In Situ IR Spectroscopy Studies of Atomic Layer-Deposited SnO2 on Formamidinium-Based Lead Halide Perovskite

Perovskite photovoltaics has achieved conversion efficiencies of 26.0% by optimizing the optoelectronic properties of the absorber and its interfaces with charge transport layers (CTLs). However, commonly adopted organic CTLs can lead to parasitic absorption and device instability. Therefore, metal oxides like atomic layer-deposited (ALD) SnO2 in combination with fullerene-based electron transport layers have been introduced to enhance mechanical and thermal stability. Instead, when ALD SnO2 is directly processed on the absorber, i.e., without the fullerene layer, chemical modifications of the inorganic fraction of the perovskite occur, compromising the device performance. This study focuses on the organic fraction, particularly the formamidinium cation (FA+), in a CsFAPb(I,Br)3 perovskite. By employing in situ infrared spectroscopy, we investigate the impact of ALD processing on the perovskite, such as vacuum level, temperature, and exposure to half and full ALD cycles using tetrakis(dimethylamido)-Sn(IV) (TDMA-Sn) and H2O. We observe that exposing the absorber to vacuum conditions or water half-cycles has a negligible effect on the chemistry of the perovskite. However, prolonged exposure at 100 °C for 90 min results in a loss of 0.7% of the total formamidinium-related vibrational features compared to the pristine perovskite. Supported by density functional theory calculations, we speculate that FA+ deprotonates and that formamidine desorbs from the perovskite surface. Furthermore, the interaction between TDMA-Sn and FA+ induces more decomposition of the perovskite surface compared to vacuum, temperature, or H2O exposure. During the exposure to 10 ALD half-cycles of TDMA-Sn, 4% of the total FA+-related infrared features are lost compared to the pristine perovskite. Additionally, IR spectroscopy suggests the formation and trapping of sym-triazine, i.e., a decomposition product of FA+. These studies enable to decouple the effects occurring during direct ALD processing on the perovskite and highlight the crucial role of the Sn precursor in affecting the perovskite surface chemistry and compromising the device performance.

. Schematic illustration of the home-built ALD reactor used for the in-situ transmission IR spectroscopy measurements. The different components of the system are highlighted and include IR source, MCT detector, ICP plasma source, and turbo-pump. Additional details can be found in a previous publication. 1 During our experimental investigation, the perovskite absorber is processed on top of a c-Si substrate. The ALD cycle used during the experimental investigation is shown schematically. The FTIR differential spectra are reported by subtracting the spectrum of the perovskite from measurement FTIR (A) or FTIR (B). 2 The first spectrum measured during each experiment is taken as reference.

IR vibrational modes aided by DFT calculations
Density functional theory (DFT) calculations were performed to confirm the attribution of the infrared modes of FA + , which are highlighted in the spectrum shown in Figure 1b. The optimized slab model of the perovskite structure is shown in Figure  SI 2a. The calculated vibrational modes with their corresponding wavenumbers are summarized in Table S1. For reference, a free-standing formamidinium cation was also calculated and taken as the simplest model for the assignment of the vibrational modes of FA + in the perovskite structure.
Based on the results, it was determined that there are four different N-H stretching modes belonging to the four H bonded to the two N in each FA + . Each of the two H atoms bonded to the same N can form one symmetric and one asymmetric combination, splitting the vibrations of each side of the FA + cation into asymmetric, ʋ as (N-H), and symmetric, ʋ s (N-H), vibrational modes, as illustrated in Figure 1d. For the free-standing FA + cation, considered thus without any influence from the perovskite surroundings, the splitting within each stretching mode is small, with 5 cm -1 and 17 cm -1 difference for ʋ as (N-H) and ʋ s (N-H), respectively, as shown in Table S1. This indicates that each of these two modes is nearly doubly degenerate, owing to the symmetric configuration of the FA + cation. In the mixed compound (CS,FA)Pb(I,Br) 3 , this symmetry is broken due to the different chemical surroundings of the two NH 2 groups present at the two sides of the FA + cation, as illustrated in Figure 1c. As a result, the splitting within the asymmetric and symmetric modes is increased to 38 cm -1 and 43 cm -1 , respectively. As a result, this leads to four different vibrational modes, ʋ as (N-H), ʋ', ʋ'', and ʋ s (N-H) are detected, as shown in Figure 1d.  These results are comparable to the assignment given by Hills-Kimball et al., where they suggested that the N-H stretching vibrations are affected by different hydrogen-bond strengths between the H in the FA + and halides present in the inorganic perovskite cage. [3][4][5] The above analysis well explains the secondary features, ' and υ υ '', measured by the experiment, which has a 45 cm -1 and 62 cm -1 difference with respect to the ʋ as (N-H) and ʋ s (N-H), respectively, as shown in Table I of the main text. In addition to the N-H stretching frequencies, also the calculated values for the mixed halide perovskite (Cs,FA)Pb(I,Br) 3 are generally in good accordance with those experimentally measured.  3 . The FA + cation, whose vibrational frequencies are studied, is highlighted by an orange dashed circle.

B. Deprotonation of FA + and supporting results from DFT simulations
As reported in Table S2, the release of HX species is independent of the temperature at which the perovskite, or its simpler constituent FAI, is exposed and begins from the deprotonation of formamidinium into formamidine. As can be seen, the reported temperature onsets for the release of decomposition byproducts, vary widely depending on the study and oscillate between 50 and 360 °C. 7-10 This process is caused by the intrinsic high reactivity of the delocalized double bond. 11 The resulting formamidine molecules, having lost one of their hydrogen bonds, are more likely to be abstracted from the perovskite surface which becomes organic-deficient. 12 To precisely compare the vibration frequencies of the deprotonated system to those of the pristine system, the optimized structure of the deprotonated system, with one Hi or HBr removed, is calculated and shown in Figure S3. The comparison between the vibrational modes of the deprotonated FA and the pristine FA + cation is given in Table S3. Noticeably, three vibrational modes are absent in the deprotonated FA system at 3378, 1718 and 1127 cm -1 . Assuming that a fraction of the formamidinium moiety is converted to formamidine due to the deprotonation, it would be expected to have negative features arise in correspondence with each of the vibrational modes of FA + and positive ones corresponding to FA. Additionally, as the DFT calculations show, some of the vibrational modes in the FA system appear to be shifted (and even overlap) with those of the pristine model.
Comparing the negative features of the experimentally measured pristine perovskite spectrum with that of the heat-treated one matches this behavior. Specifically, looking at the secondary features of ʋ as (N-H) and ʋ s (N-H) at 3355 and 3329 cm -1 , respectively, we do not detect them as negative features. This is due to, as shown by the DFT calculations, the fact that the secondary ʋ s (N-H) mode of the deprotonated system vibrates with a frequency in between the two pristine ones, thus overlapping with their negative features and resulting in a smaller contribution to these losses in the IR spectrum, similar to what we detected experimentally. These results support the hypothesis of formamidinium undergoing deprotonation, coupled with the release of HI, during ALD processing. Additionally, as shown in Figure S4, the deprotonation of the FA + cation results in the pinning of the C=N bond on the side that loses the hydrogen atom. In the IR spectrum, such change would be shown as a negative peak corresponding to the loss of the delocalized resonating C=N bond, as also determined through the DFT calculations reported above. It also would give rise to a positive peak caused by shifts in the vibrational frequencies of the remaining C=N bonds, affected by the atomic ordering and subsequent changes in the surface composition. For the remainder of the vibrational modes, their intensity is much lower than those discussed above, and their evaluation is less trivial.