betaFeedback

WorksPhysics and Astronomy

Spin-Polarized Electronic Structure and Chemical Bonding Data for 2,500+ Halide Double Perovskites

Abstract

Halide double perovskites (A2_2BB'X6_6) are a long-known class of materials that has recently been rediscovered for diverse applications, including photovoltaics, photocatalysis, and radiation detection. Their doubled unit cell provides immense chemical tunability, allowing the incorporation of magnetic ions and enabling access to a wide range of electronic-structure features, including different band-edge characters, alignments, and symmetries. Magnetic elements may further introduce spin degrees of freedom and magnetic behaviour, thereby broadening the functional landscape of these compounds. Here, we present the first comprehensive database of spin-polarised electronic-structure data for all halide double perovskites predicted to be stable by the recently introduced ττ tolerance factor by Bartel et al. The dataset focuses on the Cs2_2BB'X6_6 family, with X = I, Br, Cl, and F, and includes density of states (DOS) for >>2,500 compounds, calculated using hybrid-functional density functional theory. Among these, 719 compounds exhibit band gaps in the visible range and 118 display half-metallic character. In addition, we provide chemical-bonding analysis using lobster, which provides insights into orbital interactions across the dataset. To facilitate exploration, we further offer UMAP-based visualisations and an interactive app for systematic investigation of chemical composition, electronic structure, and magnetic properties.

Background & Summary

Halide perovskites, with the chemical formula ABX3_3 (where A and B are mono- and divalent cations, and X is a halide), crystallize in a network of corner-sharing BX6_6 octahedra. Many halide perovskites show great potential for energy and quantum applications due to their tunable properties, facile synthesis and pronounced light-matter interactions. For example, the organic-inorganic perovskite methylammonium lead iodide (MAPbI3_3)[1] has favorable optoelectronic properties and is widely used in single and tandem solar cells with record power conversion efficiencies.[2] Although most research in photovoltaics has focused on MAPbI3_3 and related mixed-site compositions,[3, 4, 5, 6] the halide perovskite family encompasses a vast number of materials, including halide double perovskites,[7] vacancy ordered perovskites,[8] dimensionally reduced perovskites,[10, 11, 12, 13] [9] and other perovskite-like structures with edge- and/or face-sharing octahedral connectivity.[14, 15] This enormous structural and compositional flexibility opens a broad material design space, allowing access to bespoke properties for a wide range of applications.

In this space, halide double perovskites (HDPs, also known as elpasolites) are a particularly interesting subclass because they accommodate two distinct B-site cations - or, in vacancy-ordered compositions, a single B-site cation alternating with a vacancy.[7] The doubling of the unit cell, leading to a chemical formula of A2BB′X6\mathrm{A}_2\mathrm{BB}^{\prime}\mathrm{X}_6, greatly expands the number of possible perovskite compositions since the two B sites can host elements with oxidation states from +I+\mathrm{I} to +IV+\mathrm{IV}.[16, 17] HDPs were first discovered in the late 19th century, and many stable compositions have since been identified and synthesized. They typically crystallize in a cubic structure (space group Fm3ˉm\mathrm{F}m\bar{3}m). With the advent of halide perovskites for photovoltaic applications, renewed interest in HDPs has led to the discovery of many new compositions, for example, Cs2AgBiBr6\mathrm{Cs}_2\mathrm{AgBiBr}_6, which is highly stable and has favorable optoelectronic properties.[18, 19, 20] Beyond photovoltaics, the compositional freedom of HDPs allows broad tuning of structural, (opto-)electronic, and magnetic properties.[21, 22, 23, 24, 25] This has generated interest in HDPs for applications ranging from photocatalysis[26] to humidity sensors[27] and X-ray detectors.[28] Furthermore, some HDPs feature robust excitonic properties due to chemical confinement effects,[29, 30, 31, 32, 33] suggesting potential for quantum information and sensing applications.

Incorporating transition- and rare-earth elements at the B sites opens a largely unexplored dimension in HDP research, introducing magnetic degrees of freedom alongside structural and electronic ones.[25, 34, 35] Although rare-earth HDPs have been studied since the 1960s, only a few compositions are known[7] and their electronic properties remain poorly understood. With the spin magnetic moment as an additional parameter, the tunability of HDPs expands, for example, allowing for spin-engineering of charge-carrier lifetimes and spintronic applications. Recent work has explored the incorporation of 3d3d transition metals in HDPs, leading to the discovery of Cs2AgFeCl6\mathrm{Cs}_2\mathrm{AgFeCl}_6, which is antiferromagnetic at very low temperatures.[36] Furthermore, first-principles studies using density functional theory (DFT) by Klarbring et al. and Singh et al. have focused on magnetism in HDPs, reporting a number of new compositions but restricting their high-throughput searches to transition metals from the 3d3d, 4d4d, and 5d5d series.[34, 35] Other computational studies on HDPs included a broader range of compositions but did not consider spin-polarized electronic structures or magnetic properties. [16, 17, 37]

To bring this material space within reach of systematic exploration, we present the first mapping of spin states and spin-polarized electronic structures across the predicted HDP landscape. To achieve this, we have implemented an automated workflow that uses DFT with hybrid-functional accuracy and generated a database of spin-polarized electronic-structure data, including an in-depth bonding analysis. One of us has recently released a quantum chemical bonding database for inorganic materials, [38] and used an extended database to demonstrate that bonding descriptors posses strong predictive power for phonon-related and elastic properties within machine learning frameworks. [39] Bonding analysis has also been show to be predictive for optoelectronic properties, [40] and has previously been applied in the context of Halide (Double) Perovskites. [41, 42, 43] Together, these results highlight the benefit of chemical bonding data in data-driven materials design.

Our database contains >2500>2500 compositions whose properties can be explored via an interactive application. We use the dimensionality reduction technique UMAP to visualize composition–property correlations in this dataset and show that these correlations are encoded in the spin-polarized projected density of states of the materials. With this, our database provides an overview of electronic properties, including spin polarization and magnetic moments, across stable HDP compositions and can be used for materials discovery, the design of new functional HDPs with tailored properties, and as an input for machine-learning studies.

Methods

Compound Selection

To map the full chemical landscape of HDPs, we first enumerate all charge-neutral HDP compositions. Including vacancy-ordered double perovskites, the 6×(−1)6 \times(-1) charge of the halide anions admits five possible cation charge combinations, illustrated in Figure 1a. Combining

Schematic representation of cationic oxidation state combinations, periodic table of predicted stable compositions, and HDP composition counts

Figure 1. (a) Schematic representation of cationic oxidation state combinations that result in charge-neutral A2BB′X6\mathrm{A}_2\mathrm{BB'}\mathrm{X}_6, where X=I−\mathrm{X}=\mathrm{I}^{-}, Br−\mathrm{Br}^{-}, Cl−\mathrm{Cl}^{-}, F−\mathrm{F}^{-}. Halides are not shown. (b) Periodic table indicating which elements are featured in at least one predicted stable composition as A-site (with green upper left corner) or as B-site (by blue lower right corner). The X-sites are indicated with pink boxes. (c) For the compounds studied here (i.e., those with A=Cs+\mathrm{A}=\mathrm{Cs}^{+}), the bar plot shows the number of HDPs for each combination of blocks of the periodic table that the two B-sites originate from (Vac = vacancy-ordered). Due to the lack of available PAWs in vasp for elements from berkelium onward, a discrepancy arises between predicted (lower-opacity bars, larger total number) and simulated (higher-opacity bars, smaller total number) HDP compositions.

all elements with viable oxidation states under these constraints, yields 47,013 candidate compositions.

We then use a tolerance-factor-based assessment as stability criterion. For perovskites, several tolerance-factor definitions exist, [44, 45] based on the original definition by Goldschmidt. [46] The Goldschmidt tolerance factor tt depends only on the ionic radii of the different species. The predictive power of tt is, however, severely limited for the heavier halides. [37] To overcome this limitation, Bartel et al. introduced a new tolerance factor τ\tau, using SISSO (sure independence screening and sparsifying operator) [47] on a database of experimentally realized perovskites. [37]

τ=rXrB−nA(nA−rA/rBln⁡rA/rB)(1)\tau= \frac{r_X}{r_B} - n_A\left(n_A - \frac{r_A/r_B}{\ln r_A/r_B}\right) \tag*{(1)}

where rir_i is the ionic radius of the ion on site ii and nAn_A is the oxidation state of the A-site cation. Shannon’s revised atomic radii [48] were used, which were accessed through the Mendeleev python package. [49] Lower τ\tau values are correlated with higher decomposition enthalpies, and Ref. 37 found that τ≤4.18\tau\le4.18 predicted perovskite stability with a 91% accuracy across all considered double perovskite compositions, and 84% accuracy when restricted to HDPs. [50]

For double perovskites, Ref. 37 computes rBr_B in Equation (1) as the arithmetic mean of the B and B’ ionic radii. This introduces an ambiguity for vacancy-ordered compositions: Basing rBr_B on the B ionic radius alone underestimates the number of stable vacancy-ordered compositions, while averaging rBr_B with a vacancy radius of zero overestimates it. We therefore apply a separate τ\tau cutoff to vacancy-ordered HDPs, set as τ=5.46\tau= 5.46, which is the highest τ\tau among vacancy-ordered HDPs in the experimental data of Ref. 51.

Besides vacancy-ordered HDPs, cryolites [7, 52] are another HDP subclass that was not considered by Bartel et al.. In cryolites the A-site cation also occupies one of the B-sites, resulting in chemical formula A3_3BX6_6. In our database cryolites are represented as Cs2_2CsBX6_6. Apart from the inclusion of vacancy-ordered HDPs with a separate τ\tau cutoff value and cryolites, our procedure for selecting compositions is the same as in Ref. 37. Running this procedure on the 47,013 potential compositions resulted in 11,823 predicted stable compositions. Figure 1b shows which elements are featured in at least one predicted stable HDP composition. Since A-site choice has limited direct influence on (opto-)electronic properties, we fix the A-site cation and vary B-site and X-site composition. Among the available A-site cations, cesium permits the largest set of charge-balanced compositions. We therefore restrict the database to Cs-based HDPs. Under this constraint, 2,953 HDP compositions are predicted to be stable. The lack of VASP PAW (Projector Augmented Wave) pseudopotentials for elements with Z≥97Z \ge97 further reduces this number to 2,680 compositions. Figure 1c provides an overview of compositional diversity in the dataset by showing the B-cation pairings for each halide of the HDP, grouped by the periodic-table blocks from which the B cations originate.

Computational Workflow

Our automated workflow is schematically depicted in Figure 2 and consists of four DFT steps using the VASP program package [53, 54, 55] with PAW pseudopotentials, [56, 57] followed by a projection onto atomic orbitals using LOBSTER. [58, 59, 60, 61, 62] In general, the recommended PAW potentials included in VASP 6.4 (potpaw.64) are used. For several lanthanides, the recommended PAW potentials treat the 4f4f electrons as frozen-core states. We refer to these as reduced-valence PAWs (RPAWs) and use them for structural relaxation to improve numerical stability. For subsequent electronic-structure calculations, where spin-related effects are of interest, we instead used PAW potentials that include 4f4f electrons as valence states. For cesium, calculations with the recommended potential (Cs_sv) failed to converge. We therefore used the Cs_sv_GW potential, which provides a more accurate description of high-energy states via additional projectors. An overview of the PAW potentials used is provided in Table 6.

Overview of the automated workflow developed for this work

Figure 2. Overview of the automated workflow developed for this work.

ElementRPAWconfig.PAWconfig.LCAO basis
LiLi_sv1s22s11s^{2}2s^{1}1s 2s1s\ 2s
BeBe2s1.992p0.012s^{1.99}2p^{0.01}2s2s
BB2s22p12s^{2}2p^{1}2p 2s2p\ 2s
NN2s22p32s^{2}2p^{3}2p 2s2p\ 2s
FF2s22p52s^{2}2p^{5}2p 2s2p\ 2s
NaNa_pv2p63s12p^{6}3s^{1}2p 3s2p\ 3s
MgMg3s23s^{2}3s3s
AlAl3s23p13s^{2}3p^{1}3p 3s3p\ 3s
SiSi3s23p23s^{2}3p^{2}3p 3s3p\ 3s
PP3s23p33s^{2}3p^{3}3p 3s3p\ 3s
ClCl3s23p53s^{2}3p^{5}3p 3s3p\ 3s
KK_sv3s23p64s13s^{2}3p^{6}4s^{1}3p 3s 4s3p\ 3s\ 4s
CaCa_sv3s23p64s23s^{2}3p^{6}4s^{2}3p 3s 4s3p\ 3s\ 4s
ScSc_sv3s23p63d24s13s^{2}3p^{6}3d^{2}4s^{1}3d 3p 3s 4s3d\ 3p\ 3s\ 4s
TiTi_sv3s23p63d34s13s^{2}3p^{6}3d^{3}4s^{1}3d 3p 3s 4s3d\ 3p\ 3s\ 4s
VV_sv3s23p63d44s13s^{2}3p^{6}3d^{4}4s^{1}3d 3p 3s 4s3d\ 3p\ 3s\ 4s
CrCr_pv3p63d54s13p^{6}3d^{5}4s^{1}3d 3p 4s3d\ 3p\ 4s
MnMn_pv3p63d64s13p^{6}3d^{6}4s^{1}3d 3p 4s3d\ 3p\ 4s
FeFe3d74s13d^{7}4s^{1}3d 4s3d\ 4s
CoCo3d84s13d^{8}4s^{1}3d 4s3d\ 4s
NiNi3d94s13d^{9}4s^{1}3d 4s3d\ 4s
CuCu3d104s13d^{10}4s^{1}3d 4s3d\ 4s
ElementRPAWconfig.PAWconfig.LCAO basis
ZnZn3d104s23d^{10}4s^23d 4s3d\ 4s
GaGa_d3d104s24p13d^{10}4s^24p^13d 4p 4s3d\ 4p\ 4s
GeGe_d3d104s24p23d^{10}4s^24p^23d 4p 4s3d\ 4p\ 4s
AsAs4s24p34s^24p^34p 4s4p\ 4s
SeSe4s24p44s^24p^44p 4s4p\ 4s
BrBr4s24p54s^24p^54p 4s4p\ 4s
RbRb_sv4s24p64d0.0015s0.9994s^24p^64d^{0.001}5s^{0.999}4p 4s4p\ 4s
SrSr_sv4s24p64d0.0015s1.9994s^24p^64d^{0.001}5s^{1.999}4p 4s 5s4p\ 4s\ 5s
YY_sv4s24p64d25s14s^24p^64d^25s^14d 4p 4s 5s4d\ 4p\ 4s\ 5s
ZrZr_sv4s24p64d35s14s^24p^64d^35s^14d 4p 4s 5s4d\ 4p\ 4s\ 5s
NbNb_sv4s24p64d45s14s^24p^64d^45s^14d 4p 4s 5s4d\ 4p\ 4s\ 5s
MoMo_sv4s24p64d55s14s^24p^64d^55s^14d 4p 4s 5s4d\ 4p\ 4s\ 5s
TcTc_pv4p64d65s14p^64d^65s^14d 4p 5s4d\ 4p\ 5s
RuRu_pv4p64d75s14p^64d^75s^14d 4p 5s4d\ 4p\ 5s
RhRh_pv4p64d85s14p^64d^85s^14d 4p 5s4d\ 4p\ 5s
PdPd4d95s14d^95s^14d 5s4d\ 5s
AgAg4d105s14d^{10}5s^14d 5s4d\ 5s
CdCd4d105s24d^{10}5s^24d 5s4d\ 5s
InIn_d4d105s25p14d^{10}5s^25p^14d 5p 5s4d\ 5p\ 5s
SnSn_d4d105s25p24d^{10}5s^25p^24d 5p 5s4d\ 5p\ 5s
SbSb5s25p35s^25p^35p 5s5p\ 5s
TeTe5s25p45s^25p^45p 5s5p\ 5s
II5s25p55s^25p^55p 5s5p\ 5s
CsCs_sv_GW5s25p65d15s^25p^65d^15p 5s5p\ 5s
BaBa_sv5s25p65d0.016s1.995s^25p^65d^{0.01}6s^{1.99}5p 5s 6s5p\ 5s\ 6s
ElementRPAWconfig.PAWconfig.LCAO basis
LaLaLa4f0.00015s25p65d0.99996s24f^{0.0001}5s^{2}5p^{6}5d^{0.9999}6s^{2}5p5p 5s5s 6s6s
CeCe_35s25p65d16s25s^{2}5p^{6}5d^{1}6s^{2}Ce4f15s25p65d16s24f^{1}5s^{2}5p^{6}5d^{1}6s^{2}4f4f 5d5d 5p5p 5s5s 6s6s
PrPr_35s25p65d16s25s^{2}5p^{6}5d^{1}6s^{2}Pr4f2.55s25p65d0.56s24f^{2.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
NdNd_35s25p65d16s25s^{2}5p^{6}5d^{1}6s^{2}Nd4f3.55s25p65d0.56s24f^{3.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
PmPm_35s25p65d16s25s^{2}5p^{6}5d^{1}6s^{2}Pm4f4.55s25p65d0.56s24f^{4.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
SmSm_35s25p65d16s25s^{2}5p^{6}5d^{1}6s^{2}Sm4f5.55s25p65d0.56s24f^{5.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
EuEu_35p65d16s25p^{6}5d^{1}6s^{2}Eu4f6.55s25p65d0.56s24f^{6.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
GdGd_35p65d16s25p^{6}5d^{1}6s^{2}Gd4f7.55s25p65d0.56s24f^{7.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
TbTb_35p65d16s25p^{6}5d^{1}6s^{2}Tb4f8.55s25p65d0.56s24f^{8.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
DyDy_35p65d16s25p^{6}5d^{1}6s^{2}Dy4f9.55s25p65d0.56s24f^{9.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
HoHo_35p65d16s25p^{6}5d^{1}6s^{2}Ho4f10.55s25p65d0.56s24f^{10.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
ErEr_35p65d16s25p^{6}5d^{1}6s^{2}Er4f11.55s25p65d0.56s24f^{11.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
TmTm_35p65d16s25p^{6}5d^{1}6s^{2}Tm4f12.55s25p65d0.56s24f^{12.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
YbYb_35p65d16s25p^{6}5d^{1}6s^{2}Yb4f13.55s25p65d0.56s24f^{13.5}5s^{2}5p^{6}5d^{0.5}6s^{2}4f4f 5p5p 5s5s 6s6s
LuLu_35p65d16s25p^{6}5d^{1}6s^{2}Lu4f145s25p65d16s24f^{14}5s^{2}5p^{6}5d^{1}6s^{2}4f4f 5d5d 5p5p 5s5s 6s6s
HfHf_pv5p65d36s15p^{6}5d^{3}6s^{1}Hf_sv5s25p65d45s^{2}5p^{6}5d^{4}5d5d 5p5p 5s5s
TaTa_pv5p65d46s15p^{6}5d^{4}6s^{1}5d5d 5p5p 6s6s
WW_sv5s25p65d56s15s^{2}5p^{6}5d^{5}6s^{1}5d5d 5p5p 5s5s 6s6s
ReRe5d66s15d^{6}6s^{1}5d5d 6s6s
OsOs5d76s15d^{7}6s^{1}5d5d 6s6s
IrIr5d86s15d^{8}6s^{1}5d5d 6s6s
PtPt5d96s15d^{9}6s^{1}5d5d 6s6s
AuAu5d106s15d^{10}6s^{1}5d5d 6s6s
HgHg5d106s25d^{10}6s^{2}5d5d 6s6s
TlTl_d5d106s26p15d^{10}6s^{2}6p^{1}5d5d 6p6p 6s6s
ElementRPAWconfig.PAWconfig.LCAO basis
PbPb_d5d106s26p25d^{10}6s^{2}6p^{2}5d 6p 6s5d\ 6p\ 6s
BiBi_d5d106s26p35d^{10}6s^{2}6p^{3}5d 6p 6s5d\ 6p\ 6s
PoPo_d5d106s26p45d^{10}6s^{2}6p^{4}5d 6p 6s5d\ 6p\ 6s
AtAt6s26p56s^{2}6p^{5}6p 6s6p\ 6s
RnRn6s26p66s^{2}6p^{6}6p 6s6p\ 6s
FrFr_sv6s26p67s16s^{2}6p^{6}7s^{1}6p 6s 7s6p\ 6s\ 7s
RaRa_sv6s26p67s26s^{2}6p^{6}7s^{2}6p 6s 7s6p\ 6s\ 7s
AcAc6s26p66d17s26s^{2}6p^{6}6d^{1}7s^{2}6d 6p 6s 7s6d\ 6p\ 6s\ 7s
ThTh5f16s26p66d17s25f^{1}6s^{2}6p^{6}6d^{1}7s^{2}5f 6d 6p 6s 7s5f\ 6d\ 6p\ 6s\ 7s
PaPa5f16s26p66d27s25f^{1}6s^{2}6p^{6}6d^{2}7s^{2}5f 6d 6p 6s 7s5f\ 6d\ 6p\ 6s\ 7s
UU5f26s26p66d27s25f^{2}6s^{2}6p^{6}6d^{2}7s^{2}5f 6d 6p 6s 7s5f\ 6d\ 6p\ 6s\ 7s
NpNp5f36s26p66d27s25f^{3}6s^{2}6p^{6}6d^{2}7s^{2}5f 6d 6p 6s 7s5f\ 6d\ 6p\ 6s\ 7s
PuPu5f46s26p66d27s25f^{4}6s^{2}6p^{6}6d^{2}7s^{2}5f 6d 6p 6s 7s5f\ 6d\ 6p\ 6s\ 7s
AmAm5f56s26p66d27s25f^{5}6s^{2}6p^{6}6d^{2}7s^{2}5f 6d 6p 6s 7s5f\ 6d\ 6p\ 6s\ 7s
CmCm5f66s26p66d27s25f^{6}6s^{2}6p^{6}6d^{2}7s^{2}5f 6d 6p 6s 7s5f\ 6d\ 6p\ 6s\ 7s

Table 6. Overview of PAW pseudopotentials used with corresponding electronic configuration, and atomic orbital basis. The first two columns indicate whether relaxation was performed with a reduced-valence PAW (RPAW) that included additional frozen electrons. Boldfaced PAW names indicate cases where the chosen PAW deviates from that recommended by VASP.

All compositions were set up in the 10-atom primitive Fm3ˉmFm\bar{3}m unit cell. The starting lattice parameter of a=7.78 A˚a = 7.78\,\mathrm{\AA} was determined from preliminary relaxation of 10 chloride compositions: We took the largest relaxed lattice parameter from this test set and expanded it by ∼10%\sim10\% to provide a conservative starting point. Fluoride, bromide, and iodide compositions were added to the workflow at a later stage; although bromide and iodide HDPs have larger equilibrium lattice parameters, using the same starting geometry did not lead to convergence issues.

For certain compositions, structural relaxation remained trapped in the high-symmetry configuration, with only minor ionic displacements before converging toward a local minimum and exhausting the maximum number of VASP ionic steps without achieving meaningful structural change. The workflow detects this outcome and restarts the relaxation with a larger initial lattice parameter (a=8.06 A˚a = 8.06\,\mathrm{\AA}). Other convergence failures were handled automatically by either tightening the electronic convergence criterion or cycling through alternative electronic minimization algorithms.

Structural relaxations were performed using PBEsol [63] with a force convergence threshold of 1 meV/A˚1\,\mathrm{meV}/\mathrm{\AA}. For these calculations, we used an increased plane-wave cutoff of 1.5ENMAX, with a minimum of 500 eV. For all subsequent DFT calculations, we used the largest default ENMAX value among the PAW potentials of the constituent elements, but not less than 350 eV. A 4×4×44 \times4 \times4 Γ\Gamma-centered kk-mesh was used throughout.

In the second step of the workflow, we enumerated all combinations of high- and low-spin states (based on the OhO_h point symmetry of the B-site) for transition metal, lanthanide, and actinide B-sites, including both parallel and antiparallel alignments of the magnetic sublattices. We did not consider non-collinear orderings, nor magnetic configurations requiring supercells beyond the 10-atom Fm3ˉmFm\bar{3}m primitive cell. Each configuration served as a starting point for a PBE[64] calculation, and the spin configuration with the lowest total energy was then selected for the remainder of the workflow. This was followed by a PBE pre-convergence run, and then the main electronic-structure calculation using the HSE06[65, 66] hybrid functional. The PBE pre-convergence and hybrid functional steps were performed using time-reversal symmetry only, to accommodate the projection in LOBSTER.[62] The electronic convergence criterion was set to 10−410^{-4} eV. The Brillouin zone was integrated using the tetrahedron method with Blöchl correction,[67] and the number of energy points for the density of states (DOS) was set to 5000.

For the final step we used LOBSTER to project the hybrid functional wavefunctions onto a linear combination of atomic orbitals (LCAO). We used the basis function set PBEVaspFit2015.[61] Using a minimal basis set, which only includes states occupied by valence electrons, is generally best practice, but several elements have additional polarization functions available. We assessed projection quality and the need for additional basis functions using the absolute charge spilling and band-overlap parameters. Our workflow started with projecting onto the minimal basis set. If the minimal basis yielded an absolute charge spilling above 3% or the orthonormalization step resulted in a matrix with a maximum deviation from identity exceeding 0.1 (indicating overlapping bands), the workflow calculated additional basis sets until either both conditions were met or it ran out of additional functions. Several additional basis-set projections were performed under the band-overlap condition, but the ultimate choice of basis set was based on the absolute charge spilling criterion alone (see Technical Validation). The band-overlaps criterion was applied as a precautionary check to flag potentially non-physical projections. Corresponding additional projections are retained in the database.

A technical subtlety concerns the cesium basis set used in the LOBSTER projections. We found that removing the Cs 6s basis functions substantially improved the projection quality for outlier systems (i.e., compositions with large absolute charge spilling, or low charge spilling but with localized charges summing to a value far from charge-neutrality). The origin of these deviations is likely related to the use of the Cs_sv_GW potential, whose projector structure differs from that assumed in the PBEVaspFit2015 basis. Excluding the 6s basis functions solved all observed projection issues and produced equal or slightly improved results for compositions that were already well behaved. Given that the Cs 6s state is expected to be unoccupied and weakly involved in bonding in these materials, we adopted the reduced cesium basis as the standard.

The energy range for the DOS and other chemical bonding quantities was automatically matched to the HSE06 DOS range. To ensure an accurate representation of the projected DOS and chemical bonding indicators, we increased the number of energy points to 6000 in LOBSTER. We invoked the COHPGenerator keyword separately for each B-X pair, with a maximum distance of 4 Å, restricting the bonding analysis to nearest-neighbor B-X interactions.

Chemical Bonding Analysis

LOBSTER projects the PAW wavefunctions onto a Slater-type orbital local basis set. The projection onto LCAO improves the quality of several localized quantities, like atomic charges [68] and site-projected DOS (pDOS), [60] compared to VASP’s standard projection method, which uses either a user-defined or PAW projector radius to define the projection sphere. LOBSTER then determines, for each atomic orbital, the kk-dependent expansion coefficients with respect to the plane-wave Bloch states, yielding the LCAO crystal orbital. Combined with the overlap integral between two atomic orbitals, Sμν=∫ϕμ∗ϕνdτS_{\mu\nu} = \int\phi_{\mu}^{*}\phi_{\nu}\mathrm{d}\tau, the coefficients lead to the first quantum-chemical bonding indicator: the projected crystal-orbital overlap population (pCOOP), giving an energy-resolved picture of where orbitals are overlapping, and which interactions are bonding/antibonding:

pCOOP⁡μν(E)=Sμν∑j,kwkRe⁡(cμ,jk∗cν,jk)⋅δ(εj(k)−E)(2)\operatorname{pCOOP}_{\mu\nu}(E) = S_{\mu\nu} \sum_{j,k} w_k \operatorname{Re}(c_{\mu,jk}^{*}c_{\nu,jk}) \cdot\delta(\varepsilon_j(\mathbf{k}) - E) \tag*{(2)}

Here μ\mu, ν\nu indicate atomic orbitals located on different atoms, εj\varepsilon_j is the energy eigenvalue of band jj, cμ,jkc_{\mu,jk} are the coefficients of the LCAOs, and wkw_k is the k-point weight. Integrating pCOOP(EE) up to the Fermi level yields the ICOOP, which corresponds to the number of electrons populating the bond. The pCOOP can be understood as an overlap-weighted density of states.

Another bonding indicator is the projected crystal-orbital Hamiltonian population (pCOHP), for which the LCAO basis is orthonormalized using Löwdin’s symmetric orthonormalization (LSO), [60, 69] yielding the orthonormalized wavefunctions χk′\chi_k^{\prime} and coefficients Ck′C_k^{\prime} respectively. The pCOHP is given by,

pCOHP⁡μν(E)=Hμν∑j,kwkRe⁡(cμ,jk′∗cν,jk′)⋅δ(εj(k)−E)(3)\operatorname{pCOHP}_{\mu\nu}(E) = H_{\mu\nu} \sum_{j,k} w_k \operatorname{Re}(c_{\mu,jk}^{\prime*}c_{\nu,jk}^{\prime}) \cdot\delta(\varepsilon_j(\mathbf{k}) - E) \tag*{(3)}

where HμνH_{\mu\nu} is the Kohn-Sham Hamiltonian reconstructed in the orthonormalized LCAO basis, and cμ,jk′c_{\mu,jk}^{\prime} the LSO-LCAO coefficients. The pCOHP integrated up to the Fermi level (ICOHP) corresponds to the energy contained in the bond, i.e., the bond strength. The pCOHP can be understood as a Hamiltonian-weighted density of states that partitions the Kohn-Sham band energy.

Replacing the overlap matrix element SμνS_{\mu\nu} or Hamiltonian matrix element HμνH_{\mu\nu} with the density matrix element PμνP_{\mu\nu} gives the crystal-orbital bonding index (COBI):[70]

COBIμν(E)=Pμν∑j,kwkRe⁡(cμ,jk∗cν,jk)⋅δ(εj(k)−E)(4)\mathrm{COBI}_{\mu\nu}(E) = P_{\mu\nu} \sum_{j,\mathbf{k}} w_{\mathbf{k}} \operatorname{Re}(c_{\mu,j\mathbf{k}}^{*}c_{\nu,j\mathbf{k}}) \cdot\delta(\varepsilon_j(\mathbf{k}) - E) \tag*{(4)}

The integrated value of the COBI, the ICOBI, corresponds to the classical bond order, with low values indicating a mostly ionic bond and values close to one (or two for double bonds) indicating a mostly covalent bond.

Additionally, LOBSTER computes orbital and gross populations, atomic charges, and Madelung energies based on both the Mulliken[71] and Löwdin[69] partitioning schemes.

For this database, we omitted the pCOOP calculation (keyword skipCOOP) to reduce storage requirements, since bonding/antibonding information is also contained in the pCOHP and COBI. The initial LCAO basis can occasionally lead to negative occupations in orbital populations and the pDOS due to numerical projection errors. To ensure physically meaningful populations and consistent results across the dataset, the analysis of the pDOS and charges presented here is based on the LSO basis. The analysis scripts also supports pDOS analysis from VASP or the original LOBSTER basis, as well as Mulliken-scheme population analysis.

Generating Data Records

As detailed above, we use the Bartel τ\tau-factor as our primary stability criterion. For comparison, the database also reports the generalized Goldschmidt (tt) factor introduced by Filip and Giustino.[17] Although the generalized tt has lower reported accuracy than τ\tau for predicting perovskite stability, we find it informative for HDPs, because it captures both the average ionic radius of the two B-sites (through the average octahedral factor μˉ=(rB+rB′)/2rX\bar{\mu} = (r_B + r_{B'})/2r_X) and the difference between them (through the octahedral mismatch Δμ=∣rB−rB′∣/2rX\Delta\mu= |r_B - r_{B'}|/2r_X).

Ref.[17] combines tt, μˉ\bar{\mu}, and Δμ\Delta\mu into eight geometric inequalities, and translate violations of these inequalities into a stability prediction via volumetric integration over uncertainties in the ionic radii. We instead compute tt, μˉ\bar{\mu}, and Δμ\Delta\mu directly from the tabulated Shannon radii and report which (if any) of the geometric limits are violated, using the notation of Ref. 17.

For the relaxed structures, we report B-X bond lengths together with the X-X distances between opposing X-anions in each octahedron. The latter remains well-defined for vacancy-ordered HDPs (where one B-site is empty and the B-X distance is undefined) and thus enables direct structural comparison across standard and vacancy-ordered compositions.

The contributions near band edges were analyzed from the projected DOS in LOBSTER (in the LSO basis). Since all A- and X-sites are symmetrically equivalent in our structural models, we summed over their respective pDOSs. We determined band-edge contributions within an analysis window of 0.5 eV below the valence band maximum (VBM) and 0.5 eV above the conduction band minimum (CBM). Within this window, all states were integrated, and for each species we computed its contribution to the integrated states and identified the dominant orbital. The contributions to the VBM and CBM were determined separately for the spin-up and spin-down channels. We selected the highest VBM and lowest CBM as the global band edges across both spin channels, and stored the data in a separate ’combined’ spin channel. Storing all three sets - spin-up, spin-down, and combined - allows us to identify spin-forbidden transitions, half-metallic compositions, and the band gap of the non-conducting spin channel in half-metals. Furthermore, we report the charges and the orbital gross population of each site to determine the site-specific spin magnetic moments. Owing to the simplified treatment of magnetism in our workflow, these values are not expected to be quantitatively accurate. They nonetheless serve as useful indicators when screening for compositions with interesting spin properties.

In addition to the standard quantities provided by the LOBSTER analysis, we introduce a bonding descriptor specific to the HDP geometry. Our bonding analysis considers only nearest-neighbor B-X bonds. Therefore, we provide analysis only for the properties of these bonds, which are aligned with the Cartesian axes in the cubic structure imposed by our 10-atom primitive cell. This means that the cubic harmonics from LOBSTER are properly aligned by default, enabling us to identify differences in bonding in the xx, yy, and zz directions.

The projected COHP (pCOHB) and pCOBI (and their integrated values) are averaged over the two bonds along each direction and over all six B-X bonds. The B-X bonds along one axis are not always the same in the positive and negative directions. To quantify this, we use the asymmetry index introduced by Belli et al. [72] and implemented in LobsterPy [73] (which we dub axial-asymmetry index to distinguish it from the other asymmetry descriptor introduced for this work):

axial_asym_index=∣Si∣=∣1Bi∑a=1BiICO..(i,a)ei,a∣(5)\mathrm{axial\_asym\_index} = \lvert\mathbf{S}_i \rvert= \left\lvert\frac{1}{B_i} \sum_{a=1}^{B_i} \mathrm{ICO}_{..}(i,a)\mathbf{e}_{i,a} \right\rvert \tag*{(5)}

where BiB_i is the number of bonds from site ii, ei,a\mathbf{e}_{i,a} is the unit vector along bond number aa originating from site ii, and ICO.. refers to any of ICOOP, ICOHP, or ICOBI. The axial-asymmetry index is the norm of the absolute differences in bond strength/index along each axis; when applied to the ICOBIs, it yields a maximum value of 0.0032 in our dataset. To quantify whether there is any asymmetry between the Cartesian axes, we introduce the directional-asymmetry index, given by:

directional_asym_index=max⁡(ICO..(x/y/z))−min⁡(ICO..(x/y/z))ICO..avg(6)\mathrm{directional\_asym\_index} = \frac{\max(\mathrm{ICO}_{..(x/y/z)}) - \min(\mathrm{ICO}_{..(x/y/z)})}{\mathrm{ICO}_{..\mathrm{avg}}} \tag*{(6)}

which gives the difference between the strongest and weakest bond relative to the six-bond average. Using the directional-asymmetry index on the ICOBIs gives a maximum value of 1.49 in our dataset.

UMAP projection

To visualize underlying trends and facilitate navigation of our database, we used the Uniform Manifold Approximation and Projection (UMAP) method, [74] a non-linear dimensionality reduction algorithm that is similar to t-distributed Stochastic Neighbor Embedding (t-SNE) [75] but with more favorable scaling. UMAP calculates the distance from a point xix_i to its NN nearest neighbors using a locally varying distance metric, producing a directed weighted graph. This graph is then symmetrized into an undirected graph. Next, a lower-dimensional (in our case, 2D) representation is initialized, and then the cross-entropy between the high- and low-dimensional representations is minimized. Two key UMAP parameters are the distance metric and the number of neighbors.

Since our aim is to visualize trends in the electronic-structure data directly, we used the pDOS and pCOHP data as input for UMAP. We applied a Gaussian smearing of 0.05 eV to these data to mitigate disproportionate contributions of very sharp peaks in the pDOS and pCOHP to the distance measure. Additionally, we restricted the input to states within 5 eV of the band edges, i.e., in the range [VBM - 5 eV, CBM + 5 eV], to exclude deep core state and high-energy conduction states that are not chemically informative. The VBM was shifted to 0 eV across the dataset. To ensure the energy axis and intervals are equivalent for all compositions, we used the binning method introduced in Refs. 76,77. Overall, this gives an energy axis ranging from -5.00 eV up to 14.83 eV with 0.1 eV intervals, resulting in a 199-dimensional vector for each pDOS/pCOHP curve selected for UMAP input. Binned pDOS and pCOHP curves were saved for each atomic site and spin channel to simplify testing different combinations of input data for the projection.

To represent the compositions in the UMAP projection, several options exist based on the computed pDOS for each site and the B-X pCOHP curves. Additionally, different embeddings can be merged (using their union (++), intersection (∗*), and contrast (−-)), combining the fuzzy edge weights between pairs of points across constituent embeddings. This allows us, for example, to compute separate embeddings from the B1 pDOS and B2 pDOS and then intersect them. Our UMAP projections are based on the Euclidean and Manhattan metrics, their standardized versions (weighted with the inverse standard deviation), the Bray-Curtis metric (∑i∣ui−vi∣/∑i∣ui+vi∣\sum_i |u_i-v_i|/\sum_i |u_i+v_i|), and the cosine metric. We calculated each metric using the nearest-neighbor values 5, 15, 25, 50, 100. As input to our projection, we tested the following combinations:

  • Total DOS (spin-up & down concatenated) (TDOS)

  • Intersection of B1-pDOS (spin-up & down concatenated) and B2-pDOS (spin-up & down concatenated) (B-pDOS)

  • Intersection of B1 and B2 spin-up pDOS (concatenated) and B1 and B2 spin-down pDOS (concatenated) (alternative B-pDOS)

  • Intersection of B1 spin-up, B1 spin-down, B2 spin-up, and B2 spin-down pDOS (separated B-pDOS)

  • Intersection of B1-X average COHP with B2-X average COHP (B-X pCOHP)

and also a normalized version of each combination, for which the area under the binned curves was set to 1.

The examples above were chosen to showcase how these data representation choices affect the projection, and are not intended to be final. Users of the database are encouraged to explore further options. For this purpose, we developed an interactive viewer using Dash (www.dash.plotly.com) that enables rapid switching between precomputed projections and coloring by computed material properties. Selecting a point in the embedding reveals the corresponding pDOS and pCOHP/pCOBI curves together with basic electronic-structure data for that specific material. This functionality helps explore different projection setups and evaluate whether the resulting embedding reflects meaningful chemical or physical trends. All precomputed projections are based on UMAP, but the interactive viewer can, in principle, be used with any dimensionality reduction algorithm, as long as the projection is two-dimensional and the projection is saved with the same composition indexing as the database and similar table header structure.

Data Records

Of the 2,680 compositions fed into the workflow, 138 encountered convergence issues, leaving our database with Data Records for 2,542 HDP compositions spanning the chemical landscape. The Data Records are separated into four categories: basic data on the input species and the LOBSTER basis functions (Table 2); structural data, including lattice vectors and interatomic distances (Table 3); band edge data, including band gaps and band-edge orbital character for spin-up, spin-down, and combined spin channels (Table 4); and bonding descriptors (Table 5). The information from Table 2 - Table 5 is combined into a single HDP_CombinedInfo.csv file for convenience, with only the ’combined’ spin channel from Table 4 included. In this combined file, the site (with orbital character) with the largest contribution to the VBM and CBM edges is also extracted, and was used to create the coloring scheme in Figure 3a.

Scatter plot of net spin magnetic moment versus band gap values and UMAP projection

Figure 3. (a) Scatter plot of the net spin magnetic moment vs. band gap values for chloride HDPs. The color of the data points is determined by the orbital character of the valence band maximum (orange, purple, blue, green corresponding to s-, p-, d-, f-orbital character, respectively), whereas the color tone indicates the orbital character of the CBM. (b) UMAP projection based on the separated B-pDOS data using the Bray-Curtis metric with n_neighbors=25, colored by the band gap values.

Variable nameDatatypeDescription
CompID_numintUnique ID number assigned to each composition, starting from 1000.
comp_name_simplestringChemical formula of composition without numbers (i.e., CsBB′XCsBB'X).
comp_name_fullstringFull chemical formula of composition (i.e., Cs2BB′X6Cs_{2}BB'X_{6}).
element.{site}stringElement symbol of {site}.
specie.{site}stringElement and oxidation state of {site} (e.g., Ag2+Ag^{2+}).
r_ionic.{site}floatShannon ionic radius in picometer associated with {site}.
tau_factorfloatBartel’s τ\tau-factor [37] for this composition.
oct_factorfloatOctahedral factor based on the average B-site ionic radius.
oct_mismatchfloatOctahedral mismatch between the two B-sites introduced in [17].
gen_t_factorfloatGeneralised tt-factor introduced in [17].
geom_stableboolWhether the compound is stable based on the geometric criteria of [17].
broken_conditionstringIndicates which geometric limit is broken if predicted non-stable by [17].
block.{B1/B2}stringBlock of the periodic table {B1/B2} comes from (s/p/d/f).
row.{B1/B2}intRow of the periodic table {B1/B2} originates from.

Table 2. Data contained in HDP_BasicInfo.csv. {site} means the quantity is given for A, B1, B2, and X separately. {B1/B2} means the quantity is reported for B1 and B2 separately.

Variable nameDatatypeDescription
lattice_a_primitivefloatLattice constant of the primitive unit cell in Å.
lattice_a_conventionalfloatLattice constant of the conventional unit cell in Å.
distance_{B1/B2}_XfloatDistance between {B1/B2} and X ions in Å.
distance_{B1/B2}_AfloatDistance between {B1/B2} and A cations in Å.
size_Oh_{B1/B2}floatSize of the {B1/B2} octahedron, measured from opposing X anions in Å.

Table 3. Data contained in HDP_StructuralInfo.csv. {B1/B2} means the quantity is given for B1 and B2 separately.

Variable nameSpinDatatypeDescription
VBM“1”, “-1”, “comb”floatValence band maximum energy (eV).
CBM“1”, “-1”, “comb”floatConduction band minimum energy (eV).
Variable nameSpinDatatypeDescription
band_gap“1”, “-1”, “comb”floatBand gap energy (eV).
VBM.totcontr.{site}“1”, “-1”, “comb”floatTotal contribution of {site} to the valence band.
VBM.orbcontr.{site}“1”, “-1”, “comb”floatHighest single orbital contribution of {site} to the valence band.
VBM.orbital.{site}“1”, “-1”, “comb”stringMain contributing orbital of {site} to the valence band.
VBM.orbchar.{site}“1”, “-1”, “comb”stringCharacter (s/p/d/f) of the main contributing orbital of {site} to the valence band.
VBM.orborder.{site}“1”, “-1”, “comb”stringPrincipal quantum number (nn) of the main contributing orbital of {site} to the valence band.
CBM.totcontr.{site}“1”, “-1”, “comb”floatRelative contribution of {site} to the conduction band.
CBM.orbcontr.{site}“1”, “-1”, “comb”floatHighest single orbital contribution of {site} to the conduction band.
CBM.orbital.{site}“1”, “-1”, “comb”stringMain contributing orbital of {site} to the conduction band.
CBM.orbchar.{site}“1”, “-1”, “comb”stringCharacter (s/p/d/f) of the main contributing orbital of {site} to the conduction band.
Variable nameSpinDatatypeDescription
CBM.orborder.{site}“1”, “-1”, “comb”stringPrincipal quantum number (nn) of the main contributing orbital of {site} to the conduction band.
spin_cbm“combined”stringSpin channel of the overall CBM.
spin_vbm“combined”stringSpin channel of the overall VBM.
spin_forbidden“combined”boolWhether the VBM and CBM are in differing spin channels, with a minimal energy difference of 50 meV.
cond_type“1”, “-1”, “comb”stringWhether the spin channel is metallic (Eg≤100E_g \le100 meV), semiconducting, insulating (Eg>4E_g > 4 eV), or half-metallic (combined channel only).

Table 4. Data contained in bandedgeInfo_lsodos.csv. Data is collected for spin-up (“1”) and spin-down (“-1”) channels separately, and combined (“comb”) using the highest VBM and lowest CBM across channels. {site} means the quantity is included for A, B1, B2, and X separately.

Variable nameDatatypeDescription
abs_charge_spillingfloatThe absolute charge spilling from the LOBSTER projection (%).
bandOverlap_existsboolWhether LOBSTER wrote the bandOverlaps file, indicating imperfect orthonormalization.
bandOverlap_maxDevfloatMaximum deviation from identity matrix recorded in bandOverlaps.
Variable nameDatatypeDescription
bandOverlap_perc_kptsfloatPercentage of k-points that exceed a deviation of 0.1 from the identity matrix (%).
popdiff.totalfloatTotal difference between spin-up and spin-down gross populations.
popdiff.B1floatSpin-up&down population difference on B1-site.
popdiff.B2floatSpin-up&down population difference on B2-site.
charge.{site}floatNet charge on {site}.
charge.totalfloatSum of local charges.
charge.balancedboolWhether the total structure is charge neutral with a tolerance of 0.1e−e^{-}.
Icohp.{B1/B2}.sumfloatSum of the 6 B-X ICOHP values on site {B1/B2} (eV).
Icohp.{B1/B2}.avgfloatAverage of the 6 B-X ICOHP values on site {B1/B2} (eV).
Icohp.{B1/B2}.avg_spinupfloatAverage of the 6 B-X ICOHP values on site {B1/B2}, based on spin-up electrons only (eV).
Icohp.{B1/B2}.avg_spindownfloatAverage of the 6 B-X ICOHP values on site {B1/B2}, based on spin-down electrons only (eV).
Icohp.{B1/B2}.{x/y/z}_avgfloatAverage of the B-X ICOHP values aligned with {x/y/z}-cartesian axis (eV).
Icohp.{B1/B2}.{x/y/z}_spinupfloatAverage of the B-X spin-up ICOHP values aligned with {x/y/z}-cartesian axis (eV).
Variable nameDatatypeDescription
Icohp.{B1/B2}.{x/y/z}_spindownfloatAverage of the B-X spin-down ICOHP values aligned with {x/y/z}-cartesian axis (eV).
Icobi.{B1/B2}.sumfloatSum of the 6 B-X ICOBI values on site {B1/B2}.
Icobi.{B1/B2}.avgfloatAverage of the 6 B-X ICOBI values on site {B1/B2}.
Icobi.{B1/B2}.avg_spinupfloatAverage of the 6 B-X ICOBI values on site {B1/B2}, based on spin-up electrons only.
Icobi.{B1/B2}.avg_spindownfloatAverage of the 6 B-X ICOBI values on site {B1/B2}, based on spin-down electrons only.
Icobi.{B1/B2}.{x/y/z}_avgfloatAverage of the B-X ICOBI values aligned with {x/y/z}-cartesian axis.
Icobi.{B1/B2}.{x/y/z}_spinupfloatAverage of the B-X spin-up ICOBI values aligned with {x/y/z}-cartesian axis.
Icobi.{B1/B2}.{x/y/z}_spindownfloatAverage of the B-X spin-down ICOBI values aligned with {x/y/z}-cartesian axis.
{Icohp/Icobi}.axial_asym_index.{B1/B2}floatNorm of absolute differences in bonding indicator along each axis (from [72]).
{Icohp/Icobi}.directional_asym_index.{B1/B2}floatRelative differences in bonding indicator between cartesian axes.
MadelungEnergyfloatTotal electrostatic stabilization energy (eV).
Sitepotential.{B1/B2}floatElectrostatic stabilization coming from site {B1/B2} (eV).

Table 5. Data contained in LobsterInfo.csv. {B1/B2} means the quantity is reported for B1 and B2 separately. {x/y/z} indicates that the quantity is determined separately along the Cartesian axes.

All calculation data, including all LOBSTER projections performed with additional basis functions, are made accessible through NOMAD[78] and the workflow was parsed with the NOMAD Utility Workflows plugin.[79] The dataset is publicly available via NOMAD.[80] Data processing was implemented using several functionalities from PyMatGen[81] and LobsterPy.[73] All scripts, the extracted data, and a list of the compositions that failed to converge are available on Zenodo.[82] The interactive UMAP viewer, including the precomputed projections, is available on GitHub[83] and is also archived on Zenodo.[84]

Data Overview

Figure 3 provides representative views of the dataset. Figure 3a shows a scatter plot of the net magnetic moments and band gaps for the Cs2[BB′]Cl6\mathrm{Cs}_{2}[\mathrm{BB'}]\mathrm{Cl}_{6} compositions, highlighting the diversity of electronic properties in this chloride-based subset of the data. Figure 3b shows a selected UMAP projection of the separated pDOS, computed using the Bray-Curtis metric. Figure 3b is colored by band gap values, revealing an underlying structure. The projection shows an isolated cluster in the top-left corner, comprising all vacancy-ordered HDP entries. Figure 3b illustrates that the pDOS alone carries enough information for UMAP to reveal underlying structure in a metric of interest and to separate a distinct subgroup of HDPs. Figure 4a and Figure 4d show the band gap and ICOBI values per element, averaged over the number of compositions in which they are incorporated on a B-site or X-site. As expected from electronegativity trends, fluorine-containing HDPs exhibit the lowest ICOBI values and the highest band gaps among the halides, reflecting their predominantly ionic character. Cl-, Br- and I-containing HDPs follow in sequence, with progressively higher ICOBI values and lower band gaps.

Periodic tables showing average band gap and average ICOBI

Figure 4. (a) Periodic table showing the average band gap for compositions associated with each element (in the case of Cs only compositions with Cs on a B-site are included). (b) Periodic table showing the average ICOBI value for each element. The figures were created using PyMatViz [85].

Technical Validation

Completeness of Chemical Landscape

Our goal was to cover the full chemical landscape of cesium HDPs, yet the database inevitably contains some gaps. Of the compositions screened by the τ\tau factor, 138 failed to converge and are absent from the dataset. Beyond convergence failures, the tolerance factor is an imperfect stability criterion. The choice of τ\tau cutoff involves a balance of false positives and false negatives. The τ\tau-factor is linked to a stability probability using Platt scaling in Ref. [37], and the cutoff of 4.18 corresponds to a 50% stability probability, so some misclassified compositions near the boundary are to be expected. One illustration of this is provided by Kent et al. who report the synthesis and analysis of Cs2NaLnI6\mathrm{Cs_2NaLnI_6} with Ln=(Ce,Nd,Gd,Tb,Dy)Ln=(\mathrm{Ce}, \mathrm{Nd}, \mathrm{Gd}, \mathrm{Tb}, \mathrm{Dy})[86], whereas the τ\tau-factor only predicts Ln=(Ce,Nd)Ln=(\mathrm{Ce}, \mathrm{Nd}) to be stable. The remaining lanthanide iodide compositions exceed the maximum τ\tau-value by 0.02 or less, yet are evidently stable.

Structural Data

We used the ICSD[87] to compare the computed lattice parameters from our database with experimental values. We found 352 entries matching the HDP chemical formula, corresponding to 160 unique compositions. Filtering for entries with the Fm3ˉmFm\bar{3}m space group resulted in a total of 284 entries for 125 unique compositions. The results of the comparison are shown

in Figure 5a. Multiple ICSD entries for the same composition are shown with a vertical bar spanning the full range of values, and the markers indicate the average. Overall, there is good agreement between our simulations and experimental values. The mean absolute error (MAE) between the averaged ICSD and computed lattice parameters is 0.075 Å. The composition with the widest spread in lattice parameters is Cs2_2AgBiBr6_6 with 24 entries measured at temperatures from 30 K to 473 K (Cs2_2AgBiBr6_6 features a structural phase transition from Fm3‾mFm\overline{3}m to I4/mI4/m at 122 K[88]). The 35 compositions without a cubic phase entry mostly have entries with the following space group symbols: P3‾m1P\overline{3}m1 (15 compositions), R3‾mHR\overline{3}mH (12 compositions), and I4/mmmI4/mmm (3 compositions). The trigonal crystal systems (P3‾m1P\overline{3}m1 and R3‾mHR\overline{3}mH space groups) have octahedra with different connectivities. The remaining 5 crystal structures are differently distorted variants of the trigonal or cubic crystal systems with lower symmetries. While we cannot compare the I4/mmmI4/mmm entries structurally to our simulations, we nevertheless analyze our bonding results within the Fm3‾mFm\overline{3}m space group. Two of the I4/mmmI4/mmm entries, Cs2_2CuHgCl6_6 and Cs2_2PdHgCl6_6, feature alternating elongated and contracted octahedra due to a Jahn-Teller distortion. [89, 90] Both also exhibit high directional asymmetry in our bonding analysis. Using the ICOBI-based directional asymmetry descriptor, we find: Cs2_2CuHgCl6_6 with an asymmetry of 0.30 on copper and 0.12 on mercury, and Cs2_2PdHgCl6_6 with 0.05 on palladium and 0.13 on mercury. Using the ICOHPs as input for the bonding descriptor yields similar results. The third I4/mmmI4/mmm entry (Cs2_2CuF6_6) is missing from our database. These examples highlight the additional insights that can be provided by bonding analysis.

Validation of the data: comparison of experimental and calculated lattice constants and histogram of charge spillings

Figure 5. Validation of the data. (a) Comparison of experimental lattice constants from the ICSD with calculated lattice constants. Only entries with Fm3ˉmFm\bar{3}m space group are included (125/160 compositions). For compositions with multiple entries, the average is shown, with error bars indicating the full range. (b) Histogram of charge spillings in LOBSTER projection, the most important quality measure. The inset shows a zoom into the black dashed box.

Bonding Analysis

Projecting the PAW wavefunction onto LCAO can lead to the loss or relocation of electrons. The extent of the relocation is measured by the absolute charge spilling[61]

SQ=1Nj∑kNkwk∑jNj∣1−Ojj(k)∣,(7)S_Q = \frac{1}{N_j}\sum_{\mathbf{k}}^{N_k} w_k \sum_j^{N_j} \left|1 - O_{jj}(\mathbf{k})\right|, \tag*{(7)}

where OjjO_{jj} are elements of the overlap matrix of the projected bands, which is summed over all occupied bands jj, and k-points with weight wkw_k. The absolute charge spilling is the most important metric to assess the quality of the projection. Figure 5b shows a histogram of the charge spillings in the dataset using the minimal basis set. None of the minimal basis projections exceed our imposed 3% charge-spilling limit. Some compositions have high BANDOVERLAP values (indicating deviations from the identity matrix during orthonormalization). High band overlap is of minor consequence to the projected values. Including additional basis functions, however, can directly affect ICOHP and ICOBI values, even if those orbitals are only fractionally occupied. Inspecting the results for the compositions with high band overlap values showed no notable irregularities. The results and analysis presented here are therefore based on the minimal basis set for all compositions. The projection quality measures are reported for all compositions (see HDP_LobsterInfo.csv). Where additional basis-function calculations were performed, the projection data are also available on NOMAD (see [80]).

Band gaps

We also compare our calculated band-gap values with experimentally measured gaps. Experimental band gap determination can be method- and sample-dependent, leading to varying band gap values for materials with nominally identical composition and structure.[91] We therefore provide only one or two sources per composition and focus on results obtained using the Tauc-plot method. The Tauc-plot method analyzes the UV-vis absorption onset and can give insight into whether a material has a direct or indirect band gap. This method has limitations, especially for (poly-)crystalline samples,[92] but we use it here due to its prevalence in the literature. Note that we do not expect HSE06 band gaps to yield exact agreement with experiment. Our aim is instead to quantify the extent to which our calculated band gaps agree with experiment, within reasonable error bars given the limitations of our methods. Discrepancies between calculated and experimental values may be caused by limitations of the HSE06 functional - which uses a fixed amount of screened exact exchange - and from our neglect of dynamic distortions, electron-hole interactions, and spin-orbit coupling, which have all been shown to substantially affect band gap predictions of halide perovskites.[93, 94, 95, 96, 97] Given these caveats, our results for selected compositions, shown in Table 1, are in good agreement with experimental values and can serve as a starting point for future calculations using more accurate methods.

CompositionCalc. EgE_g (eV)Exp. EgE_g (eV)NotesRef.
Cs2AgBiCl6Cs_{2}AgBiCl_{6}2.752.50-2.92 (indirect)single crystal[98, 99]
Cs2AgBiBr6Cs_{2}AgBiBr_{6}2.072.02 (indirect)single crystal[98]
Cs2AgInCl6Cs_{2}AgInCl_{6}2.323.53 (direct)polycrystalline[100]
Cs2AgSbCl6Cs_{2}AgSbCl_{6}2.132.54 (indirect)polycrystalline[100]
Cs2NaBiCl6Cs_{2}NaBiCl_{6}4.743.56 (indirect)nanocrystals[101]
Cs2NaBiBr6Cs_{2}NaBiBr_{6}3.923.10 (indirect)nanocrystals[102]
Cs2CuBiCl6Cs_{2}CuBiCl_{6}2.051.31 (indirect)single crystal[99]
Cs2CuSbCl6Cs_{2}CuSbCl_{6}1.621.66 (indirect)nanocrystals[103]
Cs2SnI6Cs_{2}SnI_{6}0.771.25 (optical gap)UV-vis reflectance[91]
Cs2TeI6Cs_{2}TeI_{6}1.831.59 (optical gap)UV-vis reflectance[91]

Table 1. Comparison of band gaps calculated with the HSE06 functional with experimental values.

Data Availability

The full calculation outputs can be accessed through NOMAD. [80] The extracted data described in the Data Record section can be found on Zenodo. [82] The UMAP explorer app, including all projections performed for this work is also deposited on Zenodo. [84]

Code Availability

All code developed for running the workflow, and analysing the results can be found on Zenodo together with the extracted data. [82] The code used for the UMAP explorer app is contained in the same Zenodo Archive. [84] Both Zenodo archives also contain exact python package version information in the included requirements.txt files. VASP calculations were performed using version 6.3.0. LOBSTER calculations with version 5.1.1.

Acknowledgement

L.W. and J.G. were supported by ERC Grant MultiBonds (grant agreement no. 101161771; funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.) L.W. and J.G. gratefully thank the Gauss Centre for Supercomputing e.V. (www.gausscentre.eu) for providing generous computing time on the GCS Supercomputer SuperMUC-NG at the Leibniz Supercomputing Centre (www.lrz.de) under Project No. pn73da. L.L. acknowledges discussions with David Prendergast about the UMAP method and funding by the Dutch Research Council (NWO) through Grant No. OCENW.M20.337 and VI.Vidi.223.072. We acknowledge the support of the NFDI consortium FAIRmat, funded by the Deutsche Forschungsgemeinschaft (DFG) – Project 460197019, for providing research data infrastructure and services. We thank Aakash A. Naik for adapting the NOMAD LOBSTER parsers to reflect recent LOBSTER updates and providing further technical support.

inputcharge.{B1/B2}intCharge assigned to {B1/B2} in compound generation.
used_lobbasisstringbasis{n} indicating which set of atomic orbitals was used.
used_lobbasis_funcsdict(str)Dictionary of atomic orbitals used for each element.

References

  1. [1]Kojima, A.; Teshima, K.; Shirai, Y.; Miyasaka, T. Organometal halide perovskites as visible-light sensitizers for photovoltaic cells. J. Am. Chem. Soc. 2009, 131, 6050–1.DOI
  2. [2]National Renewable Energy Laboratory Best Research-Cell Efficiencies. 2026; https://nrel.gov/media/docs/libraries/pv/cell-pv-eff.pdf.
  3. [3]Singh, T.; Miyasaka, T. Stabilizing the efficiency beyond 20% with a mixed cation perovskite solar cell fabricated in ambient air under controlled humidity. Adv. Energy Mater. 2017, 8.DOI
  4. [4]Eperon, G. E.; Stranks, S. D.; Menelaou, C.; Johnston, M. B.; Herz, L. M.; Snaith, H. J. Formamidinium lead trihalide : a broadly tunable perovskite for efficient planar heterojunction solar. Energy Environ. Sci. 2014, 7, 982–988.DOI
  5. [5]Brauer, J. C.; Tsokkou, D.; Sanchez, S.; Droseros, N.; Roose, B.; Mosconi, E.; Hua, X.; Stolterfoht, M.; Neher, D.; Steiner, U.; De Angelis, F.; Abate, A.; Banerji, N. Comparing the excited-state properties of a mixed-cation–mixed-halide perovskite to methylammonium lead iodide. J. Chem. Phys. 2020, 152, 104703.DOI
  6. [6]Bi, D.; Li, X.; Milić, J.; Kubicki, D.; Pellet, N.; Luo, J.; LaGrange, T.; Mettraux, P.; Emsley, L.; Zakeeruddin, S.; Grätzel, M. Multifunctional molecular modulators for perovskite solar cells with over 20.
  7. [7]Wolf, N. R.; Connor, B. A.; Slavney, A. H.; Karunadasa, H. I. Doubling the Stakes: The Promise of Halide Double Perovskites. Angew. Chem. Int. Ed. 2021, 60, 16264–16278.DOI
  8. [8]Maughan, A. E.; Ganose, A. M.; Bordelon, M. M.; Miller, E. M.; Scanlon, D. O.; Neilson, J. R. Defect Tolerance to Intolerance in the Vacancy-Ordered Double Perovskite Semiconductors Cs₂SnI₆ and Cs₂TeI₆. J. Am. Chem. Soc. 2016, 138, 8453–8464. Mitzi, D. B.; Wang, S.; Feild, C. a.; Chess, C. a.; Guloy, a. M. Conducting Layered Organic-inorganic Halides Containing <110>-Oriented Perovskite Sheets. Science 1995, 267, 1473.
  9. [10]Stoumpos, C. C.; Cao, D. H.; Clark, D. J.; Young, J.; Rondinelli, J. M.; Jang, J. I.; Hupp, J. T.; Kanatzidis, M. G. Ruddlesden-Popper Hybrid Lead Iodide Perovskite 2D Homologous Semiconductors. Chem. Mater. 2016, 28, 2852–2867.DOI
  10. [11]Pedesseau, L.; Sapori, D.; Traore, B.; Robles, R.; Fang, H. H.; Loi, M. A.; Tsai, H.; Nie, W.; Blancon, J. C.; Neukirch, A.; Tretiak, S.; Mohite, A. D.; Katan, C.; Even, J.; Kepenekian, M. Advances and Promises of Layered Halide Hybrid Perovskite Semiconductors. ACS Nano 2016, 10, 9776–9786.
  11. [12]Connor, B. A.; Leppert, L.; Smith, M. D.; Neaton, J. B.; Karunadasa, H. I. Layered Halide Double Perovskites: Dimensional Reduction of Cs₂AgBiBr₆. J. Am. Chem. Soc. 2018, 140, 5235–5240.DOI
  12. [13]Dyksik, M.; Duim, H.; Zhu, X.; Yang, Z.; Gen, M.; Kohama, Y.; Adjokatse, S.; Maude, D. K.; Loi, M. A.; Egger, D. A.; Baranowski, M.; Plochocka, P. Broad Tunability of Carrier Effective Masses in Two-Dimensional Halide Perovskites. ACS Energy Lett. 2020, 5, 3609–3616.DOI
  13. [14]Umeyama, D.; Leppert, L.; Connor, B. A.; Manumpil, M. A.; Neaton, J. B.; Karunadasa, H. I. Expanded Analogs of Three-Dimensional Lead-Halide Hybrid Perovskites. Angew. Chem. Int. Ed. 2020, 59, 19087–19094.DOI
  14. [15]Kamminga, M. E.; de Wijs, G. A.; Havenith, R. W. A.; Blake, G. R.; Palstra, T. T. The Role of Connectivity on Electronic Properties of Lead Iodide Perovskite-Derived Compounds. Inorg. Chem. 2017, 56, 8408–8414.DOI
  15. [16]Faber, F. A.; Lindmaa, A.; Von Lilienfeld, O. A.; Armiento, R. Machine Learning Energies of 2 Million Elpasolite (ABC₂D₆) Crystals. Phys. Rev. Lett. 2016, 117, 135502.
  16. [17]Filip, M. R.; Giustino, F. The geometric blueprint of perovskites. Proc. Natl. Acad. Sci. 2018, 115, 5397–5402.DOI
  17. [18]Slavney, A. H.; Hu, T.; Lindenberg, A. M.; Karunadasa, H. I. A Bismuth-Halide Double Perovskite with Long Carrier Recombination Lifetime for Photovoltaic Applications. J. Am. Chem. Soc. 2016, 138, 2138–2141.DOI
  18. [19]Volonakis, G.; Filip, M. R.; Haghighirad, A. A.; Sakai, N.; Wenger, B.; Snaith, H. J.; Giustino, F. Lead-Free Halide Double Perovskites via Heterovalent Substitution of Noble Metals. J. Phys. Chem. Lett. 2016, 7, 1254–1259.
  19. [20]McClure, E. T.; Ball, M. R.; Windl, W.; Woodward, P. M. Cs₂AgBiX₆ (X = Br, Cl) — New visible light absorbing, lead-free halide perovskite semiconductors. Chem. Mater. 2016, 6, 1348–1354.DOI
  20. [21]Volonakis, G.; Haghighirad, A. A.; Milot, R. L.; Sio, W. H.; Filip, M. R.; Wenger, B.; Johnston, M. B.; Herz, L. M.; Snaith, H. J.; Giustino, F. Cs₂InAgCl₆ : A New Lead-free Halide Double Perovskite with Direct Band Gap. J. Phys. Chem. Lett. 2017, 8, 772–778.DOI
  21. [22]Luo, J. et al. Efficient and stable emission of warm-white light from lead-free halide double perovskites. Nature 2018, 563, 541–545.DOI
  22. [23]Slavney, A. H.; Leppert, L.; Saldivar Valdes, A.; Bartesaghi, D.; Savenije, T. J.; Neaton, J. B.; Karunadasa, H. I. Small-Band-Gap Halide Double Perovskites. Angew. Chem. Int. Ed. 2018, 57, 12765–12770.DOI
  23. [24]Xue, J.; Wang, Z.; Comstock, A.; Wang, Z.; Sung, H. H. Y.; Williams, I. D.; Sun, D.; Liu, J.; Lu, H. Chemical Control of Magnetic Ordering in Hybrid Fe–Cl Layered Double Perovskites. Chem. Mater. 2022, 34, 2813–2823.DOI
  24. [25]Jöbbis, H. J.; Fykouras, K.; Reinders, J. W.; van Katwijk, J.; Dorresteijn, J. M.; Arens, T.; Vollmer, I.; Muscarella, L. A.; Leppert, L.; Hutter, E. M. Conduction Band Tuning by Controlled Alloying of Fe into Cs₂AgBiBr₆ Double Perovskite Powders. Adv. Funct. Mater. 2024, 34, 2306106.
  25. [26]Muscarella, L. A.; Hutter, E. M. Halide Double-Perovskite Semiconductors beyond Photovoltaics. ACS Energy Lett. 2022, 7, 2128–2135.DOI
  26. [27]Weng, Z.; Qin, J.; Umar, A.; Wang, J.; Zhang, X.; Wang, H.; Cui, X.; Li, X.; Zheng, L.; Zhan, Y. Lead-free cs2biagbr6 double perovskite-based humidity sensor with superfast recovery time. Adv. Funct. Mater. 2019, 29.
  27. [28]Pan, W. et al. Cs₂AgBiBr₆ single-crystal X-ray detectors with a low detection limit. Nat. Photonics 2017, 11, 726–732.
  28. [29]Palummo, M.; Berrios, E.; Varsano, D.; Giorgi, G. Optical Properties of Lead-Free Double Perovskites by Ab Initio Excited-State Methods. ACS Energy Lett. 2020, 5, 457–463.DOI
  29. [30]Biega, R. I.; Filip, M. R.; Leppert, L.; Neaton, J. B. Chemically Localized Resonant Excitons in Silver-Pnictogen Halide Double Perovskites. J. Phys. Chem. Lett. 2021, 12, 2057–2063.DOI
  30. [31]Cucco, B.; Katan, C.; Even, J.; Kepenekian, M.; Volonakis, G. Fine Structure of Excitons in Vacancy-Ordered Halide Double Perovskites. ACS Mater. Lett. 2023, 5, 52–59.
  31. [32]Kavanagh, S. R.; Savory, C. N.; Liga, S. M.; Konstantatos, G.; Walsh, A.; Scanlon, D. O. Frenkel Excitons in Vacancy-Ordered Titanium Halide Perovskites (Cs₂TiX₆). J. Phys. Chem. Lett. 2022, 13, 10965–10975.DOI
  32. [33]Biega, R.-I.; Chen, Y.; Filip, M. R.; Leppert, L. Chemical Mapping of Excitons in Halide Double Perovskites. Nano Lett. 2023, 23, 8155–8161.
  33. [34]Klarbring, J.; Singh, U.; Simak, S. I.; Abrikosov, I. A. Electronic structure of the magnetic halide double perovskites Cs₂(Ag,Na)FeCl₆ from first principles. Phys. Rev. Mater. 2023, 7, 044605.
  34. [35]Singh, U.; Klarbring, J.; Abrikosov, I. A.; Simak, S. I. Exploring magnetism of lead-free halide double perovskites: A high-throughput first-principles study. Phys. Rev. Mater. 2023, 7, 114404.DOI
  35. [36]Ning, W. et al. Magnetizing lead-free halide double perovskites. Sci. Adv. 2020, 6, eabb5381.DOI
  36. [37]Bartel, C. J.; Sutton, C.; Goldsmith, B. R.; Ouyang, R.; Musgrave, C. B.; Ghiringhelli, L. M.; Scheffler, M. New tolerance factor to predict the stability of perovskite oxides and halides. Sci. Adv. 2019, 5, eaav0693.DOI
  37. [38]Naik, A.; Ertural, C.; Dhamrait, N.; Benner, P.; George, J. A Quantum-Chemical Bonding Database for Solid-State Materials. Scientific Data 2023, 10, 610.DOI
  38. [39]Naik, A. A.; Dhamrait, N.; Ueltzen, K.; Ertural, C.; Benner, P.; Rignanese, G.-M.; George, J. A critical assessment of bonding descriptors for predicting materials properties. 2026; http://arxiv.org/abs/2602.12109, arXiv:2602.12109 [cond-mat].
  39. [40]Saleh, G.; Manna, L. The Predictive Power of Chemical Bonding Analysis in Materials: A Perspective on Optoelectronic Properties. J. Am. Chem. Soc. 2025, 147, 46705–46719.DOI
  40. [41]Walsh, A. Principles of Chemical Bonding and Band Gap Engineering in Hybrid Organic–Inorganic Halide Perovskites. J. Phys. Chem. C 2015, 119, 5755–5760.
  41. [42]Goesten, M. G.; Hoffmann, R. Mirrors of Bonding in Metal Halide Perovskites. J. Am. Chem. Soc. 2018, 140, 12996–13010.DOI
  42. [43]Maughan, A. E.; Ganose, A. M.; Scanlon, D. O.; Neilson, J. R. Perspectives and Design Principles of Vacancy-Ordered Double Perovskite Halide Semiconductors. Chem. Mater. 2019, 31, 1184–1195.DOI
  43. [44]Kieslich, G.; Sun, S.; Cheetham, T. An Extended Tolerance Factor Approach for Organic-Inorganic Perovskites. Chem. Sci. 2015, 6, 3430–3433.
  44. [45]Travis, W.; Glover, E. N. K.; Bronstein, H.; Scanlon, D. O.; Palgrave, R. G. On the application of the tolerance factor to inorganic and hybrid halide perovskites: a revised system. Chem. Sci. 2016, 7, 4548–4556.DOI
  45. [46]Goldschmidt, V. M. Die Gesetze der Krystallochemie. Naturwissenschaften 1926, 14, 477–485.
  46. [47]Ouyang, R.; Curtarolo, S.; Ahmetcik, E.; Scheffler, M.; Ghiringhelli, L. M. SISSO: A compressed-sensing method for identifying the best low-dimensional descriptor in an immensity of offered candidates. Phys. Rev. Mater. 2018, 2, 083802.DOI
  47. [48]Shannon, R. D.; Prewitt, C. T. Revised values of effective ionic radii. Acta Crystallogr. Sect. B 1970, 26, 1046–1048.DOI
  48. [49]Mentel, Ł. mendeleev – A Python resource for properties of chemical elements, ions and isotopes. 2014; https://github.com/lmmentel/mendeleev.
  49. [50]Bartel, C. J.; Clary, J. M.; Sutton, C.; Vigil-Fowler, D.; Goldsmith, B. R.; Holder, A. M.; Musgrave, C. B. Inorganic Halide Double Perovskites with Optoelectronic Properties Modulated by Sublattice Mixing. J. Am. Chem. Soc. 2020, 142, 5135–5145.DOI
  50. [52]Shuai, P.; Guo, Q.; Liao, L.; Xiang, M.; Su, K.; Mei, L. Structure, optical characteristics and temperature sensing performance studies of Cs₃YF₆: Er3+, Yb3+ up-conversion material with cryolite structure. J. Solid State Chem. 2022, 306, 122720.DOI
  51. [53]Kresse, G.; Hafner, J. Ab initio molecular-dynamics simulation of the liquid-metal–amorphous-semiconductor transition in germanium. Phys. Rev. B 1994, 49, 14251–14269.
  52. [54]Kresse, G.; Furthmüller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comput. Mater. Sci. 1996, 6, 15–50.DOI
  53. [55]Kresse, G.; Furthmüller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 1996, 54, 11169–11186.
  54. [56]Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 1994, 50, 17953–17979.
  55. [57]Kresse, G.; Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 1999, 59, 1758–1775.DOI
  56. [58]Dron­skowski, R.; Bloechl, P. E. Crystal orbital Hamilton populations (COHP): energy-resolved visualization of chemical bonding in solids based on density-functional calculations. J. Phys. Chem. 1993, 97, 8617–8624.DOI
  57. [59]Deringer, V. L.; Tchougréeff, A. L.; Dronskowski, R. Crystal Orbital Hamilton Population (COHP) Analysis As Projected from Plane-Wave Basis Sets. J. Phys. Chem. A 2011, 115, 5461–5466.DOI
  58. [60]Maintz, S.; Deringer, V. L.; Tchougréeff, A. L.; Dronskowski, R. Analytic projection from plane-wave and PAW wavefunctions and application to chemical-bonding analysis in solids. J. Comput. Chem. 2013, 34, 2557–2567.DOI
  59. [61]Maintz, S.; Deringer, V. L.; Tchougréeff, A. L.; Dronskowski, R. LOBSTER: A tool to extract chemical bonding from plane-wave based DFT. J. Comput. Chem. 2016, 37, 1030–1035.DOI
  60. [62]Nelson, R.; Ertural, C.; George, J.; Deringer, V. L.; Hautier, G.; Dronskowski, R. LOBSTER: Local orbital projections, atomic charges, and chemical-bonding analysis from projector-augmented-wave-based density-functional theory. J. Comput. Chem. 2020, 41, 1931–1940.DOI
  61. [63]Perdew, J. P.; Ruzsinszky, A.; Csonka, G. I.; Vydrov, O. A.; Scuseria, G. E.; Constantin, L. A.; Zhou, X.; Burke, K. Restoring the Density-Gradient Expansion for Exchange in Solids and Surfaces. Phys. Rev. Lett. 2008, 100, 136406.
  62. [64]Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865–3868.
  63. [65]Heyd, J.; Scuseria, G. E.; Ernzerhof, M. Hybrid functionals based on a screened Coulomb potential. J. Chem. Phys. 2003, 118, 8207–8215.
  64. [66]Heyd, J.; Scuseria, G. E.; Ernzerhof, M. Erratum: “Hybrid functionals based on a screened Coulomb potential” [J. Chem. Phys. 118, 8207 (2003)]. J. Chem. Phys. 2006, 124, 219906.
  65. [67]Blöchl, P. E.; Jepsen, O.; Andersen, O. K. Improved tetrahedron method for Brillouin-zone integrations. Phys. Rev. B 1994, 49, 16223–16233.
  66. [68]Ertural, C.; Steinberg, S.; Dronskowski, R. Development of a robust tool to extract Mulliken and Löwdin charges from plane waves and its application to solid-state materials. RSC Adv. 2019, 9, 29821–29830.DOI
  67. [69]Löwdin, P.-O. On the Non-Orthogonality Problem Connected with the Use of Atomic Wave Functions in the Theory of Molecules and Crystals. J. Chem. Phys. 1950, 18, 365–375.
  68. [70]Müller, P. C.; Ertural, C.; Hempelmann, J.; Dronskowski, R. Crystal Orbital Bond Index: Covalent Bond Orders in Solids. J. Phys. Chem. C 2021, 125, 7959–7970.
  69. [71]Mulliken, R. S. Electronic Population Analysis on LCAO-MO Molecular Wave Functions. J. Chem. Phys. 1955, 23, 1833.DOI
  70. [72]Belli, F.; Zurek, E.; Errea, I. A chemical bonding based descriptor for predicting the impact of quantum nuclear and anharmonic effects on hydrogen-based superconductors. 2025; http://arxiv.org/abs/2501.14420.
  71. [73]Naik, A. A.; Ueltzen, K.; Ertural, C.; Jackson, A. J.; George, J. LobsterPy: A package to automatically analyze LOBSTER runs. J. Open Source Softw. 2024, 9, 6286.DOI
  72. [74]McInnes, L.; Healy, J.; Melville, J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. 2020; http://arxiv.org/abs/1802.03426.
  73. [75]van der Maaten, L.; Hinton, G. Visualizing Data using t-SNE. J. Mach. Learn. Res. 2008, 9, 2579–2605.
  74. [78]Scheidgen, M. et al. NOMAD: A distributed web-based platform for managing materials science research data. J. Open Source Softw. 2023, 8, 5388.
  75. [79]Rudzinski, J. F.; Albino, A.; Bereau, T.; Daelman, N.; Ladines, A. N.; Mohr, B.; Pedersen, J.; Walter, L. J. NOMAD Utility Workflows. 2026; https://zenodo.org/records/20171309.DOI
  76. [80]Luc Walterbos NOMAD dataset: HalideDoublePerovskite_HSE06_Lobster. https://nomad-lab.eu/prod/v1/gui/dataset/doi/10.17172/nomad.wb9y-b8j7.DOI
  77. [81]Ong, S. P.; Richards, W. D.; Jain, A.; Hautier, G.; Kocher, M.; Cholia, S.; Gunter, D.; Chevrier, V. L.; Persson, K. A.; Ceder, G. Python Materials Genomics (pymatgen): A robust, open-source python library for materials analysis. Comp. Mater. Sci. 2013, 68, 314–319.DOI
  78. [82]Walterbos, L. Luccerboi/HDP_WorkFlow_Analysis: Archived for publication. 2026; https://zenodo.org/records/20598121.DOI
  79. [83]LWalterbos Luccerboi/HDP_UMAP_Explorer. https://github.com/Luccerboi/HDP_UMAP_Explorer, original-date: 2026-04-27T14:35:01Z.
  80. [84]Walterbos, L. Luccerboi/HDP_UMAP_Explorer: Archived for publication. 2026; https://zenodo.org/records/20598093.DOI
  81. [85]Riebesell, J.; Yang, H.; Goodall, R.; Baird, S. G. Pymatviz: visualization toolkit for materials informatics. 2022; https://github.com/janosh/pymatviz.
  82. [86]Kent, G. T.; Morgan, E.; Albanese, K. R.; Kallistova, A.; Brumberg, A.; Kautzsch, L.; Wu, G.; Vishnoi, P.; Seshadri, R.; Cheetham, A. K. Elusive Double Perovskite Iodides: Structural, Optical, and Magnetic Properties. Ang. Chem. Int. Ed. 62, e202306000.
  83. [87]Hellenbrandt, M. The Inorganic Crystal Structure Database (ICSD)—Present and Future. Crystallogr. Rev. 2004, 10, 17–22.DOI
  84. [88]Schade, L.; Wright, A. D.; Johnson, R. D.; Dollmann, M.; Wenger, B.; Nayak, P. K.; Prabhakaran, D.; Herz, L. M.; Nicholas, R.; Snaith, H. J.; Radaelli, P. G. Structural and Optical Properties of Cs2AgBiBr6 Double Perovskite. ACS Energy Lett. 2019, 4, 299–305.
  85. [89]Schröder, L.; Keller, H.-L. Darstellung und Kristallstruktur von Cs2HgPdCl6, eine Verzerrungsvariante der Chloroperowskite. Z. Anorg. Allg. Chem. 1991, 603, 69–76.DOI
  86. [90]Ji, G.; Xiao, Z. Jahn-Teller Distortion-Stabilized Halide Double Perovskites with Un-usual Rock-Salt-type Ordering of Divalent B-Site Cations. Chem. Mater. 2022, 34, 8207–8212.
  87. [91]Maughan, A. E.; Ganose, A. M.; Bordelon, M. M.; Miller, E. M.; Scanlon, D. O.; Neil-son, J. R. Defect Tolerance to Intolerance in the Vacancy-Ordered Double Perovskite Semiconductors Cs2SnI6 and Cs2TeI6. J. Am. Chem. Soc. 2016, 138, 8453–8464.DOI
  88. [92]Klein, J.; Kampermann, L.; Mockenhaupt, B.; Behrens, M.; Strunk, J.; Bacher, G. Limitations of the Tauc Plot Method. Adv. Funct. Mater. 2023, 33, 2304523.
  89. [93]Scherpelz, P.; Govoni, M.; Hamada, I.; Galli, G. Implementation and Validation of Fully Relativistic GW Calculations: Spin-Orbit Coupling in Molecules, Nanocrystals, and Solids. J. Chem. Theory Comput. 2016, 12, 3523–3544.DOI
  90. [94]Wiktor, J.; Rothlisberger, U.; Pasquarello, A. Predictive Determination of Band Gaps of Inorganic Halide Perovskites. J. Phys. Chem. Lett. 2017, 8, 5507–5512.DOI
  91. [95]Leppert, L.; Rangel, T.; Neaton, J. B. Towards predictive band gaps for halide per-ovskites: Lessons from one-shot and eigenvalue self-consistent GW. Phys. Rev. Mater. 2019, 3, 103803.
  92. [96]Leppert, L. Excitons in metal-halide perovskites from first-principles many-body per-turbation theory. J. Chem. Phys. 2024, 160, 050902.
  93. [97]Filip, M. R.; Leppert, L. Halide perovskites from first principles: from fundamental optoelectronic properties to the impact of structural and chemical heterogeneity. Electron. Struct. 2024, 6, 033002.DOI
  94. [98]Choudhary, D.; Kaur, M.; Sharma, G.; Palsaniya, R.; Loylaka, S.; Singh, S.; Verma, U.; Yadav, P.; Manendra Impact of halide variation on the optoelectronic properties of double perovskites. Sci. Rep. 2025, 15, 33629.
  95. [99]Bechir, M. B.; Houcine Dhaou, M. Lead-free double perovskite Cs₂MBiCl₆ (M = Ag, Cu): insights into the optical, dielectric, and charge transfer properties. RSC Adv. 2023, 13, 17750–17764.
  96. [100]Tran, T. T.; Panella, J. R.; Chamorro, J. R.; Morey, J. R.; McQueen, T. M. Designing indirect–direct bandgap transitions in double perovskites. Mater. Horiz. 2017, 4, 688–693.DOI
  97. [101]Jiang, W.; Sun, R.; Wang, S.; Yu, Y.; Qi, L.; Pan, K. Improved optical properties of lead-free double perovskite Cs₂NaBiCl₆ nanocrystal via K ions doping. J. Alloys Compd. 2023, 960, 170871.
  98. [102]Lamba, R. S.; Basera, P.; Singh, S.; Bhattacharya, S.; Sapra, S. Lead-Free Alloyed Double-Perovskite Nanocrystals of Cs₂(NaₓAg₁₋ₓ)BiBr₆ with Tunable Band Gap. J. Phys. Chem. C 2021, 125, 1954–1962.DOI
  99. [103]Zhou, W.; Han, P.; Zhang, X.; Zheng, D.; Yang, S.; Yang, Y.; Luo, C.; Yang, B.; Hong, F.; Wei, D.; Lu, R.; Han, K. Lead-Free Small-Bandgap Cs₂CuSbCl₆ Double Perovskite Nanocrystals. J. Phys. Chem. Lett. 2020, 11, 6463–6467.DOI

Paper details

vlnc:W_kzx4k9v3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents