Temperature-driven anion migration in gradient halide perovskites

In recent years, perovskite solar cells (PSCs) with their remarkable power conversion efficiencies as high as 24.2% have drawn the attention to design next generation photovoltaics.¹ While the majority of efficient PSCs have been organic-inorganic perovskites [where the A site cation is methylammonium (MA) or formamidinium (FA)], the instability of these perovskites in the presence of water and oxygen poses a problem for their implementation in practical devices.^2–6 All-inorganic lead halide perovskites, such as CsPbX₃ (X = I, Br, Cl), or mixed cation perovskites (A = FA, MA, Cs) show enhanced stability compared to organic-inorganic perovskites, especially with regard to temperature.^7,8

The recent development of silicon-perovskite two-junction tandem solar cells with efficiencies topping 28%, which is greater than silicon solar cells, has motivated the development of all-perovskite tandem solar cells.¹ In this context, we consider a gradient halide perovskite (GHP) structure as a potential tandem PSC due to its ability to capture photons of different energy (Scheme 1). For example, in our previous study, we reported the creation of CsPbBr_3−xI_x (0 < x < 3) gradient halide perovskites (GHPs) where a CsPbBr₃ rich region is linked to a CsPbI₃ rich region by a compositional gradient architecture.⁹ GHPs have the potential to serve as an active layer with a built in bandgap gradient. This unique band structure allows effective capture of photons across the visible spectrum and minimizes photon losses due to the bandgap gradient (Scheme 1). Higher energy photons are more efficiently absorbed in the CsPbBr₃ region, while low energy photons pass through the film until the light is eventually absorbed by the mixed halide and CsPbI₃ layers. This tandem architecture allows for photovoltaic devices that may overcome the Shockley-Quiesser limit. The built-in tandem bandgap has the added benefit of removing the losses and mismatch at junctions between active layers in traditional tandem device stacks.

The high contribution of the halide p-orbital to the valence band in lead halide perovskites causes a shift in the valence band across the GHP.^10–12 This shift in the valence band should help preferentially shuttle holes toward the CsPbI₃ rich regime. The shift in bandgap should also allow for recombination losses to be minimized since holes can be favorably driven toward the hole transport layer for extraction. This inherent charge directing capability of a gradient band structure can further minimize charge carrier recombination losses.

The highest efficiency PSC devices are mixed halide perovskites (MHPs) that implement a combination of halide ions (usually Br and I).¹³ This combination allows for absorption over the entire visible region and inhibits degradation by humidity and high temperatures.^14,15 However, halide ions have been shown to be highly mobile in MHPs, specifically under photoirradiation.^16–22 Under photoirradiation, halide ions segregate to form iodide rich and bromide rich domains. These domains act as recombination centers, lowering the overall efficiency of the PSC.^23,24 Understanding both the photo and thermally activated movement of halide ions in MHPs is important to develop strategies to overcome the photoinduced halide segregation effects in MHP based solar cells.

The mobility of the halide ions can also pose an issue for the stability of PSCs.¹⁵ In GHP films, the thermally activated movement of halide ions can lead to homogenization. However, arguments have also been made to indicate the presence of miscibility gap in attaining the mixed halide perovskites.^25,26 In order to further shed light on halide ion diffusion, we have now investigated homogenization of GHP films into MHP films at different temperatures. UV-Visible (UV-Vis) absorbance and X-ray photoelectron spectroscopy (XPS) measurements that provide new insights into the movement of halide ions in GHP films are described. The changes in the excited state properties of the gradient and homogenized films dictating the charge carrier flow are also discussed.

CsPbBr₃ nanocrystals (NCs) were synthesized using an adapted method from Protesescu et al.²⁸ A lead (ii) bromide solution was prepared in a 3 neck round bottom flask by combining 6 ml of oleic acid (Sigma Aldrich, 90% technical grade), 6 ml of 1-octadecene (Sigma Aldrich, 95%), and 0.828 g of PbBr₂ (Alfa Aesar, 99.999%). This solution was heated to 170 °C. A cesium-oleate solution was prepared in a separate round bottom flask by combining 0.306 g of cesium carbonate (Alfa Aesar, 99.9%), 2.8 ml of oleic acid, and 3 ml of 1-octadecene. This solution was heated to 150 °C. Both solutions were degassed for 1 h while being heated. After 1 h of degassing, 6 ml of oleylamine (Sigma Aldrich, 70% technical grade) was added to the lead (ii) bromide solution. Both flasks were then immediately placed under nitrogen. 4 ml of the as prepared cesium-oleate solution was injected into the lead (ii) bromide solution. The reaction was immediately quenched with an ice bath.

The NCs were washed by rinsing with 40 mL of 1-Octadecene (ODE) and centrifuged at 7800 rpms for 10 min. The supernatant was discarded, and the pellet was washed with 20 ml of ODE. The solution was centrifuged again (7800 rpms, 10 min). The supernatant was again discarded, while the pellet was rinsed with acetone (Fischer Scientific, HPLC grade), dried with an airstream, and suspended in about 5 ml of a 90/10 by volume hexane (n-hexane, Sigma Aldrich, 95% anhydrous)/heptane (n-heptane, Sigma Aldrich, 99% spectrophotometric grade) solution. The final solution was centrifuged one final time at 7800 rpms for 5 min, and the supernatant was kept.

Bulk CsPbBr₃ films were created by using a sequential deposition method described by Hoffman et al.⁹ Aliquots of CsPbBr₃ NCs were spun cast onto glass slides (cleaned with Versa-Clean, ethanol, and plasma cleaning). The NC films were annealed for 3 min at 250 °C which sintered the NCs to become bulk CsPbBr₃ films. The films were briefly washed with hexane in between deposition cycles. This process was repeated until the desired thickness was achieved (600 nm).

Gradient halide perovskite films were prepared by soaking the as made CsPbBr₃ films in a lead iodide solution. The lead iodide solution was prepared by combining 68 mg of lead (ii) iodide (Sigma Aldrich, 99.999% trace metal basis), 0.6 ml of oleylamine, 0.4 ml of oleic acid, and 20 ml of ODE at 170 °C while stirring until dissolved. The solution was cooled to 85 °C. Once at temperature, a CsPbBr₃ slide was placed in the solution and allowed to react for 45 min. The film (now a gradient perovskite) was rinsed thoroughly with hexane and dried with an air stream to remove any excess exchange solution.

In situ UV-Vis absorbance measurements were performed on a Cary 50 biospectrophotometer with the temperature controlled by a Peltier sampler holder. 2.5 cm × 0.8 cm films were placed into a 3 chambered quartz cuvette (film in center with ODE in the outer chambers to act as a heat exchange medium).

Films were sputtered with 2 nm of iridium before scanning electron microscopy (SEM) measurements to ensure that they were adequately conductive. A Magellan 400 SEM operated at 2 kV, 13 pA beam power was used for top down imaging and at 5 kV, 50 pA for cross sectional imaging. X-ray photoelectron spectroscopy (XPS) was performed on a Physical Electronics (PHI) VersaProbe II XPS.

A Clark MXR 2010 laser system paired with detection software from Ultrafast Systems (Helios) was used for transient absorption measurements. A 775 nm fundamental beam (1 mJ/pulse, FWHM = 150 fs, 1 kHz repetition rate) was passed through a 95/5 beam splitter where the 95% was frequency double to 387 nm to act as the pump beam (power density = 8 μJ/cm²). The 5% portion of the fundamental was passed through a titanium sapphire window to generate a white light super continuum which was used as the probe beam.

Bulk CsPbBr₃ films were prepared using a multistep spin coating process where layers of CsPbBr₃ QDs were deposited onto glass slides with annealing (250 °C) after deposition of each layer. Upon annealing, the CsPbBr₃ NCs are transformed into bulk perovskite with a characteristic change in absorbance of the appearance of a sharp excitonic feature around 510 nm.²⁷ The methods for synthesizing the nanocrystals as well as creating the bulk films have been reported in earlier studies.^9,28,29 The thickness of the perovskite film employed in the present study was around 600 nm (supplementary material, SI 1).

To create the GHP, the bulk CsPbBr₃ films were soaked in a lead iodide (PbI₂) solution in ODE to induce anion exchange. As bromide in the CsPbBr₃ film was exchanged with iodide, a visible color change from light-yellow to a dark brown-orange could be seen. Although the halide ions undergo exchange rapidly at the solution interface, the diffusion of iodide within the film is rather slow thus allowing us to tailor the composition of the gradient. The halide exchange was stopped at an intermediate stage (i.e., without allowing it to full conversion to CsPbI₃) after 45 min of exchange at 85 °C so that we can achieve a gradient composition.

The color change associated with the halide exchange was confirmed through UV-Vis absorbance (Fig. 1). With increasing time of exposure, the absorbance onset of the film showed a red shift along with decrease in the excitonic CsPbBr₃ peak and the appearance of a CsPbI₃ tail absorption that spans from 625 to 700 nm. This change in absorbance indicates a compositional gradient of CsPbI₃ at the surface of the film that transitions to CsPbBr₃ at the perovskite/glass interface. The evolution of partially exchanged films during iodide substitution has been systematically explored by Hoffman et al.⁹ The unique absorbance spectral features of GHPs provide a fingerprint that can be tracked through the homogenization (annealing) process.

SEM micrographs of the films taken before and after the anion exchange process [Figs. 2(a) and 2(b)] confirm that the halide ion exchange process does not appreciably affect the grains of the perovskite. Figure 2(b) shows a slight change in the surface morphology of the film, namely, the presence of smaller crystals. This small change in surface morphology can be attributed to the ligands in the anion exchange solution, which can cause the large grains to be slightly dissolved in solution and reformed during the heating process.

The GHP film was then annealed at 90 °C for 160 min in order to induce halide movement through heat. SEM images were taken 20 min into the annealing process and after the annealing process was complete [Figs. 2(c) and 2(d)]. Throughout the annealing process, the smaller crystallites become less prominent indicating bulk crystal growth during annealing as seen in our previous studies.²⁷ 

While SEM does provide information about the surface morphology of the films, the halide composition can be probed through UV-Vis absorbance spectra. MHP films have a distinct optical fingerprint where the absorption onset is dictated by the bromide to iodide ratio of the perovskite. GHPs, however, show a distinct absorbance peak at 575 nm along with a broad absorbance tail between 625 and 700 nm arising from the CsPbBr₃ region and the CsPbI₃ region, respectively. This drastic difference in absorbance features allows for UV-Vis absorbance to give information about the makeup of the film during the annealing process.

GHP films were placed in a heated cuvette and held at 90 °C, while in situ absorbance measurements were taken over time. Figure 3(a) shows the shift in absorbance of the GHP perovskite over time with heat. At time 0, absorbance corresponding to both CsPbBr₃ and CsPbI₃ regions are clearly seen. Intermediate spectra recorded at every 10 min show a spectral shift of the CsPbBr₃ peak toward higher wavelengths, while CsPbI₃ absorption diminishes. The increase in absorbance seen in the 595–615 nm wavelength region corresponds to MHP absorbance formed as a result of homogenization. After 160 min of annealing, the homogenization process becomes complete as no additional shift in absorbance could be seen. The absorbance spectra of the homogenized films show that the approximate formula of the perovskite is CsPbBr_1.5I_1.5.³⁰ 

To track the homogenization process in the above experiment, the absorbance at 600 nm (corresponding to the MHP absorbance region) was plotted vs time for annealing at 90 °C. Figure 3(b) shows the observed changes in the absorbance at 600 nm throughout the annealing process. This absorbance was fit with a monoexponential growth equation in order to find the apparent rate constant of homogenization. The fitting procedure was repeated for five other wavelengths between 595 and 615 nm. The average rate constant of homogenization at 90 °C was found to be 1.03 × 10⁻² s⁻¹.

GHP films were annealed at temperatures ranging from 50 to 80 °C, and rate constants of homogenization were determined for these temperatures as well. To verify that this process is purely due to thermal activation, a control experiment was run. The supplementary material, SI 2, shows a GHP film kept at room temperature in the dark for 16 h. A negligible change in the GHP absorbance was seen during this period, signifying that the homogenization process requires thermal activation.

At lower temperatures (for example, 50 °C), the rate of the homogenization process was significantly slower than at 90 °C, taking 960 min for homogenization to be complete (supplementary material, SI 3). For all temperatures studied, however, the trend in the change of the absorbance remains the same, showing that the homogenization process occurs in a similar fashion but with differing rates at all temperatures studied. The absorbance of the MHP region was tracked at different temperatures as described previously and can be found in the supplementary material, SI 4. These data were used to find the rate constant of homogenization for all five temperatures studied.

The Arrhenius equation, which takes the form of Eq. (1), illustrates that a linear relationship should exist between the natural log of the rate constant of a reaction and the inverse of the temperature as per the Arrhenius expression [Eq. (1)]

where k is the rate constant of homogenization, E_a is the energy of activation, R is the universal gas constant, T is the temperature in Kelvin, and A is a pre-exponential factor. An Arrhenius plot was constructed from the rate constants obtained at different temperatures, and a linear relationship is observed [Fig. 3(b)]. The energy of activation of the homogenization process taken from the slope of the Arrhenius plot was found to be 74.5 kJ/mol. This value of activation energy reflects the barrier for halide ion mobility in cesium lead halide perovskites.

The value of activation energy estimated in the present experiments is of the same relative order of magnitude as other literature values.^16,17,31,32 Interestingly, the energy of activation reported here is similar to the reported energy of activation for the exchange of iodide anions into CsPbBr₃ nanocrystals.³³ Additionally, Jin et al. reported the energy of activation for chloride ions through CsPbBr₃ nanowires to be 42 kJ/mol.³¹ A major difference between the value reported from the work of Jin et al. and our study is the different halide ions that were considered. Our value is representative of the movement of iodide and bromide ions through a CsPbX₃ film, while Jin et al. study the movement of chloride and bromide ions through nanowires. The large ionic radii of iodide compared to chloride should make it more difficult for iodide to move throughout the crystal lattice and is reflected in the larger energy of activation that was found in our experiments.

Previously, we have investigated the movement of halide ions in methylammonium based MHP.³⁴ It was found that the energy of activation of halide ion movement in these species is 51 kJ/mol, lower than our reported 74.5 kJ/mol. It has been reported that halide ion movement in MHP films under photoirradiation is suppressed in cesium containing perovskites compared to all methylammonium lead halide perovskites.¹⁵ This suppression of anion movement is reflected here due to our larger energy of activation of halide ion movement compared to our other study.

The rate of homogenization can also allow us to calculate the diffusion coefficient for all temperatures studied. From Fick’s law [Eq. (2)],

can be used to find the diffusion coefficient of halide ions, where d is the diffusion length, D is the diffusion coefficient, and t is the lifetime of the diffusion process. Using half of the thickness of the film as the diffusion length (300 nm), the diffusion coefficient for all temperatures studied can be found in Table I. These values are of the order of magnitude of 10⁻¹⁴ cm²/s.

Previous research into the movement of halide ions through perovskite films has reported similar values.^16,17,35,36 Values reported for the diffusivity of anions through CsPbBr₃/CsPbCl₃ nanowires have been shown to be on the order of 10⁻¹²–10⁻¹³ cm²/s.³⁷ While our values are smaller than the lower bound of that report, we attribute this to both lower temperatures studied in our experiment as well as the difference in halide ions. As mentioned in our discussion of the energy of activation, iodide ions are larger and therefore not able to move as easily through the lattice as chloride ions. Using impedance spectroscopy, values of halide ion mobility on the order of magnitude of 10⁻¹⁰ cm²/s for organic-inorganic perovskites reported for perovskite single crystals have been obtained. This suggests that the halide ion diffusion process is slower in cesium perovskites compared to organic-inorganic perovskites, a conclusion also supported through experimental research done into the rate of segregation of halide ions under photoirradiation.^15–20 

In order to obtain quantified values for the composition of halide ions at the surface of the GHP films, XPS measurements were conducted. Films were analyzed throughout the homogenization process with XPS to determine the ratio of bromide to iodide at the surface of the GHP films. This measurement acts as a proxy for the composition for the rest of the film. GHP films before homogenization should have primarily iodide at the surface with minimal bromide since the surface of the film is the interface at which anion exchange occurs. Throughout the homogenization process, the percent of bromide should increase until the final MHP composition is found.

Figures 4(a) and 4(b) show XPS spectra for bromide 3d and iodide 3d regions of films annealed for 0 min, intermediate annealing times, and 160 min at 90 °C. As seen through the previous absorbance measurements, the homogenization process is completed by 160 min and should therefore yield the final MHP composition of the sample. At 0 min (film before annealing), the bromide 3d peaks in the GHP film are difficult to resolve with accuracy because of its low concentration at the surface. By 20 min of annealing, however, bromide appears at the surface of the film. Throughout the annealing process, the iodide 3d peaks are seen indicating that iodide remains at the surface of the film throughout the process. It is important to note that the binding energy for the elements does not shift, indicating that no degradation has occurred during the homogenization process.

To gather quantitative information about the composition of bromide to iodide at the surface, the 3d peaks were fit using a pseudo-Voigt curve and the ratio of bromide to iodide was then calculated [Fig. 4(c)]. At 0 min of annealing, the anion composition is only ∼10% bromide at the surface, while iodide contributes to 90% of the anion sites. At the end of the homogenization process, the bromide and iodide compositions have become 45% and 55%, respectively. This is indicative of a MHP with the chemical formula CsPbBr_1.35I_1.65.

Thus far, the homogenization of GHP films has provided valuable information on the movement of halide ions through perovskite films, in general. GHPs, however, are useful light absorbing layers in their own right. Due to the high contribution of the halide p-orbitals to the valence band, charge shuttling from the CsPbBr₃ region to the CsPbI₃ region of the film should be inherent in the film. This was explored briefly by our lab previously and is expanded below.⁹ 

Transient absorption spectroscopy is a useful tool to study the excited state processes of semiconductors and dyes. After exciting the sample with a high energy laser pulse pump, the sample is probed with a white light pulse. Time-resolved difference spectra capture the recovery of the sample from the excited state to the ground state to be probed on a time scale of femtoseconds to nanoseconds.

High absorbance of the sample films at the pump wavelength of 387 nm leads to inhomogeneous absorption of the pump laser pulse. Using the absorption coefficient at the pump wavelength (387 nm), it was estimated that 90% of the excitation pulse is absorbed within ∼150 nm of the film (see the supplementary material).⁹ This penetration depth is much shorter than the thickness (600 nm) of the film. This depth of pump penetration allows for selective excitation of different areas of the film (i.e., the CsPbI₃ region of the CsPbBr₃ region at opposite ends of the film) so that the excited state processes in different regions can be studied.

The time-resolved transient absorption spectra of GHP films recorded following excitation from the iodide side and the bromide side are shown in Figs. 5(a) and 5(b), respectively. When excited from the iodide-rich side, two ground state bleaches at 584 nm and 675 nm were observed. Appearance of these two separate bleach bands indicates the change in absorbance in two independent excited species corresponding to CsPbBr₃ and CsPbI₃ regions of the film. The ground state bleach at 584 nm, which corresponds to the CsPbBr₃ region, is broader than what has been seen in the previous literature.³⁸ Here, however, the broadening of the peak is supported by our UV-Vis and XPS data that shows that iodide is incorporated throughout the film. This means that CsPbBr₃ does have small amounts of I⁻ incorporated into the bromide rich side of the film. The presence of different mixed halide species in the graded GHP film convolutes and broadens the CsPbBr₃ bleach signal. The CsPbI₃ ground state bleach is prominent in Fig. 5(a) since the CsPbI₃ is preferentially excited. Since CsPbBr₃ and CsPbI₃ have different molar extinction coefficients at the laser pump wavelength (387 nm), this leads to a difference in the magnitude of the excitation for the CsPbBr₃ and CsPbI₃ rich sides of the film.

The excitation of the bromide-rich side of the GHP film also shows two separate ground state bleaches, with the CsPbBr₃ ground state bleach being the prominent signal [Fig. 5(b)]. The formation of the CsPbBr₃ peak occurs within the duration of the excitation pulse duration and recovers within 0.5 ns. However, the CsPbI₃ ground state bleach does not appear as a fully developed band at t = 1 ps as with iodide-rich side excitation [Fig. 5(a)]. Instead, a slower growth of the bleach signal is seen until ∼100 ps indicating the formation of the iodide bleach from a process other than direct excitation. This suggests that with bromide-rich side excitation, the iodide-rich region bleach in GHPs can occur from charge carrier migration from the higher bandgap region to the low bandgap iodide-rich region. Due to the high contribution of halide ion orbitals to the valence and conduction bands, the iodide-rich region has a valence band with higher energy (Scheme 2).^10–12 This means that holes are preferentially shuttled toward the iodide-rich side of the film. Since the conduction band is isoenergetic across the structure, the electrons can also readily move toward the iodide-rich region. Therefore, the lower bandgap of CsPbI₃ serves as a sink to capture electrons and holes generated at different parts of the gradient film.

If indeed the slower growth of the CsPbI₃ bleach in Fig. 5(b) is due to charge carrier migration, we should be able to elucidate these processes through comparison of the excited state lifetimes from Figs. 5(a) and 5(b). The kinetics of the ground state bleaches at the CsPbBr₃ peak (584 nm) and the CsPbI₃ peak (675 mm) extracted from the time-resolved transient spectra of Figs. 5(a) and 5(b) are shown in Figs. 5(c) and 5(d), respectively. For CsPbI₃ side excitation, both bleach signals appear within the pulse duration (<1 ps) followed by excited state lifetime decay with different lifetimes [Fig. 5(c)]. The CsPbBr₃ bleach recovery exhibits a shorter lifetime than the CsPbI₃ bleach, thus indicating the independent disappearance of the excited state of the two different regions of the film.

When excited from the CsPbBr₃ side, we observe a different appearance of the two bleach signals [Fig. 5(b)]. The CsPbBr₃ signal appears promptly after the laser pulse excitation on the same time scale as the previous experiment in Fig. 5(c). The CsPbI₃ peak, however, shows a growth in the bleach signal that does not fully appear until ∼100 ps postexcitation. This growth time corresponds to the time scale of the decay of the bromide bleach signal. This kinetic analysis shows that the appearance of the excited CsPbI₃ in the gradient film (when excited through the bromide side) is indirect and is formed through charge carrier migration from higher bandgap regions to the low-lying bands of CsPbI₃ states.

To further correlate this charge carrier migration between the two species in the GHP film, the rise time of the CsPbI₃ signal and the decay time of the CsPbBr₃ signal were fit with a monoexponential growth and decay, respectively, for the first 100 ps after excitation. The apparent rate constant of relaxation for excited CsPbBr₃ was found to be 2.8 × 10¹⁰ s⁻¹. Since this value was calculated for the first 100 ps of data, the main contribution to the relaxation comes from charge carrier transfer and not recombination which occurs over a longer time scale. The apparent growth rate constant of excitation for CsPbI₃ was calculated to be 2.1 × 10¹⁰ s⁻¹. The direct correlation between these two values implies that the excitation of CsPbI₃ arises from charge transfer from CsPbBr₃ across the film and not direct laser excitation. This charge carrier transfer process was not seen when excitation was limited to the low bandgap region (CsPbI₃) excitation but seen only when excited from the higher bandgap (CsPbBr₃/mixed halide) excitation. This selective response shows that the charge carrier migration is unidirectional and dependent on the favorable band alignment in the GHP. Scheme 2 shows the relative band alignment for a GHP film as well as the charge carrier migration discussed in this study.

In metal halide perovskites, the A-site cation has a negligible contribution to the overall band structure of the material.^10–12 As mentioned previously, the B-site cation, in our case lead, primarily dictates the position of the conduction band. No change is seen in the conduction band position since lead is present in all species studied here, while the valence band position, which comes mostly from the p-orbital of the halide ion, shifts to higher energies as the perovskite transitions from CsPbBr₃ to CsPbI₃. This means that we can assume that holes are the primary charge carrier transferred through the material. In order to engineer tandem devices of GHPs, an architecture that places the hole transport layer on the CsPbI₃ interface would allow for the most efficient charge extraction.

Finally, transient absorption spectroscopy was also used to confirm the presence of only one species in the films after the homogenization process. Unlike in the GHP films, Figs. 6(a) and 6(b) show that only one ground state bleach is present in the posthomogenized films in both iodide side and bromide side excitation. The bleach position around 600 nm corresponds to mixed halide film. The presence of only one ground state bleach, instead of at least two as in Fig. 5, supports the conclusion drawn from UV-Vis absorbance and XPS that a MHP is formed after the homogenization process. The transient absorption measurements discussed here show a convenient way to probe the excited state processes in GHP films and track the flow of photogenerated charge carriers.

The homogenization of GHPs to form MHPs provides a useful avenue for the synthesis of bulk cesium MHPs. While cesium based MHP NCs can be easily synthesized or created through postsynthetic methods, the same is not true for bulk polycrystalline films.^28,30,39 This is due to the poor solubility of cesium salts used for synthesis and the preference for CsPbX₃ perovskites to form the nonoptically active phases.^40–43 This technique provides a useful opportunity to create cesium based MHPs for photovoltaic operations with varying halide ion concentrations.

Halide ion movement has proved to be an interesting property in lead halide perovskites since this property allows for facile anion exchange which leads to tunable perovskite compositions.^30,39,44,45 This very same property has proved to be a challenge toward attaining perovskite stability, especially for MHP under photoirradiation.^19,20,46,47 Understanding the movement of halide ions in perovskites is imperative in order to create strategies to mitigate halide ion migration. In a recent study with a MAPbBr_xI_1−x–MAPbBr₃ heterostructure, a miscibility gap at 100 °C was noted instead of a fully miscible system.²⁶ The authors used a polydimethylsiloxane (PDMS) mask to expose selective area to Br₂ and obtain a MAPbBr₃ region. It is not clear whether the surface treatment or remediation of surface defects with Br₂ treatment could have limited bromide ion diffusion. The surface treatment with trioctylphosphine oxide (TOPO) or inclusion of 2D perovskites has been shown to suppress the halide ion diffusion.^48,49 In the present study, however, we have not employed any surface treatment but only a built-in gradient structure within the CsPbBr_3−xI_x film which has allowed us to study the homogenization of halide ions without miscibility gap issues. Our study provides additional knowledge about the diffusion processes that occur in cesium based mixed halide perovskites as well as charge carrier migration within the gradient structure. The diffusion coefficient of halide ions through cesium GHP to form MHP was found to be smaller than the values found in previous studies on methylammonium based perovskites. This slower anion migration seen with incorporation of cesium in organic-inorganic MHP further explains reduced anion migration and its beneficial effect in improving stability.

Strategies to mitigate the homogenization of anions are needed in order to implement these films in optoelectronic devices. We have shown that inorganic capping layers, such as PbSO₄ can inhibit anion exchange across nanocrystals.^50,51 Efforts are underway in our laboratory to employ PbSO₄ between MHP layers to minimize the anion movement across perovskite films.

The supplementary material contains cross-sectional SEM, control experiments, additional absorbance data for different temperatures, XPS, and transient absorption spectra.

P.V.K. acknowledges the support of the Division of Chemical Sciences, Geosciences, and Biosciences, Office of Basic Energy Sciences of the U.S. Department of Energy, through Award No. DE-FC02-04ER15533. R.A.S. acknowledges the support of the Arthur J. Schmitt Leadership Fellowship and CEST-Bayer fellowship (University of Notre Dame). This is NDRL Contribution No. 5253 from the Notre Dame Radiation Laboratory. The authors would also like to thank Dr. Tatyana Orlova from the Notre Dame Integrated Imaging Facility for her helpful discussions of SEM and Dr. Jacob B. Hoffman for his helpful discussions.