Transient Energy-Resolved Photoluminescence Study of Excitons and Free Carriers on FAPbBr 3 and FAPbBr 3 /SnO 2 Interfaces

: Lead bromide perovskites have a larger band gap and are significantly more stable than their iodine counterparts, offering the perspective for higher voltage, tandem photovoltaics exceeding the Shockley − Queisser limit, and shorter time to deployment of photovoltaics. However, their efficiencies still need to be rivaling the iodine ones. Herein, the photophysics of FAPbBr 3 and the ones behind electron transfer from FAPbBr 3 to SnO 2 , one of the most effective electron transporting materials (ETMs), are reported. Time-and energy-resolved photoluminescence studies revealed the existence of two emitting states in the perovskite, which were assigned to bounded excitons and free carriers. SnO 2 extracted electrons from excitons and free carriers, with a selectivity related to the SnO 2 surface treatment. This new insight helps explain SnO 2 ’s unique qualities as an ETM to produce photovoltaics with reduced voltage losses. Furthermore, this study illustrates the importance of performing time-and energy-resolved photoluminescence to capture the intricacies of the photophysical process.


■ INTRODUCTION
Energy use is ubiquitous in almost every human activity.The rising global temperatures threaten to trigger severe climate change events that will affect everyone and everything.Mitigation strategies such as increasing coastal defenses, reducing soil erosion, and executing reforestation projects are being realized in the areas already at risk.The consensus is that to stop the worse from happening, humans need to meet their energy demands using renewable energy instead of fossil fuels.There is a consensus that electrifying all human activities that can be electrified enables the direct use of wind-and solargenerated power and reduces air pollution.Therefore, there is a significant research effort to improve and develop photovoltaics and wind-turbine technologies that can produce the necessary electricity that reached ca.26,000 TWh in 2020 and is forecast to grow annually between 2.5 and 4%. 1 Lead halide perovskites are the most researched topic concerning a novel photovoltaic technology.In 2009, Kojima et al. 2 introduced this class of materials as potential active materials for photovoltaics, reporting a device with efficiencies of around 9%.−5 Issues related to stability and the use of lead need to be addressed for their commercial deployment, but progress is being made by introducing polymeric interlayers, 6−8 engineering material dimensionality, 9,10 and using tin as an alternative to lead. 11,12odine-based perovskite systems dominate the ranking of record photovoltaic cells.−22 Transient photoluminescence (TRPL) studies are widely used to follow charge recombination decay that provides valuable insights into the charge lifetime and quenching efficiency, which can be correlated with the photovoltaic performance.This is commonly done by single-photon counting, which solely addresses the time component. 23,24−29 The results reveal the existence of two emitting states from FAPbBr 3 , which are harnessed differently by the electron transporting layer (ETL) (SnO 2 ) when its surface is treated with KI.
■ METHODS Preparation of Perovskite Samples.A colloidal dispersion of tin oxide (15% in H 2 O, Alfa Aesar) was diluted with deionized water, and the volume ratio of water to SnO 2 is 4:1.Potassium iodide (KI, 1.2 mg/mL) was added into the diluted SnO 2 dispersion for preparing the SnO 2 + KI substrate.The SnO 2 dispersion was then deposited on the glass slide by spin coating at 3000 rpm for 30 s.The film was annealed at 150 °C for 30 min followed by the UV/ozone treatment for 20 min.
1.1 M PbBr 2 was prepared in a mixture of dimethylformamide and dimethyl sulfoxide with a volume ratio of 9:1 and stirred at 60 °C overnight.The first step was spin-coating the PbBr 2 solution on different substrates at 1500 rpm for 20 s and 5000 rpm for 30 s.Then, the films were annealed at 70 °C for 1 min.The second step was depositing a FABr (65 mg/mL in methanol) solution over the PbBr 2 films (2000 rpm, 30 s), and the films were annealed at 140 °C for 25 min in the air.
Characterization of Films.UV−vis absorption was determined by a Cary 5000 spectrophotometer.Steady-state photoluminescence (PL) was conducted on an Edinburgh FS5 spectrofluorometer.
Time-resolved PL was performed on a streak camera detection system.The 515 nm excitation wavelength was generated from the fundamental 1030 nm laser (200 kHz, Jasper 10, Fluence).The pump power was adjusted to 1 mW by a graduated neutral density filter, and the angle of the excitation beam hitting the perovskite side was 45°.After being transmitted through a long-pass filter, PL emission was detected by a streak camera (C5680 + M5675, Hamamatsu).

■ RESULTS AND DISCUSSION
For this study, we prepared films consisting of the formamidinium bromide perovskite (FAPbBr 3 ) on thin glass slides with and without a coating of SnO 2 .Details of material synthesis and film preparation can be found in the Supporting Information (SI).The systems were extensively characterized in a previous publication; 30 thus, the focus of this contribution is the optical part.FAPbBr 3 on glass was used as a reference sample to depict the underlying photophysics without charge extraction.Figure 1a shows the UV−vis absorbance and PL of this film.The absorbance is dominated by a sharp absorption edge starting at ca. 545 nm and an excitonic peak at 529 nm.The emission overlaps the absorption edge and has a peak at 546 nm.
Deconvolution of the PL with a single component did not produce a good fit with R 2 = 0.972 (see Supporting Information Figure S1). Figure 1b shows the deconvolution considering two components that yielded an R 2 = 0.999.Since FAPbBr 3 is expected to be a pure orthorhombic phase, 24,30,31 one can exclude the possibility of the second peak being associated with the mixture of crystalline phases.This also supported the X-ray diffraction pattern presented in Figure S5, consistent with previously published diffractograms. 32,33nother possibility for the doublet is that it relates to defectrelated trap states. 25However, the overlapping nature of absorbance and emission features and the relatively narrow PL global peak suggest otherwise.A more plausible explanation is that emissions relate to free carriers (higher energy PL peak) and bounded excitons (lower energy PL peak), as suggested recently for the methylammonium bromide perovskite (MAPbBr 3 ). 31This establishes the exciton energy at about 60 meV (estimated from the exciton peak energy 2.343 eV minus the free-carrier emission energy ca.2.283 eV, 34 see Supporting Information Table S1), consistent with what was previously reported 35,36 and higher than its iodine perovskite counterpart. 37,38o test the hypothesis, a time-and energy-resolved TRPL study utilizing a streak camera with femtosecond pulse excitation measurements was performed.If the hypothesis is correct, one should expect different rising components for each species and different times and models for the decays.The laser pulse was fixed at 515 nm (above the band gap energy), and its duration was around 300 fs.Note that the novelty of the manuscript is to follow what happened in the early times (inaccessible to techniques, such as time-correlated singlephoton counting), where excitons and free carriers coexist and thus permit assessing whether carriers can be extracted directly from an exciton, which is particularly relevant in this system since it has an exciton energy higher than the room temperature.Later, carriers are injected as free or trapped carriers but never as excitons because they have already recombined or broken into free carriers.It is worth mentioning that perovskites are known to have long-lived and injectable states (up to 200 μs) that have been thoroughly characterized by time-correlated single-photon counting. 39o establish the optimal laser fluence that ensures a good signal-to-noise ratio and coexistence of excitons and free carriers, TPRL signal power dependence measurements were performed (Figure 2).A linear fit of the data shows that above 10 W/cm 2 , the slope is larger than unity (i.e., 1.5), consisting of co-mingling of excitons and free-carrier recombination. 40,41elow 10 W/cm 2 , the slope is closer to 2, consistent with PL related primarily to free-carrier recombination.Note that PL The Journal of Physical Chemistry C related primarily to exciton recombination would yield a unity slope.Therefore, the optimal pulse energy was established to be 18 W/cm 2 , ensuring a good signal-to-noise ratio while staying within the constant ratio of the squared population of free carriers over that of the excitons according to the Saha equation (see Supporting Information Figure S2). 42,43The Saha equation correlates the ratio of free carriers and excitons to the excitation density in a hot electron−hole plasma. 44,45he TPRL results are summarized in Figure 3 and Supporting Information Tables S1 and S2.Note that small shifts around 10−20 nm are expected when comparing transient with steady-state PL data.The fitting of the global TRPL after 1000 ps provided more substantial evidence for a two-peak component (Figure 3a and Supporting Information Table S1) with a similar peak ratio as the steady-state data.
The lower energy component kinetic trace was extracted between 565 and 575 nm to reduce the overlap with the highenergy component.The kinetic trace had a rising edge of 37 ps and a single exponential decay with τ = 520 ± 12 ps (Figure 3b and Supporting Information Table S2).The rising edge relates to the exciton lifetime.Lifetimes in tens to hundreds of picoseconds suggest that multiple excitons are being formed since single excitons have lifetimes in tens of nanoseconds. 46he extracted decay is consistent with what is expected for bounded exciton recombination since they are formed instantaneously to our time resolution and decay by simple electron−hole annihilation.
The higher energy peak is centered at 528 nm and shows considerable overlap with the exciton component.Fittings across the width of the band were performed to establish the energy range where there is a minimal contribution from the low-energy band and the signal-to-noise ratio is still good.While in the accumulated PL spectrum in Figure 3a, this seems unachievable, one can ensure that in the time domain, this energy range is easier to identify, and, in the present case, it was found to be between 520 and 530 nm.Please check the Supporting Information section explanation of the TRPL peak deconvolution for further substantiation of our claim.
The kinetic trace extracted from the 520−530 nm region was fitted with a rising component of 55 ps and a biexponential decay, namely τ 1 = 120 ± 11 ps (50%) and τ 2 = 1222 ± 69 ps (50%) (Figure 3c and Supporting Information Table S2).The rising component is more significant than the instrument response function (IRF) (ca. 25 ps), meaning this is the associated time for the emitting component to reach the maximum number of emitters before its decay, i.e., the time associated with the cooling down of charges and filling the  The Journal of Physical Chemistry C emitting state.High-quality perovskite films display larger diffusion lengths; consequently, they will take longer to fill the emitting states. 47According to our hypothesis, the higher energy component relates to free carriers, which must follow a bimolecular decay mechanism consistent with the observation. 48he data are then consistent with the proposed hypothesis.Figure 4 presents a Jablonski-type diagram summarizing the photophysics phenomena.Upon band gap excitation with photons above the energy threshold (blue arrow), the excited carriers form bounded excitons or break into free carriers within 37 ps.The bounded excitons reach their maximum population instantaneously to our temporal resolution, and their radiative decay (orange arrow) could be fitted with a single exponential decay.The free carriers take about 55 ps to their maximum.The kinetic trace of the radiative decay (green arrow) follows a bi-exponential model as expected for bimolecular recombination.
An n−i−p structure is a common perovskite photovoltaic architecture, and SnO 2 is among the best-performing ETL materials, 15,19,30,49 with an added benefit that it can be processed from solution at a low temperature.Thus, studying the material's ability to extract charge from the two identified states is pertinent.Figure 5a shows the UV−vis absorbance and PL of FAPbBr 3 deposited on SnO 2 on a glass film.As with the case of FAPbBr 3 on glass, the emission is narrow and overlaps nicely with the absorption edge, which has an identical position as FAPbBr 3 on glass.In this case, deconvolution of the PL signal could be done with a single component, yielding an R 2 = 0.998 (Figure 5b and Supporting Information Table S2).Note that the perceived small shoulder at lower energies is believed to be an artifact from a source reflection that we removed from the spectrum.However, we cannot rule out small amounts of an intermediate crystal phase. 33Yet, this signal was not visible in the TPRL data.
TRPL confirmed the observation of an accumulated PL signal after 1000 ps (Figure 6a).Also, the best fit required only a rising edge and a single exponential decay (Figure 6b).This, allied to the emission centered around 555 nm, suggests that the signal is related to bounded excitons.This finding advocates a preference for electron extraction from the freecarrier population, as depicted in the Jablonski diagram in Figure 7, leading to the suppression of this emission channel.Fitting of the emission could be done with a single exponential (τ = 1468 ± 48 ps) and a rising edge with 34 ps, consistent with bounded exciton emission (Figure 6b and Supporting Information Table S2).The ultrafast rising edge confirms once more the creation of multiple excitons. 46he SnO 2 ETL often exhibits defect-related issues associated with the perovskite/SnO 2 interface.−53 It has been found that the potassium treatment shifts the conduction band minimum and Fermi energy of SnO 2 to higher energy, resulting in a better conduction band alignment at perovskite/SnO 2 . 54,55here is also a possibility that K ions diffused into the bulk perovskite, enlarged the grain size, and passivated the grain boundaries. 55Still, our measurement is less sensitive to that compared to changes in interface charge extraction.Due to the prevalence of such a treatment to achieve high-performing photovoltaic devices, it is relevant to study if this surface treatment has any effect from where the electrons are extracted.

The Journal of Physical Chemistry C
Figure 8a shows the UV−vis absorbance and PL of FAPbBr 3 deposited on SnO 2 treated with KI after deposition on glass.Similar to the previous case, the emission is narrow and overlaps nicely with the absorption edge, which is identical to FAPbBr 3 on glass.Deconvolution of the PL signal could be done with a single component with an R 2 = 0.997 (Figure 8b).
The fitting of the global TRPL 1000 ps required two components (Figure 9a), similar to what was observed with the perovskite on glass but contrasting the result obtained for the perovskite on untreated SnO 2 .It is worth noticing that the peak ratio of the accumulated TRPL emission after 1000 ps is similar to the steady-state ratio observed on the perovskite alone.
Analysis of the kinetic traces (Figure 9b,c) revealed that the higher energy component has a rising edge with about 48 ps, and its decay requires a bi-exponential model to be fitted with τ 1 = 123 ± 19 ps (38%) and τ 2 = 1883 ± 140 ps (62%) (Figure 8c).This is once more consistent with free-carrier emission.The component at lower energy was fitted with a single exponential decay (τ = 1168 ± 36 ps) and a rising function of 16 ps that is lower than our IRF (Figure 9b, again consistent with bounded exciton emission (Supporting Information Table S2).
The results suggest that the potassium treatment that induces an upward shift of the conduction band maximum and Fermi energy of SnO 2 significantly changed the electron extraction ratio of the bounded excitons and the free electrons (Figure 10), with potential implications for the photovoltaic efficiency, which are beyond the focus of this study.It is,   The Journal of Physical Chemistry C however, important to stress that efficient electron extraction directly from excitons requires good electronic coupling and band alignment, 56 which motivates potassium-treated SnO 2 as the ideal ETL for this particular perovskite.

■ CONCLUSIONS
In conclusion, lead bromide perovskites are promising for tandem solar cells.This contribution studied the photophysics of FAPbBr 3 FAPbBr 3 /SnO 2 interfaces.TRPL revealed the existence of two emitting states associated with free carriers and bounded excitons.SnO 2 , a popular ETL, preferentially harvests electrons from the free-carrier population.However, when its surface is treated with potassium, the electrons are extracted from both states with a similar apparent probability, directly affecting overall photovoltaic efficiencies.Direct extraction of electrons from excitons requires good electronic coupling and band alignment at the interface, motivating the use of potassium-treated SnO 2 as an ETL, at least for this class of perovskites.The Journal of Physical Chemistry C

Figure 1 .
Figure 1.Steady-state UV−vis absorbance and emission upon excitation at 470 nm of FAPbBr 3 .(a) Overlapped signal of absorbance and emission and (b) deconvolution of the emission signal.

Figure 2 .
Figure 2. Power dependence of the accumulated TRPL signal of FAPbBr 3 on glass upon excitation at 515 nm with a 300 fs laser pulse and accumulated for 1000 ps.

Figure 3 .
Figure 3. TRPL of FAPbBr 3 on glass upon excitation at 515 nm with a 300 fs laser pulse and a photon energy of 18 W/cm 2 .(a) Deconvolution of the accumulated PL signal after 1000 ps; (b) kinetic decay of the signal extracted between 565 and 575 nm (bounded excitons); and (c) kinetic decay of the signal extracted between 520 and 530 nm (free carriers).Note that TRPL fittings are shown till 600 ps but performed over a 1 ns time range, ensuring at least 60−70% signal decay.

Figure 5 .
Figure 5. Steady-state UV−vis absorbance and emission upon excitation at 470 nm of FAPbBr 3 on SnO 2 .(a) Overlapped signal of absorbance and emission and (b) deconvolution of the emission signal.

Figure 6 .
Figure 6.TRPL of FAPbBr 3 on SnO 2 upon excitation at 515 nm with a 300 fs laser pulse and a photon energy of 18 W/cm 2 .(a) Deconvolution of the accumulated PL signal after 1000 ps and (b) kinetic decay of the signal extracted between 565 and 575 nm (bounded excitons).Note that TRPL fittings are shown till 600 ps but performed over a 1 ns time range, ensuring at least 60−70% signal decay.

Figure 8 .
Figure 8. Steady-state UV−vis absorbance and emission upon excitation at 470 nm of FAPbBr 3 on SnO 2 treated with KI.(a) Overlapped signal of absorbance and emission and (b) deconvolution of the emission signal.

Figure 9 .
Figure 9. TRPL of FAPbBr 3 on SnO 2 treated with KI upon excitation at 515 nm with a 300 fs laser pulse and a photon energy of 18 W/cm 2 .(a) Deconvolution of the accumulated PL signal after 1000 ps; (b) kinetic decay of the signal extracted between 565 and 575 nm (bounded excitons); and (c) kinetic decay of the signal extracted between 520 and 530 nm (free carriers).Note that TRPL fittings are shown till 600 ps but performed over a 1 ns time range, ensuring at least 60−70% signal decay.