Synthetic control over orientational degeneracy of spacer cations enhances solar cell efficiency in two-dimensional perovskites

Two-dimensional perovskites have emerged as more intrinsically stable materials for solar cells. Chemical tuning of spacer organic cations has attracted great interest due to their additional functionalities. However, how the chemical nature of the organic cations affects the properties of two-dimensional perovskites and devices is rarely reported. Here we demonstrate that the selection of spacer cations (i.e., selective fluorination of phenethylammonium) affects the film properties of two-dimensional perovskites, leading to different device performance of two-dimensional perovskite solar cells (average n = 4). Structural analysis reveals that different packing arrangements and orientational disorder of the spacer cations result in orientational degeneracy and different formation energies, largely explaining the difference in film properties. This work provides key missing information on how spacer cations exert influence on desirable electronic properties and device performance of two-dimensional perovskites via the weak and cooperative interactions of these cations in the crystal lattice.


Supplementary Tables
Supplementary  (2) 34.6431 (17) 8.7354 (2) b (Å) 8.6332 (12) 6.1150 (3) 8.7379 (2) c (Å) 8.8000 (12) 12.2922 (6)  We conducted EQE measurement on each 2D perovskite solar cells without light bias. The integrated current density from EQE is lower than the current density achieved from J-V cruves under one sun conduction. This can be caused by the existence of multiple phases in our 2D perovskite films. For a standard EQE measurement without light bias, the large band gap components are not excited, which might cause a large resistance in the cell and lower the current measured. Such phenomenon has also been seen in tandem solar cells 1-4 and Cu2ZnSnS4 solar cells. 5

Supplementary Note 2
The dynamics of n = 1, 2 and 3D phases for each 2D OIHP sample show a clear decay of n = 1 for all four 2D OIHP films (Supplementary Figure 7a), however, there is a clear increase of the absorption intensity at early stage (< 10 ps) in the decay curve of n = 2 for mF1PEA and pF1PEA 2D OIHP films (Supplementary Figure 7b). For PEA based 2D OIHP film, though there is no clear increase of the absorption intensity for n = 2, the decay of the absorption intensity (n = 2) for PEA is much slower than that of oF1PEA (Supplementary Figure 7b). This suggests that when n = 1 phase was excited, the excitons can transfer their energy to n = 2 phase in the case of PEA, mF1PEA and pF1PEA based 2D OIHP films. In contrast, oF1PEA sample shows the quickest decay of the absorption intensity (n = 2), indicating that such energy transfer (from n = 1 to n = 2) is not significant. Furthermore, for oF1PEA based 2D OIHP film, the signal for the 3D phase in the full spectrum (Fig. 2e) has a significant increase of intensity from 1 ps to 100 ps: we suspect that there is an efficient energy transfer from n=1 phase to the 3D phase due to their proximity, consistent with the PL spectrum for oF1PEA based 2D perovskite (Fig. 2b).

Supplementary Note 3
ToF-SIMS analyses were conducted using a ToF SIMS V (ION TOF, Inc. Chestnut Ridge, NY) instrument with a Cs + sputtering gun.
We tried elemental profiling method (ToF-SIMS), to further confirm the proposed phase distribution by analyzing F and organic fragments from spacer cations (Supplementary Figure  8). However, we could only observe a clear trend for mF1PEA 2D OIHP (Supplementary Figure  8c). For the rest, the substrate signal (Si) appeared at the very early stage, indicating a high etching rate (Supplementary Figure 8 a, b, d). Therefore, we could not draw a clear conclusion from ToF-SIMS result.

Supplementary Note 4
For UPS measurement, films of the OIHPs were deposited on ITO substrate by spin coating. The precursor solution contains 0.25 M PbI2 and 0.5 M ammonium iodide salt (i.e., PEAI or F1PEAI) and 9:1 DMF/DMSO is used as solvent. The precursor solutions were spun at 5000 rpm for 20 s and followed by annealing at 80 °C for 5 minutes. 3D perovskite (MAPbI3) was deposited following a previous reported procedure. 6 The kinetic energy was probed by Kratos Axis Ultra DLD Ultraviolet Photoelectron Spectrometer.
We conducted ultraviolet photoelectron spectrometer (UPS) on the film with pure n = 1 phases, i.e., PEA2PbI4, oF1PEA2PbI4, mF1PEA2PbI4, and pF1PEA2PbI4, and 3D phase (MAPbI3) to study the energy levels of these perovskite phases. Because of the gradient structure transition from n = 1 to 3D, we believe that other 2D phase such as n = 2, 3, 4, etc. should have the energy levels lying between n = 1 and 3D. As shown in Supplementary Figure S9 and Supplementary Table 4, the valence band of these n = 1 2D OIHP ranges from -5.8 eV to -6.4 eV. As comparison, we also conducted UPS on 3D perovskite samples and the valence band position is 5.8 V. Therefore, it is highly possible that a type I band alignment formed in our 2D OIHP films. However, it is worth noting that our 3D perovskite has a different valence band position compared to the work reported by others (~-5.4 eV) and UPS is a highly surface sensitive technique and the surface of perovskite films could suffer from contamination and degradation. Therefore, it's hard to draw the conclusion the real band alignment in our 2D OIHP films with multiple phases.
The T = 0 K formation enthalpies were estimated using density functional theory (DFT) calculated total energies as implemented in VASP 7, 8 by way of ΔHf 0K ≈ EA2PbI4 -(2EAI + EPbI2). DFT calculations used the Projector Augmented Wave (PAW) method 9, 10 to describe the effects of core electrons and Perdew−Burke− Ernzerhof (PBE) 11 implementation of the Generalized Approximation (GGA) for the exchange-correlation functional. The energy cutoff was set to 530 eV for the plane-wave basis of the valence electrons. The optB86b-vdW functional 12 for dispersion corrections was applied. The total energy tolerance for the electronic energy minimization was 10 -4 eV; for structure optimization, forces were minimized such that all atoms experience forces < 0.05 eV Å -1 after relaxation of the ionic coordinates and unit cell shape and volume. The product structures were idealized if the crystal structures were solved with split-site disorder, as discussed in the supporting information. The structures of fluorinated phenylethyl ammonium-iodide salts used for formation enthalpy calculations were obtained by relaxation of the reported structure of phenylethylammonium bromide, 13 after substitution of the bromine by iodine and relevant hydrogen with fluorine.
The formation enthalpies calculated for all the four 2D OIHPs are positive. Considering the relative small entropies change for solid state reaction, this suggests that 2D OIHP (n = 1) materials are not thermodynamically stable. However, this does not account for systematic errors associated with the exchange-correlation functional and finite temperature. For instance, the enthalpies might be temperature-dependent and entropic considerations may not be negligible. Last, highly idealized starting structures may also introduce systematic errors. Nevertheless, the trend of formation enthalpies is consistent with the quality of single crystal (pF1PEA2PbI4 ≈ mF1PEA2PbI4 > PEA2PbI4 >> oF1PEA2PbI4, Figure S4) and also consistent with the structure transition energy reported by Li et al. 14 Therefore, we report only the relative differences in formation energies, to approximate a fortuitous cancellation of systematic errors (Fig. 6).

Supplementary Discussion 1
We performed a thin film X-ray diffraction (XRD) experiment shown in Supplementary  Figure 11 and 12. In the XRD profiles, we observed peaks of strong intensity around 14.2° and 28.5° in all samples, consistent with previous work on similar 2D OIHP films. 15,16 Since our films contain multiple phases (both 2D and 3D phases) and these two peaks (14.2° and 28.5°) are present in XRD data for both 3D (Supplementary Figure 12) and 2D OIHPs, they can either be labelled as (111) and (202) based on 2D OIHP structure 15,17 or (110) and (220) based on 3D OIHP structure. 18 Here we labelled them as (110) and (220) for our discussion. The absence of XRD peaks below 10° in PEA or pF1PEA 2D OIHP films indicates that these two films do not have many 2D OIHP crystalline phases with the inorganic layers parallel to the substrate. 19,20 In contrast, a family of peaks from 2D phase(s) are strongly visible in the oF1PEA 2D OIHP film. Based on our single crystal test, these peaks belong to the (002) family of peaks from n = 1 phase. For the mF1PEA 2D OIHP film, we observe a stronger (002) family of peaks from 2D phases when compared to PEA and pF1PEA 2D OIHP films, but they are still weaker than those in oF1PEA 2D OIHP film. This indicates that there are 2D phases with the inorganic layers parallel to the substrate in mF1PEA 2D OIHP film, more than those in PEA and pF1PEA 2D OIHP film, but less than those in oF1PEA 2D OIHP film. We will discuss the crystal orientation further with the results from grazing incidence wide angle X-ray scattering (GIWAXS, vide infra) of all these 2D OIHP films. In addition to the different peak intensity for each film, the full-width at halfmaximum (FWHM) for both the (110) and (220) peaks of all four samples are plotted in Supplementary Figure 13. In sharp contrast with other films, the FWHMs of oF1PEA 2D OIHP are larger than those of the rest. This suggests that the crystallinity of the oF1PEA 2D OIHP film is much worse than the other three films. This further explains the low photovoltaic performance of oF1PEA 2D OIHP.

Supplementary Discussion 2
SCXRD data for pF1PEA, mF1PEA, and oF1PEA were collected at room temperature using a Bruker D8 Quest ECO diffractometer equipped with a microfocus Mo K radiation source and Photon 50 CMOS half-plate detector. Single crystals were mounted onto a glass fiber with 5minute epoxy. Bruker SAINT was used for integration and scaling of collected data and SADABS (multi-scan) was used for absorption correction. Starting models for the three compounds were generated using the intrinsic phasing method in SHELXT. 21 SHELXL2014 was used for leastsquares refinement. 22 Structures, including disorder, based on electron densities are provided, as well as idealized supercells without disorder based on chemically-reasonable bond distances and molecular configurations.
pF1PEA2PbI4 has a fully ordered structure with no split atomic positions, indicating long range order for both the inorganic [PbI6] and organic pF1PEA units, and that registry between the inorganic layers is retained. The aromatic moieties within the organic interlayer gallery face the same direction resulting in a co-aligned configuration ( Fig. 5d and 5h of the main text). This leads to short intermolecular contacts between molecules within the top and bottom layers of the interlayer gallery of 2.731 Å (H5· · · F1) and 2.816 Å (F1· · · H7). Additionally, within the top and bottom layer, neighboring pF1PEA molecules have relatively short H5· · · H4 contacts of 2.872 Å.
The mF1PEA molecules within mF1PEA2PbI4, on the other hand, pack in a different manner than pF1PEA in pF1PEA2PbI4. The crystal structure results in significant disorder that is crystallographically accommodated using a split-occupancy of atoms to capture the total electron density. However, using the electron density of the heavy atoms, the relationship to other compounds, and realistic constraints on bond distances, we have assembled an idealized structural model without disorder that describes how the mF1PEA molecules are packed in the lattice. A significant difference between mF1PEA and pF1PEA is the presence of split axial iodine atomic positions and consequentially a splitting of the ammonium groups on the mF1PEA molecules. This disorder is similar to that observed in (3-FPEA)2SnI4 where the splitting of sites was attributed to loss of registry between inorganic sheets within the material. This disorder results in a superimposed "average" structure of the two possible octahedral tilting configurations that can be adopted in ~50% probability (Supplementary Figure 16a). If one assumes that the lead is octahedrally coordinated in a relatively regular octahedron (as found in nearly all related compounds), then one finds the same tilting pattern found in pF1PEA2PbI4.
In addition to the split iodine positions, there are split positions of the mF1PEA molecules in the crystallographically-rigorous model. However, in the disordered structural model, for each of the iodine positions at a given split-site, there is a specific R-CH2-NH3 orientation that corresponds with realistic interatomic distances (Supplementary Figure 16b). This is due to the ammonium group always pointing towards the "puckered-out" orientation of the Pb-I-Pb bonds (cf., Supplementary Figure 17); the intermolecular contacts for an ammonium group facing towards the "puckered-in" direction are too short to be realistic. As neighboring octahedra must tilt-in and S39 then out to retain its structure (Supplementary Figure 17), the orientations of the R-CH2-NH3 cations must also alternate between the two possible configurations to ensure the bond distances between the inorganic and organic units are realistic. Therefore, we have generated an idealized supercell model in space group P1 that has the chemically-reasonable orientations included for both the [PbI6] inorganic framework and the mF1PEA molecules. Within a single layer of the organic interlayer gallery, neighboring aromatic moieties are rotated relative to one another to generate a herringbone configuration (Fig. 6c). Compared to pF1PEA, this packing arrangement leads to even shorter intermolecular contacts within a layer (H2B· · · F1 of 2.512 Å), but leads to longer contacts between molecules in the top and bottom of the interlayer gallery of 3.005 Å (H6· · · F1).
Lastly, oF1PEA2PbI4 represents yet another type of packing and disorder that is different from pF1PEA2PbI4 and mF1PEA2PbI4. The crystal structure, as solved, is again an "average" structure with split sites such as those seen in mF1PEA2PbI4; however, there is no site splitting within the inorganic framework (i.e., each atom is fully occupied). Instead, the disorder arises from the different orientations the aromatic moieties can adopt, which are then coupled with the ammonium groups. There are two possible orientations the oF1PEA cations can adopt within a single layer, as illustrated in red and blue in Supplementary Figure 17. Interatomic distances dictate that the aromatic moieties must orient in a colinear fashion (Fig. 5b, similar to pF1 PEA2PbI4), but the neighboring layer along the stacking direction must then be oriented ~90° relative when projecting along the c-axis (Fig. 6b and 6f). This is further illustrated in Supplementary Figure 17 where either the red or blue molecules exist in a single layer, but both colors cannot exist at the same time. Because the aromatic moieties orient in this motif, and the fact that the R-CH2-CH2-NH3 groups prefer the "J-shaped" gauche conformation relative to the inorganic sub-structural unit, the total number of possible orientations for the ammonium group is three, rather than four as seen in mF1PEA2PbI4. This is because the aromatic groups orient colinearly, and thus are always facing the same direction. This hinders the organic cations near the "puckered in" I-Pb-I geometry and allows us to assign the specific colors to the differently oriented molecules (Supplementary Figure 17). The occupancies of each molecular orientation were allowed to freely refine, which revealed a nearly perfect 50% occupancy for either of the two possible configurations (red or blue in Supplementary Figure 17). This indicates that like in mF1PEA2PbI4, each possible configuration occurs in the same total amount; however, the configurations lack ordering along the c-axis (Supplementary Figure 18). An idealized superstructure in P1 was also generated for oF1PEA2PbI4, which was used for DFT calculations.