Methylammonium-free, high-performance, and stable perovskite solar cells on a planar architecture

See allHide authors and affiliations

Science  26 Oct 2018:
Vol. 362, Issue 6413, pp. 449-453
DOI: 10.1126/science.aat3583

Staying in the black phase

Hybrid perovskite solar cells often use the more thermally stable formamidinium (FA) cation rather than methylammonium, but its larger size can create lattice distortion that results in an inactive yellow phase. Turren-Cruz et al. show that by using iodide instead of bromide as the anion (to create a redder bandgap) and an optical mix of cesium, rubidium, and FA cations, they can make solar cells with a stabilized efficiency of more than 20%. No heating steps above 100°C were needed to create the preferred black phase.

Science, this issue p. 449


Currently, perovskite solar cells (PSCs) with high performances greater than 20% contain bromine (Br), causing a suboptimal bandgap, and the thermally unstable methylammonium (MA) molecule. Avoiding Br and especially MA can therefore result in more optimal bandgaps and stable perovskites. We show that inorganic cation tuning, using rubidium and cesium, enables highly crystalline formamidinium-based perovskites without Br or MA. On a conventional, planar device architecture, using polymeric interlayers at the electron- and hole-transporting interface, we demonstrate an efficiency of 20.35% (stabilized), one of the highest for MA-free perovskites, with a drastically improved stability reached without the stabilizing influence of mesoporous interlayers. The perovskite is not heated beyond 100°C. Going MA-free is a new direction for perovskites that are inherently stable and compatible with tandems or flexible substrates, which are the main routes commercializing PSCs.

Perovskites with an ABX3 structure [A = cesium (Cs), methylammonium (MA), formamidinium (FA); B = lead (Pb), tin (Sn); X = chlorine (Cl), bromine (Br) , iodine (I)] have skyrocketed in recent years, with power conversion efficiencies (PCEs) now at 23.3%, which is close to established technologies such as gallium arsenide (GaAs), cadmium telluride (CdTe), and silicon (Si) (1). Currently, almost all high-performance perovskite solar cells (PSCs) with efficiencies >20%, including all reported world records, contain unstable MA (26).

One important trend to achieve PCEs >20% has been the usage of more complex perovskite compositions, ranging from double-cation (MAFA or CsFA) (79) and triple-cation (CsMAFA) (6) to Rb-modified perovskites (5, 1012). FA is used as the majority cation in almost all current high-efficiency PSCs because FA, compared with MA, is thermally more stable (8, 13) and has a more optimal, red-shifted bandgap because it is the largest organic cation to still fit into a three-dimensional perovskite (14). Unfortunately, the relatively large size of FA also induces a distorted lattice that results in a photoinactive “yellow phase” at room temperature, whereas the photoactive “black phase” can only be observed at elevated temperatures (8). Therefore, obtaining phase-stable, black-phase FA perovskites at room temperature has become the implicit or explicit objective of the perovskite research field. Avoiding yellow-phase impurities motivated the development of the above-mentioned optimized processing techniques and elaborate multication, multihalide mixtures. The addition of Br has been especially crucial to suppress yellow-phase impurities. Accordingly, Br often constitutes up to 20% of the precursor compositions within most high-performance PSCs (5, 6, 15, 16).

However, using the halide position to improve crystallization comes at a cost: Br blueshifts the bandgap disproportionately. We illustrate this point in Fig. 1A, where we show the Shockley-Queisser limit for the theoretical, maximum PCE and short-circuit current (JSC) as a function of the bandgap (Eg) (17). By interchanging Br with I in MAPbIxBr1–x compounds, a bandgap range from 1.58 eV (MAPbI3) to 2.28 eV (MAPbBr3), spanning 700 meV, is achieved (18).

Fig. 1 Disproportionate bandgap tuning and cation stability.

(A) Shockley-Queisser limit for the theoretical, maximum PCE and short-circuit current (JSC) as a function of the bandgap (Eg). A wide bandgap range is spanned from 1.48 eV (FAPbI3) to 1.58 eV (MAPbI3) to 2.28 eV (MAPbBr3). The cation position (FA to MA) shifts the bandgap only moderately by 0.1 eV, whereas the halide position (I to Br) incurs a disproportionate “blue penalty,” shifting the bandgap by 0.7 eV. The red open circle is the newly developed, non-MA, non-Br compound RbCsFAPbI3, which is closer to more optimal bandgaps, whereas RbCsMAFAPb(Br17I83)3 (blue open circle) (5), one of the currently highest-performing compositions, has a suboptimal bandgap and contains unstable MA. (B) Illustration of the volatile nature of the MA molecule. One reported degradation pathway is CH3NH3I into CH3I and NH3 (the I atom is omitted for simplicity) at temperatures as low as 80°C, as reported in (23). (C) By contrast, Rb, Cs, and FA are thermally more stable cations and therefore preferable for long-term stability.

In stark contrast, the cation interacts only moderately with the metal-halide cage and accordingly has a relatively small bandgap shift of only 100 meV from MAPbI3 (1.58 eV) to FAPbI3 (1.48 eV) (18). The instability of MAPbI3 is reported by various works that show film degradation because of degassing MA (13, 1922). In Fig. 1B, we illustrate the volatile nature of the MA molecule itself using the reported degradation pathway of CH3NH3I into CH3I and NH3 at temperatures as low as 80°C (the I atom is omitted in the figure for simplicity) (23). Therefore, MA remains a principled risk factor for long-term stability and should thus be avoided (although the exact time scales of MA degradation require further research). To substantiate this point further, supplementary text 1 (together with figs. S1 to S4) shows the XRD and respective ultraviolet–visible (UV-vis) spectra before and after annealing at 130°C for 3 hours of MA-containing perovskite films—MAPbI3, Cs5MAFAPbI3, MAFAPbI3, and Rb5Cs5MAFAPbI3—and non-MA perovskite films—FAPbI3, Rb5FAPbI3, Cs10FAPbI3, and Rb5Cs10FAPbI3. All MA-based perovskites exhibit strong degradation, as evidenced by the growth of the PbI2 peak at 12°, a behavior which does not occur for the non-MA perovskites. Although thin films are not full devices, they nevertheless provide a clear motivation to avoid MA in all future compositions, ensuring perovskites with intrinsic long-term stability that is required for decade-long stable solar cells. Industry cannot afford the long-term risk factor of MA. However, currently almost all reported PCEs >20%, including all those setting world records, use MA (26), whereas perovskites without MA currently have PCEs often below 20% and frequently use high-temperature steps (2, 5, 24).

Thus, a more ideal perovskite compound ought to avoid Br because of the “blue penalty” and MA for stability reasons. This leaves I as the preferred halide and the thermally more stable Rb, Cs, and FA as the preferred cations (Fig. 1C), rendering RbCsFAPbI3 perovskites particularly relevant. Hence, in this work, we explored RbxCsyFA(100–xy)PbI3 compositions. We discovered an optimized RbCsFAPbI3 perovskite (without MA and Br) with a bandgap of 1.53 eV, close to the single-junction optimum, with high short-circuit currents comparing favorably with the previously used RbCsMAFAPb(Br17I83)3 perovskite (Fig. 1A, red and blue open circles). We achieved a planar PSC with a short-circuit current of 25.06 mA cm−2 and a high PCE of 20.44% (stabilized at 20.35%) for planar PSCs, which is among the highest non-MA–containing results thus far reported. The perovskites do not require any heating beyond 100°C, rendering them compatible with perovskite/Si tandem solar cells or flexible solar cells, which are among the most attractive pathways to commercialization.

We provide x-ray diffraction (XRD), photoluminescence (PL), and UV-vis absorption data for RbxCsyFA(100–xy)PbI3 perovskite films (on a planar substrate), written for convenience as RbxCsyFAI going forward (x and y are in percentages throughout). This nomenclature refers to the precursor solutions as outlined in the supplementary materials.

We started our investigation with the double-cation compounds CsyFA(100–y)I and RbxFA(100–x)I. In supplementary text 2 (together with figs. S5 to S8), the XRD data for CsyFA(100–y)I (figs. S5 and S6) show the characteristic perovskite peak at 14° that shifts as more Cs is added. This corresponds to a blue-shift in the PL and UV-vis data as shown in Fig. 2A and as previously reported (8, 25). For RbxFA(100–x)I, the perovskite peak in the XRD data does not shift as strongly as for the Cs series (figs. S7 and S8). This is consistent with the PL and UV-vis data (Fig. 2B), in which only a very slight blue-shift occurs for small Rb quantities (Fig. 2B). As the amount of Rb is increased, only a very small blue-shift in the PL occurs. Albeit for Rb15FA85I, a more noticeable blue-shift and broadening occur in the PL, correlating with the occurrence of a second phase in the XRD at 10° (fig. S7). In addition, in contrast to CsFA, the RbFAPbI3 films require annealing at 150°C, close to the temperature needed to convert pure FAPbI3 films. This is consistent with previous work (5, 11), confirming that there is no “black phase” for RbPbI3.

Fig. 2 Inorganic cation tuning: UV-vis, PL, SEM, and XRD characterization.

(A and B) PL and UV-vis data for the (A) CsyFA(100–y)PbI3 (x = 0, 1, 5, 10, or 15%) and the (B) RbxFA(100–x)PbI3 (y = 0, 1, 5, 10, or 15%) series. (C) Corresponding top-view SEM images. Scale bar, 200 nm. (D) Corresponding cross-sectional SEM images for full devices. Scale bar, 200 nm. The last image on the right is the Rb5Cs10PbI3 composition that results later in the highest device results. (E) XRD data and (F) PL and UV-vis data for the RbxCsyFA(100–xy)PbI3 series.

From scanning electron microscopy (SEM) top-view images in Fig. 2C, we observed that CsyFA(100–y)I films, starting with 5% Cs, have a relatively ordered film morphology with large grains. By contrast, Rb, likely because of the lacking integration, has a less ordered film morphology, resembling pure FAPbI3 films. In Fig. 2C, far right, Rb5Cs10FAI, has a regular film morphology. We display specifically Rb5Cs10FAI because it shows the highest and most reproducible device results.

We fabricated full devices on a planar stack of glass/FTO/SnO2/perovskite/spiro/Au as reported previously (26). In Fig. 2D, cross-sectional SEM images are presented. The pure FAPbI3 film is highly irregular and rough. For the CsFA series, starting with Cs5FA95PbI3, the grains become monolithic throughout the film, reaching from the top to the bottom contact. For the RbFA series, for all Rb concentrations, the films resemble FAPbI3, which is not beneficial for the perovskite/hole transporter materials (HTM) interface. The Rb5Cs10FAI compound has particularly monolithic, large grains.

Last, in Fig. 2, E and F, we present XRD, PL, and UV-vis data for the Rb5CsyFA(95–y)I series. The most blue-shifted composition is for 15% Cs followed, surprisingly, by 3% Cs. Upon adding 5 and 10% Cs, the PL remains relatively close to that of pure FA-based perovskite. This is consistent with Rb and Cs interacting with one another. We posit Cs could be aiding Rb to modify the lattice more effectively, but it is also possible that for certain Rb/Cs ratios, RbCsPbI3 perovskites form (in Fig. 2E, shown is the occurrence of an additional phase at 10° for 10 and 15% Cs). These inorganic compounds could “lock up” excess I-halides and help form a passivation layer for the remaining, pure FAPbI3 perovskite that has a more optimal single-junction bandgap. This is most noteworthy for the Rb5Cs10FAI compound, which has a PL emission very close to FAPbI3, in contrast to the non–Rb-containing Cs10FAI (Fig. 2A).

From device data in supplementary text 3 (together with figs. S9 to S16), we observed relatively high-performing CsFA (figs. S9 and S11) compounds that are not matched by the analogous RbFA series (figs. S10 and S12).

The double-cation CsFA films, possibly because of the large mismatch between the small Cs and large FA, do not yield highly reproducible PSCs. Although CsFA can occasionally reach high PCEs of 19.23% (fig. S9), we show in fig. S14A that the statistical baseline is lower (5). The Rb-modification again yields higher-performing perovskites with PCEs toward 19.52% for Rb5Cs10FAI (the RbCsFAI series is provided in fig. S13). In fig. S14, we show that the best-performing Cs10FAI perovskite improves upon 5% Rb addition in terms of absolute efficiencies and especially reproducibility (an often neglected metric).

From supplementary text 3, we identified Rb5Cs10FAPbI3 as a promising composition (figs. S13 and S14) and therefore continued with this composition for the remainder of this study. We successfully red-shifted the bandgap as motivated in Fig. 1. This can be observed clearly in the external quantum efficiency (EQE) measurements in figs. S15 and S16, where the EQE follows the PL trends. Consequently, we measured a high short-circuit current density of 24.52 mA cm−2, as confirmed with EQE data that show an integrated JSC of 24.01 mA cm−2 (fig. S16B), highlighting the inorganic cation tuning as a versatile tool to achieve more optimized, MA-free bandgaps.

The planar device architecture was used because it is inherently more compatible with flexible and tandem applications (none of the processing steps exceeds 100°C). Currently, planar PSCs show relatively high PCEs >20% and up to 21.4% (all using MA) (2729), which is slightly lower than mesoporous-containing perovskites at 22.1% (also using unstable MA) (4). One reason for this is that planar PSCs are prone to exhibit shunting pathways between the highly monolithic perovskite grains (30). Accordingly, one of the main parameters lacking for planar PSCs is the fill factor (FF) (supplementary text 2). In planar PSCs, any pinhole in the perovskite film results in a shunt-path that enables a direct pathway between electron and hole transporter. By contrast, an additional meso-layer avoids direct contact between the electron and hole transporter, preventing pinholes from becoming detrimental shunt paths. Consequently, for planar PSCs, the perovskite layer itself needs to be formed nearly perfectly, and thus the processing window for planar PSCs is much narrower than for meso PSCs. In addition, we posit that the formation of pinholes is less likely with a meso-layer because of a more uniformly surface wettability for the liquid precursor in contrast to planar PSCs, for which surface wettability is a constant concern. Thus, an imperfect perovskite layer can be saved by the meso interlayer. In other words, meso PSCs are more forgiving of imperfections. Consequently, because of the challenging fabrication, planar PSCs have remained less efficient than meso PSCs. This results in fewer groups that specialize on planar PSCs, which in turn results in fewer reported results. At the moment, this is a vicious circle: Because planar PSCs do not reach world records yet, groups are less willing to switch from meso to planar, even though it is industrially more preferable (31). To change this, we aimed for high-performance, MA-free perovskites on a planar architecture, requiring modifications of the planar architecture.

Recently, much progress has been made in improving perovskite interfaces through polymeric buffer layers that help to passivate the perovskite, reduce recombination, and mitigate the negative impact from pinholes as well as block metal electrode migration at elevated temperatures (5, 32). This approach is used both for the electron-transporting layer (ETL)/perovskite and the HTM/perovskite interface—for example, C60-SAM (33), phenyl-C61-butyric acid methyl ester (PCBM) (34), and poly(methyl methacrylate) (PMMA) (32). As another example, mixing PCBM and PMMA was shown to benefit the ETL/perovskite interface (35). This method, however, still uses an undesirable high-temperature titanium dioxide (TiO2) mesoporous layer. Therefore, we explored the potential of polymeric buffer layers for a planar ETL/perovskite interface using our optimized Rb5Cs10FAI compound (supplementary text 4 and figs. S17 to S19). In fig. S17, we explored different concentrations of PCBM at the SnO2/perovskite interface, realizing that no improvement was achieved compared with the control device without polymeric layers. We then investigated different concentrations of a mixed polymeric interlayer of PCBM:PMMA at the SnO2/perovskite interface (fig. S18). For planar PSCs, we found that PCBM:PMMA with a 5:1 ratio increases the average FF with values up to 77% (when compared with the control device without polymer-modification). The reproducibility, a very critical parameter for planar devices, is also improved, as indicated by the lowered standard deviation.

Encouraged by this, we investigated the top interface between the HTL and the perovskite using a PMMA layer on devices that already have a PCBM:PMMA polymer mix at the ETL interface. We show the results for the double-polymer–modified, planar PSCs in fig. S19. The additional PMMA layer improves the device architecture further. Perovskite films with and without a PMMA layer were characterized further by using atomic force microscopy (fig. S20) and contact angle measurements (fig. S21). From these measurements, we can conclude that the surface roughness is reduced and the contact angle increased when using PMMA, indicating that the polymer layer indeed modified the interface.

Whereas the devices with a 10:0.1 PCBM:PMMA show some of the highest PCE, the 5:1 PCBM:PMMA ratio exhibits more reproducible device parameters, especially in the FF. Thus, we used double-polymer–modified PSCs (pm-PSCs) with a 5:1 PCBM:PMAA mix at the ETL/perovskite and a PMMA layer between the perovskite/HTM as the most promising system for high performances and reproducibility.

In figs. S19 and S14B, we evaluated statistically the potential of the FTO/SnO2/PCBM:PMMA/perovskite/PMMA/HTM stack. We observed in fig. S19 a similar average JSC at 24.67 mA cm−2 (control) and 24.51 mA cm−2 (polymer-modified), a similar open-circuit voltage (VOC) from 1095 to 1083 mV, and an increase of the average FF from 68 to 74%, translating into an improved average PCE from 18.20 to 19.71%. The polymer-modification improves reproducibility, which is especially noticeable in the FF.

In Fig. 3A, we show the champion device for pm-PSCs reaching a high stabilized power output of 20.35% [the current-voltage (J-V) parameters are provided in table S1]. The JSC of 25.06 mA cm−2 is in good agreement with the integrated EQE value of 24.48 mA cm−2 (Fig. 3B). (In addition, figs. S22 and S23 display a 0.5 cm2 area device and J-V parameters for differently sized masks, respectively). In Fig. 3C, we show the SEM image of a representative Rb5Cs10FAPbI3 polymer-modified device. Moreover, in Fig. 3E, we show a Tauc plot with a bandgap of 1.53 eV for Rb5Cs10FAPbI3, confirming once more that the inorganic cation tuning achieved the original goal of red-shifting the bandgap.

Fig. 3 Best-performing device, J-V, EQE, SEM, and stability.

(A) J-V curve with MPP of a Rb5Cs10FAPbI3 device with polymer layers composed of PCBM/PMMA at the SnO2/perovskite interface and PMMA at the perovskite/HTM interface. The planar device displays a stabilized power output of 20.35%. (B and C) Corresponding (B) EQE and (C) SEM image. (D) Stability for the Cs10Rb5FAPbI3 device without polymer layers (green curve) and with polymer-modification (blue curve) aged at room temperature after 1000 hours of continuous MPP tracking in a nitrogen atmosphere. (E) Tauc plot resulting in a 1.53-eV bandgap for the optimized Rb5Cs10FAPbI3 perovskite.

Stability is one of the most important obstacles remaining for PSCs to become attractive for industry. However, relatively few stability datasets exist for planar devices.

We therefore measured stability as outlined in previous works (5, 27, 36). We used the Rb5Cs10FAPbI3 composition with and without the polymer modification, applying aging at room temperature for 1000 hours of continuous maximum power point (MPP) tracking in a nitrogen atmosphere. As shown in Fig. 3D, without polymeric buffer layers, the loss is relatively high (starting at 19.54% and finishing at 15.31% PCE). In stark contrast, the polymer-modified device has a only small loss. The absolute PCE starts at 18.74%, drops quickly to 17.63% after 35 hours (sometimes compared with a “burn-in”), and then exhibits a long-term component that stays relatively constant, finishing at 17.51% PCE after 1000 hours. Thus, after the initial component, the efficiency loss is less than 2 relative %, which is among the best-reported stability data yet for planar PSCs (without using MA). The stability of the Rb5Cs10FAPbI3 compound is confirmed in figs. S24 and S25, where additional devices, including mesoporous (figs. S25 and S26), were aged under the same conditions for 500 hours. This stability data highlights the importance of the device architecture for the long-term stability of especially planar PSCs. This implies also that the mesoporous titania interlayer in the current world-record architectures not only contributes to higher performances but also higher stabilities. We posit that the higher stability is due to the meso-layer acting as a mechanic barrier against external degradation factors (such as moisture, vapors, and a diffusing metal electrode) as well as the challenges for planar PSCs to deposit a nearly perfect, compact perovskite absorber, which is needed for high performances but also to protect against the mentioned external degradation factors. Thus, stabilizing planar architectures, without the help from protective mesoporous interlayers, is one of the key challenges for perovskite research. Therefore, the contacts, such as polymeric buffer layers, will play an especially important role in fortifying planar PSCs further in the future, as clearly evidenced in Fig. 3D (37).

More generally, using only inorganic additives to phase-stabilize the thermally relatively stable FA perovskites is a compositional design strategy that may be used to stabilize intermediate bandgaps as well, which is highly attractive for tandem applications. The prospect of a perovskite/Si tandem especially has inspired much progress in the field. However, Si solar cells are stable over many decades, and therefore, any unstable perovskite component, such as MA, must be avoided if PSCs were to be combined with Si. Our work provides a direct pathway for circumventing many of the drawbacks so far observed in MA- and Br-containing perovskites. Future strategies for MA-free perovskites can also include inorganic cation tuning using combinations of K, Rb, and Cs with FA (38, 39).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S26

Table S1

Reference (40)

References and Notes

Acknowledgments: S.-H.T.-C. and M.S. thank the Adolphe Merkle Foundation for support. S.-H.T.-C. thanks CONACYT-México for support. S.-H.T.-C. and M.S. thank the Adolphe Merkle Foundation for support. A.H. thanks the Swiss National Science Foundation for financial support with the project entitled “Fundamental studies of mesoscopic devices for solar energy conversion” with project number 200021_157135/1. The authors thank E. Ochoa Martinez and J. A. Borrego Perez for technical assistance. Funding: This work was partially funded by the Adolphe Merkle Foundation (M.S.). Author contributions: M.S. conceived, designed, and led the overall project; S.-H.T.-C. fabricated all samples and devices and conducted PL, UV-vis, contact angle, and long-term aging experiments on the perovskite films and devices; M.S. conducted SEM measurements; and S.-H.T.-C. and M.S. conducted XRD measurements on the perovskite films. All authors analyzed their data. A.H. participated in the supervision of the work; M.S. directed and supervised the research; M.S. wrote the first draft of the paper; and all authors contributed to the discussion, commented, reviewed, and approved of the paper. Competing interests: None declared. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article