Cadmium‐Doping Slows Trap Emptying in Ambient‐Air Blade‐Coated Formamidinium Lead Iodide Perovskite Solar Cells

Formamidinium lead iodide (FAPbI3) in its α‐phase is among the most desirable perovskite compositions for solar cells. However, because of its transition into the yellow δ‐phase at room temperature, it is a challenge to process it in ambient air by scalable fabrication methods. Here the introduction of a trace amount of cadmium (in the form of CdI2) to FAPbI3 is reported and found that it enhances the stability of the perovskite's black α‐phase polymorph, inhibits non‐radiative recombination events, leads to pin‐hole free compact surface morphology, and improves band energy alignment. The 0.6% Cd‐doped FAPbI3 solar cells show a champion efficiency of 22.7% for 0.049 cm2 and 16.4% for cm2‐scale pixels, which, to the best of the knowledge, are among the highest for air‐ambient fully blade‐coated pure FAPbI3 solar cells with an n‐i‐p architecture. Transient absorption microscopy measurements reveal that Cd doping reduces the number of trapped charges and increases their lifetimes, promoting charge accumulation and a higher photovoltage. The study sheds light on the potential of cadmium as a homovalent dopant for the stabilization and performance enhancement of FAPbI3 performance solar cells.


Introduction
Formamidinium lead iodide (FAPbI 3 ) in its -phase is arguably one of the most desired perovskite compositions for solar cells, [1][2][3] reaching a power conversion efficiency (PCE) of 26.1%.It offers a bandgap of ≈1.5 eV, [4,5] which is the lowest among known single-Pb-based perovskites and is closest to the ideal bandgap of 1.34 eV for single-junction solar cells (Shockley-Queisser limit). [6]dditionally, it exhibits improved thermal and operational stability compared to the archetypical MAPbI 3 perovskite (MA stands for methylammonium): MAPbI 3 decomposes at ≈100 °C, [7,8] whereas FAPbI 3 remains intact up to 170 °C. [9,10]owever, -FAPbI 3 is metastable at room temperature and readily transforms into the undesired wide bandgap -phase; this phenomenon is known as polymorphism.16][17][18][19][20][21][22][23][24][25][26][27][28][29] The spin-coating method is, unfortunately, neither economical (≈1% perovskite atom economy, see Supplementary Note 1, Supporting Information) nor scalable (each point on the substrate is affected by different magnitudes of centrifugal force, solution viscosity, and surface tension, leading to non-uniform films at multi-centimeter scales). [30,31]In contrast, commercially viable processes demand near-unity perovskite atom efficiency and uniform films at large scale.Blade-coating has recommended itself as one such alternative for scaling PSCs. [32]iven the advantages of FAPbI 3 PSCs discussed above, it is now important to upscale them − a challenge yet to be addressed considering the FAPbI 3 polymorphism issue and its hypersensitivity to fabrication conditions. [33]Intermediate-phase engineering with the aid of N-Methyl-2-pyrrolidone has led to 17.8% efficient PSCs, a record value, to the best of our knowledge, among all blade-coated FAPbI 3 with n-i-p architecture. [33]e recently reported that Pb-site doping effectively stabilizes -FAPbI 3 single crystals. [34]Using Bi 3+ as a dopant with smaller ionic radii than Pb 2+ , we reported that this approach relaxed structural strain and stabilized the -phase.But Bi 3+ , unfortunately, introduced carrier recombination centers which we attributed to its heterovalent nature relative to Pb 2+ .[37] One such alternative is cadmium (Cd 2+ ) − it is of the same oxidation state is a similarly soft Lewis acid as Pb 2+ , and has a smaller atomic radius (0.95 Å vs 1.19 Å).[40][41][42][43] Cadmium's effect on the crystallinity of FAPbI 3 , and its polymorphism, remains to be known.
Here we demonstrate that a trace amount of Cd indeed stabilizes -FAPbI 3 enabling blade-coating of FAPbI 3 in ambient air.Using compositionally-graded film (CGF) optimization method, [44] we found that 0.6% Cd (relative to Pb) leads to the strongest photoluminescence of perovskite films.We then find that this exact composition also leads to the largest grain size in films with no pin holes.We also observe that Cd-doping enhances carrier lifetimes by an order of magnitude as compared to control FAPbI 3 .Further photophysical investigations revealed that trap filling is promoted by Cd doping, leading to charge accumulation and higher photovoltage.Finally, we demonstrate all-blade coated FAPbI 3 solar cells reaching 22.7% power conversion efficiency.

Results and Discussion
We first fabricated FAPbI 3 film with the gradient content of Cd 2+ via a CGF platform, as it allows making all possible binary compositions in a single experiment. [44]We showed earlier that the CGF composition changes linearly, i.e., one can accurately estimate a local composition from its location by l L (c ink1 − c ink2 ), where l is the distance from one end of the CGF film to the desired point, L is the length of the CGF (28 cm in this work), and c − the concentration additive in the inks. [44]For the CdI 2 -FAPbI 3 CGF, the two FAPbI 3 inks with (5 mol.% CdI 2 relative to PbI 2 ) and without CdI 2 were deposited on a glass substrate by slot-die coater at varying ratios enabled by a gradual change of the pump rate of inks.The appearance of the resultant CGF film with gradient Cd composition is shown in Figure 1a.
To identify the optimal content of Cd, we measured the photoluminescence (PL) spectra (Figure 1b; Figure S1, Supporting Information) using a compact spectrometer with a reflection probe, which slid through the center of the film by a robotic arm and collected data at every 3 mm interval (corresponding to a spatial resolution of ≈0.05% Cd).The PL intensity increased along the film reaching the maximum intensity at l = 3.6 cm (corresponding to ≈0.6% concentration of Cd 2+ ) and then decreasing gradually.The enhancement of PL indicates the reduction of non-radiative recombination rate in the perovskite film, which is beneficial for the performance of solar cells.
To determine the impact of Cd on the morphology of perovskite films, we blade-coated FAPbI 3 films without and with CdI 2 on SnO 2 -coated ITO (indium tin oxide)/glass substrates (we chose this substrate because it was used in solar cells discussed below).Figure 2a compares Scanning Electron Microscopy (SEM) images of the blade-coated films.The control FAPbI 3 film shows many pinholes, obscured grain boundaries as well as bright-contrast features on grains − all these defects may act as electron or hole recombination sites in the completed device.The size and density of grains increased with Cd doping and at 0.6% Cd reached full coverage, an important attribute for films in solar cells.[47][48][49] Further addition of Cd over 1% led to a further increase in grain size but also resulted in poor coverage and large pinholes.
To understand how Cd-doping increases the grain size of FAPbI 3 film, we performed X-ray diffraction (XRD) characterization of as-deposited films before and after annealing, as well as dynamic light scattering (DLS) characterization of perovskite precursor inks.We observed stronger and more oriented diffraction peaks associated with FAI-PbI 2 -solvent intermediate complexes in perovskite films containing Cd compared to those without Cd (Figure S2a, Supporting Information). [33]These complexes, when annealed, are known to produce highly crystalline perovskite films with enlarged grain size and fewer grain boundaries, as we also confirmed by SEM images (Figure 2a).Additionally, these complexes prevented the formation of -FAPbI 3 , as observed in the XRD profiles of annealed films (Figure 2c, Figure S2b, Supporting Information). [50,51]DLS data of perovskite inks indicate that the average diameter of colloidal particles is ≈2 nm (Figure S3, Supporting Information).Furthermore, the colloids increase in size with a higher concentration of CdI 2 .This reduces the number of nucleation sites, slowing down the nucleation rate during film formation. [43]XPS results in Figure S4 (Supporting Information) show the presence of Cd on perovskite films: the Cd 3d 5/2 peak at ≈406 eV increases with higher Cd content in perovskite ink.Therefore, we A critical question, when doping a material, is determining the whereabouts of the dopant.One scenario involves dopant segregation, where Cd would manifest as CdI 2 or FA 2 CdI 4 (a zero-dimensional structure with tetrahedral coordination of Cd) within the FAPbI 3 matrix. [52]To probe this scenario, we compared CdI 2 and FA 2 CdI 4 XRD profiles with FAPbI 3 containing 0−20% CdI 2 : at CdI 2 concentration of 5% and higher, we indeed observed diffraction peaks belonging to CdI 2 (Figure S6, Supporting Information); but at CdI 2 concentration of 1% and lower, we observed no discernible diffraction peaks corresponding to CdI 2 and FA 2 CdI 4 (Figure 2c), ruling out the scenario of dopant phase segregation at the low dopant concentrations relevant to this study.
Another scenario entails dopant incorporation into the crystal structure, for which we have observed some evidence.First, we noted a change in full width at half maximum (FWHM) of XRD peaks when doping FAPbI 3 with cadmium.Lattice strain analysis using Williamson-Hall plots based on the XRD FWHM indicates that structural strain decreases to the minimum with the addition of 0.6% Cd (Figure S6, Supplementary Note 2, Supporting Information).Second, XPS results showed a shift of Pb 4f 7/2 , Pb 4f 5/2 and I 3d 5/2 in Cd-doped films indicating change in the environment of these elements (Figure S4, Supporting Information).Third, we grew FAPbI 3 single crystals with and without CdI 2 .Using optimal Cd concentration in mother solutions, we observed the formation of only black FAPbI 3 phase, whereas without Cd, the yellow phase was dominant (Figure S7, Supporting Information).We then isolated the crystals, fully converted them to black phase through annealing, and left them in ambient air: after 46 days, the control FAPbI 3 crystals turned yellow, while FAPbI 3 with optimal Cd doping remained black (Figure S7, Supporting Information).The observed Cd effect on FAPbI 3 polymorphism should be due to its incorporation into crystal structure as single crystals are free of grain boundaries.In light of these observations, and Cd 2+ and Pb 2+ similar chemical properties (same oxidation state and Lewis acidity), we speculate that Cd 2+ substitutes Pb 2+ in FAPbI 3 , at least at the concentrations tested here.It is worth noting that dopant incorporation into the crystal structure should lead to changes in lattice parameters and, consequently, shifts in XRD peaks.However, assuming substitutional doping of Pb (1.19 Å) with Cd (0.95 Å), a 0.6% Cd doping would result in a diffraction peak shift of only 0.003 °, a value that is practically challenging to resolve.
The UV-vis absorption spectra show a ≈1.5 eV bandgap with no appreciable change in the band edge with Cd doping (Figure 2b).The absorption slope and tail also remained unchanged, indicating that the Urbach energy, a measure of energetic disorder, remained the same.The enhancement in the intensity of absorption peak for the 0.6% Cd-doped FAPbI 3 is attributed to the compact, pin-hole-free nature of the film.
We also assessed the stability of the films.After aging in ambient air for 30 days at an RH of 35%, the FAPbI 3 film with no Cd predominantly exhibited the -phase, while the optimal Cddoped samples showed no sign of degradation (Figure 2d; Figure S8, Supporting Information).55][56][57] Inspired by these findings, we used 0.6% Cd doped 1 m FAPbI 3 solutions to fabricate PSCs in Glass/ITO/SnO 2 /FAPbI 3 /Spiro-OMeTAD/Au configuration on 3.25 cm by 7.5 cm substrates in ambient air using air knife-assisted blade-coating method; SnO 2 and Spiro-OMeTAD layers were also deposited by blade coating method in ambient air at an RH of 35%; the gold counterelectrode was deposited by thermal evaporation.We report statistical data in Figure S9 (Supporting Information).The performance of champion PSCs for 0.049 cm 2 pixels, both with and without Cd, is shown in Figure 3a.The cell without Cd showed a PCE of 20.8% with a short-circuit current (J SC ) of 25.2 mA cm −2 , open-circuit voltage (V OC ) of 1.05 V, and a fill factor (FF) of 78.5%.Cd-doped one showed an increased PCE of 22.7% with a J SC of 25.9 mA cm −2 , V OC of 1.10 V, and FF of 79.6%.Control FAPbI 3 showed poor reproducibility due to the presence of pinholes, while >80% of target Cd-FAPbI 3 showed a PCE of over 20% (Figure 3b).Solar cells of 0.9 cm 2 active area with and without Cd showed a champion PCE of 16.41% and 13.90%, respectively (Figure S10 and Table S1, Supporting Information).Major improvement in performance arises from V OC .We also investigated the stability of unencapsulated PSCs.Consistent with the discussions above, Cd-FAPbI 3 solar cells retained 80% of their original performance following 600 h of operational stability at maximum power point in an inert atmosphere at 56 °C (Figure S11, Supporting Information).
To understand the origin of V OC enhancement, we performed time-resolved PL measurements of perovskite films on glass substrates.Consistent with the steady-state PL measurements discussed above, the Cd-FAPbI 3 films demonstrated a remarkably long PL lifetime of 1,117 ns, an order of magnitude longer than the control FAPbI 3 film (Figure 3c), indicating suppressed nonradiative recombination of charge carriers in the target film.We also performed Ultraviolet photoelectron spectroscopy characterization of the films (Figure S12, Supporting Information).We observed a systematic downshift of perovskite band positions with Cd, minimizing band misalignment with the SnO 2 electron transporter layer (Figure 3d).56]58] To gain further insights into charge transfer and trapping in the absorber layer which directly influences the photovoltaic performance, [59][60][61] we explored Transient Absorption (TA) signals from the films.We monitored a long-lived negative TA signal near the light absorption onset in the μs to ms timescales which is assigned to Ground State Bleaching (Figure S13a, Supporting Information). [62]The experimental kinetic traces were described by the sum of two components: 1) a power law decay attributed to the recombination limited by multiple trapping and release of charge carriers and 2) a second order decay which describes bimolecular recombination of free charge carriers (Figure S13b; Supplementary Note 3, Supporting Information). [63,64]The decays were parametrized by the TA signal amplitude at 1.2 μs and the time the signal decays by half (t 50% ).
Using Transient Absorption Microscopy (TAM), we explored the spatial heterogeneity in the TA signals between the control and the target samples, using a 785 nm laser diode probe with a spatial resolution of ≈50 μm.We observed higher spatial heterogeneity in both parameters for the target (0.6% Cd) sample compared to the control (Figure S14, Supporting Information).This heterogeneity can be attributed to differences in charge carrier transport and trapping, the main photophysical processes occurring on the monitored μs to ms timescales. [65]Given comparable uniformity in the deposited films between the two samples, [66]  these differences between the control and target samples can be linked to changes in compositional and structural features in the perovskite grains or at grain boundaries resulting from Cd doping. [67,68]istograms of the extracted parameters from the fits in each pixel are built to quantitatively compare the charge carrier dynamics of the control and target samples.Since our fitting model consists of the sum of two components, we can isolate the power law and the second-order components and extract their parameters separately.Starting with the second-order law component, a histogram of its amplitude (the ΔO.D.(t 0 ) associated with the second-order decay) is shown in Figure 4a, where we find that between the two samples, the Gaussian distributions almost overlap.The t 50% parameter, interestingly, shows homogenous values within and between the samples − 35 out of 36 pixels in either sample have values of ≈2.6 μs.We conclude that the fraction of excited charges that undergo bimolecular recombination in the μs to ms timescales are not affected by the Cd modification.
Strikingly, the power law component accounts for the change in the overall TA behavior.The histograms of the ΔO.D.(t 0 ) and t 50% associated with the power law decay are shown in Figure 4b  and c, respectively.In contrast to what we observed in the secondorder component, the ΔO.D. at t 0 derived from the power law varies significantly between the samples (Figure 4b,e,f).The histograms show two Gaussian distributions with the means centered at −0.0034 and −0.002 for the control and the target, re-spectively.We observe heterogeneity in the t 50% parameter of the power law component as well.The distribution of t 50% depicts two distinct Gaussian distributions with two means for the control and the target samples of 10 −5.14 s (7 μs) and 10 −4.49 s (33 μs).
Since power-law decays are characteristic of systems where an energetically exponential tail of trap states is present below near the band edge, [63,64] these observations suggest differences in charge trapping in both samples.Overall, the control FAPbI 3 sample showed a higher density of trapped charges (higher amplitude of the ΔO.D. at t 0 ) and a narrower energetic distribution of trap states (higher  value).Taken together, these results are consistent with an increased availability of trap states near the band edge.The sensitivity of the charge carrier dynamics to trapping makes the difference apparent in the TA signals while it was not observed in the Urbach tail for ground state absorption, [69] suggesting a fairly low trap state density.
In addition, the mean of the t 50% of the power law component is 5-fold shorter for the control sample (7 μs) compared to the Cd-doped target sample (33 μs).As the charge carrier lifetime is shorter in the control sample, we conclude that the trap states near the band edge likely are detrimental recombination centers that lead to reduced charge carrier densities.Our spectroscopic investigation suggests that these states are passivated by the Cddoping, leading to longer lifetimes of trapped charge carriers, and facilitating the accumulation of high energy charges.This results in an increased quasi-Fermi level splitting and ultimately an improved V oc in the optimal devices doped with Cd.

Conclusion
The development of stable and scalable FAPbI 3 perovskite solar cells has been a long-standing challenge due to the polymorphism issue and their sensitivity to fabrication conditions.Here we demonstrated a fully blade-coated FAPbI 3 PSCs with a PCE of 22.7% by doping with Cd.Using the compositionally graded film optimization method, we found an optimum doping concentration of Cd, 0.6%, which maximizes radiative recombination rates.Transient absorption microscopy indicated that the target sample exhibits less negative amplitudes (i.e., reduced trapping of charges on the μs − ms timescale) and longer lifetimes (i.e., slower recombination or trap emptying) under illumination.In other words, the trap states in the target sample are more readily filled, leading to increased charge accumulation and higher V OC .These findings open a new avenue for the development of stable and scalable FAPbI 3 PSCs, paving the way for the potential commercialization of perovskite solar cells.

Figure 1 .
Figure 1.Compositionally-Graded Film (CGF) of CdI 2 and FAPbI 3 : a) Image of the CGF film on a glass substrate with dimensions of 28 cm in length by 4 cm in width.b) Photoluminescence spectra in the color map along the CGF film at 3 mm intervals.

Figure 2 .
Figure 2. Characterization of Cd-FAPbI 3 blade-coated thin films.a) Surface SEM images.The scale bar indicates a length of 5 μm.The yellow rectangles and circles show pinholes and obscured grain boundaries.b) Absorption spectra of films.c) XRD profile of fresh films.d) XRD profile of films aged for 30 days in ambient air at relative humidity (RH) of 35%.

Figure 3 .
Figure 3. Characterization of perovskite films and solar cells.a,b) J−V curves and statistical efficiency data of FAPbI 3 PSCs with and without Cd.The inset in panel a) displays an image of perovskite solar cells.c) Transient photoluminescence of FAPbI 3 films with and without Cd.d) Energy band diagrams of n-i-p FAPbI 3 solar cells with and without Cd doped.

Figure 4 .
Figure 4. Histograms depicting frequency distributions of the parameters derived from the control (FAPbI 3 ) and target (0.6% Cd) samples: a) ΔO.D. (t 0 ) derived from second-order component b) ΔO.D. (t 0 ) derived from power law component c) log10(t 50% ) values derived from the power law component d) Parameter .e) and f) portray the spatial maps for log10(t 50% ) derived from the power law component for the control and the target, respectively.Scale bars are 50 μm.