Butanediammonium Salt Additives for Increasing Functional and Operando Stability of Light-Harvesting Materials in Perovskite Solar Cells

Organic diammonium cations are a promising component of both layered (2D) and conventional (3D) hybrid halide perovskites in terms of increasing the stability of perovskite solar cells (PSCs). We investigated the crystallization ability of phase-pure 2D perovskites based on 1,4-butanediammonium iodide (BDAI2) with the layer thicknesses n = 1, 2, 3 and, for the first time, revealed the presence of a persistent barrier to obtain BDA-based layered compounds with n > 1. Secondly, we introduced BDAI2 salt into 3D lead–iodide perovskites with different cation compositions and discovered a threshold-like nonmonotonic dependence of the perovskite microstructure, optoelectronic properties, and device performance on the amount of diammonium additive. The value of the threshold amount of BDAI2 was found to be ≤1%, below which bulk passivation plays the positive effect on charge carrier lifetimes, fraction of radiative recombination, and PSCs power conversion efficiencies (PCE). In contrast, the presence of any amount of diammonium salt leads to the sufficient enhancement of the photothermal stability of perovskite materials and devices, compared to the reference samples. The performance of all the passivated devices remained within the range of 50 to 80% of the initial PCE after 400 h of continuous 1 sun irradiation with a stabilized temperature of 65 °C, while the performance of the control devices deteriorated after 170 h of the experiment.


Introduction
The steep progress of perovskite photovoltaics over the last decade relies mainly on the impressive growth of record power conversion efficiencies up to 25.7%, being almost as high as the record PCE of monocrystalline silicon-based solar cells (26.1%) [1]. At the same time, the long-term stability of PSCs is still far from the target value of 20-25 years [2], despite significant progress in this field driven by a number of approaches like perovskite compositional engineering [3], dopant-free hole-transporting materials [4], barrier layers [5], the application of 2D perovskites [6], and 2D/3D heterostructures [7], as well as the control of surface and bulk defects in perovskite materials and interfaces through passivation with various organic or inorganic compounds [8]. Organic ammonium salts have been successfully used in the last three approaches [9][10][11]. The most prominent member of these additives is the n-butylammonium (BA + ) cation, which forms a homologous row of layered (BA) 2 (MA) n−1 Pb n X 3n+1 perovskites with a wide range of n [12]. The use of n-butylammonium-based 2D/3D heterostructures allows to create devices with record open-circuit voltage [13] and outstanding exploitation stability [14]. Along with such clear advantages, monoammonium cations demonstrate a tendency of dynamical evolution of 2D/3D interfaces under heating and light illumination, leading to PSC instability and loss of performance [15,16].
Diammonium cations should be considered as a promising alternative with various advantages, including their much lower deprotonation ability, the absence of van der Waals gaps in 2D perovskite structures and possible influence on perovskite grain and domain interconnections [6,17]. Indeed, a direct stability comparison of PSCs based on monoammonium and diammonium cations revealed the superiority of diammonium counterparts used as either spacer cations in 2D perovskites [6,18] or passivating agents in 3D structures [19]. It is known that different chemical structures and carbon chain lengths of diammonium cations play a crucial role in PSCs performance and stability [20][21][22], thus demanding a proper selection of organic compounds with desired functional groups. One of the record PSC stabilities over 5000 h under continuous 1 sun illumination has been achieved so far by using 2D perovskites with a 1,3-propanediammonium (PDA) spacer cation and cesium co-doping [23]. On the other hand, a direct comparison of diammonium cations with different chain lengths within 2D perovskites, provided by Zheng et al., revealed the superior performance and stability of PSCs with a 1,4-butanediammonium (BDA) spacer cation, as compared to PDA-based devices [21]. The same trend was reported for the case of surface passivation with diammonium halides, where BDAI 2 demonstrated better performance than PDAI 2 [22]. These results are reliable, since PDAI 2 (in contrast to BDAI 2 ) forms a non-perovskite structure with the empirical formula [PDAPbI 4 ] 15 •[PDAI 2 ] [24]. Therefore, 1,4-butanediammonium appears to be a more promising alternative to shorter PDA 2+ cations and other diammonium counterparts. However, there are very few works regarding the use of BDA 2+ in PSCs and the available results are controversial in terms of the ability to crystallize different members of the 2D (BDA)(MA) n−1 Pb n I 3n+1 compositional row; BDA 2+ localization in 3D perovskites; the role of BDA 2+ additives in performance and photothermal stability of perovskite solar cells.
Here, we provided a systematic study on the use of 1,4-butanediammonium diiodide (BDAI 2 ) either as a spacer cation in 2D (BDA)(MA) n−1 Pb n I 3n+1 perovskites or as a passivating agent in 3D MA 0 . 25 FA 0 . 75 PbI 3 and FA 0 . 85 Cs 0 . 15 PbI 3 counterparts to define the most effective method of introducing BDAI 2 into light-harvesting materials in perovskite solar cells.
Synthesis of 2D perovskite crystals. Crystals with the following nominal compositions were synthesized by the solution method described in [25]: BA 2 PbI 4 , BA 2 MAPb 2 I 7 , (BDA)PbI 4 , and (BDA)MAPb 2 I 7 . A 1.5M stoichiometric solution of the corresponding halide salts in a mixture of propylene carbonate and hydroiodic acid was kept at a constant temperature of 70 • C for 8 h. A (BDA)MAPb 2 I 7 solution was also prepared with a 1.5 M and 0.1 M concentration to study the effect of concentration on 2D perovskite phase composition. Crystallization in the (BDA)MAPb 2 I 7 solution with 25, 50, 75, and 100% excess of MAI, with respect to Pb content, was achieved with 1.5 M solutions using the same method. All the resulting crystals were dried, ground, and studied by powder X-ray diffraction.
Crystallization of (BDA)MAPb 2 I 7 and (BDA)MA 2 Pb 3 I 10 using an equilibrium system with a transfer mediator. To study the equilibrium phases in the (BDA)(MA) n−1 Pb n I 3n+1 system with n = 2 and 3, we carried out the following experiment: the mixtures of solid-state BDAI 2 + MAI + PbI 2 precursors with (BDA)(MA)Pb 2 I 7 (n = 2) and (BDA)(MA) 2 Pb 3 I 10 (n = 3) overall compositions were thoroughly ground in an argon-filled glove box (<10 ppm O 2 ; 0.1 ppm H 2 O) and then dispersed in dry toluene with a 5% amount of molecular iodine as a mass-transfer agent. The 5% amount of iodine was estimated with respect to the total lead amount. The prepared solid-liquid mixtures were continuously stirred and heated at 70 • C for 7 days to achieve thermodynamic equilibrium. After the resulting powders were washed several times with pure toluene, they were dried, and again ground in the glove box with dry air (<20% RH) just before XRD analysis.
Deposition of 3D perovskite films. MA 0 . 25 FA 0 . 75 PbI 3 perovskite films were prepared from 1.5 M solutions of MAI, FAI, PbI 2 in stoichiometric molar ratios with 15% excess of MACl in DMF/DMSO (4:1 v/v) via 1-step antisolvent spin-coating in an inert glove box. Spin-coating was performed at 6000 rpm for 30 s, where 100 µL of chlorobenzene was added dropwise at the 5th second of rotation. The following annealing step was performed at 125 • C for 20 min: FA 0 . 85 Cs 0 . 15 PbI 3 perovskite films were prepared by the same method from 1.3 M stoichiometric solution of FAI, CsI, and PbI 2 in DMF/DMSO (4:1 v/v). The spin-coating method was the same as in the previous example, except for the addition of CB at the 15th second of rotation. Annealing was also performed for 20 min at 125 • C.
Bulk passivation of both perovskite compositions was carried out following the addition of 0.2, 0.5, 1, 2.5 and 5 mol.% BDAI 2, with respect to Pb content, to the initial perovskite solutions by mixing two perovskite solutions with 0 and 5% BDAI 2 in the volume ratios that corresponded to the resulting amount of the additive.
Fabrication of perovskite solar cells. FTO-coated glass substrates were patterned via chemical etching (Zn powder + 4 M HCl) and then sequentially cleaned with detergent (2% Hellmanex solution), distilled water, and the prepared solution of concentrated H 2 SO 4 + 30% H 2 O 2 in a volume ratio of 3:1 for 10 min. After this, the substrates were washed in deionized water, dried by N 2 gas flow, and immediately transferred onto a hotplate for the deposition of the blocking TiO 2 layer. A compact TiO 2 layer was formed by spray pyrolysis at 450 • C. The precursor solution was prepared by mixing 900 µL of titanium (IV) isopropoxide, 600 µL of acetilacetone and 50 mL of isopropyl alcohol. After spraying, the substrates were annealed for 60 min at 450 • C and then left to cool down. The deposition of the SnO 2 layer was carried out according to the method stated in ref. [26]. Shortly after, 0.625 g of urea was dissolved in 500 mL of deionized water, followed by the addition of 10 µL of mercaptoacetic acid and 0.5 mL of 37% HCl. Finally, 0.1 g of SnCl 2 ·2H 2 O was dissolved in the above-mentioned solution. The glass/FTO/c-TiO 2 substrates were previously treated by ozone with UV/ozone cleaner (Osilla) for 15 min, placed vertically into a glass with the as-prepared solution, and then stored in a laboratory oven at 70 • C for 3 h. After SnO 2 deposition, the substrates were rinsed with deionized water and sonicated for 2 min to remove large agglomerates from the SnO 2 film surface. Finally, the dried substrates were transferred onto a hotplate and annealed at 180 • C for 1 h. FA 0 . 85 Cs 0 . 15 PbI 3 + x% BDAI 2 perovskite films (where x = 0.25, 0.5, 1, 2.5, and 5%) were deposited on FTO/c-TiO 2 /SnO 2 substrates, according to the previously described protocol. The spiro-OMeTAD solution was prepared by dissolving 60 mg of Spiro-OMeTAD in 598 µL of CB with 21.6 µL of TBP, 12 µL of 1.8 M solution of LiTFSI in anhydrous acetonitrile, and 4.8 µL of 0.25 M solution of co-complex FK209 in anhydrous acetonitrile. Deposition of 28 mkl of doped spiro-OMeTAD solution was carried out at 4000 rpm in an inert glove box. Finally, an Au electrode with~80 nm thickness was deposited by thermal evaporation (base pressure 10 −5 Torr) through a shadow mask.
Encapsulation of perovskite solar cells. The method for the encapsulation of PSCs was provided by ref. [27]. At the first stage, a 150-200 nm layer of molybdenum trioxide (MoO x ) was thermally evaporated under a high vacuum (P = 1 × 10 −5 mbar) on top of the as-prepared PSCs. The deposition rate was monitored via QSM sensors and maintained at 0.4-0.7 Å/s. After the evaporation of the capping layer, the solar cells were immediately transferred into a nitrogen-filled glovebox. The second encapsulation step was carried out with commercially available UV-curable polymers ("Nano Clear" by FunToDo) and a cover glass slide. For this step, 5-7 µL of epoxy was dispensed onto an active area of each device and covered by a glass slide. Immediately after this, the photopolymerization of the epoxy was carried out via UV light exposure (10 W InGaAlN LED; λ = 365-375 nm) for 1-2 min.
Perovskite solar cell characterization and stability measurements. The current densityvoltage (J-V) characteristics of solar cells were measured with a Keithley series 2450 source meter under a simulated sun AM 1.5 G (ORIEL LSH-7320 ABA LED Solar Simulator) with custom-built python-based software. The light intensity was calibrated to 1 sun (AM1.5, 100 mWcm −2 ) by unfiltered Si reference cells calibrated by Newport.
The long-term photostability of encapsulated PSCs under ambient conditions was achieved using an unfiltered sulfur plasma lamp (LG PSH 0731B), calibrated with Si reference cells to a 100 ± 10 mW/cm 2 power density. To maintain the desired device temperature, PSCs were placed onto a custom-built copper stage that was consistently heated to 65 • C. MPPT was performed using a standard perturb and observe algorithm and the voltage was updated every 5 min with a 10 mV step by a homemade electronic board with 16-channel MPPT capability.
X-Ray powder diffraction. XRD analysis was performed using a Bruker Advance D8 diffractometer with Bragg-Brentano geometry and Cu Kα radiation (λ = 15,418 Å). XRD patterns were recorded in the 3-35 • 2θ range with 0.1 s per dot and a 0.02 • step.
SEM. Scanning electron microscopy (SEM) was performed on a Zeiss Supra 40 microscope with an accelerating voltage of 1-5 kV.
PL spectroscopy. PL spectra were recorded via a home-built setup based on Thorlabs components with a 405 nm excitation laser, 10x focusing objective lens, and an Ocean Optics Flame spectrometer. The laser power was ∼200-210 µW, and the laser spot size was ∼10 µm.
Time-resolved photoluminescence spectroscopy. Time-resolved photoluminescence spectroscopy (TRPL) was performed with a laser pulse frequency of 1000 kHz and a single pulse time of 30 ps. PL decay curves were recorded on a calibrated HMP-100-50 single-photon detector (Becker and Hickl, Germany), equipped with an SPC-150 module. The PL decay curves were processed using the SPC Image software (Becker and Hickl) and OriginPro. To approximate the decay curves, we used the sum of two exponential decay functions, which was as follows: τ avg = (A 1 τ 1 + A 2 τ 2 )/(A 1 + A 2 ), where A i and τ I are the amplitude and lifetime of charge carriers for the ith component of the decay curve, respectively.
Photostability study of perovskite films. The photostability study of perovskite films was carried out according to the protocol described in [28]. The perovskite films on FTO substrates were irradiated with a white LED with 100 mW/cm 2 power density in an argon-filled glove box (<10 ppm O 2 ). All perovskite films were irradiated from the substrate side to minimize the interference of defective perovskite surfaces in light-induced degradation [29]. The characterization of perovskite films was periodically provided by PL spectroscopy under an inert atmosphere. The comparison of the photostability of different perovskite samples was carried out based on normalized PL intensity values as a function of the light-soaking time.

Results and Discussion
The synthesis of higher n members of (BDA)(MA) n−1 Pb n I 3n+1 2D perovskites has not been yet clearly presented, while published XRD and optical spectroscopy data are very contradictory [18,30]. Often, authors discuss the problem of the solution synthesis of phase-pure (BDA)(MA) n−1 Pb n I 3n+1 with n > 1, demonstrating the formation of multiplephase samples with the prevailing (BDA)PbI 4 and/or 3D perovskite phases [31][32][33]. An analysis of the open access 2D perovskites database [34] reveals that only (BDA)PbI 4 (n = 1) crystallographic data have been reported [35].
To clarify this issue, we tried to synthesize (BDA)(MA) n−1 Pb n I 3n+1 layered perovskites with n = 2 and 3 using the following two methods: crystallization from a 1.5 M solution of nPbI 2 + (n-1)MAI + BDAI 2 in an HI + propylene carbonate mixture, as previously reported in [25], and slow crystallization using an equilibrium system with a transfer mediator. At first, we achieved the solution synthesis of single crystals for (BDA)(MA) n−1 Pb n I 3n+1 and (BA) 2 (MA) n−1 Pb n I 3n+1 (BA-butylammonium) compositions with n = 1 and 2 ( Figure 1). In the case of BA-based solutions, high-quality single crystals with target n numbers were successfully obtained ( Figure 1A,B), while only the (BDA)PbI 4 phase (yellow crystals) was able to crystallize from BDA-based solutions, regardless of the initial precursors' stoichiometry ( Figure 1C,D). This behavior of the system was previously explained by the much lower solubility of the n = 1 member, due to the strong electrostatic affinity of diammonium cations towards inorganic perovskite-like layers [31]. Additional attempts to crystallize (BDA)(MA)Pb 2 I 7 perovskites from a diluted 0.1 M solution also resulted in the formation of (BDA)PbI 4 phases ( Figure S1). Hypothetically, the excess of MAI in the solution that corresponds to the (BDA)(MA)Pb 2 I 7 composition could shift the equilibrium of the system towards the crystallization of the n = 2 member. However, according to our experimental results, the BDA-based solution system does not respond to the variation in MAI:BDAI 2 stoichiometry up to 2:1 (100% excess of MAI), thus preserving (BDA)PbI 4 phase composition ( Figure S2).
degradation [29]. The characterization of perovskite films was periodically provided by PL spectroscopy under an inert atmosphere. The comparison of the photostability of different perovskite samples was carried out based on normalized PL intensity values as a function of the light-soaking time.

Results and Discussion
The synthesis of higher n members of (BDA)(MA)n−1PbnI3n+1 2D perovskites has not been yet clearly presented, while published XRD and optical spectroscopy data are very contradictory [18,30]. Often, authors discuss the problem of the solution synthesis of phase-pure (BDA)(MA)n−1PbnI3n+1 with n > 1, demonstrating the formation of multiplephase samples with the prevailing (BDA)PbI4 and/or 3D perovskite phases [31][32][33]. An analysis of the open access 2D perovskites database [34] reveals that only (BDA)PbI4 (n = 1) crystallographic data have been reported [35].
To clarify this issue, we tried to synthesize (BDA)(MA)n−1PbnI3n+1 layered perovskites with n = 2 and 3 using the following two methods: crystallization from a 1.5 M solution of nPbI2 + (n-1)MAI + BDAI2 in an HI + propylene carbonate mixture, as previously reported in [25], and slow crystallization using an equilibrium system with a transfer mediator. At first, we achieved the solution synthesis of single crystals for (BDA)(MA)n−1PbnI3n+1 and (BA)2(MA)n−1PbnI3n+1 (BA-butylammonium) compositions with n = 1 and 2 ( Figure 1). In the case of BA-based solutions, high-quality single crystals with target n numbers were successfully obtained ( Figure 1A,B), while only the (BDA)PbI4 phase (yellow crystals) was able to crystallize from BDA-based solutions, regardless of the initial precursors' stoichiometry ( Figure 1C,D). This behavior of the system was previously explained by the much lower solubility of the n = 1 member, due to the strong electrostatic affinity of diammonium cations towards inorganic perovskite-like layers [31]. Additional attempts to crystallize (BDA)(MA)Pb2I7 perovskites from a diluted 0.1 M solution also resulted in the formation of (BDA)PbI4 phases ( Figure S1). Hypothetically, the excess of MAI in the solution that corresponds to the (BDA)(MA)Pb2I7 composition could shift the equilibrium of the system towards the crystallization of the n = 2 member. However, according to our experimental results, the BDA-based solution system does not respond to the variation in MAI:BDAI2 stoichiometry up to 2:1 (100% excess of MAI), thus preserving (BDA)PbI4 phase composition ( Figure S2).  To study the thermodynamic equilibrium in the (BDA)(MA) n−1 Pb n I 3n+1 system, we carried out the following experiment: the mixtures of solid-state BDAI 2 + MAI + PbI 2 precursors with (BDA)(MA) 2 Pb 3 I 10 (n = 3) and (BDA)(MA)Pb 2 I 7 (n = 2) overall compositions were thoroughly ground in an inert glove box and then dispersed in dry toluene with a 5% amount of molecular iodine as a mass-transfer agent (Figure 2A). The mass transfer proceeds through the liquid media via the formation of liquid alkylammonium polyiodides [36,37], which play the role of highly reactive intermediate phases that quickly yield hybrid iodoplumbates with the highest thermodynamic stability. The grinding stage can initiate the solid-state formation of layered perovskite phases, which was reported earlier as a more perspective preparation method for diammonium-based 2D perovskites than solution synthesis [31]. The prepared solid-liquid mixtures were continuously stirred and heated at 70 • C for 7 days; after this, the resulting powders were washed several times with pure toluene, dried, and again ground just before XRD analysis. Diffraction patterns of the resulting powder samples are given in Figure 2B, demonstrating that the mixtures of (BDA)PbI 4 and MAPbI 3 with some amount of unreacted PbI 2 formed instead of (BDA)(MA) 2 Pb 3 I 10 and (BDA)(MA)Pb 2 I 7 phases in both cases. A relative amount of (BDA)PbI 4 and MAPbI 3 phases correlates with the initial stoichiometry; the (BDA)PbI 4 phase dominates the (BDA)(MA)Pb 2 I 7 sample ( Figure 2B(2)), while 3D perovskite is the main product for the (BDA)(MA) 2 Pb 3 I 10 sample ( Figure 2B(1)). To study the thermodynamic equilibrium in the (BDA)(MA)n−1PbnI3n+1 system, we carried out the following experiment: the mixtures of solid-state BDAI2 + MAI + PbI2 precursors with (BDA)(MA)2Pb3I10 (n = 3) and (BDA)(MA)Pb2I7 (n = 2) overall compositions were thoroughly ground in an inert glove box and then dispersed in dry toluene with a 5% amount of molecular iodine as a mass-transfer agent (Figure 2A). The mass transfer proceeds through the liquid media via the formation of liquid alkylammonium polyiodides [36,37], which play the role of highly reactive intermediate phases that quickly yield hybrid iodoplumbates with the highest thermodynamic stability. The grinding stage can initiate the solid-state formation of layered perovskite phases, which was reported earlier as a more perspective preparation method for diammonium-based 2D perovskites than solution synthesis [31]. The prepared solid-liquid mixtures were continuously stirred and heated at 70 °C for 7 days; after this, the resulting powders were washed several times with pure toluene, dried, and again ground just before XRD analysis. Diffraction patterns of the resulting powder samples are given in Figure 2B, demonstrating that the mixtures of (BDA)PbI4 and MAPbI3 with some amount of unreacted PbI2 formed instead of (BDA)(MA)2Pb3I10 and (BDA)(MA)Pb2I7 phases in both cases. A relative amount of (BDA)PbI4 and MAPbI3 phases correlates with the initial stoichiometry; the (BDA)PbI4 phase dominates the (BDA)(MA)Pb2I7 sample ( Figure 2B(2)), while 3D perovskite is the main product for the (BDA)(MA)2Pb3I10 sample ( Figure 2B(1)). Considering these results along with previously reported data, we conclude that the formation of a light-absorbing layer based on phase-pure 2D (BDA)(MA)n−1PbnI3n+1 perovskite is unfavorable. On the other hand, passivation of 3D perovskites with BDAI2 appears to be more beneficial and can be accompanied by the formation of either (BDA)PbI4 phases or BDA 2+ layers at the interfaces and/or perovskite intra and intergrain boundaries (GBs). In the case of surface passivation, the formation of (BDA)PbI4 phases on the perovskite surface was previously reported [38][39][40], but the effect of bulk passivation with BDAI2 on the structure and optical properties of 3D perovskites has not been fully discussed, despite the potential for more efficient passivation of cation and anion vacancies throughout the thickness of the perovskite film. To clarify this issue, we performed bulk passivation of MA0.25FA0.75PbI3 (MAFAPI) and FA0.85Cs0.15PbI3 (FACsPI) perovskite films with different amounts of BDAI2 and studied the effect of passivation on the morphology, crystal structure, optical properties, and photostability of perovskite films, as well as PSCs. The use of mixed-cation perovskites was motivated by their higher PCE and stability compared to the conventional MAPbI3 composition [41][42][43][44]. The amount of diammonium additives in Considering these results along with previously reported data, we conclude that the formation of a light-absorbing layer based on phase-pure 2D (BDA)(MA) n−1 Pb n I 3n+1 perovskite is unfavorable. On the other hand, passivation of 3D perovskites with BDAI 2 appears to be more beneficial and can be accompanied by the formation of either (BDA)PbI 4 phases or BDA 2+ layers at the interfaces and/or perovskite intra and intergrain boundaries (GBs). In the case of surface passivation, the formation of (BDA)PbI 4 phases on the perovskite surface was previously reported [38][39][40], but the effect of bulk passivation with BDAI 2 on the structure and optical properties of 3D perovskites has not been fully discussed, despite the potential for more efficient passivation of cation and anion vacancies throughout the thickness of the perovskite film. To clarify this issue, we performed bulk passivation of MA 0 . 25 FA 0 . 75 PbI 3 (MAFAPI) and FA 0 . 85 Cs 0 . 15 PbI 3 (FACsPI) perovskite films with different amounts of BDAI 2 and studied the effect of passivation on the morphology, crystal structure, optical properties, and photostability of perovskite films, as well as PSCs. The use of mixed-cation perovskites was motivated by their higher PCE and stability compared to the conventional MAPbI 3 composition [41][42][43][44]. The amount of diammonium additives in the perovskite solution varied in a wide range from trace 0.025 wt.% [17] to 3-8% [20,45]; therefore, we chose the following BDAI 2 concentrations for 3D perovskite bulk passivation: 0.25, 0.5, 1, 2.5, and 5%.
According to the XRD data of the MA 0 . 25 FA 0 . 75 PbI 3 + x% BDAI 2 samples ( Figure 3A), no phase impurities, such as (BDA)PbI 4 or delta polytypes of FAPbI 3 , formed after the introduction of BDAI 2 into the material. The same behavior was observed for the FACsPI + x% BDAI 2 samples ( Figure S3). Interplanar distances in the perovskite lattice were also not affected by the diammonium additive ( Figure 3B). Evidently, such small amounts of additives would not be detected by XRD; however, we observed the widening of diffraction reflexes along with an increased amount of the BDAI 2 additive, as shown in Figure 3B for the (100) perovskite reflex. This seems to correlate with the monotonic decrease in the average grain size from 407 ± 149 nm for the reference MAFAPI sample, down to 188 ± 46 nm for MAFAPI + 5% BDAI 2 ( Figure 3C-F). In the case of FACsPI perovskite composition, we also observed a pronounced decrease in grain size with the addition of ≥1% BDAI 2 , while 0.25% of diammonium salt does not noticeably affect the perovskite morphology ( Figure S4). The decrease in grain size with the BDAI 2 additive may be due to the preferential localization of diammonium cations on the GBs instead of the perovskite lattice, since a too large BDA 2+ cation cannot be incorporated into 3D perovskites without severe lattice distortions. Similar trends with decreasing grain size during perovskite passivation with diammonium salts are reported elsewhere [45,46]. the perovskite solution varied in a wide range from trace 0.025 wt.% [17] to 3-8% [20,45]; therefore, we chose the following BDAI2 concentrations for 3D perovskite bulk passivation: 0.25, 0.5, 1, 2.5, and 5%. According to the XRD data of the MA0.25FA0.75PbI3 + x% BDAI2 samples ( Figure 3A), no phase impurities, such as (BDA)PbI4 or delta polytypes of FAPbI3, formed after the introduction of BDAI2 into the material. The same behavior was observed for the FACsPI + x% BDAI2 samples ( Figure S3). Interplanar distances in the perovskite lattice were also not affected by the diammonium additive ( Figure 3B). Evidently, such small amounts of additives would not be detected by XRD; however, we observed the widening of diffraction reflexes along with an increased amount of the BDAI2 additive, as shown in Figure  3B for the (100) perovskite reflex. This seems to correlate with the monotonic decrease in the average grain size from 407 ± 149 nm for the reference MAFAPI sample, down to 188 ± 46 nm for MAFAPI + 5% BDAI2 ( Figure 3C-F). In the case of FACsPI perovskite composition, we also observed a pronounced decrease in grain size with the addition of ≥1% BDAI2, while 0.25% of diammonium salt does not noticeably affect the perovskite morphology ( Figure S4). The decrease in grain size with the BDAI2 additive may be due to the preferential localization of diammonium cations on the GBs instead of the perovskite lattice, since a too large BDA 2+ cation cannot be incorporated into 3D perovskites without severe lattice distortions. Similar trends with decreasing grain size during perovskite passivation with diammonium salts are reported elsewhere [45,46]. The optoelectronic properties of perovskite films also depend strongly on the BDAI2 amount. Time-resolved photoluminescence (TRPL) data of the MAFAPI + x% BDAI2 samples reveal a dramatic decrease in the average charge carrier lifetimes tm from 97.2 ns for the reference perovskite sample down to 20.6 ns and 11.3 ns for the samples with 1% and The optoelectronic properties of perovskite films also depend strongly on the BDAI 2 amount. Time-resolved photoluminescence (TRPL) data of the MAFAPI + x% BDAI 2 samples reveal a dramatic decrease in the average charge carrier lifetimes t m from 97.2 ns for the reference perovskite sample down to 20.6 ns and 11.3 ns for the samples with 1% and 5% BDAI 2 , respectively ( Figure 4A). The lifetime decrease is accompanied by the growing proportion of fast decay components (t 1 in the inset table in Figure 4A), which are usually associated with trap-assisted recombination on the surface and at the GBs. In contrast, a small amount of diammonium additive (0.25% BDAI 2 ) leads to an increase in the average lifetime t m up to 124,1 ns and a simultaneous rise in the slow decay component t 2 , which is thought to originate from perovskite bulk [47]. Steady-state photoluminescence (PL) data of MAFAPI + x% BDAI 2 correlate with the TRPL decay curves; perovskite PL intensity increases with 0.25% BDAI 2 with respect to the reference MAFAPI sample, but is progressively quenched with the further addition of 1, 2.5, and 5% BDAI 2 ( Figure 4B). Interestingly, FACsPI perovskite composition has higher "tolerance" to the BDAI 2 additive, with a gradual increase in PL intensity up to 1% BDAI 2 ( Figure 4C). After overcoming the threshold level of 1% BDAI 2 , FACsPI films also suffer from a dramatic loss of PL intensity, which is accompanied by the widening of emission lines ( Figure 4D, red lines). In addition, the center of the PL peaks (X C ) of both perovskite compositions shifts to shorter wavelengths up to 14 nm at the maximum BDAI 2 content of 5% ( Figure 4D, blue lines). Similar trends are observed for the FACsPI + x% BDAI 2 samples by optical absorption spectroscopy ( Figure S5). This slight increase in the perovskite bandgap could be explained by the distortions of the inorganic [PbI 3 ]sublattice [46], which is predominantly located at grain boundaries [48], the proportion of which becomes higher with the increasing amount of diammonium additive according to the morphological observations ( Figures 3C-F and S4).
Nanomaterials 2022, 12, x FOR PEER REVIEW 8 of 14 5% BDAI2, respectively ( Figure 4A). The lifetime decrease is accompanied by the growing proportion of fast decay components (t1 in the inset table in Figure 4A), which are usually associated with trap-assisted recombination on the surface and at the GBs. In contrast, a small amount of diammonium additive (0.25% BDAI2) leads to an increase in the average lifetime tm up to 124,1 ns and a simultaneous rise in the slow decay component t2, which is thought to originate from perovskite bulk [47]. Steady-state photoluminescence (PL) data of MAFAPI + x% BDAI2 correlate with the TRPL decay curves; perovskite PL intensity increases with 0.25% BDAI2 with respect to the reference MAFAPI sample, but is progressively quenched with the further addition of 1, 2.5, and 5% BDAI2 ( Figure 4B). Interestingly, FACsPI perovskite composition has higher "tolerance" to the BDAI2 additive, with a gradual increase in PL intensity up to 1% BDAI2 ( Figure 4C). After overcoming the threshold level of 1% BDAI2, FACsPI films also suffer from a dramatic loss of PL intensity, which is accompanied by the widening of emission lines ( Figure 4D, red lines). In addition, the center of the PL peaks (XC) of both perovskite compositions shifts to shorter wavelengths up to 14 nm at the maximum BDAI2 content of 5% ( Figure 4D, blue lines). Similar trends are observed for the FACsPI + x% BDAI2 samples by optical absorption spectroscopy ( Figure S5). This slight increase in the perovskite bandgap could be explained by the distortions of the inorganic [PbI3] -sublattice [46], which is predominantly located at grain boundaries [48], the proportion of which becomes higher with the increasing amount of diammonium additive according to the morphological observations ( Figures 3C-F and  S4).  The comparative study of long-term photostability of MAFAPI and FACsPI perovskite films was carried out in accordance with the previously reported approach based on continuous light soaking of perovskite films in an inert glove box and periodic registration of PL signals [28]. The resulting evolution of relative PL intensity of the perovskite samples, normalized to the first measurements, is given in Figure 4E for MAFAPI and Figure 4F for FACsPI compounds. The reference samples of both perovskites undergo a dramatic drop in PL intensity after 22 h of light soaking, as 0.25% of the BDAI 2 additive is sufficient to slow down the degradation process. The further increase in the BDAI 2 content up to 5% reveals a different response of MAFAPI and FACsPI perovskites' photostability to bulk passivation. The MAFAPI films show no further stabilization of the PL signal, demonstrating threshold-like behavior and the effectiveness of small amounts of diammonium salt for the improvement of both perovskite optoelectronic properties and long-term photostability. In turn, the FACsPI perovskite composition appears to be more sensitive to BDAI 2 content with a photostability that is proportional to the passivator amount ( Figure 4F). FACsPI + 5% BDAI 2 demonstrates enhancement of the PL signal under continuous 1 sun illumination, while 2.5% of the additive preserves 80% of the initial PL intensity. These observations, together with the steady-state PL data, motivate us to choose the FACsPI perovskite composition with BDAI 2 bulk passivation as a light absorbing material in PSCs.
Hence, we investigated the influence of BDAI 2 passivation on FACsPI-based PSC performance and stability. The devices were assembled with the following planar architecture: FTO/c-TiO 2 /SnO 2 /perovskite/spiro-OMeTAD/Au and encapsulated in an inert glove box with MoO x and epoxy resin in accordance with ref. [27]. The analysis of J-V curves ( Figure 5A), PCE statistics ( Figure 5B) and dynamics of other PSC working parameters ( Figure 5C) as a function of BDAI 2 content reveals again the threshold-like behavior of the devices, where >1% BDAI 2 content initiates an almost order of magnitude PCE decrease for the 5% BDAI 2 additive, while ≤1% BDAI 2 has a slight effect on PCE ( Figure 5A,B). A comparative analysis of basic PSC working parameters, including fill factor (FF), open circuit voltage (V OC ), short circuit current density (J SC ), and PCE ( Figure 5C), normalized to the corresponding values of the control devices, reveals a strong correlation of PCE with J SC, which agrees with the previously observed decrease of the average grain sizes and charge carrier lifetimes in perovskite films ( Figures 3C-F and S4). The absolute values of all the listed parameters are given in the Supporting Information ( Figure S6). The fill factor of PSCs tends to monotonically decrease, losing up to 30% of the initial FF value, in contrast to the increasing V OC . Usually, a V OC increase is explained by the reduction in the trap densities in the device [49], and also correlates with a slight widening of the perovskite bandgap with the addition of ≥1% BDAI 2 ( Figures S5 and 4D).
We also estimated the series and shunt resistances from quasi-steady-state J-V curves and revealed similar trends, as both resistances tended to increase with growing amounts of diammonium additive ( Figure S7). It means that BDAI 2 passivation exerts two opposite effects on hybrid perovskite materials and devices, including the positive effect of decreasing the number of shunts and the negative effect of decreasing the electric conductivity in the devices.
The long-term stability tests of encapsulated PSCs were performed under constant ∼100 mW/cm 2 white light illumination and a stabilized temperature of 65 • C, using a maximum power point tracking (MPPT) regime in an ambient environment. The resulting stabilized power outputs (SPO) of the best devices with different BDAI 2 contents, normalized to the first measurements, are shown in Figure 5D. The reference FACsPI-based solar cell undergoes rapid irreversible degradation between 140 and 190 h of the experiment, being the least stable device in the row. This is accompanied by the severe deterioration of the initial perovskite morphology ( Figure 5E). In contrast, all passivated solar cells demonstrate superior stability, remaining within the 50 to 80% range of the initial SPO after 400 h of continuous photothermal aging. The least stable device within this group appears to be FACsPI + 0.25% BDAI 2 , while all other samples with more diammonium content demonstrate higher stability. The absolute values of SPO are provided in SI ( Figure S8). Importantly, the presence of BDAI 2 even at a low level of 0.25% protects the perovskite morphology and phase from degradation on a timescale of 400 h, which is illustrated by the comparative SEM images in Figure 5E. The enlarged version of these microscopy data is given in Figure S9. Interestingly, we observed unusual behavior of the SPO during the first 50 h of the light-soaking experiment with a pronounced drop in the device performance, which was restored to the initial value (in the case of 1% BDAI 2 ) or to 80-90% of the initial SPO (for reference, 0.25%, and 0.5% BDAI 2 samples) ( Figure 5D). In the literature, similar SPO evolution was previously reported for n-i-p PSCs architectures by Jung et al. [50] and Bai et al. [51] without, however, any explanation. Notably, high enough BDAI 2 contents (≥2.5%) diminish this effect. We also estimated the series and shunt resistances from quasi-steady-state J-V curves and revealed similar trends, as both resistances tended to increase with growing amounts of diammonium additive ( Figure S7). It means that BDAI2 passivation exerts two opposite effects on hybrid perovskite materials and devices, including the positive effect of decreasing the number of shunts and the negative effect of decreasing the electric conductivity in the devices.
The long-term stability tests of encapsulated PSCs were performed under constant ∼100 mW/cm 2 white light illumination and a stabilized temperature of 65 °C , using a maximum power point tracking (MPPT) regime in an ambient environment. The resulting stabilized power outputs (SPO) of the best devices with different BDAI2 contents, normalized to the first measurements, are shown in Figure 5D. The reference FACsPI-based solar cell undergoes rapid irreversible degradation between 140 and 190 h of the experiment, being the least stable device in the row. This is accompanied by the severe deterioration of the initial perovskite morphology ( Figure 5E). In contrast, all passivated solar cells demonstrate superior stability, remaining within the 50 to 80% range of the initial SPO after 400 h of continuous photothermal aging. The least stable device within this group appears to be FACsPI + 0.25% BDAI2, while all other samples with more diammonium content demonstrate higher stability. The absolute values of SPO are provided in SI (Figure S8). Importantly, the presence of BDAI2 even at a low level of 0.25% protects the perovskite morphology and phase from degradation on a timescale of 400 h, which is illus- We suppose that the mechanism of PSC stability increase with BDAI 2 passivation could originate from the intergrain "encapsulation" by either (BDA)PbI 4 phases or BDA 2+ monolayers and the additional improvement of the interconnection between perovskite grains, owing to bidentate diammonium cations. This hypothesis is primarily supported by the electron microscopy data (Figure 5E), as well as by the increasing values of shunt resistance in the passivated devices ( Figure S7). The presence of diammonium passivation at the GBs, which are known to be the main channels of ion migration [52], could also decrease the extent of extrinsic and intrinsic ionic migration-the main factor of potentialinduced degradation in various types of solar cells, including PSCs [53].

Conclusions
In conclusion, we experimentally demonstrated that (BDA)PbI 4 is the only stable 2D perovskite phase based on 1,4-butanediammonium cations, which means that (BDA)(MA) n−1 Pb n I 3n+1 compositional row is not suitable for the use as phase-pure lightharvesting materials in PSCs. The bulk passivation of 3D ABX 3 hybrid perovskites with BDAI 2 leads to the modulation of the perovskites' microstructure and optoelectronic properties depending on the amount of diammonium additive. BDA 2+ cations seem to accumulate mostly on the perovskite grain boundaries in the form of either (BDA)PbI 4 phases or BDA 2+ -containing layers, leading to the gradual decrease in the average grain size with an increased BDAI 2 content. No phase impurities were detected in the passivated samples. The influence of BDAI 2 passivation on perovskite optoelectronic properties and device performance demonstrates a threshold behavior with a positive effect on perovskite materials below the threshold (increase in PL intensity, charge carrier lifetimes, V OC ; shunt resistance) and a strongly negative effect above the threshold (decrease in PL intensity and charge carrier lifetimes, J SC decrease; series resistance increase). In the case of BDAI 2 , the threshold value was found to be~1 mol.% for FACsPI and~0.5 mol.% for MAFAPI compositions, with respect to Pb content. On the other hand, the presence of any amount of diammonium salt leads to the sufficient enhancement of the photothermal stability of perovskite materials and devices, compared to the reference samples. The performance of the passivated devices remained within the 50 to 80% range of the initial PCE after 400 h of continuous 1 sun irradiation with a stabilized temperature of 65 • C, while the performance of the control devices deteriorated after 170 h of the experiment. The most possible stabilization mechanism of BDAI 2 -passivated PSCs is assumed to originate from the intergrain encapsulation by either (BDA)PbI 4 phases or BDA 2+ monolayers and the additional improvement of interconnection between perovskite grains, owing to bidentate diammonium cations. Therefore, 1,4-butanediammonium iodide could be considered as a promising additive for the improvement of PSC stability, but the amount of this salt should be carefully controlled to improve the performance of the final devices.
Supplementary Materials: The following supporting information can be downloaded at https: //www.mdpi.com/article/10.3390/nano12244357/s1: Figure S1: XRD patterns of 2D perovskites, prepared from (BDA)(MA)Pb 2 I 7 solutions with different concentrations; Figure S2: XRD patterns of 2D perovskite crystals, prepared from (BDA)(MA)Pb 2 I 7 solutions with different excess amounts of MAI; Figure S3: XRD patterns of FA 0 . 85 Cs 0 . 15 PbI 3 perovskite films with different BDAI 2 contents; Figure S4: SEM images of FACsPI perovskite films with different BDAI 2 contents; Figure S5: Tauc plots of FA 0 . 85 Cs 0 . 15 PbI 3 perovskite films with different BDAI 2 contents; Figure S6: Distributions of operando parameters of PSCs based on FACsPI + x% BDAI 2 light-harvesting materials; Figure S7: The dependence of R series and R shunt of PSCs on BDAI 2 content; Figure S8: Absolute SPO values of encapsulated perovskite solar cells with different BDAI 2 amounts under continuous 1 sun illumination, stabilized temperature of 65 • C, and ambient environment; Figure S9: SEM images of FACsPI perovskite films with different BDAI 2 contents before and after 400 h of PSC photothermal aging.