Unraveling the Impact of Halide Mixing on Perovskite Stability.

Increasing the stability of perovskites is essential for their integration in commercial photovoltaic devices. Halide mixing is suggested as a powerful strategy toward stable perovskite materials. However, the stabilizing effect of the halides critically depends on their distribution in the mixed compound, a topic that is currently under intense debate. Here we successfully determine the exact location of the I and Cl anions in the  CH3NH3PbBr3- yI y and CH3NH3PbBr3- zCl z mixed halide perovskite lattices and correlate it with the enhanced stability we find for the latter. By combining scanning tunneling microscopy and density functional theory, we predict that, for low ratios, iodine and chlorine incorporation have different effects on the electronic properties and stability of the CH3NH3PbBr3 perovskite material. In addition, we determine the optimal Cl incorporation ratio for stability increase without detrimental band gap modification, providing an important direction for the fabrication of stable perovskite devices. The increased material stability induced by chlorine incorporation is verified by performing photoelectron spectroscopy on a half-cell device architecture. Our findings provide an answer to the current debate on halide incorporation and demonstrate their direct influence on device stability.

Introduction

Metal halide perovskite materials have been the focus of intense investigation due to their low cost and superior photovoltaic performance, with record power conversion efficiencies reaching 23.7%.[1] Despite numerous advantages, their poor stability hinders commercialization of perovskite-based devices. To increase perovskite stability various strategies have been envisaged.[2,3] Mixing different halides (I, Br, Cl) has been shown both experimentally and theoretically to have a strong impact on the device performance and stability.[4−9] For example, a small addition of bromine[10] or chlorine[11] confers increased device stability compared to pure iodide-based (CH3NH3PbI3) devices. Addition of Cl to CH3NH3PbBr3 films improves carrier recombination lifetime[12] and increases open circuit voltage.[13,14] However, the role of Cl and the extent of its incorporation into the crystal lattice remain a topic of debate, with conflicting results reported in the literature.[15−21]

Here, we choose to mix CH3NH3PbBr3 perovskite with two different halides, I and Cl, which have different ionic radii and bond strengths with Pb. In this way, the effect of the incorporation of a larger ion (I > Br) and a smaller ion (Cl < Br) can be addressed. In this work, combining scanning tunneling microscopy (STM), density functional theory (DFT), and UV/X-ray photoelectron spectroscopy (UPS/XPS), we reveal the exact location of I and Cl anions in the perovskite structure at the atomic scale, and demonstrate the impact of halide incorporation on the material electronic properties and stability. Furthermore, we determine the optimal Cl incorporation ratio for stability increase without detrimental band gap modification.

Despite the great capability of STM and UPS as demonstrated in the current study, it is important to point out that there are some limitations associated with such surface sensitive characterization techniques. Combined with suitable DFT calculations, STM can be used to investigate mainly the surface and possibly also the subsurface structures of perovskite materials. The probing depth of photoelectron spectroscopy (PES) depends on the energy of photoelectrons, but in general it is still considered to be surface sensitive as it can only probe up to approximately 5 nm of the top surface of the sample. Therefore, our current surface science study is not an attempt to provide a full picture of what it is happening in the bulk material of a solar cell device. On the other hand, although our results may not accurately represent the  interior of the perovskite films as in real solar cell absorbers, our study does offer valuable insights on other aspects such as surface/interface properties. Several studies have reported that surface/interface properties themselves play an important role in determining perovskite solar cell performance.[22,23] For example, it has been shown that the interface between the perovskite film and the hole transport layer (HTL) is crucial for achieving high solar cell device performance.[24−27] The aim of our current study combining the use of STM, DFT calculations, and PES is therefore to obtain a better understanding of these interfacial properties by examining their electronic structure and morphology with an unprecedented high spatial resolution.

Results and Discussion

CH3NH3PbBr3 perovskite thin films were prepared on a clean Au(111) surface[28] via dual-source co-evaporation in an ultrahigh vacuum environment (see the Methods section). The thickness of the ultrathin film was approximately 4 ± 1 nm. Scanning tunneling microscopy images reveal a pair motif similar to the characteristic surface reconstruction previously observed for CH3NH3PbBr3 single crystals[29−31] showing the (010) surface termination (Figure a). The bright spots in the STM images correspond to bromine ions of the top layer perovskite structure. For pure CH3NH3PbBr3 all the bromine ions have the same height and width in the STM images.

Figure 1

Halide substitution at the perovskite surface. Scanning tunneling microscopy images of (a) CH3NH3PbBr3, (b) CH3NH3PbBr3–I, and (c) CH3NH3PbBr3–Cl perovskite surfaces. (d–f) Calculated (010) surface of the mixed-halide organic–inorganic perovskites. Inset: corresponding surface structure and unit cell. The lattice unit cell of pristine CH3NH3PbBr3 is square with the parameters A = B = 7.4 Å, and was found to not be altered by the deposition of PbI2 or PbCl2 molecules. The sizes of Br, I, and Cl ions were multiplied by a factor of 4.5 for better visibility. (g) Scheme of the substitution mechanism occurring at the surface of CH3NH3PbBr3 perovskite after deposition of PbI2 or PbCl2 molecules (only the PbI2 case is presented for clarity, but PbCl2 is following the same mechanism). Image parameters: (a) sample bias = +1.3 V, tunneling current = 80 pA; (b) +2.0 V, 120 pA; (c) −1.9 V, 19 pA; image size, (a–c) 2.3 nm × 2.3 nm.

Thereafter, PbI2 or PbCl2 was deposited on the pure CH3NH3PbBr3 film kept at room temperature (see the Methods section). Distinct protrusions with different apparent height and width appear on the perovskite film surface (Figure b,c). After deposition of PbI2, strong bright spots appear with an estimated height of 40 ± 10 pm higher than the surrounding Br ions and with larger diameters (Figure b). In contrast, after PbCl2 deposition, slightly darker spots are evident, with smaller diameters and a lower apparent height (Figure c). These dark spots appear 20 ± 10 pm lower than the neighboring bromine ions.

To unravel the origin of the dark and bright protrusions observed on the surface of the perovskite films, we performed DFT calculations. The atomic structures, as well as the projected density of states (PDOS), of the perovskite films were determined using the Vienna ab initio simulation package (VASP) code (see the Methods section). First the pair structure of the (010) top layer surface of CH3NH3PbBr3 perovskite was reproduced using DFT and taken as a reference (Figure d). Thereafter the impact of PbI2 and PbCl2 deposition on the atomic structure of the perovskites was further addressed by DFT. Different scenarios were considered, such as the adsorption of the entire PbI2/PbCl2 molecules (Figure S1), as well as the adsorption (Figure S2) or substitution of Pb, I, and Cl on the CH3NH3PbBr3 surface (Figure ). The best match between the STM experimental results and DFT simulation is found for the dissociation of the PbI2 (or PbCl2) molecule, followed by the substitution of Br by I (or Cl). The schematic drawing for the iodine case is shown in Figure g as an example. We thus assign the bright and dark protrusions observed in our STM experiment, respectively, to I and Cl ions, which substitute Br ions at the surface of the perovskite film (Figure e,f). These scenarios are the most energetically favorable ones, and the corresponding simulated STM images best reproduce our experimental data (supplementary note 1, in the SI). Thus, by depositing PbI2 or PbCl2 on a pristine CH3NH3PbBr3 film, two different mixed-halide perovskites, CH3NH3PbBr3–I and CH3NH3PbBr3–Cl, were formed.

Large-scale STM images (Figure a–c) of the surface topography show that the additional bright and dark spots observed after PbI2 and PbCl2 deposition are randomly distributed at the surface of the perovskite films without any long-range order (Figure b,c). These findings are supported by the fast Fourier transform (FFT) analyses of the STM images, which do not show any additional peak at low k-values for the mixed-halide perovskites compared to pristine CH3NH3PbBr3 (insets in Figure a–c). The distribution histogram of the different apparent heights that are measured at the surface of the three films is given in Figure d–f. While for the pristine perovskite only a single peak is present corresponding to Br ions, additional peaks show up for the mixed-halide perovskites corresponding to I and Cl ions, respectively. Surprisingly, the substitution of Br by I or Cl does not significantly change the lattice unit cell (i.e., lateral structure) of the perovskite surface. However, the vertical positions of the I and Cl were found to be different from the Br ions in pristine CH3NH3PbBr3 (Figure g–i). The iodine ions are slightly upshifted with respect to the Br plane (Figure h).

Figure 2

Atomic-resolution surface structures of mixed-halide perovskites. Large area STM images (10 nm × 10 nm) of the (010) surface of (a) the pristine CH3NH3PbBr3, (b) CH3NH3PbBr3–I, and (c) CH3NH3PbBr3–Cl mixed-halide perovskites. Scan parameters: (a) sample bias = +1.3 V, tunneling current = 80 pA; (b) +2.0 V, 120 pA; (c) −2.0 V, 100 pA. Inset: fast Fourier transform (FFT) results obtained from the topographic STM images showing no additional peak at low k-values for the mixed-halide perovskite, suggesting that there is no long-range order of the additional iodine and chlorine ions at the surface. (d–f) Histogram of the apparent height distribution of local maxima (ions). The main peaks correspond to the Br ions, while the small additional peaks are associated with iodine and chlorine ions (e and f, respectively). Inset: typical profiles obtained for the different halides (Br, I, Cl) at the perovskite surface. (g–i) Relative positions of the halide ions with respect to the top CH3NH3Br structure. The dash lines in side view indicate the differences in the height calculated by DFT for the I ion (+0.35 Å in part h) and Cl ion (−0.24 Å in part i) with respect to the Br ion. The values calculated by DFT are in good agreement with the experimentally measured differences in the STM apparent height (in parts d−f). Color code: Pb (gray), Br (brown), I (purple), Cl (green), C (black), N (blue), H (light pink).

Additionally, the iodine ions have a larger ionic radius than bromine. Both the upshift and larger radius contribute to the observed topographic height increase. In contrast, the chlorine ions are downshifted and have a smaller ionic radius compared to Br (Figure i) explaining their lower apparent height. These experimentally observed STM apparent height differences are further confirmed by the DFT calculations (Figure h,i and Table S1).

Besides the single bright and dark protrusions, which are attributed to the substitution of Br with I or Cl, we observed pairs and clusters of bright and dark protrusions (Figure S3). They are attributed to I or Cl ions that substitute multiple nearby bromine ions of the perovskite surface (Figure S4). Two types of pairs are found on the surface. Iodine and chlorine pairs can be formed either by substitution of two bromine ions within the same pair, or by the substitution of two neighboring Br ions within two adjacent pairs (Figure S3). DFT calculations show that the chemical nature of the halides strongly influences the stability of the pair. The energy barrier for the formation of two isolated Cl–Br pairs is lower than the energy formation of a single Cl–Cl pair (E2(Cl–Br) < ECl–Cl). These results suggest that at low coverage the chlorine ions prefer to form isolated Cl–Br pairs rather than clustering together to form Cl–Cl pairs or phase segregating. The same trend was found for the iodine case, i.e., E2(I–Br) < EI–I (Table S1). This is the reason why the substituting halide ions were mainly found to be randomly distributed at the surface of the perovskite, leading to the disordered surface structure observed in the experiment.

Ultraviolet photoemission spectroscopy (UPS) experiments were performed on the mixed-halide perovskites to correlate their atomic structures with their electronic properties (Figure ). The valence band (VB) spectra of the perovskite films are composed of 6 main peaks (A–F) as shown in Figure a. A strong modification of the valence band line shape can be observed when substituting bromine ions at the surface with iodine or chlorine. Based on DFT calculations, we found that the VB of pristine CH3NH3PbBr3 is composed mainly of Br p, p, and p hybridized states with a small contribution from the Pb s and p orbitals (Figure b). In addition to these states, contributions of I and Cl appear in the VB of mixed-halide perovskites. The substituting halides are responsible for the valence band modification as observed in the experiment. A higher binding energy was found for Cl as compared to Br and I (Figure b), in line with the higher stability of the chlorine-mixed perovskite (CH3NH3PbBr3–Cl) as discussed later. Additionally, the impact of halide substitution on the material work function was revealed by UPS. A work function (ϕ) of 4.77 eV was measured for pristine CH3NH3PbBr3, similar to the values reported in the literature,[32−34] while a decrease of 0.2 eV of the work function was observed after chlorine substitution (ϕ = 4.57 eV). Interestingly with iodine incorporation in the perovskite film, the work function was 4.79 eV, similar to the value of pristine CH3NH3PbBr3 (Figure S5). We found that the work function changes originate from a downshift of the vacuum level induced by a modification of the surface dipole moment after halide incorporation (Figure S6). A larger dipole moment amplitude results in a stronger downshift of the vacuum level, and thus a smaller work function (ϕI > ϕBr > ϕCl). The possibility of adjusting the work function as well as the vacuum level position through halide substitution is of prime interest for engineering energy level alignment in perovskite-based devices.

Figure 3

Valence band modification in mixed-halide perovskite. (a) Experimental ultraviolet photoemission spectroscopy (UPS) spectra of pure CH3NH3PbBr3 (orange) and mixed CH3NH3PbBr3–I (purple), CH3NH3PbBr3–Cl (green) perovskites. (b) Calculated PDOS of the pristine and mixed-halide perovskites. The thick curves represent the summation of all contributions of each orbital indicated by differently colored lines. The PDOS were obtained for the full halide substitution of the perovskite surface.

To further understand the impact of halide substitution on the electronic properties of perovskite materials and its correlation with the stability, we evaluated the changes in the decomposition energy (see the Methods section) and band gap of the CH3NH3PbBr3–I and CH3NH3PbBr3–Cl films for different iodine and chlorine ratios. As depicted in Figure a, the substitution of bromine ions by iodine at the surface of the perovskite induces a decrease of the decomposition energy (purple curve) and a small reduction of the band gap (orange curve). In contrast, Figure b shows that the incorporation of a small amount of chlorine (below 25%, green curve) leads to an increase of the decomposition energy, with no substantial band gap changes.

Figure 4

Stability versus band gap change in mixed perovskite. Decomposition energy difference between mixed-halide and pristine perovskite determined by DFT calculations (see the Methods section): (a) CH3NH3PbBr3–I with increasing iodine content (y) at the surface, and (b) CH3NH3PbBr3–Cl with increasing Cl content (z) at the surface. The calculations were performed on a 6 layer slab with only the top layer being substituted. Bottom panels show the atomic structure of the pristine, half-mixed, and completely substituted top surface of the perovskite. Color code: Br (brown), I (purple), Cl (green), C (black), N (blue), H (light pink).

An increase of the decomposition energy after chlorine substitution suggests higher stability of the perovskite film with respect to external stimuli such as temperature, X-ray beam damage, etc. This stability increase is related to the stronger bond strength of Cl–Pb compared to Br–Pb and I–Pb.[35,36] Beyond the substitution ratio of 25%, the strain induced by the incorporation of the smaller Cl ions with respect to Br counterbalances the benefit of the stronger Cl–Pb bond strength and results in a lowering of the material stability (decomposition energy). Thus, the stability in mixed-halide perovskite results from the interplay between the Pb–halide bond strength that stabilizes the material and the strain induced by the halide substitution that lowers decomposition energy. Based on our results, we suggest that, as a general trend, substitution of the original surface halide ions with a smaller ion size and a stronger Pb-halide bond strength leads to an increase of the decomposition energy of the system, which in turn increases the material stability. Our DFT finding shows a stability increase without a strong band gap modification. Although here only the substitution at the surface is considered, these results indicate that there may exist an optimal bulk Cl substitution ratio that would provide higher stability without substantially changing the band gap of the perovskite material.

To further confirm these findings experimentally, we have studied the stabilization effect of chlorine on an actual solar cell device architecture (Figure S7). First, a CH3NH3PbBr3 film with a thickness of approximately 300 nm was spun on a TiO2/FTO/glass substrate (FTO, fluorine-doped tin oxide; see the Methods section), which resembles half (up to the perovskite layer) of a standard perovskite solar cell structure. Thereafter, the half-cell architecture was transferred into the UHV system, where chlorine was incorporated into the perovskite film by  PbCl2 vacuum evaporation. The degradation was then investigated by X-ray photoelectron spectroscopy (XPS) for the half-cell samples with different Cl substitution ratios at the surface (Figure ). First, the evolution of the Pb 4f peak was monitored for the pure CH3NH3PbBr3 film without the incorporation of Cl. Figure a shows that initially (at T = 0 h), the spectrum is composed of only two peaks at 137.9 and 142.8 eV, attributed to the spin–orbit split Pb 4f7/2 and Pb 4f5/2 peaks, later referred to as Pb(2+). After 4 h, additional peaks appear at lower binding energies associated with the lower oxidation Pb(0) state. Such a state is associated with deterioration of the perovskite film.[37] Therefore, in the present study we use the Pb(0) peak evolution as an indicator of the perovskite degradation. A progressive and strong increase of the Pb(0) peak intensity is observed as a function of time, indicating that the pure CH3NH3PbBr3 film undergoes a fast degradation. In contrast, much slower degradation (i.e., a significantly reduced Pb(0) peak intensity after a given time) was observed for the samples with the additional PbCl2 evaporation (Figure b,c). The degradation process is further slowed down as the amount of chlorine substitution increases. When 12% of the Br ions are substituted by Cl at the perovskite surface, the Pb(0) peak emerged after 28 h, while it only appears after 116 h for a 18% substitution ratio (Figure d). For comparison, the surface substitution ratio at the perovskite surface in the STM image of Figure c is about 26%. Substitution ratios were determined by measuring the Cl:Br chemical ratio in XPS. A quantitative analysis of the remaining Pb(2+) species (i.e., undegraded Pb ions on the surface) after a given time is presented in Figure e. The results demonstrate experimentally that the incorporation of Cl improves stability of perovskite films. It has been reported that CH3NH3PbBr3 shows significantly better stability than CH3NH3PbI3.[38] Here in this study, we have shown that, with the surface Br ions of CH3NH3PbBr3 partially substituted by Cl ions, the half-cell sample stability gets further enhancement. Such a result is expected to have strong implications for device applications. Although the current study focuses on the CH3NH3PbBr3 case, the stability increase induced by Cl incorporation is expected to be even stronger in the CH3NH3PbI3 perovskite case, due to a larger difference in bond strength between Pb–Cl and Pb–I when compared to Pb–Cl and Pb–Br. Finding the balance between stability increase and band gap modification in Cl-incorporated CH3NH3PbI3 is expected to play an important role in optimizing CH3NH3PbI3 perovskite-based devices.

Figure 5

Beneficial aspects of halide mixing on perovskite stability. Time evolution of the Pb 4f peak obtained by X-ray photoelectron spectroscopy (XPS) on a device architecture (a) without addition of chlorine, (b) with 12% of Cl at the surface, and (c) with 18% Cl at the surface. (d) Appearance time of Pb(0) feature with respect to the Cl substitution ratio at the surface. (e) Time evolution of the percentage of Pb(2+) remaining for the different Cl substitution ratio at the surface.

Conclusion

We have determined for the first time the surface atomic structure of CH3NH3PbBr3–I and CH3NH3PbBr3–Cl mixed-halide perovskites in real space and correlate it to the stability of the corresponding compound. Using scanning tunneling microscopy combined with density functional theory we find that Br ions are substituted by I or Cl ions. The substituted halides are mainly randomly distributed at the surface, showing a disordered surface structure with no phase segregation. For the low substitution ratio (below 25%), I and Cl substitutions have different effects on the work function and stability of the mixed perovskites. Our studies on half-cell device architectures demonstrate that the stability is increased after incorporating 12–18% of Cl, which agrees with the calculated increase of the decomposition energy. Our results suggest that there is an optimal ratio of chlorine substitution which provides an increase in stability without significant band gap modification. This balance is an important direction for future device research, to address the perovskite stability issue, which remains as one of the main challenges for industrial applications.

Methods

Surface Characterization

CH3NH3PbBr3 perovskite films were grown on a clean Au(111) single crystal using dual-source co-evaporation, following a previously reported method.[28] Vacuum deposition techniques have been used to fabricate perovskite solar cell devices with similar power conversion efficiencies to those fabricated by solution-based deposition methods.[39−41] In addition, the use of vacuum deposition techniques ensures the sample’s cleanliness and high purity (e.g., free from solvent molecules), which facilitates further atomically precise characterization with scanning tunneling microscopy. First, the gold sample was cleaned in an ultrahigh vacuum environment (∼1 × 10–9 Torr) with several cycles of Ne+ sputtering followed by subsequent annealing at 773 K for 5 min. Thereafter, CH3NH3PbBr3 was obtained by dual-source co-evaporation of CH3NH3Br (MABr) and PbBr2 molecules at 361 and 498 K, respectively, for 10 min. During deposition, the gold surface was kept at 130 K to ensure the adhesion of MABr, and then annealed at room temperature for at least 3 h. Mixed perovskites were obtained by sublimating PbI2 and PbCl2 on the CH3NH3PbBr3 film from Knudsen cells held at 525 and 560 K, respectively. During sublimation, the sample was kept at room temperature. Low-temperature scanning tunneling microscopy (LT-STM) was used to characterize the atomic-scale surface structures of the different perovskite ultrathin films. The STM measurements were performed at 5 K. A cut Pt/Ir tip was used to acquire the STM images. The bias voltage was applied to the sample. X-ray photoelectron spectroscopy (XPS) and ultraviolet photoemission spectroscopy (UPS) were performed in the same ultrahigh vacuum (UHV) setup as the STM measurements, preventing any contamination from air.

Density Functional Theory

The calculations were performed using the VASP code[42−44] with the projector augmented wave (PAW)[45] potential and Perdew–Burke–Ernzerhof (PBE)[46] exchange-correlation functional. For the unit cell, Brillouin zone (BZ) sampling was done using a 4 × 4 × 4 Monkhorst–Pack grid. The CH3NH3PbBr3 (010) surfaces were modeled using (2 × 2) slabs containing six atomic layers. The size of the supercell was (16.24 × 16.54 × 35.52) Å. A large vacuum thickness of 20 Å along the direction normal to the surface was employed to separate surfaces from their periodic images. The bottom four layers of atoms were kept fixed at their optimized bulk positions. All other atoms were fully relaxed until the forces on each atom were smaller than 0.01 eV/Å. The nonlocal van der Waals interactions were evaluated with the optB86-vdW functional.[47] A Γ-centered Monkhorst–Pack sampling of 3 × 3 × 1 was adopted to study the electronic properties of the surfaces of CH3NH3PbBr3 (MAPbBr3). From the electronic ground-state structure, STM images were calculated with the p4vasp code.

To evaluate the stability of the mixed perovskite, we consider the following reaction:where X is Cl or I, and y/3 is its ratio. The decomposition energy of the mixed perovskite 2MAPbBr3–X is calculated asThe stability of the perovskite is characterized by the decomposition energy (Edec). When the decomposition energy is larger, the material is more stable.

Device Fabrication

The patterned FTO glass (Pilkington, TEC-8, 8Ω/sq) was cleaned with detergent and DI water and sonicated with ethanol in an ultrasonic bath for 20 min. UV-ozone was treated for 15 min prior to use. The TiO2 (ca. 50 nm) layer was deposited by spin-coating at 2000 rpm for 20 s (acceleration was 500 rpm/s) using a diluted TiO2 solution in 1-butanol.[48] The deposited TiO2 film was annealed at 550 °C for 30 min. The CH3NH3PbBr3 film was prepared via the adduct method.[49] 0.367 g of PbBr2 (TCI) and 0.112 g of CH3NH3Br (Dyesol) were dissolved in 0.325 mL of DMF with 225 μL of DMSO at 60 °C. The CH3NH3Br3 solution was spin-coated on the TiO2 substrate at 4000 rpm for 20 s, and 0.3 mL of diethyl ether was poured when 10 s remained. Then, it was dried on a hot plate at 100 °C for 20 min.