WorksPhysics and Astronomy
Steric engineering of point defects in lead halide perovskites
Abstract
Due to their high photovoltaic efficiency and low-cost synthesis, lead halide perovskites have attracted wide interest for application in new solar cell technologies. The most stable and efficient ABX perovskite solar cells employ mixed A-site cations, however the impact of cation mixing on carrier trapping and recombination – key processes that limit photovoltaic performance – is not fully understood. Here we analyse non-radiative carrier trapping in the mixed A-cation hybrid halide perovskite MACsPbI. By using rigorous first-principles simulations we show that cation mixing leads to a hole trapping rate at the iodine interstitial that is eight orders of magnitude greater than in the single cation system. We demonstrate that the same defect in the same material can display a wide variety of defect activity – from electrically inactive to recombination centre – and, in doing so, resolve conflicting reports in the literature. Finally, we propose a new mechanism in which steric effects can be used to determine the rate of carrier trapping; this is achieved by controlling the phase and dynamical response of the lattice through the A-site composition. Our findings elucidate crucial links between chemical composition, defect activity and optoelectronic performance, and suggest a general approach that can help to rationalise the development of new crystalline materials with target defect properties.
mance, and suggest a general approach that can help to rationalise the development of new crystalline materials with target defect properties.
Introduction
Organic-inorganic lead halide perovskites (OLHPs) have attracted wide interest for their application in optoelectronics. Single junction halide perovskite solar cells now exceed 25% power conversion efficiency, [1] with the most stable and efficient devices employing mixed organic (methylammonium, MA and formamidinium, FA) or inorganic (Cs, Rb) cations on the -site of the perovskite structure. -site cation engineering is primarily used to improve thermal and chemical stability;[2] the impact of -site mixing on defect activity is not fully understood, despite this being crucial for the development of devices with increased efficiency.[3]
The -site cation indirectly determines various optoelectronic, transport and defect properties through an influence on the symmetry and dynamic response of the crystal lattice.[3, 4] Although defect formation and activity is sensitive to the exact system under consideration, the common observation across OLHPs is that halide ions form abundant, mobile point defects which are active in carrier trapping and recombination.[5, 6] Furthermore, the bonding in OLHPs is relatively weak, leading to an easily distorted (‘soft’) lattice and large lattice relaxation after carrier capture at a defect site.7,8
Large lattice relaxation is not commonly observed in all-inorganic optoelectronic materials. It is more akin to what occurs in organic molecular materials,9 where steric engineering through the incorporation of bulky ligands is used to restrict the vibrational modes associated with lattice relaxation after electronic excitation.9,10 Here we propose that a similar approach can be used to rationalise the design of defect-tolerant OLHPs. Using first-principles quantum chemical calculations, combined with methods adapted from the field of organic semiconductors, we show that cation mixing in can be used to adjust the hole trapping rate at the iodine interstitial by eight orders of magnitude. Furthermore, using symmetry mode analysis we demonstrate that defect activity can be tuned by controlling phase formation through the steric size of the A-site cations. This provides a new route for engineering defect properties without altering the metal–halide chemistry that is beneficial for photovoltaic performance.
Methods
Electronic structure calculations
The underlying electronic structures were calculated using density functional theory (DFT) as implemented in VASP11 using a plane wave basis set with an energy cutoff of eV. A gamma centered Monkhorst–Pack mesh was used for the Brillouin zone integration. The interstitial was placed in a -atom supercell. We calculated for the iodine interstitial in a atom unit cell and found it to be converged within . A-site ordering for the mixed cation systems were determined using Special Quasi-random Structures12 as implemented in the ICET code.13
Ground state geometries were found using the PBEsol functional14 with a force cutoff of eV . The interpolated geometries were generated using custom code (available in an online repository15) and the Atomic Simulation Environment.16 The potential energy surface was calculated using the screened-exchange HSE06 functional17 with and spin–orbit coupling. The total energy cutoff was eV. We used a uniform reduction factor to evaluate HSE06 energies at the gamma point only. We calculated the neutral defect formation energy in using the full k-point grid and found it to be converged within eV per formula unit. We employed a delta self-consistent field approach to constrain the occupation of the defect states near energy level crossings. The electron–phonon coupling term was derived from wavefunctions calculated at the same level of theory.
Defect properties
The formation energy of a defect in charge state is given by
where is the total energy of the defect lattice in charge state , is the total energy of the pristine lattice, is the chemical potential of species and is the number of atoms that are added or removed. , , and were calculated using DFT, as outlined in the previous section. includes a correction term for charged defects, which was calculated using sxdefectalign with a value of for the static dielectric constant. 18,19 More details on the methodology as applied to hybrid halide perovskites have already been published in Reference. 8
A quantum mechanical treatment of electron capture was performed using the open-source CarrierCapture package20 which builds on the approach outlined in Reference.21 In this model the carrier capture coefficient for capture from an initial state i to a final state f is given by
where is the total change in energy, is the supercell volume, is the energetic degeneracy of the final state, is the electron-phonon coupling matrix element, is the overlap of the vibrational wavefunctions , and the Dirac ensures that there is conservation of energy. In practice the Dirac term is replaced by a smearing function; for the calculations in this study this is a gaussian function of width 0.01 eV. is the thermal occupation of the vibrational state . The electron-phonon coupling term was calculated using the Nonrad package.22 Further details of the methodology can be found in the literature. 21
Geometry and symmetry analysis
Bond lengths and bond angles were analysed using the Atomic Simulation Environment.16 Crystal structures were visualised using vesta.23 Isodistort was used for symmetry mode analysis, with MAPbI in the parent cubic phase as a reference structure. The rotational motion of the MA molecule was not considered in this analysis; all A-sites were modelled as point particles. The phonon mode amplitudes were normalised to the parent cell volume to allow comparison between different compositions.
Results
Carrier capture rates with molecular rotations
We focus our analysis on the iodine interstitial defect in the negative () and neutral () charge states as these have been found to be most detrimental to solar cell efficiency.24 The iodine interstitial is a negative-U defect so it follows that is metastable.[5] [8, 25, 26, 27] However can still be formed through electron capture at or hole capture at . As shown in Figure 1a, the neutral iodine interstitial bonds with a lattice iodine to produce a molecular H-centre with a trapped hole.27 After electron capture or hole release the negative charge state is formed in a split-interstitial configuration. This is accompanied by tilting and distortions of the inorganic octahedra to accomodate the I-I bond lengthening. The inorganic structural changes are coupled to MA rotations, similar to the behaviour observed during thermal phase transitions.28
To model the kinetics of charge capture at a defect site we must consider the coupling of electronic and atomic structure. We map the potential energy surface (PES) between two charge states as a function of a collective coordinate
where the sum is over atoms with mass and a displacement from equilibrium of .
For inorganic and hybrid materials the standard procedure is to conduct a linear interpolation between each atomic position of the two equilibrium structures, 21,29,30 and this is the procedure that has been previously applied to hybrid perovskites. 8,25,26,31 Contributions from the rotation of the molecular cation have previously been ignored as linear interpolation adjusts the intramolecular bond lengths to give unphysical energies.8
We introduce an interpolation method (which we will term ‘Kabsch interpolation’) most commonly used for molecular materials, but transferred here for the first time to a hybrid inorganic-organic material. To interpolate the molecular species we first express atomic positions in the initial and final geometries as a set of vectors, and respectively. We use the Kabsch algorithm to calculate an optimal rotation axis and angle that maps between and whilst minimising the root mean squared deviation between each vector pair. 32 This allows us to rotate the molecule around using a linear interpolation of . Finally, we combine this with a linear interpolation of both the molecular centre of mass and inorganic framework. This method generates a more accurate energy surface that allows us to consider molecular translations and rotations. Unphysical energies are avoided as the molecule is treated as a rigid object that cannot be deformed.
Figure 1b shows configuration coordinate diagrams for the transitions in MAPbI, using standard linear interpolation (in grey) and Kabsch interpolation (in colour). We use the Heyd-Scuseria-Ernzerhof hybrid functional17 alongside spin-orbit coupling to obtain accurate defect energetics. corresponds to the equilibrium configuration of and corresponds to that of . A nonradiative recombination process beings at with an electron at the conduction band minimum (CBM) and hole at the valence band maximum (VBM). This is represented with the orange dash line in Figure 1b. After electron capture there is a transition to (blue dot line); in the semiclassical picture this requires overcoming the energy barrier . Subsequent hole capture over the energy barrier to (green solid

Figure 1. Carrier capture processes in single and mixed cation perovskites. Predicted energies are also given in Table S1. (a) Crystal structures for: i) a pristine (defect-free) lead halide perovskite material; ii) the neutral iodine interstitial in a H-centre configuration with a localised hole; iii) the negative iodine interstitial in a split-interstitial configuration. (b) Configuration coordinate diagram for the neutral and negative iodine interstitial in MAPbI. Each scatter point represents a DFT calculated total energy. The solid grey lines are the potential energy surfaces (PES) generated using a lower accuracy interpolation method that describes translations only (no molecular rotations). (c) Nonradiative carrier capture coefficients for the iodine interstitial in MAPbI. Calculated for electron capture at a neutral iodine interstitial (yellow solid line), hole capture at a negative iodine interstitial (blue solid) and electron capture followed by hole capture (red dash) (d) Ratio of electron capture rate (from the conduction band) and hole emission rate (into the valence band) for the neutral iodine interstitial in MAPbI. (e) Configuration coordinate for the neutral and negative iodine interstitial in MACsPbI. The grey dot line is the PES of the negative iodine interstitial in MAPbI, given for comparison. (f-g) As in (c-d), but for MACsPbI.
line) completes the recombination cycle. For Kabsch interpolation the energy surface is softened as energy is dissipated through rotations of the MA cation, resulting in a significant reduction of (0.15 eV to 0.025 eV) and (0.92 eV to 0.63 eV). In semiclassical models the capture rate has an exponential dependence on the ratio of barrier height to , so changes of meV, as seen here, can have a significant impact on defect activity.
For accurate predictions of capture rates, we use a quantum chemical theory to calculate coefficients for electron capture () and hole capture (). 21,29 This moves beyond the semiclassical picture to include the strength of coupling between the defect state and CBM (, for electron capture), or the defect state and VBM (, for hole capture). It also considers the vibronic overlap between each PES, thus allowing quantum tunnelling below the classical barrier. The capture coefficients determine the capture rate at a defect. To take electron capture at a neutral defect as an example,
where is the neutral defect density and is the electron density. We consider electron capture followed by hole capture, the total rate of which is quantified using
We note that this does not correspond to the total rate of non-radiative recombination at the iodine interstitial, as we do not consider non-radiative recombination with hole capture as the initial step.
Figure 1c shows that although electron capture is fast at 300 K, the non-radiative recombination process is limited by the slow rate of hole capture. Importantly, competing with the first step of this process (electron capture) is hole emission from the localised state associated with (green solid line) to a delocalised state in the valence band (forming , blue dot line). This process does not require photoexcitation so can happen ‘in the dark’. The ratio of the electron capture rate to the hole emission rate is given by:
where is the density of occupied states in the valence band and is the electron concentration. 33 Assuming a hole effective mass value of 34 and that the electron concentration is , 35 hole emission at the neutral iodine interstitial will occur faster than electron capture at temperatures above 160 K (Figure 1d). It is important to highlight that once the negatively charged iodine interstitial is formed, whether through electron capture or hole emission, it is limited by the slow rate of hole capture ( at 300 K) and is electrically inactive.
Carrier capture in mixed A-cation systems
We now investigate the impact of cation mixing on nonradiative trapping and recombination processes. We consider the mixed cation system and compare this against the control case of . Initially we focus our analysis on the mixed A-cation system , which is close to the Cs concentration reported to be optimal for device efficiency. 36 We find that a H-centre defect is formed with charge localisation around the iodine dimer (Figure S1), indicating that the basic defect activity is comparable to .
Figure 1e shows that the total lattice relaxation is suppressed through Cs incorporation ( to ). We note that, despite the reduction, this lattice relaxation is still relatively large as is typically less than for all-inorganic materials. 29,30 Analysis of the defect structure for each charge state shows that the reduction can be primarily attributed to reduced displacements of Pb and I. In particular, the change in Pb-I-Pb bond angle after charge capture is reduced, suggesting that octahedral rotations are suppressed in the mixed cation materials (Table S3).
fig:1e also shows an increase in the neutral to negative charge transition level ( to ). The charge transition level corresponds to the difference in total energy of defect states and evaluated at their equilibrium configurations and , respectively, and referenced to the valence band edge of the host material. 37
The charge transition level (CTL) is a key parameter for classifying defect activity in the semi-classical defect model. For example, defects within a few of the band edge typically show single carrier trapping and de-trapping behaviour, whilst ‘deep’ defects towards the middle of a band gap may successively trap both carrier species and form a site for non-radiative recombination. 33 We note here that the CTL is distinct from the Kohn-Sham defect eigenvalue as the former accounts for structural relaxation after charge capture or release. The Kohn-Sham defect eigenvalue, often referred to simply as ‘defect level’, has been monitored in previous studies which show that there are large energy fluctuations over 1 eV. 38–40 After cation mixing we find that CTL increases so that the two charge states become close to thermodynamic equilibrium. As a result, fast hole trapping and de-trapping behaviour is expected.
We find that the shape of the PES is largely unchanged after Cs mixing. This is especially true at small displacements around the equilibrium structures, where there is an equal softening of the harmonic PES after Cs incorporation for both the negative (43 cm to 35 cm) and neutral (58 cm to 50 cm) charge states. For comparison, the DX-Centre in GaAs has an effective harmonic frequency of 81 cm. The increase in , combined with a relatively rigid displacement of the and PES in - space, leads to a reduction in the hole capture barrier (0.63 eV to 0.23 eV) and a small increase in the electron barrier height (0.025 eV to 0.045 eV). The reduction in indicates that there will be an increased rate of hole trapping at the negatively charged iodine interstitial after cation mixing.
To quantify the impact of cation mixing we analyse the same multiphonon capture and recombination processes as outlined in Section . We find that the defect state couples more strongly with the valence band () compared to the conduction (); this is expected as the VBM has a larger density of electronic states derived from iodine -orbitals (Figure S2). This, combined with the reduction in , results in a carrier recombination process that is more balanced between electron and hole capture, and that shows two clear regimes. At temperatures below 200 K the process is limited by low vibronic overlap between occupied states of the and energy surfaces, leading to slow hole capture at and a recombination coefficient that is strongly temperature dependent. At higher temperatures the small makes electron capture at the the limiting process, and has a reduced dependence on temperature (Figure 1f).
As in the single cation material, we must also consider hole emission from the localised state associated with . We find that that the rate of hole emission exceeds the rate of electron capture across the whole temperature range (Figure 1g). Furthermore, in contrast to the single cation material, hole trapping is also fast with at 300 K; an increase by over eight orders of magnitude compared to single cation . We conclude that will form a site for successive hole trapping and de-trapping in this system.
Steric engineering of point defect properties
The increase in after cation mixing can be rationalised using the concept of reorganisation energy from electron transfer theory. 41,42 As shown in Figure 2a, is the free energy change associated with the relaxation between equilibrium geometries:
Note that by definition is always negative. and are related through the expression , 37 where is the Franck-Condon absorption energy for a vertical transition with no change in lattice geometry:

Figure 2. Relationship between defect properties for the series of materials. (a) Schematic showing key defect properties: reorganisation energy , absorption energy and charge transition level (b) Relationship between reorganisation energy and . The purple circle, green triangle, orange cross and blue square denote and compositions respectively. Values for the single cation phase are taken from Reference.26 The upper dash line is a first-order polynomial fit to the mixed cation data, the lower dash line is a fit to the single cation data. The dot line is a fit to all scatter points. (c) Relationship between charge transition level and . The dash line is a fit to the single cation data. The dot line is a fit to all scatter points. The inset shows energies calculated with the pbesol exchange-correlation functional (empty scatter points) and the hybrid HSE06 functional with spin-orbit coupling (filled scatter points).
For relaxation after electron capture in , so that there is a thermodynamic penalty for re-forming the neutral charge state. After Cs incorporation there is a reduction in and, as a result, the magnitude of is also reduced. This leads to an increase in so that the two charge states become close to thermodynamic equilibrium.
Reorganisation energy is a key parameter that controls charge mobility in organic electronics and the efficiency of organic photovoltaic devices. 43 Across a wide range of molecular systems there a correlation between the charge transfer reaction coordinate and ; a smaller geometric relaxation is associated with a smaller reorganisation energy. This relationship enables charge transport engineering, whereby the size of groups within a molecule and/or molecular packing density is used to determine the extent of lattice relaxation and, following this, .10,44
In order to evaluate the extension of steric engineering to lead halide perovskites we begin by investigating the relationship between and for carrier capture at the iodine interstitial in across a range of stoichiometries ( and ). At higher stoichiometries () we found that the valence band maximum increases so that there is no localised defect state for hole capture within the band gap, in agreement with previous reports for .45
In Figure 2b we show that after Cs incorporation is reduced relative to the single cation system ( phase) for all compositions. Note that to allow comparison between different A-site compositions we scale by where is the volume of the defect supercell and is the volume of the control supercell ( in the phase). We find that there is a strong linear correlation between and for the mixed A-cation systems (upper dash line). This analysis includes the point , as required by the definition of reorganisation energy.
We combine our results with data from the literature at the same level of theory for the single cation system.26 This enables us to consider three lattice relaxation processes in , each corresponding to a different regime of lattice relaxation: i) a vertical transition (no relaxation, ); ii) relaxation in the phase (large relaxation, ); and iii) relaxation in the phase (giant relaxation, ). Figure 2b confirms a linear correlation between and for the single cation systems (lower dash line).
Across all data points (single and mixed cation materials) we find a linear correlation (dot line). We also note that for all materials is in the range 1 eV–2 eV, which is significantly larger than that typically found for rigid organic molecules (0.1 eV–0.2 eV)43 or point defects in 2D materials ( eV).46
To understand how may be used to tune the the charge state and activity of we investigate the relationship between and . In Figure 2c we show a negative correlation between and (dot line), indicating that can be tuned through variation of . As in Figure 2a, if we confine our analysis to a single composition and value, the trend becomes stronger (dash line).
The inset in Figure 2c compares values calculated using the semilocal pbesol exchange-correlation functional without spin-orbit coupling (empty scatter points), and the HSE06 exchange correlation functional with spin-orbit coupling (HSE06-SOC, filled scatter points). Across all materials we observe a systematic increase in at the higher HSE06-SOC level of theory. This is due to a shift in the the predicted electronic band edge, as has been observed in previous studies.⁵ Our results suggest that once the change in for a single cation system is known, this can be applied as a correction term to lower accuracy predictions for related mixed cation systems.
Figure 2c provides insight into the significant 0.7 eV discrepancy between previously published values for .⁸,²⁶ Our results identify that the source of this discrepancy is the perovskite phase used for modelling, and the influence this has on the predicted : whilst Reference⁸ uses the high symmetry pseudo-cubic phase, Reference²⁶ uses the lower symmetry orthorhombic phase, which is formed from condensation of the and phonon modes associated with octahedral tilting. If a subset of the distortions associated with lattice relaxation in the cubic phase is not available in the orthorhombic phase, this will lead to the observed reduction in and increase in .
Our analysis suggests that a symmetry lowering mechanism might be responsible for the reduction observed after cation mixing. To test this hypothesis we use symmetry mode analysis.⁴⁷ This allows us to decompose the structural changes after A-cation mixing into the normal phonon modes of the single cation cubic structure, giving insight into the specific atomic displacements that are contributing to any lattice distortion. Figure 3a shows that for all three mixed cation systems the most dominant phonon modes is . This is a polar zone centre displacement corresponding to an off-centering of Pb and I. The second most dominant phonon modes is . This is a zone boundary mode and a dominant component of thermal phase transitions in perovskite materials.⁴⁸ It corresponds to rigid rotations of the PbI octahedra, and consequently describes motions of I only. The amplitudes for all phonon modes are given in Figures S3–S5.

Figure 3. Symmetry mode analysis of MACsPbI with the cubic perovskite phase used as a reference (parent) structure. a) Amplitude of the two most dominant phonon modes and . Green (Pb) and orange (I) are used to indicate the contribution from each atomic species; the A-site species is not displaced for either mode. b) Schematic of the atomic displacements that correspond to each phonon mode. For easier visualisation of the octahedral tilt patterns the iodine atoms are not shown.
Inspection of the polarisation vector shows that this mode is primarily a rigid shift relative to the A-site cation and so does not lead to a reduction in volume. In contrast, mapping along the mode reduces the metal-halide-metal bonding angle and the cubo-octahedral volume around the A-site. This is in agreement with our measured volumes and Pb-I-Pb bond angle after Cs incorporation (Table S2). We conclude from this that the reduced steric size of the Cs cation (1.81 Å) compared to MA (2.70 Å) leads to a volume contraction primarily mediated via condensation of the octahedral tilting mode.
Group theoretical analysis confirms that displacement along reduces the symmetry from cubic to tetragonal (), which is in agreement with the structures reported for FA/Cs and FA/MA/Cs mixed cation systems 49 50. These experimental observations of mixed A-cation perovskites in the tetragonal phase suggest two important points. Firstly, that this symmetry lowering behaviour is common across other systems where cation mixing leads to a reduction in unit cell volume. Secondly, that any dynamic disorder leading to an effective pseudo-cubic phase is suppressed; the octahedral rotations are ‘locked-in’ and do not time-average to a higher symmetry structure. The latter point is supported by molecular dynamics simulations which show that a low concentration of Cs or Rb in suppresses octahedral tilting. 51,52
Discussion
Our calculations show that Cs incorporation leads to a significant increase in the rate of non-radiative trapping and de-trapping at the iodine interstitial. At first this may appear to contradict the well-established improvement in performance for mixed cation perovskite materials, which are used in cells with the highest power conversion efficiencies (PCEs). [3] [53] However for high-performance PV materials, both high PCE and good stability (chemical, thermal and mechanical) are a prerequisite. Figure 1 suggests that the high PCEs for mixed A-cation perovskites do not derive from a reduced rate of non-radiative trapping or recombination in the bulk (in-grain) material as-synthesised, but follows from other well-documented factors including increased (photo-)stability [2, 3] [54, 55] and phase purity. 56,57
Our results are in agreement the relatively limited experimental characterisation of mixed MA/Cs perovskite materials. Microwave photoconductivity decay measurements show that Cs-Br incorporation reduces carrier lifetime in (, 0.1 and 0.15) perovskites films immediately after deposition. 58 The films were then kept in ambient conditions and the carrier lifetime was measured hourly for nine hours; whilst the carrier lifetime of films decreased over this time period, the carrier lifetime of the Cs/MA compounds increased. In a separate study, powder x-ray diffraction shows that incorporation of Cs in leads to a reduction in unit cell volume and increased tilting of the inorganic cage. 59 Time-resolved photoluminescence measurements confirm that there is a decrease in the non-radiative recombination lifetime for the MA-rich compounds (, 0.4) compared to the pure MA compound. For increasing Cs content () the non-radiative lifetime is restored, suggesting that the dominant defect-mediated recombination channel has been removed. This supports our calculations showing that there is no active hole trapping state at higher Cs proportion ().

Figure 4. Schematic illustration outlining the impact of steric engineering and crystal phase on I defect activity. is the neutral to negative charge transition level, is a measure of lattice relaxation between the equilibrium geometries of I and I. The control case of single cation MAPbI in the pseudo-cubic phase shows the largest lattice relaxation after carrier capture, leading to an electrically inactive defect (orange illustrations). A-site mixing with a smaller cation reduces lattice relaxation () and the hole capture barrier (), leading to fast hole trapping and de-trapping (blue). Full symmetry lowering to the orthorhombic phase leads to further reductions in and , and a charge transition level that is in the electronic band gap (green).
Our results demonstrate that CTLs are highly sensitive to , and that the same defect in the same material can show a wide variety of defect behaviour. As shown schematically in Figure 4, the iodine interstitial can vary from being electrically inactive (with in the valence band) through being a site for non-radiative hole trapping and de-trapping (with ), to being a site for non-radiative recombination (with towards the middle of the band gap). This conclusion is supported by time resolved photoluminescence measurements of MAPbI which show that regions with greater compressive strain are associated with increased non-radiative decay. 60 In addition, first-principles molecular dynamics studies demonstrate that the Kohn-Sham eigenvalues associated with defect states in hybrid halide perovskites can oscillate by as much as 1 eV, reinforcing our finding that defects in this system have an unusually high level of sensitivity to lattice distortions. 38–40
Our results also reveal the possibility of tuning defect properties through control of , which is a new approach to defect engineering for hybrid and inorganic materials. Heterovalent doping to adjust the Fermi level is firmly established, 61 as is adjusting the chemical potentials of reactants during growth to increase the formation energy (and thus decrease the concentration) of harmful defects. 62 Instead, we propose isovalent doping (in this case, at the perovskite A-site) to determine the available lattice relaxation pathways (), which in turn adjusts the CTL and carrier capture barriers. This approach is enabled by a material system which shows unusually large lattice relaxation and associated re-organisation energies. In this study we find that the doping-induced volume contraction leads to a symmetry-lowering distortion which ‘locks in’ the accepting phonon modes that are active in non-radiative carrier recombination. However there are other mechanisms beyond this which could used to engineer the available lattice relaxation pathways; for example, applied hydrostatic pressure and epitaxial growth have been shown to induce phase transformations in perovskite materials.63,64
Conclusions
We have analysed non-radiative carrier trapping in single and mixed A-cation perovskite systems. We have introduced an interpolation method for hybrid materials that describes the coupling between electronic states and molecular rotations, and applied this to for accurate predictions of defect activity. We find that cation mixing leads to a significant increase in the rate of hole trapping and de-trapping compared to the single cation system in the pseudo-cubic phase. Importantly, we find a linear relationship between and , demonstrating that the same defect in the same material can display a wide range of defect activity depending on e.g. film morphology. The reduction in is associated with a symmetry-lowering distortion; in the case of mixed A-site cations this is induced through the decreased steric size of Cs compared to MA. Furthermore, our results suggest that can be used as an additional parameter to tune carrier trapping activity in crystalline materials. This is a general approach for materials that display large lattice relaxation after carrier capture, and may be extended to other hybrid organic-inorganic materials including metal-organic frameworks.
Acknowledgement
The author thanks Marc Etherington, Séan Kavanagh, Jarvist Frost and Aron Walsh for scientific discussions. This work used the Oswald High Performance Computing facility operated by Northumbria University (UK). Via our membership of the UK’s HEC Materials Chemistry Consortium, which is funded by EPSRC (EP/R029431), this work used the ARCHER2 UK National Supercomputing Service (http://archer2.ac.uk).
Data Availability Statement
The custom analysis code and data used to generate the figures and tables in this work are openly available at https://dx.doi.org/10.5281/zenodo.7689415. Input and output files for the the HSE06-SoC total energy calculations are openly available at https://dx.doi.org/10.17172/NOMAD/2023.02.27-2.
Supporting Information Available
Supporting Information: DFT calculated energies, charge density distribution, geometric analysis data, electronic density of states, and symmetry mode analysis (PDF)
Graphical TOC Entry

Figure 6 (PDF p. 28).
Steric engineering of point defects in lead halide perovskites: Supplementary Information
DFT calculated energies
In Table 1 we present the DFT calculated energies used in this work. Kabsch interpolation as outlined in the main text was used to generate the intermediate structures. All values are given for calculations using the screened-exchange HSE06 functional with and spin–orbit coupling [3]. For more information please see the methods section in the main text.
| MAPb | MACsPbI | MACsPbI | MACsPbI | |
| -1077.90 | -982.71 | -887.56 | -697.17 | |
| -1080.38 | -984.49 | -889.23 | -698.84 | |
| -1079.55 | -983.42 | -888.00 | -697.68 | |
| -1077.57 | -982.33 | -886.89 | -696.45 | |
| 1.01 | 1.06 | 1.10 | 1.20 | |
| 2.50 | 2.68 | 2.87 | 2.89 | |
| -0.057 | 0.044 | 6E-3 | 0.057 | |
| 5.3E-3 | 4.3E-3 | - | - | |
| 7.2E-3 | 6.2E-3 | - | - | |
| 0.63 | 0.23 | - | - | |
| 0.025 | 0.045 | - | - |
Table 1. DFT calculated energies (in eV) for the materials studied in this work. is the total energy of the pristine (defect-free) supercell. and are the total energies of the defect supercell in the neutral and negative charge states (not including any charged defect corrections), respectively. is the total energy of the negative charge state evaluated at the equilibrium geometry of the neutral charge state. is the image-charge correction for the negatively charged defect. and describe the potential energy surface (PES) curvature for the neutral and negative charge states, respectively. They assume a harmonic PES, which is only valid near equilibrium geometries. and are the energetic barriers for hole capture at and electron capture at respectively.
The charge transition level , reorganisation energy and Franck-Condon vertical transition energy were calculated from the values in the table above, using the expressions given in the main text.
Charge localisation at the H-centre defect

Figure 1 (PDF p. 30). Charge density plot for MACsPbI with a neutral iodine interstitial defect in the H-centre configuration (). The plot has been visualised using vesta with an isosurface level of 3.8E-4 [5]. For clarity only part of the full supercell crystal structure is shown.
Analysis of the pristine crystal structures
In Table 2 we present a summary of geometric measures for the MACsPbI materials in this work. All crystal structures were relaxed using DFT and the pbesol exchange-correlation functional; for more information please see the methods section in the main text. Bond angles were measured using ASE [4]. Ellipsoidal properties were calculated using Pieface [1].
| (Å) | (Å) | (deg.) | ||
| 0 | 250.1 | 132.7 | 168.2 (±2.8) | 0.0047 |
| 0.125 | 245.2 | 134.0 (±0.25) | 161.9 (±8.1) | 0.0042 (±0.0081) |
| 0.25 | 240.3 | 135.1 (±0.47) | 156.2 (±7.7) | -0.0046 (±0.0274) |
| 0.5 | 239.8 | 134.1 (±0.66) | 157.6 (±9.0) | 0.0050 (±0.0168) |
Table 2. Geometric measures for MACsPbI materials. The first column specifies the proportion of Cs on the A-site. The second column gives the total volume per formula unit. The third column specifies the mean average volume of the 16 minimum bounding ellipsoid fitted to each inorganic PbI octahedra. The fourth column specifies the mean average of the 48 Pb-I-Pb bond angles. The fifth column specifies the mean average shape factor of the 16 minimum bounding ellipsoids, where a perfect sphere has a shape factor of zero; for more details please see Reference [1]. Standard deviations are given in brackets. In the case of there is only one ellipsoid and three Pb-I-Pb bond angles to consider.
Analysis of lattice distortions after electron capture at
In Table 3 we present a summary of the lattice distortions induced after charge capture at an iodine interstitial in materials. All crystal structures were relaxed using DFT and the pbesol exchange-correlation functional; for more information please see the methods section in the main text. Bond angles and bond lengths were measured using ASE [4].
| 0 | 36.77 | 35.16 | 10.75 | 6.03 (7.03) | 0.054 (0.103) |
| 0.125 | 25.94 | 25.09 | 6.612 | 4.78 (5.01) | 0.041 (0.080) |
| 0.25 | 26.30 | 23.17 | 12.46 | 3.61 (4.74) | 0.035 (0.070) |
| 0.5 | 28.96 | 26.62 | 11.42 | 3.77 (3.63) | 0.028 (0.030) |
Table 3. A summary of lattice distortions after charge capture at an iodine interstitial in materials, calculated using DFT. The first column specifies the proportion of Cs on the A-site. Columns 2–4 give values calculated for all atomic species, the inorganic Pb-I framework and the A-site species. Column 5 gives the mean average absolute change of the 48 Pb-I-Pb bond angles after charge capture. Column 6 gives the mean average absolute change of the 48 Pb-I bond lengths after charge capture. In the case of there are only three Pb-I bond lengths and three Pb-I-Pb bond angles to consider.
Electronic density of states

Figure 2 (PDF p. 31). Electronic partial density of states analysis for , calculated using DFT. Electronic bands from a supercell calculation were unfolded using PyProcar [2].
Symmetry Mode Analysis
The mode amplitudes are given in Å, and all phonon modes with a displacive mode amplitude above 0.1Å are shown. The reference (parent) structure is perovskite in the cubic phase. The rotational motion of the MA molecule was not considered in this analysis; all A-sites were modelled as point particles. This mode amplitudes were analysed using Isodistort [6].

Figure 3 (PDF p. 32). Symmetry mode analysis for MACsPbI.

Figure 4 (PDF p. 33). Symmetry mode analysis for .

Figure 5. Symmetry mode analysis for .
References
- [1]CUMBY, J., AND ATTFIELD, J. P. Ellipsoidal analysis of coordination polyhedra. Nature Communications 8, 1 (Feb 2017), 14235.DOI
- [2]HERATH, U., TAVADZE, P., HE, X., BOUSQUET, E., SINGH, S., MUÑOZ, F., AND ROMERO, A. H. Pyrocar: A python library for electronic structure pre/post-processing. Computer Physics Communications 251 (2020), 107080.
- [3]HEYD, J., SCUSERIA, G. E., AND ERNZERHOF, M. Hybrid functionals based on a screened coulomb potential. The Journal of Chemical Physics 118, 18 (2003), 8207–8215.DOI
- [4]LARSEN, A. H., MORTENSEN, J. J., BLOMQVIST, J., CASTELLI, I. E., CHRISTENSEN, R., DULAK, M., FRIIS, J., GROVES, M. N., HAMMER, B., HARGUS, C., HERMES, E. D., JENNINGS, P. C., JENSEN, P. B., KERMODE, J. J., KITCHIN, J. R., KOLSBJERG, E. L., KUBAL, J., KAASBJERG, K., LYSGAARD, S., MARONSSON, J. B., MAXSON, T., OLSEN, T., PASTEWKA, L., PETERSON, A., ROSTGAARD, C., SCHIØTZ, J., SCHÜTT, O., STRANGE, M., THYGESEN, K. S., VEGE, T., VILHELMSEN, L., WALTER, M., ZENG, Z., AND JACOBSEN, K. W. The atomic simulation environment—a python library for working with atoms. Journal of Physics: Condensed Matter 29, 27 (2017), 273002.DOI
- [5]MOMMA, K., AND IZUMI, F. VESTA3 for three-dimensional visualization of crystal, volumetric and morphology data. Journal of Applied Crystallography 44, 6 (Dec 2011), 1272–1276.DOI
- [6]STOKES, H. T., HATCH, D. M., CAMPBELL, B. J., AND TANNER, D. E. ISODISPLACE: a web-based tool for exploring structural distortions. Journal of Applied Crystallography 39, 4 (2006), 607–614.DOI