Thin Film Stoichiometry and Defects Management for Low Threshold and Air Stable Near-Infrared Perovskite Laser

Abstract

While significant efforts have been devoted to optimize the thin-film stoichiometry and processing of perovskites for applications in photovoltaic and light-emitting diodes, there is a noticeable lack of emphasis on tailoring them for lasing applications. In this study, it is revealed that thin films engineered for efficient light-emitting diodes, with passivation of deep and shallow trap states and a tailored energetic landscape directing carriers toward low-energy emitting states, may not be optimal for light amplification systems. Instead, amplified spontaneous emission (ASE) is found to be sustained by shallow defects, driven by the positive correlation between the ASE threshold and the ratio of carrier injection rate in the emissive state to the recombination rate of excited carriers. This insight has informed the development of an optimized perovskite thin film and laser device exhibiting a low threshold (≈ 60 µJ cm⁻²) and stable ASE emission exceeding 21 hours in ambient conditions.

1 Introduction

Lead halide perovskites have emerged as potential candidates for next-generation on-chip light amplification devices due to their remarkable optoelectronic properties. Promising performances based on self-assembled microcavity,^[1-3] distribute feedback cavity,^[4-6] distribute Bragg reflector cavity,^{[7, 8]} and thin film random lasers^{[9, 10]} have been shown. Despite many reports having already achieved optically pumped lasing in perovskite devices based on light-emitting diodes (LEDs) structure,^[11-13] electrical operation in perovskite laser devices is still a goal that has not been reached. The operating philosophy significantly affects the aspects to select a proper active layer material. A good LED device requires a material with high radiative recombination efficiency, and thus a material with maximized Photoluminescence quantum yield (PLQY).^[14] Yet, for selecting an ideal gain medium material for laser devices, the crucial factor in selecting a suitable low-threshold laser material lies in how easily population inversion can be achieved, where the excited carrier could easily accumulate at the band edge.

The development of perovskite gain medium suffers from an insufficient understanding of the correlation between thin film chemical composition and excited carrier dynamics. In the research of perovskite luminescence performances, Shockley–Read–Hall (SRH) model is often used to interpret the semiconductor thin film behavior,^{[15, 16]} where time-resolved photoluminescence spectroscopy (tr-PL) and PLQY measurements are widely applied to analyze the carrier recombination dynamics. However, in the development of optical gain medium, the threshold for stimulated emission is influenced not only by the recombination rate but also by the carrier relaxation within the excited state. In this work, we investigate formamidinium lead iodide (FAPbI₃) perovskite thin films, a state-of-the-art model system for light emitting diodes, where trap density management can be achieved by tuning the FA⁺ cation stoichiometry.^{[17, 18]} We show that excess formamidinium iodide (FAI) salt has a role in the passivation of carrier traps, thus on the enhancement of PLQY, especially in low excitation density regimes. However, for high excitation densities close to the amplified spontaneous emission (ASE) threshold condition, we discovered that the threshold increases corresponding to the passivation of shallow traps. By expanding on the three-level system model for stimulated emission, we clarify that the excess FAI salt has a passivation effect on both shallow and deep traps, but for laser applications where the system works at high excitation densities, the shallow trap levels have an affirmative effect for achieving low threshold stimulated emission.

Our goal, with this work, is to provide a complete model to illustrate stimulated emission processes and evaluate the performance of a perovskite gain medium as a function of its chemical composition, trap nature and thin film morphology.

2 Results and Discussion

2.1 Optical Properties of Perovskite Thin Film

In this work, we refer to a maturely developed FAPbI₃ perovskite films (see the Experimental Section for details) for perovskite LED applications,^[19] where trap density can be controlled through adding excess of FA⁺ cation to passivate the traps.^[17-19] FAPbI₃ perovskite thin films are deposited from perovskite precursors of PbI₂ (0.3 m) mixed with different concentrations of additional FAI salt (marked as FAx, here x denotes the molar ratio between FAI and PbI₂ in perovskite precursors). Figure S1 (Supporting Information) shows X-ray diffraction (XRD) patterns of the perovskite films with different amounts of FAI. All the perovskite diffraction peaks can be assigned to their cubic phases. The perovskite films of FA 1.2 and FA 1.6 show high intensity of (100) peak (at around 13.9°). It decreases when higher FAI concentrations (FA 2.0 and FA 2.5) are used, which would be the result of forming low-dimensional phases perovskites in the films.^[20]

We then investigated the photoluminescence (PL) properties of these films. We measured the power-dependent PLQY to investigate the role of trap carrier states since it allows us to examine the recombination dynamics in the deep trap-limited, bimolecular, and Auger–Meitner recombination regions. A PLQY close to 0 indicates poor emission.^[17] Then, increasing the excitation density, the recombination dynamics align with band-to-band bimolecular recombination, and the critical point on the curve represents the region where most traps are filled by the photocarriers and band-to-band and Auger–Meitner recombination processes kick in.

In Figure 1a, within the low carrier density regime, we note a rise in the absolute PLQY. This observation suggests that carrier recombination is predominantly governed by monomolecular trap-mediated processes. There is also a noticeable shift of the peak by one order of magnitude toward lower carrier density as the excess FAI concentration increases. This shift indicates a decrease in deep trap density. Furthermore, the absolute PLQY values are depicted in Figure 1a, exhibiting an augmentation in PL intensity with higher FAI concentration. These findings align with previous studies,^[17] which suggest that the introduction of excess FAI salt can diminish deep trap density and thereby improve spontaneous emission characteristics, which is advantageous for high-efficiency LEDs.

Figure 1

Optical properties of FAPbI₃ of different excess FAI concentration. a) Absolute PLQY as a function of excitation carrier density. b) Power-dependent PL spectra of FA 1.6 sample, the ASE peak arise at around 60 µJ cm⁻³. c) PL intensity as a function of excitation carrier densities. (Figure plotted with a constant offset of 1 to show the change of the threshold carrier density value.) d) ASE threshold excitation carrier density as a function of excess FAI concentration (Excitation wavelength: 670 nm, all measurement done under ambient condition).

We further increase the excitation carrier density to characterize the ASE performance. Figure 1b shows the PL spectra of the FA 1.6 sample with the increase of optical pump power, showing a strong ASE character with a threshold of 60 µJ cm⁻². Besides, we discovered the ASE threshold carrier density doubles from 1.6 × 10¹⁸ to 3.5 × 10¹⁸ cm⁻³, when the FAI concentration increases from 1.2 to 2.5. According to Figure 1a, the samples with a higher concentration of additional FAI show a lower deep trap density, which may be expected to have a lower ASE threshold. Yet, in Figure 1c,d we examine the power-dependent PL measurement in different FAI concentration samples, finding an unexpected trend. The carrier density threshold as a function of excess FAI concentration is shown in Figure 1d: the ASE threshold carrier density does not decrease according to our expectation, but instead, increases with excess FAI concentration.

2.2 Shallow Trap-Assisted Amplified Spontaneous Emission

To comprehend the role of traps in the perovskite stimulated emission process, it is crucial to first clarify in which regime deep traps and shallow traps are active. Figure S2 (Supporting Information) displays the power law plot of PL intensity, finding a slope of 3/2. This suggests that the radiative recombination process in the low excitation density region corresponds to the interaction between the deep trap, free carriers, and shallow donors or acceptors.^{[16, 21]} This finding implies that a certain amount of carrier trap-assisted recombination occurs in the system at low excitation density through the deep traps. However, as the carrier density rises, deep traps are gradually filled, notably with a carrier lifetime so long that remain filled in the steady states,^{[22, 23]} and only shallow traps remain active. Therefore, we investigated whether shallow trap states are involved in the stimulated emission process, happening at high excitation densities close to ASE threshold regime. Through photothermal deflection spectroscopy (PDS) measurements (Figure 2a) we calculate the Urbach energy U_e for different samples and plot it as a function of FA excess concentration in Figure S3 (Supporting Information). Notably, we observe that as the excess FA cation concentration increases, the U_e decreases. The U_e value represents the amount of shallow trap states, influenced by the crystal disorder. Furthermore, the ASE peak position represents the energy gap between the inverted populated states and the ground states. In Figure 2b, we present the ASE spectra at excitation density close to the ASE threshold, demonstrating a blueshift of the ASE peak position with an increase of excess FAI concentration. Next, to decode the relation between the trap states and stimulated emission states, we compare the differences in U_e with the ASE peak shifts for various concentrations of excess FAI samples, using the FA 2.5 sample as baseline. Figure 2c shows a strong correlation between the ASE peak position and U_e. To exclude that the ASE peak shift is due to the tuning of the bandgap, we compare spontaneous PL emission spectra below treshold, confirming that they do not show any spectral shift (Figure S4, Supporting Information). This clarifies that the ASE emitting states are controlled by the depth of the shallow trap states, as depicted in Figure 2d. In samples with little excess FAI concentration, the U_e tail is deeper in the bandgap, and the ASE peak is also found at a smaller photon energy. In samples with the larger excess of FAI concentration, the shallow Urbach tail corresponds to the blueshift of the ASE peak position.

Figure 2

Urbach energy tail and origin of ASE state. a) Absorption spectra measured by PDS. b) ASE spectra of samples with different excess FAI concentration (Excitation source: 670 nm, P ≈ 1.2P_th). c) Comparison between the Urbach energy difference and ASE peak energy shift, normalized to sample FA 2.5. d) Scheme showing the ASE state is highly dependent on the Urbach energy difference.

This investigation demonstrates the crucial role of trap states in the FAPbI₃ ASE process. The carriers trapped by the shallow defect states, when deep defects are filled, become a significant source for stimulated emission as the carrier density increases, acting like a carrier reservoir for achieving population inversion.

2.3 Three-Level Laser Rate Equation Derivation

To gain deeper insights into the impact of trap passivation strategies on stimulated emission performance, it is imperative to develop a comprehensive carrier relaxation dynamics model and integrate it into the laser rate equation framework. For 3D FAPbI₃ perovskite, the classic three-level laser model is adopted (sketch in Figure 3a) E_VB, E_CB, and E_ex represent the energy levels of the top of the valance band (VB), the bottom of the conduction band (CB), and transient photoexcited states, respectively; N_VB, N_CB, and N_ex represent the corresponding densities of carrier populations.

Figure 3

Relaxation lifetime curve comparison between FA 1.6 and FA 2.5 samples. a) The three-level laser model, the carrier is excited to super-bandgap excited state then relax to band edge in a short time, then accumulated at the band edge. b,c) TA spectra of FA 1.6 and FA 2.5 samples. d) The GSB signal of FA 1.6 and FA 2.5, plotted as a function of pump-probe delay. e) The relation between the lifetime ratio factor β and the ASE threshold carrier density plotted as a function of excess FAI concentration (Excitation wavelength: 670 nm, carrier density: close but below threshold condition, FA 1.6 ≈ 1.5 × 10¹⁸ cm⁻³, FA 2.5 ≈ 3.3 × 10¹⁸ cm⁻³).

The transient photocarriers could be attributed to hot carriers’ relaxation or funneling of carriers from higher laying energy states. We can match these parameters to the three-level laser model shown in Figure S5 (Supporting Information). The transient photoexcited states are the first to be populated after photoexcitation, then the excited photocarriers quickly relax to the CB edge with a relaxation rate γ_ex, and finally the carriers accumulated at the CB edge relax back to the VB with a relaxation rate γ_rec. W_p is the pump excitation rate.

We derive the laser rate equation in detail in Note S1 (Supporting Information). We define the factor $$\beta \equiv \ \frac{{{{\gamma }_{rec}}}}{{{{\gamma }_{ex}}}} = \ ( {\frac{{{{T}_{ex}}}}{{{{T}_{rec}}}}} )$$, which describes the ratio between the injection and recombination rate of CB carrier density.

At the steady state condition, the rate equation can be expressed as

$$$$\begin{equation}\left( {{{N}_{CB}} - {{N}_{VB}}} \right)/N = \frac{{{{W}_p}{{\gamma }_{{\mathrm{ex}}}}\left( {1 - \beta } \right) - 1}}{{{{W}_p}{{\gamma }_{{\mathrm{ex}}}}\left( {1 + 2\beta } \right) + 1}}\ \end{equation}$$$$(1)

The threshold condition happens when carriers accumulated at the CB edge surpass the VB populated carriers: N_CB − N_VB > 0,

$$$$\begin{equation}{{W}_p}{{\gamma }_{rec}} \ge \frac{1}{{1 - \beta }}\end{equation}$$$$(2)

In the ideal case of a three-level system, the transient excited carriers relax immediately (T_ex = 0), and accumulate at the CB edge, thus β = 0. In other words, the lower the β, the easier it is to achieve population inversion.

The transient absorption (TA) spectra and dynamics of thin films with different excess FAI concentration are shown in Figure 3b,c. The samples are pumped at 670 nm (close but below threshold condition, FA 1.6 ≈ 1.5 × 10¹⁸ cm⁻³; FA 2.5 ≈ 3.3 × 10¹⁸ cm⁻³) to obtain near band edge excitation. We can obtain T_rec by monitoring the value of the ground state bleach (GSB) signal as a function of pump-probe delay, which corresponds to the lifetime of carriers that accumulated at the bottom of the CB. The lifetime of the CB carriers can be calculated by exponentially fitting the decay of the GSB peak. On the other hand, the injection rate of carriers towards the bottom of the CB, expressed as $${{\gamma }_{ex}} = \frac{1}{{{{T}_{ex}}}}$$, can be obtained by fitting the rise time of the GSB signal. The GSB dynamics are compared in Figure 3d. We find that FA 1.6 thin films exhibit a longer GSB lifetime compared to FA 2.5 thin film. This phenomenon can be attributed to the high concentration of active shallow traps in the FA 1.6 sample under high excitation density (FA 1.6 ≈ 1.5 × 10¹⁸ cm⁻³; FA 2.5 ≈ 3.3 × 10¹⁸ cm⁻³) which delays the recombination lifetime, serving as a carrier sink for population inversion.^[24] On the other hand, the rise time is remarkably similar, showing similar injection dynamics.

Finally, to quantitatively compare the ASE threshold, we calculate the carrier relaxation lifetime ratio β factor by dividing the two lifetime values. We can see from Figure 3e the lifetime ratio β holds the same increasing trend as ASE threshold carrier density with increasement of excess FAI concentration (detailed parameter shown in Table S1, Supporting Information), indicating that the model monitors effectively the stimulated emission process, and can thus be used as a straightforward methodology to quantitatively compare the threshold of new perovskite materials. This model could also be used to compare threshold condition under different pump excitation wavelengths, as shown in Note S2 (Supporting Information).

2.4 Low-Dimensional Phases and Energy Transfer

The use of excess FAI salt in FAPbI₃ not only is used as a trap passivation strategy, but it also induces the formation of low dimensional phases which needs to be further considered in the system. Detailed formation mechanism is depicted in Note S3 (Supporting Information). The absorption spectra from FA 1.6 and FA 2.5 thin films are shown in Figure S6 (Supporting Information). The extra excitonic absorption features at 500 and 600 nm correspond to the absorption of the low-dimensional FA_n+1PbI_3n+1 phases results from the high excess FAI concentration.

By performing TA spectroscopy, we monitor the photogenerated carrier dynamics in a time range between 100 fs and 3 ns within the perovskite thin film. Figure 4a,b illustrates the TA spectra of FA 1.6 and FA 2.5 samples, photoexcited at 343 nm to provide enough energy to excite the higher lying energy states. In Figure 4a, the TA spectra of the FA 1.6 thin film is shown. We observe two photobleaches (PB) at around 480 nm (PB1) and 780 nm (PB2). The origin of PB1 is still unclear, which requires more investigation on electron spectroscopy, it might be an indirect transition from the neighboring energy valley or higher energy bands, or origin from dual VB, CB.^[25-27] PB2 represents the GSB signal related to the band edge photocarrier population. For the FA 2.5 sample, the TA spectra are shown in Figure 4b. We observe multiple additional photobleach signals at 550 nm (2D-1) and 600 nm (2D-2), corresponding to the photobleach of low-dimensional phases. However, such phenomenon is not observed in the FA 1.6 sample. The presence of low-dimensional phases provides a funneling path that facilitates the transfer of photocarriers from the state corresponding to PB1 to the bottom of the CB.^[20]

Figure 4

UV excited TA spectra of FA 1.6/FA 2.5 samples. a,b) TA spectra of the corresponding photobleaching, excitation density ≈ 2 × 10¹⁸ cm⁻³. c,d) Comparison of the PB2 GSB signal as a function of pump-probe delay time between FA 1.6 and FA 2.5 samples, under different excitation density (c: low ≈ 1.5 × 10¹⁷ cm⁻³, d: high ≈ 2 × 10¹⁸ cm⁻³).

To carefully investigate how the energy funneling effect affects our system, we monitor the GSB signal to obtain the information of the band edge carrier population. The population of the band edge photocarriers could be contributed by various sources, including: (1) hot carriers cooling and (2) carrier funneling from the low-dimensional phases. As they happen simultaneously, it is not easy to isolate each contribution from the data. To better resolve these two processes, the power dependence of the PB2 dynamic of the two samples is performed and shown in Figure 4c,d to prove the existence of a carrier funneling effect. When we excite at lower excitation density (≈1.5 × 10¹⁷ cm⁻³, shown in Figure 4c), the hot carrier population relaxes without a bottleneck effect slowing down the carrier recombination.^[28] Also, as seen in Figure S7 (Supporting Information), the bandwidth of the PB2 in both samples barely changes even at early times, indicating that the contribution of hot carrier cooling is relatively small. Here, we observe a continuosly increase of PB₂ in FA2.5 samples, which could be due to the carrier injection from the energy funneling effect. Conversely, we did not observe such a phenomenon in the FA 1.6 sample. However, at high excitation density (close but below ASE threshold ≈ 2 × 10¹⁸ cm⁻³, Figure 4d) the PB2 signal of both samples shows a similar growth trend. This shows that under high excitation density, the hot carrier cooling becomes the dominating process. As the funneling effect is hidden by the hot carrier cooling effect, we can no longer identify the funneling effect independently in this regime.

From this experiment we learn that the presence of the low-dimensional FA_n+1PbI_3n+1 phase in high FAI concentration sample functions as an energy funnel, allowing carriers to transfer from PB1 to PB2 under low excitation density (≈1.5 × 10¹⁷ cm⁻³). This effect potentially explains why in the LED device research, the excess FAI samples provide a higher external quantum efficiency.^{[17, 20, 29]} On the other hand, the funneling effect is not a major contribution under high excitation density (≈2 × 10¹⁸ cm⁻³), while hot carrier injection of the CB edge is the main source led to population inversion at high excitation energies.

2.5 Lasing Performance in a DFB Cavity

According to the results shown up to this point, we have selected FA 1.6 as the best FA⁺ composition for lasing application due to its morphological and optical properties. Considering the refractive index (Figure S8, Supporting Information) of the perovskite thin film, we fabricated a device based on a distributed feedback (DFB) cavity with a DFB periodic ranging from 315 to 355 nm. The DFB period, according to the Bragg condition, is λ_B = 2Λn_eff/m, where λ_B is the target resonance wavelength, Λ is the grating period, n_eff is the effective refractive index, and m is the grating order.^[5] To achieve a surface emitting DFB device, a second-order DFB cavity is selected. To show the lasing performances of the DFB cavity devices, we performed the power dependent micro-PL measurement in Figure 5a. A sharp single mode lasing emission peak has been observed at 798 nm. In Figure 5b, we plot the PL intensity and full width half maximum (FWHM) as a function of pump energy. A sharp decrease in FWHM and nonlinear increase of PL is observed with a threshold of 13 µJ cm⁻² (ASE threshold without DFB cavity ≈ 60 µJ cm⁻²).

Figure 5

Lasing characteristic of FAPbI₃ DFB cavity diode. a) Power-dependent PL spectra of the DFB device, clear single-mode lasing peaks have been shown. The inset shows the device stack. b) The PL intensity and FWHM as a function of pump power density, a threshold of 13 µJ cm⁻² is shown. c) PL spectra of the bare film and DFB structured device. ASE in the film shows a FWHM (w) of 8.3 nm, lasing in the DFB device shows a FWHM of 1.4 nm (excitation wavelength: 515 nm). d) Linearly polarized chart of the lasing signal. e) The ASE stability measurement of the film under ≈5 × 10¹⁸ cm⁻³ excitation carrier density, 670 nm excitation wavelength. f) ASE peak intensity as a function of excitation time (h). The dot line corresponds to the spontaneous emission peak intensity.

Then, in Figure 5c we compare the PL spectra of the bare film and that of the DFB based device the FWHM is significantly reduced from 8.3 to 1.4 nm within the DFB cavity, indicating the achievement of lasing. In Figure 5d, the PL signal also presents a polarization feature with a degree of linear polarization of ≈66.2%, which is considered as one of the coherent features in laser. In terms of photostability, the film shows promising ASE stability of more than 21.5 h, in fact as shown in Figure 5e, the ASE peak remains after 21 h of continuously pulsed laser excitation (1 kHz repetition rate), without further encapsulation, under ambient conditions. Figure 5f presents the ASE peak intensity as a function of excitation time, showing excellent photo-stability and robust material against intense photoexcitation, oxygen, and moisture.

3 Conclusion

In this work, we studied the stimulated emission process in a maturely developed FAPbI₃-based perovskite thin film for light emission applications. The optimization of thin films for light emitting diodes sees the use of excess FAI salt. This allows the passivation of deep and shallow carrier trap states and the formation of multidimensional phases which leads to an improvement of the PL efficiency. We show that these features do not lead to an ideal thin film for lasing applications as well. In fact, being the target to accelerate and sustain population inversion, and working at high density excitation regime, we find that the presence of shallow energy states is critical to obtain a low threshold, stable laser.

Finally, thanks to optimization of the thin film stoichiometry we fabricated a DFB based diode showing optically pumped lasing characteristics with a single-mode emission and polarization factor. An exciting long ASE stability of 21 h, in air, is shown. Such progresses contribute to a deeper understanding of the underlying mechanisms and provide insights for optimizing the design and fabrication of perovskite-based lasers. The ability to control the concentration of FAI and its impact on optical properties and carrier dynamics offers promising avenues for improving the performance and efficiency of perovskite laser devices.

4 Experimental Section

Materials

Colloidal ZnO nanoparticles were synthesized via a solution precipitation method in a previous work.^{[30, 31]} FAI was obtained from Dyesol. Other chemicals were obtained from Sigma-Aldrich.

Perovskite Precursor

The perovskite precursors were prepared by dissolving CsI, FAI, PbI₂, and 4,4′-diaminodiphenyl sulfones (DDS) in N,N-dimethylformamide (DMF) solution with different amount of excess FAI. The concentration of all perovskite precursors was kept at 0.3 mol mL⁻¹.

Film Preparation

Indium tin oxide (ITO)-coated glasses were cleaned using an ultrasonic bath for 15 min with deionized water and 2-propanol (IPA) sequentially. After dried by high-speed nitrogen flow, ITO substrates were treated by UV–ozone for 15 min. Then, colloidal ZnO nanoparticles were spin-coated on the ITO substrates at 4000 rpm for 30 s in air, followed by thermal annealing at 100 °C for 2 min. Polyethylenimine ethoxylated (0.05 wt% in isopropyl alcohol) was spin-coated on the prepared ZnO substrates at 5000 rpm for 30 s, followed by thermal annealing at 100 °C for 10 min. Next, the substrates were transported into a nitrogen-filled glovebox. The perovskite precursors were spin-coated on the PEIE-ZnO substrates at 4000 rpm for 30 s, followed by thermal annealing at 100 °C for 5 min.

Optical Characterization

All the optical measurements were performed under ambient conditions without further sample encapsulation, the excitation density dependent PLQY measurement was pumped by a 670 nm pulsed laser generated by a Pharos femtosecond pulsed laser and an external noncollinear optical parametric amplifier (pulse width ≈ 30 fs, 500 kHz repetition rate, ORPHEUS-N, Light conversion) and then collected with a Maya 2000 Pro visible spectrometer and a 680 nm long pass filter. The fluence-dependent ASE measurements were performed under the same configuration but reduced excitation repetition of 1 kHz to prevent sample from photodegradation. The absolute values were measured with an integrating sphere, then the relative values were calibrated.

The PDS measurements were performed with a NKT supercontinuum laser and a 632 nm He-Ne laser, the sample was placed in a cuvette filled with tetradecafluorohexane, the optically chopped supercontinuum laser filtered by a monochromator was aligned and focused onto the sample, the He-Ne laser was carefully aligned in parallel to the sample surface then into a photodetector. The detector and the optical chopper were connected to a lock-in amplifier. Absorption signals were obtained when the sample absorbed certain wavelength of light, the generated heat would transiently change the refractive index of the surrounding tetradecafluorohexane solution. Causing a deflection of the He-Ne laser, the amplitude of the deflection corresponded to the absorption of the light.

Femtosecond TA spectroscopy was performed with a home-built pump-probe setup, a femtosecond laser (1030 nm, pulse width ≈ 300 fs, repetition rate = 2 kHz, Pharos, light conversion) was split into two beams, one was focused on to a 5 mm sapphire plate for white light generation, then focused on the sample as the probe beam, which was finally detected by a monochromator and a CCD camera. The rest of the beam passed either through an external harmonic generator (Hiro, Light conversion) to generate 343/515 nm beam (for pumping, respectively, with third or second harmonic), through a noncollinear optical parametric amplifier system (ORPHEUS-N, Light conversion) for pumping in the 670 nm. Pumping at 343 nm was to excite higher lying states but also provided significant excess energy to the generated carriers, while 670 nm was to excite the 3D perovskite at near band edge to exclude any carrier funneling effect. The pump repetition rate was then halved to 1 kHz by an optical chopper. The pump beam was then aligned onto an optical delay stage (8MT195, Standa), which allowed to perform the pump-probe experiment from few femtoseconds delays to 10 ns. The pump was then noncollinearly focused on the sample and spatially overlapped with the probe beam.

Acknowledgements

The research received funding from the European Union's Horizon 2020 research and innovation program MSCA-ITN PERSEPHONe under Grant Agreement No. 956270. H.W. was funded by Marie Skłodowska-Curie Fellowship No. 101105123. G.F. was funded by the ERC project SOPHY under Grant Agreement No. 771528.

Open access publishing facilitated by Istituto Italiano di Tecnologia, as part of the Wiley - CRUI-CARE agreement.

Conflict of Interest

The authors declare no conflict of interest.

Author Contributions

C.S.J.W.: experiment design and perform, sample preparation, data analysis, manuscript writing-review and editing. A.O.: technical support in PDS experiment, data analytical discussion, manuscript review and editing. G.F.: technical support in TA spectroscopy, data analytical discussion, manuscript review and editing. H.W.: supervision, sample preparation, data analytical discussion, manuscript writing-review and editing. A.P.: supervision, data analytical discussion, funding acquisition, manuscript writing-review and editing.

Open Research

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.