Improving the Stability and Efficiency of Perovskite Solar Cells by a Bidentate Anilinium Salt.

We report improvement in the perovskite solar cell efficiency and stability after passivation with an organic molecule decorated with two anilinium cations. We compare this salt with its neutral analog and found that the change in the electron density distribution upon protonation and the presence of the halide anion are key to explaining the better passivation ability of the salt. In addition, we show that the counteranion has a significant impact on the performance of the device.

Hybrid organic–inorganic perovskites (HOIP) are promising materials for fabricating efficient solar cells because of the high optical absorption coefficient, direct bandgap, long lifetime, diffusion length, and mobility of the charge carriers.[1] HOIP adopts a cubic-like structure (Figure ) with the general formula ABX3, where A is a monovalent organic/inorganic cation, e.g., formamidinium (FA+), methylammonium (MA+), and cesium (Cs+), B is a divalent metal (e.g., Pb2+ and Sn2+), and X a halide anion (I–, Br–, Cl–). One of the outstanding properties of these materials is their defect tolerant ability.[2] This is because, in the bulk, most of the low-formation energy defects form shallow levels near the band edges,[3,4] becoming populated because of thermal energy. Therefore, those intrinsic defects are not critical to impact the solar cell performance. Nevertheless, the solution-processed methods used to fabricate perovskite films result in dangling bonds on the surface and in a polycrystalline material with grain boundaries (GB) over the film. These two types of defects are responsible for decreasing the device’s efficiency: From one side, the dangling bonds increase the density of trap states on the surface,[5] in addition to work as a nonradiative recombination center for the photogenerated charge carriers, thereby reducing the open-circuit voltage (VOC) of the perovskite solar cell (PSC).[6−8] On the other hand, GB can facilitate the moisture and oxygen permeation,[9] act as nonradiative recombination centers,[10−12] and cause hysteresis behavior of the current density vs. voltage (J–V) curve.[13] In this context, it is clear that the surface is a critical point of PSC performance and stability. For this reason, interface engineering strategies are an important key. An approach to obtaining the desired passivating effect lies in the post-treatment of the film using conjugated organic molecules.[14−16] Playing with the Lewis acid and base chemistry of the functional groups, it is possible to passivate positively and negatively charged defects.[2,17−21]

Figure 1

Schematic illustration of the molecules designed and their interaction with an ABX3 perovskite (in the scheme, A, B, and C cations are the blue, red, and green spheres, respectively).

Also, there is an explicit concern about the PSC production under sustainable and cost-effective conditions.[22−25] An approach to reach these goals is the replacement of the toxic halogenated solvents used as antisolvent to prepare perovskite film. Among the strategies, we highlight the development of gas quenching methods,[26−29] which triggers the perovskite crystallization by blowing an inert gas on the film. Combining this innovative method with postpassivation strategies, it is possible to obtain efficient, stable, and more sustainable devices.

Herein, we report the synthesis of a newly designed organic molecule, the 3,3′-((9H-fluoren-9-ylidene)methylene)dibenze-naminium (FM-NHX, Figure ), decorated with two anilinium cations, designed to passivate FAPbI3-based PSCs produced by the gas quenching method. When compared to the neutral molecule analog, the salt presents a superior passivation effect, leading to efficiencies of ∼19% and stability for more than 90 days.

To design our passivating molecules (Figure ), we were inspired by two commonly used structural moieties: fluorene (1) and aniline (2). The former is known for its planar conjugated backbone and electron donation properties.[30] The latter is a Lewis base, able to interact with Lewis acid defects on the perovskite. We coupled the aniline with the fluorene using the Corey–Fuchs reaction,[31] yielding the FM-NH(Figure ). This molecule presents clear characteristics of aggregation-induced emission (AIE) luminogen: a feature that can be explored in the passivation of PSC. Further, the protonation of FM-NH with hydrogen halides (HX, where X is Cl, Br, or I) yielded anilinium salts (FM-NHX). Sections 1 and 3 in the Supporting Information show the detailed experimental procedure and the full characterization, respectively, of the molecules synthesized. Note that the amino group in our molecules is attached directly to the conjugated backbone. As a result, the electron-donating and hydrogen-bonding (HB) ability of the −NH2 group is weakened compared to alkylamine molecules because of the resonance effect.[32] A weaker HB was reported to benefit the defect passivation of perovskite films.[32] In addition, annilinium-based salts were found to improve the physical contact between the perovskite and the hole transport material (HTM), in the case of nip-type PSC.[33] In both neutral and protonated molecules, the formation of HB with iodide is expected to heal ionic vacancies in the film.[34,35]

We investigated the efficiency of the passivation effect of FM-NH in the FAPbI3-based films. To this end, we assembled PSCs with an architecture FTO/SnO2/Cs0.05FA0.95Pb(I0.95Br0.05)3/modification/Spiro-OMeTAD/Au. Scanning electron microscopy (SEM) images reveal the formation of an amorphous film of FM-NH coating the perovskite (Figure S14). Despite this, we observed that the perovskite passivation with FM-NH in a concentration range from 4 to 16 mM does not cause any impact on the solar cell performance (Figure S15a). However, the changes in the contact angle of the film denote a tendency of the neutral molecule to decrease moisture permeation (Figure S15b). Such improvement can be explained in terms of the orientation of the molecules: From our Density Functional Theory (DFT) calculations, the fluorene moiety on FM-NH points outward to the surface, creating hydrophobic protection, whereas the polar −NH groups interact with perovskite directly (Section 6 in the Supporting Information).

With the aim of improving the efficiency while keeping similar hydrophobic protection, we protonated the neutral molecule with HX, yielding a class of FM-NHX salts. These salts were used to passivate the perovskite film in the same way as the neutral molecule. The prepared devices showed reduced hysteresis (Table S9); hence, to simplify the discussion, we present only the results of the backward scan measurements (Figure a, b). In general, the post passivation with the salts increases the efficiency of the PSCs, as shown in Figure a. A table summarizing the device parameters obtained after 10 days of stabilization is shown in Section 9 in the Supporting Information. The main improvements were observed in the VOC, which we attribute to healing of the surface defects. Also, the fill factor (FF) increases slightly (about 2%, on average), which can be assigned to a smoother interface with the HTM. In particular, we note that the VOC improvements are dependent on the halide nature, with the highest average values found for bromine- and iodide-based salts. Such a dependence is also evident in the time-resolved photoluminescence (TRPL) measurements: from the control to FM-NHBr and FM-NHI, the TRPL increases from 591 to 601 and 1125.4 ms, respectively. Those results indicate that the halide anion impacts the defect passivation.[36] In the literature, bromide and iodide have been shown to cause band bending at the interface[37] and the passivation of halide vacancies,[38] respectively, which can explain the VOC and charge carrier lifetime improvements obtained in our work. As a result of the passivation, the highest average efficiency (18.36%) was achieved for the salt with iodide, followed by chloride (17.91%), and bromide (17.37%). The average efficiency of the control device was equal to 17.15%. Our champion device, obtained after FM-NHI passivation, shows VOC, FF, short-circuiting current (JSC), and efficiency equals 1.07 V, 23.51 mA cm–2, 77.82%, and 19.35% (Figure b), respectively.

Figure 2

(a) Statistics of the device performance with the post-treatment with the FM-NHX salts. (b) J–V curve of the champion device efficiency. (c) Time-resolved photoluminescence (TRPL) decay curves of the perovskite film modified with FM-NHX onto a glass substrate.

To understand how the salts modify the surface, we characterized the films by X-ray diffraction (XRD) and Fourier transform infrared (FTIR) spectroscopy (Figures S16 and S17). From XRD, we cannot find any significant difference between the control and passivated films. On the other hand, FTIR shows the emergence of a band between 2800 and 3000 cm –1, which can be ascribed to the aromatic C–H stretching of the salt. Because the X-ray can penetrate deeply into the film, our results revealed that the modifications affect only the surface, with no (or undetectable) impact on the bulk of the perovskite.

Further, we used X-ray photoemission spectroscopy (XPS) to probe the surface modifications and investigate the interaction between the perovskite and the passivating molecules. From the C 1s high-resolution spectra of the pristine film, we can see an intense peak at ∼288.5 eV corresponding to the oxidation of the organic component of the perovskite structure (FA molecules) by oxygen and/or moisture (Figure S18a).[15] Interestingly, after the passivation with the FM-NHX, the intensity of this peak is highly suppressed, indicating that the organic salts help to decrease the film degradation.

Also, XPS can give insights into toward the formation/passivation of iodide vacancies (VI). This type of defect is very detrimental to PSC’s long-term efficiency because it triggers the degradation of the perovskite in several ways. For instance, the formation of superoxide species (O2 + e– → O2–) is facilitated by VI, which causes irreversible perovskite degradation.[39] In addition, iodide vacancies facilitate moisture permeation by solvating the Pb species.[40] To study VI by XPS, we need to look at the Pb 4f high-resolution spectra. Once VI is present, the electron left by the released I atom (2I– → I2(g) + 2e–) is mainly located at the vacant site,[41] allowing it to reduce the uncoordinated Pb2+ ion to yield metallic lead (Pb2+ + 2e– → Pb0). The formation of Pb0 results in a small shoulder close to the main peaks of Pb 4f7/2 and Pb 4f5/2 in the XPS spectra.[42] In our case, it was possible to identify those shoulders only in the case of the pristine and FM-NH-treated films at an energy equal to 137.8 and 142.6 eV (Figure S18c).[43,44] However, in the cause of neutral molecule, the area of the Pb0-related peaks decreased by 32% regarding the pristine film, suggesting a passivating effect conferred by FM-NH due to both the lone pairs on the nitrogen atom and the electron cloud over the molecule. The main changes are seen upon salt treatment, in which the intensity of Pb0 signals is highly suppressed, meaning that these molecules retard the formation of metallic lead. Because it is one of the products generated by VI formation the absence of the signal corresponding to Pb0 suggests the passivation of iodide vacancy.[45] Note that only the salts can effectively prevent the formation of Pb0. This can be explained by the presence of the halide counteranion and by the nature of the molecular orbital distribution, which changes significantly upon protonation (Figure S11). Both of these effects allow the salt to have an electron-donation behavior, which can lead to the passivation of electron acceptor-type of defects, e.g., VI.[42,46] Further evidence of this passivating effect is seen by the shift of Pb 4f peaks toward lower energy, suggesting an increased electron density over the Pb2+ conferred by the salt molecules. The mitigation of VI is one of the reasons behind the improved VOC of the salt-treated device.

The surface passivation is also expected to impact the morphology of the perovskite film. Using atomic force microscopy (AFM) measurements (Figure a), we verified that the average roughness (Ra) changes from 67 nm for the control to 52, 58, and 55 nm for FM-NHX, with X = Cl, Br, and I, respectively, suggesting a smoother and more uniform surface. Further, SEM measurements revealed that the grain size of our perovskite films is larger than 1 μm, even for the control sample, which is a reflection of their good quality (Figure b). In addition, SEM shows a less defined grain boundary upon passivation, which can be attributed to the filling of the intergrain voids by the molecules because of its soft molecular backbone. In principle, a smooth surface could benefit the contact between the perovskite layer and the HTM (in our case, Spiro-OMeTAD).[47,48]

Figure 3

(a) AFM and KPFM merged images, (b) SEM, and (c) contact angle measurements of the perovskite film treated with FM-NHX (where X = Cl, Br, I). (d) Evolution of the solar cell parameters upon aging. The devices were stored in an N2-filled box, with humidity <30%. The VCPD in panel a was obtained as an average between the 50 000 pixels of the image. The scale bar in panel b is 1 μm.

Concomitantly with the AFM measurement, we performed Kelvin probe force microscopy (KFPM) to map the surface potential. Briefly, in KPFM, we measured the contact potential difference (VCPD) between the sample and the AFM tip. This quantity can be correlated with the Fermi energy (EF) of the sample and the tip as follows: VCPD ∝ EF, sample – EF, tip. Hence, it is possible to probe how the EF of the surface is affected by the interaction with the molecules. A more detailed discussion about the KPFM can be found in Section 13 in the Supporting Information. The KPFM maps (Figure a) show that the average VCPD changes from 520 mV for the pristine film to 447, 488, and 553 mV for FM-NHX, with X = Cl, Br, and I, respectively, revealing the impact of the counteranion in the modulation of the surface EF. At a first approximation, we can conclude that the EF of the films modified with FM-NHCl and FM-NHBr are shifted upward compared to the pristine film. On the other hand, the salt with iodide presented a downward shift. As suggested by Chen et al.,[49] a deep EF can allow a better energy match with the highest occupied molecular orbital (HOMO) of the Spiro-OMeTAD, facilitating the charge extraction and decreasing the bimolecular charge carrier recombination at the interface. This result is in line with the TRPL measurements and the device’s performance, which point out an enhancement in the average lifetime and VOC, respectively, upon the treatment with FM-NHI. Combining the results, we can conclude that the better passivation effect observed for the salt with iodide is due to a combined effect of VI passivation, improved electron density over the fluorene moiety, and decreased bimolecular recombination at the surface.

Despite finding that the I-based salt resulted in the best performance, other works in the literature showed that Br–-[37] or Cl–-based[50] salts confer a superior passivation effect. The conflicting results reported by different groups reveal that it is quite hard to predict the exact role of the counteranion in the overall device efficiency as well as in the electronic and photophysical properties of the perovskite. This is because the formation of a passivating or 2D layer onto the film depends[51] on the perovskite composition, the purity of the precursors, the method of thin-film preparation, and the device architecture. Because all these parameters influence the defect concentration and surface energy of the perovskite, the role of the counteranion is not a predictable parameter, and variations are thus expected.

Further, we verified the effect of the salts in retarding the degradation of the PSC by tracking the device’s efficiency for 90 days (Figure d). The postpassivation, especially for FM-NHI, was effective at preventing device degradation, keeping average efficiencies near 18%. This effect is assigned to the decreased moisture permeation into the film. Also, the grain boundary and ionic vacancy passivation,[52] as well as the suppressed oxidation of the organic components in the perovskite film (see XPS results), could help decrease film degradation. The increased hydrophobicity conferred by the FM-NHX salts can be seen by the contact angle measurements, which increase from 47 to 92° upon the salt treatment (Figure c). It is worth noting that, at the same concentration (i.e., 8 mM), the neutral molecule, FM-NH, has a smaller effect on the contact angle when compared to the protonated molecules (Figure S15b and Figure c). These observations suggest that the cationic molecules interact strongly with each other when compared to the neutral analog, which is in line with a recent study on the interaction of organic cations by π-stacking interactions.[53] Such a conclusion is also supported by the fact that the electron density is higher on the fluorene moiety in the salt when compared to the FM-NH, benefiting the intermolecular π–π interactions (Figure S5a).

Finally, to understand in more detail how the protonated and neutral molecules interact with the perovskite surface, we created models varying (i) the particular orientation of the NHx groups in the molecules, (ii) the relative orientation of the molecules on top of the perovskite slab, and (iii) the termination of the perovskite surface. The details of the calculation and the results obtained were shown in Section 6 in the Supporting Information. We studied eight models: four containing the neutral covering a 2 × 2 × 3 FAPbI3 neutral slab where the FAI salt was removed from the surface, allowing the molecules to interact with Pb and I atoms (Figure S12), and the other four containing the protonated molecules covering a 2 × 2 × 3 FAPbI3 negatively charged slab where only the FA was removed, allowing FM-NH to effectively replace the FA molecules (Figure S13). In each group of four, the models were differentiated by the rotation of the aniline/anilinium group regarding the fluorene moiety, as well as the distribution of the molecules on the surface. This was done to ensure proper sampling of the different possible configurations and to evaluate possible energetic preferences.

With the calculations, we determined the average energy associated with the interaction, adsorption, and distortion, Ei, Ea, and Ed, respectively, measured as meV/Å2. We found that these values amount to -15.43, −12.61, and 2.81 for FAPbI3+FM-NH, and −110.73, −91.86, and 18.88 for FAPbI3+FM-NH. These results are consistent and show that the protonated molecules effectively establish an ionic bond with the surface, whereas the neutral molecule has only weak van der Waals (vdW) interactions. Thus, the fact that the interaction of the first is seven times stronger is coherent. This number would probably be higher if not for the increased distortion associated with the interaction of the protonated molecule with the perovskite. Using an alkylammonium chain, instead of the amine connected directly with the aromatic ring, for instance, should decrease this distortion, since there will be more degrees of freedom for the molecule to adopt a configuration with minimized impact on the organization of the lead octahedra on the surface. However, a deeper discussion on this topic goes beyond the scope of this article.

To further understand how molecule orientation and position affect total energy, we evaluated the dispersion energy associated with the different components of the system, as well as the dispersion energy directly associated with the interaction between perovskite and molecules. To do so, we calculate the optimized geometry structure of the different isomers of FM-NH and FM-NH (Section 5 in the Supporting Information). Those isomers differ from the relative position of the amine group, yielding three analogs: up–up, up–down, and down–down (see Figure S7). We found that the down–down configuration had slightly stronger dispersion energy (0.5 eV). Consequently, it possesses a stronger vdW contribution associated with the molecules stacking on the perovskite surface. This stacking was more compact than the up–up stacking, for instance, as it brought the fluorene moieties closer. This better packing of the molecule in the down–down configuration benefits π–π stacking interactions, which supports the formation of a hydrophobic protective film that decreases moisture permeation into the perovskite. Additionally, the dispersion associated with the interaction between molecules and surface was shown to be stronger in the FAPbI3+FM-NH systems (2.37 eV on average, against 1.53 eV for FAPbI3+FM-NH systems), which shows that the stronger interaction observed is a combination of Coulombic and vdW interactions.

In summary, we prepare FAPbI3 perovskite by the gas quenching method using a bidentate anilinium salt as a post-treatment of the film. With this approach, we obtained efficiencies up to 19% stable for more than 90 days. We demonstrated that the use of a salt derived from an AIE-luminogen, when compared to its neutral counterpart, was effective in passivating iodide vacancies on the surface of the film, in addition to creating a better protective barrier against moisture permeation. We found that the outperformance of the salt is related to the localization of the electron wave function on the fluorene moiety, which is beneficial to the passivation of iodide vacancies along with improving molecule–molecule interaction. We also showed that the counteranion of the salt substantially affects the EF of the surface. Hence, the design of new passivating agents should also include the selection of the best counteranion, which will depend on several parameters highlighting the environment of the amino group in the molecule, and the composition of the perovskite used. In future works, we intend to focus on the optimization of the perovskite layer, aiming to improve the efficiency of the device beyond the 19% obtained here.