Compositional engineering of organic-inorganic halide perovskite materials using mixed cations and/or mixed halides shows potential to achieve stable photovoltaic and optoelectronic devices with high efficiency and tunable energy levels. [1][2][3][4] Mixed cations consisting of methylammonium (MA), formamidinium (FA), cesium (Cs), or rubidium (Rb) are often used in perovskite solar cells (PSCs), with FA being the major component-with the molar fraction varying from 0.75 to 0.85-due to its desired bandgap, photo stability, and reproducibility. 1,[4][5][6][7] Although high performance has been achieved in these mixed-cation halide PSCs, their long-term operational stability still presents a critical challenge. 8,9 Even after eliminating environmental factors, e.g. humidity and oxygen, through encapsulation, the inherent chemical reactivity and volatility of organic cations remain major factors in halide perovskite material degradation under light and heat. 8,[10][11][12][13] The degradation of MAPbI 3 has been widely studied, [14][15][16][17][18][19] and the comprehensive understanding of its instability and decomposition mechanisms results in efforts to eliminate MA from halide perovskite compounds. 6,7,20 In PSCs, the perovskite, irrespective of composition, has been reported to degrade through interfacial reactions with neighbouring materials, 7,12,[21][22][23][24][25][26][27][28][29] yielding lower device stability and performance. Although recently a few studies have been performed on the decomposition of FAPbI 3 by themselves, 10,18,[30][31][32][33][34] there is little to no understanding of the interface-induced degradation in FA-based perovskites. Thus, to evaluate the stability of this material as a potential absorber in PSCs, it is imperative to identify possible interfacial reactions between the FA cation and contact layer materials.
While previous work focused on the thermal stability and degradation mechanism in formamidinium iodide (FAI, HC(NH 2 ) 2 I) and FAPbI 3 by themselves, here we investigate the thermal stability and decomposition pathway of FAI in contact with NiO, a commonly used hole transport layer material, 27,35 or TiO 2 , a commonly used electron transport layer material. 36 In our previous work, we showed that the inorganic component of the perovskite materials-PbI 2 -does not undergo any change in this temperature range (o400 1C), both by itself or in contact with metal oxides. 22 The instability of halide perovskites primarily arises from the decomposition of the organic component. Hence, studying FAI degradation will provide an understanding of FAPbI 3 stability. The thermal stability and degradation reactions are studied using thermogravimetric analysis complemented with differential scanning calorimetry (TGA-DSC), as well as the temperature-programmed desorption technique combined with mass spectrometry (MS) and Fourier transform infrared spectroscopy (TPD-MS-FTIR) for simultaneous detection and unequivocal identification of gas-phase decomposition products. The solid decomposition products are examined by X-ray diffraction (XRD). Density functional theory (DFT) modelling is employed to explain the experimental results. This combination allows us to construct an accurate delineation of the decomposition pathways.
All chemicals in this study were used as received: FAI (499.5%, Greatcell Solar Materials), NiO (o50 nm, 99.8%, Sigma-Aldrich), TiO 2 (50 nm, 99.9+%, US Research Nanomaterials, Inc.), and NH 4 I (99.999%, Alfa Aesar). The NiO and TiO 2 powders were dried in a vacuum oven at 250 1C and 150 1C, respectively, prior to mixing with FAI.
The TGA-DSC analysis was carried out in an SDT Q600 (TA Instruments). The powder samples of FAI and FAPbI 3 in contact with metal oxides were prepared by mixing FAI/FAPbI 3 with NiO and TiO 2 in 1 : 1 molar ratio using a vortex mixer (Vortex 3, IKA Works, Inc.) for 1 min. The NH 4 I in contact with NiO was also prepared in a 1 : 1 molar ratio. From the prepared samples, B8-9 mg of the powder was placed in an alumina pan and heated from 25 1C to 450 1C at a rate of 10 1C min À1 under 100 sccm of N 2 gas.
The thermal decomposition studies were performed using the TPD-MS-FTIR technique described in our previous work. 22 The samples were prepared as above. For TPD-MS-FTIR experiments, the weight of FAI in all samples was kept constant at 40 mg. For the 1 : 1 molar ratio FAI + oxide experiments, 57 mg FAI + NiO and 59 mg FAI + TiO 2 powders were used. For the 1 : 4 molar ratio experiments, 110 mg FAI + NiO and 114 mg FAI + TiO 2 powders were used. In 1 : 1 molar ratio NH 4 I + NiO experiments, 61 mg was used. The samples were placed in a quartz tube, sandwiched between quartz wool. Prior to the analysis, the sample cell, heated gas lines, MS (Vision 1000-C, MKS Instruments, Spectra Products), and FTIR (Thermo Nicolet Nexus 670) were thoroughly purged using He gas at a flow rate of 30 sccm for 30 min to minimize environmental contributions such as H 2 O, O 2 , and CO 2 . Then the samples were heated from 25 1C to 400 1C at 10 1C min À1 under 30 sccm He flow at atmospheric pressure. The gaseous products were carried to FTIR and MS instruments for simultaneous in situ gas-phase analysis. The FTIR spectra were recorded with a resolution of 4 cm À1 in the range of 650-4000 cm À1 at 5 s intervals. The mass analysis was carried out by scanning sequentially from m/z = 2 to 300 and detected with a Faraday cup.
To identify the phase and composition of decomposed solid products, we mimicked the decomposition reactions by heating powders of FAI, FAI + NiO, or FAI + TiO 2 , to 100 1C, 150 1C, 200 1C, 250 1C, and 300 1C, sequentially-holding for 10 min at each temperature-on a hot plate (Thermo Scientific) inside a N 2 purged glove box (Plas-Labs, Inc.) to prevent environmental contributions. The powder XRD data were collected on samples at each temperature using a Rigaku Ultima III diffractometer (40 kV/44 mA) equipped with Cu Ka radiation (l = 1.5406 Å) over a 2y range from 101 to 501 with a step size of 0.021 and a scan speed of 21 min À1 . The crystalline phases were determined by comparing the experimental XRD patterns with the powder diffraction files (PDFs).
DFT modelling was employed to calculate adsorption and reaction energies using the VASP program package. 37,38 We used plane wave basis expansions with an energy cutoff of 400 eV, and the Perdew-Burke-Ernzerhof (PBE) type generalized gradient approximation (GGA) for the exchange-correlation. 39 The corevalence interaction was considered by selecting the projectoraugmented wave (PAW) method. 40 All atomic positions and lattice were relaxed until residual forces and the energy change were less than 0.01 eV Å À1 and 10 À6 eV, respectively, to obtain unit cell configurations. Spin-polarized DFT and Hubbard U (U = 6.2 eV for Ni 3d electrons) correction was used to calculate NiO systems. 41,42 The Tkatchenko-Scheffler method with iterative Hirshfeld partitioning was employed to reflect van der Waals interactions. 43,44 NiO surface structures were made by multiplying converged unit cell lattice structures. Monkhorst-Pack sampling using 2 Â 2 Â 1 and 3 Â 3 Â 1 G-centered grids was used to calculate the NiO and TiO 2 surfaces, respectively. NiO/TiO 2 surfaces consisted of 128/135 atoms, and the vacuum layer was set to be larger than 12 Å. The positions of atoms below half of the slabs were fixed to mimic the surface structure. 3p3d4s and 3d4s were considered as valence states of Ti and Ni, respectively.
The adsorption energies were estimated by subtracting the surface and the adsorbed molecule energies from the total system energy. Therefore, the more negative the adsorption energies, the stronger the binding on the oxide surface.
We tried 10 different initial configurations to find optimum adsorbed structures for each case. The atomic configurations obtained from DFT calculations are shown in Fig. S2 (ESI †).
The decomposition energies of FAI on metal oxide surfaces were calculated by the evaluation of binding energies during the reactions. For example, the energy change of (HC(NH 2 ) 2 I)* = HCN* + HI* + NH 3 * (* denotes molecules attached to the surface) on NiO was evaluated as given in eqn (2):
We performed TGA-DSC analysis to determine the thermal stability of neat FAI powders, which is compared to when FAI is mixed with dried NiO or TiO 2 nanoparticles. For neat FAI (Fig. 1a), the onset of weight loss (black) occurs at B250 1C (blue dashed line) with almost complete weight loss (99.5%) by B320 1C. Two endothermic features are observed in the DSC data (Fig. 1a, red) during the FAI decomposition when it is by itself. Our neat FAI TGA-DSC results are in agreement with previous reports. 18,29 When FAI is in contact with NiO (Fig. Although interactions between FA cations and inorganic Pb-I matrices are thought to be stronger, resulting in higher structural stability in FAPbI 3 perovskite, 18 our results explicitly show that the physical contact between perovskite and metal oxides does induce intrinsic instability in these materials. In particular, contact with NiO substantially lowers the thermal stability of FAI and FAPbI 3 alike. Thus, the similar thermal stability behaviour between FAI and FAPbI 3 validates our rational to perform further degradation studies using FAI.
To identify the volatile decomposition products associated with thermal events observed in TGA-DSC, we performed TPD-MS-FTIR experiments with neat FAI, FAI + NiO, and FAI + TiO 2 powders. Such simultaneous detections of evolved gases by FTIR and MS help to accurately identify molecular species as ionization probability and fragmentation into smaller ions in MS complicate the analysis while many organic moieties have overlapping vibrational frequencies in FTIR. The top panel in Fig. 2 shows FTIR temperature profiles representing infrared absorption intensity versus temperature for evolved gases. Comparing the observed IR spectra of the gas species released from the decomposition to NIST database, 45 we assign the evolved gases at 967 cm À1 , 3600-3800 cm À1 , 739 cm À1 , and 1551 cm À1 to ammonia (NH 3 , black), water (H 2 O, blue), hydrogen cyanide (HCN, red), and sym-triazine ((HCN) 3 , green), respectively. Note that the higher wavenumber region is used for H 2 O to avoid the overlap with NH 3 signals in the 1500-1600 cm À1 region. The full FTIR line spectra at different temperatures for the three samples are shown in Fig. S4 (ESI †). In MS, a molecule can have several fragments with different mass to charge ratio (m/z) values. By comparing the intensity ratios at different m/z of all detected ions to NIST database, 46 we identify the released gases to be NH View Article Online which gases begin to evolve, to be B280 1C. While T d of the neat FAI observed here is closer to Perez et al.'s work (260 1C), 30 Ma et al. reported it to be at 245 1C. 18 The T d differences in these three works could arise from the different ramping rates used in heating the samples and the geometry of the experimental apparatus. 17 In contrast, when FAI is in contact with NiO (Fig. 2b ande), gases begin to evolve at B200 1C, 100 1C lower than T d of neat FAI. When FAI is in contact with TiO 2 (Fig. 2c andf), HCN starts to appear at B260 1C.
Correlating the TPD results with TGA-DSC results, we can attribute the first endothermic peak in DSC (Fig. 1a, blue dashed line) to bulk decomposition of neat FAI, releasing sym-triazine and HCN gases simultaneously at T d B 280 1C as detected by both FTIR and MS (Fig. 2a andd). In addition, FTIR (Fig. 2a) also shows NH 3 evolution at a higher temperature B340 1C, indicating that bulk FAI decomposition does not produce NH 3 directly; this temperature corresponds to the second endothermic peak in DSC. From these results, we can infer that the bulk decomposition of neat FAI occurs via a twostep process. Previous work on neat FAI thermal decomposition also reported sym-triazine, HCN, and NH 3 as the gaseous products, but no detection of HI or I 2 . 18,30 We also did not observe HI in FTIR or MS (Fig. 2a andd) as HI is known to adhere to cold surfaces in the apparatus. 10 The fact that NH 3 is not detected by MS (Fig. 2d) can be attributed to its low concentration, i.e. below the MS detection limit; FTIR is sensitive to the N-H symmetric deformation mode, but the intensities of these peaks are extremely low in this case (Fig. S4a, ESI, † black dotted rectangle in the zoomed-in view).
In the case of FAI + NiO, we observe two distinct degradation processes, with gaseous products of NH 3 , H 2 O, sym-triazine, and HCN (Fig. 2b ande). At 200 1C, NH 3 and H 2 O are released simultaneously, corresponding to the first endothermic peak in DSC (Fig. 1b, blue dashed line), while sym-triazine and HCN only begin to evolve at B270 1C, which aligns well with the second endothermic peak in DSC. Because the hightemperature evolved gases are the same as neat FAI and also occur at a similar temperature, we attribute this process to bulk decomposition of FAI. The low-temperature event, during which NH 3 and H 2 O are released, is a new degradation pathway that is not previously known. To accentuate the interfacial effects, we increased the molar ratio of FAI to NiO to 1 : 4. With excess NiO, both FTIR (Fig. S6a, ESI †) and MS (Fig. S6c, ESI †) results are dominated by the NH 3 and H 2 O evolution at 200 1C, confirming that the low-temperature process arises from the interaction of FAI with NiO. At the same time, the sym-triazine signal is low in FTIR and not observed in MS, indicating that a very small amount is produced whereas HCN is detected using both techniques. It is noteworthy that the T d B200 1C and the released NH 3 and H 2 O gas products are the same as observed for the decomposition of MAI in contact with NiO. 22
This journal is © The Royal Society of Chemistry 2020
Mater. Adv.
For FAI + TiO 2 samples, although there is no significant change in T d from neat FAI, HCN is released first at a lower temperature of B260 1C while sym-triazine is released at B280 1C (Fig. 2c andf). Similar to neat FAI, FTIR shows evidence of NH 3 released at B340 1C (Fig. 2c and Fig. S4c, ESI †), in agreement with the DSC result (Fig. 1c) where we observed a high temperature endothermic peak at Z300 1C. With the increased molar ratio of FAI : TiO 2 to 1 : 4, T d remains the same as 1 : 1 FAI : TiO 2 , but much more HCN is produced compared to sym-triazine (Fig. S6b andd, ESI †).
Further insight into the degradation process can be gained by examining the solid decomposed products using XRD. The XRD patterns taken before heating and after heating at each temperature are shown in Fig. S7 (ESI †). The XRD patterns of neat FAI, FAI + NiO, and FAI + TiO 1b), further substantiating the reaction between FAI and NiO, i.e. the interfacial reaction. NiI 2 has been identified previously as the reaction product of halide perovskite and NiO. 12,21,22 On the other hand, in the XRD pattern of FAI + TiO 2 (Fig. 3, bottom panel), no peaks of TiI 4 , only TiO 2 reflections (blue dumbbells), are observed at 250 1C. Thus, the absence of TiI 4 is consistent with the weight loss in TGA of FAI + TiO 2 (Fig. 1c), which corresponds to the total weight of FAI in the sample and supports the fact that no reaction occurs between TiO 2 and FAI.
Unlike MAI, the decomposition pathway and products of FAI are currently still not well understood. Although there are broad agreements on two overall reactions, 18,30,32 (3a) and (3b) there are conflicting reports on the identity and formation of Based on their FTIR results and DFT calculations, Ma et al. suggested that sym-triazine formed first (reaction (3a)) and proposed that HCN was generated from the decomposition of sym-triazine by the attack of hydrogen radicals from HI. 18 Juarez-Perez et al. also found sym-triazine as the major decomposition product in their MS data which is released above 250 1C. 30 On the other hand, Akbulatov et al. studied thermal degradation between 200 1C and 300 1C, and proposed that reaction (3b) occurred first followed by the trimerization of HCN under basic conditions to yield 2-aminomalononitrile, which has the same m/z = 81 as sym-triazine. 32 Because Akbulatov et al. only performed MS, they could not distinguish 2-aminomalononitrile from sym-triazine. These two molecules have very different characteristic vibrational frequencies because 2-aminomalononitrile contains CRN while symtriazine does not. FTIR results by both Ma et al. and this work show no evidence of the spectroscopic signature of CRN in malononitrile at B2190 cm À1 . 47 Moreover, aminomalononitrile is unstable; it would have reacted quickly with HCN to produce diaminomaleonitrile. 48 The relation between FA and symtriazine was also studied previously under the context of sym-triazine synthesis from FACl; the authors reported that high reactivity of FA arising from its negligible steric effect from H atom favours the sym-triazine route while nitrile formation occurs when H is replaced by neutral groups. 49 Therefore, based on these reports and our combined FTIR and MS results, the assignment of the gas species to sym-triazine is valid.
To explain our experimental observations and to elucidate the degradation pathways in neat FAI and when FAI contacts NiO or TiO 2 , we employed DFT calculations. The energy changes for decomposition reactions in the gas phase and on oxide surfaces are calculated with adsorption energy data sets and are represented in Fig. 4 and Table S1 (ESI †). Here our calculation results shed light on the conflicting reports of whether sym-triazine or HCN is formed first. Fig. 4 shows that, for a given sample, neat FAI, FAI + NiO, or FAI + TiO 2 , the reactions that produce sym-triazine (green dashed lines) have lower energies than equivalent reactions that produce HCN (red dashed lines). Thus, sym-triazine is a thermodynamically favored product over HCN, i.e. reaction (3a) dominates over (3b). The middle panel in Fig. 4 shows that contact with NiO (pink pentagons) and TiO 2 (blue dumbbells) lowers the energies of both reaction (3a) and (3b) compared to neat FAI (black circles). The reaction energies decrease as FAI 4 FAI + TiO 2 4 FAI + NiO, consistent with the decrease in T d observed in TPD-FTIR-MS (Fig. 2), i.e. 280 1C for FAI 4 260 1C for FAI + TiO 2 4 200 1C for FAI + NiO. Note that our calculations present the thermodynamic energy changes for selected reactions, but not the activation energy barrier heights. Thus, the negative energy of reaction (3a) on NiO indicates that the reaction is thermodynamically favoured but does not mean that it will occur spontaneously.
While Perez et al. and Akbulatov et al. proposed NH 4 I in FAI or FAPbI 3 decomposition, 30,32 our experimental results are the first to unambiguously identify its presence in the decomposition products. The XRD results of partially decomposed FAI after 250 1C heat treatment (Fig. 3, top panel) show the existence of sym-triazine (green inverted triangles) and NH 4 I (maroon squares). Our DFT calculations (Table S1, ESI †) show that when FAI decomposes, the formation of NH 4 I from NH 3 and HI is energetically favorable (À0.45 eV). In this case (black circle), the lowest energy reaction produces sym-triazine + NH 4 I (0.24 eV, green dashed line, right column), while the reaction that produces sym-triazine + NH 3 + HI, i.e. reaction (3a), has a higher energy of 0.69 eV (green dashed line, middle column). Moreover, the sym-triazine + NH 4 I reaction requires less energy than the equivalent reaction that produces HCN + NH 4 I (1.2 eV, red dashed line, right column). We therefore propose that FAI first transforms to sym-triazine and NH 4 I. sym-triazine and HCN readily desorb at B280 1C in the TPD experiment (Fig. 2a andd), while NH 4 I is still a solid at this temperature. As the temperature further rises, NH 4 I undergoes complete decomposition to NH 3 + HI by 340 1C with NH 3 being detected (Fig. 2a). The TGA-DSC of neat NH 4 I shown in Fig. S8 (ESI †) provides further evidence that NH 4 I decomposes Z300 1C. Comparing the DSC curves (red) of NH 4 I with FAI (Fig. 1a), the endothermic peak at B330 1C matches well with the second endothermic peak in FAI, supporting our hypothesis that FAI decomposition occurs in two steps with the release of NH 3 (Fig. 2a) as the result of NH 4 I decomposition. Since the reaction to form NH 4 I is highly exothermic, the reformation of solid NH 4 I on the colder parts of the experimental apparatus can occur, which has been suggested previously. 32 In neat FAI and FAI + TiO 2 TPD experiments, we observed white deposits lining the exhaust capillary of the cell. Comparing the XRD pattern of this white deposit (Fig. S9, ESI, † orange) to neat NH 4 I (Fig. S9, ESI, † maroon), it is identified as NH 4 I. Furthermore, NH 4 I TPD results show no NH 3 or HI (m/z = 128) gases (Fig. S10a,ESI †) and a large amount of white deposits in the exhaust capillary of the sample cell (Fig. S10b,ESI †). Thus, based on these results, we determine that the neat FAI undergoes decomposition via sym-triazine + NH 4 I first and NH 4 I further decomposes to NH 3 + HI at higher temperature, with the possibility of NH 4 I reformation on colder surfaces.
A notable difference in neat FAI vs. FAI + oxides is the decomposition via NH 4 I vs. NH 3 + HI. Since solid NH 4 I is not stable on either NiO (0.21 eV) or TiO 2 (0.67 eV) surfaces (Table S1, ESI †), it dissociatively adsorbs as NH 3 * and HI* on the oxide surfaces instead of as NH 4 I*. This difference in the adsorption characteristics on oxide surfaces explains the presence of NH 4 I (maroon squares) in the XRD of neat FAI (Fig. 3, top panel), while a lower amount is found in that of FAI + NiO (Fig. 3, middle panel) and none in that of FAI + TiO 2 (Fig. 3, bottom panel). Hence, the decomposition of FAI on NiO or TiO 2 follows decomposition reaction (3a). We next analyse the effects of oxide surfaces on FAI decomposition pathways based on the adsorption energies of molecules summarized in Table 1. As shown in Fig. 4, FAI decomposition into sym-triazine, NH 3 , and HI (reaction (3a)) on NiO has a significantly lower reaction energy (À0.09 eV, green dashed lines, middle column) compared to neat FAI decomposing into sym-triazine and NH 4 I (0.24 eV, green dashed lines, right column). The low energy of reaction (3a) for FAI + NiO is in good agreement with the experimental results, where we also observed NH 3 (not NH 4 I)
This journal is © The Royal Society Chemistry 2020
Mater. Adv.
and H 2 O at 200 1C, and sym-triazine and HCN at 270 1C (Fig. 2b ande). The low desorption temperature of NH 3 (200 1C) on the NiO surface is attributed to its low binding energy (À0.77 eV, Table 1) and hence it can be readily desorbed from the surface. On the other hand, the HI gas that must be formed as a decomposition product and adsorbed on the surface as HI* reacts further with NiO producing H 2 O and NiI 2 . H 2 O desorbs from the surface as water vapor and is detected in both FTIR and MS (Fig. 2b and e, blue), and NiI 2 remains as a solid decomposed product as observed in XRD (Fig. 3, middle panel, violet diamonds). Thus, the detection of NiI These results suggest that the intrinsic stability is dictated by the oxide, rather than the perovskite. While interfacial decomposition is evident for FAI + NiO with lower T d , the released gases at 200 1C are only NH 3 and H 2 O. The reason why sym-triazine is not detected at 200 1C along with NH 3 is due to its higher adsorption energy than NH 3 * (À0.87 eV vs. À0.77 eV, Table 1). Thus, we observe sym-triazine from both interfacial and bulk decomposition starting at 270 1C as a result of its strong adsorption energy on the NiO surface. While HCN has a lower adsorption energy (Table 1), the reason that HCN is not observed at low temperature is because on the NiO surface, the reaction energy for FAI decomposing to produce sym-triazine is lower than HCN (À0.09 vs. 0.75 eV, Table S1, ESI †). Therefore sym-triazine, not HCN, is the decomposition product at 200 1C. At higher temperatures, configurational entropy favors HCN over symtriazine. The free energies of HCN and sym-triazine can be estimated by considering entropy contribution using the values from JANAF table: ÀTDS @227 1C = À1.15 eV. 50 The entropy of sym-triazine is assumed to be 1/3 of the entropy of HCN. The free energy difference then becomes 0.07 eV at 227 1C and À0.11 eV at 327 1C. This implies that HCN can be generated from sym-triazine between 227 1C and 327 1C, without revoking the attack of hydrogen radicals from HI previously proposed. 18 However, the formation energy for HCN from sym-triazine is higher by 0.84 eV on NiO (Table S1, ESI †) compared to 0.58 eV for FAI + TiO 2 . Therefore, HCN evolves at 270 1C along with symtriazine on the NiO surface. The lower reaction energy along with lower T d supports the dominance of the interfacial reaction when FAI is in contact with NiO. Based on these results, the interfacial decomposition reaction of FAI + NiO can be written as: (4) In contrast to NiO, the energy FAI decomposition on the TiO 2 surface according to reaction (3a) is 0.42 eV (green dashed lines, middle column) which is slightly higher than neat FAI decomposing into sym-triazine and NH 4 I (0.24 eV, green dashed lines, right column), so FAI + TiO 2 mostly follow the bulk FAI decomposition pathway. Experimentally, we do not observe a significantly different T d (Fig. 1c, 2c andf). Similar to neat FAI, sym-triazine, HCN, and NH 3 are the decomposed gaseous products. The decomposition of FAI + TiO 2 should also produce HI. However, there is no reaction between TiO 2 and HI* as the formation enthalpy of TiI 4 is less negative compared to NiI 2 (À0.9 eV vs. À2.4 eV), 51,52 and is consistent with no TiI 4 in the XRD (Fig. 3, bottom panel). Moreover, sym-triazine and NH 3 are released at higher temperatures of 280 1C and 340 1C, respectively, while HCN is detected at a lower temperature of 260 1C. This is because the adsorption energies of sym-triazine* (À1.42 eV), NH 3 * (À1.43 eV), and HI* (À2.15 eV) are quite strong on TiO 2 compared to the NiO surface, which explains why we cannot detect them under 280 1C. On the other hand, the HCN* binding energy (À0.83 eV) is significantly lower than the other three molecules and HCN formation energy from sym-triazine is also lower (À0.58 eV), and readily desorbs from the TiO 2 surface.
It is noteworthy that the HCN formation is also affected by the oxide surface. We see that a higher amount of HCN is formed on both NiO and TiO 2 surfaces compared to neat FAI. This is because the formation energy of HCN from sym-triazine on NiO (À0.84 eV) and TiO 2 (À0.58 eV) surfaces is lower compared to neat FAI (À0.94 eV). In addition, the adsorption energies of HCN on both NiO and TiO 2 surfaces are lower than sym-triazine (Table 1). Thus, the lower formation energy compared to FAI coupled with lower adsorption energy of HCN on oxide surfaces explains the larger amount of HCN detection on the FAI + TiO 2 (Fig. 2f) and FAI + NiO samples (Fig. 2e), in particular when FAI is mixed with excess oxides (1 : 4 molar ratios, Fig. S6,ESI †).
Finally, the decomposition of FAI to gas-phase FA and HI (Fig. 4, orange solid lines and Table S1, ESI †) for all three cases is unfavourable, consistent with TPD results where we did not observe FA (m/z = 44) in any samples.
In conclusion, we have shown that interfacial interaction between the perovskite and metal oxide contact layer can trigger degradation and induce instability in PSCs. The bulk decomposition of FAI occurs at 250 1C via a two-step decomposition process: FAI first decomposes to sym-triazine and NH 4 I B280 1C, and then NH 4 I further decomposes to NH 3 and HI Z 300 1C. Among the two oxides, NiO is much more reactive and T d is lowered to 200 1C compared to the bulk T d of neat FAI. The interfacial reaction between FAI and NiO releases NH 3 and H 2 O at 200 1C while producing NiI 2 as a solid decomposed product; on the other hand, sym-triazine, from both interfacial and bulk decomposition of FAI, is released at 270 1C due to its strong adsorption energy on the NiO surface. The interfacial decomposition temperature reported here is similar to that of MAI in contact with NiO, 22 indicating the fundamental importance of oxide transport layer materials on perovskite device stability. FAI adsorbed on the TiO 2 surface slightly lowers T d , but the stability of TiO 2 , relative to NiO, prevents chemical reactions from taking place. The similar thermal stability behaviors of FAI and FAPbI 3 further emphasize that the degradation studies should not be performed on the perovskite material alone, but should also consider the chemical reactivity of perovskites with materials that they might come into contact with, so that strategies to propel PSCs towards commercialization can be developed.
This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.
Fig. 3 XRD patterns of neat FAI (top panel), FAI + NiO (middle panel), and FAI + TiO 2 (bottom panel) after heating to 250 1C (brown) for 10 min. Peaks associated with crystalline NH 4 I, sym-triazine, NiI 2 , and TiO 2 are marked by maroon squares, green inverted triangles, violet diamonds, and blue dumbbells, respectively.