Defect Engineering to Achieve Photostable Wide Bandgap Metal Halide Perovskites

Bandgap tuning is a crucial characteristic of metal-halide perovskites, with benchmark lead-iodide compounds having a bandgap of 1.6 eV. To increase the bandgap up to 2.0 eV, a straightforward strategy is to partially substitute iodide with bromide in so-called mixed-halide lead perovskites. Such compounds are prone, however, to light-induced halide segregation resulting in bandgap instability, which limits their application in tandem solar cells and a variety of optoelectronic devices. Crystallinity improvement and surface passivation strategies can effectively slow down, but not completely stop, such light-induced instability. Here we identify the defects and the intragap electronic states that trigger the material transformation and bandgap shift. Based on such knowledge, we engineer the perovskite band edge energetics by replacing lead with tin and radically deactivate the photoactivity of such defects. This leads to metal halide perovskites with a photostable bandgap over a wide spectral range and associated solar cells with photostable open circuit voltages.

M etal-halide perovskites (MHPs) are wonder semiconductors endowed with chemical flexibility associated with a variety of optoelectronic properties, such as the possibility of tuning their bandgap. 1 This allows creating a wide library of materials to be integrated in diversely purposed devices. An enormous impact is expected by the development of perovskite/silicon and perovskite/ perovskite tandem solar cells, which require stable MHPs with a bandgap of >1.7 eV. 2,3 However, this is actually a daunting task: the typical substitution of iodide with bromide in the MHP lattice can not only promote the bandgap over 1.7 eV but also introduces the instability of light-induced halide segregation, ultimately leading to the formation of iodide-rich and bromide-rich phases. 4 Earlier reports discussed about a defect-driven phenomenon for this unusual behavior, which is consistent with the well-known facile ion migration repeatedly observed in MHPs. 5−9 Confirming the defect-induced nature of the process, crystallinity improvement and surface passivation delivering high quality polycrystalline thin films (with expected lower defect densities) have indeed shown a reduced tendency of mixed-halide perovskites toward halide segregation. 10−12 Significantly, these light-induced ion migration phenomena are an intrinsic characteristic of both pureiodide and mixed-halide lead perovskites, showing up in the former as light-induced decomposition of the thin film, 13−16 while in the latter they give rise to the spectacular bandgap variation following halide demixing. 6,8,9,17−19 We investigated mixed-halide lead perovskites and exploited a well-known strategy for improving crystallinity and enlarging the grain size of the polycrystalline thin film (i.e., use of Pb(SCN) 2 -containing additive) which slows down the halide photosegregation. 20,21 This allowed us to isolate and characterize the charge carrier dynamics associated with bandgap (de) stabilization. The electronic trap states and the associated defects that trigger halide demixing are identified, discovering that light-induced instabilities in pure-iodide and mixed-halide lead perovskites have a common origin. While the defect density can be minimized by ad hoc passivation or crystallinity improvement strategies, the entire suppression of defects is hard, being related to a thermodynamic property of the material. Thus, armed with the knowledge acquired on the electronic activity of such defect, we rather propose a strategy to inactivate it by shifting the band edge of the MHP to a region where the electronic state associated with the defect moves outside the semiconductor bandgap. We attained this by entirely replacing lead with tin, where the energy bands upshift related to the different metal orbital states eventually allow us to achieve photostable wide bandgap MHPs.
The Cs 0.17 FA 0.83 Pb(I 0.5 Br 0.5 ) 3 [FA = HC(NH 2 ) 2 + ] perovskite is first investigated, which has an ideal bandgap of 1.85 eV for the top absorber in a perovskite/perovskite tandem. 22 The perovskite thin films are prepared with 0 to 5 mol % Pb(SCN) 2 -containing additive in the precursor solution (see Methods in the Supporting Information for details). The average grain size is greatly enlarged from ∼200 nm to ∼2 μm by increasing the quantity of Pb(SCN) 2 from 0 to 5 mol %, as shown by the scanning electron microscopy (SEM) images in Figures 1a-b and S1a-e. The grain size enlargement matches with the highly enhanced crystallinity, as indicated by the increase in the intensity of (001) main X-ray diffraction (XRD) peak (Figures 1c and S1f-g) of the perovskite phase. The improvement in grain size and crystallinity is possibly due to the delayed thin film crystallization induced by Pb(SCN) 2 . 23 We further explore the surface chemical environment of the thin films prepared without and with the additive by conducting the X-ray photoemission spectroscopy (XPS) measurements. Figure 1d shows the XPS peaks of Cs 3d, I 3d, N 1s, S 2p, Pb 4f, Cs 4d, and Br 3d. The XPS results show the presence of SCN − on the surface of the perovskite film prepared with SCN − addition because of the emergence of an additional N 1s peak at 398.3 ± 0.2 eV and S 2p doublet with S 2p 3/2 and S 2p 1/2 peaks at 162.7 ± 0.2 and 163.9 ± 0.2 eV, respectively, belonging to SCN − . 24 The relative atomic concentration of each element is calculated from the area of the XPS peaks, after normalization to the relative sensitivity factors for each photoemission process, and the results are   26 Combining XPS with XRD results, we suggest that almost all the residual SCN − ions in the perovskite substitute Br − on the grain surface. Figures 2a-b show the evolutions of the photoluminescence (PL) spectra of both the pristine reference (prepared without additive) and the additive containing (prepared with 5 mol % Pb(SCN) 2 ) samples by a 450 nm diode laser (50 mW/cm 2 ), operating in a continuous mode for 2 min. With this photon dose, the photoinduced halide segregation is prominent in the pristine reference sample, which manifests itself with the appearance of an additional red-shifted PL peak (Figure 2a). 4 This is due to the funnelling of photoexcited carriers into the formed iodide-rich regions, which exhibits a lower bandgap than the starting mixed material 5 (1.61 vs. 1.85 eV, respectively). In the additive containing sample, as expected, 20 halide segregation is lifted out on the investigated time scale (Figure 2b). Thus, we take this model system to identify the optoelectronic processes associated with the observed phenomenology. The transient photocurrent (TPC) is first measured for both samples before halide segregation occurs (see PL spectra tracing during the experiments in Figure S2a). By monitoring TPC, we are not limited to radiative recombination but any free carrier dynamics in the semiconductor can be monitored. 27,28 The TPC trace of the reference sample shows that part of the carrier population decay in the ns time window, while part of the population has an extremely slow dynamic (>10 μs) (Figure 2c), which is not observed by monitoring its band-to-band radiative recombination via time-resolved PL (TRPL) ( Figure S2b). This implies that long-lived mobile photocarriers do not relax radiatively. Thus, the observed dynamics in TPC are associated with a long-lived trapped carrier, 27,28 which disappears in the additive containing sample showing suppressed halide segregation.
The same experiment is also carried out on pure-iodide Cs 0.17 FA 0.83 PbI 3 thin films, synthesized without and with Pb(SCN) 2 . Identical to the mixed halide scenario, the longlived dynamics observed in the TPC trace disappears after using the additive in film preparation ( Figure S3a). This observation is further consolidated by monitoring the photobleaching dynamics of the Cs 0.17 FA 0.83 PbI 3 thin films prepared without and with Pb(SCN) 2 via transient absorption ( Figure  S3b). Under-coordinated iodine defects, e.g., iodine interstitials and lead vacancies, were previously identified as photochemically active deep traps in MAPbI 3 . 27,29 The peculiar iodine redox chemistry leads to extremely long (>μs) deactivation kinetics of slowly trapped electrons on positively charged interstitial iodine (I i + , basically a coordinated I 3 − moiety corresponding to oxidized iodine), 27   trapping-recombination, that quickly deplete the carrier population in the semiconductor. Carrier trapping at such defects leads to the formation of neutral I i 0 species (basically, a coordinated I 2 − radical) whose lifetime mirrors that of trapped carriers. The long lifetime of the I 2 − radical species (when electrons are trapped) leads to a significant probability of forming I 2 moieties through bimolecular processes (e.g., 2I 2 − → I 2 + 2I − ) or capture of a second carrier (e.g., I 2 − + 1h→ I 2 ); I 2 can be eventually expelled to the surface and grain boundaries, activating material degradation in the pure-iodide perovskite. 13 Overall, based on the observations collected from both mixed halide and pure iodide model systems, we can conclude that the trapping of carriers at interstitial defects which eventually lead to the formation of I 2 molecules is associated with the halide segregation process. Thus, we can also claim a common origin for phase segregation in mixed-halide perovskites and photodegradation in pure-iodide perovskites. It is rooted in the formation and diffusion of I 2 thanks to the presence of long-lived dynamics of trapped carriers with a sufficiently long lifetime compatible with ionic dynamics and material transformation.
The direct connection between the defect activity linked to the long-lived carrier trap and photoinduced halide segregation can be pinned up by monitoring the evolutions of the PL spectrum of the mixed-halide reference sample under a pulsed laser with varied interpulse delay (Figures 2d and S4). The halide segregation rate at each interpulse delay is derived based on the evolution of the PL spectrum, and the method is described in the Supporting Information ( Figure S4). The subbandgap feature of the segregated phase only becomes prominent for interpulse delays <20 μs, i.e., when the subsequent excitation pulse reaches the sample at a delay matching the lifetime of the long-lived traps (Figures 2c-d and   S5). Accordingly, halide segregation rate shoots up super linearly when, upon photoexcitation, the species formed by carrier trappings start piling up in agreement with bimolecular/ bielectronic reactions (Figure 2d). In the sample prepared with Pb(SCN) 2 , where the long-lived carrier trap dynamics is virtually not observed, bandgap instabilities do not appear even at a short interpulse delay of 2 μs ( Figure S6).
We have so far identified the electronic dynamics and the related halide defect chemistry responsible for bandgap destabilization in lead halide perovskites. While this may support the development of a targeted strategy for crystallinity improvement/surface passivation, it also highlights the intrinsic instability related to halide photochemistry. Based on this picture, we propose a radical but viable solution that can entirely deactivate photochemically active halide defects, instead of reducing them, by pushing the perovskite band edges toward higher energy levels. One effective approach to push the valence band edge upward, consists in substituting tin for lead in the mixed halide perovskite lattice.
Since the peculiar halide photochemistry in conjunction with low ion migration barriers plays a major role in the phase segregation process, 6 the defect activities of interstitial halides are investigated in lead mixed-halide perovskite (MAP-bI 1.5 Br 1.5 ) and in the corresponding tin counterpart (MAS-nI 1.5 Br 1.5 ) by hybrid density functional theory (DFT) calculations. We focus on MA-based perovskites, rather than the experimentally employed CsFA compounds, for simplicity, considering the exiguous impact of A-site cations on the perovskite defect chemistry. We calculate defect formation energies (DFEs) and ionization levels of halide interstitials in both types of perovskites. As anticipated, our calculations show an increase of the valence band (VB) energy, i.e., a decrease of the ionization potential, of ∼0.9 eV in the tin perovskite compared to the lead counterpart (Figures 3a-b). The calculated bandgaps of the mixed-halide lead and tin perovskites are 1.98 and 1.85 eV, respectively, consistent with experimental PL maxima (∼1.85 and 1.63 eV, respectively, see Figures 2a and 4b).
The equilibrium structures of halide interstitials in various charge states of MAPb(I 0.5 Br 0.5 ) 3 closely resemble those in the pure MAPbI 3 27 (Figures 3c-e); both I i + and Br i + are stable in a relatively wide Fermi level range, up to 0.75 and 0.45 eV above the VB maximum (VBM), respectively. Notably, positive Br i + tends to bind two lattice I − ions to form an I−I−Br trimer with Br−I and I−I bond distances of ∼2.8 Å, clearly indicating the oxidation of a lattice iodide instead of the bromide interstitial. The lower stability of the oxidized Br i + species with respect to I i + is in line with the higher oxidation potential of bromide (1.07 V) compared to iodide (0.54 V), making bromide oxidation more energetically demanding. Iodine can indeed be selectively expelled from mixed-halide lead perovskites upon electrochemical bias. 30 I i + can trap electrons through the (+/0) transition while I i − can trap holes through the (−/0) transition placed, respectively, at 1.13 and 0.37 eV above the VBM (Figure 3b). In both cases, trapping leads to the formation of the long-lived I 2 − dimer species, which may further react leading to the formation of I 2 . 27 Br i − , on the other hand, can trap holes through the shallow (0/−) transition located at 0.09 eV above the VBM (Figure 3b) in line with the higher bromide stability discussed above.
The defect chemistry of MASn(I 0.5 Br 0.5 ) 3 is remarkably different. The higher VB energy of the tin perovskite, i.e. its lower ionization potential, prevents the formation of positive halide interstitials and only negative interstitials are thermodynamically stable across the Fermi level range spanned by the bandgap, leading to thermodynamic transitions resonant with the VBM (Figures 3a-b).
The deactivation of trapping activities of halide interstitials by tin substitution suggests that iodide oxidation may only be active in lead but not in tin halide perovskites. This is confirmed by investigating the energetics of forming a I 2 molecule through carrier trappings at iodine defects, followed by its exclusion to the surface and forming a trimer with a terminal iodide in both lead and tin halide perovskites 31 (Figures 3f-g). In MAPb(I 0.5 Br 0.5 ) 3 the split I i + + I i − couple, i.e. the stable form of I 2 in the bulk lattice, is 0.73 eV less stable than I 2 adsorbed at the surface (Figure 3f), highlighting that I 2 formation and expulsion at the surface is thermodynamically favored. On the contrary, in MASn(I 0.5 Br 0.5 ) 3 , the surface adsorbed I 2 molecule is only metastable, being 0.47 eV less stable than 2I i − + 2h + (Figure 3g), indicating that I 2 formation and expulsion is not thermodynamically viable anymore due to the higher VBM which stabilizes holes residing in the VB.
In agreement with the results of DFT calculations, the I 2 formation and expulsion is not detected when the MASn-(I 0.5 Br 0.5 ) 3 perovskite film is kept under a continuous white LED (100 mW/cm 2 ) for about 30 h (Figures S7a-b). Importantly, we did not probe any change in the dark conductivity of the tin halide perovskite before and after illumination ( Figure S8). On the other hand, I 2 is detected when the same test is performed on Cs 0.17 FA 0.83 Pb(I 0.5 Br 0.5 ) 3 thin films ( Figure S7b). Most importantly, Sn-perovskites do not show I 2 formation even when the volatile MA cation is used, which is known to lead to thin films with lower optoelectronic quality than Cs-containing compositions. 32 It is also found that the lead based perovskite Cs 0.17 FA 0.83 Pb-(I 0.5 Br 0.5 ) 3 prepared with Pb(SCN) 2 shows much reduced I 2 formation, in agreement with its reduced halide demixing discussed above.
Furthermore, we fabricated MASn(I 1−x Br x ) 3 and FASn-(I 1−x Br x ) 3 thin films with x ranging between 0.17 and 0.67 (top-view SEM images of the films are shown in Figures S9− 10) and checked their photostability. As expected, films prepared with higher bromide content show a significant blueshift of the PL emission, allowing for the fabrication of tin perovskite films with bandgaps tuned over the 1.37−1.82 eV range (Figures 4a and S11a). We monitored the evolutions of the PL of all samples under continuous illumination (50 mW/ cm 2 provided by a 450 nm diode laser) for up to 30 min (Figures 4b, S11b-d and S12). The PL spectra of all films are remarkably stable, even when they are excited with very high excitation intensities of 500 mW cm −2 ( Figure S13). Even materials with high bromide fraction do not show bandgap changes over time, independently of the cation composition. Temperature may have a nontrivial effect on halide segregation: while a higher temperature favors entropically driven mixing, 33 it can also provide the necessary activation energy to allow defect migration. 34 We conducted a comprehensive temperature-dependent analysis of MASn-(I 1−x Br x ) 3 . In Figures 4c and S14, we see that between 0 and 85°C all perovskite compositions are stable under a 530 nm diode laser (30 mW/cm 2 ).
Finally, we explored the impact of deactivation of halide segregation on the photostability of the tin-based mix-halide perovskite solar cells. The Cs 0. 17 (Figures 5d-e), confirming the excellent phase stability of Sn-based mixed-halide perovskite, enabling very stable V OC of the working device.
In conclusion, we have demonstrated that light-induced instabilities in pure-iodide and mixed-halide lead perovskites have common origin, being rooted in the photochemical activities of iodine-defects activated by the deep VB of the compounds. While in pure iodide lead perovskites iodidedefects cause decomposition, 16 in mixed-halides they induce additional bandgap instabilities. Both degradation pathways can be slowed down, but not entirely suppressed, by using reducing agents and surface passivation treatments, indicating that they are intrinsic and thermodynamically favored phenomena. Surprisingly, tin halide perovskites, which are extremely sensitive to oxidation, turn out to be remarkably photostable, allowing us to fabricate wide bandgap materials with target characteristics for tandem solar cells. Key to such an achievement is the higher lying VB of tin halide perovskites than the lead counterparts, which, while on the one hand leads to a high propensity to tin oxidation, on the other hand destabilizes iodide oxidation and eliminates halide demixing. Overcoming thermodynamically favored light-induced instabilities by deactivation of defect activities through VB engineering is thus the key toward segregation-free mixedhalide perovskites.