Monitoring Phase Separation and Dark Recovery in Mixed Halide Perovskite Clusters and Single Crystals Using in-situ Spectro-Microscopy

Abstract

Mixed halide perovskites (MHP) are a group of semiconducting materials with promising applications in optoelectronics and photovoltaics, whose bandgap can be altered by adjusting the halide composition. However, the current challenge is to stabilize the light-induced halide separation, that undermines the device’s performance. Herein we track down the phase separation dynamics of CsPbBr_1.2I_1.8 MHP single cubic nanocrystals (NCs) and clusters as a function of time by in-situ fluorescence spectro-microscope. The particles were sorted into Group 1 and Group 2 using initial photoluminescence intensities. The phase separation followed by recovery kinetics under dark and photo blinking analysis suggests that Group 1 behaved more like single NCs and Group 2 behaved like clusters. Under the 0.64 W/cm² laser illumination, the phase shifts for single NCs are 3.4±1.9 nm. The phase shifts are linearly correlated with the initial photoluminescence intensities of clusters, suggesting possible inter-particle halide transportation.

Graphical Abstract

Semiconducting lead halide perovskites have attracted enormous attention due to their unique optical versatility,¹ high photoluminescence quantum yields (PLQY),² and facile synthesis.³ Their crystal structure has the general formula of ABX₃ (where A is CH₃NH₃⁺, Cs⁺ or FA⁺, etc.; B is Pb²⁺ or Sn²⁺, etc.; and X is I⁻, Br⁻ or Cl⁻),⁴ named from their same crystal structures as the naturally occurring perovskite CaTiO₃. The first lead perovskite reported for solar cell application is organic-inorganic hybrid perovskites CH₃NH₃PbBr₃ and CH₃NH₃PbI₃.⁵ A = Cs⁺ is all-inorganic perovskites and crystals with mixed Xs are called mixed halide perovskites (MHPs). MHP’s halide stoichiometry can be adjusted accordingly to tune bandgaps, presenting a wide range of promising performances in photovoltaics and optoelectronics such as LED,^6–8 lasers,⁹ and photodetectors.^10,11

A major challenge of perovskite materials is their instability issue. A variety of conditions, such as polymorphic transitions, decomposition, hydration, and oxidation affect the stability of metal halide perovskites.^12,13 All-inorganic perovskites are more resistant to moisture but suffer more from crystal lattice stress than their organic-inorganic hybrid counterparts.^14,15 Cs in MHP offers a recipe to enhance its versatility ¹² and extended emissive lifetimes than single halide ones.^{6,8,10,11,16–18} Recently reported that substitution of Rb⁺ with Cs⁺ destabilizes the perovskite crystal structure by enhancing the octahedral tilting.¹⁹ In addition to degradation by oxygen,^20–22 and humidity,^23–25 light²⁶ driven phase separation has been an important observation in MHP. Under constant illumination or charge carrier injection, MHP undergoes halide species separation into different I-rich and Br-rich regions.¹² Consequently, this ion migration leads to the display of contrasting bandgap energies compared to that of the original mixture. It is believed that the initiation of halide migration occurs by the accumulation of photoexcited charge carriers on grain boundaries providing a driving energy or an internal bias allowing halide anion mobility,²⁷ without which the material is entropically stable and phases can be remixed in the dark.²⁸ The charge carriers in the de-mixed phases have different energies defined by the local bandgap and prefer to funnel to the phase with a lower bandgap.²⁶ This effect was first observed by Hoke et al. in 2015, discovering a red-shifted photoluminescence (PL) emission peak with a noticeable increase in intensity.²⁶ UV-Vis absorption peak is also shifted to a lower energy region. Halide ion concentration has a major effect on the shift observed in absorption.¹⁸ Previous work has reported that the mixing entropy causes the recovery to the original place.²⁹ On the other hand, MHP degrades irreversibly, when oxidized halide molecules are eliminated via evaporation.³⁰ As a solution, passivating the perovskite film with polymers is required. Phase separation can either impair the device’s performance or enhance its efficiency, but the mechanism of which is still under debate.¹³ XRD²⁶ and UV absorption measurements^18,31–33 estimate that photo-induced phase separation is carried out by as low as ~1% of total halide species.³⁴

In this report, we track down the phase separation and dark remixing of MHP single nanocrystals (NCs) and clusters spun-coat on glass substrates over time using a home-built spectro-microscope. We synthesized CsPbBr_1.2I_1.8 MHP NCs using a conventional hot injection method with slight modifications.³⁵ The detailed procedure is listed in the Supporting Information. The purified MHP NC colloidal solution shows a red color under room light and an orange color under UV light (Figure 1a, 1b). To confirm the composition of MHP NCs, an energy-dispersive X-ray spectroscopic (EDX) analysis was performed. The corresponding peaks are shown in Figure S1 in the Supporting Information. Details of EDX spectra values measured in weight% and atomic% are listed in Table S1 in the Supporting Information. The EDX data shows an atomic % ratio among Cs:Pb:Br:I = (0.8±0.1):(1.5±0.1):(1.1±0.2):(1.5±0.3) that are consistent with the theoretical values 1:1:1.2:1.8 with a smaller Cs: Pb ratio than expected. The MHP composition was further analyzed using inductively coupled plasma optical emission spectroscopy (ICP-OES). The elemental ratio of Cs:Pb:Br:I was calculated to be (0.9±0.1):(1.0±0.2):(1.2±0.2):(1.7±0.2), consistent with the theoretical values. The detailed analysis is listed in Table S2 in the Supporting Information. Transmission electron microscopy (TEM) images show that single NCs are cubic shaped with an average edge size of 41.6±7.9 nm (Figure 1c, 1d, Figure S2 in the Supporting Information). Sparse CsPbBr_1.2I_1.8 MHP NCs were prepared by spin-coating the purified solution on a base piranha-cleaned borosilicate glass microscope coverslip. At this stage, NCs were aggregated to form clusters with different sizes. Scanning electron microscopy (SEM) images confirm sparsely located single NCs and clusters at 200–500 nm range (Figure 1e -1h). They are passivated by spun-coat thin polystyrene film. The sample is then measured under a home-built inverted spectro-microscope equipped with an electron multiplying charge-coupled device (EMCCD) camera to acquire PL measurements (Figure S3 in the Supporting Information). The structure and spectra calibration of the microscope have been described before.³⁶ A 473 nm continuous wave (CW) laser is used as the excitation light with power density tuned to 0.64 W/cm². A narrow slit is placed at the image plane, which extracts a thin section of the image with spatially resolved bright spots. Then a transmission diffraction grating that is placed in front of the EMCCD camera bends the light.¹³ The corresponding image of both the 0^th and 1^st order diffraction light is captured by the camera. The first-order image is then used to reconstruct a spectrum of the selected spots (Figure 2). All the PL measurements were obtained in total internal reflection fluorescence (TIRF) mode. Time-dependent PL movies were created using 0.1 s exposure time.

Figure 1.

Photos of purified CsPbBr_1.2I_1.8 MHP NCs (a) under room light and (b) under UV light, (c, d) TEM images of NCs dried on TEM grids, and (e-h) SEM images of spun coat sample.

Figure 2.

Spectro-microscopic (a) images and (b) corresponding spectra of a MHP spot before and after 600 s of continuous illumination, taken from a movie with a frame time of 0.1 s. On the images, the dashed line at the 0^th order image shows the slit dimension and the dashed line at the first order image shows the peak center at time 0 s. The insets are the zoom-in images. The image dimension is 43 μm, 83 nm/pixel. The color scale bar shows the photocounts. The spectral wavelength accuracy is ±1 nm and the resolution is ~ 6 nm estimated from the calibration spectra of noble gas lamps.

Here, we measured the phase separation dynamics and dark recovery kinetics of CsPbBr_1.2I_1.8 MHP single NCs and clusters spun coat on the substrate surface (Figure 3a). Thirty-seven particles from ten different movies were selected with the criteria that there are no competing spots nearby spatially. Figure 3b displays six particles to demonstrate the time-dependent PL evolution. These six particles are numbered P1, P2, etc. First, the MHP thin film was irradiated continuously for 600 s with the excitation beam. The relatively weak laser power 0.64 W/cm² is chosen because a stronger laser induces significant photothermal remixing in our previous report.¹³ Time-dependent PL spectra of these particles are shown in Figure 3c. The PL spectra of the bigger brighter MHP clusters are observed initially at ~ 630 nm, with smaller NCs emitted at ~ 610 nm consistent with the literature report.³⁵ Although both Br-rich and I-rich domains are created during the phase separation, the strong laser will excite the Br-rich phase showing a blue shift,^35,37 while the weak laser will excite the I-rich phase showing the redshift of the PL peak.^13,38 Throughout the time of illumination, the PL intensity decreased drastically as shown in the PL intensity profile (Figure 3d) then peak shifts during <100 s (Figure 3c). The recovery kinetics of the peak shift over time are monitored by turning off the laser after 600 s and re-flashing for 1 s every 180 s to track the recovery to the original MHP PL peak in the dark. The kinetic spectra evolution over time under illumination and recovery in the dark for P1-P6 are attached in Figure S4 in the Supporting Information. The recovery lifetime of each particle was determined by fitting the data points using a single exponential decay model: $P=A{e}^{-kt}+P_0$ where $P$ is the peak position at time $t,k$ is the rate constant, and $P_0$ is the initial peak position, and $A=P_{\mathrm{m}\mathrm{a}\mathrm{x}}-P_0$ is the maximum amplitude of the peak shift. The final peak position after 600 s of separation is taken as the initial (t = 0 s) decay peak position. (see Figure 5a, 5b, and section 2 in the Supporting Information for the details). These results are also summarized in Table S3 in the Supporting Information. The MHP particle size was correlated with the initial PL intensity of the emission peak to get an idea of how big it was. We assume that the number of nanocrystals in each cluster is proportional to their initial PL intensity before light enhancing, quenching, or shifting the PL spectra.³⁹ Hence, the initial PL spectra for the MHP particles were obtained by flashing the laser beam to the sample and acquiring the PL signal at t = 0 s. The initial intensity of the photocounts of the peak and the peak position of the wavelength of the maximum are listed in Table S3 in the Supporting Information.

Figure 3.

(a) A typical PL image of spun coat MHP particles. (b) PL images of six selected spots (P1-P6) with PL intensity normalized to the maximum value of the six. (c) Time-dependent PL spectra of these CsPbBr_1.2I_1.8 MHP particles under 0.64 W/cm² illumination showing the PL peak evolution in 0–200 s (zoomed in from the 600 s data for better contrast). The redshift is illustrated by the bending of the PL peak towards the longer wavelength on the right. The zoomed-in image of the first 2 s of illumination is attached below the dashed line for each spot. The arrow shows the initial PL peak at t = 0 s. The color bar symbolizes the normalized PL intensity. (d) The PL intensity of the six MHP particles over time integrated from (c).

Figure 5.

Peak center of selected (a) Group 1 and (b) Group 2 particles under continuously excited power density of 0.64 W/cm² from t = 0 s to 600 s (left) and dark recovery (right) by blocking the laser beam and exposing for 1s every 180 s. Correlations between the recovery lifetime (τ) of MHP particles with (c) initial PL intensity of emission peak, and (d) maximum phase shift. The red dots and dashed circle on the right are Group 1 particles while the purple dots and dashed circle highlight the Group 2 particles.

Based on the initial PL intensities, the particles were grouped into two categories using 95% confidence. Group 1 consists of particles that have the lower initial PL intensities between 1000–6000 photocounts, and the ones beyond the upper limit are included in Group 2. Next, we correlated the amplitude of the peak position shift after 600 s of illumination with the initial PL intensity of each spot. A strong positive correlation was observed between peak shifts and initial PL intensities (Figure 4b). Group 2 particles exhibited a contrastive red shift compared to the ones in Group 1. In the photo blinking analysis, we classified the upper part, the middle part, and the lower part of the PL trajectory and added two thresholds in between to divide the ON, GRAY, and OFF states following the literature suggestion (Figure S6 in the Supporting Information).⁴⁰ Most of the Group 1 particles showed blinking with OFF states, while the ones in Group 2 showed relatively smooth decays in intensity (Figure S7 in the Supporting Information). Lastly, we correlated the recovery lifetimes with initial PL intensity and phase shift. Group 2 behaved linearly while Group 1 particles exhibited variations.

Figure 4.

(a) Initial spectra at time 0 s of all selected particles. (b) Correlation between initial PL intensity and photo-induced phase separation after 600 s of laser illumination. The red dots and dashed circle are Group 1 particles while the purple dots and dashed circle highlight the Group 2 particles. (c) A Group 2 particle shows a smooth decay with time. (d) The PL intensity switches between the “ON” and “OFF” states of photo blinking of a selected Group 1 particle.

We hypothesize that Group 1 and Group 2 are consolidated with the behavior of single NCs and clusters, respectively based on two reasons. First, the initial peak intensities of Group 1 particles are normally distributed at 3000±1050 photocounts while those in clusters are higher. Second, Group 1 particles blink, while each particle in Group 2 should blink normally but overall, they blink at different frequencies such that the PL curve is smooth. In our previous study, larger particles at a few hundred nanometers blink at different locations that are still significantly blinking overall.⁴¹ To confirm if Group 2 are clusters rather than bigger fused particles, further investigation such as SEM correlation is needed in the future. We further observed that Group 1 particles show small peak shifts in 3.4±1.9 nm (Figure 4b) with various initial peak positions from 610–620 nm (Figure 5a). Group 2 particles have initial peak positions of ~ 630 nm. They have significant peak shifts from 13–36 nm showing a linear correlation with the initial peak intensity (Figure 4b). The peak shift recovery lifetime of Group 1 particles scattered around 0–1000 s while Group 2 clusters have their lifetime linearly correlated with their initial peak intensity (Figure 5c, 5d) as well as the amplitude of the phase shift. The large noise on recovery data of Group 1 particles (Figure 5a) and the big variation in lifetimes (Figure 5c) are further analyzed. Within the 1 s observation time, a significant phase shift is observed for single NCs while phase stability is seen for clusters in a single frame of 0.1 s (Figure S11 in the Supporting Information).

In our previous work, a fast initial phase shift in millisecond time scale was observed for MAPb(I_1−xBr_x)₃ thin films that also exist as aggregates in contrast to the peak shifts < 100 s in CsPbBr_1.2I_1.8 clusters and single NCs.^13,28 We don’t see the PL from the Br-rich domain which is consistent with literature reports done under relatively low excitation power.^42,43 This suggests that a large part of photogenerated charge carriers is funneled to the lower band-gapped I-rich regions to undergo radiative recombination. Besides, the binding energy associated with ionization of the longer and weaker Pb-I bond ^33,44–46 is low compared to that in the Pb-Br bond.^47,48 The reason for the initial drop in PL intensity could be non-radiative Auger recombination centers creating trap states that dissipate excitons before emitting photons,^35,49 consistent with our previously reported trap formation dynamics on single-crystal perovskite nanorods.⁴¹

The high phase stability of single NCs is expected according to the literature-reported lower threshold of particle size for a significant phase shift to be 46±7 nm at room temperature, a balance between bias-induced halide migration ^50,51 and diffusion-induced thermal remixing.^52,53 However, it is surprising that the Group 2 clusters show significant phase shifts given each particle is protected with a ligand mixture of oleic acid and oleyl amine with clear gaps observed among particles under TEM images. The single NC peak shift kinetics varies from particle to particle and is expected to be affected by factors such as surfactant defects and crystal defects.^54–56 Careful investigation is needed in the future to measure the recovery lifetimes of Group 1 NCs. Surprisingly, a strong correlation between the size and the lifetime of recovery is observed for Group 2 clusters.

We hypothesize that in Group 2, the ligand shells of the NCs partially open during the aggregation and make the perovskite NCs connected, forming a network. The halide ions penetrate through the interphase of the crystal network, to a certain extent, thus the net phase shift is comparatively higher and takes ~600 s to stabilize. While single NCs have relatively stable PL intensities, the PL of clusters drops significantly and quickly in 0.5 s (Figure 3c). This quick PL drop in cluster is unlikely from local surface defects which should be the same among particles but rather supports our hypothesis that an inter-particle network has been established to provide an additional pathway for carrier separation. Both the recovery lifetime and the maximum peak shift are positively correlated with the size of the network. In the future, colocalization of PL images with SEM images and more statistics will offer direct evidence on the correlation between phase shift and aggregation.

In conclusion, when we measure the phase separation and recovery of MHP single NCs and clusters, a surprising collective effect for clusters is observed. The sizes of the clusters, the recovery lifetime of PL peak shifting, and the maximum peak shift are all positively correlated with each other. These results suggest that the aggregated NCs opened their ligand shells to form a MHP network that allows long-distance ion migration under excitation. This could be an important factor to consider and require further quantification for MHP nanoparticle/quantum dots-based solar cells, LEDs, lasers, and photodetectors.