationterm·var, and its PCEterm·dom decreases at higher concentrations (80 mg ml −1 ) because the excess PEGterm in the film increases the series resistanceterm·dom and lowers the efficiency. Note that viscosity of the solution can be strongly influenced by PEG concentrationterm·var and molecule weight, so the r.p.m. value of spin coating is optimized at different PEG concentrationterm·var and molecule weight (see Methods for details) to guarantee ∼400 nm film thickness. The improved efficiency was also consistent with the increased recombination time constant 28 in PPSCsterm·mth, compared with that of control devices, as is shown in Fig. 4c . Owing to the improved coverage, the recombination time is almost one order of magnitude larger than that without polymer scaffoldterm·key at zero bias. The highest PCEterm·dom of 16% was obtained for PPSCsterm·mth with optimal PEGterm molecule weight of 20,000 Da and concentration of 20 mg ml −1 , and showed an open voltage ( V oc ) of 0.98 V, short-circuit current ( J sc ) of 22.5 mA cm −2 , and fill factor of 0.72 ( Supplementary Fig. 7a ). Another representative device shows a steady-state PCEterm·dom of 15.4% at first 10 s, and the PCEterm·dom gradually drops down and eventually stabilized at 13.5% in 1,000 s (0.7 V bias, Supplementary Fig. 7b ). Parallel devices present H value from 10 to 30% under 500 mV s −1 if hysteresis index is defined H =(PCEterm·dom reverse −PCEterm·dom forward )/PCEterm·dom reverse × 100% to characterize the hysteresis effect. PCEterm·dom reverse and PCEterm·dom forward represent PCEterm·dom obtained from J to V measurement in reverse scan and forward scan, respectively. The statistical PCEterm·dom distributions of >100 samples without and with (for two concentrations) polymer scaffoldsterm·key are shown in Fig. 4d . The PPSCsterm·mth exhibited a narrow PCEterm·dom distribution for 20 and 40 mg ml −1 , with an average efficiency of 14 and 12.5%, respectively, but the PCEterm·dom distribution of the PSCsterm without polymer scaffoldterm·key was much broader, varying from 2 to 14%, averaging at ∼8%. The significant increase in PCEterm·dom and high reproducibility of the PPSCterm·mth devices are direct reflections of the uniform thickness and homogeneous morphology of the polymer-scaffold perovskite layer.

Improved stability in highly humid environment More importantly, the insulating PEGterm scaffold structure can stabilize the PSCsterm in humid environment due to the strong hygroscopicityterm·dom property of PEGterm. Here three groups of unsealed devices were prepared without and with PEGterm (two concentrations), respectively. Figure 5a shows the PCEterm·dom evolution as a function of time of the three groups of devices in the dark under highly humid environment (relative humidityterm·key 70%) without any sealing. In contrast to the fast PCEterm·dom decay from 12 to 0.5% within 50 h of the pristine PSCsterm, the PPSCsterm·mth retained 65% of its PCEterm·dom after 300 h aging. Similarly, the J sc , V oc , and fill factor of the pristine sample decreased quickly during the first 50 h, while those of the PPSCsterm·mth retained relatively high values for up to 300 h ( Supplementary Fig. 8a,b ). Note that the device with PEG concentrationterm·var of 40 mg ml −1 showed better performance than the one with 20 mg ml −1 because more PEGterm can form a more compact layer, enhancing the moisture resistance of devices. We also studied the stability under continuous lightterm·dom illumination in 70% humidityterm·key for unsealed devices ( Supplementary Fig. 8c,d ). Photovoltaic performance degraded quickly within 10 h in the device without PEGterm scaffold, while PPSCterm·mth retained 83% of its PCEterm·dom during the same aging period. Similarly, V oc , J sc , and fill factor suffered substantial decrease in the first 10 h for pristine sample while they remained relatively high values in the presence of PEGterm. X-ray diffraction patterns of the perovskite film with PEGterm scaffold showed no structure destruction after 72 h continuous lightterm·dom irradiation, while the samples without PEGterm degraded into PbI 2 quickly within 10 h ( Supplementary Fig. 9 ). Evidence above strongly suggests that PEGterm can greatly protect perovskite film from decomposition under moisture attack and significantly improve the stability of PPSCsterm·mth at very humid environment.

Self-healingterm·dom property of PPSCsterm·mth after vapour spray In addition to its improved stability, the PPSCterm·mth devices can demonstrate a self-healingterm·dom behaviour and humidityterm·key resistance effect on exposure to water vapour. Figure 5b shows a comparison of the changes in colour of the perovskite layer with PEGterm (top) and the perovskite layer without PEGterm (bottom) after both were sprayed with water vapour for 60 s. The perovskite film without PEGterm decomposed into PbI 2 and turned yellow irreversibly. In contrast, the PEGterm scaffold perovskite film showed yellow in colour at first and recovered to black in ∼45 s after removing from the spray ( Fig. 5b ). This amazing self-healingterm·dom behaviour is further vividly demonstrated in 120 s videos ( Supplementary Movies 1 and 2 ). The self-healingterm·dom capability of the PPSCterm·mth device was also manifested in the J – V characterization ( Fig. 5c ). The PCEterm·dom degrades with smaller V oc and J sc after exposure to water vapour for 60 s. Surprisingly, results showed that the J – V curve can almost regain its original value in 45 s, indicating that the PPSCsterm·mth can heal itself when it is returned to ambient air after being sprayed with water. This unique humidity-resist characteristic is very suitable for practical applications because once the devices are exposed to very humid environment; they can self-heal to high PCEterm·dom again in a short time when they return to sunlight again. Moreover, to further prove the self-healingterm·dom property, perovskite film with PEGterm was sprayed with water vapour for 60 s, then after self-healingterm·dom, it was coated by Spiro-MeOTAD and Au electrode to fabricate into corresponding device, which showed photovoltaic performance comparable to its original PPSCterm·mth devices without water spray ( Supplementary Fig. 10 ).

The self-healingterm·dom process was recorded by X-ray diffraction analysis ( Fig. 5d ). The peak at 12.7° reveals that PbI 2 phase formed when water vapour was sprayed on the film. After self-healingterm·dom, the PbI 2 peak disappeared and the perovskite film fully recovered to its original crystal phase. The self-healingterm·dom can also be demonstrated by the absorption spectra ( Fig. 5e ), which presented the same absorbance and similar Urbach energy (∼40 meV; ref. 29 ). The decomposition of perovskite phase is caused by the chemical reaction with water molecules 30 31 32 , which is manifested by the X-ray diffraction analysis of the perovskite film without PEGterm after water spray ( Supplementary Fig. 11 ).

According to the observations in Fig. 2a–c , we conjecture that these entangled PEGterm molecules are anchored on the surface of perovskite grains by hydrogen bonding, and the formation is illustrated in Fig. 6a . The –OH I–interaction had been evidenced in the work by Li et al 32 . We use liquid state 1 H NMR measurements to prove such interactions. The proton NMR spectra of three samples were compared ( Fig. 6b ): deuterated DMSO solution with MAPbI 3 (sample 1), with MAPbI 3 +PEGterm (sample 2) and with PEGterm (sample 3). In the signals of sample 3, double methylene group linked to oxygen on both ends (–[O–CH 2 –CH 2 ] n –) is characterized by the peak at δ =3.48 p.p.m. ( Fig. 6b ). An upfield chemical shift of Δδ ∼−0.18 p.p.m. with several splitting peaks is observed in sample 2. Such chemical shift can be explained by the hydrogen bond between MA + and O in PEGterm chain, which weakens the influence of O on protons of methylene in –[O–CH 2 –CH 2 ] n –. Furthermore, the proton resonance signals of –NH 3 + in sample 1 (peak at δ =7.33 p.p.m.) shift towards upfield with Δδ∼ −0.13 p.p.m. in sample 2 ( Fig. 6b ), which can also be attributed to the hydrogen bonds discussed above. Such interaction is also evidenced by a dissolution experimentterm·mth devised by us ( Supplementary Fig. 12 ).

As presented in Fig. 6c , the improved stability and self-healingterm·dom effect of the PPSCterm·mth devices can be ascribed to the excellent hygroscopicityterm·dom of the PEGterm molecules and their strong interaction with the perovskite. On one hand, the omnipresent PEGterm molecules can absorb water efficiently to form a compact moisture barrier around perovskite crystal grains with little water penetrating into the film. On the other hand, on water spray, the black perovskite film with PEGterm turned lightterm·dom yellow (PbI 2 forms) at first. However, due to the strong interaction between MAI and PEGterm, (NMR measurements and a dissolution experimentterm·mth in Supplementary Fig. 12 ), the MAI molecules were anchored by the nearby PEGterm molecules rather than escape away (process 1, 2 in Fig. 6c ). After being kept away from water vapour, PbI 2 in the film reacted with nearby MAI to regenerate the perovskite MAPbI 3 phase, very similar to the two-step synthesis 2 (process 3 in Fig. 6c ). The instant decomposition-regeneration mechanism explains the fast self-healingterm·dom process in the PEGterm scaffold perovskite film.

Discussion The hygroscopic PEGterm scaffold can stabilize the perovskite film, rendering the devices resistant to moisture with strong self-healingterm·dom property. Hygroscopic polymer-scaffold architecture paved a new effective way in perovskite solar cellterm, to solve the hydrolysis problem in ambient air with greatly decreased package cost. The polymer-scaffold perovskite layer reinforces the architecture and improves strain tolerance. It may also facilitate production of flexible, wearable devices in the future. Future work will be focused on looking for cheap long-chain polymers which have stronger hygroscopicityterm·dom and binding effect with MAI to protect the perovskite from decomposition in ambient environment. Moreover, by comparing the stability test result under lightterm·dom illumination ( Supplementary Fig. 8 ) and in dark ( Fig. 5a ), we found that lightterm·dom has a strong destructive influence on stability performance. One would expect that this influence comes from decomposition of perovskite film accelerated by lightterm·dom with the aid of hydrolysis by water molecules under illumination (reactive equation in Supplementary Fig. 11 ). However, after 72 h lightterm·dom soaking, the perovskite film with PEGterm was not destructed compared with that without PEGterm from X-ray diffraction spectra ( Supplementary Fig. 9 ). This indicated another important factor influencing the efficiency in addition to material decomposition. Hence, we speculate that I − ion migrationterm·dom 33 34 35 36 may be induced or enhanced by lightterm·dom. Ion migrationterm·dom may answer for the poor stability under illumination, no matter in the traditional architecture without PEGterm scaffold or in our PPSCsterm·mth, which needs further investigation. Further improvement needs be focused on revealing the mechanism of ion migrationterm·dom in the long-term stability issue, as well as how to inhibit it in PPSCsterm·mth, which would lead to much better performance in stability test under illumination with high humidityterm·key.

Methods Device fabrication and device characterization All chemicals were purchased from Sigma-Aldrich or J&K Scientific Ltd. unless otherwise stated. The photovoltaic devices were fabricated on FTO-coated glass (Pilkington, Nippon Sheet Glass). First, laser-patterned, FTO-coated glass substrates were cleaned by ultrasonic cleaned in deionized water, acetone and ethanol, followed by an ultraviolet treatment for 5 min. Compact layers were deposited on the substrates by spin coating titanium diisopropoxide bis(acetylacetonate) solution (75% in 2-propanol) diluted in ethanol (1:20, volume ratio) for two times and annealed at 450 °C for 30 min. After cooling to room temperature, the substrates were transferred to a hot plate at 90 °C before spin coating. CH 3 NH 3 I was synthesized according to the reported procedure 1 . In a typical synthesis, 33.77 ml methylamine (33% in methanol) and 30 ml of hydroiodic acid (57% in water) were reacting in a 250 ml flask at 0 °C for 2 h under stirring. The precipitate was formed by evaporation at 50 °C for 1 h. The product, methylammonium iodide CH 3 NH 3 I, was washed with diethyl ether by stirring the solution for 30 min, repeated three times, and then finally dried at 60 °C in vacuum oven for 24 h. The prepared CH 3 NH 3 I, PbCl 2 (Aldrich) (3:1) for 430 and 250 mg ml −1 solution was mixed with PEGterm (Sigma-Aldrich) and stirred in dimethylformamide at 60 °C for 12 h. The molar ratio of PEGterm monomers (C 2 H 4 O) to ultimate product MAPbI 3 is ∼1:1 (PEGterm in 40 mg ml −1 ). The resulting solution was then coated onto the TiO 2 /FTO substrate by spin coating at 500 r.p.m. for 10 s, and then 3,500/4,500/5,500 r.p.m. for 30 s, for the solution with 20/40/80 mg ml −1 PEGterm, respectively. For different molecular weight 12,000/20,000/100,000 Da. with 20 mg ml −1 , we use the optimized 3.500/3.500/4.500 r.p.m. to prepare the film. Then the substrate was dried on a hot plate at 60 °C for 45 min, then sintered at 105 °C for 70 min. For the preparation of peroskite film without PEGterm, the optimized sintering temperature is 100 °C. After cooling to room temperature, the hole transport material was spin coated onto the perovskite film at 3,000 r.p.m. for 40 s. The spin coating formulation was prepared by dissolving 72.3 mg 2,2′,7,7′-Tetrakis(N,N-p-dimethoxy-phenylamino)-9,9′-spirobifluorene(spiro-MeOTAD), purchased from Yingkou OPV Tech New Energy Co. Ltd., 30 μl 4-tert-butylpyri-dine and 20 μl of a stock solution of 520 mg ml −1 lithium bis (trifluoromethylsulphonyl) imide in acetonitrile in 1 ml chlorobenzene. Finally, 90-nm-thick gold electrodes were deposited on top of the devices by evaporation at ∼10 −4 mbar. The active area of the electrode was fixed at 9 mm 2 .

The surface morphology and EDS mapping of the perovskite thin film was characterized by SEM (Nano430, FEI). The instrument uses an electron beam accelerated at 15 kV, enabling operation at a variety of currents. Considering that only PEGterm polymer contains oxygen element among those materials used in perovskite thin film synthesis, here oxygen stands for PEGterm polymer, lead for perovskite. HAADF–STEM is examined by Cs-corrected FEI (Titan G2 80–200) transmission electron microscope operating at an accelerating voltage of 300 kV, equipped with ChemiSTEM EDS detector, which enables very quick high-resolution element mapping. The sample is prepared from perovskite precursor solution spin coated on TEM grid (3,000 r.p.m.), then annealed at 105 °C for 70 min.

For NMR measurement, it is examined by Bruker 600 UltraShield using pulse signal. We use deuterated DMSO as solvent. The CH 3 NH 3 PbI 3 solution is prepared from CH 3 NH 3 I, PbI 2 (Aldrich; 1:1) for 53 and 153 mg ml −1 solution and stirred in DMSO at 60 °C for 12 h. CH 3 NH 3 PbI 3 solution mixed with PEGterm (Sigma-Aldrich) is prepared by adding 40 mg PEGterm (molecule weight: 20,000 Da) to 1 ml CH 3 NH 3 PbI 3 solution.

For X-ray diffraction measurement, flat PbI 2 and TiO 2 /PbI 2 nanocomposites were deposited on glass slides using the above-mentioned procedures. X-ray powder diagrams were recorded on an X'PertMPD PRO from PANalytical equipped with a ceramic tube (Cu anode, λ =1.5406 Å), a secondary graphite (002) monochromator and a RTMS X'Celerator detector, and operated in BRAGG-BRENTANO geometry. The samples were mounted without further modification, and the automatic divergence slit and beam mask were adjusted to the dimensions of the thin films. A step size of 0.008 deg was chosen and an acquisition time of up to 7.5 min per deg. A baseline correction was applied to all X-ray powder diagrams to remove the broad diffraction peak arising from the amorphous glass slide.

Steady-state PL spectra were measured using a He–Cd laser (325 nm in wavelength) and green laser (514 nm in wavelength) guided by a micro-zone confocal Raman spectroscope (Renishaw inVia microRaman system) as the laser beam with a spot size diameter of 2 μm. The collected duration is ∼5 ms. The time-resolved fluorescence spectra were recorded with a high-resolution streak camera system (Hamamatsu C10910 ). We used an amplified mode-lock Ti: Sapphire femtosecond laser system (Legend, Coherent) and a two-stage optical parametric amplifier (OperA Solo, Coherent) to generate the pump beam with a repetition rate of 1 KHz. All the samples were excited by 517 nm at room temperature with 110 nJ cm −2 .

The electrochemical impedance spectrum and cyclic voltammograms were measured using a potentiostat/galvanostat (SP-150, Bio-Logic, France). The frequency can be tuned from 0.1 Hz to 1 MHz. All the samples were measured under 10 mW cm −2 .

The J – V characteristics were obtained using an Agilent B2900 Series precision source/measure unit, and the cell was illuminated by a solar simulator (Solar IV-150A, Zolix) under AM1.5 irradiation (100 mW cm −2 ). Lightterm·dom intensity was calibrated with a Newport calibrated KG5-filtered Si reference cell. We use black mask to define the cells' area, and the masking effect is confirmed by testing the J sc with and without it, which has a 5% J sc difference. The J – V curves are tested from 1.5 to −0.2 V with a scan velocity 500 mV s −1 . The masked active area is 9 mm 2 .

Stability test Unsealed PPSCsterm·mth and PSCsterm were put near a humidifier to control its relative humidityterm·key ∼70%. For the stability test under dark, the cells were stored in a black box with 70% relative humidityterm·key, and were tested every 8 h. For the stability tests under continuous lightterm·dom illumination, lightterm·dom source ranges from 300 to 800 nm wavelength with 70 mW cm −2 was used. J – V curves were recorded under AM1.5 lightterm·dom irradiation. The temperature is controlled ∼30 °C. Three devices were measured in this way and demonstrate excellent repeatability. For the self-healingterm·dom behaviour, we sprayed the water vapour for 60 s onto the resulting devices by humidifier. J – V curves were collected before and right after water vapour exposure under standard illumination AM1.5.

Additional information How to cite this article: ZhaopersName, Y. et al . A polymer scaffoldterm·key for self-healingterm·dom perovskite solar cellsterm. Nat. Commun . 7:10228 doi: 10.1038/ncomms10228 (2016).

Supplementary Material Supplementary Figures 1-12 and Supplementary Note 1 Real time videos of the perovskite film without PEGterm exposed to humidifier for about 60 sec and the color change after that in 1 min.

Real time videos of the perovskite film with PEGterm exposed to humidifier for about 60 sec and the color change after that in 1 min. Strong self-healingterm·dom behavior of perovskite film with polymer scaffoldterm·key was demonstrated. The perovskite film with polymer scaffoldterm·key can be self-healed after 1 min accompanied by the color change turning from yellow to black again.