Light-Induced Halide Segregation in 2D and Quasi-2D Mixed-Halide Perovskites

Photoinduced halide segregation hinders widespread application of three-dimensional (3D) mixed-halide perovskites. Much less is known about this phenomenon in lower-dimensional systems. Here, we study photoinduced halide segregation in lower-dimensional mixed iodide-bromide perovskites (PEA2MAn–1Pbn(BrxI1–x)3n+1, with PEA+: phenethylammonium and MA+: methylammonium) through time-dependent photoluminescence (PL) spectroscopy. We show that layered two-dimensional (2D) structures render additional stability against the demixing of halide phases under illumination. We ascribe this behavior to reduced halide mobility due to the intrinsic heterogeneity of 2D mixed-halide perovskites, which we demonstrate via 207Pb solid-state NMR. However, the dimensionality of the 2D phase is critical in regulating photostability. By tracking the PL of multidimensional perovskite films under illumination, we find that while halide segregation is largely inhibited in 2D perovskites (n = 1), it is not suppressed in quasi-2D phases (n = 2), which display a behavior intermediate between 2D and 3D and a peculiar absence of halide redistribution in the dark that is only induced at higher temperature for the quasi-2D phase.


S2
precursor solution was spin coated at room-temperature at 5000 rpm for 45 s, followed by thermal annealing at 100 °C for 10 min.

Film characterization
UV-vis-NIR absorption spectra were recorded by using a PerkinElmer Lambda 1050 UV-vis-NIR spectrophotometer. Photoluminescence spectra were recorded by using an Edinburgh Instruments FLSP920 double-monochromator luminescence spectrophotometer. XRD patterns were recorded by using a Bruker 2D phaser (Cu Kα radiation, λ = 1.5406 Å): measurements were performed in the range 3-40° with a step size 0.02° and collection time of 1 s. 2D GIWAXS measurements were performed with a Ganesha 300XL+ system from JJ X-ray equipped with a Pilatus 300K detector (pixel side 172 µm × 172 µm). The X-ray source was a Genix 3D Microfocus sealed tube X-ray Cu-source with integrated monochromator. The wavelength used was 1.5406 Å. The detector moves in a vacuum chamber with sample-to-detector distance (SDD) varied between 0.115 m and 1.47 m depending on the configuration used, as calibrated using silver behenate (d001 = 58.380 Å).
For GIWAXS, the SDD was 115.4 mm. The angle dependent 2D GIWAXS were conducted via controlling incident angle from 0.1-0.5° with a 0.1° interval.
For time-and illumination-dependent photoluminescence measurements, the samples were mounted into a custom-built sealed holder to maintain an inert atmosphere. To induce halide segregation, samples were illuminated with blue (Thorlabs M405L4, 405 nm), and where mentioned green (Thorlabs M530L3, 530 nm), light-emitting diodes at angles of about 45° to the surface normal. The green light was filtered by a 600 nm short-pass filter and the LEDs were driven by a Thorlabs DC4104 driver. During spectrum acquisition, the blue LED was used as the excitation source for photoluminescence measurements and the green LED was switched off. To track the photoluminescence behavior when the sample was stored in the dark, the blue LED was used intermittently (exposure time ~500 ms) as the excitation source and was otherwise switched S3 off. The photoluminescent light collected at 90° from the surface normal, filtered by a 645 nm longpass filter, was focused onto an optical fiber connected to the spectrometer (Avantes Avaspec-2048x14) operated on a custom-built code in the LabVIEW environment. To measure photoluminescence behavior at elevated temperatures, the sample was loaded onto a heating stage (Linkam THMS 600) continuously flushed with nitrogen gas.
Solid-state NMR experiments were performed at room temperature on an 11.7 T Bruker Advance III spectrometer using 1.3 mm outer-diameter rotors. Isotropic 207 Pb spectra were obtained by summing the rows of sheared 2D PASS spectra, which separate the isotropic and anisotropic 207 Pb chemical shifts in a two-dimensional experiment. Five-pulse PASS spectra were acquired at 24 kHz MAS, 1 with a single rotor period for the PASS block, a radiofrequency amplitude of 250 kHz, eight increments in the indirect dimension, and a recycle delay of 0.1 s.
Between 48,000 and 186,000 scans were acquired per increment, depending on the sample. Spectra    Table S1), each with a full-width-at-halfmaximum of 220 ppm taken from the experimental spectrum of PEA2PbI4. In the bottom spectra, the weights of each configuration are given by a single binomial distribution of halides within the nominal composition of the sample (x). The middle spectra represent a model where the sample is assumed to be comprised of two sets of Pb(BrxI1−x)6 octahedra with different average compositions (x1, x2) in different proportions, that averaged together give the nominal sample composition (for (a), x = 73% × 0.20 + 27% × 0.70 = 0.33; for (b), x = 58% × 0.25 + 42% × 0.85 = 0.5). The weights of each configuration are calculated from the binomial distribution for the two sets then combined via the relative proportions. The weights of each configuration are calculated from the binomial distribution for the two regions then combined via the relative proportions. The experimental spectra (top) are as in Figure 1b. Although this simple model does not exactly match experiment, which could also result from limitations in the chemical shift calculations, it nevertheless supports the hypothesis of halide clustering. Importantly, the simulated spectra with a random halide distribution are not consistent with the experimental results.     Figure S8. Unnormalized PL spectra for the data displayed in Figure 5. The data is stacked and plotted in log scale for clarity. Table S1. The 18 possible configurations of [PbIaBr6−a] octahedra for n = 1 layered mixed-halide PEA2Pb(BrxI1−x)3 perovskites, assuming the two axial sites are equivalent, the four equatorial sites are equivalent, but the axial and equatorial sites are inequivalent. The probability of each configuration for an x = 0.5 sample with a random binomial distribution is shown. The calculated isotropic shieldings are converted to isotropic chemical shifts by interpolating between the experimental shifts of the pure bromide and pure iodide samples (δ = 5147.4 -0.57266σ).