REVIEW

www.advtheorysimul.com

Theoretical Progress on the Relationship between the

Structures and Properties of Perovskite Solar Cells
Wei Geng, Chuan-Jia Tong, Yanning Zhang, and Li-Min Liu*

Later, Lee et al. found that solar cells can

Organic–inorganic lead halide perovskite solar cells have attracted maintain excellent efficiency of 10.9% when
great attention due to their high conversion efficiency, easy preparation, electrically conducting TiO2 was replaced
and low cost. In spite of impressive development, some fundamental by insulating Al2 O3 . This finding indicated
scientific issues regarding perovskites have not been resolved, such that perovskite itself might act not only as
a light absorber but also as a good elec-
as the inclusion of toxic Pb, their poor stability under moisture, oxygen, heat,
tron transporter.[3] Jeon et al. used mixed-
and/or light conditions, and their hysteretic current–voltage behavior. In the halide CH3 NH3 Pb(I1-x Brx )3 to achieve a
past few years, theoretical approaches have been extensively and successfully high PCE of 16.2% through solution-
applied to the investigation of perovskite solar cells. Here, the current processing.[4] In 2015, Yang et al. achieved
theoretical progress in perovskite photovoltaic applications is summarized a PCE exceeding 20% with high-quality
HC(NH2 )2 PbI3 films.[5] Most recently, a
from a theoretical perspective, including important theoretical results in
new record of PCE = 25.2% has been set
fundamental structures and properties, computational design of perovskites for the perovskite solar cell (PSC),[6] which
by high throughput calculation and machine learning, hysteresis-related is comparable to their commercialized
ion migration, compositional substitution and mixing, surface reconstruction, counterparts.
mechanisms of degradation and passivation, interface carrier recombination, It is generally accepted that the excep-
and the distribution and rotation of organic cations caused polarization and tionally high PCE achieved with perovskite-
based solar cells result from the per-
domain.
ovskite’s outstanding properties, such as
direct and tunable bandgaps,[7–10] high
light-absorption coefficient,[11–13] ultralong
carrier diffusion length,[14–16] high carrier
1. Introduction mobility, long carrier lifetime, and low
The organic–inorganic lead halide perovskite has attracted great nonradiative recombination rates.[17,18] In addition, photovoltaics
attention as an unprecedented class of semiconductors for opto- devices based on halide perovskites are easy to fabricate with ex-
electronic applications. Organic–inorganic hybrid CH3 NH3 PbX3 tremely low costs.[19,20] Halide perovskites are also promising in
(X = Br, I) perovskites were used as sensitizing materials in liq- other optoelectronic applications such as light-emitting diodes
uid dye-sensitized solar cells by Kojima et al. in 2009, with a low (LEDs),[21–25] photodetectors,[26–28] lasers,[29] nonlinear optics,[30]
power conversion efficiency (PCE) of 3.8%.[1] In 2012, to solve and optospintronics.[31]
the corrosion problems, Kim et al. replaced the liquid electrolytes Multiple reviews and perspectives[32–40] have been devoted to
with solid-state spiro-MeOTAD as the hole transport layer and perovskite photovoltaics about their stability, composition, work-
deposited perovskites on the crystalline TiO2 surface. Unexpect- ing mechanisms, degradation, synthetic approaches, and physi-
edly, the PCE was increased to 9.7%, boosting the researches on cal properties, from both experimental and computational points
perovskite materials and their photovoltaic (PV) applications.[2] of view.[41–48] In the present review, we focus on the progress
in theoretical methods that bridge between the structures and
the unique properties of perovskites materials. In Section 2,
we discuss the fundamental structures and properties of bulk
Dr. W. Geng, Prof. Y. Zhang perovskites, including tolerance factor and stabilities, theoreti-
Institute of Fundamental and Frontier Sciences cal method on band structure, electronic properties and screen-
University of Electronic Science and Technology of China
Chengdu, Sichuan 610054, China
ing of perovskites by high throughput calculation, and machine
Dr. W. Geng, Dr. C.-J. Tong, Prof. L.-M. Liu
learning. In Section 3, defect and doping for perovskites are dis-
School of Physics cussed, for example, hysteresis related ions migration, compo-
Beihang University sitional substitution, and mixing. The Section 4 reviews the in-
Beijing 100191, China terface sciences of perovskites, including surface reconstruction,
E-mail: [email protected] degradation and passivation, interface charge transport, and car-
rier recombination. In Section 5, we describe the distribution and
The ORCID identification number(s) for the author(s) of this article rotation of organic cations as a factor for perovskite performance
can be found under https://doi.org/10.1002/adts.202000022
and stability.
DOI: 10.1002/adts.202000022

Adv. Theory Simul. 2020, 3, 2000022 2000022 (1 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Wei Geng received his Ph.D.

(2013) from Lanzhou Univer-
sity. He then worked as a post-
doctoral fellow in Prof. Li-Min
Liu’s group at Beijing Computa-
tional Science Research Center
from 2013 to 2016 and worked
as a special associate research
fellow at Sun Yat-sen University
from 2016 to 2019. Since Jan-
uary 2020, he has worked at the
Institute of Fundamental and
Frontier Sciences, University
of Electronic Science and Technology of China. His recent
research has focused on photovoltaic materials and nanocat-
alysts.

Li-Min Liu received his Ph.D.

from the Institute of Metal Re-
Figure 1. The schematic crystal structures of a) 𝛼, b) 𝛾, c) 𝛽, and d) 𝛿 search, Chinese Academy of
phases of MAPbI3 .[49] Reproduced with permission. Copyright 2014, John Sciences in 2006. During his
Wiley & Sons, Inc.
Ph.D. study, he visited Queen’s
University of Belfast, UK, for
2. Fundamental Structure Parameters and a year. He then worked in the
Properties of Bulk Perovskites Fritz-Haber-Institut (Ger-
Perovskite, the name of mineral CaTiO3 , generally represents a many), University College
class of compounds with the same crystal structure as CaTiO3 , London (UK), and Princeton
and its chemical formula is written as ABX3 . For a halide per- University. During 2011–2017,
ovskite, A is a relatively large inorganic cation like Cs+ or an or- he took a tenure-track position
ganic cation like methylammonium (CH3 NH3 + /MA+ ) and for- in Beijing Computational Sci-
mamidinium (HC(NH2 )2 + /FA+ ), B is usually a divalent cation ence Research Center. Since September 2017, has been at the
like Pb2+ , and X is halogen anion, such as I− , Br− , and Cl− . In School of Physics, Beihang University. His research interests
ABX3 , each B cation is surrounded by six X anions, forming a focus on nanocatalysts, TiO2-based photocatalysis, electro-
[BX6 ]n− octahedron. The corner-sharing [BX6 ]n− octahedra con- catalysis, and fuel cells.
stitute a 3D framework that consists of alternating ─B─X─B─X─
Yanning Zhang works on den-
bonds along all three directions. At finite temperature, as shown
sity functional theory studies
in Figure 1, transitions between cubic (𝛼 phase), tetragonal (𝛽
and predictions on the struc-
phase), and orthorhombic (𝛾 phase) structures[43,49] can occur for
tural and physical properties
most of the perovskites. During these phase transitions, B─X─B
of novel functional materials,
bond angles are changed and the corner-sharing BX6 octahedra
through close collaborations
become distorted. The transitions temperatures of MAPbI3 are
with experimental measure-
330 K for the 𝛼→𝛽 transition and 160 K for the transition from
ments. Her main research
𝛽 to 𝛾 phase.[50] For FAPbI3 , CsPbI3 , and CsSnI3 ,[51–53] a non-
interests include energy con-
perovskite 𝛿 phase has been discovered; the 𝛼 to 𝛿 phase transi-
version and storage materials,
tion involves breaking of the B─X bond, rather than the distor-
magnetic materials and thin
tion of B─X─B bond angle. Geometrical structures of 𝛽 [54] and
films, small molecules on vari-
𝛾 [51] phases for MAPbI3 can be derived from X-ray diffraction.
ous substrates.

2.1. Tolerance Factors

The structure of the ABX3 perovskite can be abstractly indicated Where rA , rB , and rX are the ionic radius at A-, B- and X-site, re-
by an empirical Goldschmidt tolerance factor[55] expressed as spectively. Another empirical parameter of the octahedral factor,
following, which is related to the stability and morphology of 𝜇, is also used to reflect the stability of the BX6 octahedral build-
ABX3 , ing unit,
r + rX
t = √ (A ) (1)
2 rB + rX 𝜇 = rB ∕rX (2)

Adv. Theory Simul. 2020, 3, 2000022 2000022 (2 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

In general, a stable ABX3 should satisfy the requirements of Table 1. The calculated band gaps of MAPbI3 for different phases under
0.4 < 𝜇 < 0.9 and 0.7 < t < 1.1.[33,42,43] While t lies between different levels, compared with the experimental one. Here, vdW means
van der Waals correlation, SR represents scalar relativistic, SOC means
0.9 and 1.0 for the cubic structures, the small t in the range of SOC.
0.71–0.90 leads to perovskites in the tetragonal and orthorhom-
bic phases. The large t in the range of 1.01–1.1, usually caused Phase Method Band gap [eV] Reference
by the larger A cations, such as organic cations containing 2–
3 C─C/C─N bonds, leads to the breakdown of the 3D ABX3 𝛽 Exp. 1.55 [1]

network to the 2D (A2 BX4 ), 1D (A3 BX5 ), or even 0D (A4 BX6 ) 𝛼 PBE 1.57 [63]

perovskites. Several review articles have summarized the low- 𝛽 PBE 1.66 [63]

dimensional perovskites.[56–58] It was found that the reduced ef- 𝛽 PBE (vdW-DF) 1.60–1.82 [64]

fective mass scales almost linearly with the size of the cluster and 𝛾 PBE (vdW-DF) 1.9 [64]

the ion emissions of Pb2+ in 0D perovskites. 𝛽 LDA 1.51 [66]

Although these structural factors can provide useful intuition 𝛼 GW-SOC 1.67 [72]
to rationalize the stability of some perovskite-derived structures,
𝛼 SR-DFT 1.34 [73]
it should be noted that these empirical parameters, derived from
𝛽 SR-DFT 1.66 [73]
hard-sphere packing models involving uncertain ionic radii, may
𝛽 HSE06 2.02 [74]
be too simple and unreliable to capture the full subtleties of inter-
atomic interaction. One of the challenges in utilizing these struc- 𝛽 HSE06-SOC 1.05 [74]

tural factors is the estimation of the effective ionic radii for the 𝛽 PBE-SOC 0.63 [74]

molecular cations, especially the nonspherical molecular cations.

In this aspect, Becker et al.[59] proposed a new computational ap-
proach, in which the radius is determined as the scaled sphere ra- 2.2. Band Structures
dius resulting from rotation of the electron density of the molec-
ular cation around the center of mass, using NH4 + as a scaling The electronic properties of MAPbI3 as well as their derivatives
reference. They calculated the tolerance factor t and octahedral have been widely studied by using the density functional the-
factor 𝜇 for 486 permutations of ABX3 organic–inorganic hybrid ory (DFT) calculations with various functionals, as shown in
compounds, which were composed by 18 organic monoammo- Table 1. Generally, 𝛽-MAPbI3 exhibits a direct bandgap within
nium cations, 9 divalent metal ions, and three halides. Based on the range from 0.63 eV by PBE-SOC to 2.02 eV by the hybrid
these, 106 hitherto unknown hypothetical compounds were pre- HSE06 functional. The valence band maximum (VBM) is mainly
dicted as the stable 3D perovskites at ambient conditions. contributed by the antibonding states of Pb-6s and I-5p orbitals,
It was reported that due to the insufficient accuracy of the tol- and the conduction band minimum (CBM) is dominated by Pb-
erance factor, the success rate for distinguishing perovskite and 6p orbitals.[63] This unique band structure is in contrast to those
nonperovskite with tolerance factors was only 74%, a value which of the conventional semiconductors such as GaAs and CdTe, in
was even lower for compounds containing heavier halides.[60] which the CBM primarily dominated by s orbital, whereas the
Sun et al.[61] calculated the decomposition energies of 138 per- VBM primarily dominated by p orbital. Yin et al.[49] considered
ovskite compounds and proposed a new descriptor (𝜇 + t)𝜂 to that the lone pair of 6s electrons in Pb account for unusual prop-
indicate the stability of perovskites, where t, 𝜇, and 𝜂 are the tol- erties for halide perovskite. The different distributions of MA+
erance factor, the octahedral factor, and the atomic packing frac- in 𝛽-MAPbI3 also lead to a band gap fluctuation of 0.22 eV. The
tion, respectively. The new descriptor showed a linear correlation effect of distributions of MA+ will be discussed in later section.
with decomposition energies, and predicted the relative stability It indicates that although the MA+ does not directly contribute
within an accuracy rate of ≈90% for the known compounds. Uti- electronic states in the band edge, they still play important roles
lizing this new structural descriptor, another 69 perovskite com- via weak interaction. The band structures and gap values of 𝛽
pounds were successfully predicted. and 𝛾 phases are similar to those of 𝛼 phase, whereas the band
To improve the accuracy of tolerance factor, Bartel et al.[62] pro- structure of 𝛿 phase is much different because the broken Pb─I
posed an even more complex tolerance factor (𝜏), which is defined bonds in the 𝛿 phase weaken the coupling between Pb-s and I-p.
as However, the inclusion of the nonlocal van der Walls (vdW) inter-
action between organic ions and the inorganic matrix decreases
( )
rX rA ∕rB the band gap difference between 𝛽 and 𝛾 phase to 0.3 eV.[64]
𝜏 = − nA nA − ( ) (3) Accurate prediction for band structures is of great impor-
rB ln rA ∕rB tance in understanding the electronic structures of perovskites.
It is widely known that the generalized gradient approximation
where nA is the oxidation state of A, ri is the ionic radius of ion i (GGA) tends to underestimate the band gaps of semiconductors
with rA > rB , and 𝜏 < 4.18 implies perovskite. Based on sure inde- due to the self-correlation error of electrons and the inherent lack
pendence screening and sparsifying operator, the new tolerance of derivative discontinuity. There are a few cases that GGA predict
factor (𝜏) achieved an overall high accuracy (over 90%) for an ex- band gaps in good agreement with experiments, for example, the
perimental dataset of 576 ABX3 materials (X = O2 − , F− , Cl− , Br− , calculated band gap of MAPbI3 about 1.6 eV with PBE-GGA[63–66]
I− ) and 918 A2 BB′X6 . The new tolerance factor was applied to ex- agrees well with the experimental results.[1,3] However, these suc-
amine 23 314 new double perovskites (A2 BB′X6 ) candidates and cesses are mostly due to the fortuitous error cancellation. The
rank these compounds by their relative stabilities as perovskites. neglecting of relativistic effects on the heavy element Pb such

Adv. Theory Simul. 2020, 3, 2000022 2000022 (3 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

as spin-orbit coupling (SOC), can lead to overestimation of the sidering GW or SOC, indicating significance of many body effect
band gap. The importance of SOC in Pb halide perovskites was and SOC.[67,76,77]
firstly highlighted by Even et al.,[67] who found that the Pb 6p- The carrier mobility can be approximately calculated based on
derived CBM exhibits a very large splitting after including SOC, the deformation potential theory,[78]
and gives an obvious large bandgap correction about 1.1 eV. How-
ever, we can see that the PBE-SOC shows smaller bandgap in an- (8𝜋 )1∕2 ℏ4 eCij
other work as shown in Table 1. 𝜇ij = ( )5∕2 ( )3∕2 2 (5)
3 mij ∗ kB T Dij
Advanced computational methods, such as hybrid functionals
and GW, were also used to assess the band gaps approximation.
where e is the element charge, kB is the Boltzmann constant, T
Hybrid functional by combination of Hartree–Fock exchange and
is the temperature, and Cij , mij *, and Dij are the stiffness ten-
LDA/GGA exchange correlation, such as HSE06,[68,69] has been
sor, effective mass tensor, and the deformation potential tensor,
widely used to calculate the band gap of semiconductors.[70–74]
respectively. Cij are obtained from linear least-squares fitting us-
The band gap of 𝛽-MAPbI3 is 2.02 eV with the hybrid HSE06
ing the stress–strain relationship. Dij can be obtained based on
functional, which is obviously larger than the experimental one,
the linear relationship between eigenvalue and strain. The car-
and the HSE-SOC gives a small band gap of 1.05 eV. The many-
rier mobility can also be predicted based on the Bloch-Boltzmann
body perturbation theory GW approximation,[75] which expresses
Theory (BBT), which involves the calculation of the conductiv-
the self-energy operator as a product of the one-body Green’s
ity tensor and carrier relaxation time. The BBT approach is more
function (G) with the screened coulomb interaction (W)), usu-
complicated than the deformation potential theory approach. The
ally also provides highly accurate results for electronic properties
interested readers might consult the reports of Motta et al.[79] and
of semiconductors. Umari et al.[76] developed an effective GW
Filippetti et al.[80]
method incorporating SOC and the relativistic effects to model
Based on the Marcus theory, Meng et al.[81] calculated the in-
the properties of MAPbI3 and other perovskites. Interestingly,
tramolecular vibration (namely, internal recombination energy
the calculated band gap is 1.50 eV with GW-SOC, in good agree-
𝜆), intermolecular electronic coupling integral V, and then the
ment with experiment. It should be noted that GW is rather time-
anisotropy of carrier mobility of MAPbI3 . They found that the
consuming, which greatly hampers the design and prediction of
electrons and holes exhibited consistency in transport orienta-
new types of perovskites. Thus it is still a big challenge to ac-
tions along the (010), (101), and (110) direction, but inconsistency
curately predict the band gaps of perovskites with some decent
along the (111) and (001) direction. The hole mobility was ex-
computational approaches.
tremely low along the (010) or (110) direction, while it changes
from 0.09 to 0.2 cm2 V−1 s−1 along the (111), (001), and (101)
2.3. Photovoltaic Properties direction. The electronic mobility ranged from 0.02 to 350 cm2
V−1 s−1 and was especially high along the (001) direction. Using
Organic–inorganic Pb halide perovskite exhibit superior photo- deformation potential theory within the carrier-acoustic phonon
voltaic properties related to their excellent device performance. scattering regime, Lee et al.[78] suggested that the carrier mobili-
Here we summary the recent progress in the related theoretical ties of hybrid halide perovskites may be reduced relative to the in-
methods and computational details. organic perovskite with same-sized A-site cations, primarily due
to the octahedral tilting and the increase of concomitant charge
effective mass.
2.3.1. Effective Mass and Carrier Mobility

Experimentally, Xing et al.[14] reported that the photo generated 2.3.2. Optical Absorption
carriers for MAPbI3 perovskites have ultra long lifetimes (up to 1
𝜇s) and diffusion lengths (up to 1 𝜇m). These unique properties The experimental results confirm that MAPbI3 show extremely
are considered to be related to the drift velocity of electrons and high optical absorption coefficients in the order of 105 cm−1 . The
holes. To evaluate them, the direction dependent effective masses optical absorption coefficients can be calculated with the follow-
of carriers are calculated according to the following equation by ing equation[82,83]
fitting the band edge, [√ ]1∕2
√
𝛼 (𝜔) = 2𝜔 𝜀1 (𝜔)2 + 𝜀2 (𝜔)2 − 𝜀1 (𝜔) (6)
[ ]−1
𝜕 2 𝜀 (k)
m ∗ = ℏ2 (4)
𝜕k2 where 𝜔 is the frequency, 𝜀1 (𝜔) and 𝜀2 (𝜔) are real and imaginary
parts of complex dielectric function, respectively. The imaginary
where m* is the effective mass, ℏ is the reduced Planck constant, part of the dielectric function represents direct inter-band tran-
𝜖(k) is the band edge eigenvalues along the wave vector k. The sitions between occupied and unoccupied electronic states. The
lower effective masses imply more efficient charge-carrier trans- real part of dielectric function can be obtained from with the
port. Giorgi et al.[77] reported the theoretical estimation of effec- Kramers–Kronig[84] relations.
tive masses of me * = 0.23 m0 and mh * = 0.29 m0 (m0 is the free Yin et al.[49] summarized the mechanisms of optical absorp-
electron mass) for pseudo-cubic (𝛼) MAPbI3 including SOC ef- tion for perovskite-based solar cell and compared with the first
fect. It should be noted that the effective mass calculated by GW and second generation solar cells. The optical absorption of a
or SOC is lower than the predicted effective mass without con- semiconductor is fundamentally determined by the transition

Adv. Theory Simul. 2020, 3, 2000022 2000022 (4 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

matrix elements and the joint density of states (JDOS). According 2.3.4. Vibrational Properties
to the Fermi’s golden rule, the optical absorption formula can be
calculated by the following equations, In order to reveal the nature of the lattice vibrations (phonons),
first principles calculations have been widely explored for the
( (⇀) (⇀) )
2𝜋 | ̂ |2 2 vibrational modes in MAPbI3 , which benefits the understand-
∫ |v|H|c| 𝛿 E k − E k − ℏ𝜔 d3 k (7)
ℏ | | 8𝜋 3 c v ing of the chemical environments and the optoelectronic prop-
erties. Brivio et al.[85] reported the vibrational frequencies for 𝛼,
̂ is the transition matrix of the valence band v to the
where v|H|c 𝛽, and 𝛾 phases of MAPbI3 , and emphasized the important role
conduction band c and the integration is over the whole recipro- of the interplay between the molecular orientation and the oc-
cal space. For simplicity, the transition matrix is approximated by tahedral networks in the structural transformations. The calcu-
considering independent of k lated phonon dispersion shows that the low-frequency modes
are associated with the inorganic (PbI3 − )n network and the high-
( (⇀) (⇀) ) frequency modes are associated with the organic MA+ cation.
2𝜋 | ̂ |2 2
|v|H|c| ∫ 𝛿 E k − E k − ℏ𝜔 d3 k (8) Despite the low-frequency modes and the high-frequency modes
ℏ| | 8𝜋 3 c v
are separated in energy, significant coupling between them is ob-
served in the calculated phonon. Pérez-Osorio et al.[86] reported
where the second term is the JDOS. Then the theoretical maxi- a detailed analysis of the vibrational eigen modes and eigen fre-
mum efficiency could be achieved as a function of the thickness quencies calculated within density functional perturbation the-
of the absorber layer by using the improved Shockley–Queisser ory, including a comprehensive factor group analysis. They di-
model. The maximum quantum efficiency could be calculated by vided the vibrational frequencies into three regions, 800–3100
numerically maximizing J × V with the standard AM1.5G solar cm−1 , 140–180 cm−1 , and < 140 cm−1 , corresponding to the vi-
spectrum. The J–V curve of an ideal p-n junction is described as brations of the MA+ , the vibrations of the cations and the internal
( ( )) vibrations of the PbI3 , respectively.
J = Jsc − J0 1 − exp eV∕kB T (9)

where Jsc is the short circuit current, J0 is the reverse saturation 2.3.5. Thermal Conductivity and Thermoelectric Properties
current, e is electronic charge, V is the voltage between elec-
trodes, kB is Boltzmann constant, and T is the temperature. Jsc Thermal transport plays an essential factor in determining the
and J0 are described as efficiency and lifetime of energy conversion devices. Most of the
solar energy is converted into heat in the form of phonons by heat
∞ dissipation. The thermal stress would lead to delamination of the
Jsc = e 𝛼 (E ) Isun (E ) dE (10) solar cells and decreases of efficiency. Therefore, understanding
∫0
the thermal properties of perovskite is beneficial for rational de-
∞
signing of solar cells. Besides, potential thermoelectric applica-
J0 = e𝜋 𝛼 (E ) Ibb (E, T ) dE (11)
∫0 tions of hybrid perovskites are also of great interests due to the
low thermal conductivity and the high mobility of charge carriers.
where 𝛼(E) is the optical absorption coefficient at photonic en- He and Galli[87] evaluated the thermal transport properties of
ergy E, Isun (E) is the sunlight photon flux at energy E, Ibb (E, T) is MABI3 (B = Pb and Sn) based on the Kane single band model.
the black-body spectrum at temperature T. They considered that the large carrier mobilities of MABI3 origi-
nate from a combination of the small effective carrier masses and
the relatively weak carrier-phonon interaction. The large values
2.3.3. Exciton Binding Energy of the Seebeck coefficients, which is a measure of the magnitude
of an induced thermoelectric voltage in response to a tempera-
The exciton bind energy (Eb ) can be calculated based on a simple ture difference across that material, are attributed to the multide-
Wannier exciton model[45] generation of conduction and valence bands. They computed the
thermoelectric figure of merit (ZT) for MABI3 and found values
𝜇e4 in the range of 1 to 2, depending on the value of the ionic thermal
Eb = (12) conductivity.
2ℏ𝜀2∞
Based on the relaxed crystal structure of MAPbI3 from DFT
where 𝜖 ∞ is the high-frequency dielectric constant and the re- calculations, Qian et al.[88] performed ab initio molecular dynam-
duced effective mass 𝜇 is related to the effective mass of electron ics simulations to obtain the relationship between interatomic
and hole forces and atomic displacements, and developed an empirical
potential field. To extract the temperature dependent thermal
( ) conductivity equilibrium molecular dynamics simulations were
1 1
𝜇 = 1∕ + (13) performed with the empirical potential field based on the Green–
me mh
Kubo (GK) relation, which can be calculated by integrating the
The exciton bind energy describes the binding energy for the exci- time-dependent heat-flux autocorrelation functions by
∞
tons generated immediately after photon excitation (without lat- 1
k = J (t) × J (0) dt (14)
tice polarization process involved). 3VkB T 2 ∫0

Adv. Theory Simul. 2020, 3, 2000022 2000022 (5 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Where T and V are the temperature and volume, J(t) and J(0) are ration of interesting crystal structures with various properties
the heat flux vectors at time t and 0, respectively, and are defined predicted by calculations, such as tunable bandgap, Rashba
as band splitting, and high stability against moisture, oxygen, and
heat.[97–101] As shown in Figure 2, the 2D (C4 H9 NH3 )2 PbX4 (X
1 ∑N ( ) 1 ∑N ∑N ( )
= Cl, Br, and I) hybrid perovskites showed the layer depen-
J= ki + pi vi + r Fij × vj (15)
2 i=1 2 i=1 j≠1 ij
dent properties.[102] For example, the effect mass of hole mh *
increases in proportion to the decreased layer thickness, the
where N is the total number of atoms, ki and pi are the kinetic and absorption peaks redshift from 502 nm for the single layer to
potential energies, vi is the velocity vector of atom i, and rij and 519 nm for the sextuple layer, and converge at 522 nm for bulk
Fij are the distance vector and interatomic force between atoms (C4 H9 NH3 )2 PbI4 .
i and j. The results showed a sudden jump in thermal conduc- Considering the multitude of organic molecules that can be
tivity at about 330 K from 0.57W mK−1 of tetragonal structure introduced into the A site of perovskites, many different kinds
to 1.80W mK−1 of the cubic structure. They attributed the lower of the hybrid low dimension perovskites can be achieved. Due
thermal conductivity of tetragonal structure to the lower phonon to the quantum and dielectric confinements, the low dimension
group velocities and the stronger anharmonicity. Wang et al.[89] perovskites often show large electron–hole binding energy (Eb )
obtain a similar result by using the same method, they further on the order of hundreds of meV, as well as the free-exciton and
noted that the rotational degree of MA+ played a governing role self-trapped excited states upon photoexcitation. These proper-
in the anisotropy of thermal conductivity at low temperatures. ties make them proper materials for luminescence, LEDs and
The virial contribution from short- and long-range intermolecu- lasers, which were summarized in other review articles.[103–106]
lar nonbonded electrostatic interactions dominates total thermal
transport, while those from intramolecular interactions are neg-
ligible. 2.6. High Throughput and Machine Learning for Screening of
Perovskites

2.4. Effect of Pressure and Strain Artificial intelligence (AI) or machine learning (ML) are now used
from daily life to scientific research in diverse areas such as med-
External pressure and strain, as a green and facile strategy, can be ical diagnosis, chemical reactions, and novel materials design in-
applied to regulate the band gap and properties of perovskites.[90] cluding perovskites and their modified materials. Over the past
Experimentally, the external pressure or strain can be achieved decades, the perovskite data have been accumulated explosively,
via a diamond anvil cell or growth of epitaxial thin films. Compu- both experimentally and computationally. For instance, halide
tationally, the external pressure or strain can be applied to simula- double perovskites with a formula of A2 B(I)B(III)X6 , which is
tions via volumetric strain, axial strain, layer mismatch, etc. Com- resulting from disproportionation of the divalent B(II) cation of
putational studies of the tunable properties such as electronic ABX3 into monovalent B(I) and trivalent B(III) cations, offers up
structures under external pressure for MAPbX3 (X = I, Br, Cl) to 104 candidates.[107] Not to mention triple-A double-X,[108,109]
and all-inorganic halide perovskites CsMX3 were reported.[91–94] quadruple-A double-X[110] perovskite have been used to improve
Using first-principles calculations, Liu et al.[95] found that CsPbI3 the performance and stability in experiments. Because of the
under appropriate strain engineering exhibited ferroelectric and huge data size and dimension, it becomes extremely challenging
topological orders simultaneously, namely a ferroelectric topolog- to screen proper perovskite and perform data analyses for pho-
ical insulator. The bulk polarization was stabilized by the metallic toelectric applications with traditional computational methods,
topological surface states, because it created an intrinsic short- which derives materials properties deductively on the basis of
circuit condition as innate metallic electrodes, allowing auto- physical laws, for example, solving the Kohn–Sham equation. In-
matic dissipation of surface charges. Suppressing PbI6 cage ro- stead, the inductive data-driven ML bypassing the complex quan-
tation is key for achieving a ferroelectric topological insulator. tum mechanics pictures the feature–properties mapping. Based
Zhang et al. showed that the inorganic framework deforms un- on the pre-computed materials database, efficient algorithms are
der external strains and the embedded organic MA+ rotates cor- deployed for data mining (DM) to learn the hidden relationship
respondingly. A band gap oscillation and a new phase were ob- between various physical parameters with performance. It can
served in response to the external strain.[96] not only greatly accelerate materials design with high accuracy,
but also provide capabilities to predict the outcome of a future
observation.
2.5. Low Dimensional Perovskites Xu et al.[111] developed a procedure to identify formability of
perovskites with the stoichiometry of ABX3 and A2 B’B”X6 by us-
In the ABX3 , the big A-site cation could destabilize the 3D B-X ing the crystal structures from the Materials Project database.[112]
network, leading to the 2D layered structure. The 2D perovskites The criterions include cations coordination number (whether
with the layered structure can be regarded as a slice of the 3D equal to 6), the topology of corner-sharing BX6 octahedra and
perovskite by breaking the corner-shared octahedral structure. the space group. This procedure achieved about 90% accuracy
The 2D perovskites can be further decomposed into 1D and 0D for formability prediction. In another work,[113] using DFT re-
perovskites. The structural flexibility and tunability as well as the sults of 354 perovskites as a training set, three statistic models,
significant quantum confinement effects of the low dimension kernel ridge regression (KRR), k-nearest neighbor, and support
perovskites provide a rich and fertile “playground” for the prepa- vector machine were tested to establish ML mapping between

Adv. Theory Simul. 2020, 3, 2000022 2000022 (6 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 2. a,b) Crystal structure for 2D (C4 H9 NH3 )2 PbX4 (X = Cl, Br, I) perovskites. c,d) Crystal structures viewed along the [001] and [010] directions. e)
Thermodynamically stable ranges of Pb and I chemical potential under the equilibrium growth conditions. f) Surface energies of C4 H9 NH3 -T and PbI-T
as a function of layer number. Dashed lines showed the surface energy of MAPbX3 with MAI-T termination for comparison. g) The effective masses of
electron (me */me ) and hole (mh */me ), exciton binding energy (Eeb ) and h) absorption coefficients with different layer numbers.[102] Reproduced with
permission. Copyright 2017, American Chemical Society.

the thermodynamic stability and features. The ML model was gorithm evolves from the combination of boosting methods and
trained based on the feature input of the ionic radii (RA , RB , and regression trees. The accuracy of the regression is improved by
RX ) and property output of decomposition energies (ΔHD ); fea- gradually reducing the algorithm generated by the training pro-
tures of electronegativity had also been checked but it exhibited cess. The final regression algorithm is the weighted sum of sev-
obvious overfitting. The KRR model showed better performance. eral weak regression algorithms obtained from each training. It
Then this ML model was independently validated by the exper- is suitable for effectively mining features, feature engineering,
imental results of 246 double perovskites (beyond the training and establish nonlinear mapping between the input features and
set of 354) and used to predict the ΔHD for total 14 190 addi- outputs. A “last-place elimination” procedure was employed to
tional halide perovskites and mixed-halide perovskites. The ML exclude the less relevant features from 30 initially considered fea-
predictions are in good agreement with DFT-calculated results tures in the GBR analysis. In this procedure, the initial 30 fea-
for 74 perovskites randomly selected in 14 190 halide double per- tures were ranked by GBR algorithm according to the relative
ovskites that were not previously reported. importance, and the least one was discarded. This elimination
Based on the DFT calculations of ABX3 chalcogenide, halide step was repeated for 29 times. They found that the model score,
and hybrid perovskites, Park et al.[114] also carried out data R2 , reaches the maximum value when the feature set includes 14
analyses with both supervised and unsupervised machine features, such as tolerance factor, octahedral factor, total number
learning methods. The supervised machine learning employed of ionic charge, electronegativity, electron affinity, ionic polariz-
random forest (RF) regression to assess how some physical ability, sum of the s and p orbital radii, iron radii of B site atoms,
and chemical attributes affect the octahedral deformation and ionic polarizability, iron radii, the highest occupied molecular or-
the bandgap. Random Forest can better tackle properties in a bital, and the lowest unoccupied molecular orbital (LUMO) of A
non-linear way compared to the general linear method. The site cations, and p orbital electron of X site atoms. They also com-
considered attributes include the ionic radius, the Goldschmidt pared the performance of GBR with those of other five ML regres-
tolerance factor, the octahedral factor, the electronegativity, the sion algorithms, such as KRR, support vector regression, Gaus-
electron affinity, the ionization energies, the period number, the sian process regression, decision trees regression, and multilayer
group number, and the atomic number. Unsupervised machine perceptron regression based on three indexes of R2 , Pearson co-
learning, namely the t-SNE cluster analysis, was performed to efficient (r), and mean square error, and found GBR algorithm
classify similar compounds. presented more reliable results. The 6 perovskites screened out
Lu et al.[115] obtained 346 hybrid organic–inorganic perovskites by ML were further checked using first-principle calculations for
from high-throughput first-principles calculations as the input the thermal and environmental stability. Finally, they proposed
data, and then they performed ML process showed in Figure 3 that C2 H5 OInBr3 , C2 H5 OSnBr3 , and C2 H6 NSnBr3 could prevent
to discover perovskites with appropriate bandgaps, outstanding the hydration degradation.
thermal and environmental stabilities. A flexible non-parametric There are also other reports about high-throughput calcu-
statistical algorithm, gradient boosted regression (GBR), is im- lations or ML for perovskite coating materials, such as the
plemented in an open source package, scikit-learn,[116] to ana- screening perovskites with gradient-boosted regression tree[117]
lyze the dataset and construct a predictive model. The GBR al- or artificial neural network,[118] predicting electron transmission

Adv. Theory Simul. 2020, 3, 2000022 2000022 (7 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 3. a) Design framework for lead-free perovskites. The blue boxes represent the material screening process based on the ML algorithm. The green
boxes show electronic properties and stability calculated by DFT. b) High-throughput calculations of perovskites. The combination of 11 small organic
molecular species (A-site) and 32 divalent metals (B-site), and X is a typical halide.[115] Reproduced with permission. Copyright 2018, Springer Nature.

coefficients between MAPbI3 and metal electrodes, and so in open-circuit voltage (Voc ). However, deep level transient
on.[119–126] Now the artificial intelligence or machine learning spectroscopy experiments showed that the charge trap densities
could even surpass human intelligence in some fields. For ex- of MAPbI3 are in the range of 1015 –1016 cm−3 ,[127] lower than
ample, they have defeated the top player in chess and go games, those of conventional semiconductors such as polycrystalline
and we hope artificial intelligence can work as top chemists Si (1013 –1014 cm−3 ), CdTe (1013 –1015 cm−3 ), and CIGS (1013
and physicists to design and optimize the high performance cm−3 ). Several theoretical studies have been performed to unveil
perovskite in the near future. the unique defect tolerance nature. Yin et al.[49] systemati-
cally studied the possible intrinsic point defects of 𝛼-MAPbI3
including vacancies (VMA , VPb , VI ), interstitial (MAi , Pbi , Ii ),
3. Defects and Doping of Halide Perovskites cation substitutions (MAPb , PbMA ), and antisite substitutions
(MAI , PbI , IMA , IPb ) by DFT. They investigated the formation
3.1. Formation and Effect of Defects energies of these defects as functions of Fermi level under
I-rich/Pb-poor, moderate, and I-poor/Pb-rich conditions, re-
For semiconductor applications in photoelectricity, native de- spectively. They also showed that the Fermi level position,
fects that inevitably exist in crystalline play prominent roles, when the defect charge states are changed by electrons gain or
because they usually create the deep level gap states to act as lose, determined the electrical effect of the defects (Figure 4).
carrier traps and Shockley-Read-Hall nonradiative recombi- Their results suggested that the dominant defects with low for-
nation centers, leading to reduced carrier lifetime and losses mation energies only produce shallow levels, in contrast to the

Adv. Theory Simul. 2020, 3, 2000022 2000022 (8 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 4. Transition energy levels of point defects in CH3 NH3 PbI3 . The formation energies for the neutral defects are shown in parentheses.[49] Repro-
duced with permission. Copyright 2014, John Wiley & Sons, Inc.
Table 2. Summary of selected references of DFT work on the activation energies ΔEa of defect migrations in perovskites APbX3 . V stands for vacancy
defect, subscript i represents impurity defect; AFM and FM represent antiferroelectric and ferroelectric; < > show the directions of migration.

ΔEa [eV] VX VA VPb Xi Ai Reference

MAPbI3 0.08 0.46 0.80 0.08 [ 146]

MAPbBr3 0.09 0.56 [ 146]

MAPbI3 0.32<100> 0.44<111> 0.57<110> 0.89<001> 0.32<100> [ 147]

FAPbI3 0.55<100> 0.48<111> 0.61<110> 0.61<001> [ 147]

MAPbI3 AFM 0.38 0.57 0.20 0.28 [ 150]

MAPbI3 FM 0.26 0.66 0.21 0.24 [ 150]

MAPbI3 0.342 0.735 [ 151]

MAPbBrx I3-x 0.456–0.522 (I)***0.326 (Br) 0.801 [ 151]

MAPbI3 AFM 0.06–0.65 [ 155]

MAPbI3 FM 0.12–0.52 [ 155]

deep level, which account for the generation of free carriers and ticular charge state. Therefore, it is feasible to prevent their for-
thus are desirable for photovoltaics. As MAPbI3 is high iconicity, mation by controlling the equilibrium carrier concentrations to
when defects disrupt connections between the atoms, the energy make sure that the Fermi energy position is about 0.3 eV away
levels on valence and conduction states do not change signifi- from the band edges. Whalley et al.[137] also suggested that the
cantly. In addition, they also suggest that the MAPbI3 exhibits polyanion nature of iodine leads to the formation of V-center
bipolarly doping property with variable conductivity[128] from (2X− + h+ → X2 − ) and H-center (X− + Xi − + h+ → X2 − ) that acted
p-type to n-type depending on the chemical potentials, which as hole traps.
is in line with the experiment findings for p-type[129,130] and
n-type MAPbI3 [131] at different growth conditions. The similar
properties have been found in 𝛽-MAPbI3 [132,133] and CsPbBr3 .[134] 3.2. Ion/Defect Migration
Besides the point defects, there are reports about other types
of common defects, such as Schottky defects (equal numbers of The ion or defect migration in perovskites has been proposed to
positive and negative vacancies) and Frenkel defects (equal num- correlate many peculiarities,[138] including current–voltage (J–V)
bers of vacancies and interstitials of the same ion). Kim et al.[135] hysteresis,[139–142] switchable photovoltaic,[143] photo-induced
suggested that the formation energies of PbI2 vacancies were far giant dielectric constant,[144] and electric-field-driven reversible
lower than those of MAI vacancies; the localized states within the phase conversion.[145] Some related works on the activation en-
band gap were not found either. Agiorgousis et al.,[136] however, ergies of defect migrations in perovskites are shown in Table 2.
found that the formation of Pb dimer and I trimer with strong Azpiroz et al.[146] obtained a low activation barrier about 0.1 eV
covalent bonds in MAPbI3 , in contrast to the point defects and for the migration of I vacancies and interstitials, corresponding
Schottky defects, created deep levels within the band gap. These to a migration time of < 1 𝜇s, which is too fast to explain the slow
defects show significantly lower formation energy only in par- J–V hysteresis on the time scale of seconds. While the activation

Adv. Theory Simul. 2020, 3, 2000022 2000022 (9 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 5. Mechanism for light-induced annihilation of Frenkel defects. Left: calculated energy profile for formation of a VI + /Ii − pair as a function of the
distance between VI + and Ii − at the ground state (GS, red line) and lowest excited state (e/h, blue line). The perfect crystal corresponds to the zero
VI + /Ii − distance. Right: corresponding structures.[152] Reproduced with permission. Copyright 2016, The Royal Society of Chemistry.

energies for the MA and Pb vacancies migration were 0.5 and in MAPbI3 at finite temperatures. They considered that the I dif-
0.8 eV, respectively, and response times were on the order of fusion is by far the dominant mechanism for ionic transport in
tens of ms to minutes, suggesting that the migrations of the MA MAPbI3 at 300–450 K. For I interstitials and vacancies, diffusivi-
and Pb vacancies were responsible for the J–V hysteresis. Using ties are ultrahigh (7.4 × 10−7 and 4.3 × 10−6 cm2 s−1 at 300 K) and
nudged elastic band (NEB) method, Haruyama et al.[147] reported single activation energies are 0.24 and 0.1 eV, respectively.
that the activation energies of the I− migrations in tetragonal Beside the intrinsic defects mentioned above, Egger et al.[154]
MAPbI3 and trigonal FAPbI3 were 0.44 and 0.48 eV, respectively, investigated another kind of defect, interstitial hydrogen based
and the activation energies for the MA and FA migrations on DFT calculations. They showed that the interstitial hydrogens
were 0.57 and 0.61 eV, respectively. They suggested that the were located at different sites depending on the charge state. The
replacement of MA+ with larger cation molecules might be a further calculations indicated a low migration barrier for proton
promising way to suppress J–V hysteresis and to prevent aging of transfer, which might eventually allow for ionization enhanced
perovskites. Similar results on ion migration activation energies defect migration.
were reported by others.[148,149] The dominant defects could vary, Tong et al.[155] then investigated the diffusion of the I− , MA+ ,
depending on the growth conditions and methods.[150] As Yang and Pb2+ ions in the anti-ferroelectric and ferroelectric struc-
et al. suggested, the diffusion of MA has a much smaller barrier tures, respectively. The results showed that the energy barriers
through the “interstiticaly” mechanism compared to that of the were greatly affected by the dipole moments of the MA species.
vacancy-assisted mechanism. An increased activation energy for For example, the energy barriers of the most mobile I− ion range
the I vacancy diffusion from 0.34 to 0.46 eV with Br substitution from 0.06 to 0.65 eV. The positively charged MA+ and Pb2+ ions
was reported by Zhang et al.[151] prefer to diffuse along the dipole direction, whereas the neg-
To understand the mechanism of the ion or defect migration, atively charged I− ion strongly prefers the opposite direction.
the molecular dynamics studies also were carried out at the fi- Therefore, the arrangement of MA+ can effectively modulate the
nite temperatures. Mosconi et al.[152] employed Car–Parrinello current–voltage hysteresis intensity by the ion migration. The
molecular dynamics (CPMD) to investigate the VI+ ∕Ii− Frenkel barriers of I, MA, Pb were showed in Table 2.
pairs and the mechanism of defect migration under light irradi-
ation are showed in Figure 5. They found in the ground state,
the energy of intimate VI+ ∕Ii− couple is only 0.27 eV above that 3.3. Compositional Substitution and Mixing Perovskites
of the perfect crystal with an activation energy of 0.31 eV. When
taking the entropy into account, the estimated Gibbs free energy Compositional mixing or doping can be used to tune the phys-
formation of VI+ ∕Ii− couple were about 0.13 eV at room temper- ical and electronic properties of perovskites, such as conductiv-
ature. At the excited state, VI+ ∕Ii− couple is less stable than the ity and mobility. It is also considered a feasible way to tackle the
perfect crystal by 0.49 eV. They further considered that MA could stability and toxicity. Some combinations of cations or halides
assist defect migration by means of local reorientation. Delugas for perovskites have been demonstrated superiority over single
et al.[153] performed classical molecular dynamics with the clas- ion perovskites, in terms of both efficiency and stability. Because
sical interatomic potential to study the diffusion of point defects the doping has been described in somewhere else,[34,48] here we

Adv. Theory Simul. 2020, 3, 2000022 2000022 (10 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Table 3. Some recent works on the substitution and mixing perovskites.

Perovskites Methods Calculations Reference

𝛼-RbPbI3, 𝛼-CsPbI3, 𝛿-RbPbI3, 𝛿-CsPbI3 PBE(-SOC), HSE06(-SOC) Band aliment, dielectric tensors, born [156]

effective charge tensor

Li, Na, K, Rb doped FAPbI3 and FAPbI3 B3LYP Band gap, chemical shift, Raman and [157]

IR spectra
MA1-x FMAx PbI3
(FMA = [CH2 F-NH3 ]+ [CHF2 -NH3 ]+ ) PBE Band structure, formation energy, [158]

iodine vacancy diffusion

FA1-x Csx PbI3 HSE06 Tolerance factors, relative energy [159]

CsPbX3-x Yx (X, Y = I, Br, Cl) PBE-SOC, HSE06-SOC Band structure, formation energy [160]

CH3 NH3 PbI3-x Clx SR-SOC Band structure [161]

CsPbX3 (X = I, PF6 , and BF4 ) HSE06-SOC Band structure, tolerance factor, [165]

effective mass
NH3 NH3 ZrS3 , ABX3 (A = Ca, Sr, and Ba, B HSE06 Band structure, optical absorption [166]

= Ti, Zr, and Hf, X = S and Se)

MASnx Pb1-x I3 SR-DFT, SOC-GW Band aliment, optical absorption [170]

CH3 NH3 SnX3 (X = Cl,Br,I) vdW-DF Band structure, deformation energy, [171]

ealctic constant, effective masses,

carrier mobility
RbxCs1-x SnI3 PBE, HSE06-SOC Band structure, surface energy, band [172]

aliment
CH3 NH3 GeI3 PBE(-D3), PBEsol(-D3), HSE06(-D3) Band structure, effective mass [173]

Cs2 CuInX6 , Cs2 AgInX6 (X = Cl, Br, I) PBE, HSE06 Band structure, decomposition energy [174]

Cs2 In(I)M(III)X6 (M = Bi, Sb; X = Halogen) PBE, HSE06 Band structure, decomposition energy [17]

mainly focus on the recent calculations on the substitution and et al.[162] showed the potential possibilities for pseudohalide an-
mixing perovskites summarized in Table 3. ions, such as SCN− , BH4 − , N3 − , HCOO− , and C6 H6 S− , being in-
The A site is commonly occupied by large and monovalent corporated in perovskites; Other pseudohalide anions, such as
cations such as organic MA+ , FA+ , inorganic Cs+ , Rb+ .[156] BF4 − and PF6 − ,[165] have also been demonstrated. A review arti-
Suzuki et al.[157] reported the theoretical investigation of incor- cle focused on chalcogenide perovskites[166] reported that replace-
poration of alkali metal cations (Li+ , Na+ , K+ , and Rb+ ) into ment of the alkaline-earth cations at A site by molecular cations
FAPbI3 , which showed the decreased Gibbs free energy with the is stable for chalcogenide perovskites and they show compara-
tuned lattice constants and the deviation of charge distribution. ble optical absorption properties with halide perovskites. Thus,
El-Mellouhi et al.[158] suggested that the optimal fluorination of chalcogenide perovskites may provide new route toward solving
the MA+ cation has a tendency to enhance the thermodynamic the issues of instability against moisture and inclusion of toxic
stability without significant impacts on the band gap due to the Pb.
weak partial hydrogen bonds between MA+ and the inorganic Pb-free perovskites, resulting from substitution of the toxic
framework. Meanwhile, fluorination could significantly reduce Pb at the B site, have been recently summarized on several re-
the diffusion of iodine vacancies under applied bias voltage. Li view articles.[41,167–169] Two ways are used for substitution of Pb:
et al.[159] suggested that Cs doped FAPbX3 , FA0.85 Cs0.15 PbI3 , with one is the substitution by divalent metals such as Sn,[170–172]
a tuned tolerance factor, could enhance the stability and sup- Ge,[173] Cu(II), and the other is the heterovalent substitution
pressed the humidity-triggered 𝛼-to-𝛿 phase transition in FAPbI3 by B(I)/B(III) pairs (B(I) = Cu(I), Ag,[174] In;[175] B(III) = Bi,
films. Sb[175] ). The non-toxic perovskites containing two kind of B
For X site, perovskites with mixing halides, such as I− , Br− , atoms could be double perovskites, and even more multiple per-
and Cl− , are widely studied.[160–164] Yin et al.[160] performed first- ovskites. Till now, Pb-free perovskites cannot fully replace Pb
principles and Monte Carlo (MC) calculations on mixing CsPbX3 based perovskites due to structural stability and lower photo-
(X = I/Br, Br/Cl, I/Cl). Based on the curve of alloy formation en- voltaic performance.[41] More efforts must be made to further im-
ergies, they proposed that it is difficult to form I/Cl mixing per- prove their instability issue.
ovskite, but easy to form I/Br or Br/Cl mixing perovskites near
the room temperature. The mixing perovskites showed partic-
ularly stable energy at the I/Br or Br/Cl ratio of 1/3, which is 4. Surface and Interface of Perovskites
attributed to the joint effects of strains and coulomb energies.
Colella et al.[161] showed that the Cl incorporation was allowed The PSC devices usually employs layered structures, where the
only at very low concentrations (below 3–4%) for MAPBI3 , but it perovskite forms the interface with molecules and semicon-
could dramatically improve the charge transport. Besides, Walker ductors that act as electron/hole transport material (HTM) or

Adv. Theory Simul. 2020, 3, 2000022 2000022 (11 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

passivation layers. The surface or interface composition and

structures play crucial roles for device functionality, perfor-
mance and even more device stability. The contact between
perovskite and adjacent materials governs the charge transfer,
electron/hole injection, and recombination. Experimental re-
sults, such as impedance spectroscopy, suggest that interfacial
recombination is the major factor for photovoltage loss.[176] On
the other hand, the degradation of perovskite, which involves a
series of complex chemical reaction processes, always initiates at
the surface and can be suppressed by surface passivation. Nev-
ertheless, arising from the arrangement of different elements,
molecules, and even defects, the complicated composition and
structures of interfaces magnify the intricacy and make the
mechanism more unclear. Many works on interfaces between
perovskite and molecules as well as crystals have been reported
Figure 6. a) The optimized structures of 5-layer slabs with (001) surface of
and summied.[177,178] We focus on here the theoretical progress MAPbI3 perovskite, left are MAI-T and right are PBI2-T surfaces. b) Their
in surface and interface of perovskites on this section. band gaps as a function of slab thickness. c) Absorption coefficients with
different slab thickness.[180] Reproduced with permission. Copyright 2015,
The Chinese Ceramic Society, published by Elsevier B.V.
4.1. Surface Structures and Reconstruction

It is noted that the investigations of perovskites surfaces are usu- DFT calculations suggested that the reorientation of surface MA
ally based on a finite set of unit cells in the slab model. There dipoles weakened the surface polarity and strengthened the inter-
are mainly two types of slab models: one with the same chem- actions with surface iodine anions. This leads to the iodine dimer-
ical termination for upper and down sides, and the other with ization and thus stabilize the surface structures. Che et al.[183] in-
different terminations. The former leads to a non-stoichiometric vestigated the reconstruction for (001) surface of the orthorhom-
formula as compared to the formula of the unit cell, and the later bic MAPbBr following the above STM results, and they found
introduces a surface polarization. It should be carefully consid- the existence of the Rashba effect both in valence and conduction
ered whether the calculated properties are affected by the choices bands, due to the inversion symmetry breaking in the presence
of models. of strong SOC. The caused band splitting are 0.59 eV Å (0.24 eV
Haruyama et al.[179] investigated the structures of the (110), Å) for the CBM (VBM) of the zigzag phase and 0.44 eV Å (0.27 eV
(001), (100), and (101) surfaces of 𝛽-MAPbI3 by DFT calcula- Å) for the CBM (VBM) of dimer phase.
tions. They found that a vacant PbI2 termination (with broken
PbIx polyhedron) was energetically more favorable than the flat
PbI2 termination under thermodynamic equilibrium conditions, 4.2. Interaction with Environmental Species
and the (110) and (001) surfaces were the most stable among
these surfaces. They suggested that both the terminations could Many environmental factors affect the stability of the PSC. Hu-
coexist on (110) or (001) surfaces and maintain the electronic midity, oxygen, and light radiation are the most critical environ-
characteristics of bulk MAPbI3 without midgap states. However, mental factors that are responsible for the degradation of PSC.
Geng et al.[180] found the MAI terminated surfaces is relatively Exposure of unpackaged MAPbI3 to air with humidity leads to
more favorable than the PBI2 terminated surfaces in equilibrium rapid irreversible decomposition of perovskite with formation of
growth conditions (see Figure 6), and the formation of the latter various products, such as PbI2 , I2 , HI, MAI, and so on.[184,185]
ones could only be achieved under PBI2 rich chemical environ- Computational works were performed to investigate the nature
ment. Furthermore, these two types of surfaces exhibit different of the interaction between water and perovskite, so as to uncover
electronic properties. The PbI2 terminated surface show smaller the mechanism of the degradation or solvation processes of per-
band gaps compared with the bulk MAPbI3 due to the surface ovskite.
dangling Pb-p states, whereas the MAI terminated surface show Mosconi et al.[186] presented CPMD simulations for the MAI-
layer dependent band gaps that decrease with the increase of and PbI2 -terminated (001) surfaces of tetragonal MAPbI3 . They
number of layers. Wang et al.[181] studied the low-Miller-index sto- found that the MAI-terminated surface underwent a rapid sol-
ichiometric surfaces of orthorhombic MAPbI3 including (100), vation process due to the interaction of water molecules with
(010), (001), (110), (101), (011), and (111) surfaces with differently Pb sites. A nucleophilic substitution of I by H2 O was observed
coordinated “defected” Pb-centered octahedra, and they found accompanying with the releasing of MA molecules. The PbI2 -
that both defected Pb-centered octahedral structure and coordi- terminated surfaces were less sensitive to the presence of interfa-
nation number of I greatly affect the stabilities of orthorhombic cial water due to the particularly shorter and stronger Pb─I bonds
MAPbI3 surfaces. on these facets. However, the percolation of a water molecule
For MAI terminated (001) surfaces, its surface reconstruction into the PbI2 -terminated surfaces could be the initial step for the
has been confirmed by scanning tunneling microscopy (STM). new hydrated phase formation. They predicted that the hydra-
The coexistence of pristine zigzag phase and a reconstructed sur- tion of both surfaces was exergonic, with the formation energies
face phase with iodine dimmers are found.[182] Corresponding of −0.49 and −0.44 eV, respectively.

Adv. Theory Simul. 2020, 3, 2000022 2000022 (12 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 7. The calculated energy profile of water molecules move from a) outside of the surface to the first atom layer and then b) the sub-surface by
NEB technique.[187] Reproduced with permission. Copyright 2015, American Chemical Society.

Tong et al. also investigated the interaction between water The mechanism for oxygen-induced photodegradation was
molecules and the (001) surface of MAPbI3 .[187] Their first- also investigated by combination of experimental and computa-
principles results indicated that the adsorption energy of per tional techniques. Aristidou et al.[191] suggested that the photoex-
water molecule increased from 0.30 to 0.58 eV as the number of cited carriers could reach the O2 molecules that were incorpo-
water molecules increased from 1 to 8 due to the clustering of rated into the MAPbI3 surface to form the reactive superoxide
water molecules via hydrogen bonds, indicating the hydrophilic- ions (O2 − ), making the decomposition reaction thermodynami-
ity of the surface. Next, they predicted the trajectory for water cally favorable. DFT simulations indicated that the formation of
molecules moving from above the surface to the first atom layer O2 − species by the direct electron transfer from the perovskite
of the surface and then to the sub-surface by NEB technique. to oxygen is energetically favorable and iodide vacancies are the
The calculated energy barriers for these two steps were 0.04 most preferred donor sites. Upon adsorption, the bond length be-
and 0.31 eV, respectively, and the relative energies were −0.18 tween O atoms increases from 1.22 to 1.33 Å and analysis of the
and −0.39 eV, respectively, as shown in Figure 7, indicating electron density shows the spin-polarized surface species.
that the penetration of water molecule is thermodynamically
favorable and kinetic feasible. Then, first principle molecular
dynamics (FPMD) was carried out with the NVT ensemble at 4.3. Adsorption of Other “Extrinsic” Species
the temperature of 300 K. The first step of penetration was
clearly observed but the second step is absent due to the larger In addition to the environmental species discussed above, per-
barriers. Finally, they calculated the density of states and optical ovskite interfaces interact with many other species, such as pas-
absorbance spectrum of water on the outside, first layer and sivating films and charge transport layers. Most of these extrinsic
sublayer, respectively. The result showed that the bandgap are species are added artificially in order to improve the stability, elec-
greatly increased from 1.58 to 1.62 eV and 1.74 eV, as well as a tron or hole transport, and so on. It is important to understand
weakened optical absorption. the nature and effect of interactions between a variety of extrinsic
Koocher et al.[188] performed DFT calculations with an implicit molecular species or crystal line and perovskite surfaces.
solvation model for both MAI- and PbI2 -terminated surfaces. A combined experiments and theory was carried out for molec-
They suggested that the infiltration of water molecules into the ular passivation treatment of perovskite films.[192] Small sized
hollow site of the surface depends on the orientation of surface “moisture resistive” passivation molecules, including aniline (A-
MA+ , and a high water concentration may facilitate degradation FAPbI3 ), benzylamine (BA-FAPbI3 ), and phenethylamine (PA-
through the lattice distortion. By comparing the DFT simulated FAPbI3 ) modified FAPbI3 films were fabricated. The moisture-
with experimental STM images, Liu et al.[189] showed that the de- resistance of these materials were compared by exposing them in
fects on MAPbBr3 surface were mainly Br vacancies or Br-MA di- air with 50 ± 5% RH. The unmodified FAPbI3 film changed from
vacancies. Further calculations indicated that the adsorption en- black 𝛼 to yellow 𝛿 phase within 72 h and completely decomposed
ergies of water, oxygen, and acetonitrile significantly increased at within 90 h; while for the A-FAPbI3 and PA-FAPbI3 , the conver-
these vacancies due to the formation of hydrogen bonds between sion was reduced, yet still pronounced. In contrast, BA-FAPbI3
the adsorbates and the defective surface. The FPMD[190] showed exhibited excellent stability, showing no phase change even after
similar results by examination of the ground and excited states 4 months. In order to understand this phenomenon, they con-
of nanoclusters (MAPbI3 )8 and (MAPbI3 )8 ·nH2 O. The electronic ducted DFT calculations to evaluate the interaction between wa-
excitation tends to weaken Pb─I bonds in both nanoparticles. At ter molecules and the perovskite interface. The corresponding
the excited state of (MAPbI3 )8 ·nH2 O, the bond order could be formation energies for A-FAPbI3 , BA-FAPbI3 , and PA-FAPbI3
maximally decreased by up to 20% with water molecules involved were −0.19, −0.26, and −0.49 eV per molecule, verifying the ther-
in the electronic excitation. modynamic stability of the amine modification. The calculated

Adv. Theory Simul. 2020, 3, 2000022 2000022 (13 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 8. Optimized structures of a) MAI/anatase, b) PbI/anatase, c) MAI/rutile, and d) PbI/rutile as well as the e) PDOS in the same sequence from
up to down.[195] Reproduced with permission. Copyright 2016, Springer Nature.

water adsorption energy (Ead ) of FAPbI3 , A-FAPbI3 , BA-FAPbI3 , As shown in Figure 8, the MAPbI3 (001) had the relatively small
and PA-FAPbI3 on the surface were −0.58, −0.54, −0.40, and lattice mismatch (<5%) against anatase (001) and rutile (001).
−0.40 eV, respectively, indicating a water-repelling effect of the The PbI2 terminated surfaces showed larger binding energy
phenylalkylamine passivation. Furthermore, BA-FAPbI3 showed than that of MAI terminated surfaces, duo to the stronger inter-
that water is far away from the PbI2 layer than those of A-FAPbI3 face bonds formed from the dangling Pb atoms than the bonds
and PA-FAPbI3 . Thus, water molecules hardly penetrate into the formed from the MA. For TiO2 , the rutile is more energetically fa-
subsurface of FAPbI3 . vorable than anatase because the better atom arrangement on the
The Lewis base thiophene passivated halide perovskite for interface and smaller lattice mismatch. Furthermore, the PBI2 -
both MAI- and PbI2 -terminated surfaces were investigated by rutile interface showed a close band aliment both at VBM be-
Zhang et al.[193] They found the passivation for both surfaces tween I-p/O-p and at CBM between Pb-p/Ti-d, exhibiting delo-
were energy favorable with a slight large adsorption energy for calized features according to the electron localization function.
PbI2 -terminated surface. Hirshfeld charge analysis confirmed These characteristics might originate from the interface charge
the electrical neutrality of the thiophene passivated surface. transfer and the intrinsic electric field.
Projected density of states (PDOS) indicated that the thiophene Quarti et al.[196] performed DFT simulations on the hetero-
molecule contributed to the edge of the valence band. The junctions comprising MAPbI3 and C60 . For MAI-terminated sur-
MD simulation depicted the water repellence of the thiophene face, the CBM of the perovskite lies above the LUMO of the C60 ,
molecule. They further confirmed that the charge transfer accompanying a spontaneous charge transfer from the perovskite
characters of the surfaces could be manipulated by replacing a to the C60 . Meanwhile, the C60 shows an upshift of its electronic
hydrogen atom with electron donating group methoxy (─OMe) levels due to the charge transfer, leading to a reduction of the off-
or electron withdrawing group carboxylic acid (─COOH), and set between the perovskite CBM and the C60 LUMO. For PbI2 -
chlorine (─Cl). Torres and Rego[194] calculated the adsorption of terminated surface, the CBM of the perovskite lies below the
methoxybenzene (CH3 OC6 H5 ), which is a fragment of the most LUMO of the C60 , instead of hindering electron injection. The
commonly used HTM spiro-MeOTAD, on the perovskite sur- C60 showed a downshift of its electronic levels, leaving the elec-
face to clarify the attachment mechanisms of HTM. The results tron injection from the perovskite to the C60 .
demonstrated that methoxybenzene formed stable adsorption on Mosconi et al.[197] investigated the MAPbI3-x Clx (110) on
the (001) PbI6 octahedra terminated surface with an adsorption anatase TiO2 (101) by first-principles computations. They found
energy of ≈38 kJ mol−1 . that MAPbI3-x Clx showed enhanced binding energy compared to
MAPbI3 , indicating the stabilization of the interface by chlorine
mixing. The stronger interaction between MAPbI3-x Clx and
4.4. Interfaces between Perovskite and Hole or Electron TiO2 leaded to a slight TiO2 conduction band energy upshift
Transporting Materials and a stronger interfacial coupling between the Ti-d and Pb-p in
the conduction band states, which could enhance the electron
Another important aspect is the energy level alignment between injection.
the perovskite and the HTM or electron transporting (ETM) ma- Mosconi and Angelis et al. reported another two works with
terials. Geng et al. performed first-principles calculations to study combination of experiments and DFT calculations.[198,199] It was
the energy level alignment between MAI/PbI2 terminated sur- indicated a surface dipole of 28 D pointed from TiO2 to the
face and TiO2 anatase/rutile (the most commonly used ETM).[195] perovskite in the supercell, corresponding to ≈4.8 D per nm2 of

Adv. Theory Simul. 2020, 3, 2000022 2000022 (14 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

TiO2 surface. They also suggested that the mixed chlorines were trapping surface defects under I-rich conditions is energetically
favorably located on the surface, rather than being dispersed favorable than that under Pb-rich conditions.
into the perovskite interior. The surface mixed chlorine induced Another way to evaluate the carrier trapping and recombina-
a stronger band bending toward the TiO2 surface, leading to the tion is time-dependent density functional theory (TDDFT), which
electrons extraction with holes being partly pushed in the oppo- could depict the pattern of electron transport in the excited state
site direction. Feng et al.[200] investigated the interface between and calculated the lifetime of excited state instead of energy levels
CH3 NH3 Snx Pb1-x I3 and TiO2 , and found a charge accumulation as disscussed above. Liu and Prezhdo[205] studied the nonradia-
at the polar interface that is associated with the potential drop. tive electron-hole of Cl doped MAPbI3 by combination of nona-
Beside the interface between perovskite and other materi- diabatic molecular dynamics (NAMD) and TDDFT (Figure 9). By
als, the interface between different perovskite surfaces, namely decoherence correction to fewest switches surface hopping, they
the grain boundaries, were also investigated. It is known that obtained the charge carrier relaxation times of 42.8 and 124.6 ps
electronic properties of polycrystal perovskite are different from for MAPbI3 and MAPbI2 Cl, respectively, in agreement with avail-
those of single crystal, which can be reflected through the cal- able experimental data. The luminescence line width and vibra-
culated surface boundary. McKenna[201] found that the Σ3(111) tional modes were also calculated. They considered that two rea-
twin boundary was extremely stable in pure FAPbI3 with no elec- sons were responsible for the reduction of the recombination rate
tron or hole trapping states and relatively small barriers (<100 for the Cl doping. First, the doped chlorine atoms lead to the re-
meV) to transport of electrons and holes. However, in the FA/Cs duced nonadiabatic (NA) electron-phonon coupling. Since the Cl
and I/Br mixed-ion situation, the formation of an I- and Cs-rich atoms contribute little to the band edge states, and the wave func-
secondary phase was found at twin boundaries. The hole trap- tion overlap, which determines the NA coupling matrix element,
ping states from the segregated phase were likely to enhance the was reduce by the Cl doping. Second, due to the lighter atomic
electron–hole recombination and lead to the reduced open-circuit mass, Cl atoms introduce higher-frequency modes into the lat-
voltage in solar cell devices. Park et al.[202] studied the Σ5(130) tice, which accelerates loss of quantum coherence due to elastic
grain boundary model of CsPbI3 generated by rotation of per- electron-phonon scattering, then shortens the coherence time.
ovskite supercells with a formation energy of 1.45 eV nm−2 with It is noted that the organic MA cations, although do not con-
respect to orthorhombic CsPbI3 . Iodine interstitials in either the tribute to electron and hole charge states and influence the charge
negative or positive charge state were placed in the grain bound- recombination directly, their motions were coupled with the Pb-I
ary model to form split interstitials (Ii − ), I-I-I trimers (Ii + ), or Pb- vibrations, which determined the nonradiative electron–hole re-
I bonds with undercoordinated Pb atoms. The calculated charge combination. In the latter section, we will introduce how the dis-
transition levels showed that in the case of the split interstitials tribution of MA cations plays important roles in the solar cell per-
and I-I-I trimer, the gap states were introduced in the energy level, formance including charge recombination.
promoting carrier recombination. Halide mixing with Br allows
Ii − to be partially replaced by shallower Bri − and therefore par-
tially reduces the nonradiative recombination. 5. Distribution and Rotation of Organic Cations
In organic–inorganic hybrid lead iodide perovskites, the organic
cations, for example, the most representative MA+ , bring some
4.5. Charge Carrier Transport, Trapping, and Recombination unique features in comparison with the inorganic ones. The or-
ganic cations could be in irregular shapes instead of a sphere,
Trapping of charge carriers on their surfaces or grain boundaries and can exhibit orientations; the differently charged moieties of
is detrimental for PSC performance because of the nonradiative the cation can lead to dipoles, for which the directions are associ-
recombination. Energy levels calculation associated with the de- ated with the orientations and rotational motions of the cations.
fect states could evaluate the charge trapping. Du[132] calculated As mentioned above, the previous works conformed that the or-
the structures of iodine split interstitials on the equatorial site at ganic cations play indirect roles via the effects of the ionic ra-
+1 and −1 charge states. According to the energy levels calcula- dius, H bond interaction, and vibration coupling on the inorganic
tions, this defect is the only deep trap that leads to non-radiative lattice, next we discuss how they affect the electronic properties
recombination among native point defects including vacancies, and carrier transport in more significant ways. Some theoreti-
interstitials, and antisites. However, the carrier trapping at the io- cal works investigated the distribution and rotation of organic
dine interstitial involves kinetic barriers, leading to the reduction cations,[208–214] and found that rotational relaxation time in the
of capture cross section. The mixed halides with Cl introduced tetragonal and cubic phases are 32.7 and 15.2 ps, respectively.[208]
shallower acceptor Cli − , which is not an effective hole trap, in The rotational motions of MA cations suppressed the thermal
comparison with Ii − . Therefore, the mixed halides significantly conductivity.[213] The energy landscape and the band structure are
reduce carrier trapping and non-radiative recombination. Simi- relatively insensitive to the A-cation orientation.[210]
lar results were also found by Buin et al.[203] Uratani et al.[204] The distribution of MA could lead to ferroelectric or anti-
studied the defects on MAI, flat and vacant terminated (001) sur- ferroelectric structures as shown in Figure 10a. Frost et al.[215]
faces of MAPBI3 . They found that under the I-rich conditions suggested that the MAPbI3 exhibited spontaneous electric
excessive I atoms on flat and vacant surfaces lead to deep states. polarization of 38 𝜇C cm−2 , and the internal junctions of fer-
While under the Pb-rich conditions, excessive Pb atoms on both roelectric domains enhanced the separation of electron and
flat and vacant surfaces and I atom vacancies on vacant surfaces hole pairs, and reduced their recombination. The studies of
are responsible for the carrier trapping. The formation of carrier- ferroelectric MAPbI3 [216,217] showed that the ferroelectric struc-

Adv. Theory Simul. 2020, 3, 2000022 2000022 (15 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 9. a) Diagram of the energy levels and charge-transfer processes inside the perovskite material.[206] Reproduced with permission. Copyright 2017,
American Chemical Society. b) Electron–hole recombination dynamics across the MAPbI3 (100)/anatase (001) interface with and without doping by Cl,
Br, or Sn.[207] Reproduced with permission. Copyright 2015, American Chemical Society.

ture was the most stable, and the calculated polarization of MA cations. The structures with opposite dipole moments in-
ferroelectric structure by different research groups ranged from duces the fluctuation of the electrostatic potential between differ-
≈8 to ≈3 𝜇C cm−2 .[218–220] Frost et al.[215] address the origin ent layers. They further investigated the localization behavior of
of high-PCE of perovskites in the spontaneous electric polar- the band edge states along the xy direction and localization inter-
ization in addition to the basic optoelectronic properties, such val, which all could be controlled by the special MA distribution.
as band gap, optical absorption, carrier effective masses. They It is shown that the overlap integral between CBM and VBM was
considered that the presence of ferroelectric domains results in decreased for the structures with localized charge density, indi-
separation of photoexcited electron and hole pairs and reduction cating better separation of the holes and electrons. However, the
of recombination. The photoferroic effect could be exploited structures with localized charge density were slight less stable
to generate a higher open circuit voltage and contribute to than the structures with uniform MA orientations. It is possi-
the current–voltage hysteresis. They further investigated the ble to introduce different charge center instead of origin I atoms
equilibrium ferroelectric domain structure and polarizability by between the MA dipoles, making the structures with localized
Monte Carlo (MC) and FPMD.[221] They found that the domain charge density more stable. This could be achieved by the defect
structure is strongly dependent on the temperature (Figure 10b) engineering or doping. For instance, the I vacancy on the z axis of
and the applied electric field. They also performed Monte Carlo the octahedra stabilizes the structures with the opposite dipole;
simulation accounting for the dipole–dipole interaction energies in converse, the opposite orientation of MA dipole decreases the
as well as incorporating the effect of cage deformation (strain) formation energy of I vacancy on the z axis.
or lattice screening of the dipole–dipole interactions.[222] After Jankowska and Prezhdo[224] investigated the electron–
temperatures cooled from 750 to 50 K, the results illustrated that hole recombination for the ferroelectric layered perovskite,
the dipoles begin to spontaneously form either antiferroelectric (benzylammonium)2 PbCl4 , by combination of NAMD and
or ferroelectric domains dependent on the strain term energy time-domain DFT (TD-DFT). They found that the ferroelectric
(Figure 10c). Tong et al. demonstrated that the diffusion barriers phase exhibited slower electron–hole recombination than the
of the I− , MA+ , and Pb2+ ions are greatly affected by dipole mo- paraelectric phase due to the suppression coupling of the elec-
ments of the MA species, which further affected the hysteresis tronic subsystem to the high frequency phonons. Nan et al.[225]
induced by iodine ion/vacancy migration in antiferroelectric and investigated the many-body excited states of MAPbI3 perovskites
ferroelectric domains (Figure 10d). in the defect free state, point defect state, and Cl-doped state by
To fully understand the effect of the arrangement of the MA their developed TD-DFT method that treated electron–hole and
cations on the properties, Geng et al. constructed 18 initial tetrag- electron–nuclei interactions on the same footing. The charge
onal structures with different orientations of MA, which were di- density for the relaxed structure of the lowest excited state
vided to five groups according to projection of the dipole along showed that in the pristine MAPbI3 , the excited-state charge
the z direction (Figure 11).[223] It is interesting that the groups density distribution is dictated by the interplay between the
with the alternately opposite dipole moments of MA+ exhibit organic cations and the inorganic cages and largely affects by the
band edge localization in different layers along z direction and dynamic orientation of the MA cations. When a lead vacancy is
smaller band gaps. A new DOS peak from I atoms also appear located in the hole-rich region, the charge densities from delocal-
at the VBM of structures for MAPbI3 with oppositely oriented ization in the ground-state became rather localized in the excited
MA cations, but is absent for MAPbI3 with uniformly oriented state. Around the trapped hole, a small polaron exciton can be

Adv. Theory Simul. 2020, 3, 2000022 2000022 (16 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 10. a) Schematic of dipoles order and domains.[215] Reproduced with permission. Copyright 2014, American Chemical Society. b) Monte Carlo
simulation of MA+ dipole alignment at 0, 100, 300, and 1000 K.[221] Reproduced with permission. Copyright 2014, American Institute of Physics. c)
Monte Carlo simulation of MA+ reorientation after cooling to 50 K accounting for dipole–dipole interactions and cage deformation energies K = 25 and
100meV.[222] Reproduced with permission. Copyright 2015, Springer Nature. d) Microscopic mechanism of hysteresis induced by iodine ion/vacancy
migration in antiferroelectric and ferroelectric domains.[155] Reproduced with permission. Copyright 2017, American Chemical Society.

formed with a large polaronic relaxation energy Epol of ≈386 delocalized state, and then caused a release of transiently trapped
meV. In contrast, in the situation where the Pb vacancies are far charge carriers (Figure 12).
from the hole-rich regions, in both ground-state and excited-state
and excitations charge density are delocalized with a small Epol 6. Conclusion and Outlook
of ≈21 meV. They found that with a collective reorientation
of dipolar MA around the defects, the NH3 end of MA turned In above sections, the theoretical and computational progresses
toward the negatively charged defects, which is energetically on PSC have been presented and summarized. The theoretical
favorable. Then the localized deep trap states induced by the Pb approaches have achieved great successes in establishing the
vacancies are healed (delocalized). The rotational dynamics of relationship between the structures and properties, providing
the MA cations is at a picosecond scale, which is faster than the experimentally inaccessible information, providing atomistic
nonradiative decay process at a nanosecond scale. Therefore, understanding on the microscopic mechanisms for solar cell
they suggested that in the presence of Pb defects, the carrier applications. The results show that the ion migration in per-
diffusion proceeded via a sequence of events. The energy driven ovskites accounts for the J–V hysteresis. The degradation of
structural rearrangements triggered a dynamic crossover of perovskite involves a sequence of surface physical and chemical
excitations charge density, from a spatially confined state to a process and could be protected by passivation. The interface

Adv. Theory Simul. 2020, 3, 2000022 2000022 (17 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

Figure 11. The optimized structures of MAPBI3 a) with uniform MA distribution (A1) and b) with the opposite dipole moments (C1). The VBM partial
charge densities of c) A1 and d) C1. e) PDOS of I atoms and Pb atoms for the structures A1 (upper) and C1 (bottom). f) Electrostatic potential of 2*2*2
supercell of A1 and C1 with dash lines show the fluctuation trend.[223] Reproduced with permission. Copyright 2019, John Wiley & Sons, Inc.

Figure 12. Schematic picture of the MA orientation and Cl doping healing mechanisms.[225] Reproduced with permission. Copyright 2017, John Wiley
& Sons, Inc.

charge carrier recombination could be regulated by the surface machine learning is an effective approach to predict the new
structures. Theoretical investigations are also used to design and perovskite materials, which will greatly speed up the discovery
screen the new perovskite materials through high throughput the high-efficient perovskite materials. To end up this, we should
calculation and machine learning. not only put more attentions to develop the new algorithm and
However, some certain challenges and debates remain due database, but also explore the factors that affect the performance
to the limitation or inaccuracy of calculations. First of all, it of perovskites. For example, the intrinsic relationship and cou-
is still difficult to achieve the accurate electronic and atomic pling between the different factors that affect the properties,
structures with the cheap and reliable approaches, which is vital for example, the defect and the orientation of MA should be
to predict the new type of perovskite materials. Thus the new considered. Third, the trapping of charge carriers is vital for PSC
approach at the different levels should be developed. Second, performance because of the nonradiative recombination, while

Adv. Theory Simul. 2020, 3, 2000022 2000022 (18 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

the dynamics processes, especially the exited states, are still not [14] G. Xing, N. Mathews, S. Sun, S. S. Lim, Y. M. Lam, M. Grätzel, S.
fully understood. Thus more works should be carried out on Mhaisalkar, T. C. Sum, Science 2013, 342, 344.
the charge transfer and recombination with the first-principles [15] S. D. Stranks, G. E. Eperon, G. Grancini, C. Menelaou, M. J. P. Alco-
molecules dynamics, TDDFT, and NAMD. We hope that this cer, T. Leijtens, L. M. Herz, A. Petrozza, H. J. Snaith, Science 2013,
342, 341.
review could serve as a useful reference and inspire further
[16] Q. Dong, Y. Fang, Y. Shao, P. Mulligan, J. Qiu, L. Cao, J. Huang, Sci-
theoretical and computational studies on lead halide perovskites.
ence 2015, 347, 967.
[17] C. S. Ponseca, T. J. Savenije, M. Abdellah, K. Zheng, A. Yartsev, T.
Pascher, T. Harlang, P. Chabera, T. Pullerits, A. Stepanov, J. P. Wolf,
Acknowledgements V. Sundström, J. Am. Chem. Soc. 2014, 136, 5189.
[18] C. Wehrenfennig, G. E. Eperon, M. B. Johnston, H. J. Snaith, L. M.
This work was supported by the science challenge project (TZ2018004) Herz, Adv. Mater. 2014, 26, 1584.
and National Natural Science Foundation of China (Grant Nos. 11974037, [19] T. Matsui, J. Y. Seo, M. Saliba, S. M. Zakeeruddin, M. Grätzel, Adv.
21803078, 51861130360, U19300402). L.-M.L. was supported by the Fun- Mater. 2017, 29, 1606258.
damental Research Funds for the Central Universities. L.-M.L. also ac- [20] H. J. Snaith, J. Phys. Chem. Lett. 2013, 4, 3623.
knowledges the support of Newton Advanced Fellowship under the grant
[21] H. Cho, S. H. Jeong, M. H. Park, Y. H. Kim, C. Wolf, C. L. Lee, J. H.
number of NAFR1180242. This research was supported by a Tianhe-2JK
Heo, A. Sadhanala, N. Myoung, S. Yoo, S. H. Im, R. H. Friend, T. W.
computing time award at the Beijing Computational Science Research
Center (CSRC). Lee, Science 2015, 350, 1222.
[22] X. Lian, X. Wang, Y. Ling, E. Lochner, L. Tan, Y. Zhou, B. Ma, K. Han-
son, H. Gao, Adv. Funct. Mater. 2019, 29, 1807024.
[23] M. Yuan, L. N. Quan, R. Comin, G. Walters, R. Sabatini, O. Voznyy,
Conflict of Interest S. Hoogland, Y. Zhao, E. M. Beauregard, P. Kanjanaboos, Z. Lu, D.
H. Kim, E. H. Sargent, Nat. Nanotechnol. 2016, 11, 872.
The authors declare no conflict of interest. [24] L. N. Quan, F. P. G. de Arquer, R. P. Sabatini, E. H. Sargent, Adv.
Mater. 2018, 30, 1801996.
[25] X. K. Liu, F. Gao, J. Phys. Chem. Lett. 2018, 9, 2251.
Keywords [26] X. Xu, C. C. Chueh, P. Jing, Z. Yang, X. Shi, T. Zhao, L. Y. Lin, A. K. Y.
Jen, Adv. Funct. Mater. 2017, 27, 1701053.
density functional theory, perovskites, solar cells [27] Y. C. Kim, K. H. Kim, D. Y. Son, D. N. Jeong, J. Y. Seo, Y. S. Choi, I. T.
Han, S. Y. Lee, N. G. Park, Nature 2017, 550, 87.
Received: February 4, 2020 [28] H. Wei, D. DeSantis, W. Wei, Y. Deng, D. Guo, T. J. Savenije, L. Cao,
Revised: April 7, 2020 J. Huang, Nat. Mater. 2017, 16, 826.
Published online: June 26, 2020 [29] G. Xing, N. Mathews, S. S. Lim, N. Yantara, X. Liu, D. Sabba, M.
Grätzel, S. Mhaisalkar, T. C. Sum, Nat. Mater. 2014, 13, 476.
[30] J. C. Johnson, Z. Li, P. F. Ndione, K. Zhu, J. Mater. Chem. C 2016, 4,
4847.
[1] A. Kojima, K. Teshima, Y. Shirai, T. Miyasaka, J. Am. Chem. Soc. 2009, [31] C. Zhang, D. Sun, Z. V. Vardeny, Adv. Electron. Mater. 2017, 3,
131, 6050. 1600426.
[2] H. S. Kim, C. R. Lee, J. H. Im, K. B. Lee, T. Moehl, A. Marchioro, S. J. [32] W. A. Dunlap-Shohl, Y. Zhou, N. P. Padture, D. B. Mitzi, Chem. Rev.
Moon, R. Humphry-Baker, J. H. Yum, J. E. Moser, M. Grätzel, N.-G. 2019, 119, 3193.
Park, Sci. Rep. 2012, 2, 591. [33] A. O. El-Ballouli, O. M. Bakr, O. F. Mohammed, Chem. Mater. 2019,
[3] M. M. Lee, J. Teuscher, T. Miyasaka, T. N. Murakami, H. J. Snaith, 31, 6387.
Science 2012, 338, 643. [34] A. K. Jena, A. Kulkarni, T. Miyasaka, Chem. Rev. 2019, 119, 3036.
[4] N. J. Jeon, J. H. Noh, Y. C. Kim, W. S. Yang, S. Ryu, S. I. Seok, Nat. [35] J. Shamsi, A. S. Urban, M. Imran, L. De Trizio, L. Manna, Chem. Rev.
Mater. 2014, 13, 897. 2019, 119, 3296.
[5] J. Liu, Y. Shirai, X. Yang, Y. Yue, W. Chen, Y. Wu, A. Islam, L. Han, [36] L. N. Quan, B. P. Rand, R. H. Friend, S. G. Mhaisalkar, T. W. Lee, E.
Adv. Mater. 2015, 27, 4918. H. Sargent, Chem. Rev. 2019, 119, 7444.
[6] U.S. National Renewable Energy Laboratory (NREL), https://www. [37] L. M. Pazos-Outón, T. P. Xiao, E. Yablonovitch, J. Phys. Chem. Lett.
nrel.gov/pv/cell-efficiency.html (accessed: July 2019). 2018, 9, 1703.
[7] G. E. Eperon, S. D. Stranks, C. Menelaou, M. B. Johnston, L. M. [38] X. Cao, L. Zhi, Y. Jia, Y. Li, K. Zhao, X. Cui, L. Ci, D. Zhuang, J. Wei,
Herz, H. J. Snaith, Energy Environ. Sci. 2014, 7, 982. ACS Appl. Mater. Interfaces 2019, 11, 7639.
[8] Y. Fang, Q. Dong, Y. Shao, Y. Yuan, J. Huang, Nat. Photonics 2015, 9, [39] P. Schulz, ACS Energy Lett. 2018, 3, 1287.
679. [40] K. Liu, Y. Jiang, Y. Jiang, Y. Guo, Y. Liu, E. Nakamura, J. Am. Chem.
[9] J. Song, J. Li, X. Li, L. Xu, Y. Dong, H. Zeng, Adv. Mater. 2015, 27, Soc. 2019, 141, 1406.
7162. [41] Z. Xiao, Z. Song, Y. Yan, Adv. Mater. 2019, 31, 1803792.
[10] N. Wang, L. Cheng, R. Ge, S. Zhang, Y. Miao, W. Zou, C. Yi, Y. Sun, [42] Z. Xiao, Y. Yan, Adv. Energy Mater. 2017, 7, 1701136.
Y. Cao, R. Yang, Y. Wei, Q. Guo, Y. Ke, M. Yu, Y. Jin, Y. Liu, Q. Ding, [43] W. J. Yin, J. H. Yang, J. Kang, Y. Yan, S. H. Wei, J. Mater. Chem. A
D. Di, L. Yang, G. Xing, H. Tian, C. Jin, F. Gao, R. H. Friend, J. Wang, 2015, 3, 8926.
W. Huang, Nat. Photonics 2016, 10, 699. [44] J. Huang, Y. Yuan, Y. Shao, Y. Yan, Nat. Rev. Mater. 2017, 2, 17042.
[11] S. De Wolf, J. Holovsky, S. J. Moon, P. Löper, B. Niesen, M. Ledin- [45] X. Zhou, J. Jankowska, H. Dong, O. V. Prezhdo, J. Energy Chem. 2018,
sky, F. J. Haug, J. H. Yum, C. Ballif, J. Phys. Chem. Lett. 2014, 5, 27, 637.
1035. [46] D. Meggiolaro, F. De Angelis, ACS Energy Lett. 2018, 3, 2206.
[12] M. A. Green, A. Ho-Baillie, H. J. Snaith, Nat. Photonics 2014, 8, 506. [47] S. Yun, X. Zhou, J. Even, A. Hagfeldt, Angew. Chem., Int. Ed. 2017,
[13] M. D. McGehee, Nat. Mater. 2014, 13, 845. 56, 15806.

Adv. Theory Simul. 2020, 3, 2000022 2000022 (19 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

[48] R. F. Berger, Chem. - Eur. J. 2018, 24, 8708. [88] X. Qian, X. Gu, R. Yang, Appl. Phys. Lett. 2016, 108, 063902.
[49] W. J. Yin, T. Shi, Y. Yan, Adv. Mater. 2014, 26, 4653. [89] M. Wang, S. Lin, Adv. Funct. Mater. 2016, 26, 5297.
[50] J. M. Ball, M. M. Lee, A. Hey, H. J. Snaith, Energy Environ. Sci. 2013, [90] G. Liu, L. Kong, W. Yang, H. k. Mao, Mater. Today 2019, 27, 91.
6, 1739. [91] C. Grote, R. F. Berger, J. Phys. Chem. C 2015, 119, 22832.
[51] T. Baikie, Y. Fang, J. M. Kadro, M. Schreyer, F. Wei, S. G. Mhaisalkar, [92] M. Faghihnasiri, M. Izadifard, M. E. Ghazi, J. Phys. Chem. C 2017,
M. Graetzel, T. J. White, J. Mater. Chem. A 2013, 1, 5628. 121, 27059.
[52] C. C. Stoumpos, C. D. Malliakas, M. G. Kanatzidis, Inorg. Chem. [93] Y. Ying, X. Luo, H. Huang, J. Phys. Chem. C 2018, 122, 17718.
2013, 52, 9019. [94] Y. Huang, L. Wang, Z. Ma, F. Wang, J. Phys. Chem. C 2019, 123, 739.
[53] I. Chung, J. H. Song, J. Im, J. Androulakis, C. D. Malliakas, H. Li, A. [95] S. Liu, Y. Kim, L. Z. Tan, A. M. Rappe, Nano Lett. 2016, 16, 1663.
J. Freeman, J. T. Kenney, M. G. Kanatzidis, J. Am. Chem. Soc. 2012, [96] L. Zhang, W. Geng, C. j. Tong, X. Chen, T. Cao, M. Chen, Sci. Rep.
134, 8579. 2018, 8, 7760.
[54] Y. Kawamura, H. Mashiyama, K. Hasebe, J. Phys. Soc. Jpn. 2002, 71, [97] Z. Q. Ma, Y. Shao, P. K. Wong, X. Shi, H. Pan, J. Phys. Chem. C 2018,
1694. 122, 5844.
[55] V. M. Goldschmidt, Naturwissenschaften 1926, 14, 477. [98] J. Yin, P. Maity, M. De Bastiani, I. Dursun, O. M. Bakr, J. L. Brédas,
[56] G. Giorgi, K. Yamashita, J. Phys. Chem. Lett. 2016, 7, 888. O. F. Mohammed, Sci. Adv. 2017, 3, e1701793.
[57] J. Almutlaq, J. Yin, O. F. Mohammed, O. M. Bakr, J. Phys. Chem. Lett. [99] L. N. Quan, M. Yuan, R. Comin, O. Voznyy, E. M. Beauregard, S.
2018, 9, 4131. Hoogland, A. Buin, A. R. Kirmani, K. Zhao, A. Amassian, D. H. Kim,
[58] K. Chondroudis, D. B. Mitzi, Chem. Mater. 1999, 11, 3028. E. H. Sargent, J. Am. Chem. Soc. 2016, 138, 2649.
[59] M. Becker, T. Klüner, M. Wark, Dalton Trans. 2017, 46, 3500. [100] N. N. Som, P. M. W. P. Sampath, S. D. Dabhi, V. Mankad, S. Shinde,
[60] W. Li, Z. Wang, F. Deschler, S. Gao, R. H. Friend, A. K. Cheetham, M. L. C. Attygalle, P. K. Jha, Sol. Energy 2018, 173, 1315.
Nat. Rev. Mater. 2017, 2, 16099. [101] J. Yin, P. Maity, L. Xu, A. M. El-Zohry, H. Li, O. M. Bakr, J. L. Brédas,
[61] Q. Sun, W. J. Yin, J. Am. Chem. Soc. 2017, 139, 14905. O. F. Mohammed, Chem. Mater. 2018, 30, 8538.
[62] C. J. Bartel, C. Sutton, B. R. Goldsmith, R. Ouyang, C. B. Musgrave, [102] D. Wang, B. Wen, Y. N. Zhu, C. J. Tong, Z. K. Tang, L.-M. Liu, J. Phys.
L. M. Ghiringhelli, M. Scheffler, Sci. Adv. 2019, 5, eaav0693. Chem. Lett. 2017, 8, 876.
[63] E. Mosconi, A. Amat, M. K. Nazeeruddin, M. Grätzel, F. De Angelis, [103] M. I. Saidaminov, O. F. Mohammed, O. M. Bakr, ACS Energy Lett.
J. Phys. Chem. C 2013, 117, 13902. 2017, 2, 889.
[64] W. Geng, L. Zhang, Y. N. Zhang, W. M. Lau, L. M. Liu, J. Phys. Chem. [104] Y. Lin, Y. Bai, Y. Fang, Q. Wang, Y. Deng, J. Huang, ACS Energy Lett.
C 2014, 118, 19565. 2017, 2, 1571.
[65] T. Umebayashi, K. Asai, T. Kondo, A. Nakao, Phys. Rev. B 2003, 67, [105] H. Lin, C. Zhou, Y. Tian, T. Siegrist, B. Ma, ACS Energy Lett. 2018, 3,
155405. 54.
[66] A. Filippetti, A. Mattoni, Phys. Rev. B 2014, 89, 125203. [106] M. D. Smith, B. A. Connor, H. I. Karunadasa, Chem. Rev. 2019, 119,
[67] J. Even, L. Pedesseau, J. M. Jancu, C. Katan, J. Phys. Chem. Lett. 2013, 3104.
4, 2999. [107] F. Giustino, H. J. Snaith, ACS Energy Lett. 2016, 1, 1233.
[68] J. Heyd, G. E. Scuseria, M. Ernzerhof, J. Chem. Phys. 2006, 124, [108] G. Volonakis, A. A. Haghighirad, H. J. Snaith, F. Giustino, J. Phys.
219906. Chem. Lett. 2017, 8, 3917.
[69] J. Heyd, G. E. Scuseria, J. Chem. Phys. 2004, 121, 1187. [109] M. Saliba, T. Matsui, J. Y. Seo, K. Domanski, J. P. Correa-Baena, M.
[70] F. Brivio, K. T. Butler, A. Walsh, M. van Schilfgaarde, Phys. Rev. B K. Nazeeruddin, S. M. Zakeeruddin, W. Tress, A. Abate, A. Hagfeldt,
2014, 89, 155204. M. Grätzel, Energy Environ. Sci. 2016, 9, 1989.
[71] A. Amat, E. Mosconi, E. Ronca, C. Quarti, P. Umari, M. K. Nazeerud- [110] T. Duong, Y. Wu, H. Shen, J. Peng, X. Fu, D. Jacobs, E. C. Wang,
din, M. Grätzel, F. De Angelis, Nano Lett. 2014, 14, 3608. T. C. Kho, K. C. Fong, M. Stocks, E. Franklin, A. Blakers, N. Zin, K.
[72] M. R. Filip, F. Giustino, Phys. Rev. B 2014, 90, 245145. McIntosh, W. Li, Y. B. Cheng, T. P. White, K. Weber, K. Catchpole,
[73] F. Brivio, A. B. Walker, A. Walsh, APL Mater. 2013, 1, 042111. Adv. Energy Mater. 2017, 7, 1700228.
[74] P. P. Sun, Q. S. Li, L. N. Yang, Z. S. Li, Nanoscale 2016, 8, 1503. [111] Q. Xu, Z. Li, M. Liu, W. J. Yin, J. Phys. Chem. Lett. 2018, 9, 6948.
[75] Y. Ping, D. Rocca, G. Galli, Chem. Soc. Rev. 2013, 42, 2437. [112] A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S.
[76] P. Umari, E. Mosconi, F. De Angelis, Sci. Rep. 2015, 4, 4467. Cholia, D. Gunter, D. Skinner, G. Ceder, K. A. Persson, APL Mater.
[77] G. Giorgi, J. I. Fujisawa, H. Segawa, K. Yamashita, J. Phys. Chem. 2013, 1, 011002.
Lett. 2013, 4, 4213. [113] Z. Li, Q. Xu, Q. Sun, Z. Hou, W. J. Yin, Adv. Funct. Mater. 2019, 29,
[78] J. H. Lee, Z. Deng, N. C. Bristowe, P. D. Bristowe, A. K. Cheetham, 1807280.
J. Mater. Chem. C 2018, 6, 12252. [114] H. Park, R. Mall, F. H. Alharbi, S. Sanvito, N. Tabet, H. Bensmail, F.
[79] C. Motta, F. El-Mellouhi, S. Sanvito, Sci. Rep. 2015, 5, 12746. El-Mellouhi, Phys. Chem. Chem. Phys. 2019, 21, 1078.
[80] A. Filippetti, A. Mattoni, C. Caddeo, M. I. Saba, P. Delugas, Phys. [115] S. Lu, Q. Zhou, Y. Ouyang, Y. Guo, Q. Li, J. Wang, Nat. Commun.
Chem. Chem. Phys. 2016, 18, 15352. 2018, 9, 3405.
[81] G. Meng, Y. Feng, X. Song, Y. Shi, M. Ji, Y. Xue, C. Hao, J. Electroanal. [116] Machine Learning in Python, https://scikit-learn.org/stable/
Chem. 2018, 810, 11. modules/ensemble.html#regression (accessed: December 2007).
[82] S. Saha, T. P. Sinha, A. Mookerjee, Phys. Rev. B 2000, 62, 8828. [117] L. Li, Y. You, S. Hu, Y. Shi, G. Zhao, C. Chen, Y. Wang, A. Stroppa, W.
[83] M. Gajdoš, K. Hummer, G. Kresse, J. Furthmüller, F. Bechstedt, Phys. Ren, Appl. Phys. Lett. 2019, 114, 083102.
Rev. B 2006, 73, 045112. [118] O. Allam, C. Holmes, Z. Greenberg, K. C. Kim, S. S. Jang,
[84] A. B. Kuzmenko, Rev. Sci. Instrum. 2005, 76, 083108. ChemPhysChem 2018, 19, 2559.
[85] F. Brivio, J. M. Frost, J. M. Skelton, A. J. Jackson, O. J. Weber, M. T. [119] C. Kim, G. Pilania, R. Ramprasad, J. Phys. Chem. C 2016, 120, 14575.
Weller, A. R. Goñi, A. M. A. Leguy, P. R. F. Barnes, A. Walsh, Phys. [120] A. Jain, O. Voznyy, E. H. Sargent, J. Phys. Chem. C 2017, 121, 7183.
Rev. B 2015, 92, 144308. [121] S. Chakraborty, W. Xie, N. Mathews, M. Sherburne, R. Ahuja, M.
[86] M. A. Pérez-Osorio, R. L. Milot, M. R. Filip, J. B. Patel, L. M. Herz, Asta, S. G. Mhaisalkar, ACS Energy Lett. 2017, 2, 837.
M. B. Johnston, F. Giustino, J. Phys. Chem. C 2015, 119, 25703. [122] S. Körbel, M. A. L. Marques, S. Botti, J. Mater. Chem. C 2016, 4, 3157.
[87] Y. He, G. Galli, Chem. Mater. 2014, 26, 5394. [123] S. Körbel, M. A. L. Marques, S. Botti, J. Mater. Chem. A 2018, 6, 6463.

Adv. Theory Simul. 2020, 3, 2000022 2000022 (20 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

[124] Y. Li, K. Yang, Energy Environ. Sci. 2019, 12, 2233. [156] J. Brgoch, A. J. Lehner, M. Chabinyc, R. Seshadri, J. Phys. Chem. C
[125] H. C. Wang, P. Pistor, M. A. L. Marques, S. Botti, J. Mater. Chem. A 2014, 118, 27721.
2019, 7, 14705. [157] A. Suzuki, T. Oku, Appl. Surf. Sci. 2019, 483, 912.
[126] H. Hafiz, A. I. Khair, H. Choi, A. Mueen, A. Bansil, S. Eidenbenz, J. [158] F. El-Mellouhi, S. N. Rashkeev, A. Marzouk, L. Kabalan, A. Belaidi,
Wills, J. X. Zhu, A. V. Balatsky, T. Ahmed, npj Comput. Mater. 2018, B. Merzougui, N. Tabet, F. H. Alharbi, J. Mater. Chem. C 2019, 7,
4, 63. 5299.
[127] S. Heo, G. Seo, Y. Lee, D. Lee, M. Seol, J. Lee, J. B. Park, K. Kim, D. [159] Z. Li, M. Yang, J. S. Park, S. H. Wei, J. J. Berry, K. Zhu, Chem. Mater.
J. Yun, Y. S. Kim, J. K. Shin, T. K. Ahn, M. K. Nazeeruddin, Energy 2016, 28, 284.
Environ. Sci. 2017, 10, 1128. [160] W. J. Yin, Y. Yan, S. H. Wei, J. Phys. Chem. Lett. 2014, 5, 3625.
[128] T. Shi, W. J. Yin, F. Hong, K. Zhu, Y. Yan, Appl. Phys. Lett. 2015, 106, [161] S. Colella, E. Mosconi, P. Fedeli, A. Listorti, F. Gazza, F. Orlandi, P.
103902. Ferro, T. Besagni, A. Rizzo, G. Calestani, G. Gigli, F. De Angelis, R.
[129] L. Etgar, P. Gao, Z. Xue, Q. Peng, A. K. Chandiran, B. Liu, M. K. Mosca, Chem. Mater. 2013, 25, 4613.
Nazeeruddin, M. Grätzel, J. Am. Chem. Soc. 2012, 134, 17396. [162] B. Walker, G. H. Kim, J. Y. Kim, Adv. Mater. 2019, 31, 1807029.
[130] W. A. Laban, L. Etgar, Energy Environ. Sci. 2013, 6, 3249. [163] G. R. Berdiyorov, F. El-Mellouhi, M. E. Madjet, F. H. Alharbi, F. M.
[131] J. You, Z. Hong, Y. Yang, Q. Chen, M. Cai, T. B. Song, C. C. Chen, S. Peeters, S. Kais, Sol. Energy Mater. Sol. Cells 2016, 148, 2.
Lu, Y. Liu, H. Zhou, Y. Yang, ACS Nano 2014, 8, 1674. [164] J. Hieulle, X. Wang, C. Stecker, D. Y. Son, L. Qiu, R. Ohmann, L.
[132] M. H. Du, J. Mater. Chem. A 2014, 2, 9091. K. Ono, A. Mugarza, Y. Yan, Y. Qi, J. Am. Chem. Soc. 2019, 141,
[133] J. Kim, C. H. Chung, K. H. Hong, Phys. Chem. Chem. Phys. 2016, 18, 3515.
27143. [165] C. H. Hendon, R. X. Yang, L. A. Burton, A. Walsh, J. Mater. Chem. A
[134] J. Kang, L. W. Wang, J. Phys. Chem. Lett. 2017, 8, 489. 2015, 3, 9067.
[135] J. Kim, S. H. Lee, J. H. Lee, K. H. Hong, J. Phys. Chem. Lett. 2014, 5, [166] Y. Y. Sun, M. L. Agiorgousis, P. Zhang, S. Zhang, Nano Lett. 2015,
1312. 15, 581.
[136] M. L. Agiorgousis, Y. Y. Sun, H. Zeng, S. Zhang, J. Am. Chem. Soc. [167] S. Chatterjee, A. J. Pal, J. Mater. Chem. A 2018, 6, 3793.
2014, 136, 14570. [168] A. E. Maughan, A. M. Ganose, D. O. Scanlon, J. R. Neilson, Chem.
[137] L. D. Whalley, R. Crespo-Otero, A. Walsh, ACS Energy Lett. 2017, 2, Mater. 2019, 31, 1184.
2713. [169] X. G. Zhao, J. H. Yang, Y. Fu, D. Yang, Q. Xu, L. Yu, S. H. Wei, L.
[138] K. Miyano, M. Yanagida, N. Tripathi, Y. Shirai, J. Phys. Chem. Lett. Zhang, J. Am. Chem. Soc. 2017, 139, 2630.
2016, 7, 2240. [170] E. Mosconi, P. Umari, F. De Angelis, J. Mater. Chem. A 2015, 3,
[139] H. J. Snaith, A. Abate, J. M. Ball, G. E. Eperon, T. Leijtens, N. K. 9208.
Noel, S. D. Stranks, J. T. W. Wang, K. Wojciechowski, W. Zhang, J. [171] L. Z. Wang, Y. Q. Zhao, B. Liu, L. J. Wu, M. Q. Cai, Phys. Chem. Chem.
Phys. Chem. Lett. 2014, 5, 1511. Phys. 2016, 18, 22188.
[140] A. Dualeh, T. Moehl, N. Tétreault, J. Teuscher, P. Gao, M. K. [172] Y. K. Jung, J. H. Lee, A. Walsh, A. Soon, Chem. Mater. 2017, 29, 3181.
Nazeeruddin, M. Grätzel, ACS Nano 2014, 8, 362. [173] D. Umadevi, G. W. Watson, ACS Omega 2019, 4, 5661.
[141] E. L. Unger, E. T. Hoke, C. D. Bailie, W. H. Nguyen, A. R. Bowring, T. [174] Z. Xiao, K. Z. Du, W. Meng, D. B. Mitzi, Y. Yan, Angew. Chem. 2017,
Heumüller, M. G. Christoforo, M. D. McGehee, Energy Environ. Sci. 129, 12275.
2014, 7, 3690. [175] Z. Xiao, K. Z. Du, W. Meng, J. Wang, D. B. Mitzi, Y. Yan, J. Am. Chem.
[142] B. Chen, M. Yang, S. Priya, K. Zhu, J. Phys. Chem. Lett. 2016, 7, 905. Soc. 2017, 139, 6054.
[143] Z. Xiao, Y. Yuan, Y. Shao, Q. Wang, Q. Dong, C. Bi, P. Sharma, A. [176] I. Zarazua, J. Bisquert, G. Garcia-Belmonte, J. Phys. Chem. Lett. 2016,
Gruverman, J. Huang, Nat. Mater. 2015, 14, 193. 7, 525.
[144] E. J. Juarez-Perez, R. S. Sanchez, L. Badia, G. Garcia-Belmonte, Y. S. [177] P. Schulz, D. Cahen, A. Kahn, Chem. Rev. 2019, 119, 3349.
Kang, I. Mora-Sero, J. Bisquert, J. Phys. Chem. Lett. 2014, 5, 2390. [178] L. K. Ono, Y. Qi, J. Phys. Chem. Lett. 2016, 7, 4764.
[145] Y. Yuan, Q. Wang, Y. Shao, H. Lu, T. Li, A. Gruverman, J. Huang, Adv. [179] J. Haruyama, K. Sodeyama, L. Han, Y. Tateyama, J. Phys. Chem. Lett.
Energy Mater. 2016, 6, 1501803. 2014, 5, 2903.
[146] J. M. Azpiroz, E. Mosconi, J. Bisquert, F. De Angelis, Energy Environ. [180] W. Geng, C. J. Tong, Z. K. Tang, C. Yam, Y. N. Zhang, W. M. Lau, L.
Sci. 2015, 8, 2118. M. Liu, J. Materiomics 2015, 1, 213.
[147] J. Haruyama, K. Sodeyama, L. Han, Y. Tateyama, J. Am. Chem. Soc. [181] Y. Wang, B. G. Sumpter, J. Huang, H. Zhang, P. Liu, H. Yang, H.
2015, 137, 10048. Zhao, J. Phys. Chem. C 2015, 119, 1136.
[148] C. Eames, J. M. Frost, P. R. F. Barnes, B. C. O’Regan, A. Walsh, M. S. [182] L. She, M. Liu, D. Zhong, ACS Nano 2016, 10, 1126.
Islam, Nat. Commun. 2015, 6, 7497. [183] X. Che, B. Traore, C. Katan, M. Kepenekian, J. Even, Phys. Chem.
[149] S. Meloni, T. Moehl, W. Tress, M. Franckevičius, M. Saliba, Y. H. Lee, Chem. Phys. 2018, 20, 9638.
P. Gao, M. K. Nazeeruddin, S. M. Zakeeruddin, U. Rothlisberger, M. [184] C. C. Boyd, R. Cheacharoen, T. Leijtens, M. D. McGehee, Chem. Rev.
Graetzel, Nat. Commun. 2016, 7, 10334. 2019, 119, 3418.
[150] J. H. Yang, W. J. Yin, J. S. Park, S. H. Wei, J. Mater. Chem. A 2016, 4, [185] I. Deretzis, E. Smecca, G. Mannino, A. La Magna, T. Miyasaka, A.
13105. Alberti, J. Phys. Chem. Lett. 2018, 9, 3000.
[151] T. Zhang, H. Chen, Y. Bai, S. Xiao, L. Zhu, C. Hu, Q. Xue, S. Yang, [186] E. Mosconi, J. M. Azpiroz, F. De Angelis, Chem. Mater. 2015, 27,
Nano Energy 2016, 26, 620. 4885.
[152] E. Mosconi, D. Meggiolaro, H. J. Snaith, S. D. Stranks, F. De Angelis, [187] C. J. Tong, W. Geng, Z. K. Tang, C. Y. Yam, X. L. Fan, J. Liu, W. M. Lau,
Energy Environ. Sci. 2016, 9, 3180. L. M. Liu, J. Phys. Chem. Lett. 2015, 6, 3289.
[153] P. Delugas, C. Caddeo, A. Filippetti, A. Mattoni, J. Phys. Chem. Lett. [188] N. Z. Koocher, D. Saldana-Greco, F. Wang, S. Liu, A. M. Rappe, J.
2016, 7, 2356. Phys. Chem. Lett. 2015, 6, 4371.
[154] D. A. Egger, L. Kronik, A. M. Rappe, Angew. Chem., Int. Ed. 2015, 54, [189] Y. Liu, K. Palotas, X. Yuan, T. Hou, H. Lin, Y. Li, S. T. Lee, ACS Nano
12437. 2017, 11, 2060.
[155] C. J. Tong, W. Geng, O. V. Prezhdo, L. M. Liu, ACS Energy Lett. 2017, [190] L. Zhang, M. G. Ju, W. Liang, Phys. Chem. Chem. Phys. 2016, 18,
2, 1997. 23174.

Adv. Theory Simul. 2020, 3, 2000022 2000022 (21 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 25130390, 2020, 9, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adts.202000022 by Egyptian National Sti. Network (Enstinet), Wiley Online Library on [05/05/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.advtheorysimul.com

[191] N. Aristidou, C. Eames, I. Sanchez-Molina, X. Bu, J. Kosco, M. S. [209] A. L. Montero-Alejo, E. Menéndez-Proupin, D. Hidalgo-Rojas, P.
Islam, S. A. Haque, Nat. Commun. 2017, 8, 15218. Palacios, P. Wahnón, J. C. Conesa, J. Phys. Chem. C 2016, 120,
[192] F. Wang, W. Geng, Y. Zhou, H. H. Fang, C. J. Tong, M. A. Loi, L. M. 7976.
Liu, N. Zhao, Adv. Mater. 2016, 28, 9986. [210] J. S. Bechtel, R. Seshadri, A. V. der Ven, J. Phys. Chem. C 2016, 120,
[193] L. Zhang, F. Yu, L. Chen, J. Li, Appl. Surf. Sci. 2018, 443, 176. 12403.
[194] A. Torres, L. G. C. Rego, J. Phys. Chem. C 2014, 118, 26947. [211] A. Mattoni, A. Filippetti, M. I. Saba, P. Delugas, J. Phys. Chem. C
[195] W. Geng, C. J. Tong, J. Liu, W. Zhu, W. M. Lau, L. M. Liu, Sci. Rep. 2015, 119, 17421.
2016, 6, 20131. [212] A. Mattoni, A. Filippetti, M. I. Saba, C. Caddeo, P. Delugas, J. Phys.
[196] C. Quarti, F. De Angelis, D. Beljonne, Chem. Mater. 2017, 29, 958. Chem. Lett. 2016, 7, 529.
[197] E. Mosconi, E. Ronca, F. De Angelis, J. Phys. Chem. Lett. 2014, 5, [213] T. Hata, G. Giorgi, K. Yamashita, Nano Lett. 2016, 16, 2749.
2619. [214] J. Even, L. Pedesseau, C. Katan, J. Phys. Chem. C 2014, 118, 11566.
[198] S. Colella, E. Mosconi, G. Pellegrino, A. Alberti, V. L. P. Guerra, S. [215] J. M. Frost, K. T. Butler, F. Brivio, C. H. Hendon, M. van Schilfgaarde,
Masi, A. Listorti, A. Rizzo, G. G. Condorelli, F. De Angelis, G. Gigli, A. Walsh, Nano Lett. 2014, 14, 2584.
J. Phys. Chem. Lett. 2014, 5, 3532. [216] G. Volonakis, F. Giustino, J. Phys. Chem. Lett. 2015, 6, 2496.
[199] V. Roiati, E. Mosconi, A. Listorti, S. Colella, G. Gigli, F. De Angelis, [217] S. Ghosh, D. Di Sante, A. Stroppa, J. Phys. Chem. Lett. 2015, 6, 4553.
Nano Lett. 2014, 14, 2168. [218] Z. Fan, J. Xiao, K. Sun, L. Chen, Y. Hu, J. Ouyang, K. P. Ong, K. Zeng,
[200] H. J. Feng, T. R. Paudel, E. Y. Tsymbal, X. C. Zeng, J. Am. Chem. Soc. J. Wang, J. Phys. Chem. Lett. 2015, 6, 1155.
2015, 137, 8227. [219] A. Stroppa, C. Quarti, F. De Angelis, S. Picozzi, J. Phys. Chem. Lett.
[201] K. P. McKenna, ACS Energy Lett. 2018, 3, 2663. 2015, 6, 2223.
[202] J. S. Park, J. Calbo, Y. K. Jung, L. D. Whalley, A. Walsh, ACS Energy [220] A. Filippetti, P. Delugas, M. I. Saba, A. Mattoni, J. Phys. Chem. Lett.
Lett. 2019, 4, 1321. 2015, 6, 4909.
[203] A. Buin, P. Pietsch, J. Xu, O. Voznyy, A. H. Ip, R. Comin, E. H. Sargent, [221] J. M. Frost, K. T. Butler, A. Walsh, APL Mater. 2014, 2, 081506.
Nano Lett. 2014, 14, 6281. [222] A. M. A. Leguy, J. M. Frost, A. P. McMahon, V. G. Sakai, W. Kockel-
[204] H. Uratani, K. Yamashita, J. Phys. Chem. Lett. 2017, 8, 742. mann, C. Law, X. Li, F. Foglia, A. Walsh, B. C. O’Regan, J. Nelson, J.
[205] J. Liu, O. V. Prezhdo, J. Phys. Chem. Lett. 2015, 6, 4463. T. Cabral, P. R. F. Barnes, Nat. Commun. 2015, 6, 7124.
[206] W. Li, J. Liu, F. Q. Bai, H. X. Zhang, O. V. Prezhdo, ACS Energy Lett. [223] W. Geng, Q. Hu, C. J. Tong, Z. K. Tang, L. M. Liu, Adv. Theory Simul.
2017, 2, 1270. 2019, 2, 1900041.
[207] R. Long, O. V. Prezhdo, ACS Nano 2015, 9, 11143. [224] J. Jankowska, O. V. Prezhdo, J. Phys. Chem. Lett. 2017, 8, 812.
[208] C. Goehry, G. A. Nemnes, A. Manolescu, J. Phys. Chem. C 2015, 119, [225] G. Nan, X. Zhang, M. Abdi-Jalebi, Z. Andaji-Garmaroudi, S. D.
19674. Stranks, G. Lu, D. Beljonne, Adv. Energy Mater. 2018, 8, 1702754.

Adv. Theory Simul. 2020, 3, 2000022 2000022 (22 of 22) © 2020 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim