Vapour-Deposited Cesium Lead Iodide Perovskites: Microsecond Charge Carrier Lifetimes and Enhanced Photovoltaic Performance.

Metal halide perovskites such as methylammonium lead iodide (MAPbI3) are highly promising materials for photovoltaics. However, the relationship between the organic nature of the cation and the optoelectronic quality remains debated. In this work, we investigate the optoelectronic properties of fully inorganic vapour-deposited and spin-coated black-phase CsPbI3 thin films. Using the time-resolved microwave conductivity technique, we measure charge carrier mobilities up to 25 cm2/(V s) and impressively long charge carrier lifetimes exceeding 10 μs for vapour-deposited CsPbI3, while the carrier lifetime reaches less than 0.2 μs in the spin-coated samples. Finally, we show that these improved lifetimes result in enhanced device performance with power conversion efficiencies close to 9%. Altogether, these results suggest that the charge carrier mobility and recombination lifetime are mainly dictated by the inorganic framework rather than the organic nature of the cation.

Ever since the first reports on perovskite-based solar cells,[1,2] huge efforts have been made to both improve their photovoltaic performance[3,4] and gain insight into fundamental optoelectronic properties of metal halide perovskites.[5,6] Initially, the most intensively investigated metal halide perovskite was methylammonium lead iodide (MAPbI3), containing an organic cation (i.e., MA) that fills up the voids of the inorganic framework built from corner-sharing PbI6 octahedra.[6] The rotational freedom of the dipolar organic cation has been a topic of interest[7,8] and was proposed to play a key role in properties of metal halide perovskites.[8−11] For instance, it was suggested that electron mobility is driven by dynamic disorder of the MA[9] and that metal halide perovskites owe their unique band structure to collective orientations of MA.[8,10,11] Although optoelectronic properties have been correlated with the orientation of the organic cation,[12] there is still a lack of experimental work truly revealing the interplay between the organic cation and the fundamental properties of metal halide perovskites.[13,14] To this end, replacement of MA with inorganic cesium (Cs) ions is crucial to separate the role of the cation’s dipole moment from that of the lead iodide framework.[15] Although “mixed ion” perovskites comprising multiple cations can be easily made,[4,16] fully replacing the organic cation with Cs+ is experimentally challenging because Cs-based precursors are only poorly soluble in archetypical solvents used for solution processing. Additionally, a yellow nonperovskite structure is the stable phase of CsPbI3 at room temperature.[17−19] However, it was recently shown that the black perovskite phase of CsPbI3 can be obtained as a metastable phase at room temperature[17,18] and that solar cells with efficiencies close to 10% can be obtained using CsPbI3 produced by co-evaporation of its precursors under vacuum.[20] This improved quality of bulk CsPbI3 perovskites now opens up the possibility to study their optoelectronic properties and finally enables a rational comparison to their MA-containing analogues.

In this work, we use time-resolved microwave conductivity (TRMC) measurements to investigate the charge carrier dynamics in polycrystalline CsPbI3 thin films prepared using vapour deposition and different solution-based routes. Although perovskites are well-known to be easily processed from solution, vapour deposition is an ideal method to prepare model systems of lead halide perovskites[21−23] because the film thickness can be precisely tuned and the resulting samples show reproducible quality. We find that the charge carrier mobilities in vapour-deposited polycrystalline CsPbI3 films reach values around 25 cm2/(V s), which is very similar to values previously reported for lead iodide perovskites containing an organic cation such as MA or formamidinium (FA).[21,24,25] Furthermore, we observe impressively long charge carrier lifetimes of several μs in the vapour-deposited CsPbI3 films, while both electrons and holes are immobilized within tens of ns for their spin-coated analogues. We attribute this difference to a significantly higher defect density of the latter, acting as traps to mobile charge carriers. Finally, we show that the extended charge carrier lifetimes of vapour-deposited CsPbI3 result in a substantial enhancement of photovoltaic performance with respect to their spin-coated analogues.

Thin CsPbI3 films of varying thicknesses were prepared using vapour deposition of its precursors. Specifically, CsI and PbI2 powders were heated to their sublimation temperatures under vacuum (see the Experimental Methods section for more details). Although our approach is similar to previously reported co-evaporation methods,[20,22,26] we build up the CsPbI3 film in a very controlled fashion by alternating the deposition of thin CsI (∼2 nm) and PbI2 (∼2.5 nm) layers until the desired thickness is obtained.[21] The absorption spectrum of an as-deposited film on quartz is shown in Figure a. A relatively weak onset is found below 750 nm, with a steep rise in absorption at 520 nm. These results suggest that a mixture of yellow (nonperovskite) and black (perovskite) CsPbI3 has formed during the deposition. That is, while its yellow phase is most stable up to 310 °C,[17] the black perovskite phase of CsPbI3 is only metastable at room temperature.[18] In order to induce the yellow-to-black phase transition, the as-deposited films were annealed to 310 °C in a nitrogen-filled glovebox until these turned black. Then, the CsPbI3 films were rapidly cooled to room temperature on a metal surface to “freeze” the crystals in the black phase. We note that while this black phase is highly unstable under ambient conditions, it can be maintained for at least a few days under nitrogen (see Figure S1 in the Supporting Information). Therefore, to prevent conversion back to the yellow nonperovskite phase, the samples were not exposed to air at any time before and during the optoelectronic characterization.

Figure 1

(a) Fraction of absorbed photons (FA) as a function of excitation wavelength for a thin (260 nm) vapour-deposited CsPbI3 film before (yellow) annealing and absorption (red, solid line) and emission spectra (red, dotted line) after annealing. (b) Wavelength-dependent absorption coefficient (red) of CsPbI3, calculated using the transmission and absorption spectra for a vapour-deposited (solid line) and spin-coated (dotted line) film. The absorption coefficient of vapour-deposited MAPbI3 (black line, data from ref (21)) is added for comparison. (c–f) Background-subtracted X-ray (Co κα radiation, λ = 1.79 Å) diffraction (XRD) patterns of vapour-deposited CsPbI3 films before (c) and after (e) annealing, taken at room temperature using an airtight sample holder. XRD patterns of the yellow nonperovskite phase at 298 K taken from ref (28) (d) and black perovskite phase taken from ref (17) (f), both simulated for a 1.79 Å Co X-ray source. ● denotes the yellow phase, ■ denotes the black phase, ◊ denotes PbI2, and * denotes SnO2.

The red line in Figure a shows the absorption spectrum of the CsPbI3 film after annealing, which shows a sharp onset at 730 nm, consistent with conversion to the black phase. The annealed films show an emission peak centered at 727 nm (1.71 eV); see the dotted lines in Figure a. Initially, the as-deposited layers are composed of relatively small grains (<50 nm), which increase in size up to a few microns upon annealing (Figure S2 in the Supporting Information). We used the absorption and transmission spectra, recorded using an integrating sphere, of the 260 nm thick CsPbI3 film to determine its wavelength-dependent absorption coefficient α (see also the Experimental Methods section). The results are shown in Figure b, in which the α of MAPbI3 (ref (14)) is added for comparison. This shows that the α of CsPbI3 steeply rises below 730 nm until a value of 4 × 104 cm–1 is reached, after which it increases gradually to 2 × 105 cm–1 at 400 nm. Similar to MAPbI3, these values are relatively high,[27] so that thin (∼300 nm) layers of CsPbI3 are already sufficient to absorb most of the visible light at wavelengths shorter than the excitonic absorption peak.

The small absorption feature at around 500 nm hints toward the presence of either PbI2 or yellow CsPbI3 being left in the film after annealing. To further investigate the crystal structure and phase purity of the films, we measured X-ray diffraction (XRD) patterns of the films as deposited (Figure c) and after annealing (Figure e), using an airtight sample holder. The reference patterns of the yellow nonperovskite (Figure d)[28] and black perovskite (Figure f)[17] phases are added for comparison. We observe that the as-deposited film (Figure c) shows reflections corresponding to the black phase (denoted by squares) as well as reflections characteristic of the yellow nonperovskite phase (denoted by circles), confirming that the absorption spectrum (Figure a) represents a mixture of yellow and black CsPbI3. After annealing, the reflections corresponding to the yellow nonperovskite phase are no longer present and the XRD pattern matches those reported in the literature (Figure e).[17,29] Additionally, we observe that, in general, the reflections are much broader in the as-deposited film (Figure c) than those after annealing (Figure e). This is consistent with enlargement of the crystalline domains as observed with atomic force microscopy (see Figure S1). Both before and after annealing, we observe reflections indicative of PbI2 (diamonds) (see also Figure S3), which is most likely the origin of the absorption feature at around 500 nm in the black CsPbI3 (Figure b).

The TRMC technique[30] was used to investigate the mobilities and recombination dynamics of photoexcited charge carriers in CsPbI3 thin films frozen in the black phase. Figure a shows the intensity-normalized photoconductance (ΔG) of a 260 nm thick CsPbI3 film, as a function of time after pulsed excitation at 600 nm. The initial rise of ΔG is attributed to the generation of mobile charges, while the decay represents their immobilization due to trapping or recombination. The product of charge carrier generation yield (φ) and the sum of their mobilities (∑μ) is obtained from the maximum signal height. In view of the low exciton binding energy of ∼15 meV compared to the thermal energy at room temperature,[31] we assume that φ is close to unity. Thus, as clearly shown in Figure a, the sum of electron and hole mobilities in the vapour-deposited CsPbI3 film is ∼25 cm2/(V s). Interestingly, this φ∑μ is very comparable to what we and others have previously measured in planar MAPbI3 and FAPbI3 films using similar techniques.[21,25,32] This observation suggests that the mobility in metal halide perovskites is mainly dictated by the inorganic framework rather than the nature of the monovalent cation.

Figure 2

(a,b) TRMC traces for (a) a vapour-deposited CsPbI3 thin film with thickness of 260 nm and (b) a 350 nm CsPbI3 film spin-coated from a DMF/DMSO solution. Note that the horizontal time scales of (a) and (b) are different. An excitation wavelength of 600 nm was used, and the laser intensity was varied to generate initial charge carrier densities ranging from 1015 to 1017 cm–3. The dotted lines are fits to the experimental data (solid lines). (c) Half lifetime as a function of the initial charge carrier density, corresponding to the CsPbI3 thin films shown in (a) and (b).

As clearly visible from Figure a, the charge carrier lifetimes in the vapour-deposited CsPbI3 (annealed) film reach values on the order of tens of microseconds at a charge carrier density of 1015 cm–3. This is exceptionally long and only comparable to high-quality MAPbI3.[33] Upon increasing the initial charge carrier density to 1017 cm–3, the lifetime gradually decreases. This trend is characteristic for higher-order recombination between electrons and holes, for instance, second-order band-to-band recombination.[32] Upon increasing the thickness of the vapour-deposited films, both the mobility and lifetime remain constant; see Figure S4. This suggests that the optoelectronic quality of vapour-deposited polycrystalline CsPbI3 films is not dependent on their thickness, which enables us to prepare electronically homogeneous layers with various thicknesses.

To investigate whether the mobilities and lifetimes observed in the vapour-deposited samples could be generalized to polycrystalline CsPbI3 films, we repeated the TRMC measurements on spin-coated layers (see also XRD patterns in Figure S3). Figure b shows the result for a (∼350 nm) film, spin-coated using a stoichiometric precursor solution in a mixture of N,N-dimethylformamide (DMF) and dimethyl sulfoxide (DMSO) and annealed to 310 °C to convert to the black phase; see the Experimental Methods for further details. We find that the mobilities in these spin-coated CsPbI3 layers are close to 20 cm2/(V s), which is very similar to those for the vapour-deposited CsPbI3 (Figure a). The charge carrier lifetimes, on the other hand, are significantly shorter in the spin-coated samples for equivalent excitation densities, reaching maximum values of 0.2 μs. These observations suggest that all mobile charges are rapidly immobilized in trap states, indicating a relatively high trap density for these type of solution-processed CsPbI3 films. Most importantly, an increasing lifetime with increasing charge carrier densities is observed up to 4 × 1016 cm–3 (Figure b), while the lifetime decreases again when the number of charge carriers is further increased. We interpret this as filling of trap states until these are saturated, resulting in an enhancement of the lifetime with increasing charge carrier densities.[21]

To further visualize these observations, we have plotted the time at which the signal has reduced to half of its initial value (i.e., the half lifetime τ1/2) against the excitation density in Figure c. This clearly shows the decrease in lifetime with increasing charge densities for the vapour-deposited films (squares), which is typically observed in regimes where higher-order recombination dominates. In contrast, in the spin-coated CsPbI3 (open circles), the lifetime of free charges initially increases (<4 × 1016 cm–3), which we attribute to gradual saturation of trap states upon increasing the charge carrier density. Then, only when the latter starts exceeding the trap density (>4 × 1016 cm–3) will higher-order recombination dominate. Considering that for the spin-coated samples free charges are rapidly immobilized into trap states, the electron and hole diffusion lengths are expected to be significantly shorter than those in the vapour-deposited films, for which long-lived free charges are generated.

In order to quantify processes such as charge carrier trapping and second order (i.e., band-to-band) recombination, we use a previously developed global kinetic model to fit the experimental TRMC data (see also Scheme S1 in the Supporting Information).[32,34] Because our measurements do not enable us to separate the contributions of electrons and holes, we initially assumed that the traps are electron-selective and that free electrons and holes have balanced mobilities based on their similar effective masses.[35] The fits are added as dotted lines in Figure , using the kinetic parameters listed in Table .

Table 1

Kinetic Parameters Used to Fit the Experimental TRMC Data Shown in Figure , Listing Rate Constants for Second-Order Recombination (k2), Trap Filling (kT), Trap Depopulation (kD), the Trap Density (NT), and Mobilities for Electrons (μe) and Holes (μh) for CsPbI3 Thin Films Prepared via Different Routes

	vapour-deposited	spin-coated
k2 (cm3 s–1)	1.3 × 10–10	1.2 × 10–9
NT (cm–3)	9.0 × 1014	1.1 × 1016
kT (cm3 s–1)	1.0 × 10–9	6.0 × 10–8
kD (cm3 s–1)	2.5 × 10–11	9.0 × 10–11
∑μ (cm2/(V s))	26	23

As shown in Figure a,b, this relatively simple kinetic model gives an excellent description of the basic features observed in the experimental decays over a wide range of excitation densities. By using a single set of kinetic parameters characteristic for each sample, we can accurately determine the rate constants for second-order recombination and trap-assisted recombination. From here, we extract that the trap densities (NT) in the spin-coated films are on the order of 1016 cm–3, while these are only 9 × 1014 cm–3 for the vapour-deposited CsPbI3 layers. Furthermore, the trapping rate (kT = 10–9 cm3 s–1) is much higher than the trap depopulation rate (kD = 2.5 × 10–11 cm3 s–1), so that at low charge carrier densities electrons get relatively rapidly trapped and subsequently slowly recombine with the free holes (see also Figure S5). At charge carrier densities above ∼1015 cm–3, their decay is dominated by band-to-band recombination, for which we find a second-order recombination rate constant (k2) as low as 1.3 × 10–10 cm3 s–1. From here, we conclude that the relatively long lifetimes are a combination of slow recombination of free holes with trapped electrons (at charge densities < 1015 cm–3; see also Figure S5) and slow second-order recombination between free electrons and holes (charge denisities < 1017 cm–3). Interestingly, the rate constants for both second-order recombination and trapping are on the same order of magnitude as previously reported for MAPbI3,[24,32] once again suggesting that the cation actually plays a minor role in the charge carrier dynamics.

Interestingly, the experimental data for the spin-coated sample could only be fitted with the current model assuming that one of the charge carriers (for instance, the hole) is already immobilized within the instrumental response time of 3 ns. Then, its countercharge (i.e., electron) gets trapped within the time frame shown in Figure b (∼200 ns). Therefore, it seems likely that both electrons and holes will be rapidly trapped in the spin-coated samples, which makes it challenging to collect both charges efficiently when used in a device configuration.

Finally, in line with the higher trap density in the spin-coated CsPbI3 films, we find a substantial reduction in the PL emission intensity as compared to the vapour-deposited samples (see Figure S6). We note that in both cases, even at charge densities above the trap density, the external photoluminescence quantum efficiencies (PLQEs) are very low (≪ 1%) and PL lifetimes are relatively short (Figure S6), which we are currently investigating.

To further investigate the relationship between the preparation route of CsPbI3 and their optoelectronic quality, we constructed planar heterojunction solar cell devices in the configuration of FTO/SnO2/CsPbI3/HTM/Ag. Here, the HTM is a composite of polymer-wrapped single-walled carbon nanotubes and spiro-OMeTAD (see the Experimental Methods),[36] which enables good hole extraction without oxidation steps that require air exposure of the solar cell[37] and temporarily helps to inhibit moisture and air ingress during cell characterization in air. The results for the best-performing devices with vapour-deposited and with spin-coated CsPbI3 are shown in Figure , and the statistics of 20 devices for each preparation route are summarized in Table . Consistent with their superior charge carrier lifetimes, the devices based on vapour-deposited CsPbI3 (260 nm) show significantly enhanced photovoltaic performance with respect to the spin-coated layers. This results in power conversion efficiencies (PCEs) close to 9% with a highest stabilized power output (SPO) of 7.8% (see Figure b), which is among the highest values reported for fully inorganic perovskite-based devices.[19,38] In contrast, spin-coated CsPbI3 devices gave J–V measured efficiencies of up to 6.4%, which stabilized at 4.3%, which to our knowledge is already the best J–V scan efficiency for devices from spin-coated polycrystalline CsPbI3 thin films.[20] Higher efficiencies of up to 10% have been reported with films composed of CsPbI3 nanocrystals, where suppression of the black-to-yellow phase instability occurs.[38]

Figure 3

(a) Current–voltage (J–V) scans for the highest-efficiency spin-coated (in red) and vapour-deposited (in black) devices. Reverse scans (VOC to JSC) are shown with solid lines, and forward scans (JSC to VOC) are shown with dotted lines. Dark J–V scans are added as dotted lines. (b) SPO of PCE measurements for the same devices, measured at constant voltage.

Table 2

Statistics from J–V Data from Reverse Scans of 20 Devices from Each Preparation Method (8 devices for SPO)a

	vapour-deposited		spin-coated
parameter	best cell	mean ± std. dev.	best cell	mean ± std. dev.
JSC (mA cm–2)	13.0	12.6 ± 1.2	14.4	11.6 ± 2.6
VOC (V)	1.00	0.95 ± 0.06	0.80	0.74 ± 0.13
FF	0.68	0.61 ± 0.05	0.56	0.51 ± 0.07
PCE (%)	8.80	7.27 ± 0.96	6.40	4.4 ± 1.5
SPO (%)	7.8	6.0 ± 1.3	4.3	2.5 ± 1.1
SPO ratio	0.89	0.82 ± 0.09	0.67	0.56 ± 0.14

Stand. dev. is the standard deviation; SPO ratio is the ratio of the SPO with the reverse J–V scan for the same device.

The improved efficiency of the devices employing vapour-deposited CsPbI3 is derived from their significantly higher values for the open-circuit voltage (VOC) and fill factor (FF), which both indicate a higher-quality perovskite layer for vapour-deposited CsPbI3. On average, also the short-circuit current (JSC) values are higher for the devices employing vapour-deposited CsPbI3, even though the perovskite layers are thinner than those for the spin-coated CsPbI3. These improvements are in line with the extended charge carrier lifetimes and lower trap densities observed for the vapour-deposited films (Figure and Table ). Additionally, we find that the standard deviation (20 devices from each method) for the J–V scan characteristics is substantially lower for the vapour-deposited films, which represents a significantly improved reproducibility for this method of perovskite layer deposition. When these devices are held at their maximum power point, we find the SPO is significantly closer to the J–V scan efficiency for the vapour-deposited films, which represents lower hysteresis in the device and is another indication of improved material quality.

Interestingly, the presence of PbI2 in the vapour-deposited CsPbI3 does not seem to harm its overall performance and might even turn out to be beneficial, as previously observed for MAPbI3.[39] Additionally, we note that the devices were exposed to air during deposition of the electrodes and the J–V scans; therefore, these may have suffered from a slight loss of black-phase CsPbI3. However, as these devices are already comparable in efficiency to CsPbI3 devices in the literature,[20] our results emphasize conclusively that vapour deposition yields a higher-quality inorganic perovskite material than solution-based processing methods. In view of its optoelectronic properties, it seems likely that optimizing the device configuration and fabrication together with the quality of vapour-deposited CsPbI3 itself could result in devices with much higher efficiencies. Further, given its band gap of 1.7 eV, CsPbI3 is a suitable candidate to be used in tandem configuration with a Si-based bottom cell,[40] thereby having the potential to boost the efficiencies of commercially available technologies using a robust and thermally stable perovskite absorber.

To summarize, we have investigated the optoelectronic properties of fully inorganic CsPbI3 perovskite thin films, prepared using vapour-deposited and solution-based methods. To freeze the CsPbI3 crystals in their black perovskite phase, the films were heated to 310 °C followed by rapid cooling. Then, the charge carrier mobilities and recombination lifetimes were analyzed using the TRMC technique. We found that the mobilities are around 25 cm2/(V s), both in vapour-deposited and spin-coated CsPbI3 films. Furthermore, in vapour-deposited CsPbI3, we observed lifetimes on the order of tens of microseconds, which is exceptionally long for a metal halide perovskite thin film. For the spin-coated CsPbI3, we found that all free charges are immobilized within 200 ns after pulsed illumination, which we interpret as rapid trapping of electrons and holes. Upon fitting the experimental TRMC data to a previously reported kinetic model,[32,34] we extracted a trap density of 1.1 × 1016 cm–3 for spin-coated films and 9 × 1014 cm–3 for vapour-deposited CsPbI3. Additionally, for the latter, we found that the rate constant for second-order recombination is 1.3 × 10–10 cm3 s–1 and, hence, comparable to both MA- and FAPbI3.[24,25] Importantly, these observations suggest that the organic or dipolar nature of the cation actually plays a minor role in both the mobility and band-to-band recombination of free charges in metal halide perovskites. However, in view of the somewhat higher trap densities and inferior phase stability of fully inorganic CsPbI3 perovskites, the presence of MA+ and/or FA+ might still be crucial for these perovskites to be suitable for high-performance solar cells. On the other hand, it seems likely that postsynthetic treatments could improve the optoelectronic quality and phase stability of CsPbI3.

Finally, we show that the enhanced mobilities and lifetimes in vapour-deposited CsPbI3 are reflected in its performance as a photoactive layer in solar cell devices, boosting the maximum SPO from 4.3 to 7.8% (average of 2.5–6.0%). We suspect that the relatively low PCEs throughout the literature compared to MAPbI3 are related to poor structural stability of black CsPbI3 in combination with the limited effort on optimizing the quality of the inorganic perovskite absorber layer and devices, rather than an intrinsic limitation of fully inorganic metal halide perovskites.

Experimental Methods

Sample Preparation. Vapour Deposition. Thin films of CsPbI3 on quartz substrates were prepared by sequential physical vapour deposition of the precursors PbI2 (≥99%, Sigma-Aldrich) and CsI (≥99.999%, Sigma-Aldrich) in a stoichiometric ratio. Therefore, an adapted evaporation machine (ATC Orion 4 - AJA International, Inc.) with a deposition controller (SQC-310 Inficon) and thermal evaporation controller (TEC-15A) was used. The PbI2 and CsI powders were put into quartz crucibles and positioned in the vacuum chamber. After the pressure was reduced to 10–5–10–6 mbar, the substrates (plasma-cleaned quartz for optical and structural characterization and SnO2 on ITO glass for device fabrication) were introduced into the vacuum chamber. Then, the crucible containing the PbI2 precursor was heated to 240–260 °C until a deposition rate of ∼0.8 Å/s was reached, as indicated by a quartz microbalance. The CsI was heated to 390–400 °C to reach a rate of ∼0.9 Å/s. CsPbI3 was then obtained by alternating deposition of 2.5 nm PbI2 and 2 nm CsI (with 5 s in between), which was repeated until the desired total thickness was reached. Finally, the films were annealed at 300 °C until these turned black. This was followed by rapid cooling on a cold metal surface.

Spin-Coating. Substrates were cleaned by sequential sonication in Hellmanex, deionized water, acetone, and isopropanol and dried with nitrogen. Immediately before spin-coating, the substrates were treated with oxygen plasma for 10 min. Solution preparation, film formation, and annealing were carried out in a nitrogen-filled glovebox. The solutions were stirred until dissolved and filtered with a 2.7 μm GF filter before spin-coating. The precursors were dissolved in a mixture of 0.65:0.35 DMF/DMSO at 0.8 M (1:1 CsI (Alfa Aesar) + PbI2 (TCI)). Films were formed by spin-coating the precursor solution dynamically using a two-step program: 1000 rpm followed by 2000 rpm, with anisole solvent quenching after 30 s. The films were then dried at 45 °C for 10 min. The films were heated at 310–320 °C until black and cooled rapidly on a cold metal surface.

The precursors were dissolved in DMF at 0.5 M (1:1 CsI (Alfa Aesar) + PbI2 (TCI)). Hydriodic acid was added at 35 μL per mL of solution prior to spin-coating. Films were formed by spin-coating the precursor solution dynamically using a two-step program: 1000 rpm followed by 2000 rpm. The films were then annealed at 100 °C for 10 min.

Solar Cell Device Fabrication. Fluorine-doped tin oxide (FTO)-coated glass substrates (Pilkington, 7 Ω □–1) were patterned using etching by 2 M HCl and Zn powder. The etched substrates were cleaned by sequential sonication in Hellmanex, deionized water, acetone, and isopropanol and dried with nitrogen. The clean substrates were then treated with oxygen plasma for 5 min before depositing the electron transport layer (ETL). The ETL was prepared by spin-coating at 3000 rpm (200 rpm/s acceleration) a solution of SnCl4·5H2O (0.05 M in IPA) that had been stirred for 30 min. The SnCl4·5H2O layer was then dried at 100 °C for 10 min and annealed at 180 °C for 60 min to form a compact layer of SnO2. Perovskite layers were then deposited and annealed as for optical measurements. The hole transport layers (HTLs) were prepared by spin-coating 200 μL of a solution of polymer-wrapped single-walled carbon nanotubes dropwise at 3000 rpm.[36] Once dry, this was infiltrated with undoped spiro-OMeTAD by spin-coating a solution of spiro-OMeTAD (LumTec, 85 mg/mL in chlorobenzene with 33 μL/mL tert-butylpyridine additive) at 2000 rpm. The HTL and the active layer at the bottom of the device were removed before deposition of the bottom and top electrodes. Silver electrodes (80 nm) were deposited by thermal evaporation in an evaporator that opens to air.

Solar Cell Device Characterization. The performance of the solar cell devices was measured under simulated AM 1.5 sunlight generated with a class AAB ABET solar simulator calibrated to give simulated AM 1.5 of 100.0 mW cm–2 equivalent irradiance, using an NREL-calibrated KG5 filtered silicon reference cell. The mismatch factor was calculated to be 1.02 between 300 and 900 nm for MAPbI3. The current–voltage curves were recorded with a sourcemeter (Keithley 2400, USA). The devices were masked with a metal aperture defining the active area (0.092 cm2) of the device and measured in a light-tight sample holder to minimize any edge effects and to ensure that the reference cell and test cell are located during measurement in the same spot under the solar simulator. To avoid cross-talk between devices on the same substrate, the active material was removed between devices before measurement. The devices were unencapsulated and therefore exposed to ambient conditions during the measurement.

Optical Characterization. Absorption Measurements. Absorption and transmission spectra were recorded with a PerkinElmer Lambda 1050 spectrophotometer equipped with an integrated sphere. The thin films were placed in front of the sphere to measure the fraction of transmitted light (FT) and at an angle of 10° inside of the sphere to detect the total fraction of reflected and transmitted photons (FR+T). From here, we calculated the fraction of absorbed light (FA)

The fraction of reflected light (FR) was determined from

The absorption coefficient α is often calculated from the transmission spectrum usingwhere IL/I0 equals FT for a sample of thickness L with negligible reflection. However, because thin perovskite films are highly reflective, α was obtained from

Photoluminescence Quantum Efficiency (PLQE) Measurements. Perovskite films were placed in an integrating sphere and were photoexcited using a 532 nm continuous-wave laser. The laser and the emission signals were measured and quantified using a calibrated Andor iDus DU490A InGaAs detector for the determination of PL quantum efficiency. PLQE was calculated as described in ref (41).

Photoconductance Measurements. The thin films on quartz substrates were placed in an airtight microwave cell inside of an N2-filled glovebox. The TRMC technique was used to measure the change in microwave (8.5 GHz) power after pulsed excitation (repetition rate 10 Hz) of the samples at 600 nm.[30] Neutral density filters were used to vary the intensity of the incident light. The illuminated sample area was ∼2.5 cm2. The time-resolved change in conductance ΔG(t) was obtained from the photoexcitation-induced change in microwave power ΔP(t), which are related by a sensitivity factor K

The rise of ΔG is limited by the width of the laser pulse (3.5 ns fwhm) and the response time of our microwave system (1 ns). The slow repetition rate of the laser of 10 Hz ensures full relaxation of all photoinduced charges to the ground state before the next laser pulse hits the sample. Before and during the photoconductance measurements, the samples were not exposed to moisture and air to prevent degradation.