Brightly Luminescent and Moisture Tolerant Phenyl Viologen Lead Iodide Perovskites for Light Emission Applications

Lead halide perovskites are outstanding materials for optoelectronics, but they typically feature low stability against external agents. To overcome this drawback, LHPs based on quaternary ammonium cations, such as phenyl viologen lead iodide (PhVPI), were found to be promising candidates, being water-resistant and thermally stable. In this Letter, the optoelectronic properties of the PhVPI are investigated by a combined experimental–theoretical approach. Although the as-prepared material is photoluminescence-inactive, a short thermal (5 min @ 290 °C) or laser annealing turns PhVPI into a highly luminescent material, in the 600–1000 nm range. The PhVPI PL emission was characterized at different annealing conditions, and the structural evolution following thermal treatments was investigated by means of X-ray diffraction, Raman, and NMR spectroscopies. Besides this, the electronic structure and emission properties were investigated by density functional theory simulations. The intense optical emission and high stability make PhVPI an intriguing material for applications related to light-emitting devices.

Synthesis. PhVPI was synthesized according to the procedure described in a previous work. 1 Thermal annealings were performed by treating the sample at 290 °C for 5 min. in air (ramp time from RT to 290 °C: 1 h).
X-ray diffraction. Powder X-ray diffraction analysis was performed by using a Malvern Panalytical X'Pert Pro MPD diffractometer equipped with a Ni-filtered Cu Kα source (λ=1.54184 Å) and an ultrafast RTMS X'Celerator detector. The determination of the amorphous content in the as prepared sample and in the thermally annealed sample was performed by the internal standard method 2 by mixing carefully weighted amounts of each sample with NIST 676a standard reference material (corundum phase Al2O3) and analysing the diffraction patterns of the mixtures with the Rietveld method for quantitative phase analysis (QPA) using the MAUD software package. 3 NMR spectroscopy. 1 H NMR spectral analyses of as prepared and thermally annealed PhVPI were performed on samples dissolved in DMSO-d6 using a 400 MHz Bruker Avance III spectrometer. 1 H-NMR spectra showed phenyl-viologen characteristic multiplets at δ 9.69 and 9.06 (assignable to protons of the pyridyl moiety in ortho and meta position relative to the aza group, respectively) in both the as-prepared and annealed sample (at 290 °C). The signals of the organic cation showed no changes after annealing. However, signals belonging to the solvent (DMF), from which the compound was formerly precipitated, were visible in the as-prepared sample, while disappearing after annealing, demonstrating the evaporation of the residual solvent.
Raman spectroscopy. For Raman measurements, the excitation laser was provided by a single frequency Nd:YVO4 lasers (DPSS series by Lasos) with emission wavelength equal to 532.2 nm. The Raman signal was spectrally analysed by means of a 750-mm focal length monochromator Acton SP750 equipped with a 1200 grooves/mm grating and detected by a back-illuminated Si CCD Camera (model 100BRX) by Princeton Instruments. The laser light was filtered out by a very sharp high-pass Razor edge filter (Semrock). The Raman spectral resolution was 0.7 cm -1 . A 20 objective (Olympus) with NA=0.4 was employed to excite and collect the light, in a backscattering configuration. The laser spot can be modelled as a gaussian with an experimentally measured σ = 0.75±0.02 µm.
Photoluminescence measurements. For photoluminescence (PL) measurements, the same experimental conditions used for Raman were employed (same excitation laser, same objective, backscattering configuration, Razor edge filter). In this case, the signal was spectrally analysed by means of a 200-mm focal length monochromator Isoplane 160 by Princeton Instruments, equipped with a 150 grooves/mm grating. The same CCD camera used for Raman was employed to detect the signal. The PL spectral resolution was 0.6 nm.
Photoluminescence spatial distribution. The PL spatial distribution was studied by projecting the PL signal onto a Si-CCD chip. This was done by performing measurements in a backscattering configuration, as discussed above. The collected signal was projected onto the chip by means of a long-focal-length lens and by keeping the slits opened. A long-working distance 100× objective (Zeiss, NA= 0.75) was used to excite the sample and collect the signal.
Time-resolved PL measurements. Time-resolved PL measurements were performed separately in the PL regions below 1.31 eV and above 1.91 eV. A supercontinuum laser at 530 nm was employed to excite the PL signal. An avalanche photodetector from MPD was employed to collect the signal.
UV-VIS spectroscopy. UV-VIS spectra in diffuse reflectance mode were acquired with a Shimadzu (Japan) UV2600 UV-Vis spectrophotometer equipped with an ISR-2600 Plus integrating sphere.
BaSO4 powder was used as reflectance standard.  Table S1 and Figure S9 of SI.
Self-trapping processes have been studied in the 4x2x2 supercell of PhVPI by using the CP2K package. 9 The hybrid PBE0 functional 10 was used by including dispersion interactions through the the revised Vydrov-van Voorhis nonlocal van der Waals density functional. 11 Cell parameters were kept fixed at the experimental values. DZVP basis set 12 and norm-conserving Goedecker-Teter-Hutter pseudopotentials 13 have been used with a cutoff on the charge density of 300 Ry. The auxiliary density matrix method 14 has been used to accelerate hybrid functional calculations. Thermodynamic ionization levels (TIL) of self-trapped electrons and holes were calculated by using the following expression 15 are the energies of self-trapped systems in the different state of charge q, εVB is the valence band of the pristine system and E q corr are finite-size supercell corrections for charged defects.
Makov-Payne corrections have been applied by using the static dielectric tensor of the perovskite εxx=8.6, εyy=6.6, εzz=5.7, estimated by following the approach of Umari et al. 16 PL emission energies have been calculated by simulating vertical transitions between the excited (trapped) and the ground state potential energy surfaces at the excited state equilibrium geometry.

Computational modelling of self-trapping processes in PhVPI
To investigate the nature of localized excitons, the self-trapping of excited charge carriers in the pristine PhVPI was simulated within the supercell approach. Calculations were performed in the 4×2×2 supercell of PhVPI by using the hybrid PBE0 functional 10 Figure S8). The localization of the hole in proximity of the STE increases the spatial overlap between the electron and the hole and their radiative recombination probability. 17 On the other hand, the self-trapped hole (STH) shows two different minima. In the less stable configuration, the hole partially localizes on a single iodide row in the supercell by leading to a small lattice rearrangement and a relatively deep (+/0) TIL placed at 0.14 eV above the VBM. An optical emission from this semi-localized state at 1.76 eV is calculated, close to STE emission.
The most stable configuration of the self-trapped hole, however, is associated to the formation of a Vk center, i.e. two iodines bound at ~3.3 Å to form a I2species in the inorganic channel (see Figure   4c of the Manuscript). This configuration is more stable than the delocalized hole by 0.53 eV and introduces a deep (+/0) TIL placed at 0.50 eV above the VBM. The large reorganization energy accompanying the hole trapping process indicates that the formation of the Vk center is a potential non-radiative recombination channel, even though a thermodynamic barrier to displace iodides from lattice sites and to form the Vk center is expected. Figure S1. Comparison between the calculated (red) and experimental (dark grey) Raman (left) and IR (right) spectra. The experimental IR spectrum was taken from ref. 1 The Raman spectra were acquired by laser excitation at 532.2 nm in a confocal microscope setup, as detailed in the Methods section. Three distinct regions in the phonon spectra can be identified: i) a low frequency region < 150 cm -1 associated to bending and stretching of the inorganics partially coupled to the torsion of the organic cation; ii) an intermediate frequency range between 150-1700 cm -1 associated to torsion, bending and stretching of the organic cation rings; iii) a high frequency range around 3000 cm -1 associated to the C-H stretching in the organic cation. Notably, some peaks in the IR 1700-3000 cm -1 range of the experimental spectrum are absent in the simulated spectrum. We ascribe these peaks to solvent impurities in the as-synthesized sample. Figure S2. PL rise upon laser annealing in PhVPI. Top: Normalised spectra acquired with excitation power Pexc = 50 µW, as a function of the annealing time, and for three different annealing powers. Bottom: Integrated intensity behaviour as a function of the cumulative annealing time, obtained by the spectra shown above. Indeed, the same qualitative behaviour is observed in the three cases, with an initial rise of the PL intensity followed by its quenching. By increasing the annealing power, the maximum intensity is observed for shorter annealing times.    Figure S5. Comparison between the PL emission from the thermally-annealed and laser-annealed samples and InP. PL spectra acquired with Pexc = 0.12 µW in the most efficient thermally-annealed (annealed at 290 °C for 5 minutes) and laser-annealed samples we obtained. The spectra are compared to that of a high-quality InP epilayer, showing comparable peak intensity. Indeed, the PL spectra from the PhVPI are much broader, implying a much larger integrated intensity.     calculations. Convergence on the number of bands and cutoff on exchange (EXXRLvcs) and dielectric matrix (NGsBlkXd) have been carried out in the plasmon-pole approximation to reduce computational effort. In all cases the same number of bands have been used to calculate dielectric matrix and correlation energy. A smaller band-gap is obtained by using a real axis full frequency approach. Based on convergence tests the following computational setup for (EXXRLvcs / NGsBlkXd / number of bands / frequency steps) has been used (30 Ry, 3 Ry, 1024, 150).

Computational setup
Direct energy gap at Γ (eV)