Organic-inorganic halide perovskites have demonstrated superior optoelectronic properties, making them great candidates in various applications.1-3 In particular, perovskite solar cells (PSCs) have leapfrogged several technologies in power conversion efficiencies. 4 For these halide perovskite materials to become viable technologies, stability is a critical issue. Many works have examined the environmental stabilities of PSCs, in particular due to oxygen, humidity, and UV light. 5 To this date, only a few studies have examined the intrinsic thermal6-10 and photo11-12 stability of organic-inorganic halide perovskites. Degradation of CH3NH3PM3 (MAPbE) by itself can proceed in two paths: CH3NH3Pbl3(s) = Pbl2(s) + HI(g) + CH3NH2W

(1) or CH3NH3Pbl3(s) = Pbh(s) + CH,I(g) + NH,(g).

(

Both reactions are endothermic with approximately the same entropic contributions, with reaction (2) favored thermodynamically.8,10 Experimental results depend on ambient conditions: in vacuum, release of HI + CH3NH2 (reaction (1)) is observed at low temeparture,11-13 but reaction (2) is reported for experiments performed under inert carrier gas at high temperatures and atmospheric pressure.6,9,14 Furthermore, reaction (1) is believed to be reversible, while reaction (2) is irreversible and the true degradation pathway.12 Because the bulk degradation temperature of MAPbE or CH3NH3I (MAI) is > 250 °C,6,9 above the typical processing temperature of PSCs, the intrinsic degradation studies have garnered less attention from researchers than environmental effects. However, even at low temperatures, thermodynamically driven reactions still take place, albeit at a slower rate. The temperature of gas release onset indicates the material's stability.

To date the intrinsic stability studies focused on the degradation mechanism of halide perovskites by themselves, i.e. bulk decomposition. However, in devices, the halide perovskite is in contact with other materials. For example, NiO has been used as the hole transport layer in both solar cells and LEDs.15,16 TiO2 and SnO2 are widely used as the electron transport layer in PSCs.17 Reactions between MAPbR and oxides have been reported previously,18-20 with a recent report of MAI reacting with NiO as low as 120 °C.20 With improved packaging the detriment of environment can be greatly reduced, but the interactions between halide perovskites and contact materials cannot be eliminated as contact layers are integral to the devices. Thus, understanding possible reactions from interfacial interaction is critical to address the intrinsic stability of perovskite devices.

In this study, the thermal stability and degradation pathways in MAPbb and MAI, when in contact with NiO, TiO2, or SnO2, are examined using a combined temperature programmed desorption-mass spectrometry-Fourier transform infrared spectroscopy (TPD-MS-FTIR)

technique. The simultaneous MS and FTIR measurements are essential for unambiguous identification of evolved gases, because MS detects ions and are complicated by ionization and fragmentation probabilities while FTIR measures the characteristic vibrational frequencies associated with specific moieties in neutral molecules, but often exhibits similar or overlapping spectra. The experimental results are augmented by thermodynamic calculations and density functional theory (DFT) modelling. The experimental and computational methodology are described in Supporting Information.

We first compare the thermal stability and decomposition pathway in pure phase MAPbR powders21 (Figure S1) to MAPbR mixed with NiO, TiO2, or SnO2 nanoparticles. Dry powders are used to avoid effects due to complexing with solvents.7,14 The powder samples also accentuate the interfacial decomposition pathway (Table S1). The TPD was performed in flowing He gas (30 sccm, 760 Torr) to minimize exposure to oxygen or water vapor, i.e. environmental effects. Figure 1a shows the two-dimensional (2D) contour plots of FTIR results: When MAPbl? is mixed with NiO, it is clear that gases begin to evolve at ~ 270 °C, ~ 95 °C lower, while there is no significant change in Tth when mixed with Ti02 (370 °C) and a small reduction with Sn02 (Tth ~ 320 °C). The FTIR spectra at the temperature with maximum signals are shown in Figure S3. Figure 1b shows MS results (total gas output over the entire temperature range) for the four cases, with evolved gases identified by their mass to charge ratio (m/z) according to NIST database22 and summarized in Table S2. The TPD profiles for the four cases are shown in Figure S4. For neat MAPbF, both FTIR and MS detected only NH3 and CH3I, suggesting that the bulk decomposition proceeds through the reverse N-methylation reaction6,23

(reaction ( 2)). However, when MAPbF is mixed with NiO, H2O and CH3NH2 become the decomposition products, signaling a completely different degradation pathway, i.e. acidic hydrolysis reaction. When mixing with TiO2, MAPM3 decomposition produces the same gases as neat MAPbF, but when mixing with SnO2, the decomposition products are the same as MAPbF + NiO, albeit the Tth is ~ 50 °C higher. The excellent agreements between FTIR and MS provide unambiguous evidence that the degradation pathway can be altered when MAPbI3 is in contact with metal oxides, and depends on the metal oxides. The lower reaction temperature at the MAPbI3/NiO interface indicates lower intrinsic stability of devices when NiO is used.

To understand these experimental results, we perform thermodynamic free energy calculations based on the reaction suggested in Ref. 20. They presented that water vapor and nickel(II) iodide could be produced by the hydrogen iodide reaction with the NiO: % NiO(s) + HI(g) = % NiI:(s) + % %O(g) By combining the above equation with equation (1), we obtain

a reaction now possible when MAPbb is mixed with NiO. Similarly, when MAPbb is mixed with TiO2, it is possible to have CHaNHsPb#) + % TiO:(s) = % %O(g) + CH,NH2(g) + Pbh(s) + % TiI4(s),

and with SnO2, it is possible to have CHaNHsPb#) + % SnO2(s) = % H2O(g) + CH3NH2W + PbI2(s) + % SnI4(s).

(

Free energy change (AG° ) for reactions (1)-( 5) are evaluated using values from experimental data,8 NIST-JANAF Thermochemical Tables,24 and Materials Project25 at ambient temperature, and are compared in Table S3. For bulk decomposition of MAPbL, reaction (2) (AG° = 91 kJ/mol) is favored over reaction (1) (AG° = 126 kJ/mol). Note that if reactions (3)-( 5) take place, the evolved gases become H2O and CH3NH2, rather than NH3 and CH3I. It is clear that reaction Since none of the PbI2, metal oxides by themselves, or NiO mixed with PbI2 shows any gas evolution (Figure S5), the instability of MAPM3 primarily arises from MAI decomposition.

Thus, we further perform TPD-MS-FTIR experiments using MAI. To investigate the competition between bulk degradation occurring in neat MAI and interfacial degradation that happens when MAI is in contact with oxide, the molar ratio of MAI:NiO is varied from 1:0, 1:0.25, to 1:4.

Figure 2a shows the 2D contour plots of FTIR spectra at different temperatures and Figure 2b shows the MS results. As the results show, MAI alone The similar behaviors between MAI and MAPbF, whether alone or mixed with NiO or TiO2, further provide justification for performing DFT modelling using MAI to understand why degradation pathway of is changed when contacting with NiO, but remains the same when contacting TiO2. In the DFT calculations, we start with MAI molecule in the gas phase and compare the two possible intrinsic degradation pathways:

or CHsNHsI(g) = NHs(g) + CHsI(g). ( 7)

The results show that MAI decomposition in the gas phase is endothermic with 64 kJ/mol and 26 kJ/mol for reaction ( 6) and ( 7), respectively. Thus, decomposition to NH3 and CH3I (reaction (

where * denotes molecules attached to surfaces, now depend on the type of oxide substrates (Figure S10). The adsorption energies of CH3NH3I, HI, CH3NH2, CH3I, and NH3 and reaction energies of ( 8) and ( 9) (calculated from adsorption energies) are summarized in Table 1. The calculation results indicate that the MAI decomposition into HI + CH3NH2 is favored over the NH3 + CH3I pathway on NiO and TiO2 surfaces. On NiO, the reaction energies of ( 8) and ( 9) are -9.35 kJ/mol and 50.8 kJ/mol, respectively (Figure 3a). Thus, the preferred degradation pathway of MAI is altered when it is adsorbed on the NiO surface. This change in the reaction pathway has profound impact on the stability of MAPM3. Second, CH3NH2 is the desorbed gas from reaction (8) because HI binding on oxide surface is much stronger (higher adsorption energy), consistent with experiments where no HI gas is observed. On NiO, CH3I* has the lowest adsorption energy of -47.4 kJ/mol. If reaction (9) were to occur on NiO, we would have detected CH3I in the evolved gases, which is not the case. This observation further supports the change in reaction pathway at the interface. The dissociation of HI* to H* on O and I* on Ni results in the formation of H2O, which desorbs as water vapor and hence detected in both FTIR and MS, while NiI2 remains as a solid by-product (Figure S11a). The lower Tth when mixed with NiO than with SnO2 arises from the more negative formation enthalpy of NiI2 compared to SnI4 (Table S3).24 Lastly, molecular binding on TiO2 is much stronger than on NiO, indicating that these molecules will remain on TiO2 surface. This difference in MAI binding on NiO vs. TiO2 surface arises from the presence of density of states (DOS) of MAI (green) near the valence band edge for NiO (red, Figure 3b), while the same states are deeper for TiO2 (blue, Figure 3c). Hence, once a monolayer is built up on TiO2, further adsorption or reaction cannot proceed, i.e. the surface becomes inert. Thus, the only remaining degradation pathway is bulk decomposition, as shown in films compared to powder samples as films have higher surface-to-volume ratio. This result indicates that the interfacial degradation in perovskite devices could be significant at typical processing temperatures.

In summary, we show that interfacial reaction between oxide transport layers and perovskite active layer can lower the intrinsic stability of MAPbF. DFT calculations reveal that the energetics of surface reaction favor the alternative reaction pathway. Our experimental results and thermodynamic data substantiate that the stability decreases in the following order: MAPbb