Research ArticleENGINEERING

Polymer-modified halide perovskite films for efficient and stable planar heterojunction solar cells

See allHide authors and affiliations

Science Advances  23 Aug 2017:
Vol. 3, no. 8, e1700106
DOI: 10.1126/sciadv.1700106


The solution processing of polycrystalline perovskite films introduces trap states that can adversely affect their optoelectronic properties. Motivated by the use of small-molecule surfactants to improve the optoelectronic performance of perovskites, we demonstrate the use of polymers with coordinating groups to improve the performance of solution-processed semiconductor films. The use of these polymer modifiers results in a marked change in the electronic properties of the films, as measured by both carrier dynamics and overall device performance. The devices grown with the polymer poly(4-vinylpyridine) (PVP) show significantly enhanced power conversion efficiency from 16.9 ± 0.7% to 18.8 ± 0.8% (champion efficiency, 20.2%) from a reverse scan and stabilized champion efficiency from 17.5 to 19.1% [under a bias of 0.94 V and AM (air mass) 1.5-G, 1-sun illumination over 30 min] compared to controls without any passivation. Treating the perovskite film with PVP enables a VOC of up to 1.16 V, which is among the best reported for a CH3NH3PbI3 perovskite solar cell and one of the lowest voltage deficits reported for any perovskite to date. In addition, perovskite solar cells treated with PVP show a long shelf lifetime of up to 90 days (retaining 85% of the initial efficiency) and increased by a factor of more than 20 compared to those without any polymer (degrading to 85% after ~4 days). Our work opens up a new class of chemical additives for improving perovskite performance and should pave the way toward improving perovskite solar cells for high efficiency and stability.


Hybrid perovskite solar cells (HPSCs) show great potential as a renewable energy source due to their exceptional optoelectronic properties and the possibility of low-cost solution processing to reduce manufacturing cost and energy input (16). Starting from 3.8% in 2009 (7), the state-of-the-art HPSC power conversion efficiency (PCE) has surged to a certified state-of-the-art value of more than 22.1% within a short time span, owing to the recent advancements in interface modification (811), morphology manipulation (1217), composition engineering (1821), etc. It has been shown that an ultimate efficiency of more than 30% and a maximum VOC of 1.33 V can be theoretically achieved for the methylammonium lead tri-iodide (MAPbI3) systems (22). However, charge recombination via nonradiative channels, such as defects or trap states, constitutes a major energy loss mechanism, which sets the upper limits for practically achievable PCE of MAPbI3 solar cells (23). Thus far, as literature reports, the highest efficiency of a MAPbI3 HPSC is around 20% with a VOC of ~1.11 V (24, 25). Thus, further boosting the HPSC toward their ultimate device efficiency relies on both fundamental understanding and advanced strategies to suppress nonradiative energy loss channels, such as charge recombination via defect states in the bulk or interfaces of the HPSC.

Defect states are well known to have a significant effect on device performance in many thin-film semiconductor photovoltaic technologies (2628). For instance, abundant trap states can exist at the grain boundaries in semiconductors such as a polycrystalline Si thin film, where dangling bonds form deep defect states and significantly affect charge recombination and transport in photovoltaic devices (29). Because HPSCs are processed from solution at low temperature, defects in the polycrystalline perovskite films are likely unavoidable. These defects can lead to nonradiative energy loss, decreasing radiative efficiency, which may reduce the VOC and device performance below the Shockley-Queisser limit. For example, the grain boundaries of perovskite films show optoelectronic properties that are different from the bulk and exhibit shorter charge carrier lifetime and weaker photoluminescence (PL) intensity (30). Studies using Kelvin probe force microscopy and conductive atomic force microscopy have shown that the grain boundaries and the bulk have different electrical potentials (31) and that grain boundaries have faster ion migration than the bulk (32). However, contrary results have also been reported to claim benign or even beneficial grain boundaries (31, 3335). Notably, the chemical composition at the grain boundaries varies depending on the processing methods, which can be the most plausible explanation to the debate on the effects of grain boundaries (20, 25, 36). Through theoretical studies, Haruyama et al. investigated the influence of different terminations at the grain boundaries and interstitials and found that the grain boundary composition terminated by either the iodide or the ammonium ion can significantly affect the optoelectronic properties of the perovskite grains (3739). Tailoring the chemical composition of perovskite films can be an effective strategy to passivate defects and improve optoelectronic properties and overall device performance, which has been successfully demonstrated via small molecules, such as pyridine (40), thiophene (40), tri-n-octylphosphine oxide (41), and fullerenes (42, 43). However, the design rules for the selection of an efficient passivating agent for high HPSC device performance are lacking. Moreover, passivation of perovskite films, especially within the bulk film, using long-chain polymers that might coordinate with Pb2+ sites has rarely been reported (44), likely due to the challenges in film fabrication in the presence of a high–molecular weight polymer. However, compared to small molecules, polymers can form more stable and reliable interactions with the perovskite grains and could conceivably cross-link grains to improve film stability, as has been proposed for alkylphosphonic acid ω-ammonium chlorides (45).

Here, we successfully demonstrate a highly efficient and stable HPSC, which exhibits a champion reverse-scanned efficiency of 20.2% [averaged efficiency of 18.8 ± 0.8% and stabilized efficiency of 19.1%, with bias of 0.94 V, under AM (air mass) 1.5-G, 1-sun illumination] and long shelf lifetime, by chemically tailoring the perovskite film composition. We explore using polymers bearing different functional groups and propose that the binding interaction between the perovskite and the polymers is a key parameter affecting the passivation. The highest VOC of the HPSC by poly(4-vinylpyridine) (PVP) passivation reaches 1.16 V (1.14 ± 0.02 V) with the planar MAPbI3 system [Eg of ~1.63 eV from electroabsorption spectroscopy (46); absorbing edge of 795 nm], and thus, a champion small VOC deficit (0.47 V; Eg/e-VOC) is achieved in the MAPbI3 system, which is also among the lowest of all types of perovskite solar cells (47). In addition, the HPSC with PVP passivation shows less hysteresis, a stable efficiency of up to 19% at a bias of 0.94 V under AM 1.5-G, 1-sun illumination, and an improved shelf lifetime of up to 90 days, retaining more than 85% of the initial efficiency. The influence of different polymers on the carrier dynamics and trap-state densities of perovskite films is studied via transient PL, transient photovoltage (TPV)/transient photocurrent (TPC) decay, and electronic impedance spectroscopy (IS) measurements.


Figure 1A shows the chemical structure of the polymers and a schematic of their possible interaction with perovskite. The branched polyethyleneimine (b-PEI), PVP, and polyacrylic acid (PAA) polymers were selected because of their possibility for different chemical interactions with the perovskite film: The b-PEI is rich in (possibly protonated) amino groups, PVP is rich in Lewis base pyridine groups, and PAA is rich in carboxylic acid groups. The lone electron pairs at the N or O of b-PEI, PVP, and PAA can be delocalized to the 6p empty orbits of Pb2+ to form coordination bonds with different bonding strengths (48, 49). The bonding energy between the polymers and the MAPbI3 was estimated using density functional theory (DFT) (see the Supplementary Materials for the calculation method details in the vacuum state), and the absorption energy for amino, pyridine, and carboxylic groups to the MAPbI3 crystalline slab is calculated to be 51.88, 37.70, and 15.71 kJ/mol, respectively (Table 1). The strong interactions between the polymer and the perovskite can also be confirmed by the observation after adding the polymers into the perovskite precursor solutions (fig. S1A). For example, blending the b-PEI solution into the PbI2/methylammonium iodide (MAI) solution immediately forms a precipitate at the bottom of the vial, and all the lead component is absorbed by the b-PEI, rendering the upper liquid colorless. A gelation structure is formed by adding the PVP polymer (fig. S1B), but the solution remains clear with PAA. From the DFT calculation (vacuum state), we predicted that the interaction between the polymers and the perovskite could be a dative bond between the Pb2+ and functional groups on the polymer (that is, pyridine and amino), and formation of complex molecules of perovskite and polymers can also be expected.

Fig. 1 Polymer-passivated perovskite films.

(A) Chemical structures of b-PEI, PAA, and PVP. (B) Schematic diagram for incorporating the polymer into the perovskite films. SEM images of (C) homogeneous PbI2 film without slow growth, (D) mesoporous PbI2 film with slow growth, and (E) perovskite film processed from slowly grown PbI2 with PVP polymer. (F) FTIR spectra of perovskite, PVP, and perovskite/PVP films. a.u., arbitrary units. (G) FTIR spectra of perovskite, PAA, and perovskite/PAA films. (H) GDOES measurement of perovskite and perovskite/PAA films.

Table 1 Absorption energy and carrier dynamic properties of different functional group–modified perovskite films.

View this table:

Unfortunately, the strong interactions between the polymers and the perovskite precursor make it challenging to form a homogenous perovskite films with these polymers via the conventional one-step or two-step method. To resolve this problem, here, we use an interdiffusion method (50), as illustrated in Fig. 1B. The key to develop the polymer-coordinated perovskite films is the formation of a mesoporous PbI2 layer (Fig. 1D), which is created by a slow growth process (51). That is, the half-dried PbI2 (spin-coated for 20 s) substrates were kept in a glass petri dish for 15 min, during which the transparent PbI2 turned cloudy, consistent with the formation of a mesoporous structure, as has been previously reported. The polymers were then blended with the MAI in isopropanol (IPA) solution to form perovskite films with different polymers. We propose that during the perovskite crystallization step, induced by dripping the MAI/polymer solution onto the films, the micropores in the PbI2 films allow the infiltration of the polymers into the PbI2 film (Fig. 1E). After MAI/polymer dripping, the perovskite films containing polymers were formulated via thermal annealing. In the conventionally processed PbI2 films, the films show homogeneous texture, and most of the polymers are expected to segregate at the perovskite surface due to the “filter” effect of the pinhole-free homogeneous film (Fig. 1C), which might form a barrier for charge extraction.

The existence of polymers and the formation of different coordination bonds within the perovskite film can be probed by Fourier transform infrared spectroscopy (FTIR). The samples for FTIR measurement were prepared with the same procedure as the device fabrication but with a higher polymer content (10:1) to increase the signal. A NaCl substrate was used to hold the films for the measurement. The main spectral feature to be considered in the PVP film is the deformation modes (52) of the pyridine rings at 1597 cm−1, as shown in Fig. 1F. The same absorption feature is observed in the PVP-perovskite films, but the feature peak moves to 1604 cm−1, presumably due to the increased rigidity (53). This shift would be consistent with the formation of coordination bonds by sharing the lone electron pair on the N to the empty 6p orbit of Pb2+. A distinct absorption peak appeared at 1711 cm−1 for the intrinsic PAA film (53), and this peak is shifted to 1722 cm−1 due to coordination with the perovskite (Fig. 1G), where the lone electron pair on the carboxylic group is delocalized to the empty orbit of Pb2+. Because large amounts of ammonium groups exist in the perovskite film, it is very difficult to distinguish the feature absorptions of b-PEI and MA+ (fig. S2). However, the interaction between the perovskite and the b-PEI polymer should be strong according to the DFT calculations. The location of the polymers in the perovskite film was further examined with glow-discharge optical emission spectroscopy (GDOES) measurements, which is a technique used to measure the chemical composition as a function of depth with parts-per-million sensitivity (54). PAA was selected as a model due to the existence of O atoms, which should not be present in the MAPbI3 perovskite. Thereafter, the existence of PAA can be traced by probing the oxygen content in the perovskite film. As shown in Fig. 1H, the O signal intensity in the perovskite film with PAA is greater than the O baseline intensity observed in several bare perovskite films throughout the film. Although sample-to-sample variations in adventitious O2 and H2O could also account for the different O baseline of the PAA films, we neglected the possibility because we found that the increased O baseline far exceeded the sample-to-sample variation for similar batches of bare perovskite film processed and measured under identical conditions (fig. S3). These results confirm that the polymers were incorporated throughout the film with mesoporous PbI2. The film morphology of the perovskite film with and without different polymer modifications was also examined by scanning electron microscopy (SEM). As shown in fig. S4, all the perovskite films with and without polymers exhibit similar morphology, with large grain size around 700 nm and smooth surface without any pinholes. The unchanged crystallinity of perovskite films with and without polymers was confirmed by the x-ray diffraction (XRD) patterns, which are similar for the perovskite with and without different polymers (fig. S5). Therefore, we conclude that incorporation of polymers into the perovskite films via the interdiffusion method has little influence on the film morphology but that it is likely that the polymers are being incorporated at some level throughout the depth of the films.

HPSC devices were fabricated using the perovskite films with and without polymer incorporation. Figure 2A shows the device architectures. SnO2 (11) and 2,2′,7,7′-tetrakis(N,N-di-4-methoxyphenylamino)-9,9′-spirobifluorene (Spiro-OMeTAD) (6) were applied as the electron transport layer and hole transport layer (HTL), respectively. Cross-sectional images of the devices, shown in Fig. 2B, show a homogeneous pinhole-free 550-nm perovskite film fully covered by a 200-nm Spiro-OMeTAD layer. The current-voltage relation (I-V) characteristic curves of the HPSC devices with and without different polymers are shown in Fig. 2C, and the corresponding device parameters are summarized in Table 2. The best HPSC without passivation shows a high reverse-scanned PCE of 18.02% (averaged efficiency of 16.9 ± 0.7% from more than 70 devices), with a VOC of 1.09 V, a JSC of 21.82 mA/cm2, and a fill factor (FF) of 0.76. With PVP treatment, the champion HPSC shows a significant improvement in reverse-scanned PCE to 20.23% (averaged efficiency of 18.8 ± 0.8% from more than 90 devices), where the highest VOC is 1.16 V (averaged value of 1.14 ± 0.02 V), JSC is 21.74 mA/cm2, and FF is 0.81. This change represents a more than 10% improvement in PCE compared to the champion film without PVP; furthermore, the average device performance increased by 10%. Notably, the HPSC devices with PVP treatment reach a high VOC of 1.16 V (fig. S6), which is the highest VOC reported for MAPbI3-based HPSC we are aware of to date. The ultrahigh VOC enables an unprecedented low VOC deficit of 0.47 V for the MAPbI3 system, considering a band gap of 1.63 eV, as measured from electroabsorption spectroscopies [external quantum efficiency (EQE) response edge of ~795 nm (fig. S7)] (46). The low VOC deficit obtained from HPSC devices with PVP passivation suggests a significant suppression of nonradiative charge recombination, as will be discussed below. The influence of different amounts of PVP content on the device performance is also studied, and the I-V characteristic curves are shown in fig. S8. The best device performance is obtained when the PVP/MAI ratio is 1:1000 by weight %. Less PVP content will cause insufficient passivation, and excessive PVP can potentially cause charge transport problems. To investigate whether the improvement in PCE was due to the passivation in the bulk of the perovskite or only at the interfaces with HTL, we fabricated the HPSC device with a thin PVP polymer on top of the perovskite film. With PVP only present on top of the perovskite, the device performance showed only a small improvement in reverse-scanned PCE from 18.05% (averaged 16.9 ± 0.7%) to 18.61% (17.6 ± 0.5%, averaged from 15 devices) (fig. S9). This result implies that the incorporation of PVP into the bulk and/or its presence during the microporous film growth is critical to the observed performance enhancement.

Fig. 2 CH3NH3PbI3 solar cell devices with different polymer converted from microporous PbI2.

(A) Schematic diagram of the device architecture. (B) Cross-sectional image of the solar cells. ITO, indium tin oxide. (C) I-V characteristics, (D) EQE spectra, (E) steady-state output power, and (F) shelf time of the solar cells with and without PVP treatment.

Table 2 Device parameters of polymer-modified perovskite solar cells.

RS, reverse scan; FS, forward scan.

View this table:

The HPSCs also showed appreciable improvements with PAA treatment, where the highest reverse-scanned PCE of 19.64% (averaged value of 17.9 ± 0.7% from more than 50 devices) was achieved, with a JSC of 21.53 mA/cm2, a VOC of 1.15 V, and an FF of 0.79. In contrast, however, the HPSC device made with b-PEI treatment exhibited a significant drop in performance, with the highest reverse-scanned PCE dropping to 16.67% (averaged value of 15.8 ± 0.7%, from over 50 devices), with a JSC of 21.94 mA/cm2, a VOC of 1.07 V, and an FF of 0.71. Variation in device performance is also shown in the device histogram in fig. S10.

Figure 2D shows the external quantum efficiency of the HPSC with different polymers. The EQE profiles are similar for all the different films with the highest EQE of 93% at ~450 nm. The integrated JSC from the EQE using the AM 1.5 reference spectra reaches 21.5 mA/cm2, which is close to that from the I-V measurement (21.7 mA/cm2; PVP-modified HPSC). The I-V hysteresis problem is typically observed in planar heterojunction HPSC (54, 55), which casts doubt on the device efficiency obtained from I-V scan. The mechanism of I-V hysteresis is still under investigation, and it has been demonstrated that the hysteresis is most likely associated with the ionic nature of perovskite, such as ion migration (5658). In our case, the HPSC device performance deviation between the forward and backward scan is also evaluated. As shown in fig. S11, all the HPSC devices show the I-V hysteresis problem to different extents. The device parameters with different scan direction and different polymers are summarized in Table 2. The HPSC device with no polymers shows a forward-scanned PCE of 15.52%, retaining 86% of the reverse-scanned efficiency (18.05%), but enhanced to 18.80% with PVP passivation, retaining 92% of the reverse-scanned efficiency (20.23%). The forward-scanned efficiency of PAA- and b-PEI–modified perovskite solar cells retained 93 and 90% of that from the reverse scan, respectively. It is evident that the hysteresis problem is alleviated with incorporation of polymers.

The appearance of hysteresis makes it hard to judiciously evaluate the device performance simply through the I-V curve scan (5558). On the contrary, steady-state output power mimicking the real working condition of solar cells under a bias light and voltage is a more reliable method to characterize the authentic performance of HPSC (18). Figure 2E illustrates the steady-state output power of HPSCs under ambient conditions (20° to 25°C; relative humidity of 25 to 40%) under continuous AM 1.5-G, 1-sun illumination and a constant bias voltage. As shown, HPSCs without any polymer treatment show a maximum output power of 17.5 mW/cm2 (or stabilized PCE of 17.5%) under 0.88 V and degrades to 16.2 mW/cm2 after 30 min. The HPSC with PVP treatment exhibits a maximum output power of 19.1 mW/cm2 (or stabilized PCE of 19.1%) under 0.94 V, and little drop is observed after 30 min (18.9 mW/cm2). Note that the champion stabilized efficiency of 19.1% represents one of the highest among planar heterojunction MAPbI3 HPSCs (24, 25). These results verify the enhanced device performance and, moreover, the improved stability under working condition of an HPSC device with PVP passivation.

Furthermore, the shelf life stability of the HPSCs was also evaluated. The HPSC devices for shelf time stability tests were encapsulated by curable epoxy glue and a glass cover, stored in a glove box, and measured under ambient condition (20° to 25°C; relative humidity of 25 to 40%). As shown in Fig. 2F, the HPSC without the PVP surface passivation degrades to 59% of the initial efficiency after 2500 hours, whereas the HPSC with PVP passivation retains more than 85% of the original efficiency after 2500 hours. We speculate that because the polymers can bind multiple interfaces, the improved working stability and shelf time can be possibly attributed to cross-linking of the grains, as has been proposed for treatment with alkylphosphonic acid ω-ammonium chlorides (45). It could also be that the polymer simply helps prevent access of water to the perovskite or that it actually reduces the mobility of some of the perovskite components (44). In any event, the origin of the improved lifetime is beyond the scope of this study but should prove an interesting topic for future investigation.

Next, the carrier dynamics of the perovskite films were investigated to gain more insight into the mechanism of HPSC device performance changes with the different polymer treatments. It has been demonstrated that charge recombination via nonradiative recombination is the most significant energy loss channel at carrier densities comparable to 1-sun illumination conditions (23). Notably then, Fig. 3A shows that the PL intensity increased by ~700% with PVP and by ~650% with PAA but dropped by 70% with b-PEI. These changes provide strong evidence that the polymer treatments are altering the nonradiative decay rates in a manner consistent with the observed changes in device performance: PVP and PAA treatments both result in increased PL (reduced nonradiative losses) and increased device performance, whereas b-PEI results in a decrease in PL (increased nonradiative losses) and a concomitant decrease in device performance.

Fig. 3 Carrier dynamics.

(A) PL intensity spectra and (B) time-resolved PL (TRPL) decay spectra of perovskite films with different polymer incorporation. The fluence for the TRPL measurement is around 11 nJ/cm2, and the excitation density is around 1014 cm−3. Yellow line represents the fitted curve using the stretched exponential function. (C) TPV and (D) TPC of perovskite solar cells with different polymers.

In addition, we measured PL by exciting the samples from different sides, that is, the top surface of the perovskite film (front side) and the glass side (back side). As shown in fig. S12, the perovskite films processed from mesoporous PbI2 and compact PbI2 show similar PL intensity when excited from the top side and the glass side. For the perovskite/PVP films processed from compact PbI2 (where we would expect the polymer to reside predominantly at the film surface), the PL intensity is significantly greater when excited from the perovskite/PVP side compared to when excited from the glass side. On the other hand, the perovskite/PVP film processed from mesoporous PbI2 (where we hypothesize that the PVP will be able to penetrate the film thickness during growth) showed similar PL intensity when excited from either side. Because optical simulations show that excitation at 640 nm will be primarily absorbed by the first 250-nm perovskite layer, we interpret these results as evidence to support the hypothesis that the PVP treatment of the mesoporous PbI2 films results in a change in the electronic properties throughout the entire film, whereas the compact film is only affected at the interface with the PVP.

We also measured the transient PL response of the perovskite films with different polymers. The fluence for the TRPL measurement is around 11 nJ/cm2 with a frequency of 1 MHz, and the excitation density is around 1014 cm−3. The PL transient data further support our proposal that films treated with PVP and PAA show improved optoelectronic properties with reduced levels of nonradiative recombination. As shown in Fig. 3B, the PL decays were fitted with a stretched exponential equation (Embedded Image) (30, 41, 59, 60). The reference perovskite film without any polymer showed a characteristic PL lifetime of τc ~ 89 ± 2 ns and stretching exponent β ~ 0.86 ± 0.02, which is consistent with literature (59). With the polymer treatments, the PL transient decays exhibit longer lifetimes and less dispersion in decay rates for PVP and PAA, with τc ~ 275 ± 3 ns and β ~ 0.95 ± 0.02 for PVP and τc ~ 172 ± 1 ns and β ~ 0.93 ± 0.01 for PAA, and a reduction in lifetime and increased dispersion with b-PEI treatment, with τc ~ 8 ± 0.3 ns and β ~ 0.66 ± 0.01. We calculated the average lifetime ⟨τ⟩ of the stretched exponential distribution according to the equation Embedded Image. Following the equation, we report an averaged lifetime of 96, 281, 178, and 11 ns for control perovskite, perovskite/PVP, perovskite/PAA, and perovskite/b-PEI films, respectively. The increase in lifetime for PVP and PAA passivation indicates reduction in nonradiative recombination in the perovskite film, whereas the shortened lifetime for b-PEI suggests introduction of nonradiative recombination centers within the perovskite film, which is consistent with the PL intensity trends.

During the operation of a real HPSC device, photogenerated carriers are collected at electron and hole-collecting electrode interfaces, for example, Spiro-OMeTAD/perovskite (for holes) or SnO2/perovskite (for electrons), and there exists the possibility for additional nonradiative decay pathways introduced by the electrode/perovskite interfaces (61). The separated photocarriers can either recombine via surface trap states or be swept out of the device for electrical power generation. In this scenario, the techniques of TPV and TPC decay can provide additional information about carrier dynamics during device operation. A bias light with intensity of around 1 sun is applied to follow the kinetics at carrier densities approximating those from the working condition of HPSCs, and a laser pulse (pulse width of 4 ns and a repetition frequency of 10 Hz) is used to generate small disturbance on VOCVOC of ~5 meV). The decay of TPV traces the fates of photocarriers at a small perturbation above open-circuit conditions, where the separated photocarriers mainly recombine via defect or trap states. Thereafter, the TPV lifetime, taken as the time at which a small photovoltage decays to 1/e of its peak intensity, can be extracted to learn about the charge recombination rate. As shown in Fig. 3C, the TPV lifetime of HPSC without any polymer passivation is ~1.35 μs and increases to 1.95 and 2.69 μs with PAA and PVP passivation, respectively. However, the carrier lifetime reduces to 0.9 μs with b-PEI incorporation. The elongated carrier lifetime of HPSC with PVP and PAA passivation is again consistent with our proposal that treatment with these polymers during growth reduces the density of defect states in the film. The decay of TPC is recorded under short-circuit condition, where the photocarriers swept out of the device can quickly recombine via the external circuit. The decay of TPC illustrates the photocarrier transit time across the bulk perovskite and the electrode interfaces. As shown in Fig. 3D, the photocarrier transit time in HPSC without any polymer passivation is ~0.75 ± 0.02 μs, and little change is observed with PAA (0.75 ± 0.02 μs) or PVP (0.72 ± 0.01 μs) passivation. However, the transit time becomes longer with b-PEI modification (0.85 ± 0.03 μs). We attributed the elongated transit time to the increased charge traps in the b-PEI–modified perovskite films, which decrease the effective carrier mobility. The TPV/TPC results are again broadly consistent with the conclusions from the PL measurements and, in turn, with the trends observed in device performance variation with different polymer treatments.

Electronic impedance measurements (fig. S13) were carried out to determine the origin of the difference in the photovoltage and FF upon polymer treatment. As shown in Fig. 4 (A and C), under the same VOC, the devices with PAA and PVP demonstrated larger interfacial charge transfer resistances and longer lifetimes, being consistent with the photovoltaic performance for the two devices with higher VOC and FF (62, 63). This result has been confirmed by TPV decay measurements of the carrier lifetime shown in Fig. 3C. As shown in Fig. 4B, the capacitance C exhibits an exponential increase with VOC. Therefore, the interfacial trap density of states [DOS; Embedded Image, where Nt,0 is the total density of the localized state, and T0 has the unit of temperature and reflects the width of the trap distribution function below the edge of the conduction band (EC), a high value of T0 implying a broad DOS] distribution function can be plotted as shown in Fig. 4D. Using an exponential function to fit the capacitance curves in Fig. 4B gives T0 around 1649 K for the reference device, 1365 K for PAA-device, 1461 K for PVP device, and 1737 K for b-PEI device. The DOS in the range of ~1016 cm−3 indicates the high quality of the perovskite films investigated in this study. As shown in Fig. 4D, if we set a given DOS for four samples, then an increase of electron energy level can be observed in the order of b-PEI < reference < PVP < PAA. As schematically illustrated in Fig. 4D, a narrower DOS distribution suggests that PAA and PVP reduce the density of defects/trap states.

Fig. 4 Electronic IS measurements.

Plots for (A) charge transfer resistance (R) and (B) the correlated capacitance (C), (C) lifetime (τ), and (D) DOS versus electron energy level in various devices.


To summarize our experimental result, we demonstrate that the incorporation of polymer additives into the perovskite film is an effective strategy to improve both the device efficiency and the stability of perovskite solar cells. The mechanism regarding the device performance variations is attributed to the changes of trap states in the whole perovskite film with different polymers, as unveiled by the PL intensity, carrier dynamics, and impedance study. The variation in device performance and other optoelectronic properties is correlated to different interactions of the functional groups with perovskite (fig. S14), based on which possible rules guiding the passivating mechanism for perovskite is proposed. As shown, the perovskite solar cells show the best performance with the pyridine group, having a binding energy of 37.70 kJ/mol with Pb2+. With low coordination energy (carboxylic group, 15.71 kJ/mol), the passivation effect becomes relatively weak but still shows significant improvement compared to the reference. Thereafter, coordination bonds, especially those with large binding energy with the perovskite film (or Pb2+), would benefit the trap-state passivation. Defect states can possibly be passivated by chemical groups that donate electron density, likely by sharing lone electron pairs (41) presumably with the empty 6p orbitals of undercoordinated Pb2+ sites. The decrease in device performance with high–coordination energy b-PEI seemed to contradict with our conclusion. However, further analysis from proton nuclear magnetic resonance (1H NMR) measurements (fig. S15) revealed that the amino group on b-PEI would change to an ammonium group when blended with MAI due to the strong basicity of amino group (pKa of ~10; the pKa of pyridine is ~5). The protonation of b-PEI could potentially change the interactions from a coordination-type bonding (amino-Pb2+) to ionic bonding (CH3NH3+-I). Thereafter, although passivation effect is observed with amino group (64), the formation of ionic bonding might introduce trap states due to the electron-withdrawing effect of ammonium to I, which weakens the I–Pb bond. Further investigation is needed to find the detailed mechanism.

Another challenging issue with perovskite solar cells is their stability. By introducing a small cross-linking molecule, Li et al. (45) demonstrated high device performance with high stability. Ideally, a polymer cross-linker would be more effective in reinforcing the stability of the perovskite film. However, the strategy of using polymers for efficient passivation and crystal cross-linking is limited by processing because the strong interaction between the electron-donating groups and the perovskite precursors makes it difficult to form a smooth film. Therefore, only polymers with weak interactions with the perovskite have been incorporated in previous studies (65, 66). Using a mesoporous scaffold–facilitated interdiffusion method, we have successfully incorporated polymers with different functional groups into the films. The mesoporous PbI2 films were developed via slow growth, and the polymers along with MAI were interdiffused into the pores of the PbI2 film, forming a coordination bond with the perovskite crystals during crystallization. Above all, by incorporating the optimal coordinating polymer (PVP) into the perovskite film, the perovskite solar cells reverse-scanned device efficiency improves to 20.2% (18.8 ± 0.8%; stabilized efficiency of 19.1%; under bias of 0.94 V and AM 1.5-G, 1-sun illumination for 30 min) and the shelf time stability to more than 3 months (retaining 85% of the original efficiency).



The MAI iodide was home-synthesized, and the lead iodide was purchased from Alfa Aesar and was used as received. The Spiro-OMeTAD was purchased from Lumitech Corp. and was used as received. Other materials not specified were from Sigma-Aldrich.

Device fabrication

The patterned ITO glass substrates were cleaned using an ultrasonicator in an acetone, water, and IPA bath. A SnO2 layer was formed on the surface of the ITO substrate. SnCl2·2H2O was dissolved in ethanol (11 mg ml−1) in a vial and aged by vigorously stirring for 1 hour at room temperature and ambient atmosphere, leaving the vial uncapped. The SnCl2 solution was spin-coated twice on ITO glass at 3000 rpm for 30 s. A SnO2 thin film with an approximate thickness of 30 nm was formed after annealing the spin-coated film in ambient atmosphere at 150°C for 30 min and 180°C for 60 min. After cooling at room temperature, the glass/ITO/SnO2 substrates were transferred into an ultraviolet-ozone cleaner for 15 min and then left in a dry-air glove box. The perovskite films were deposited by a sequential two-step solution process. Hot PbI2 solution (600 mg/ml; in dimethylformamide) was spun onto the glass/ITO/SnO2 substrates at 3000 rpm for 40 s. For fabrication of mesoporous PbI2, the samples were taken out from the spin coater after 20 s and left in a petri dish for 15 min. After annealing on a 75°C hot plate for 15 min, the mesoporous PbI2 substrates were dripped with MAI/MACl solution (70:7 mg/ml; in IPA) by spin-coating to form the PbI2-MAI complex films. The complex films were put on a 135°C hot plate for 15 min. The Spiro-OMeTAD solution [80 mg/ml; in chlorobezene with 3% 4-tert-butylpyridine and 9 mg of lithium bis(trifluoromethanesulfonyl)imide] was spun onto the perovskite film as a hole conductor. The devices were completed by evaporating 80-nm gold in a vacuum chamber (base pressure, 5 × 10−4 Pa).

Device characterization

The device efficiencies were measured using a Keithley SourceMeter under AM 1.5-G, 1-sun irradiance, as generated by a thermo oriel solar simulator. The light intensity was adjusted by a KG-5 Si diode, which can be traced back to the National Renewable Energy Laboratory. The devices were measured in ambient air at temperature of 22°C. Different scan directions (reverse scan, from 1.2 to −0.2 V; forward scan, from 0 to 1.2 V) were recorded at a scan rate of 0.5 V/s for the hysteresis study. The device pixel had the dimension of 0.2 cm by 0.7 cm, and the accurate device area was defined by an aperture of 0.1031 cm2. The device photocurrent showed little variation with the aperture area. In addition, for the steady-state output measurement, the solar cells were put under the simulated AM 1.5-G, 1-sun illumination to record the photocurrent under the bias of 0.88 and 0.94 V. The EQE was measured on an Enlitech measurement system. The shelf lifetime of the devices was measured in ambient air (relative moisture, 20 to 40%; temperature, 20° to 25°C) and by storing them in a N2 glove box.

TPV decay measurements

A white light bias was generated from an array of diodes (Molex 180081-4320) to simulate a 1-sun working condition. A pulsed red dye laser (rhodamine 6G; 590 nm) pumped by a nitrogen laser (LSI VSL-337ND-S) was used as the perturbation source, with a pulse width of 4 ns and a repetition frequency of 10 Hz. The perturbation light intensity was attenuated to keep the amplitude of transient VOCVOC) below 5 mV so that ΔVOC << VOC. Voltage dynamics were recorded on a digital oscilloscope (Tektronix DPO4104B), and voltages at open circuit were measured over a megohm and a 50-ohm resistor, respectively.

PL and TRPL measurements

A Horiba Jobin Yvon system with a red-light source with an excitation at 640 nm was used to conduct PL measurements. TRPL spectra were obtained using a PicoHarp 300 with time-correlated single-photon counting capabilities. A picosecond diode laser provided excitation at a wavelength of 640 nm with a repetition frequency of 1 MHz (PDL 800B).

Scanning electron microscopy

An emission Nova 230 NanoSEM was used to obtain SEM images with an electron beam acceleration in the range of 500 V to 30 kV.

FTIR spectra

The FTIR spectra (4000 to 500 cm−1) were recorded on a Jasco FT/IR-6100 FTIR with samples prepared on silicon substrates.

Electronic IS measurement

IS measurements were performed using the Hewlett Packard 4284A Precision LCR Meter providing voltage modulation in the desired frequency range. The running program was edited with MATLAB R2003. The Z-view software (v2.8b, Scribner Associates Inc.) was used to analyze the impedance data. The IS experiments were performed at a constant temperature of 20°C under illumination with various intensities. The impedance spectra of the perovskite devices were recorded at potentials in the range of VOC ± 0.1 V at frequencies ranging from 20 Hz to 1 MHz, the oscillation potential amplitudes being adjusted to 30 mV. The ITO electrode was used as the working electrode, and the Au electrode (CE) was used as both the auxiliary electrode and the reference electrode.

DFT simulation

Our first-principles calculations were carried out using the Vienna ab initio simulation package (VASP) (67), a DFT approach using the projector-augmented wave method (68, 69). Calculations were performed using the Perdew-Burke-Ernzerhof version of the generalized gradient approximation (70) for exchange correlation. Surface slabs were modeled as (001)-terminated slabs of tetragonal structure. In our calculation, eight asymmetric interchanging MAI-PbI2 monolayers with 30 Å vacuum added on top of the slab surface were used. Dipole slab correction was applied to remove artificial dipole-dipole interaction across periodic images in VASP. To demonstrate the passivation effect of polymers of different functional groups, we concentrated on the PbI2-terminated surface because its surface energy is small (37). Adsorption energies were calculated usingEmbedded Image

GDOES measurement

Depth-dependent compositional profiles were collected with a Horiba GD-Profiler 2 using a high-power radio frequency argon plasma in a 4-mm-diameter anode. The plasma was operated at 30 W and a pressure of 450 Pa. Oxygen (O) and lead (Pb) were detected using the 130- and 406-nm atomic emission lines, respectively.


The 1H NMR spectra were recorded in solution of dimethyl sulfoxide–d6 on a Bruker DRX300 NMR spectrometer with tetramethylsilane as the internal standard.


Supplementary material for this article is available at

fig. S1. Interactions between the perovskite precursors ( PbI2 and MAI) and polymers.

fig. S2. FTIR spectra of perovskite (PVSK), polyethylenimine (PEI), and perovskite/polyethylenimine (PVSK + PEI) films.

fig. S3. GDOES (O element) of perovskite films processed and measured under the same condition.

fig. S4. Surface morphology of perovskite films.

fig. S5. XRD patterns of perovskite film with and without different polymers.

fig. S6. I-V characteristic curves of perovskite solar cells with PVP grain boundary passivation.

fig. S7. EQE spectra of perovskite solar cells versus photon energy.

fig. S8. I-V characteristic of perovskite solar cells with different amount of PVP.

fig. S9. I-V characteristic of perovskite solar cells (made from compact PbI2) with and without PVP.

fig. S10. Histogram of device efficiency distribution of perovskite solar cells with different polymers.

fig. S11. I-V hysteresis behavior of perovskite solar cells with and without different polymer passivation.

fig. S12. Depth-dependent PL behavior of perovskite film with or without PVP.

fig. S13. A typical Nyquist plot for a PVP device at an open-circuit voltage under 1-sun illumination in the frequency range of 20 Hz to 1 MHz.

fig. S14. Binding energy–dependent device performance.

fig. S15. Interactions between MAI and the polymers.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank C. Wang for the DFT calculation and X. Shi for 1H NMR measurement. Funding: This work was financially supported by a grant from the Office of Naval Research (grant no. N000141110250), University of California, Los Angeles (UCLA) Internal Funds, UC-Solar Program (fund number MRPI 328368), China Scholarship Council (No. 201506165038), and the Enli Tech (Taiwan) for donating the EQE measurement system to UCLA. Author contributions: L.Z. conceived the ideas, rationalized the experiments, performed the data analysis, and prepared the manuscript. H.G. conducted the device fabrication and measurement. S.D. measured the PL. R.D. and B.D. measured the FTIR spectroscopies. N.D. performed the SEM measurement. D.W.D., S.J., and D.S.G. helped with the GDOES measurement and data analysis on PL. M.W. carried out the C-V, TPV, and TPC measurements and analyzed the data. Y.Y. guided and supervised the whole study. All authors discussed the results and commented on the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article