Article https://doi.org/10.

1038/s41467-023-43436-3

Design principles for sodium superionic

conductors

Received: 30 August 2023 Shuo Wang1,5, Jiamin Fu2,3,5, Yunsheng Liu1, Ramanuja Srinivasan Saravanan1,
Jing Luo2, Sixu Deng2, Tsun-Kong Sham 3, Xueliang Sun 2 & Yifei Mo 1,4
Accepted: 9 November 2023

Motivated by the high-performance solid-state lithium batteries enabled by

Check for updates lithium superionic conductors, sodium superionic conductor materials have
great potential to empower sodium batteries with high energy, low cost, and
1234567890():,;
1234567890():,;

sustainability. A critical challenge lies in designing and discovering sodium

superionic conductors with high ionic conductivities to enable the develop-
ment of solid-state sodium batteries. Here, by studying the structures and
diffusion mechanisms of Li-ion versus Na-ion conducting solids, we reveal the
structural feature of face-sharing high-coordination sites for fast sodium-ion
conductors. By applying this feature as a design principle, we discover a
number of Na-ion conductors in oxides, sulfides, and halides. Notably, we
discover a chloride-based family of Na-ion conductors NaxMyCl6 (M = La–Sm)
with UCl3-type structure and experimentally validate with the highest reported
ionic conductivity. Our findings not only pave the way for the future devel-
opment of sodium-ion conductors for sodium batteries, but also consolidate
design principles of fast ion-conducting materials for a variety of energy
applications.

Superionic conductor (SIC) is a unique type of materials exhibiting While Na-SICs beta-alumina14 and NASICON15 achieved high Na
exceptionally high ionic conductivities, serving as critical components ionic conductivities back in 1970s, there are many discoveries of Li-SIC
in a wide range of devices for energy storage and conversion, including families in the past two decades (Supplementary Figs. 1 and 2),
solid-state batteries, solid-oxide fuel cells, solid-oxide electrolyzers, including Li-garnet Li7La3Zr2O12 (LLZO) oxides16, Li10GeP2S12 (LGPS)1,
and ceramic membranes1–4. Among them, the lithium SICs, which Li7P3S11 sulfides17, Li-argyrodite Li6PS5X (X = Cl, Br, I)18, Li3MX6 (M = Y,
replace liquid electrolytes in lithium-ion batteries as solid electrolytes, Sc, In, Er, etc., X = Cl, Br) halides19–21, with high Li ionic conductivities
enable the next-generation solid-state lithium batteries with improved σRT on the order of 1–10 mS/cm at room temperature (RT). By under-
safety, high energy density, and long cycle life5–7. Thanks to the standing the Li+ diffusion mechanisms in these SICs22–24, scientists
abundance and low cost of sodium resources, sodium SICs also attract established multiple design principles for Li-SICs, for example, based
great interest as solid electrolytes for solid-state sodium batteries8–10 on body-centered cubic (bcc) anion framework25, concerted
and sodium-sulfur batteries11–13, which are economical and sustainable migration26, and corner-sharing crystal structural framework27. By
alternatives to current lithium-ion batteries. However, only a few successfully employing these design principles, many Li-SICs, includ-
materials are known as SICs, impeding the further development of ing LiZnPS4, LiTaSiO5, LiGa(SeO3)2, were discovered from first princi-
these novel energy technologies. A long-standing challenge in mate- ples computation and experimentally verified27–29.
rials science is how to rationally design and discover SIC materials with Despite the rapid advancement in the field of Li-SICs, the devel-
high ionic conductivities. opment and discovery of Na-SICs have been greatly lagging. Given the

1
Department of Materials Science and Engineering, University of Maryland, College Park, MD 20742, USA. 2Department of Mechanical and Materials Engi-
neering, University of Western Ontario, London, ON N6A 5B9, Canada. 3Department of Chemistry, University of Western Ontario, London, ON N6A 5B7,
Canada. 4Maryland Energy Innovation Institute, University of Maryland, College Park, MD 20742, USA. 5These authors contributed equally: Shuo Wang, Jiamin
Fu. e-mail: [email protected]; [email protected]

Nature Communications | (2023)14:7615 1

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Article https://doi.org/10.1038/s41467-023-43436-3

chemical similarities between Li+ and Na+, a common approach of conductors NaxMyCl6 (M = La–Sm), is discovered and successfully
developing Na-SICs is to make Na counterparts of known Li SICs, but synthesized with σRT of 1.4 mS/cm, the highest among sodium halides.
the outcomes are underwhelming. For example, while Li-SIC LGPS
(σRT = 12 mS/cm) is a major breakthrough1, its Na-analogy Na10SnP2S12 Results
in the same structure only achieved a much lower ionic conductivity Different site preferences for Na+ and Li+ in crystal structures
σRT of 0.4 mS/cm.30 For another inspiring discovery of halide Li-SIC We first analyze and compare the structures of representative Li+ and
Li3MX6 (M = Y, Sc, In, Er, etc., X = Cl, Br) with high ionic conductivity Na+ SICs (Fig. 1), and identify the differences in the local geometry of
σRT > 1 mS/cm,19–21 the Na-halide counterpart Na2ZrCl6 with the same Li+/Na+ sites in forming the diffusion channels. There are significant
hexagonal close-packed (hcp) Cl-anion framework as Li3YCl6 only differences between the Li+/Na+ site coordination number (CN), which
exhibits a limited σRT of 0.02 mS/cm.31 Moreover, Li-garnet16 and Li- is the number of the nearest neighbor anions (Methods). Whereas Li+
argyrodite18,22, which are successfully demonstrated as solid electro- occupies and migrates among tetrahedral (Tet) sites in the LGPS Li-SIC,
lytes for solid-state lithium batteries with excellent cell Na+ occupies sites with higher CNs of ≥ 5 in Na-SICs, such as beta-
performances32,33, have no Na counterparts. For Na SICs, W-doped alumina (CN = 6–8), NASICON (CN = 5, 6, 8), Na3SbS4 (CN = 8) (Fig. 1),
Na3SbS434 has a reported σRT of 32 mS/cm, but its structures differ and Na3PS4 (CN = 6) (Supplementary Fig. 3). The preferences of Na+ for
significantly from its Li-counterparts. As clearly indicated by these higher CN can be understood by the larger Na+ radius (rNa+ = 1.02 Å)
facts, the crystal structures that yield the low energy barriers for Li+ and compared to Li+ (rLi+ = 0.76 Å) according to Pauling’s rules35. Among all
Na+ diffusion are different, but are not yet understood. It is not clear Na- and Li-containing oxides, sulfides, and chlorides, the preference
how the knowledge derived from known Na-SICs can be utilized to for high-CN Na+ sites is general as confirmed by our analyses (Sup-
design and discover more Na-SICs. One cannot simply duplicate the plementary Fig. 4).
design principles for Li-ion conductors to discover Na-ion conductors.
Given many recent discoveries in Li-SICs with σRT higher than Why the design principles for Li-ion conductors cannot be
1–10 mS/cm, the discovery of Na-SICs is an under-explored opportu- duplicated for Na-ion conductors?
nity. In comparison to Li-SICs, no oxide structures of Na-SICs with Here we illustrate the design principles for Li-ion conductors cannot be
σRT > 1 mS/cm was discovered since 1970s (Supplementary Table 1), duplicated for Na-ion conductors, because of the different preferred
and the halide Na-SICs still show σRT two orders of magnitude lower site-coordination of Na+ and Li+. We investigate and simulate the
than Li-halides31. While Li-SICs empower solid-state lithium batteries energy barrier of Li+ and Na+ migration in the model systems with fixed
with ever-improving cell performances, the limited availability of Na- bcc, hcp, and face-centered cubic (fcc) anion sublattices with no cation
SICs has been impeding the development and innovation of sodium (Fig. 2a–c). In a bcc anion framework, such as in LGPS and Li7P3S11, Li+
batteries. Thus the design principles applicable to Na-SICs are urgently sits at tetrahedral (Tet) sites and migrates by crossing an anion-triangle
needed. bottleneck to another face-sharing Tet site, giving a low energy barrier
In this study, we first reveal the fundamental differences between of 0.12 eV for Li+ migration (Fig. 2a). Proposed and demonstrated by
the crystal structures and diffusion mechanisms of Li+ versus Na+ by Wang et al.25, the bcc anion framework as a design principle for Li-SICs
analyzing Li- and Na-conducting oxides, sulfides, and halides. Through has successfully led to the discovery of a Li-SIC LiZnPS429. However, in
this understanding, a unique feature of fast Na-ion conductors is Na-containing compounds, because of the strong preference of high-
identified, and is then formulated as a design principle. Applying our CN Na+ sites (Supplementary Fig. 4), it would be difficult to form per-
design principle in a high-throughput computational screening, we colation diffusion channels solely by Tet Na+ sites, as desired in the bcc-
discover over a dozen of structural families of Na-ion conductors with anion-framework design principle. There has been no Na-containing
high ionic conductivities. In particular, a halide family of Na-ion material with the bcc-anion framework.

Fig. 1 | The ion diffusion channel in Li/Na-ion conductors. The Li+/Na+ sites (green) coordinated with O2-/S2-/Cl- anions (yellow) connected to form the diffusion channel in
representative a–c sodium and d–f lithium SICs.

Nature Communications | (2023)14:7615 2

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Article https://doi.org/10.1038/s41467-023-43436-3

Fig. 2 | Lithium-ion and sodium-ion diffusion in model anion sublattices. volume as the real materials, O3-type LiCoO2 (Oct: 12.0 Å3) and NaCoO2 (Oct: 15.7
a–c The diffusion pathways (left) and corresponding energy profile (right) for Å3) for fcc O2−, Li2ZrCl6 (Oct: 23.8 Å3) and Na2ZrCl6 (Oct: 30.9 Å3) for hcp Cl-,
single Li+ (green) and Na+ (red) migration in fixed a body-centered cubic (bcc) S2−, Li10GeP2S12 (Tet: 7.4 Å3) and Na10SnP2S12 (Tet: 9.6 Å3) for bcc S2−. d–f The energy
b face-centered cubic (fcc) O2−, c hexagonal close-packed (hcp) Cl− anion sublattice. barrier of Na+ (red) and Li+ (green) migration as a function of site volume in the fixed
The fixed anion lattice is set to have the octahedral (Oct) and tetrahedral (Tet) site d bcc S2−, e fcc O2−, f hcp Cl- anion sublattice.

For the close-packed fcc and hcp anion sublattices, as com- Na2ZrCl6 with the same hcp Cl-anion framework as Li3YCl6, as
monly found in Li-chlorides and bromides SICs Li3MX6, Li+ migration reported in experiments31.
has low energy barriers of 0.2–0.3 eV (Supplementary Figs. 7 and 8), In summary, low-barrier Na+ migration is difficult to be realized in
which are sufficiently low for fast Li-ion conductors36. The Na+-con- these typical materials structures given by the design principles for Li-
ducting O3-type NaMO2 (M = Co, Mn, Ni, etc.) also has an fcc anion ion conductors (Fig. 2d–f). While the mechanisms, such as concerted
sublattice, in which Na+ occupies octahedral (Oct) sites (CN = 6) and migration23,26, poly-anion rotation37,38, and phonon lattice effect39,40,
migrates between Oct sites through intermediate Tet site (CN = 4). may further facilitate ion diffusion (see section “High-throughput
The intermediate Tet-site has a higher site energy than the Oct sites, discovery for Na-ion conductors”), having the structural framework
as the high CN preference of Na+ makes Tet-sites unfavorable, thus with the flat energy landscape is the first requirement to have fast ion
giving a high migration barrier of >1.0 eV along the Oct-Tet-Oct conductors. Compared to Li+, Na+ diffusion in solid crystal structures
pathway (Fig. 2b and Supplementary Figs. 7 and 8). In the hcp anion have two critical limitations. (1) Na+ is unfavorable in the Tet (low-CN)
sublattice, as in Na2ZrCl6, this Oct-Tet-Oct pathway is also required sites, but typical structures give migration pathways with intermediate
for 3D conduction. In 1D Oct-Oct pathways along the c-axis, Na+ Tet sites, causing a high migration-energy barrier (Figs. 2b and 3). (2)
migrates among face-sharing Oct sites across an anion-triangle Na+ requires a larger bottleneck size of the diffusion channel. As
bottleneck (Fig. 2c). This anion-triangle bottleneck has an appro- quantified in our analyses on the percolation radius pr, which is the
priate size for Li+ migration but is too small for Na+. As a result, the maximum radius of a sphere that can percolate the structure across at
hcp Cl- anion sublattice gives a low Li+ migration barrier of ~0.3 eV in least one dimension (Methods), Na SICs have much larger percolation
Li3MX6 halide Li-SICs, but gives a high Na+ migration barrier of radii (pr = 0.88–1.28 Å) than Li SICs (pr = 0.52–0.72 Å) (Supplementary
~0.9 eV (Fig. 2c). This explains the limited Na-ion conduction in Fig. 4 and Supplementary Table 2). As shown above (Fig. 2), typical

Nature Communications | (2023)14:7615 3

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Article https://doi.org/10.1038/s41467-023-43436-3

Fig. 3 | Design principles for sodium-ion conductors. a The schematics illustrate in the fixed O-anion sublattice at the lattice volume of 19.3 Å3 per O2−. b The com-
(lower) the Prism-Prism pathway between face-sharing high-CN sites in P2-type parison of the percolation radii pr and the average CNs of Li/Na sites along the
NaMO2, in comparison to (upper) the Oct-Tet-Oct pathway with intermediate Tet diffusion channels in Li-ion and Na-ion conductors showing a clear distinction.
site in O3-type NaMO2, with (middle) the calculated energy profile for Na+ migration

crystal structures cannot offer such large bottlenecks for low-barrier their diffusion channels, and the Na2ZrCl6 with hcp anion framework
Na-ion diffusion. How do the structures of Na-SICs overcome these has a small bottleneck size (pr = 0.73 Å). By contrast, all Li-SICs have
limitations to achieve fast Na-ion conduction? low average CNs (Fig. 3b). For example, LGPS and Li7P3S11 with face-
sharing Tet sites (average CN = 4) (Figs. 1 and 3b) show the highest
Design principles for Na-ion conductors ionic conductivity, according to the bcc-anion-framework design
Here we identify and propose the feature of face-sharing high-CN sites principle25. Other non-bcc Li-SICs, such as LLZO and Li3ScCl6, have
for fast Na-ion conductors. Specifically, the Na+ diffusion channel higher average CN because of the non-Tet sites along the diffusion
should be comprised of only high-CN sites (CN ≥ 5) that are face- pathways, and in general show ionic conductivities not as high as those
sharing with a large bottleneck size to form percolation in the crystal with bcc-anion framework (Fig. 3b). This comparison chart (Fig. 3b)
structure. While high-CN Na+ sites are general (Supplementary Fig. 4), clearly demonstrates the key feature of face-sharing high-CN sites in
having these high-CN sites be face-sharing is unique among crystal Na-SICs highly distinct from Li-SICs.
structures (Fig. 1 and Supplementary Fig. 3). For example, in Na beta-
alumina (Fig. 1), the Na+ occupies the trigonal prismatic (CN = 6) or High-throughput discovery for Na-ion conductors
capped trigonal prismatic sites (CN = 7, 8), which are connected via Here, our design principle is employed in a high-throughput compu-
face-sharing O2--rectangle with a large bottleneck size (pr = 1.28 Å). tation screening to discover Na SICs (Fig. 4a), among Na-containing
Similarly, the P2-type NaCoO2, (Figs. 1 and 3a) which is reported to have oxides, sulfides, and chlorides in the Inorganic Crystal Structure
high RT ionic conductivity up to 6 mS/cm,41 also has a diffusion net- Database (ICSD)42. In addition to basic checks and practical con-
work of equivalent prismatic (Prism) Na sites connected by face- siderations of materials (Methods), two screening criteria following
sharing O2−-rectangle with a large bottleneck size (pr = 0.92 Å) (Fig. 3a). our design principle were employed: (1) the diffusion channel consists
The Prism-Prism pathway has a low energy barrier of 0.26 eV for Na+ of only high-CN sites (CN ≥ 5) that are face-sharing and are connected
migration in the fixed O2- anion sublattice model of the P2-type NaMO2 within a distance of 3.1 Å for oxides and 3.5 Å for sulfides/chlorides; (2)
(Fig. 3a). By contrast, for the structures with the high-CN sites not face- a large percolation radius pr > 0.85 Å for sulfides/chlorides and
sharing, e.g. the fcc anion sublattice as in O3-type NaMO2, there is pr > 0.90 Å for oxides. Using these screening criteria, all known sodium
unfavorable Tet site as intermediate along the diffusion pathway, SICs, including NASICON, beta-alumina, Na3SbS4, Na3PS4, and
causing a high energy barrier of 0.76 eV at the same lattice volume of Na11Sn2PS12, are identified (Supplementary Table 3), and 35 unique
P2 (Fig. 3a). Face-sharing high-CN sites allow a direct Na+ hopping structures are discovered as candidates of Na-SICs. These candidate
among equivalent high-CN sites with a large bottleneck size and structures are further studied by aliovalent substitution to tune Na+
without an unfavorable intermediate Tet site (Fig. 3a), hence over- content, in order to enhance ionic conductivity through differ-
coming the aforementioned limitations of Na+ diffusion. ent mobile-ion concentration24,28,29 or activating concerted migration
In Fig. 3b, we compare Na-ion and Li-ion conductors for their mechanisms23,29,43 (“Methods”). Those substituted materials with good
average CN of Li+/Na+ sites along the diffusion channels and their stability (energy above hull <100 meV/atom) are then evaluated for
bottleneck sizes (pr). The known Na-SICs, such as NASICON, beta-alu- ionic conductivity using ab initio molecular dynamics (AIMD) simula-
mina, and Na3SbS4, have high average CNs ≥ 6 along their diffusion tions. Other mechanisms that may facilitate ion migration, such as
channels in their structures and also have large bottleneck size concerted migration26,37, cooperative polyanion rotation37,38, and
(pr ≥ 0.88 Å) (Fig. 3b). By contrast, other Na-containing materials that phonon effects39,40, if occur in the corresponding materials, would be
do not exhibit face-sharing high-CN sites and large bottleneck size captured in the AIMD simulations. Among them, 19 Na-SICs with
show lower conductivity. For example, O3-type NaMO2 structure and σRT > 0.1 mS/cm are discovered (Table 1). Given the large statistical
Na3YCl6 have lower average CN due to the intermediate Tet-site along variances of diffusivities and the extrapolation of the Arrhenius

Nature Communications | (2023)14:7615 4

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Article https://doi.org/10.1038/s41467-023-43436-3

Fig. 4 | Discovery of sodium superionic conductors. a High-throughput screen- Na0.86SmTa0.43Cl6, NaLa0.95Ta0.43Cl6, NaCe0.83Ta0.5Cl6, and NaNd0.83Ta0.5Cl6. f The
ing of Na-containing oxides, sulfides, and chlorides. b–d The crystal structures of Rietveld refinements of synchrotron-based diffraction pattern and g the fitting
representative discovered sodium SICs and (inset) their diffusion channels consist result of the pair distribution function of NaLa0.95Ta0.43Cl6.
of face-sharing high-CN sites. e The high experimental Na+ conductivities σRT of

relations, the results of AIMD simulations should only be interpreted as simulations, both exhibited fast Na+ conduction along the 1D channels
the confirmation of these structure frameworks are fast Na-ion con- with a low activation barrier of 0.13 eV and 0.15 eV, respectively
ducting if an appropriate level of aliovalent doping can be achieved. All (Table 1 and Supplementary Fig. 14). Experimentally, we successfully
these structures of discovered Na-SICs show the unique feature of synthesized this family of Na3xM2-xCl6-containing (M = La, Ce, Nd, Sm)
face-sharing high-CN sites (Fig. 4b–d), confirming our design principle. halide Na+ conductors using ball-milling methods (Methods), and
Among the discovered oxides Na-SICs, Na0.67Ti0.33Ga4.67O8 (doped acquired high ionic conductivities of 1.2, 1.4, 0.22, 0.15 mS/cm at 25 °C
from ICSD-34196), from the tunneled alkali titan-gallate family44, has a for Na0.86SmTa0.43Cl6, NaLa0.95Ta0.43Cl6, NaCe0.83Ta0.5Cl6, and
low activation barrier (Ea = 0.14 eV) and a high conductivity (σRT = 8.8 NaNd0.83Ta0.5Cl6, respectively (Fig. 4e). The X-ray diffraction pattern
mS/cm) in AIMD simulations (Supplementary Fig. 12). Its crystal struc- (XRD) verified the UCl3-type structure as the dominant phase (Sup-
tural framework consists of connected GaO4 tetrahedra and GaO6/TiO6 plementary Fig. 16). The synchrotron-based XRD refinement results
(Fig. 4b), and Na+ occupies capped triagonal prismatic sites (CN = 7), and pair distribution function of NaLa0.95Ta0.43Cl6 (Fig. 4f, g and
which are face-sharing with O-rectangle bottlenecks with a large bot- Supplementary Tables 12 and 13) confirmed the dominant crystalline
tleneck (pr = 1.07 Å). In another discovered Na-SIC Na1.33Mg0.67Ti7.33O16 phase NaLa1.67Cl6, and additionally NaTaCl6. The NaTaCl6 is formed as
(doped from ICSD-50764)45, Na+ migrates between equivalent eight- secondary phase in the inter-grain, and is known to have relatively
coordinated sites through a large O-rectangle bottleneck (pr = 1.28 Å) lower ionic conductivity (σRT = 0.045 mS/cm). The NaLa0.95Ta0.43Cl6
(Fig. 4c), resulting in fast 1D diffusion (Ea = 0.30 eV and σRT = 1.7 mS/cm). composite of NaLa1.67Cl6 and NaTaCl6 achieved the highest reported
While the fast ion conduction is confirmed in the bulk phases of these σRT of 1.4 mS/cm (Fig. 4e and Supplementary Fig. 17), a significant
materials, those materials that have 1D fast diffusion channels may be improvement over the previous halide Na-ion conductor Na2ZrCl6 with
susceptible to the blocking effect of defects and grain boundaries5,46,47, σRT of 0.02 mS/cm31. Given that the secondary phase NaTaCl6 has a
which deserve future studies. relatively lower ionic conductivity, the bulk phase of NaLa1.67Cl6
A halide family NaxMyCl6 with UCl3-type structure48 with a wide should have high ionic conductivity, in good agreement with the AIMD
range of compositions (M = lanthanides, x = 0–1, y = 1.67–2) is dis- simulations (Supplementary Fig. 17). This discovery of a chloride Na-
covered as Na-SICs. In contrast to Li3MX6 (M = Y, Sc, In, Er, etc., X = Cl, SIC family with the highest reported σRT is a strong validation of our
Br) halide Li-SICs with closed-packed anion framework, the structure design principle for fast Na-ion conductors.
of NaxMyCl6 exhibits face-sharing octahedral Na+ sites forming 1D
diffusion channel along c-axis (Fig. 4d). These octahedral sites are Discussion
distorted, thus enlarge the Cl-triangle bottleneck (pr = 1.16 Å). We As the key underlying mechanism of our design principle, the unique
evaluate NaSm2Cl6 and NaLa1.67Cl6 for Na+ diffusion. In AIMD feature of face-sharing high-CN sites gives direct ion-migration

Nature Communications | (2023)14:7615 5

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Article https://doi.org/10.1038/s41467-023-43436-3

Table 1 | Sodium superionic conductors predicted by AIMD simulations

ICSD-IDs Original composition Origin/dopant Doped composition Ehull Ea (eV) σ at 300 K Error bound [σmin,
(meV/atom) (mS/cm) σmax] (mS/cm)
78919 Na7Y2P7O24 P5+/Mo6+ Na6.5Y2Mo0.5P6.5O24 23 0.23 ± 0.07 0.5 [0.03, 9.4]
50764 Na1.7Cr1.7Ti6.3O16 Cr3+/Ti4+, Mg2+ Na1.33Mg0.67Ti7.33O16 31 0.30 ± 0.04 1.7 [0.3, 9.3]
34196 Na0.7Ti0.3Ga4.7O8 Ga3+/Ti4+ Na0.67Ti0.33Ga4.67O8 20 0.14 ± 0.05 8.8 [1.5, 53]
108816 NaTi2Ga5O12 Ga3+/Ti4+ Na0.67Ti2.33Ga4.67O12 10 0.26 ± 0.05 0.4 [0.04, 3.3]
155643 Na0.8Ti1.2Ga4.8O10 Ga3+/Ti4+ Na0.67Ti1.33Ga4.67O10 26 0.26 ± 0.05 0.8 [0.1, 5.0]
163234 Na2V3O7 V4+/V5+ Na1.33V3O7 40 0.34 ± 0.06 0.1 [0.01, 1.4]
262512 Na3Nb4As3O19 O2−/F− Na2Nb4As3O18F 43 0.24 ± 0.07 1.0 [0.06, 17.9]
39248 NaTiPO5 – – 29 0.19 ± 0.06 4.3 [0.3, 64]
39788 NaGeSbO5 Ge4+/P5+ Na0.75Ge0.75P0.25SbO5 29 0.18 ± 0.05 7.2 [0.7, 76]
239705 Na2ZnGe2S6 Ge4+/Zn2+ Na2.25Zn1.125Ge1.875S6 20 0.25 ± 0.06 0.7 [0.08, 6.9]
300175 Na5InS4 In3+/Sn4+ Na4.5In0.5Sn0.5S4 46 0.35 ± 0.04 0.2 [0.03, 0.8]
33236 Na6ZnS4 Zn2+/Ga3+ Na5.5Zn0.5Ga0.5S4 23 0.32 ± 0.05 0.2 [0.02, 1.8]
62579 Na5FeS4 Fe3+/In3+, Sn4+ Na4.75In0.75Sn0.25S4 62 0.28 ± 0.04 1.9 [0.4, 9.1]
234888 Na3ZnGaS4 Zn2+/Ga3+ Na2.75Zn0.75Ga1.25S4 42 0.25 ± 0.06 0.5 [0.04, 6.2]
200983 Na0.7Ti0.3Cr0.7S2 Cr3+/Ti4+ Na0.67TiS2 3 0.23 ± 0.04 2.2 [0.4, 14]
74928 NaSm2Cl6 – – 53 0.13 ± 0.03 131 [41 430]
Sm2+/La3+ NaLa1.67Cl6 9 0.15 ± 0.03 56.5 [13, 240]
69343 Na2MgCl4 Mg2+/Er3+ Na1.33Mg0.33Er0.67Cl4 44 0.35 ± 0.06 0.1 [0.01, 1.3]
11165 Na2Ti3Cl8 Ti2+/Ti3+ Na1.67Ti3Cl8 0 0.30 ± 0.07 0.2 [0.01, 3.5]

pathways among equivalent, favorable Na+ sites, with small CN changes distortion of mobile-ion sites is a key feature of Li-/Na-SICs, bringing
and a large bottleneck, thus giving a low energy barrier (Fig. 3a). By multiple beneficial effects for fast ion diffusion.
contrast, in the structures with Na+ sites that are not face-sharing, the Overall, these insights consolidate previous understandings on
diffusion pathways include intermediate Tet sites, which are unfavor- multiple features and mechanisms about the crystal structure frame-
able for Na+ and thus cause high energy barriers (Figs. 2c and 3a). In works of SICs. While this work focuses on the crystal structural fra-
comparison, in the Li-SICs with a bcc anion framework, the diffusion meworks that give low energy barriers for ion diffusion, more complex
channels of face-sharing Tet-sites25, which are generally favorable for Li+, mechanisms e.g. concerted migration, cooperation motion, and
lead to the lowest energy barrier (Figs. 2a and 3b)25. The high-CN pre- phonon-assisted hopping, may be further added to devise more
ference of Na+ versus Li+ explains why the structures of Li- and Na-SICs detailed design principles in the future. Furthermore, the effects of
are different and why the optimal Li-SIC structures cannot be duplicated grain boundaries in addition to the bulk-phase fast ion conduction
as Na-SICs (Figs. 1 and 3b). Regardless of the coordination preferences should be further studied in the future. As demonstrated in designing
of the mobile-ions, the crystal structures of fast ion conductors should Na-SICs, these generalized design principles can be extended and
form direct migration pathways among equivalent favorable sites to applied for other types of fast ion conductors, facilitating the future
minimize energy barrier. Here the design principles and desired fea- design and discovery of SICs across vast materials space.
tures for fast Li+ and Na+ conductors are consolidated and generalized.
Furthermore, this generalized design principle can unify the Methods
understanding of Li-SICs with or without bcc-anion-framework. For DFT calculation
example, in the LLZO garnet (Fig. 1), the distorted Oct-Li sites can be All the calculations were carried out using Vienna Ab initio Simulation
considered as a split into two Tet-Li sites and thus form face-sharing Package (VASP)51 based on density functional theory (DFT) using
Tet-sites channel for Li+ diffusion (Fig. 1), similar to that in LGPS, thus Perdew–Burke–Ernzerhof (PBE)52 generalized gradient approximation
shows relatively flat potential energy landscape. By considering this (GGA) described by the projector-augmented-wave (PAW) approach.
distorted site splitting into multiple Tet sites, other non-bcc-anion- The convergence parameters used in all static DFT calculations were
framework Li-SICs, such as LLZO, NASICON, Li6PS5X (X = Cl, Br, I) (Fig. 1 set to be consistent with the Materials Project53.
and Supplementary Fig. 3), can be understood as face-sharing tetra-
hedral sites as in the bcc-anion-framework21. The mechanisms of site Diffusion in fixed anion sublattice model
distortion are also attributed to the flattening of the energy landscape We performed the nudged elastic band (NEB) calculations to evaluate
by raising site energies as reported in Li-SICs26,27,43,49. In addition, this the migration of a single Li+/Na+ in a fixed bcc, fcc, and hcp anion
splitting of a large local site volume into multiple equivalent sites of sublattice of O2−, S2−, and Cl− with no other cations as in ref. 21. The
mobile-ions within a close distance causes the feature of enlarged Li+ anions were fixed, and the background charge was set to maintain the
sites proposed by He et al.34, which promotes the frustration and dis- correct valence states of mobile-ion and anions (i.e. Li+, Na+, O2−, S2−,
ordering of the overall Li+ sublattice43,50. In this study, we find in the Na- and Cl−). The supercell models (54 atoms for bcc, 32 atoms for fcc, 36
SICs that site distortion has another beneficial effect, that is, enlarging atoms for hcp) and Γ-centered 2 × 2 × 2 k-point grids were used. Static
the bottleneck size, which is critical for the diffusion of large-radius relaxation of mobile ion (i.e. Li+ and Na+) at initial and final sites within
Na+. For example, in NASCION, the Na+ site is in the six-coordinated the fixed anion sublattice used an energy convergence criterion of 10−5
sites of distorted antiprism that form a face-sharing O-triangle bot- eV and a force convergence criterion of 10−2 eV/Å. A total of seven
tleneck with a large bottleneck size of pr = 1.03 Å. The triangle bottle- images interpolated between initial and end structures were used for
neck size is enlarged by the distortion of Oct sites, in contrast to the the NEB calculations. In the NEB calculations, the anions were fixed,
non-distorted Oct-sites in the hcp anion sublattice (e.g., in Na2ZrCl6) and the energy and force convergence criterion remained the same as
which has a small bottleneck size (pr = 0.73 Å). Therefore, the the static relaxation of initial and final structures.

Nature Communications | (2023)14:7615 6

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Article https://doi.org/10.1038/s41467-023-43436-3

The lattice parameters of the bcc, fcc, and hcp anion model were other via face sharing within the cut-off distance to form percolation.
set to have the same site volume as representative Li-SICs (i.e., O3-type The cut-off distance was set to 3.1 Å for oxides and 3.5 Å for sulfides
LiCoO2, Li2ZrCl6, Li10GeP2S12) and Na-SICs (i.e., O3-type NaCoO2, and chlorides.
Na2ZrCl6, Na10SnP2S12) as shown in Fig. 2a–c. The fixed anion sub- Step 4. Unique structures. We grouped the compounds with the
lattices extracted from real materials of P2- and O3-type NaCoO2 are same crystal structural framework regardless of the cation species
used in Fig. 3a. To consider the effects of changing volumes on Li+/Na+ using the structure matching algorithm in Pymatgen55. In the candidate
migration, the models with a varying range of lattice parameters and list (Supplementary Tables 5, 7, and 9), one compound was evaluated
lattice volume following the volume distribution of the Li/Na-con- to represent a structural framework.
taining compounds in the ICSD (Supplementary Fig. 6) were also Step 5. Practical consideration. We excluded the compounds
conducted (Fig. 2d–f and Supplementary Figs. 7 and 8). containing Au, U, (CO3)2-, (SO4)2-. We excluded all known Na-SICs and
cathode materials, and the Na-Al-O ternary systems. For AIMD simu-
Topological analysis of crystal structure frameworks lations, we excluded the compounds with large supercells containing >
The topological analysis was performed to identify percolation radius, 300 atoms.
Na sites, site coordination, and diffusion network using in Zeo++54 as in Step 6. AIMD screening. For each crystal structural framework
the previous study43. The topological analysis was performed to the identified in the candidate list, the representative compound was
crystal structural framework by removing all Li/Na ions from the selected to evaluate Na+ diffusion. Aliovalent substitution was per-
crystal structures. The analyses were performed using the crystal ionic formed to change Na content in the compound. For each framework,
radii for each ion species in Pymatgen55. For sites with mixed partial Na content was changed to have the ratio of the number of Na ions
occupancy of multiple cations, the smallest ionic radius of these over the total number of identified Na sites to the target range of
cations was used. The percolation radius was calculated as the max- 0.3–0.7. The energy above the hull for doped composition was cal-
imum radius of a sphere that can percolate the crystal structure across culated, and those with good stability of the energy above the hull
at least one direction. <100 meV/atom were further evaluated for Na+ diffusion.
The Na sites in the crystal structural framework were identified as AIMD simulations were first performed at two temperatures,
follows. The structural framework was decomposed by the 900 K and 1150 K for oxides and sulfides and 700 K and 900 K for
Voronoi–Dirichlet partition algorithm using the ions as the centers of chlorides, as an initial screening following Ref. 43 The materials that
the polyhedrons as implemented in Zeo++54. A Voronoi node was the melted during AIMD simulations were excluded. For materials with
vertices shared by the polyhedrons and corresponded to the center of extrapolated Na+ conductivity >0.1 mS/cm at 300 K were further stu-
a local void, which may be possible mobile-ion sites. These Voronoi died by AIMD simulations at more temperatures. The ionic con-
nodes were further screened by the chemical environments suitable ductivity was evaluated according to the Arrhenius relation,
for Na+ occupancy using the criteria based on the coulombic repulsion

and bond valence (BV). Any nodes close to other non-Na cations were
E a
σT = σ 0 exp ð1Þ
excluded, with the cutoff distance set to 2.6 Å for oxides/sulfides and kBT
2.8 Å for chlorides. The BV was calculated for each Voronoi node, and
those with BV in the range of 0.3–1.5 were kept. The distance cut-off where σ is conductivity, T is temperature, Ea is the activation energy, σ0
and BV range were determined by analyzing Na-containing oxide, is the pre-exponential factor, and kB is the Boltzmann constant
sulfide, and chlorides (Supplementary Figs. 10 and 11). Finally, the Na
nodes were grouped into a site if their distance was less than 1.6 Å. Ab initio molecular dynamics simulation
The site coordination number for a given possible Na site was We performed AIMD simulations to study ionic diffusion in supercell
calculated as the number of neighboring anions (i.e. O2−, S2−, and Cl−) models with lattice parameters near or larger than 10 Å. Non-spin mode,
using Voronoi decomposition with solid angle weights as implemented a single Γ-centered k-point, and a time step of 2 fs were used. In each
in CrystalNN algorithm56. The site volumes were calculated by con- simulation, the initial temperature was set to 100 K and then the struc-
structing a convex hull of identified coordinating anions (Supple- tures were heated to the target temperatures at a constant rate by
mentary Fig. 5). velocity scaling during a period of 2 ps. All simulations adopted the NVT
ensemble with Nosé-Hoover thermostat57. The diffusivity D was calcu-
High-throughput screening of Na-ion conductors lated as the mean square displacement (MSD) over the time interval Δt,
Step 1. Basic Materials Check. We firstly filtered all the Na-containing
oxide, sulfide and chloride compounds in the ICSD (2017 version), and 1 X N
D= h½r ðt + 4t Þ
ri ðt Þ2 it ð2Þ
excluded the compounds with any of following attributes: binary 2Nd4t i = 1 i
compounds; containing more than one anion species; co-occupancy of
Na with other elements; containing anions with disordering or partial where d is the dimension of the diffusion, N is the total number of
occupancy; containing water molecules; having elements with no diffusion ions, ri(t) is the displacement of the i-th ion at time t. The
valence or abnormal valences that have no ionic radii information in ionic conductivity was calculated according to the Nernst-Einstein
the defaulted table of pymatgen55; the compounds are not charge relationship,
neutral; containing more than 300 atoms. After this step of basic
materials checks, there were a total of 2673 oxides, 153 sulfides, and 80 nq2
σ= D ð3Þ
chlorides for further screening. kBT
Step 2. Percolation radius. Considering the bottleneck size of
diffusion channels in the crystal structure, we identify the structures where n is the mobile ions volume density and q is the ionic charge.
with large percolation radii of pr > 0.9 Å for oxides and pr > 0.85 Å for Given that the ion hopping is a stochastic process, the statistical
sulfides and chlorides. The cut-off values were identified based on the deviations of the conductivity were evaluated according to the scheme
key parameters of the crystal frameworks of known Na-ion conductors established in our previous work58. The total time duration of AIMD
(Supplementary Table 2). simulations was within the range of 40–400 ps until the total mean-
Step 3. Face-sharing high-CN sites. For a structure to pass the square-displacement reach over 1000–2000 Å2 and the ionic diffusiv-
screening, the diffusion network should only consist of high-CN Na ity converged within a relative standard deviation between 20% and
sites with CN ≥ 5. These high-CN Na sites should also connect with each 40% for most data points.

Nature Communications | (2023)14:7615 7

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Article https://doi.org/10.1038/s41467-023-43436-3

Experimental methods 7. Zhao, Q., Stalin, S., Zhao, C. Z. & Archer, L. A. Designing solid-state
Synthesis. All preparation processes and sample treatments were electrolytes for safe, energy-dense batteries. Nat. Rev. Mater. 5,
carried out in an Ar-filled glovebox (O2 < 1 ppm, H2O < 1 ppm). The 229–252 (2020).
family of Na3xM2-xCl6-contained halide conductors (M = La, Ce, Nd, Sm) 8. Zhao, C. et al. Solid-state sodium batteries. Adv. Energy Mater. 8,
were synthesized by ball-milling the starting materials of LaCl3 (Sigma 1703012–1703012 (2018).
Aldrich, 99.9%), SmCl3 (Sigma Aldrich, 99.9%), NdCl3 (Sigma Aldrich, 9. Zhou, W., Li, Y., Xin, S. & Goodenough, J. B. Rechargeable sodium
99.8%), CeCl3 (Sigma Aldrich, 99.9%), TaCl5 (Sigma Aldrich, 99.9%) and all-solid-state battery. ACS Central Sci. 3, 52–57 (2017).
NaCl (Sigma Aldrich, reagent grade) according to the stoichiometric 10. Zhang, Z. et al. Na11Sn2PS12: a new solid state sodium superionic
ratios. For ball-milling synthesis, the mixture of precursors was sealed conductor. Energy Environ. Sci. 11, 87–93 (2018).
in a zirconia jar (100 mL) under vacuum and was mechanically milled 11. Ellis, B. L. & Nazar, L. F. Sodium and sodium-ion energy storage
using a plenary high-energy ball-milling machine (PM200, RETSCH) batteries. Curr. Opin. Solid State Mater. Sci. 16, 168–177 (2012).
with zirconia balls (ϕ = 5, 7, and 10 mm, balls/precursors = 40:1 w/w) 12. Wei, S. et al. A stable room-temperature sodium–sulfur battery. Nat.
for 60 cycles. The ball-milling process included 10-min milling and Commun. 7, 1–10 (2016).
5-min resting for each cycle. The ball-milling speed was 500 rpm. The 13. Hueso, K. B., Armand, M. & Rojo, T. High temperature sodium bat-
as-prepared samples were collected in the glovebox for ionic con- teries: status, challenges and future trends. Energy Environ. Sci. 6,
ductivity measurements. 734–749 (2013).
14. Yao, Y.-F. Y. & Kummer, J. T. Ion exchange properties of and rates of
Structure characterization ionic diffusion in beta-alumina. J. Inorg. Nucl. Chem. 29, 2453–2475
Lab-based X-ray diffraction (XRD) measurements were performed (1967).
on Bruker AXS D8 Advance with Cu Kα radiation (λ = 1.5406 Å). 15. Goodenough, J. B., Hong, H.-P. & Kafalas, J. A. Fast Na+-ion transport
Kapton tape was used to cover the sample holder to prevent air in skeleton structures. Mater. Res. Bull. 11, 203–220 (1976).
exposure. Synchrotron-based powder diffraction patterns were 16. Murugan, R., Thangadurai, V. & Weppner, W. Fast lithium ion con-
collected using the Brockhouse High Energy Wiggler beamline at duction in garnet‐type Li7La3Zr2O12. Angew. Chem. Int. Ed. 46,
the Canadian Light Source (CLS) with a wavelength of 0.3497 7778–7781 (2007).
and 0.2077 Å. The samples were loaded into 0.8 mm inner 17. Yamane, H. et al. Crystal structure of a superionic conductor,
diameter polyimide capillaries and sealed with epoxy in an Ar-filled Li7P3S11. Solid State Ion. 178, 1163–1167 (2007).
glove box. The X-ray diffraction Rietveld refinement and pair dis- 18. Deiseroth, H. J. et al. Li6PS5X: a class of crystalline Li‐rich solids with
tributed function fittings were conducted by GSAS-259 and PDFgui an unusually high Li+ mobility. Angew. Chem. Int. Ed. 47, 755–758
software60. (2008).
19. Liang, J. et al. Site-occupation-tuned superionic LixScCl3+x halide
EIS measurements of ionic conductivity solid electrolytes for all-solid-state batteries. J. Am. Chem. Soc. 142,
The temperature-dependent ionic conductivities of prepared solid 7012–7022 (2020).
electrolytes were obtained via the EIS measurements of model cells on 20. Li, X. et al. Air-stable Li3InCl6 electrolyte with high voltage com-
a multichannel potentiostat 3/Z (German VMP3). The temperature patibility for all-solid-state batteries. Energy Environ. Sci. 12,
range was between −25 and 55 °C. The applied frequency range was 2665–2671 (2019).
1 Hz to 7 MHz and the voltage amplitude was 20 mV. The test cell was 21. Asano, T. et al. Solid halide electrolytes with high lithium-ion con-
fabricated as follows: 100–120 mg of the electrolytes were pressed ductivity for application in 4 V class bulk-type all-solid-state bat-
(~300 MPa) into a pellet (diameter: 1 cm, thickness ~0.5–0.7 mm). teries. Adv. Mater. 30, 1803075–1803075 (2018).
About 5 mg of acetylene black carbon was then spread onto both sides 22. Morgan, B. J. Mechanistic origin of superionic lithium diffusion in
of the pellet and pressed with ~150 MPa. anion-disordered Li6PS5 X argyrodites. Chem. Mater. 33, 2004–2018
(2021).
Data availability 23. Deng, Y. et al. Structural and mechanistic insights into fast lithium-
All data are provided in the paper and its Supplementary Information. ion conduction in Li4SiO4–Li3PO4 solid electrolytes. J. Am. Chem.
Additional information is available from the corresponding authors Soc. 137, 9136–9145 (2015).
upon request. Source data are provided with this paper. 24. Catlow, C. R. A. Atomistic mechanisms of ionic transport in fast-ion
conductors. J. Chem. Soc. Faraday Trans. 86, 1167–1176 (1990).
Code availability 25. Wang, Y. et al. Design principles for solid-state lithium superionic
The code used in this study is available from the corresponding author conductors. Nat. Mater. 14, 1026–1031 (2015).
upon request. 26. He, X., Zhu, Y. & Mo, Y. Origin of fast ion diffusion in super-ionic
conductors. Nat. Commun. 8, 15893–15893 (2017).
References 27. Jun, K. et al. Lithium superionic conductors with corner-sharing
1. Kamaya, N. et al. A lithium superionic conductor. Nat. Mater. 10, frameworks. Nat. Mater. 21, 924–931 (2022).
682–686 (2011). 28. Xiong, S. et al. Computation‐guided design of LiTaSiO5, a new
2. Zhou, Y. et al. Strongly correlated perovskite fuel cells. Nature 534, lithium ionic conductor with sphene structure. Adv. Energy Mater.
231–234 (2016). 9, 1803821–1803821 (2019).
3. Janek, J. & Zeier, W. G. A solid future for battery development. Nat. 29. Richards, W. D., Wang, Y., Miara, L. J., Kim, J. C. & Ceder, G. Design
Energy 1, 16141–16144 (2016). of Li1+2xZn1−xPS4, a new lithium ion conductor. Energy Environ. Sci.
4. Wachsman, E. D. & Lee, K. T. Lowering the temperature of solid 9, 3272–3278 (2016).
oxide fuel cells. Science 334, 935–939 (2011). 30. Richards, W. D. et al. Design and synthesis of the superionic con-
5. Famprikis, T., Canepa, P., Dawson, J. A., Islam, M. S. & Masquelier, C. ductor Na10SnP2S12. Nat. Commun. 7, 11009–11009 (2016).
Fundamentals of inorganic solid-state electrolytes for batteries. 31. Kwak, H. et al. Na2ZrCl6 enabling highly stable 3 V all-solid-state
Nat. Mater. 18, 1278–1291 (2019). Na-ion batteries. Energy Storage Mater. 37, 47–54 (2021).
6. Bachman, J. C. et al. Inorganic solid-state electrolytes for lithium 32. Dixit, M. B. et al. Polymorphism of garnet solid electrolytes and its
batteries: mechanisms and properties governing ion conduction. implications for grain-level chemo-mechanics. Nat. Mater. 21,
Chem. Rev. 116, 140–162 (2016). 1298–1305 (2022).

Nature Communications | (2023)14:7615 8

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Article https://doi.org/10.1038/s41467-023-43436-3

33. Tan, D. H. S. et al. Carbon-free high-loading silicon anodes enabled 57. Nosé, S. A unified formulation of the constant temperature mole-
by sulfide solid electrolytes. Science 373, 1494–1499 (2021). cular dynamics methods. J. Chem. Phys. 81, 511–519 (1984).
34. Hayashi, A. et al. A sodium-ion sulfide solid electrolyte with 58. He, X., Zhu, Y., Epstein, A. & Mo, Y. Statistical variances of diffusional
unprecedented conductivity at room temperature. Nat. Commun. properties from ab initio molecular dynamics simulations. npj
10, 5266–5266 (2019). Comput. Mater. 4, 18–18 (2018).
35. Shannon, R. D. Revised effective ionic radii and systematic studies of 59. Toby, B. H. & Von Dreele, R. B. GSAS-II: the genesis of a modern
interatomic distances in halides and chalcogenides. Acta Crystallogr. open-source all purpose crystallography software package. J. Appl.
Sect. A Cryst. Phys. Diffr. Theor. Gen. Crystallogr. 32, 751–767 (1976). Crystallogr. 46, 544–549 (2013).
36. Wang, S. et al. Lithium chlorides and bromides as promising solid- 60. Farrow, C. et al. PDFfit2 and PDFgui: computer programs for studying
state chemistries for fast ion conductors with good electrochemical nanostructure in crystals. J. Phys. Condens. Matter 19, 335219 (2007).
stability. Angew. Chem. Int. Ed. 58, 8039–8043 (2019).
37. Adams, S. & Rao, R. P. Structural requirements for fast lithium ion Acknowledgements
migration in Li10GeP 2S12. J. Mater. Chem. 22, 7687–7691 (2012). Y.M. acknowledge the funding support from National Science Founda-
38. Zhang, Z. & Nazar, L. F. Exploiting the paddle-wheel mechanism for tion Award# 2118838 and the computational facilities from the University
the design of fast ion conductors. Nat. Rev. Mater. 7, 389–405 (2022). of Maryland supercomputing resources. X.S. thanks the funding support
39. Ren, Q. et al. Extreme phonon anharmonicity underpins superionic from the Natural Sciences and Engineering Research Council of Canada
diffusion and ultralow thermal conductivity in argyrodite Ag8SnSe6. (NSERC), Canada Research Chair Program (CRC), and University of
Nat. Mater. https://doi.org/10.1038/s41563-023-01560-x (2023) Western Ontario. X.S. also appreciates the help of the BXDS beamline at
40. Muy, S., Schlem, R., Shao-Horn, Y. & Zeier, W. G. Phonon–ion Canadian Light Source.
interactions: designing ion mobility based on lattice dynamics. Adv.
Energy Mater. 11, 2002787 (2021). Author contributions
41. Mo, Y., Ong, S. P. & Ceder, G. Insights into diffusion mechanisms in Y.M. oversaw the overall project. Y.M. and S.W. conceived the project.
P2 layered oxide materials by first-principles calculations. Chem. S.W. designed and performed the computation and analyses. J.F., J.L., and
Mater. 26, 5208–5214 (2014). S.D. fabricated the samples of the materials and carried out the experi-
42. Hellenbrandt, M. The inorganic crystal structure database (ICSD)— ments, characterizations, and data analyses, supervised by X.S. and T.S.
present and future. Crystallogr. Rev. 10, 17–22 (2004). S.W. and Y.M. prepared the manuscript with the help of all authors. All
43. He, X. et al. Crystal structural framework of lithium super‐ionic authors contributed to the discussions and revisions of the manuscript.
conductors. Adv. Energy Mater. 9, 1902078–1902078 (2019).
44. Armand, M. B., Chabagno, J. M. & Duclot, M. J. in Fast Ion Transport Competing interests
in Solids (eds Vashishta, P., Mundy, J. N. & Shenoy, G. K.) 131 (North The authors declare no competing interests.
Holland, 1979).
45. Michiue, Y., Sato, A. & Watanabe, M. Low-temperature phase of Additional information
sodium priderite, NaxCrxTi8–xO16, with a monoclinic hollandite Supplementary information The online version contains
structure. J. Solid State Chem. 145, 182–185 (1999). supplementary material available at
46. Dawson, J. A., Canepa, P., Famprikis, T., Masquelier, C. & Islam, M. S. https://doi.org/10.1038/s41467-023-43436-3.
Atomic-scale influence of grain boundaries on Li-ion conduction in
solid electrolytes for all-solid-state batteries. J. Am. Chem. Soc. 140, Correspondence and requests for materials should be addressed to
362–368 (2018). Xueliang Sun or Yifei Mo.
47. Dawson, J. A. et al. Toward understanding the different influences
of grain boundaries on ion transport in sulfide and oxide solid Peer review information Nature Communications thanks the anon-
electrolytes. Chem. Mater. 31, 5296–5304 (2019). ymous reviewers for their contribution to the peer review of this work. A
48. Zachariasen, W. H. The UCl3 type of crystal structure. J. Chem. Phys. peer review file is available.
16, 254–254 (1948).
49. Di Stefano, D. et al. Superionic diffusion through frustrated energy Reprints and permissions information is available at
landscape. Chem 5, 2450–2460 (2019). http://www.nature.com/reprints
50. Zhang, Y. et al. Unsupervised discovery of solid-state lithium ion
conductors. Nat. Commun. 10, 1–7 (2019). Publisher’s note Springer Nature remains neutral with regard to jur-
51. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio isdictional claims in published maps and institutional affiliations.
total-energy calculations using a plane-wave basis set. Phys. Rev. B
54, 11169–11169 (1996). Open Access This article is licensed under a Creative Commons
52. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50, Attribution 4.0 International License, which permits use, sharing,
17953–17953 (1994). adaptation, distribution and reproduction in any medium or format, as
53. Jain, A. et al. Commentary: The Materials Project: a materials gen- long as you give appropriate credit to the original author(s) and the
ome approach to accelerating materials innovation. Apl. Mater. 1, source, provide a link to the Creative Commons licence, and indicate if
11002–11002 (2013). changes were made. The images or other third party material in this
54. Willems, T. F., Rycroft, C. H., Kazi, M., Meza, J. C. & Haranczyk, M. article are included in the article’s Creative Commons licence, unless
Algorithms and tools for high-throughput geometry-based analysis indicated otherwise in a credit line to the material. If material is not
of crystalline porous materials. Microporous Mesoporous Mater. included in the article’s Creative Commons licence and your intended
149, 134–141 (2012). use is not permitted by statutory regulation or exceeds the permitted
55. Ong, S. P. et al. Python Materials Genomics (pymatgen): a robust, use, you will need to obtain permission directly from the copyright
open-source python library for materials analysis. Comput. Mater. holder. To view a copy of this licence, visit http://creativecommons.org/
Sci. 68, 314–319 (2013). licenses/by/4.0/.
56. Pan, H. et al. Benchmarking coordination number prediction algo-
rithms on inorganic crystal structures. Inorg. Chem. 60, 1590–1603 © The Author(s) 2023
(2021).

Nature Communications | (2023)14:7615 9

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

[email protected]