WorksPhysics and Astronomy
Optoelectronic Properties of Chalcogenide Perovskites by Many-Body Perturbation Theory
Abstract
Chalcogenide perovskites have emerged as non-toxic and stable photovoltaic materials, acting as an alternative to lead halide hybrid perovskites having similar optoelectronic properties. In the present work, we report the electronic and optical properties of chalcogenide perovskites AZrS (A=Ca, Sr, Ba) by using the density functional theory (DFT) and many-body perturbation theory (MBPT viz. GW and BSE). This study includes excitonic analysis for the aforementioned systems. The exciton binding energy (E) is found to be larger than that of the halide perovskites, as the ionic contribution to dielectric screening is negligible in the former. We also observe a more stable charge-separated polaronic state as compared to that of the bound exciton. Finally, on the basis of direct gap and absorption coefficient, the estimated spectroscopic limited maximum efficiency (SLME) of the solar cells is large and suggests the applicability of these perovskites in photovoltaics.
Phone: +91-11-2659 1359
Graphical TOC Entry

Figure 1.
Inorganic-organic (IO) hybrid halide perovskites have emerged as an efficient compound semiconductors alternative to conventional materials used in photovoltaics. [1] [2, 3, 4] The power conversion efficiency (PCE) of solar cell based on IO hybrid perovskites has increased from 3.8% to 25.5% in the last decade. [1] [4] Nevertheless, the concerns regarding long term stability and toxicity of lead restrict their commercialization. 5,6 In pursuit of alternative materials having similar kind of optoelectronic properties, chalcogenide perovskites have been investigated in last few years. Sun Sun et al.⁷ first theoretically reported the optoelectronic properties of chalcogenide perovskites, which can be utilized in solar cells. Subsequently, other theoretical studies have also characterized different chalcogenide perovskites for high solar cell efficiency and photoelectrochemical water splitting. 8–12 Many chalcogenide perovskites such as (, , ), have been synthesized experimentally as well, 8,13–19 which are stable and consist of earth-abundant non-toxic elements. These chalcogenide perovskites have high optical absorption, optimal photoluminescence and good charge carrier mobility, which suggest the possibility of their usage in various optoelectronic devices. 16,20–22
Zr-based chalcogenide perovskites contain the d-orbital character, wherein the states are less localized than , resulting in large absorption coefficient and small effective mass of the charge carriers in these compounds. 7 Therefore, many experimental and theoretical studies have been performed on Zr-based chalcogenide perovskites (, where is alkaline earth metal).
Note that, the charge separation in solar cell gets hugely influenced by formation of excitons. Therefore, the operation mechanism of a solar cell highly depends on it as these excitons thermally dissociate into free electrons and holes, giving rise to the required free-charge transport. However, until date, due to huge computational cost, any detailed study with adequate accuracy of the excitonic properties is not very well known. Therefore, it is of profound interest to employ advanced theoretical methodologies for accurate understanding of the excitonic properties that will sufficiently correlate with the experimental studies to disentangle the scientific insights of excitons. Despite several theoretical studies on the chalcogenide perovskites, investigation of the optical properties using the excited-state methods remain unexplored. In view of this, presumably for the first time, we have reported the excitonic properties of the chalcogenide perovskites.
In this Letter, we have done a systematic study of electronic and optical properties of chalcogenide perovskites (A=Ca, Sr, Ba) using ground- and excited-state methods. First, we have employed density functional theory (DFT)23,24 with semi-local PBE25 exchange-correlation () functional to optimize the crystal structures. In order to study electronic structure, we have calculated atom-projected electronic partial density of states (pDOS) using hybrid functional HSE06.26 Subsequently, we have determined the optical properties using excited-state method viz. many-body perturbation theory (MBPT). The Bethe-Salpeter equation (BSE)27,28 has been solved to get the electronic contribution to dielectric function on top of single-shot @PBE.29,30 Further to investigate the ionic contribution to dielectric function, density functional perturbation theory (DFPT) has been used. Finally, using the quasiparticle (QP) band gap and optical properties, the maximum theoretical photoconversion efficiency has been determined by calculating the spectroscopic limited maximum efficiency (SLME)31 metric.
Here, we have considered the distorted orthorhombic phase of chalcogenide perovskites (A=Ca, Sr, Ba) having the space group 13 (see Figure 1a). In addition, we have considered the needle-like phase of (see Figure 1b), since the same has been found in two crystallized phases (–phase, which is needle-like and -phase, which is distorted-perovskite phase).14 The lattice parameters of the optimized structures calculated using PBE functional are given in Table 1. These are in close agreement with previous experimental results.13,14 We have calculated the electronic pDOS of the aforementioned configurations using HSE06 functional as shown in Figure 1[c-f]. The valence band maximum (VBM) is mostly contributed by S 3p-orbitals, whereas the conduction band minimum (CBm) is mainly from Zr 4d-orbitals. Rest of the orbitals have small contribution at VBM and CBm. As we change the cation from Ca to Ba, the peaks in pDOS become more narrow and sharper at the CBm (see Figure 1c, 1d and 1f), which indicates that electronic nonradiative lifetime becomes shorter as we go down the group. This is due to the fact that narrow peaks in the pDOS signify large number of carrier relaxation path, and hence, shorter carrier lifetimes. 32,33
| Configurations | (Å) | (Å) | (Å) |
| CaZrS₃ | 6.56 (6.54) | 7.06 (7.03) | 9.63 (9.59) |
| −SrZrS₃ | 3.84 (3.83) | 8.63 (8.53) | 13.99 (13.92) |
| −SrZrS₃ | 6.78 (6.74) | 7.16 (7.11) | 9.82 (9.77) |
| BaZrS₃ | 7.03 (7.03) | 7.16 (7.06) | 10.01 (9.98) |
Table 1. Calculated lattice parameters of AZrS₃ (A=Ca, Sr, Ba) perovskites. The experimental values are provided in brackets. For distorted perovskites viz. CaZrS₃, −SrZrS₃, and BaZrS₃, the experimental values are from Ref.13 For −SrZrS₃, experimental values are from Ref.14
−SrZrS₃ has the narrowest and sharpest peak in the pDOS at CBm (see Figure 1e). Further, with the change in A-cation species, there is shift in CBm, which has altered the band gap. Moreover, as we go down the group from Ca to Ba, the bands become more dispersive at the CBm (see Figure S1). Hence, the effective mass of the electron decreases (see Table 2). Contrastingly, the effective mass of hole is not affected much. We find −SrZrS₃ to have the smallest electron and hole effective masses. Therefore, these perovskites are expected to have better charge carrier transport as indicated by the smaller values of effective masses.
| Configurations | |||
| CaZrS₃ | 0.503 | 0.588 | 0.271 |
| −SrZrS₃ | 0.323 | 0.545 | 0.203 |
| −SrZrS₃ | 0.440 | 0.580 | 0.250 |
| BaZrS₃ | 0.411 | 0.587 | 0.242 |
Table 2. Effective mass of electron, hole and reduced mass (in terms of free-electron mass ) of chalcogenide perovskites along −Z high symmetry path

Figure S1. Electronic band structure of (a) CaZrS₃, (b) α-SrZrS₃, (c) β-SrZrS₃, and (d) BaZrS₃ using PBE functional.
All the considered chalcogenide perovskites exhibit direct band gap at high symmetry point, which is the desired property for effective absorption. The band gaps are underestimated by semi-local functional PBE (see Table 3), which is due to the well known self-interaction error. The hybrid functional HSE06 corrects the band gaps that are well in agreement with the experimental values (see Table 3). Therefore, the electronic structure can be well described by the HSE06. However, the latter is less accurate in predicting the optical features of the systems.34 Therefore, the MBPT based GW-BSE method has been used to compute the optical response, which explicitly considers the electron-hole interaction.35 Since the single shot GW () calculation depends on its starting point, we have
| Configurations | PBE | HSE06 | @PBE | Experimental |
| CaZrS | 1.24 | 2.04 | 2.29 | 1.9015 |
| −SrZrS | 0.60 | 1.40 | 1.60 | 1.5216 |
| −SrZrS | 1.22 | 2.05 | 2.32 | 2.0516 |
| BaZrS | 1.06 | 1.87 | 2.10 | 1.8316 |
Table 3. Band gap (in eV) of chalcogenide perovskites

Figure 1 (PDF p. 6). Schematic crystal structure of orthorhombic (a) AZrS (A=Ca, Sr, Ba) in distorted phase and (b) −SrZrS in needle-like phase. Electronic partial density of states (pDOS) of (c) CaZrS, (d) −SrZrS, (e) −SrZrS, and (f) BaZrS using HSE06 functional.
validated it by calculating the optical response of BaZrS, i.e., the imaginary part of the complex dielectric function (see Figure 2). The first peak represents the optical transition corresponding to the band gap. The peak position, which is underestimated by PBE (1.06 eV), is improved by GW on top of both the PBE and HSE06, which are at 2.10 and 2.32 eV, respectively (see Figure 2). Notably, the QP gaps computed using GW are overestimated in comparison to experimental band gap, since it does not take into account the exciton binding energy. Further, the gaps are improved by solving the BSE. The peak positions obtained using BSE@GW@PBE and BSE@GW@HSE06 are at 1.88 and 2.02 eV, respectively (see Figure 2). The former is in close agreement with the experimental band gap, whereas the latter is overestimated. Hence, the PBE is accurate than the HSE06 as a starting point for calculating the optical properties using MBPT approach. The QP gaps of AZrS (A=Ca, Sr, Ba) calculated using GW@PBE are provided in Table 3. Figure 3 shows the imaginary part of the dielectric function [Im ()] and oscillator strength calculated using BSE@GW@PBE. In the same figure, Im () calculated using GW@PBE is also shown. The exciton binding energy (E) can be computed from this figure, as the E is the difference between QP band gap (GW@PBE peak position) and optical band gap (BSE@GW@PBE peak position).

Figure 2. Imaginary [Im ()] part of the dielectric function for BaZrS with light polarization perpendicular to c-axis (), obtained using different level of theories viz. PBE, HSE06, GW@PBE, GW@HSE06, BSE@GW@PBE and BSE@GW@HSE06. First peak corresponds to the band gap of BaZrS.

Figure 3. Spatially averaged imaginary part of the dielectric function for (a) CaZrS₃, (b) −SrZrS₃, (c) −SrZrS₃, and (d) BaZrS₃ obtained using G₀W₀@PBE and BSE@G₀W₀@PBE. Peaks with turquoise color represent the oscillator strength.
Hence, from Figure 3, the of the first bright exciton for CaZrS₃, −SrZrS₃, −SrZrS₃ and BaZrS₃ are 0.23, 0.54, 0.25 and 0.21 eV, respectively. As per the BSE eigenvalue analysis, we have found that a dark exciton (optically inactive) also exists in case of BaZrS₃ below the bright exciton. Moreover, several dark excitons exist in case of −SrZrS₃. The for the lowest energetic dark exciton is 1.53 and 0.22 eV for −SrZrS₃ and BaZrS₃, respectively. Furthermore, the oscillator strength for all the considered perovskites is mainly distributed within the spectral window of 2-4 eV and matches well with the excitonic peak positions (see Figure 3). It signifies the high recombination between electron and hole in the considered energy range. Moreover, using the exciton binding energy, band gap, dielectric function and reduced mass, several excitonic parameters can be determined 36 such as excitonic temperature () and radius () given in Table 4. The lifetime () is inversely related to the probability of wavefunction () for electron-hole pair at zero separation (for further details of excitonic parameters relations, see Section VIII in SI). Therefore, the for the considered perovskites are in the order: -SrZrS > BaZrS > CaZrS > -SrZrS.
| Excitonic parameters | CaZrS | -SrZrS | -SrZrS | BaZrS |
| (eV) | 0.23 | 0.54 | 0.25 | 0.21 |
| (K) | 2669 | 6267 | 2901 | 2437 |
| (nm) | 0.79 | 0.93 | 0.73 | 0.92 |
| 0.65 | 0.40 | 0.82 | 0.41 |
Table 4. Excitonic parameters for chalcogenide perovskites
The high in comparison to halide perovskite37 can be understood from the ionic contribution to dielectric screening. It has been recently shown that if the calculated using vertical transition is much greater than the energy of the longitudinal optical phonon mode (), then the ionic contribution to dielectric screening is negligible and hence, does not alter the .38 In case of chalcogenide perovskites, , which can be seen from the ionic contribution to dielectric function (see Figure 4). Therefore, the lowering of by the ionic screening can be excluded. Further, by employing Wannier-Mott approach as well, we have calculated the . According to this model, the is related to reduced mass of the charge carriers () and effective dielectric constant () as follows:

Figure 4. Ionic contribution to dielectric function for (a) CaZrS, (b) –SrZrS, (c) –SrZrS, and (d) BaZrS obtained using DFPT.
where, is the Rydberg constant. Here, the lies in between the static value of dielectric constants as contributed by electrons and ions. The electronic and ionic static dielectric constants provide the upper and lower bounds to the exciton binding energy. For CaZrS, -SrZrS, -SrZrS and BaZrS, the electronic static dielectric constants are 4.06, 3.55, 3.44 and 4.19, respectively, which are calculated using BSE. The respective ionic static dielectric constants are 57.07, 20.86, 74.51, and 99.74 calculated using DFPT (see Figure 4). For BaZrS, the calculated value of ionic static dielectric constant is in close agreement with previous experimental results.39 From the reduced mass (provided in Table 2), and the static dielectric constants, we have determined the upper and lower bounds of using Equation (1), which are listed in Table 5 (also shown in Figure S2). The upper bounds are in good agreement with the calculated by taking the difference of GW and BSE peak positions, except for –SrZrS. Thus, the electronic contribution is more prominent than ionic contribution in dielectric screening for chalcogenide perovskites.
| Configurations | Upper bound (eV) | Lower bound (meV) |
| CaZrS | 0.22 | 1.13 |
| –SrZrS | 0.22 | 6.35 |
| –SrZrS | 0.29 | 0.61 |
| BaZrS | 0.19 | 0.33 |
Table 5. Upper and lower bounds on exciton binding energy for chalcogenide perovskites

Figure S2. Exciton binding energy () of (a) CaZrS₃, (b) −SrZrS₃, (c) −SrZrS₃, and (d) BaZrS₃, as a function of dielectric constant. The intersection of the curve with vertical dashed line defines the upper bound obtained using electronic static dielectric function and horizontal dashed line defines the lower bound obtained by ionic static dielectric function.
Further, we have determined the electron-phonon coupling using the Fröhlich model.40,41 In this model, the electron moving through the lattice interact with the polar optical phonons via Fröhlich parameter , given by
where and are the electronic and ionic static dielectric constants, respectively. is Planck’s constant and is the speed of light. The characteristic frequency is determined from the multiple phonon branches using athermal ‘B’ scheme of Hellwarth et al. 42 The calculated values of are provided in Table 6. From these, the reduction in QP gap can be determined.38 For CaZrS, –SrZrS, –SrZrS and BaZrS, the QP gap is lowered by 0.24, 0.23, 0.33 and 0.27 eV, respectively. On comparing these values with , we infer that except for –SrZrS, the charge-separated polaronic state is more stable than the bound exciton. Furthermore, the computed polaron mobilities are given in Table S3 (see section IX in SI).
| Configurations | (cm) | ||
| CaZrS | 142.03 | 4.51 | 8.29 |
| –SrZrS | 152.74 | 3.56 | 8.49 |
| –SrZrS | 117.86 | 5.61 | 11.1 |
| BaZrS | 107.51 | 4.68 | 9.53 |
Table 6. Electron-phonon coupling parameters for chalcogenide perovskites
| Configurations | (cm/Vs) | (cm/Vs) |
| CaZrS | 8.16 | 5.96 |
| −SrZrS | 18.77 | 6.81 |
| −SrZrS | 6.84 | 3.76 |
| BaZrS | 11.35 | 5.58 |
Table S3. Polaron mobilities () of CaZrS, −SrZrS, −SrZrS, and BaZrS at K
We have observed that, the AZrS () perovskites exhibit large absorption coefficient and direct band gap in visible region. These two make them interesting materials for photovoltaic applications. Therefore, we have calculated the spectroscopic limited maximum efficiency (SLME),31,43 which has been proved to be a good metric to determine the maximum efficiency that an absorber material can reach in a single-junction solar cell. SLME is an improved version of Shockley and Queisser (SQ) efficiency,44 as it takes into account the nature of band gap, the shape of absorption spectra and the material-dependent nonradiative recombination losses, in addition to the band gap. The standard solar spectrum, material’s band gap and the absorption coefficient are given as input for the SLME calculation. Figure 5 shows the calculated SLME of (A=Ca, Sr, Ba). Except for -, the SLME becomes constant for a layer thickness greater than . - requires a thicker absorption layer () for maximum efficiency. This is due to the fact that absorption onset in case of - is at larger value than the band gap. SLME values at absorber layer thickness are 21.33%, 25.45%, 21.19%, and 25.02% for , -, - and , respectively (see Figure 5). The theoretically predicted SLME of (25.02%) is in good agreement with previously reported theoretical efficiency (% at thickness).8 These values of SLME are encouraging for their photovoltaic applications.

Figure 5. Spectroscopic limited maximum efficiency of AZrS3 (A=Ca, Sr, and Ba).
In conclusion, we have determined the electronic and optical properties of chalcogenide perovskites (A = Ca, Sr, Ba) by state-of-the-art ground- and excited-state methods. The effective mass of electron has been observed as decreasing down the group from Ca to Sr, thereby enhancing the charge carrier transport. Subsequently, the optical band gap is well reproduced by solving the Bethe-Salpeter equation (BSE). The exciton binding energies for , -, - and are computed as 0.23, 0.54, 0.25 and 0.21 eV, respectively. In addition, by calculating the electron-phonon coupling parameters, we have observed that the charge-separated polaronic state is more stable than the bound exciton. We also report negligible ionic contribution to the effective dielectric screening that determines the exciton binding energy. Finally, the calculated spectroscopic limited maximum efficiency (SLME) suggests their usage in photovoltaics.
Computational Methods
The density functional theory (DFT) 23,24 calculations have been performed as implemented in the Vienna ab initio simulation package (VASP). 45,46 The ion-electron interactions in all the elemental constituents are described using projector-augmented wave (PAW) potentials.46,47 All the structures are optimized using generalized gradient approximation viz. PBE25 exchange-correlation () functional until the forces are smaller than 0.001 eV/Å. The electronic self consistency loop convergence is set to 0.001 meV, and the kinetic energy cutoff is set to 500 eV for plane wave basis set expansion. A -grid of is used for Brillouin zone integration, which is generated using Monkhorst-Pack 48 scheme. The effective mass has been calculated by SUMO49 using parabolic fitting of the band edges. Advanced hybrid functional HSE0626 is used for the better estimation of band gap. Note that spin-orbit coupling (SOC) has not been taken into account because it negligibly affects the electronic structure of considered chalcogenide perovskites (see Table S1). In order to determine optical properties, Bethe-Salpeter equation (BSE) 27,28 has been solved on top of single shot GW29,30 () calculations. The initial step for calculation is performed by the PBE functional. The polarizability calculations are carried out on a grid of 50 frequency points. The number of unoccupied bands is set to eight times the number of occupied orbitals (for convergence of empty states, see Table S2). -centered -grid has been used, for BSE calculations. To construct the electron-hole kernel for BSE calculations, 24 occupied and 24 unoccupied states have been used. The convergence for the same is shown in Figure S3. To check the convergence with respect to Brillouin zone sampling, model-BSE (mBSE) has been done (see Figure S4). We have found that there is a negligible shift of the lower energy peak on increasing the -grid. Further, we have seen that the chalcogenide perovskites are optically active along all the three directions, signifying minute anisotropy of dielectric function in chalcogenide perovskites (see Figure S5). The ionic contribution to dielectric function has been calculated using density functional perturbation theory (DFPT) with -grid generated using Monkhorst-Pack scheme.
| Configurations | Without SOC | With SOC |
| 1.24 | 1.22 | |
| 0.60 | 0.59 | |
| 1.22 | 1.19 | |
| 1.06 | 1.01 |
Table S1. Band gap (in eV) of , , , and using PBE functional
| Number of bands | CaZrS | –SrZrS | –SrZrS | BaZrS |
| 320 | 2.39 | 1.63 | 2.42 | 2.20 |
| 480 | 2.32 | 1.60 | 2.35 | 2.12 |
| 640 | 2.29 | 1.60 | 2.32 | 2.10 |
| 720 | 2.29 | 1.60 | 2.32 | 2.10 |
| 800 | 2.30 | 1.61 | 2.32 | 2.09 |
Table S2. Band gap (in eV) of CaZrS, –SrZrS, –SrZrS, and BaZrS using @PBE with different number of bands

Figure S3. Imaginary part of electronic dielectric function with light polarization perpendicular to c-axis () for BaZrS with different number of valence (NO) and conduction bands (NV) used in electron-hole interaction kernel.

Figure S4. (a) Model fitting for model-BSE (mBSE). (b) Spatially average imaginary [Im ()] part of the dielectric function for BaZrS with different -grid using mBSE. Imaginary part using GW-BSE is shown for reference by orange color. Calculated values of inverse of the static ion-clamped dielectric function and the screening length parameter are used in mBSE.

Figure S5. Imaginary part of electronic dielectric function for with light polarization along all three lattice vectors. For other chalcogenide perovskites as well, the anisotropy in dielectric function is negligible.
Acknowledgement
MK acknowledges CSIR, India, for the senior research fellowship [grant no. 09/086(1292)/2017-EMR-I]. AS acknowledges IIT Delhi for the financial support. DG acknowledges UGC, India, for the senior research fellowship [grant no. 1268/(CSIR-UGC NET JUNE 2018)]. SB acknowledges the financial support from SERB under core research grant (grant no. CRG/2019/000647). We acknowledge the High Performance Computing (HPC) facility at IIT Delhi for computational resources.
Supporting Information Available
Electronic band structure of chalcogenide perovskites using PBE functional, exciton binding energy () of chalcogenide perovskites as a function of dielectric constant, effect of spin-orbit coupling (SOC) on band gap, convergence of number of bands used in calculations, convergence of number of valence (NO) and conduction bands (NV) used in electron-hole interaction kernel for BSE calculations, imaginary part of dielectric function for with different number of -grid using model-BSE (mBSE), imaginary part of dielectric function for with light polarization along the three lattice vectors, relations of excitonic parameters with exciton binding energy (), dielectric constant () and reduced mass (), polaron mobility of chalcogenide perovskites.
Phone: +91-11-2659 1359. Fax: +91-11-2658 2037.
Supplemental Material
Electronic band structure of chalcogenide perovskites using PBE functional
Exciton binding energy () of chalcogenide perovskites as a function of dielectric constant
Effect of spin-orbit coupling (SOC) on band gap
Convergence of number of bands used in calculations
Convergence of number of valence (NO) and conduction bands (NV) used in electron-hole interaction kernel for BSE calculations
Imaginary part of dielectric function for BaZrS with different number of -grid using model-BSE (mBSE)
Imaginary part of electronic dielectric function for BaZrS with light polarization along the three lattice vectors
Relations of excitonic parameters with exciton binding energy (), dielectric constant () and reduced mass ()
Polaron mobility of chalcogenide perovskites
Electronic band structure of chalcogenide perovskites using PBE functional
Exciton binding energy () of chalcogenide perovskites as a function of dielectric constant
Effect of spin-orbit coupling (SOC) on band gap
Convergence of number of bands used in calculations
The number of bands used in the calculations are 720.
Convergence of number of valence (NO) and conduction bands (NV) used in electron-hole interaction kernel for BSE calculations
Imaginary part of dielectric function for BaZrS with different number of -grid using model-BSE (mBSE)
In model-BSE (mBSE), the dielectric function is replaced by the model local function given below:
where is the static ion-clamped dielectric function in high frequency limit calculated using GW@PBE. is the screening length parameter, determined by fitting the at small wave vector (q) with respect to and G is the reciprocal lattice vector.
Imaginary part of electronic dielectric function for with light polarization along the three lattice vectors
Relations of excitonic parameters with exciton binding energy (), dielectric constant (), band gap () and reduced mass ()
The exciton temperature () is determined as follows:
where is the exciton binding energy and is the Boltzmann constant. The exciton radius () is calculated as follows:
where is the free electron mass, is the reduced mass, is the static effective dielectric constant (here, the electronic dielectric constant has been taken since the ionic contribution to dielectric screening is negligible), is the exciton energy level (n=1 provides the smallest exciton radius) and nm is the Bohr radius. The excitonic lifetime () is inversely proportional to the probability of wavefunction for electron-hole pair at zero seperation (). The () is determined as follows:
Polaron mobility of chalcogenide perovskites
At temperature , the Hellwarth polaron mobility can be determined from the static dielectric constants, effective mass and optical phonon frequency.[1]
References
(1) Frost, J. M. Calculating polaron mobility in halide perovskites. Phys. Rev. B 2017, 96, 195202.
References
- [1]Frost, J. M. Calculating polaron mobility in halide perovskites. Phys. Rev. B 2017, 96, 195202DOI