Lead-free Pseudo-three-dimensional Organic-inorganic Iodobismuthates for Photovoltaic Applications

Two organic-inorganic iodobismuthates, C 5 H 6 NBiI 4 ( [py][BiI 4 ] ) and C 6 H 8 NBiI 4 ( [mepy][BiI 4 ] ), have been prepared with their structures revealed by single-crystal X-ray diffraction. One-dimensional BiI 4- anionic chains built by edge-sharing BiI 6 octahedra were found in both materials; short I…I, I…C contacts and hydrogen bonding give rise to three-dimensional intermolecular interactions. Both compounds are semiconductors, with band gaps of around 2.0 eV, and the contribution from the organic moieties to the conduction band minimum has been derived by density functional theory. Solid-state optical and electrochemical studies performed on powders and thin films were carried out, and their stabilities under ambient environment have been demonstrated. Used as the absorber layer in printable mesoscopic solar cells without hole transport material has led to efficiencies up to 0.9%, showing a promising new approach towards the development of lead-free third-generation photovoltaic materials.


Introduction
During the last few years, the rapid emergence of organicinorganic metal halide perovskites solar cells has led to a new era for research into solar energy utilization. [1][2][3][4][5][6][7][8][9] Lead-based perovskite solar cells have been extensively studied, and are central to the extraordinarily rapid progress of powerconversion efficiency (PCE) from 3.8% in 2009 1 to more than 20% today. [8][9] Crucially, this is achieved via low-cost perovskite material preparation and facile fabrication, making them promising in the future photovoltaic market. However, one fact that may hinder the full application of this kind of metal-halide solar cell is the toxicity of lead. Concerns over the leakage of lead and the improper disposal of cells may restrict application in certain areas. To address the toxicity issue, developments on non-toxic metal-halide photovoltaics are now being explored. Tin, also a group 14 element with similar atomic radius, has recently been studied to substitute lead within novel photovoltaic absorber materials. Perovskites including CH3NH3SnX3 (X = Cl, Br, I) and Cs3SnX3 (X = Br, I) in solar cell devices have achieved more than 6% PCE. [10][11][12] However, Tinbased perovskites exhibit extreme sensitivity to humidity and oxygen, which is a severe hindrance for their further application. Alternatively, bismuth-halide compounds have been proven to exhibit a photovoltaic effect, [13][14][15][16][17][18][19] with PCE over 1% achieved by Cs3Bi2I9 17 and by AgBi2I7 19 . In all the solar cell applications mentioned above, a FTO/TiO2/Perovskite/holetransport material (HTM)/gold classical structure was adopted, which suffers a lot from the high cost of HTM and gold counter electrode.
In general, iodobismuthates have attracted attention due to their semiconducting properties, non-toxicity, air stability and solution processibility. Great structural diversity of the anionic iodobismuthate motif can be found, ranging from the simplest [BiI6] 3to the poly-nuclear [Bi8I30] 6-. 20 Most iodobismuthate structures are built from zero-dimensional Bi/I anions. Among them [Bi2I9] 3-complexes have been extensively investigated, including their crystal structures, optical properties and quantum physical properties. [20][21][22][23] Different cations including CH3NH3 + , 14,15,24 , NH4 + , 23 Cs + , 18,22,25 Rb + , 26,27 and K + 27 have been studied, which may be used to aid the rational design of photovoltaics. So far, much focus has been put on the zerodimensional [Bi2I9] 3structural unit in solar cells. Low dimensional materials are likely to give rise to limited capability for charge carrier transport, leading to relatively low PCE. We propose that alongside the well-studied zero-dimensional [Bi2I9] 3complexes, the lesser-studied iodobismuthates based on extended inorganic networks are also important candidates to explore for lead-free solar absorbers, such as 1D infinite chain structures, including [Bi2I7 2-]n, [Bi3I11 2-]n, [BiI4 -]n, and [BiI5 2-]n. [28][29][30][31][32] In this article, two organic-inorganic iodobismuthates,  30 where the inorganic cations have almost no electronic interaction with the 1D BiI4framework, we have demonstrated that the organic entities are playing an active role in intermolecular interactions and the frontier orbital, which makes them pseudo-3D materials regarding the dimension of charge transfer. Moreover, we have made cost-effective solar cell devices adopting HTM-free mesoscopic structure with carbon counter electrode, reaching the best efficiency of 0.9%. In this initial study, we show this approach as a new strategy towards the development of higherdimensionality iodobismuthate frameworks, which presents a comparable order of magnitude of efficiency with the other best iodobismuthate examples.  (6)°, and the angles of I (bridging)-Bi-I (bridging) in each BiI6 octahedra are generally smaller than those of I (terminal)-Bi-I (terminal). The deviation of the geometric parameters indicates that the BiI6 octahedron is distorted, which is a common characteristic of iodobismuthates. 24,33 In [mepy][BiI4], [BiI4]anionic chains propagate along the a-axis with methylpyridinium cations arranged along the edge-sharing BiI6 chain direction. Shorter bond lengths for Bi-I bond (2.9098(8) Å -2.9195(8) Å) are again found for the terminal iodine atoms than that with bridging iodine atoms (3.1027(9) Å -3.3029(7) Å) in the slightly distorted BiI6 octahedra. The bond angles for I-Bi-I range from 87.620(19)° to 94.30(2)° with smaller I (bridging)-Bi-I (bridging) than that with terminal I. Bi-I (bridging)-Bi angles range from 92.21(2)° to 93.75(2)°. The distortion of BiI6 octahedra in the two title iodobismuthate compounds probably originates from the repulsion of bismuth atoms between the adjacent [BiI6] 3building blocks. 24 Interesting features for both compounds regarding intermolecular interactions can be observed when viewing along different axis (Figure 3, S1, S2), and the corresponding short contact distances are listed in table 1. In the crystal structure of [py][BiI4], short I…I contacts exist between two adjacent [BiI4]chains in the ab plane, where the shortest distance (3.823 Å) is less than the sum of their Van der Waal's radii (3.98 Å). Hydrogen bonding with iodine and Van der Waal's interactions between iodine-iodine and iodine-carbon can also be found when viewed along the a-or c-axis (Figure 3, S1). However, less intermolecular interactions can be observed in

Crystal structures and intermolecular interactions
There are short I…C contacts between the terminal iodine atoms and carbon atoms of the methylpyridinium rings, and hydrogen bonding involving the terminal iodine atoms run perpendicularly to the chain plane. Compared to [py][BiI4], no short I…I contacts were found. This can be attributed to the methyl group on the organic cation, which increases the volume of organic component, separating the distance between the anionic chains.     Figure 3. The sum of the Van der Waal radii are 3.96 Å for I…I, 3.68 Å for C…I, and 3.18 Å for N(C)-H…I.

Electronic Band structure calculations
Band structure and projected density of states (DOS) calculations were obtained using CASTEP 16.11 34 to investigate the variation (dispersion) in band energies with respect to high symmetry k-points and directions that correspond to strong intermolecular interactions. For both [py][BiI4] and [mepy][BiI4], DFT methods were employed with the PBE functional and the Tkatchenko-Scheffler (TS) dispersion correction scheme, using the coordinates for each atom obtained from optimized single crystal geometry. Projected DOS for the p-orbitals of bismuth, iodine, carbon and nitrogen were calculated using the OptaDOS 35 package, with the Fermi level set to 0 eV. Crystal structure packing diagrams, together with k-point vectors used for both compounds are shown in Figure S4, and the band structures together with projected DOS diagrams can be found in Figure 4. , relatively large dispersions in the band energies can be seen. Specifically, high dispersions exist along vectors H-Y and G-Z, which correspond to the direction along the infinite [BiI4]chains in real space. More interestingly, there is some small dispersion along the vector path Z-D, which corresponds to the interaction of iodine atoms between the anionic inorganic chains. Minor, but noticeable, dispersion also occurs along Y-G that arises from the short I…C and I…H contacts. For [mepy][BiI4], the valence band is generally flat. Small but noticeable dispersions along G-H and E-C in the valence band coincide with the one-dimensional chain direction. Thus, one aspect of valence band dispersion is contributed by the σbonding interaction of iodine 5p states with the bismuth 6p states, which leads to the infinite Bi-I covalent bonding that builds the chain structure. The other valence band interaction stemming from the iodine 5p states gives rise to the short I…I, I…C and I…H contacts. Therefore, dispersion in the valence band of [py][BiI4] mainly originates from the inorganic entities, and the intermolecular interaction of inter-chain iodine atoms is crucial for increasing the dispersion of the band structure.
However, Louvain et al 29 claimed that due to the non-spherical distribution of the bismuth 6s orbital, it is not possible to have ideal antibonding interactions for bismuth with the surrounding six iodine atoms. This is interpreted as the imperfect geometry of the BiI6 octahedron hindering the Bi-I antibonding interactions, which leads to the relatively small band dispersion compared to lead perovskite.
As presented in Figure 4, the conduction band minimum (CBM) for both [py][BiI4] and [mepy][BiI4] is almost entirely derived from atomic orbital contributions of nitrogen and carbon from the organic moieties. Although the overall band dispersion is not very significant, it is still worth noting that the cationic organic entities are playing an essential role in the CBM to increase the electronic dispersion of the band structure. Hence, participation of organic entities at band edges were found for both compounds, which may be a strategy that can be further exploited to enable higher charge transport ability in more than one dimension. Therefore, the relatively small amount of dispersion observed in the frontier orbitals of [mepy][BiI4] is derived from the lack of intermolecular interaction, which correlates with the crystal structure. However, by taking advantage of extended 1D frameworks together with the high dimensional intermolecular interactions, [py][BiI4] can be regarded as a pseudo-three-dimensional material due to its higher dimensionality in the X-ray structure and resulting band structure.

We have prepared thin films of [py][BiI4] and [mepy][BiI4]
by a one-step spin-coating method (see the Experimental section). X-ray diffraction was carried out to compare with the theoretical XRD pattern of the single-crystal phase to explore the crystallinity and phase of each material in solid-state thin films ( Figure 5). Both materials as thin films are consistent with the calculated powder XRD pattern, indicating they are predominately in the same crystallographic phase. There was no indication of the potential impurities BiI3, I2 or elemental Bi ( Figure S7). However, strong preferred orientation on the (1 2 0) plane was observed in [py][BiI4], which corresponds to the plane where the sheet of iodobismuthate chains propagates in a parallel manner to the substrate ( Figure S3). Preferred orientations on planes (0 1 1) and (1 3 2), along with other Bragg peaks in accordance with the simulated XRD pattern were found for [mepy][BiI4]. In addition, the stability of thin films under ambient conditions was also proven by observing XRD patterns on thin-films, which indicated that no major structural changes occurred after one-week exposure to air in the dark. Some slight improvement in crystallinity was observed over a week, possibly due to complete loss of some residual solvent in the film.
Diffuse reflectance spectra were measured for both compounds in powder form, as well as for spin-coated samples on glass slides, in order to estimate their optical band gaps, as converted into Tauc Figure 6), which compare favourably with the values obtained by simulation (DFT is well known to underestimate band gap energies) 37

. The observed variation in band gap for [py][BiI4] and [mepy][BiI4]
is mainly caused by the differences in the organic cations present, that lead to a lack of extended inter-chain interactions for the latter compound. Slightly higher band gap values were observed for the compounds when presented as thin films, (Figure S8), which is likely due to less-reliable measurements caused by the limited thickness of the thin film. Compared to the reported 1D (BiI4)compounds based on metal cations 30 which have band gap values of 1.70 -1.76 eV, it is apparent that both 1D (BiI4)anionic chains and the organic entities are important in determining the overall optical properties of the material.  To determine the stoichiometry of elements and probe the electronic properties of the film surface, X-ray photoelectron spectroscopy (XPS) was carried out for each compound spincoated on fluorine-doped tin oxide (FTO) glass. The survey spectrum of each film shows that the elements detected include Bi, I, N, O, Sn and C ( Figure S9). Both Sn and O originate from the FTO glass substrate. Figure 7 shows the intensity of Bi 4f and I 3d core-level high-resolution photoemission for both compounds, where the binding energy is referenced to the Fermi level. The characteristic signals for Bi and I are Bi 4f7/2(158 eV), 4f5/2 (163 eV), I 3d5/2 (620 eV) and 3d3/2 (631 eV). However, some metallic bismuth was found as signals of Bi 4f7/2 (156 eV) and Bi 4f5/2 (161 eV), indicating some reduction of iodobismuthates. We speculate that this may occur during the annealing step and in future work more careful control of this step may be required. We note that the absence of diffraction peaks for elemental Bi (vide supra) suggests its presence in amorphous or very small particle form. The existence of metallic bismuth was also observed in other literature work. 14

To study the materials electrochemically, [py][BiI4] and [mepy][BiI4]
were spin-coated on FTO conducting glass slides as working electrodes. Both of the compounds exhibited irreversible first oxidation and first reduction processes ( Figure  S10), which is likely related to some dissolution of the film after the redox step. The method from Crespilho et al 38 was used to estimate the ionization energy (IE) and the electron affinity (EA) from the experimentally-determined redox potentials . The electrochemically determined IE and EA energy levels of each compound are summarized in table 3, and the corresponding energy level diagram is shown in Figure 8. It's worth noting that the electronically determined Eg values is consistent with the computational and optical results, which further proves the validity of the energy level measurements.

Device Characterization
The title compounds were tried as light harvesters in solar cells. The device used in this study is based on the FTO/(TiO2/ZrO2/Carbon)/perovskite structure, with the crosssection SEM image shown in Figure 9. This type of fully-printable and low-cost solar cell was first developed by H.W. Han's group. 39,40,41 Mesoscopic TiO2, ZrO2 and carbon were printed on FTO glass layer by layer, followed by infiltrating the perovskite solution into the scaffold. No HTM was employed as was presently done in all high efficiency perovskite solar cells, so the compound is transporting both electrons and holes here. It was found that some [py][BiI4] start to crystallize on the surface of the carbon film before they reach TiO2, thus resulting in a poor reproducibility of the devices ( Figure S12). The problem of filling can be minimized by choosing the proper solvent and optimizing the fabrication technique, however, it is beyond the scope of this article. Figure 10 shows the performance of a [py][BiI4] solar cell with relatively good filling, giving a the PCE obtained of 0.90% (with Voc=0.62 V, Jsc=2.71 mA cm -2 , FF=0.54, reverse to forward scan). The hysteresis effect was evaluated by performing a reverse scan (from VOC to JSC) and a forward scan (from JSC to VOC), and substantial hysteresis has been found. In contrast, no photovoltaic effect was observed for [mepy][BiI4], which can be attributed to many reasons, e.g., poor crystallization in the mesoscopic structure, low carrier density and relatively large bandgap. For future research, it is apparent that more attention should be paid to the cationic part of the iodobismuthate complex, especially those with more electronwithdrawing and electronically delocalized properties.

Conclusion
Synthesis, structural characterization, electronic characterization and evaluation in solar cells have been carried out for two novel organic-inorganic iodobismuthate compounds, to assess their potential as third-generation leadfree photovoltaics absorbers. The crystallographic structures for both [py][BiI4] and [mepy][BiI4] feature one-dimensional [BiI4]infinite anionic chains, while the organic entities, protonated aromatic heterocycles, play an active role in promoting intermolecular interactions and have energies close to the conduction band minimum, which supports our strategy towards enhanced psuedo-three-dimensional charge carrier transport ability. The energy levels of each compound were experimentally and computationally determined, and found to have band gaps of about 2.0 eV. Both compounds are solution-processable, which provides potential low-cost manufacturing via spin-coating methods for thin film preparation. Although some difficulty was encountered in achieving highlyreproducible PV results, we demonstrated that the efficiency of the perovskite solar cell without HTM can approach 0.9%, which is comparable with the other best Bi-iodide based cells reported. 17,19 This was achieved within the approach of a fullyprintable cell using a cell structure not explored before for iodobismuthate materials. Considering the fact that no HTM was used in the structure, this performance is very competitive. Overall, this study provides a new perspective in designing leadfree photovoltaics, by seeking to take advantage of both the anionic chain framework and the organic cation for enhanced electronic properties. Our further studies will be aimed at optimising the design of the organic cation to further promote strong intermolecular interactions.

Single crystals of [py][BiI4] and [mepy][BiI4]
were crystallized from solvent layering of ethanol solution of BiI3 and water solution of pyridinium or methylpyridinium iodide, and red plate/block-shaped crystals were formed at the interface of these two solutions by slow reaction in an NMR tube.

Thin films of [py][BiI4] and [mepy][BiI4]
were formed on glass/conducting FTO glass substrates by spin coating, using Laurell WS-6505-6NPP-LITE spin coater. A solution of each compound was prepared by mixing equimolar organic moieties (1.0 M for pyridinium iodide, methylpyridinium iodide) and BiI3 in the same solvent (DMF:DMSO=7:3) and stirring at room temperature for 3h. For spin-coating, 60μL of solution was carefully deposited over the entire surface of a 1.5 x 1.5 cm glass/conducting FTO glass slide. The spin-coating process was performed at 2000 rpm for 1 min, or a thicker layer was made with 1000rpm for 45s on the glass for powder XRD measurements. In both cases, this was followed by annealing on a hot plate at 100 ⁰C in a dry nitrogen glove box (relative humidity less than 30%, at room temperature), yielding a reddish-orange coloured film for For each compound, a suitable-sized crystal was selected and mounted on a MITIGEN holder in Paratone oil on a Rigaku Oxford Diffraction SuperNova diffractometer. The data collection was carried out under T = 120.0 K. Olex2 42 was used when solving the structure with the ShelXS 43 structure solution program. Direct methods were used for two crystals. For each compound, the model was refined with version 2014/6 of ShelXL 43 using Least Squares minimisation. All non-hydrogen atoms were refined anisotropically, and positions of hydrogen atoms were calculated geometrically and refined using the riding model.

Powder XRD
X-ray diffraction on powder samples (PXRD) was recorded for the compounds over 2θ scattering angle of 5-50 degrees. The samples were spin-coated on glass slides (see details in the film formation section) for measurements, and monochromatic CuKα radiation with wavelength of 1.5406 Å was used. The measurement was performed at room temperature on Bruker D8 Advance diffractometer, with a rotation speed of 8°/min, and an increasement of 0.05° on 2θ scale.

X-ray Photoelectron Spectroscopy
X-ray photoelectron spectra were obtained at the National EPSRC XPS Users' Service (NEXUS) at Newcastle University. All the spectra were acquired using a Kratos Axis Nova XPS spectrometer. The samples were manipulated using clean plastic tweezers and immobilised on top of a clean aluminium platen using carbon tape. Al-Kα radiation was used as the excitation source.

Computational Details
Density functional theory (DFT) based computational studies were carried out using the CASTEP 16.11 computational package 34 BiI4]), such that of dEtot/dEcut ≤ 0.003 eV per atom. Unit cell parameters were fixed at experimental values and structures were considered optimised once the standard convergence criteria were reached (Maximum change in system energy 2.0 × 10 -5 eV/atom, maximum force 0.05eV/ Å, and maximum root-mean square (RMS) atomic displacement 0.002 Å). Brillouin zone sampling for each structure was achieved using Monkhorst-Pack grids such that the separation between k-points was less than 0.08 Å -1 . Band structure plots were then calculated with respect to k-vectors linking high symmetry points, and also via selected k-points corresponding to strong intermolecular interactions, as indicated in Fig S4. Density of States plot (DOS) for each crystal structure were generated by OptaDOS, 36 which also permitted the projection to obtain the partial DOS plots, which permitted the contributions from different elements to be ascertained.

UV-Vis diffuse reflectance
Diffuse reflectance measurements were carried out for a spin-coated thin film and powders of each compound, on a Jasco V-670 spectrophotometer with SpectaManager software. The measurements were performed at room temperature in the range of 250-850 nm. The Kubelka-Munk function was used to analyse the data collected from diffuse reflectance measurements, and values of direct band gap of each compound were constructed by Tauc plots. The validity of assumptions on allowed direct lowest energy transitions was considered together with the result from DFT calculations.

Electrochemistry measurement
Electrochemical studies were carried out in dichloromethane, and 0.3 M N(C4H9)4PF6 was used as the supporting electrolyte. The solution was degassed with N2 for 15 minutes before each measurement. A three-electrode cell with an Autolab Type III potentiostat was used, and the results were analysed on GPES electrochemical software. The spincoated compounds on FTO conducting glasses were used as working electrodes; Pt was used as counter electrode; Ag/AgCl was used as reference electrode. In the cyclic voltammetry measurement for each compound, scan rates were varied between 20 and 200 mV/s for every redox process, and the potential of the ferrocenium/ferrocene (Fc+/Fc) couple was used as an internal standard, taken to be at 0.63 V against NHE.

Solar cell fabrication
Two detached electrodes on FTO-coated glass were formed by laser etching, and then ultrasonically cleaned with detergent, deionised water, ethanol and UV-ozone treatment. A compact layer of TiO2 was deposited on FTO via spray pyrolysis at 450 °C from a titanium diisopropoxide bis(acetyl acetonate) solution Then screen-printing method was used to deposit a 0.4 μm mesoporous TiO2 layer, a 1.5 μm ZrO2 spacer layer, and a ~10 μm carbon layer on the substrate. Details of slurries preparation and screen-printing parameters can be found in the literature. 39 Both TiO2 and ZrO2 layers were sintered at 500°C for 30 minutes; the carbon layer was sintered at 400°C for 30 minutes. The same precursor solutions were used as in film formation. 4 μL precursor solution was dripped from the carbon layer and penetrated into the mesoscopic triple layers. The solar cells were then heated at 40°C overnight with a glass cover, to allow a slow crystallization process.