Photoluminescence Study of the Photoinduced Phase Separation in Mixed-Halide Hybrid Perovskite CH3NH3Pb(BrxI1−x)3 Crystals Synthesized via a Solvothermal Method

We systematically synthesized mixed-halide hybrid perovskite CH3NH3Pb(BrxI1−x)3 (0 ≤ x ≤ 1) crystals in the full composition range by a solvothermal method. The as-synthesized crystals retained cuboid shapes, and the crystalline structure transitioned from the tetragonal phase to the cubic phase with an increasing Br-ion content. The photoluminescence (PL) of CH3NH3Pb(BrxI1−x)3 crystals exhibited a continuous variation from red (768 nm) to green (549 nm) with increasing the volume ratio of HBr (VHBr%), corresponding to a variation in the bandgap from 1.61 eV to 2.26 eV. Moreover, the bandgap of the crystals changed nonlinearly as a quadratic function of x with a bowing parameter of 0.53 eV. Notably, the CH3NH3Pb(BrxI1−x)3 (0.4 ≤ x ≤ 0.6) crystals exhibited obvious phase separation by prolonged illumination. The cause for the phase separation was attributed to the formation of small clusters enriched in lower-band-gap, iodide-rich and higher-band-gap, bromide-rich domains, which induced localized strain to promote halide phase separation. We also clarified the relationship between the PL features and the band structures of the crystals.


Results and Discussion
To analyse the phase structure transformation of the mixed-halide perovskite crystals with increasing V HBr % from 0 to 100%, the XRD patterns of CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals are shown in Fig. 1(a), and the magnified patterns in 2θ = 27.5 −31° are shown in Fig. 1(b). According to the XRD patterns in Fig. 1(a), the CH 3 NH 3 PbI 3 and CH 3 NH 3 PbBr 3 crystals with V HBr % for 0% and 100% have a tetragonal phase structure with the I4/mcm space group and a cubic structure phase with the Pm m 3 space group, respectively, which agree with previous reports 20 . The diffraction peaks show a systematic shift to higher scattering angles with increasing V HBr %, which indicates a decrease of the unit cell size with increasing bromine content. Because the gradual substitution of the larger I atoms with the smaller Br atoms decreases the lattice spacing. The XRD peaks in Fig. 1(a) are also relatively sharp and no peaks of impurities were detected, thus these materials are good crystals and high purities. CH 3 NH 3 PbI 3 (V HBr % = 0%) has two peaks located at 28.15° and 28.44°, as shown in Fig. 1(b), which are indexed to (004) and (220) planes of the tetragonal phase. The (004) diffraction peak gradually disappears and finally merges into a single peak upon increasing V HBr % above 20%, corresponding to the (200) plane of the cubic phase, which confirms that the symmetry of phase structure improve. Further substitution of I ion with Br ions into the tetragonal phase of CH 3 NH 3 PbI 3 causes the systematic shift of the (200) peak towards higher scattering angle. In other words, the tetragonal (pseudo-cubic) phase can transform into the cubic phase with increasing V HBr %. It ascribe to that the smaller halide ion radius is favourable for the formation of the cubic structure, which is accepted as a criterion for the distortion of the PbX 6 octahedra 16 .
The lattice parameter a of the CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals indexed by pseudo-cubic or cubic symmetry as a function of V HBr % is shown in Fig. 2. The a gradually decreases from 8.89 Å to 5.94 Å with increasing V HBr %, which confirms that the lattice spacing decreases with increasing Br ions. Moreover, the slope displays an obviously abrupt change from the tetragonal to cubic phase upon increasing V HBr % from 10% to 20%. The a of CH 3 NH 3 Pb(Br x I 1−x ) 3 exhibits a linear relationship above 20% of V HBr %, as shown in Fig. 2. This linear trend Figure 1. (a) The XRD patterns of CH 3 NH 3 Pb(Br x I 1−x ) 3 obtained with V HBr % for 0%, 10%, 20%, 40%, 50%, 60%, 80%, and 100%, (b) The XRD patterns of CH 3 NH 3 Pb(Br x I 1−x ) 3 magnified in 2θ from 27.5° to 31° (the subscript c is defined as cubic phase, and the subscript t is defined as tetragonal phase).
satisfies Vegard's Law 31,32 , and thus the lattice parameter changes linearly with composition of the perovskite. In general, single-phase mixed-halide perovskite CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals were synthesized by the facile solvothermal method.
To further analyse the morphologies and compositions of the CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals, SEM images of crystals with V HBr % for 0%, 50%, and 100% and the corresponding EDS spectra are shown in Fig. 3. Furthermore, the morphologies of crystals with others V HBr % for 20%, 40%, 60%, and 80% are shown in Fig. S1. The SEM images show that all of the as-synthesized crystals retain cuboid shapes. Length of side of the cuboid shapes become gradually shorten from 3-5 μm to 1-3 μm with increasing the V HBr %, which implies the crystal structure variation. The EDS spectra with V HBr % for 0%, 50%, and 100% show that the composition ratio of I + Br to Pb is 2.796, 2.98, and 2.719, respectively, which are slightly different from the previous reports 12 , because the iodine or bromine atoms can possibly escape and metallic Pb can separate from the perovskite crystal under the experimental conditions of EDS 33 . The composition ratio of Br to I + Br is 50% in Fig. 3(b), which indicates that V HBr % agree well with the predicted × (the composition ratio of Br to I + Br).
To further confirm the compositions of CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals with V HBr % from 0 to 100%, we measured the XPS spectra of the crystals, the XPS full scan spectra and detailed spectra of Pb 4f, I 3d and Br 3d are provided in Fig. S2. The average compositions were calculated using the XPS peak areas of I 3d, Br 3d, the area of I 3d shorten and Br 3d increase, which confirm that the composition percent of I lacked and Br increased depending on volume ratio of HBr. Table 1 shows the elemental composition in CH 3 NH 3 Pb(Br x I 1−X ) 3 crystals with increasing V HBr % from 0 to 100%, the composition ratio of Br to I + Br increased from 0 to 1. The results again show that the V HBr % is nearly equal to the composition ratio of Br to I + Br, that is x. It should be noted that this is the first demonstration of the synthesis of mixed-halide perovskite CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals in the entire composition range via solvothermal method.
The varied of composition could be influenced the band gap or optical properties of CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals, we measured the PL spectra of crystals with increasing V HBr %, as shown in Fig. 4(a). The PL peaks of the crystals for pure CH 3 NH 3 PbI 3 and CH 3 NH 3 PbBr 3 with V HBr % for 0% and 100% point at 768.09 nm and 548.82 nm, respectively, corresponding to bandgaps (E g ) of 1.61 eV and 2.26 eV. A systematic shift of the PL spectra for CH 3 NH 3 Pb(Br x I 1−x ) 3 to shorter wavelengths was observed with increasing V HBr %, which declare that the E g can be tuned from 1.61 eV to 2.26 eV by adjusting the halide content confirming in Table 1. And the colours of the crystals also change correspondingly from dark brown for CH 3 NH 3 PbI 3 to brown-red for CH 3 NH 3 Pb(Br x I 1−x ) 3 and then to yellow for CH 3 NH 3 PbBr 3 upon increasing the Br ions, as shown in Fig. S3. Furthermore, the PL spectrum of the crystals at 50% V HBr % shows two emission peaks in Fig. 4(a), which imply that the crystal possibly comprises two phases 25 .
The E g variation with V HBr % in CH 3 NH 3 Pb(Br x I 1−x ) 3 is plotted in Fig. 4(b). According to previous studies 16 , the E g nonlinear variation with the composition x (V HBr %) in the alloy can be expressed by the following quadratic equation (eq. 1): g  3  3  x1 x 3  g  3  3  3  g  3  3  3   g  3  3  3  2 where b is the bowing parameter, which depends on the properties of the inter-substitutional atoms 34 . The bowing parameter illustrates the fluctuation degree in the crystal field or the nonlinear effect arising from the anisotropic nature of binding 35 . A least-squares fit (red line) of the E g in Fig. 4(b) yields bowing parameter of b = 0.53 eV, resulting in eq. 2. The experimental values of E g agree well with values of the least-squares fit below 50%, as shown in Fig. 4(b), which shows that CH 3 NH 3 PbI 3 and CH 3 NH 3 PbBr 3 have good miscibility. Above 50%, CH 3 NH 3 PbBr 3 is predominant in growth of the crystals, which confirms in Table S1.
To further understand the origin of the PL feature, Fig. 5 shows the PL spectra of the crystals with V HBr % for 40%, 50%, and 60% under sequential illumination. Initially, the perovskite with 40% displays an emission peak at   Similarly, for crystals with 50%, we found that the emission peak at 549 nm grows greatly in intensity with prolonged illumination, as shown in Fig. 5(b), which ascribe to enriched bromine content. In addition, the emission from the Br-rich region is significantly weaker than that of the iodide-rich region, as shown in Fig. 5(a,b), because charge carriers caused by illumination are quickly transported and accumulated at the iodide-rich region 26 .
When increasing the V HBr % to 60%, however, the peak at 547.2 nm decays, and another peak at 716.1 nm grows with prolonged illumination, as shown in Fig. 5(c). The increase in intensity of the lower-band-gap PL peak suggests that these iodide-enriched regions (defects), which act as recombination centre traps, have higher luminescence efficiency than the rest of the perovskite crystals. In other words, I-rich regions serve as the primary charge-carrier recombination sites or irrespective of the carrier generation site in the mixed-halide systems. Similar arguments of charge transfer between Br-rich and I-rich regions as well as trap-initiated recombination have been proposed in earlier studies 25 .
According to reports by Slotcavage 36 and the experimental results, we speculated that halide phase separation of CH 3 NH 3 Pb(Br x I 1−x ) 3 in small halide-enriched domains is induced by sequential illumination, and the localized strain further promotes halide phase separation, which based on halide migration and possibly caused by photo-excited charge interactions. And the halide migration in perovskites is thought to occur through halogen vacancies 37 . This instability may limit the achievable voltages resulting in degraded performances of related photovoltaic devices.
It is again confirmed that halide phase separation occurs in the time-resolved photoluminescence (TRPL) measurements, as shown in the inset in Fig. 5. The TRPL spectra show that the higher-energy band (red line) decays more rapidly than the lower-energy band (black line), which indicates that the initially formed mixed-halide perovskite, with 40 to 60% bromide content, are comprised of two species or two phases. To testify the range of V HBr % for photoinduced phase separation, we have synthesized CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals with different V HBr % and treated them with continuous illumination. The results demonstrate that the crystals display phase separation with V HBr % from 40% to 60%, as shown in Fig. 5, the crystals with V HBr % = 20% and 80% did not display phase separation, as shown in Fig. S4. Moreover, we should state here that the photoinduced change of PL spectra never return to original status after keeping sample for several hours in dark at room temperature, which is inconsistent with previous reports 25,36,38 . It ascribe to that the crystals have the larger crystallite size and higher crystalline quality, which reduce ion migration while enhancing the stability of perovskite materials 39 . Furthermore, the above experimental observations in the optoelectronic properties with various halide contents provide insight into the tuneability of mixed-halide perovskite. To clarify the relationship between the PL feature and the band structure, in this work, we using first-principle calculations study the band structure under variable doped composition conditions, based on experimental lattice parameters. Considering the composition of the unit cell, we focused on the pseudo-cubic phase of the mixed-halide materials of CH 3 NH 3 Pb(Br x I 1−x ) 3 with x = 0.333 and 0.667, and used the band gap approximation of them to fit the result of band gap of CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals with V HBr % for 40% and 60% 40,41 . Figures 6 and S5 illustrate the results of the band structure, partial density of states (PDOS) and total density of states (DOS) for the CH 3 NH 3 Pb(Br x I 1−x ) 3 with x = 0.333 and 0.667. It is found that the valence band maximum (VBM) originate mainly from the strong interaction of the Br-4p, I-5p, Pb-6s and Pb-6p states, the conduction band minimum (CBM) is mainly composed of Pb 6p states for x = 0.333 and 0.667. Moreover, the addition of Br introduces Br 4p states in the VBM whereas the VBM mixed contribution from I 5p and Br 4p in x = 0.333 and 0.667. And the Pb 6p contribution at CBM is unchanged at all systems. The results also show that the E g of the bromine doped by 33.3% and 66.7% are 1.53 eV and 1.61 eV in Fig. 6, respectively. However, the range of band gap of the Br doped by 33.3% and 66.7% are 1.53-3.22 eV and 1.61-3.60 eV. Thus, they can be interpreting the PL patterns of the bromine doped by 40% and 60% corresponding with E g values (1.7~2.32 eV and 1.73~2.26 eV, respectively), as shown in Fig. 5(a,c).
In summary, The mixed-halide hybrid perovskite CH 3 NH 3 Pb(Br x I 1−x ) 3 (0 ≤ x ≤ 1) crystals have been systematically synthesized by a solvothermal method through adjusting concentration of Br ions. The XRD indicated that the crystalline structure transitioned from the tetragonal phase to the cubic phase with the introduction of Br ions, and the crystals have higher crystallinities and purities. The SEM showed that all the as-synthesized crystals retain cuboid shapes. Furthermore, PL peaks of the CH 3 NH 3 Pb(Br x I 1−x ) 3 crystals could be tuned from 768 nm to 549 nm, corresponding to a variation in the bandgap from 1.61 eV to 2.26 eV. Moreover, CH 3 NH 3 PbI 3 and CH 3 NH 3 PbBr 3 had good miscibility below 50% about V HBr %. Notably, CH 3 NH 3 Pb(Br x I 1−x ) 3 (0.4 ≤ x ≤ 0.6) crystals obviously appear phase separation by prolonged illumination due to the formation of small clusters enriched with lower-band-gap, iodide-rich and higher-band-gap, bromide-rich domains, which induced localized strain to promote halide phase separation. In addition, the electronic band structures of the crystals were used to explain many of peaks in the PL patterns with V HBr % about 40% and 60%. Meanwhile, modifying the perovskite morphology and crystallinity greatly improved the stability.

Experiments Methods
Synthesis. All chemical reagents (analytical grade) were directly used without further purification and were supplied by Sigma-Aldrich.
The similar experimental process used here had been reported in our previous work 33 . Pb(Ac) 2 ·3H 2 O (60 mg, Ac − = CH 3 COO − , 99.9%) was completely dissolved in a mixed solution of hydroiodic acid (HI, 45% in water) and hydrobromic acid (HBr, 40% in water). Then, 30 mL of isopropanol (IPA, 99.9%) was added and stirred for 5 min, and 0.3 mL of a methylamine solution (CH 3 NH 2 , 30% in water) was added dropwise. The mixture was further stirred for 5 min and then put into 50 mL stainless steel Teflon-lined autoclave, and was sealed and heated in furnace at 150 °C for 4 h, after cooling naturally to room temperature. The precipitates were collected and washed with isopropanol by centrifugation at room temperature, and then were dried under vacuum at 60 °C for 4 h. We mixed the solution in various volume ratios of HI and HBr, and V HBr % is defined as the ratio of V HBr :V HI+HBr (the total volume of HI and HBr is 1 mL).
Characterization. The structures of the products were investigated by X-ray diffraction (XRD, X'TRA) using Cu Kα radiation (λ = 0.1542 nm). The X-ray tube voltage and current were set at 40 kV and 40 mA, respectively. The morphologies and elemental analyses of the products were observed by field-emission scanning electron microscopy (FE-SEM, JSM-7000 F) in energy-dispersive spectroscopy (EDS) mode. X-ray photoelectron spectroscopy (XPS, PHI 5000 Versa Probe) was used to identify the elemental compositions of the products, and the resolution of the spectrometer was chosen to be 0.6 eV with a pass energy setting of 40 eV. The photoluminescence (PL) spectra of the products were recorded on a HORIBA iHR 320 fluorescence spectrophotometer with an excitation wavelength of 375 nm at room temperature. The 375 nm line of a picosecond pulsed laser diode (PicoQuant PDL 800-D) was used as the excitation light source for time-resolved PL measurements, and the PL decays were recorded by a time-correlated single-photon counting module and a picosecond event timer (PicoHarp 300)

Computational Methods
The band structures of the pseudo-cubic phase CH 3 NH 3 Pb(Br x I 1−x ) 3 are calculated within the framework of density functional theory by using the CASTEP package. Norm-conserving pseudopotentials and Perdew-Burke-Ernzerhof (PBE) functional with the generalized gradient approximation (GGA) were used to model the electron-ion interactions and exchange-correlation potential, respectively 42,43 . We focused on the pseudo-cubic phase of the mixed-halide materials of CH 3 NH 3 Pb(Br x I 1−x ) 3 with x = 0.333 and 0.667. The high cutoff energy for the plane-wave basis is set at 750 eV and the Brillouin zone is sampled by a 5 × 5 × 5 k-point sampling grid. The convergence tolerance of maximum force, maximum displacement and energy were 0.01 eV/Å, 5.0 × 10 −4 Å and 5.0 × 10 −6 eV/atom, respectively. These parameters were controlled to ensure convergence 40 .