The presence of CH₃NH₂ neutral species in organometal halide perovskite films

Solar cells based on organometal halide perovskite have been evolving rapidly. The newly achieved efficiencies exceeding 20%¹ make perovskite solar cells a promising candidate to compete with even the highest efficiency single crystal Si cells.^2–5 Organometal halide perovskite shows several attractive properties for photovoltaic applications, such as wide absorption photon range (∼800 nm),² weak exciton-binding energy (<50 meV),⁶ and high carrier mobilities (electron ∼ 8 and hole ∼ 12 cm²/V s).^7,8 To further improve device performance, it is crucial to understand the exact chemical structure and the precise role of each interface in devices.⁹ Especially, several research groups studied the role of the organic cation CH₃NH₃⁺ for a better understanding about the material stability and degradation mechanism.^10,11 Although a few theoretical studies have been reported,^12–14 experimental studies are somewhat lagging behind. If we understand the exact chemical structure and the precise role of each interface in perovskite solar cells, we can establish an optimized fabrication method, taking into consideration necessary interface treatment on electron and hole transport layers, minimizing degradation factors, optimized material design, etc.

In this study, we used near-edge x-ray absorption of fine structure (NEXAFS) and high-resolution x-ray photoelectron spectroscopy (HRXPS) with synchrotron radiation to understand the exact chemical structure of organometal halide perovskite at the surface as well as in the interior of the bulk film. We found the presence of CH₃NH₂ neutral species at surface and in the interiors of organometal halide perovskite. Based on our experimental results and density functional theory calculations, we propose that the neutral CH₃NH₂ species mainly reside either on the surface or at grain boundaries in the interior of CH₃NH₃PbI₃ films.

The patterned, fluorine-doped tin oxide glass (FTO Pilkington, 7 Ω/◻) was prepared with HCl and Zn powder and cleaned. On this substrate, we deposited a 100-nm layer of compact layer (cl) of TiO₂ using spray pyrolysis with a precursor solution of acetylacetone, Ti (IV) isopropoxyde, and anhydrous ethanol (3:3:2) and then performed post-annealing at 480 °C on a hotplate. Basically, perovskite thin film fabrication followed a modified version of the two-step method of Burschka et al.¹⁵ First, a solution of PbI₂ in dimethylformamide (460 mg ml⁻¹) was prepared and stirred at 70 °C for ≥2 h. We performed spin-coating at 6000 rpm for 30 s on the cl-TiO₂ substrates, previously heated at 70 °C. After the spin coating, the PbI₂ layer was dried at 70 °C for 20 min. Then, a 20 mg/ml methylammonium iodide (MAI) solution in 2-propanol (IPA) was prepared and kept at 70 °C. PbI₂ films were dipped in the MAI solution for 30 s with gentle shaking of the substrates. After dipping, substrates were rinsed in copious IPA and dried immediately by spinning them on a spin-coater, after which they were annealed for 20 min on the hot plate at 70 °C. To confirm the film thickness and surface morphology, we used an AFM (MFP-3D Asylum Research) and scanning electron microscopy (Quanta 250 FEG, FEI). X-ray diffraction (D8 Discover, Bruker Corporation) and UV-Vis spectrophotometry (Evolution 600 UV-Vis spectrophotometer, Thermo Scientific) were used to determine the crystalline structure and optical bandgap, respectively. All NEXAFS measurements were performed at room temperature and at the beamline 4D of Pohang Light Source II (PLS II), Korea. We used the total electron yield (TEY) and partial electron yield (PEY) detection mode for NEXAFS spectra by recording the sample current normalized to a signal current, which was measured simultaneously using a gold mesh in ultrahigh vacuum (<10⁻⁹ Torr). In this case, a p-polarized (≈85%) synchrotron photon beam, with an energy in the range of 279–320 eV and a spectral energy resolution of ΔE = 150 meV, had probing depths of 6 nm and 25 nm, for PEY and TEY modes, respectively. For HRXPS measurements, we performed two experiments using synchrotron radiation (52–860 eV) and a monochromated AlKα (1486.6 eV) source. HRXPS spectra were obtained at the 10D HRXPS beamline of PLSII. Photon energy was varied from 50 eV to 860 eV to obtain high-quality XPS spectra. Photoelectron signals were recorded with a PHOIBOS 150 electron energy analyzer equipped with a two-dimensional charge-coupled device (2D CCD) detector (Specs GmbH), collecting photoelectrons from the surface normal. The base pressure of the main chamber was maintained below 9.8 × 10⁻¹¹ Torr. In the case of lab source, we used the monochromated AlKα (1486.6 eV) and a multi-channel plate analyzer with 180° hemispherical and 165 mm mean radii in the AXIS Ultra DLD (Kratos Surface Analysis). The energy resolution was <0.5 eV in both techniques. Chemical states of all samples were obtained measuring the valence region and the Pb 4f, I 3d, C 1s, and N 1s core-levels spectra. O 1s core-level was also measured to monitor any contamination during the sample transfer. O 1s core-level intensity was below the detection limits of the instruments. Binding energies were calibrated with reference to the Au 4f_7/2 level (84.0 eV).¹⁶ We fitted both C and N 1s core-level peaks using Doniach-Sŭnjić curves convoluted with Gaussian distribution of 0.5 eV full-width at half-maximum.¹⁷ Background due to inelastic scattering was subtracted by the Shirley (or integral) method.¹⁸ All calculations were performed using density functional theory (DFT) as implemented in the code of the CASTEP.¹⁹ We used ultrasoft Vanderbilt-type pseudopotentials²⁰ and the exchange-correlation functional of the Perdew-Burke-Ernzerhof expression revised for solids (PBEsol) in the generalized gradient approximation (GGA).^21,22 Electronic wave functions were expanded by plane waves with an energy cut-off of 280 eV. The k-space integration was performed using a 3 × 3 × 1 mesh in the Brillouin zone (BZ) of a three-dimensionally periodic 3 × 3 × 1 supercell. The geometry optimization for the tetragonal phase was carried out until the Hellmann-Feynman force acting on the atoms was smaller than 0.03 eV/Å. To obtain the most stable atomic structure for (CH₃NH₃,CH₃NH₂)PbI₃, various initial atomic configuration were considered and all the atoms in the supercell were fully relaxed without any symmetry constraint. Also, we used the Mulliken charge analysis method to analyze charge populations. The binding energy E_safe had the following form:

where $m$ is the number of CH₃NH₃ molecules, $n$ is the number of CH₃NH₂ molecules, $Ebindingm,n$ is the binding energy of CH₃NH₃ molecule m and CH₃NH₂ molecule n in the PbI₃ 3 × 3 × 1 unit cell, $Etotalm,n$ is the total energy of CH₃NH₃ molecule m and CH₃NH₂ molecule n in the PbI₃ 3 × 3 × 1 unit cell, $EPbI3$ is the total energy of the PbI₃ 3 × 3 × 1 unit cell, $ECH3NH3$ is the energy of CH₃NH₃ single molecule, $ECH3NH2$ is the energy of CH₃NH₂ single molecule, $EbindingCH3NH3$ is the binding energy of CH₃NH₃ single molecule, $EbindingCH3NH2$ is the binding energy of CH₃NH₂ single molecule, $ECH3NH3+PbI3total$ is the total energy of 16 CH₃NH₃ molecules in the PbI₃ 3 × 3 × 1 unit cell, and $ECH3NH2+PbI3total$ is the total energy of 16 CH₃NH₂ molecules in the PbI₃ 3 × 3 × 1 unit cell

In these forms, we considered various ratios of CH₃NH₃ and CH₃NH₂ molecules in the tetragonal-phase PbI₃ structure with 16 kinds of mixing conditions. In the equation, negative and positive values of E_safe represent stable and unstable structures of (CH₃NH₃,CH₃NH₂)PbI₃, respectively.

The solution-prepared CH₃NH₃PbI₃ (MAPbI₃) film was measured by NEXAFS with synchrotron radiation. We measured both partial (PEY, probing-depth of ∼6 nm) and total electron yield (TEY, ∼25 nm) modes to see whether electronic structure differed between surface and bulk organometal halide perovskite.²³ Different chemical states are observed only in PEY mode (Figure 1(a)) at the binding energy of ∼285 eV. The rest of the spectra are almost the same between the PEY and TEY mode measurements in solution-prepared MAPbI_3. The most significant absorption features are the two shoulders at the binding energy of ∼290 eV (Figure 1(a)). To understand these surface and bulk features, we performed curve-fittings for both of them (Figure 1(b)). We found σ*(CH₃-NH₃⁺), σ*(CH₃-NH₂), and σ*(C*=C) at 290.3, 292.8, and 301.4 eV in both measurements.²³ The carbon chemical states C*-M (Metal) and π* (C*=C) were found at 282.9 and 285.5 eV in the PEY mode.²³ In the TEY mode, however, we could not observe any of these chemical states in the energy range from 280.0 to 285.5 eV.

The photoelectron escaping depth in a thin film depends on the energy of incident photons.²⁴ To investigate chemical structure differences between surface and bulk, we performed high-resolution x-ray photoelectron spectroscopy with synchrotron radiation and a monochromated Al Kα source. The Pb 4f_7/2 and I 4d_5/2 core-levels spectra at binding energies of 138.7 and 49.7 eV show no significant change in chemical states or shapes for all photon energies measured (Figure 2(a)). There is only a single chemical state for each of these peaks. In the case of C 1s core-level spectra, however, we observed different shapes when different incident photon energies (370, 620, and 1486.7 eV) were used (Figure 2(b)). Using photo energy of 370 eV, the C 1s core-level spectrum shows some tail states around 282.5 eV that do not appear in the case of 620 or 1486.7 eV photon energy. The escaping depth of photoelectrons (inelastic mean free path) with a photon energy of 370 eV is ∼1 nm, while the escaping depth for 620 or 1486.7 eV are ∼2 nm and ∼5 nm, respectively.²⁴ The presence of these tail states only for low photon energy of 370 eV suggests that these tail states are associated with the surface of the solution-prepared MAPbI₃ thin film. In the case of N 1s core-level spectra, we could not observe any significant surface states (Figure 2(b)). To understand chemical states of C and N 1s core-levels, we performed curve fittings and obtained several chemical states of C1 (287 eV), C2 (285.4 eV), S1 (283.2 eV), S2 (281.2 eV), N1 (403.3 eV), and N2 (402.7 eV) (Figure 2(b)). We propose that the chemical states of C1, N1 and C2, N2 are from CH₃-NH₃⁺ and CH₃-NH₂.^25,26 S1 and S2 are originated from the surface states of C*-M (Metal) and C*=C bonding.^27,28 These HRXPS results are consistent with the NEXAFS results.

Finally, we found CH₃NH₂ in the organometal halide perovskite with different chemical states between bulk and surface. On the other hand, Pb and I exhibited the same single chemical states in both locations. If CH₃NH₂ resides inside the perovskite crystal structure, it is expected to introduce more than one chemical states of Pb or I. To determine the location of CH₃NH₂ in the perovskite, we performed the air exposure experiment (Figure S2 in the supplementary material).²⁹ It has been reported that water (H₂O) in air degrades perovskite.^30,31 Specifically, CH₃NH₃⁺ becomes CH₃NH₂.^30,31 If the C1 and C2 chemical states are originated from CH₃NH₃⁺ and CH₃NH₂, respectively, it is expected that the ratio between C1 and C2 intensities decreases after air exposure. We performed the air exposure experiment at 24 °C and 30% humidity and then measured the sample with HRXPS after 1 and 3 days (Figure S2 in the supplementary material).²⁹ Indeed, the ratio between C1 and C2 intensities decreased after air exposure (Figure S2 in the supplementary material),²⁹ which is consistent with the proposal that the C2 chemical state is originated from CH₃NH₂ neutral species.

Based on these experimental observations, we propose that CH₃NH₂ exists in solution-prepared organometal halide perovskites. In addition, one important question arises regarding the location of CH₃NH₂ in CH₃NH₃PbI₃ films. Therefore, we further investigated the charge populations, density of states (DOS), and system stability of CH₃NH₃ and CH₃NH₂ molecules in PbI₃ using DFT-based simulations. In these calculations, we considered the tetragonal-phase PbI₃ structure (Figure 3(a)).³² First, we investigated the atomic charge states and the C-N bond length of the isolated CH₃NH₃ and CH₃NH₂ molecules (Figure 3(b)). Basically, negative or positive values of charge populations are represented by electron accumulation or depletion relative to what it had. Charge populations of CH₃NH₃ and CH₃NH₂ molecules are changed in the tetragonal-phase PbI₃ in comparison with those of isolated CH₃NH₃ and CH₃NH₂ (Figures 3(c) and 3(d)). In particular, the change of charge population of I atoms is very sensitive to the presence of CH₃NH₃ or CH₃NH₂ molecules in PbI₃. We used the binding energies (E_safe) for the stability of possible structures when CH₃NH₃ and CH₃NH₂ coexist in PbI₃ (Table S1 in the supplementary material).²⁹ The calculated results of E_safe show positive values for all configurations of (CH₃NH₃, CH₃NH₂)PbI₃. This indicates that CH₃NH₂ is unstable in the interior region of PbI₃, and instead, it should be located at grain boundaries of thin polycrystalline films of CH₃NH₃PbI₃. In this case, CH₃NH₂ cannot directly interact with the basic building blocks of PbI₃. This conclusion is consistent with the experimental observation of a single chemical state for Pb and I in HRXPS measurements. In addition, DFT calculations of the DOS show different band gaps for CH₃NH₃PbI₃ and CH₃NH₂PbI₃; the band gap of CH₃NH₃PbI₃ was ∼1.70 eV, while CH₃NH₂PbI₃ exhibits no band gap (Figure S3 in the supplementary material).²⁹ The DFT result of CH₃NH₂PbI₃ with no band gap is inconsistent with the experimental result of optical band gap of ∼1.55 eV (Figure S1 in the supplementary material).²⁹ This also suggests that CH₃NH₂ cannot reside stably inside the CH₃NH₃PbI₃ crystal structure, but rather resides mainly on the surface or at grain boundaries in the interior of thin polycrystalline films of CH₃NH₃PbI₃. A direct correlation between the concentration of CH₃NH₂ in perovskite cells and their solar cell performance would be particularly helpful. But, unfortunately, it is challenging to accurately determine the total concentration of CH₃NH₂ for a bulk film of perovskite with a typical thickness of ∼300 nm as those used in perovskite solar cells. Nevertheless, such a correlation can be partially inferred by the study that investigated the effects of annealing conditions on perovskite solar cell performance.³³ In that study, the solar-cell efficiency of 10.7% was obtained for perovskite solar cells using perovskite films with post-annealing at 115 °C in N₂ environment, while the efficiency was substantially lower (6.7%) with post-annealing at 105 °C.³³ Note that the two types of perovskite solar cells were otherwise prepared in an identical way. Presumably, post-annealing at 105 °C was not sufficient to remove the majority of CH₃NH₂ from the perovskite film and, therefore, could be a likely cause responsible for the lower efficiency.³³ Regarding the origin of CH₃NH₂, it is possible that after perovskite films are prepared, a certain amount of unreacted CH₃NH₃I still remains in perovskite films even after annealing. CH₃NH₃I can further decompose to CH₃NH₂, leading to the presence of CH₃NH₂ in perovskite films. In addition, when CH₃NH₃PbI₃ perovskite films degrade, CH₃NH₂ can also be generated as a product, which can be another origin of CH₃NH₂ in perovskite films.

In conclusion, we found CH₃NH₂ neutral species on the surface and in the interior of CH₃NH₃PbI₃ films prepared using a typical solution method. Based on the experimental results and density functional theory calculations, we conclude that the CH₃NH₂ neutral species is mainly located on the surface or at grain boundaries in the interior of CH₃NH₃PbI₃ films, which is expected to affect device performance and degradation. The observation of different chemical states on the surface of CH₃NH₃PbI₃ films such as C-metal and C*=C bonding is also discussed.

This work was supported by funding from the Energy Materials and Surface Sciences Unit of the Okinawa Institute of Science and Technology Graduate University and JSPS KAKENHI Grant No. 15K17925. Y.M.L and H.-K.L are grateful for financial support from the Basic Science Research Program of the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2015R1C1A2A01054543 and NRF-2011-0006744). The experiment at the PLSII was supported, in part, by MEST and POSTECH. Calculations were supported by the Nano Material Technology Development Program (2012M3A7B4049888) through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning (MSIP), and the Priority Research Center Program (2010-0020207) through NRF funded by the Ministry of Education (MOE). The authors thank Steven D. Aird, the Technical Editor at Okinawa Institute of Science and Technology Graduate University (OIST) for valuable suggestions in revising the manuscript.