br Surface roughness characterisation The surface morphology
Surface roughness characterisation The surface morphology was characterised by Atomic Force Microscopy (AFM). The top view of glass, TiO2 compact layer and CH3NH3PbI3 thin film AFM images are shown in Fig. 1. The surface roughness of glass and TiO2 compact layer is negligible compared with CH3NH3PbI3 thin film. The median surface roughness of CH3NH3PbI3 thin film is 50nm and has been considered in the optical modelling, included as error bars representing measurement uncertainty.
Optical measurement and modelling Ellipsometry was carried out using a J.A. Woollam M-2000 Spectroscopic Ellipsometer in the wavelength range of 370–1690nm. All ellipsometry data in this study are collected from three incident angles 45°, 50° and 55°. The reflection (R) and transmission (T) measurements were carried out using a Varian Cary UV–vis–NIR spectrophotometer at normal incidence. Ellipsometry data and T were obtained for borosilicate glass and TiO2 on glass in the 0.8–3eV. R for the borosilicate glass and TiO2/Glass are not required, as the ellipsometry has been carried out in reflection mode and T in the 1–4eV range provides complementary information due to the different path length the light travels in the sample. WVASE® is used to model the optical properties of the individual borosilicate glass and TiO2 layer based on the ellipsometry and transmission data for these layers . The parameters are then fed into the model for the multi-layer stack to develop the optical model for the perovskite.
Glass substrates Cauchy model and two Gaussian dispersions (see Table 1 for parameters) were used to model the GSK343 of the 2.8mm thick borosilicate glass (to be later used for the CH3NH3PbI3 film modelling) and the second glass substrate (to be later used for the CH3NH3PbICl film modelling) in the transparent region and above 3eV respectively. Fig. 2(a); (b); and (c) shows the experimental and modelled amplitude component Ψ; phase difference Δ; transmission T of the borosilicate glass and the second glass substrate (except phase difference Δ for glass not shown as it is zero for the whole wavelength range) respectively which are used to determine the real (ε1) and imaginary (ε2) parts of its dielectric constants, see Fig. 2(d). A surface SiO2 layer on the bottom of 2.8mm thick borosilicate glass has been modelled to be 106nm thick using optical constants from Palik .
Tauc–Lorentz dispersion (see Table 2 for parameters) has been used to model the TiO2 thin film by combining Tauc bandedge with Lorentz broadening function . The fittings and corresponding optical constants are shown in Fig. 3. Fig. 3(a); (b); and (c) shows the experimental and modelled amplitude component Ψ; phase difference Δ; transmission T of the TiO2 on borosilicate glass respectively which are used to determine the real (ε1) and imaginary (ε2) parts of its dielectric constants, see Fig. 3(d). The TiO2 film thickness has been determined to be 44nm.
Perovskite The optical properties and films thicknesses of borosilicate glass and TiO2 layer were extracted and were fixed in the simulation of CH3NH3PbI3 perovskite properties. Two Psemi-Triangle (PSTRI) oscillators were used to describe the electronic transitions at absorption peaks, and Gaussian oscillators were used for the other regions. While for vapour-deposited CH3NH3PbICl perovskite film on glass by Wehrenfennig, optical properties of glass (measured and modelled in previous section) with a thickness of 1.7mm  were used to model the CH3NH3PbICl perovskite properties. The R and T were digitised from literature between 0.8–2.5eV, and two PSTRI oscillators were able to reproduce the experimental results in this range without the use of other oscillators at higher energy. Tables 3 and 4 list the parameters used for modelling ε2. A, E and B represent the amplitude, centre energy and broadening of each oscillator respectively. WL and WR stand for the endpoint positions relative to centre energy position, while AL and AR are the relative magnitudes of the left and right control points compared to amplitude (A).