Optical and photoconductivity analysis of LDP thin films

The optical properties of the low-n LDP thin films do not depend on the material crystallization. Indeed, the different samples absorption spectra are identical, displaying a sharp onset at 2.1eV (as calculated from the Tauc plot in Supplementary Fig. ³) and a clear excitonic peak at 567nm (Fig. 2a), typical of a n=2 LDP. We note that the horizontal sample shows a minor shoulder at longer wavelengths, indicative of a small fraction of n=3 phase, which also appears in the steady state photoluminescence (PL) spectra together with the peak related to the n=2 phase (see Supplementary Fig. ⁴). To gain a demonstration of the material stability, we monitor the PL peak position exposing the films under continuous 1 Sun AM1.5G illumination and thermal stress at 60°C (Supplementary Fig. ⁵). In both cases no variations are observed, indicative of a high robustness and phase integrity against UV light and thermal stressors.

Open in a separate window

Fig. 2

Optical and photoconductivity measurements.

a Normalized UV-Vis absorbance spectra of horizontal (in yellow) and vertical (in red) LDP thin films. b Electronic band structure of TMA₂MAPb₂I₇ from DFT simulations with the inset showing the electron and hole effective masses along the three conventional unit cell axes. Photo-induced sheet THz conductivity dynamics of the c horizontal (in yellow) and d vertical (in red) LDP thin films. The excitation fluence (excitation intensity per laser pulse) was varied from 5.3 μJ cm⁻² up to 53.8 μJ cm⁻² in a similar manner for both types of LDP thin films. e Effective electron-hole sum mobility estimates extracted from the conductivity values immediately post photoexcitation. Such analysis corroborates the vertical alignment of the inorganic chain as detected by X-rays, resulting in a corresponding boost in the charge-carrier mobility.

To investigate the effect of vertical alignment on the charge-carrier transport properties, we start by analysing the electronic structure of TMA₂MAPb₂I₇ (shown in Fig. 2b), obtained using density functional theory (DFT) calculations. The primitive cell of TMA₂MAPb₂I₇, along with the Brillouin zone showing the high symmetry directions, are shown in Supplementary Fig. ⁶. TMA₂MAPb₂I₇ has a direct band gap at the Γ-point. The flat bands along the high symmetry direction Γ-Y indicate heavy carriers perpendicular to the PbI₆ plane, while the more dispersive bands along Γ- Δ₀, and Γ-Z directions suggest lighter carrier transport within the plane. This is corroborated by the effective mass values (inset Fig. 2b) along the three axes of the conventional unit-cell, which are higher along the vertical b-axis and moderate along the horizontal a- and c-axes for both electron and hole carriers. The calculated effective masses are in good agreement with previous reported values for LDPs²⁶. These results demonstrate that carrier transport can be significantly enhanced by growing these LDPs with vertical alignment.

In order to further assess how the different orientations of the LDP planes impact charge-carrier transport at the nanoscale level, we employ optical-pump THz-probe photoconductivity spectroscopy. A sub-ps laser pulse at 3.1eV excites the LDP films, and the resulting photoconductivity is recorded for a range of fluences, as shown in Fig. 2c and d for samples with horizontal and vertical layer orientation with respect to the substrate, respectively. We may further extract the effective electron-hole sum mobility from the conductivity recorded immediately after photoexcitation, under knowledge of the absorbed photon density (see methods section). We note that since the electric field of the THz probe pulse oscillates at right angle with respect to its propagation through the substrate, such mobility values reflect charge-carrier motion within the thin-film plane, parallel to the substrate. Figure 2e and f show that these films indeed exhibit the expected behaviour for highly oriented LDP films. A high in-plane mobility of 4.9 cm²V⁻¹s ⁻¹ is observed when the THz electric field oscillates fully along the LDP planes, for the horizontal samples. For the vertical sample, on the other hand, the E-field vector lies on average half along the LDP planes, and half in the direction vertical to the planes. The observed mobility of 2.2 cm² V⁻¹ s ⁻¹ is indeed lowered by nearly a factor of 2, as would be expected given the insignificant mobility of charges across spacer layers (in line with the effective mass values in Fig. 2b), confirming the highly oriented nature of the LDP films^27,28.

Insights into chlorine effects on vertical crystallisation

We now focus our attention on the fundamental question: why does the addition of Cl have such a strong impact on guiding the vertical crystallization? For standard MAPbX₃ perovskites, it is well known that MACl-assisted perovskite crystallization favours the α-phase (with respect to non-perovskite phases, as in the case of formamidinium-based perovskites)^29,30, and that Cl sublimates during the process³¹. In order to verify the presence of Cl and map its content, we performed time-of-flight secondary ion mass spectroscopy (ToF-SIMS), as shown in Fig. 3a and b for the horizontal and vertical samples, respectively. For the horizontal sample Cl is absent, comparable in intensity to the background signal. In contrast, in the vertical sample there is a gradient distribution of Cl which maximizes at the indium tin oxide (ITO)/perovskite interface. In addition, for both horizontal and vertical LDPs, the CN, I and S signals are uniformly distributed across the whole film thickness.

Open in a separate window

Fig. 3

Chloride inclusion in the crystal structure.

ToF-SIMS analysis of a horizontal and b vertical thin films deposited onto ITO substrates, showing an increasing Cl distribution from the surface towards the bottom edge of the vertical sample. c GIWAXS pattern of horizontal (in yellow) and vertical (in red) LDP thin films in the 0.5–1nm⁻¹ q range. d Simulated phase diagram for the thermodynamic analysis of Cl-incorporation at I-sites in TMA₂MAPb₂I₇ and, e b-lattice parameter for 1121 different DFT optimized structures randomly and uniformly distributed over 0 < x<0.21 and 0.79 < x<1 for x in I_1–xCl_x. For clarity, to show the stability region (white, marked S), we have broken the x-axis at x=0.21 and 0.79, between which the unstable region (pink, marked U) continues. The metastable region is indicated by the blue shadowing and marked with M. f Unit cells of x=0.21 Cl at I-sites with the maximum (top) and minimum (bottom) b-lattice parameters. The spacer cations are omitted for clarity.

To further deepen our understanding on Cl inclusion in the film, we conducted a combined experimental and computational modelling analysis, whose results are reported in Fig. 3c–f. First, by closely comparing the GIWAXS profiles of the two samples, a shift in the (0k0) peaks is observed (Fig. 3c). The pattern of the vertical sample can be indexed by a slightly larger orthorhombic unit cell, especially along the axis perpendicular to the octahedra, with axes 8.5 × 42.5 × 9.0 Å (with respect to 8.8 × 41.5 × 8.8 Å for the horizontal sample). The expansion of the long axes can be the result of Cl incorporation inside the crystal unit, leading to structural distortion and lattice expansion. While Cl inclusion in the unit cell is known to be energetically unfavourable in MAPbX₃ perovskites, the situation appears to be different for low-n LDPs, due to their layered, quantum-confined structure. Cl partially substitutes I in the equatorial position, creating a strain in the octahedra and thus elongating the unit cell along the vertical axes. Such vertical compression induces a switch in crystal orientation, responsible for the vertical alignment. A similar structural effect, where Cl incorporates in the equatorial position substituting I, has been recently demonstrated in LDP quantum confined nanocrystals, strongly corroborating our results³².

To verify our hypothesis and gain additional atomic-scale insights into Cl incorporation in TMA₂MAPb₂I₇, we have used DFT-based methods to investigate the thermodynamic stability and miscibility of Cl-mixing at I-sites. The positive mixing energies for all the structures between 0 < x<0.21 and 0.79 < x<1 in I_1–xCl_x (Supplementary Fig. ⁷) indicate a low tendency for Cl/I mixing, although the structures near the end points (x=0, 1) can be accessible at finite temperatures. The thermodynamic stability phase diagram is shown in Fig. 3d. Three important features emerge. First, a stability region exists close to the I-rich (x=0) and Cl-rich (x=1) regions over 300K, which grows with an increase in temperature. This indicates that Cl can be incorporated into I-sites in small amounts (~2% at 373K). Second, small metastable regions exist within 10% of the I-rich and Cl-rich regions (x<0.1 and x>0.9) for the entire temperature range. Being thermodynamically metastable, these states may be accessed and made kinetically stable via alternate synthesis methods. Third, the I/Cl-mixing states are unstable in the wide range of 0.1 < x<0.9. Overall, these results show that it is possible to incorporate Cl at I-sites, but only in small amounts.

Next, we examined the b-lattice parameter variation of all the structures between 0 < x<0.21 and 0.79 < x<1 in I_1–xCl_x to investigate the structural distortion upon Cl addition (Fig. 3e; further analysis of the structural and energetic variation with Cl content is shown in Supplementary Figs. ^8,9). Interestingly, compared to the structures with the minimum b-lattice parameter at each concentration, the structures with the maximum b-parameter (i) are energetically favoured, (ii) have a preferred equatorial distribution of Cl at I-sites (Fig. 3f), and (iii) have larger spacer cation distances. Examining the energetically most stable structures at each concentration, Cl ions are found to be predominantly distributed at the equatorial I-sites. The structures with the highest b-lattice parameter have lower unit cell volumes due to compression along the a- and c-lattice parameters. Analysis of the N-C-C and C-S-C bond angles of the spacer cations from 0 <x<0.21 indicate that there is no major stretching of the spacer cation ring, whereas the terminal N-C-C bonds are affected by the presence of highly electronegative Cl^- ions in close vicinity. In summary, the equatorial distribution of Cl is energetically preferred, which causes compressive strain on the unit cell along the a- and c-lattice parameters. This creates steric hindrance among the spacer cations, leading to an elongation of the b-parameter, which is in line with our GIWAXS results.

Finally, by testing the effects on crystal orientation of other Cl^- sources such as thiophenemethylammonium chloride (TMACl) and lead chloride (PbCl₂) (Supplementary Fig. ¹⁰), we could ultimately confirm the role of the anion on triggering verticalization.

Low-dimensional perovskite thin film and solar cell fabrication

The precursor solution (0.7M) for the horizontal TMA₂MAPb₂I₇ (n=2) LDP was prepared by mixing stoichiometric ratio of TMAI, MAI and PbI₂ powders in DMF. In the case of the vertically oriented LDP, 70mol% of MAI was replaced by MACl, without changing the total stoichiometric amount of MA in the solution. After complete dissolution, the solution was subjected to a thermal annealing step of 2h at 60°C. The solution was deposited onto glass substrates via single-step spin coating procedure at 4000rpm, 2000 rpm s⁻¹ acceleration for 40s. 120µl of toluene was dropped onto the spinning substrate at 20s before the end of the programme for antisolvent procedure. Subsequently, samples were annealed at 100°C for 30min.

For PSCs fabrication, prepatterned ITO/glass substrates were sequentially cleaned with acetone and 2-propanol (IPA) by ultrasonication for 15min in each solvent. ITO/glass substrates were then dried with N₂ and treated with oxygen plasma at 100mW for 10min. Then, substrates were moved into a N₂-filled glovebox for subsequent layers deposition. MeO-2PACz self-assembly monolayer was deposited by spin coating (3000rpm, 30s, 1500rpms⁻¹) a 0.335mgmL⁻¹ solution in ethanol, followed by annealing at 100°C for 10min. The LDP layer was fabricated following the same procedure employed to prepare thin films. A PCBM solution (20mgmL⁻¹ in chlorobenzene) was deposited onto the perovskite by dynamic spin coating at 2000 rpm, with an acceleration of 1000rpms⁻¹ for 30s, followed by 10min annealing at 100°C. Finally, 20nm of SnO₂ were deposited by atomic layer deposition (ALD, Picosun) and 80nm of Ag were thermally evaporated, producing pixels with 0.045 cm² active area (shadow mask aperture area estimated by optical microscopy).

For semi-transparent devices, 100nm of IZO was sputtered (Angstrom EvoVac) onto the SnO₂ buffer layer as top contact electrode (pixels active area ~0.1 cm²).

Grazing incident wide-angle X-ray scattering

GIWAXS experiments have been performed at the MINA instrument at the University of Groningen. The instrument is built around a Cu rotating anode source, delivering a high-power X-ray beam with a photon energy of 8keV ($\lambda$=0.15413nm). The samples have been measured in GI geometry, using an incident angle of 1°, which ensures full penetration of the sided LDL films. The films were measured in a primary vacuum to avoid air scattering and sample degradation during the measurements. GIWAXS patterns were collected using a Vantec500 Bruker detector with 0.068 × 0.068mm x mm pixel dimension placed 10cm away from the sample. The angular range was calibrated using the diffraction peaks from a standard Al₂O₃ sample and data analysis was conducted using the GIXSGUI Matlab-based programme³⁹.

Pole figure

Pole Figure (PF) technique is based on the acquisition of XRD data by varying the incidence angle along two different axes to build PF maps. In such way, different orientations of crystalline samples are proved and compared with the Wulff Stereographic Projection (WSP) calculated from single crystal data. Pole Figure (PF) analysis was performed with the detector set in 0D mode, using 0.04rad soller slits and 0.28° parallel plate collimator (PPC) on the diffracted beam. This configuration allows a maximum sample tilt equal to χ=85°. Simulated pattern for the TMA₂MAPb₂I₇ n=2 LDP perovskite was obtained with CrystalDiffract 6.9.4 from CrystakMaker Software Ltd. Pole Figure analysis was carried out using X’Pert Texture 1.3 software from PANalytical. The Wulff stereographic projections (WSP) were created using WinWulff 1.6 from JCrystalSoft.

UV-Vis absorption

The absorption spectra of the perovskite thin films were collected using a UV-NIR spectrophotometer from Perkin Elmer (Lambda 1050+).

Photoluminescence measurements

The photoluminescence (PL) spectra of the different samples were collected using a customized optical setup⁴⁰ in which the excitation was provided by pulsed/CW laser at 470nm (PicoQuant), while the luminescence was analysed with an interferometer (GEMINI by Nireos) and recorded with a single photon detector (IDQuantique) coupled with a Time Tagger (Swabian Instrument).

Computational modelling

First-principles simulations were performed within the framework of density functional theory (DFT) as implemented in the Vienna Ab-initio Simulation Package (VASP)^41–43 using a projector augmented-wave (PAW)⁴⁴ basis. The generalized gradient approximation (GGA) exchange-correlation functional of Perdew Burke-Ernzerhof (PBE)⁴⁵ was used for all the structures. VASP-recommended PAW pseudopotentials were used for all the atomic species (C:2s²2p², N:2s²2p³, H:1s¹, S:3s²3p⁴, Pb:5d¹⁰6s²6p², I:5s²5p⁵, and Cl:3s²3p⁵). Grimme’s DFT-D3 dispersion correction⁴⁶ was considered due to the presence of organic molecules. A plane wave energy cut-off of 500eV and a Brillouin zone sampling using Γ-centred k-mesh are employed done REMOVE DONE for all the calculations. A k-point grid of 3 × 1 × 3 with a Gaussian smearing of 0.1 was used for structure optimization whereas a k-point grid of 6 × 2 × 6 with Gaussian smearing of 0.01 was used to obtain accurate charge densities to calculate the electronic structure. The cell volume, shape, and atomic positions were fully relaxed using the conjugate gradient algorithm until the force on each atom converged below 0.01eVÅ⁻¹. The self-consistent energy convergence of 10⁻⁶eV was used for electronic degrees of freedom.

The carrier effective masses were calculated using the Effective Mass Calculator (EMC) code⁴⁷. We used the Site-Occupation Disorder (SOD) code⁴⁸ to generate symmetrically inequivalent structures in the orthorhombic space group Aba2 (#41) for 0 < x<0.21 and 0.79 < x<1 in I_1–xCl_x. It is unfeasible to enumerate the thousands of symmetrically inequivalent structures for 0.21 < x<0.79 in I_1–xCl_x. Therefore we used the icet code⁴⁹ to generate randomly occupied structures in this compositional range. The thermodynamic stability phase diagram for Cl mixing at I-sites was calculated using the Generalized Quasi-Chemical Approximation (GQCA)⁵⁰, which has been effectively utilized in the thermodynamic analysis of perovskite materials^51–53. 235 lowest energy-optimized structures distributed uniformly over each possible concentration from 0 < x<1 in I_1–xCl_x were used to generate the thermodynamic stability phase diagram. 1121 different structures chosen randomly and distributed uniformly between 0 < x<0.21 and 0.79 < x<1 in I_1–xCl_x were optimized to analyse the preferred distribution of Cl and their effects on the lattice distortion.

THz photoconductivity and mobility measurements

An amplified laser system (Spectra Physics, MaiTai – Ascend – Spitfire), with a 5kHz repetition rate, centre wavelength of 800nm and pulse duration of 35fs is used to generate the THz radiation using a spintronic emitter. The THz probe is then focused onto the sample, overlaid with a 400nm excitation pump that is generated using a Beta Barium Borate (BBO) crystal. The THz radiation transmitted through the sample is then detected via free-space electro-optical sampling in a ZnTe (110) crystal of thickness 200 μm⁵⁴.

The variations in the THz transmission ΔT T⁻¹ through the sample can therefore be measured, and temporally resolved by varying the time delay between the optical pump and the probing THz pulses in order to obtain the temporal evolution of the photo-induced conductivity of the sample^51,52.

Additionally, for low enough laser fluences and charge-carrier densities, the sum mobility of free charge carriers will have a linear dependence on photoconductivity values and can therefore lead to an extraction of the intrinsic effective sum mobility based on the early time conductivity values post photoexcitation^55,56.

We also ascertained that the THz mobility values extracted within the first picoseconds indeed solely reflected the motion of charge carriers within the n=2 LDP phase, by also recording time-resolved photoluminescence spectra, using a time-correlated single photon counting (TCSPC) device following excitation by a 398nm picosecond pulsed diode laser (PicoHarp, LDH-D-C-405M). These PL measurements showed that over the first nanoseconds of charge-carrier lifetimes for which the THz photoconductivity transients were recorded, the PL solely reflected emission from an n=2 LDP phase.

Time-of-flight secondary ion mass spectroscopy

ToF-SIMS analyses were acquired with ToF-SIMS 5–100, with the pulsed primary ions from a Cs⁺ (2keV) ion gun for sputtering and a Bi⁺ ion beam (30keV) for analysis.

Transmission Electron Microscopy

Transmission electron microscopy (TEM)-based experiments were conducted in both parallel beam mode high-resolution TEM (HRTEM) and converge beam scanning TEM (STEM) mode. The HRTEM experiments were performed in Cs spherical aberration image-corrected Thermo Fisher Titan 60–300 Cubed TEM microscope. In STEM mode, High-Angle Annular Dark-Field STEM HAADF-STEM) imaging combined with Energy Dispersive Spectroscopy (STEM-EDS) was executed for elemental mapping on Cs-aberration-probe-corrected Thermo Fisher Titan 60–300 Cubed TEM microscope operating at 300kV. TEM data processing was performed using Gatan™ Digital Micrograph and Thermo Scientific™ Velox suites. For the TEM study, cross-sectional electron-transparent lamella was prepared in a focused ion beam (FIB) equipped scanning electron microscope (SEM-FIB Helios G4 DualBeam, FEI) with the help of an EasyLift nanomanipulator and Ga ion source. To protect the region of interest during FIB, two types of protective coatings were deposited: a 0.5µm layer of Pt coating deposited by the e-beam, followed by a 3µm layer of Pt coating deposited by the ion beam for final protection. Step-by-step ion beam milling procedure with beam current (2.4, 0.44, 0.26, 0.045, 0.025nA for 30 to 5kV) was carried out by cutting and thinning down lamella to 60nm while decreasing beam current (till 0.025nA at 2kV) to avoid ion beam damage. Additionally, a low current cleaning process (5–2kV, 81-28 pA) was performed to remove any potential contamination.

Photovoltaic device characterization

Current density (J)-voltage (V) measurements were performed in ambient conditions under simulated AM1.5G light with an intensity of 100mWcm⁻² (Wavelabs-Sinus 70). Intensity was calibrated using a calibrated Si reference cell (KG-3 filter Centronics LCE-50). Cells were scanned using a Keithley 2450 source-metre backward and forward from 1.5V to −0.1V, with a 100 mVs⁻¹ scan speed (no dwell time or preconditioning applied). A shadow mask with an aperture of 0.03 cm² (measured through optical microscopy) was employed during the measurement. Operational stability tests were carried out by maximum power point tracking (MPPT) under AM1.5G illumination (100mWcm⁻²) in N₂ atmosphere at room temperature. External quantum efficiency (EQE) measurements were performed with a Cicci Research s.r.l ARKEO steady-state tests module. Wavelength scan range was set between 300 and 900nm with a scan step of 10nm.

Average visible transmittance and light utilization efficiency computation

AVT has been calculated using Eq. 1 through the widely reported approach⁵⁷:

where λ is the wavelength, T is the device light transmission, P is the photopic response function, and S is the AM1.5G solar irradiation.

LUE has been calculated as the product of AVT value and device power conversion efficiency (PCE), as presented in Eq. 2:

The authors thank the University of Pavia through the programme “Dipartimenti di Eccellenza 2023–2027” and Fondazione Cariplo Economia Circolare 2021 Project (FLHYPER, no 20201067), funded under the “Circular Economy-2020” call. G.G. acknowledges the “HY-NANO” project that received funding from the European Research Council (ERC) Starting Grant 2018 under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement no. 802862) and the GOPV project (CSEAA_00011) which received funds from Bando Ricerca di Sistema - CSEA - TIPO A Piano triennale 2019 −2021 Decreto direttoriale 27 ottobre 2021 del Ministero della Transizione Ecologica; MASE-(ex MITE). The authors acknowledge the bilateral project CRG2022 - HERO2D/3D Perovskite Heterojunction Solar Cells: Pairing Performance with Stability. The authors acknowledge funding by the Engineering and Physical Sciences Research Council (EPSRC); U.K. M.S.I. and V. acknowledge the Engineering and Physical Sciences Research Council (EPSRC) Programme Grant (EP/X038777/1) for a postdoctoral fellowship and are grateful to the UK’s HEC Materials Chemistry Consortium (EP/X035859/1) for the use of the ARCHER2 high-performance computing facilities. The authors acknowledge the use of the KAUST Core Labs facilities for some parts of device fabrication and TEM measurements. The authors are grateful for the technical support in TOF-SIMS measurement from Nano-X vacuum interconnected nanotech workstation (Nano-X) of Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences (SINANO). L.M.H acknowledged financial support by the EPSRC. G.G.acknowledges Dr. Stefano Toso (IIT, Italy) for useful discussion.