You are on page 1of 25

Review

www.advenergymat.de

Li10GeP2S12-Type Superionic Conductors: Synthesis, Structure,


and Ionic Transportation
Yuki Kato, Satoshi Hori, and Ryoji Kanno*

Since it has been a major challenge to


Ever since the first report on Li10GeP2S12 (LGPS) in 2011, its unique structure elucidate the cause of such unusual phe-
and exceptionally high lithium conductivity (>1 × 10−2 S cm−1) have attracted nomena in these solids, research efforts
extensive interest, especially for applications in solid-state ionics and bat- have focused on understanding the origin
of superionic conduction. Alternatively,
teries. Herein, studies of LGPS and its modifications are reviewed with a
there have also been tremendous efforts to
focus on the synthesis, structure, and ionic transportation of LGPS. For mate- synthesize new materials showing supe-
rial synthesis, the relationships between LGPS and its precursor compounds rionic conductivity, in order to examine
such as Li3PS4 and Li4GeS4 are discussed. A technique for single-crystal the maximum diffusivity in solids. Fur-
growth and a family of LGPS-type materials that are chemically or structurally thermore, devices utilizing fast diffusion
in solids are being developed for practical
related to LGPS are then described. The crystal structure of LGPS is analyzed
applications in industry. Research and
from the viewpoints of tetrahedral framework units, anion sublattice, and development have thus advanced the field
Li distributions; furthermore, the conduction mechanism is qualitatively of solid-state ionics.[1–3]
analyzed. Subsequently, ionic transportation in LGPS is studied quantitatively. To enable fast diffusion in solids,
The origin of the high conductivity is discussed in terms of the activation the compound in either crystalline or
energy, diffusion coefficient, and its related parameters; and these factors amorphous form must possess not only
migration pathways inside itself but
are compared to those of other non-LGPS-type conductors. Then, the battery
also a structural framework containing
applications are briefly summarized to indicate the potential merits of using such pathways. In addition, the interac-
LGPS-type solid electrolytes with high lithium conductivity. Any remaining tions between mobile ions and the lattice
issues and possible research directions that have emerged from the afore- should be small enough to allow the ions
mentioned studies are finally highlighted. to migrate. Thus, monovalent ions with
relatively small ionic radii are generally
considered favorable for migration.
Among all potential mobile species,
1. Introduction relatively large number of conductors have been reported for
H+, Li+, Ag+, Cu+, F−, and O2− ions. Among them, Ag+- or
Superionic conductors are a group of solids composed of rigid Cu+-related chemical systems have the highest conductivities;
frameworks, in which the ions can migrate in a manner similar RbAg4I5[4] and its Cu+-modification (Rb4Cl16I7Cl13)[5] show ionic
to that in liquid electrolytes. Faraday observed superionic conduc- conductivities greater than 2 × 10−1 S cm−1 at room tempera-
tion for the first time in lead fluoride, wherein some constituting ture (RT). Nevertheless, Li+ conductors are more suitable for
ions move as in liquids, although the entire solid is composed of energy device applications since lithium batteries would offer
tight chemical bonding. After his discovery, similar unique phe- high energy density. Therefore, Li+ ionic conductors, especially
nomena were observed in various other materials, such as α-AgI, those enabling the cooperative motion of Li ions with activation
stabilized zirconia, and β-alumina, which use silver, oxygen, and energies as low as 20 kJ mol−1, have been investigated for a long
sodium, as the respective electrical charge carriers. time for use as electrolytes in all-solid-state lithium batteries.
Figure 1 shows the chronological development of Li+ conduc-
tors. Prior to the 1990s, there were reports on β-Li alumina,[6]
Dr. Y. Kato LISICON (LIthium SuperIonic CONductor)-type oxides,[7]
Toyota Motor Corporation
1200 Mishuku, Susono, Shizuoka 410-1107, Japan
Li3N,[8] polyethylene oxide (PEO)-based polymer,[9] and oxide[10]
Dr. S. Hori, Prof. R. Kanno
or sulfide glasses.[11] Although the subsequent related studies
Institute of Innovative Research (IIR) focused on applying these materials to solid-state cells, no one
All-Solid-State Battery Unit has reported the successful operation of solid-state batteries
Tokyo Institute of Technology using superionic conductors, and no materials has exhibited
4259 Nagatsuta, Midori, Yokohama 226-8502, Japan both a high conductivity and a wide electrochemical window.
E-mail: kanno@echem.titech.ac.jp
Consequently, solid-state batteries were not widely acknowl-
The ORCID identification number(s) for the author(s) of this article
can be found under https://doi.org/10.1002/aenm.202002153.
edged as a prospective battery configuration.
Later studies led to the discovery of materials with Li con-
DOI: 10.1002/aenm.202002153 ductivity over 1 × 10−2 S cm−1 at RT. In 2011, Li10GeP2S12

Adv. Energy Mater. 2020, 2002153 2002153 (1 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

2. Synthesis
2.1. Overview of Material Diversity in LGPS-Type Phases

Figure 2a summarizes the elements that are potential constitu-


ents of LGPS-type crystals together with prospective candidates
worth considering in the future. Figure 2b depicts all the chem-
ical systems in which LGPS-type materials have been reported.
For tetrahedral units that comprise the framework of the entire
crystal, there are two crystallographic positions—M1(2b) and
M2(4d)—in the tetrahedral center, each of which can be occu-
pied by more than one cation (see Figure 8). Sulfur atoms are
typically positioned at the apexes of the tetrahedral unit, and
numerous studies have demonstrated that the sulfur atoms can
be replaced by oxygen, selenium, and halogen atoms. In terms
of diffusive species, previous studies have reported Na-based
derivatives in which Li ions are replaced by Na ions.
There are several possible compositions for the LGPS-type
phases. Elements other than transition metals have been exten-
sively investigated as constituents of the LGPS framework.
The transition metals are not suitable as electrolytes due to
their redox activity, and therefore they are overshadowed in
Figure 2b. From a crystal chemistry perspective, whether transi-
tion metals are compatible with the LGPS-type structure is an
interesting research topic. There are several elements that can
possibly occupy the tetrahedral position surrounded by S atoms,
despite the absence of reported LGPS structures with such ele-
ments. In addition, no elements, other than phosphorus have
Figure 1. Chronological history of developments in Li ionic conductors. been reported to occupy the 2b site. Previous studies have not
Conductivity at RT and year of publication are indicated for representative
material groups; red: sulfide; yellow: oxide; grey: (gel) polymer composite;
documented the reason(s) why only P atom is used or whether
blue: halide, nitride, and hydride. Subcategories for each group are rep- other elements may be used. Hence, the materials with LGPS-
resented by distinctive symbols. Presented conductivity values were col- type structure requires future investigations.
lected from published works given in Note S1, Supporting Information. As described above, it is possible to modify the framework
tetrahedron to some extent by various cations and anions.
(LGPS)[12] was reported as the first lithium conductor with a Thus, researchers have been able to design a variety of mate-
lithium conductivity comparable to those of organic electrolytes rials with the LGPS-type structure, as schematically shown in
(≈10−2 S cm−1). In 2016, its modification Li9.54Si1.74P1.44S11.7Cl0.3 Figure 2c. In the following section, we briefly review the his-
was reported to show an even higher conductivity.[13] This torical background on the discovery of LGPS, and describe the
exceptionally high conductivity is ascribed to the unique crystal following interrelated studies—the phase diagram study of
structure that differs entirely from any other structural types Li3PS4–Li4GeS4 pseudo-binary systems, the synthesis of phase-
reported prior for LGPS. These features have attracted consid- pure LGPS powder, and single-crystal growth. We also describe
erable research interest on solid-state ionics and device applica- the systematic synthesis research performed for several chem-
tions, particularly on all-solid-state batteries. ical systems.
This specialized review on LGPS focuses on the synthetic
chemistry and conduction mechanisms based on the crystal
structure, rather than on the electrochemical applications, 2.2. Historical Background
because excellent reviews of their applications have already
been published in recent years.[14–18] Those reviews detailed The discovery of LGPS dates back to the development of thio-
reactivity at the interface in solid-state cells, composite elec- LISICON by Kanno et al. in 2000.[19] In the thio-LSICON group,
trode systems, and hardness of solid electrolytes. Li3.25Ge0.25P0.75S4 was reported as the first crystalline sulfide
The present review begins with the historical background on whose lithium conductivity at RT exceeded 1 × 10−3 S cm−1.
the discovery of LGPS, followed by various synthesis chemistries These researchers focused on a group of materials termed
depending on the composition and process. The crystal structure LISICONs,[7] which were one of the most prominent dis-
of LGPS is described in detail. Finally, the conduction properties coveries in the 1970s. The design guidelines were as follows:
are reviewed and the conduction mechanism is discussed. The 1) the mobile ions should have a suitable size for the conduction
studies reviewed herein have demonstrated advanced synthetic pathways in the lattice, 2) there should be disorder in a mobile
chemistry, revealed the phase–property relationships, applied ion sublattice, and 3) highly polarizable mobile ions and
crystal chemistry to the static behavior of ions, and determined anion sublattices are preferable. LISICONs are based on the
the ion dynamics using theoretical calculation or experiments. γ-Li3PO4-type crystal structure wherein various elements can be

Adv. Energy Mater. 2020, 2002153 2002153 (2 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 2. Various materials categorized in the LGPS family. a) Constituent elements reported thus far and the prospective candidates worth consid-
ering. The latter group of elements was selected because the sulfides containing tetrahedral units with the element and four sulfur atoms are reported
(e.g., in Ag3PS4, Ag atoms form AgS4 with four neighboring sulfur atoms). b) Materials map of the electrolyte with the LGPS-type structure. From its
history of development and similarity of composition, thio-LISICONs may be considered as the parent materials of LGPS. Derivatives are categorized
into two large groups: cation-substituted materials and materials with an anion-substituted composition wherein sulfur atoms are assumed to be
replaced by doped anions, such as halogens. c) Concept of material design for the LGPS phase from thio-LISICONs. Scaffolds composed of tetrahedral
units change entirely by the addition of different-sized tetrahedral units with various elements.

Adv. Energy Mater. 2020, 2002153 2002153 (3 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Meanwhile, a thio-LISICON (Li3.25Ge0.25P0.75S4) was success-


fully applied for all-solid-state cells[26,27] as an electrolyte after
vibration milling to reduce the particle size. After the milling
process, its conductivity remained unchanged, but some dif-
fraction peaks disappeared in its XRD pattern. Since this obser-
vation did not follow the general trend for solid electrolytes,
Kanno et al. recognized that this was an anomalous pheno­
menon found in the Li4GeS4–Li3PS4 system; therefore, the
phase diagram study was necessary. During the analysis of
these phase relationships, LGPS was finally discovered.

2.3. Synthesis of L10GeP2S12 and Solid Solutions


in Li4GeS4–Li3PS4

In the first report in 2011, the original Li10GeP2S12 was synthe-


sized by mechanical mixing of lithium sulfide, phosphorous
sulfide, and germanium sulfide, followed by heat treatment
Figure 3. Comparison of XRD patterns for a) x = 0.75 and b) x = 0.55 in of the mixture.[12] Based on this method, which is one of the
the Li4−xGe4−xPxS4.[22,57] The former was synthesized at 700 °C and reported standard protocols for laboratory-level research, numerous syn-
as thio-LISICON, whereas the latter was LGPS synthesized at 550 °C and thesis methods have been developed, as listed in Table 1.[12,13,28–47]
all of the peaks were indexed by P42/nmc (space group No. 137). A subsequent study reported that synthesis temperatures
more than 400 °C are required for the crystallization of the
­accommodated by forming solid solutions with lithium vacan- LGPS phase from precursors obtained by mechanochemical
cies and inter-lattice lithium. In 2000, Kanno et al. performed treatment.[48] In-depth temperature-composition studies have
a systematic and comprehensive synthesis of the Li2S–GeS2– been performed to obtain high-purity LGPS phases and com-
Ga2S3 pseudo-ternary system. They revealed that the phases prehend the Li4GeS4–Li3PS4 pseudo-binary system.[49]
within a specific composition range showed high conductivity, Figure 4 shows the phase diagram for this system proposed
and these conductive phases had a crystal structure similar to by Hori et al.[49] The highly conductive thio-LISICON Region II
the LISICON-type oxides. This study was the first example in was found to be a mixture of the orthorhombic thio-LISICON
which the oxygen sublattice in LISICONs was replaced with phase, that is, β-Li3PS4 with partial substitution by Ge, and the
sulfur. Such replacement resulted in a significant increase LGPS phase with a tetragonal lattice (Figure 3b compares the two
in the lithium conductivity due to the larger ionic radius and XRD patterns). In the pseudo-binary system, Li4GeS4 (β′ phase)
polarizability of sulfur. This novel material group was named and Li3PS4 (β phase) exist as the main crystal phases with
thio-LISICON. γ-Li3PO4-type crystal structures and Li10GeP2S12 (G phase). The
The search for new ionic conductors was based on the G phase is crystallized as a single phase for the Li10GeP2S12
modification of Li4SiS4 and Li4GeS4.[20,21] For example, composition as one end member, thus forming solid solutions
Li4−xSi1−xPxS4,[21] Li4−xGe1−xPxS4,[22] and Li4−2xZnxGexS4[19] in the ≈45–67 mol% Li3PS4 composition range. The G phase is
were used as vacancy-doped systems (Li+ + M4+ ↔ M″5+ or determined to be stable up to ≈600 °C, with a slight increase in
2Li+ ↔ Zn2+), whereas Li4+xSi1−xAlxS4[21] and Li4 +xGe1−xGaxS4[19] the formation compositional range in terms of Ge as the tem-
were used as interstitial ion-doped systems perature increases. In a certain composition range, there is a
(M4+ ↔ Li+ + M′3+). In 2004, thio-LISICON related materials two-phase mixture of G and the liquid phase between 600 and
were also discovered in Li2S–P2S5 binary systems.[23] Of all the 650 °C. Above 650 °C, a two-phase region of β′ and the liquid
chemical systems discussed above, the Li4−xGe1−xPxS4 system phase exists. Further increasing the temperature results in
(x = 0.75; Li3.25Ge0.25P0.75S4) exhibited a particularly high ionic complete melting.
conductivity (2.2 × 10−3 S cm−1),[22] which was as high as those The phase diagram determined the synthesis method of
of the sulfide glasses with the highest-class conductivities at LGPS. A high-purity single-phase sample was obtained for
that time.[24,25] A series of studies on thio-LISICONs have indi- slightly Ge-rich compositions with respect to the original one,
cated that crystalline sulfides could possibly have a high Li con- using the conventional solid-state reaction similar to that in the
ductivity comparable with the glassy counterparts. original report. Figure 3b shows that all peaks in the XRD pat-
Figure 3a shows a representative X-ray diffraction (XRD) pat- tern of a representative pure sample are indexed by P42/nmc
tern reported for the Li4−xGe1−xPxS4 system. Previous studies (space group for LGPS-type structures), with no peaks from
found additional small reflections that cannot be ascribed to the secondary crystalline phases. For such single-phase samples,
original orthorhombic Li4GeS4 structure. Therefore, the com- the transmission electron microscopy (TEM) image shown in
plex XRD patterns for the Li4−xGe1−xPxS4 system were initially Figure 5 indicates that there are no contrasts other than those
refined by assuming a monoclinic superlattice, and were clas- due to differences in the orientation.[50] Furthermore, all diffrac-
sified as Region I (0 ≤ x ≤ 0.6), Region II (0.65 ≤ x ≤ 0.75), and tion spots in Figure 5 are indexed, thus suggesting the absence
Region III (0.8 ≤ x ≤ 1.0).The complex XRD patterns had not of an amorphous phase in the samples identified as phase-
been fully elucidated at that time. pure using XRD. Conversely, secondary phases are frequently

Adv. Energy Mater. 2020, 2002153 2002153 (4 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Table 1. Synthesis protocols of LGPS-related materials in powder or single-crystal form.

Composition Starting materials Pre stepa) Temperature [°C] Time [h] Coolingb)
[12]
Li10GeP2S12 Li2S, P2S5, GeS2 VM 550 8 NC
Li10SnP2S12[28] Li2S, P2S5, Li4SnS4 N/A 600 48 SCc)
Li10GeP2S12[29] Li2S, P, Ge, S BM 420 24 NC
[29]
Li7GePS8 Li2S, P, Ge, S BM 450 (4 h) → 550 °C (1 h) NC
Li11Si2PS12[30] Li2S, P, Si, S BM 550 96 NC
Li10+δSi1+δP2−δS12[31] Li2S, P2S5, SiS2, BM 550 72 NC
Li10+δSn1+δP2−δS12[31] Li2S, P2S5, SnS2 BM 550 72 NC
Li10GeP2S12[32] Li2S, P, Ge, S BM 500 36 NC
Li10.3Al0.3Sn0.7P2S12[33] Li2S, P2S5, Sn, Al, S N/A 550 72 NC
Li10Si0.3Sn0.7P2S12[33] Li2S, P2S5, Si, Sn, S N/A 550 72 NC
Li9.54Si1.74P1.44S11.7Cl0.3[13] Li2S, P2S5, SiS2, LiCl BM 475 8 NC
Li9.6P3S12[13] Li2S, P2S5, P BM 230–260 4 NC
Li10Ge1−xSnxP2S12[34] Li2S, P2S5, GeS2, SnS2 BM 500 20 NC
Li10+δ(SnySi1−y)1+δP2−δS12[35] Li2S, P2S5, SiS2, SnS2 BM 550 24 NC
Li10Ge(P1−xSbx)2S12[36] Li2S, P2S5, Sb2S5, GeS2 BM 475 24 NC
Li9.42Si1.02P2.1S9.96O2.04[37] Li2S, P2S5, SiO2 VM 1000 5 Quench
Li10GeP2S12−xOx[38] Li2S, Li2O, P2S5, GeS2 BM 550 24 NC
Li10SiP2S12−xOx[39] Li2S, P2O5, P2S5, SiO2, SiS2 BM 550 48 NC
0.75Li3PS4-0.25Li3PO4[40] Li2S, P2S5, Li3PO4 N/A 950 1 Quench
Na10SnP2S12[41] Na2S, P2S5, SnS2 BM 700 12 SCd)
[42]
Na11Sn2PS12 Na2S, P2S5, SnS2 N/A 700 5 SCe)
Na11Sn2PS12[43] Na3PS4, Na4SnS4 N/A 600 N/A SCf)
[44]
Li10GeP2S12 Li2S, P, Ge, S BM 750 N/A SC
Li10SnP2S12[28] Li2S, P2S5, Li4SnS4 N/A 600 48 SCg)
Li10GeP2S12[45] Li2S, P, Ge, S N/A 400 (8 h) → 670 °C (1 h) SCh)
[42]
Na11Sn2PS12 Na2S, P2S5, SnS2 N/A 700 2 SCi)
[46] j)
Li10GeP2S12 Li2S, P2S5, GeS2 L 550 24 NC
Li10GeP2S12[47] Li2S, P2S5, GeS2 Lk) 240 2 NC

a)
VM, vibration milling; BM, ball milling; L, liquid phase synthesis; b)SC, slow cooling; NC, natural cooling; c)1 °C h−1; d)0.1 °C min−1; e)Cooling to 400 °C at 0.2 ○C min−1
then cooling to RT at 2 °C min−1; f)10 °C h−1; g)1 °C h−1; h)Cooling to 580 °C at 1 °C h−1 followed by quenching; i)Cooling to 250 °C at 0.075 °C min−1; j)Li2S–GeS2 reaction
in methanol → dry; this precursor was added to THF together with Li2S–P2S5. After drying the THF, the LGPS precursor was obtained; k)LGPS, synthesized via solid-state
reaction, was mixed in hydrazine. The thin-film LGPS was obtained after drying the hydrazine and annealing.

included in numerous previous reports.[13,51,52] In those cases, form a liquid phase due to the slow solid-to-solid transforma-
compounds in the Li2S–P2S5–GeS2 pseudo-ternary system tion. This renders it difficult to grow a single crystal in the melt
appear as impurity phases (Figure 6). Examples of these impu- at the original composition (Li10GeP2S12). However, it is still
rity phases include Li3PS4 and Li4GeS4 (both are the end mem- possible to select small single crystals of LGPS from the multi-
bers of the phase diagram), as well as Li4P2S7 (Li4P2S6), Li7PS6, phase mixture. Kuhn et al. melted Li10GeP2S12 and extracted
and Li2GeS3.[19,20,53–56] Reports have postulated that the samples single crystals from the G- and β-phase mixtures for structural
contained amorphous phases, whose amount influences the analysis.[44] Similarly, Bron et al. succeeded in the struc-
ionic conductivity when synthesis temperatures below 550 °C tural analysis of Li10SnP2S12 single crystals with sizes of
are used.[48] These reports indicate that secondary phase com- 0.25 mm × 0.23 mm × 0.05 mm.[28] Using the self-flux
ponents in the sample are partially responsible for differences method based on the phase diagram, Iwasaki et al. suc-
in the reported conductivity and thermal stability.[32,49,51,57] cessfully synthesized large single crystals (greater than
­Consequently, a more precise synthesis method is needed to 1 mm3) of LGPS (Figure 7).[45] They employed an Li3PS4-
better determine the inherent physical properties. rich (85 mol%) composition, which is significantly dif-
The phase diagram shows that the G phase at a composi- ferent from the original Li10GeP2S12, to ­precipitate an LGPS
tional range of 45–67 mol% Li3PS4 partially melts above 550 °C crystal directly from the melt. Guided by the liquidus
and decomposes incongruently above 650 °C, which indicates line on the phase diagram, LGPS crystals were formed at
that phase separation occurs when the sample cools slowly to 630 °C upon melt cooling, followed by continuous growth at

Adv. Energy Mater. 2020, 2002153 2002153 (5 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 4. Phase diagram for the Li3PS4–Li4GeS4 pseudo-binary system.[49] The horizontal axis of the molar ratio in Li3PS4 corresponds to x in
Li4xGe4−xPxS4 or δ in Li10+δ Ge1+δP1−δS12 where δ = 2 − 3x, which represent the chemical formula for thio-LISICONs or LGPS-type solid solutions, respec-
tively. The notation L represents the liquid phase, whereas α, β, γ, G, and β´ indicate solid solutions of α-Li3PS4, β-Li3PS4, γ-Li3PS4, Li10GeP2S12, and
Li4GeS4, respectively.

the expense of the melt until the crystals finally quenched to 2.4. Synthesis of LGPS Structures with Other Chemistries
RT to avoid β-phase formation. XRD measurement was per-
formed to precisely determine the crystal structure, while direct Many derivatives have been reported as members of the LGPS-
impedance measurements were performed to investigate the type material group. This section will describe each system.
anisotropic nature of the Li ion conductivity in LGPS.[45] The synthesis protocols are summarized in Table 1.
In addition to single crystal growth, some synthetic approaches
involved mixing in solvent without mechanical mixing.[46,47]
Machida et al. used two different solvents to obtain LGPS due
to differences in the reactivity of P2S5 and GeS2 in organic sol-
vents.[46] Wang et al. prepared LGPS in the form of a film using
hydrazine as solvent, although this solvent used is extremely
hazardous. For thin film configurations such as those prepared
by pulsed laser deposition, tetragonal crystalline phase with the
Li–Ge–P–S composition has not been obtained thus far.[58]

Figure 5. TEM observations for single-phase LGPS. Reproduced with


permission.[50] Copyright 2014, Tokyo Insititute of Technology. The right Figure 6. Compounds in the pseudo-ternary diagram for Li2S–GeS2–P2S5,
panel displays a wide region of a particle, whereas the left panel shows which frequently form secondary phases when synthesizing LGPS.
the diffraction spots of a local area. All of the spots are indexed by space Such compounds include Li4GeS4, Li2GeS3, Li7PS6, Li3PS4, and Li4P2S7
group of LGPS (P42/nmc). (Li4P2S6).[19,20,53–56]

Adv. Energy Mater. 2020, 2002153 2002153 (6 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

This composition was synthesized based on the notion that


the valency of P changes from pentavalent to tetravalent,[61]
and the structural analysis confirmed that this material has
the same structural type as LGPS. The Li9.6P3S12 phase can
be considered identical to the thio-LISICON analogue II phase
(80Li2S–20P2S5) reported by Hayashi et al.,[62] since they have
similar XRD patterns and electrical conductivity. Both materials
were synthesized by over 10 h of ball milling, followed by heat
treatment at temperatures as low as 240–260 °C. The low heating
temperatures (<300 °C) are below the synthesis temperature for
the Si, Ge, and Sn systems (450–550 °C), and the resultant low
crystallinity and a significant amount of the amorphous phase
render detailed structural analysis difficult. Therefore, the
actual structural composition of these crystalline phases still
requires elucidation. Presumably, when the valences of P and S
in the crystal are +5 and −2, respectively, the structural compo-
Figure 7. Single-crystal LGPS. Reproduced with permission.[45] Copyright
2019, the American Chemical Society. sition is approximately Li3+5δP1−δS4. However, γ- or β-Li3PS4 was
obtained at the stoichiometric composition of Li3PS4.[54] With
2.4.1. Si and Sn Systems respect to issues on compositional complexity in the Li–P–S
phases, Machida et al. reported that the Li2S–P2S5–P2S3 pseudo-
In 2013, Bron et al. reported the Sn analogue (Li10SnP2S12) of ternary system contains a crystalline phase different from that
LGPS.[28] In 2014, the Si analogue (Li11Si2PS12) of the LGPS-type reported in the Li2S–P2S5 binary system.[63] Identifying the crys-
material was synthesized by a high-pressure synthesis route.[30] talline phases in the Li2S–P2S5–P2S3 pseudo-ternary system and
During the same period, a series of solid solutions of Sn- and clarifying the chemical state of P atom in the crystalline phases
Si-derivatives were synthesized via the conventional synthesis will allow us to determine the exact chemical composition in the
route for LGPS-type materials by Hori et al.[31] These solid-solu- crystal structure of the LGPS-type Li9.6P3S12 phase.
tion systems can be written as Li10+δM1+δP2−δS12 (M = Si, Sn, Ge).
The value of δ indicates the range of the solid solution, which 2.4.3. Halogen Doping
varies depending on the constituent elements (M). The values
of δ are as follows: 0.20 < δ < 0.43 (Si), 0 < δ < 0.35 (Ge), and Halogen elements help to form the LGPS phase when used
−0.25 < δ < 0 (Sn). These differences are related to the size of the as additives, thus providing a highly ionic conductive phase
structural unit, such as [M]S4 and [Li]S6. The LGPS ­lattice cannot with reduced grain boundary resistance. In the Si system, the
be maintained when their sizes exceed the suitable range. ­synthesized Li9.54Si1.74P1.44S11.7Cl0.3 exhibited the highest ion
Hori et al. reported Li3PO4 as a side phase in the Si modifi- conductivity at RT (2.5 × 10−2 S cm−1) among the reported Li-ion
cation of LGPS; however, such oxides were not detected in the conductors.[13]
Sn or Ge systems. Interestingly, a recent study conducted by This material was synthesized at a heating temperature
Kim et al. reported that the oxygen substitution effect stabilized of 475 °C, with a highly crystallized LGPS phase and a small
the LGPS phase in Li–Si–P–S(–O) systems.[39] These studies amount of argyrodite as the secondary phase.[13] Subsequent
implied the possibility that a small amount of oxides contami- studies reported that this LGPS-type phase demonstrates a
nated the starting materials and might influence the final prod- core–shell morphology with Si- and Cl-rich compositions at the
ucts after synthesis. shell side. Although Rietveld analysis using high-flux synchro-
The framework ([M]S4) can comprise two or more M ele- tron and neutron diffraction patterns indicated that Cl occupies
ments; for example, M = Ge/Si,[59] Ge/Sn,[34,59] and Si/Sn[33,35] crystallographic sites in the LGPS-type structure,[13] it is difficult
have been reported thus far. Particularly in Si/Sn, the precise to confirm this due to the small amount of chlorine. Thus, fur-
search demonstrated that selecting suitable M elements dou- ther detailed analyses on this material are required.
bles the reported conductivity of simply substituted Si or Sn In the halogen doped Li–P–S system, compounds whose
analogues (>1 × 10−2 vs ≈5 × 10−3 S cm−1), thereby enabling a XRD patterns are similar to those for LGPS-type Li9.6P3S12
complete replacement of the expensive Ge with a cheaper ele- phase were previously reported for the iodide system. Such
ment without compromising the properties.[35] compounds in the bromide-doped system are also currently
under investigation.[64] However, detailed structures have not
yet been clarified for both systems. Additional studies are
2.4.2. Li–P–S Systems required to detect the positions of doped halogen elements in
the crystalline phases or amorphous phases surrounding them.
Metalloid elements such as Si, Ge, and Sn may not be desir-
able in terms of electrochemical stability since they are possibly
reduced at low electrochemical potential around 0 V (vs Li/Li+), 2.4.4. Oxygen Substitution
whereas the Li–P–S systems are more stable.[60] In this regard,
a system having the constituent element M substituted by P In contrast to the halogen-doped system, the oxygen atom
(Li9.6P3S12) was first highlighted as an LGPS-type material.[13] noticeably substitutes the sulfide atom in the crystal structure;

Adv. Energy Mater. 2020, 2002153 2002153 (7 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

the existence of solid solutions was confirmed by structural and SnS2 as starting materials and heat-treated the mixture at
analysis based on synchrotron XRD data.[37,38,65] Previous 700 °C.[42] In both cases, the specimen was crystallized in tetrag-
studies have systematically investigated oxygen substitution onal lattices with the I41/acd space group, where LGPS-type
in the LGPS structure, proposing that the oxygen-substituted structural units can be found (Figure 8b). The structural rela-
LGPS-type phases show improved electrochemical stability tionship will be discussed in the next part.
against reduction at low potential, even though they suffer from
a tolerable decrease in conductivity. Such properties change
depending on the degree of oxygen substitution.[37,38,65,66] 2.4.6. Other Chemistries
A study on the oxygen substitution in Li10GeP2S12 was
conducted by Sun et al. using Li2O as an oxygen source
­ The 4d site of LGPS can be occupied by Al3+ and Sb5+.[33,36]
in ­synthesis.[38] The substitution range (0 ≤ x ≤ 0.6) in Aliovalent substitution of M 4+ → Li+ + Al3+ enables con-
Li10GeP2S12−xOx was determined by observation of the con- trol of the lithium content in the LGPS structure, which may
tinuous lattice parameter change. When the oxygen-substituted impact the Li–Li interactions and thus the ionic conductivity
LGPS was used in the solid-state cell with Li–In anode and of the materials. Since there are limited reports on the triva-
LiCoO2, the cell exhibited improved cycle performance com- lent species in LGPS-type structures, those reports could be
pared to that using the non-doped counterpart, without com- the foundation of LGPS chemistry. A recent study conducted
promising the ionic conductivity (on the order of ≈10−2 S cm−1). by Liang et al. detailed Sb-substituted LGPS.[36] They success-
Kim et al. reported on the structural evolution with respect fully introduced Sb5+ into the 4d site using Sb2S5 as an anti-
to the degree of oxygen substitution in Li10SiP2S12−xOx.[39] They mony source in synthesis.[36] Interestingly, small amounts of
observed oxysulfide PS4−zOz units in the LGPS structure in the Sb (x = 0.075 in Li10Ge(P1−xSbx)2S12) significantly reduced the
Li–Si–P–S–O system using NMR and XRD analyses. Prior to generation of hydrogen disulfide under ambient atmosphere,
their study, as early as 2005, Takada et al. identified a meta- with no deterioration of the ionic conductivities. Furthermore,
stable phase with tetragonal symmetry and a composition of partial substation of sulfur by selenium was also reported for
0.75Li3PS4–0.25Li3PO4, which corresponds to Li9P3S9O3.[40] the Li–Ge–Sn–P–S system.[68] These examples indicate that
In 2016, Suzuki et al. confirmed the existence of solid solu- there are still numerous opportunities to control the properties
tions with LGPS-type structures in the Li–P–S–O system.[65] of LGPS-type phases owing to the ability of LGPS to function as
The LGPS-type phase in the Li–P–S–O system was stabilized a structural template.
by quenching the molten mixture from >700 °C. Interestingly,
the tetragonal LGPS structure could not be obtained in slowly
3. Structure
cooled samples, thus indicating that the LGPS structure could
be metastable in the Li–P–S–O system. The structure type of LGPS was discovered as a new
In the oxygen-related system, there is an obvious relationship one, whereas other important structure types in lithium
between the solubility limit and the structure. Oxygen d ­ isplaces thiophosphate systems originate from already known com-
­
sulfur at one specific site among three crystallographically equiv- pounds. For example, Li7P3S11,[69] lithium argyrodites,[70] and
alent sites. This selective substitution is similar to Li-argyrodites thio-LISOCON are derivatives of Ag7P3S11, silver argyrodite
such as Li6PS5Cl. When Cl is introduced to the high-tempera- (Ag8GeS6), and LISICON (γ-Li3PO4), respectively. The following
ture Li7PS6-argyrodite phase, selective substitution occurs at the sections describe unique characteristics of the crystal struc-
specific sulfur site that does not constitute the PS4 tetrahedra. tures found in LGPS. We begin with a structural overview, fol-
With regard to the solid-solution range of oxygen for the Li–P–S lowed by historical changes in the structural models proposed
system, Li18P6S16O8 would be obtained, assuming that all the by experiments and theoretical calculations. After subsequently
sulfur atoms present at the specific substitution sites were describing the arrangements of tetrahedral units and sulfur-
replaced by oxygen in Li18P6S24 (compositional formula per unit sublattices, this part concludes with a discussion about the Li
lattice). This composition (Li18P6S16O8) can be considered as the distribution, which is closely correlated to the Li+ conductivity.
maximum for the solid solution, and it correlates with the experi-
mentally confirmed maximum composition (Li18P6S16.8O7.2). This
supports the selective oxygen substitution at the specific sulfur 3.1. Structural Overview
site, which was observed by structural analysis.[67]
The crystal structure of LGPS can be described separately using
two components: 1) the skeleton composed of tetrahedral units
2.4.5. Sodium Conductors and 2) the Li ions distributed in the scaffold. Figure 8a1–a4
depict the crystal structure of LGPS determined using XRD
The existence of the Na variant of the LGPS-type structure was measurements performed on a high-quality single crystal.[45]
predicted by Richard et al. and experimentally synthesized in In the tetragonal lattice (P42/nmc No. 137) with a = 8.66402 Å
2016.[41] In 2017, the sodium superionic conductor Na11Sn2PS12 and c = 12.5830 Å, germanium and several of the phosphorus
was reported by two different groups using different synthesis atoms share the Ge1/P1 position (Wyckoff position: 4d), while
routes.[42,43] Duchardt et al. synthesized Na4SnS4 and Na3PS4 the remaining phosphorous atoms are located at the P2(2b)
separately, and then mixed the two compounds in a speci- position. These elements form [Ge1/P1]S4 and [P2]S4 tetrahe-
fied stoichiometric ratio and heated at 600 °C.[43] On the other dral units with four surrounding sulfur atoms, thus providing
hand, Zhang et al. synthesized the compound using Na2S, P2S5, a rigid scaffold in which the Li atoms are distributed. The four

Adv. Energy Mater. 2020, 2002153 2002153 (8 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 8. Characteristics of the LGPS-type structure.[30,42,45] a1– a4) Crystal structure of the original LGPS. a1) Perspective views of the tetrahedral
skeleton and Li sites with unit cell. a2) Connections between polyhedral units with Li sites as their center such as [Li1]S4 and [Li3]S4. a3) [Li3]S4 and
[Li2]S6. a4) [Li1]S4 and [Li4]S6. A section of the unit cell is also shown in (a3) and (a4). b1,b2) LGPS-type structural unit found in multiple unit cells in
Na11Sn2PS12. b1) Perspective view and b2) top view. Red frame indicates the LGPS-type structural unit found in multiple unit cells. c1–c3) Difference
in the arrangements of [P]S4 and [M]S4 units observed in Li10GeP2S12 with M = Ge (c1), Li11Si2PS12 with M = Si (c2), and Na11Sn2PS12 with M = Sn (c3).
In the Na11Sn2PS12, [P]S4 indicated by a blue arrow points in a direction different from the corresponding ones in the Li10GeP2S12 and Li11Si2PS12.

Li sites were identified by previous diffraction studies: Li1(16h), means of sophisticated diffraction measurements combined
Li2(4d), Li3(16h), and Li4(4c). with detailed analysis and theoretical calculation. The measure-
ments used for the structure determination have been updated
from powder XRD to single-crystal XRD and then powder neu-
3.2. Sophistication of Analytical Methods to Reveal tron diffraction at several temperatures. In the first report,[12] ab
Structural Details initio solution was used to determine the arrangement of [Ge]S4
and [P]S4 (see Figure 8a) by analyzing the high-flux synchrotron
The crystal structure model herein is obtained by continu- XRD powder data of Li10GeP2S12 using the FOX program,[71] fol-
ously updating and detailing the initial report from 2011, by lowed by the Rietveld refinement of powder neutron diffraction

Adv. Energy Mater. 2020, 2002153 2002153 (9 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

data to detect Li positions. The Li4 site was missing in the neutron powder diffraction. Single-crystal neutron ­diffraction,
initial stage,[12] which was first highlighted by theoretical cal- however, can function as a reliable method for accurate char-
culations.[72] Its position was confirmed by subsequent meas- acterization of Li atoms in the crystal structure, due to the rela-
urements using XRD on a single crystal that was picked up[44] tively large amplitude of the coherent neutron scattering length
and Rietveld refinement of powder neutron diffraction com- of Li. By applying this approach, further in-depth investigations
bined with the maximum entropy method (MEM) to visualize into the Li/vacancy ordering/disordering will become possible.
the Li positions.[57] Two subsequent reports showed a minor dif- Such in-depth investigations have been performed for LISICON-
ference in the site splitting of Li2 (4d to 8f) or Li3 (8f to 16h): type oxides, where superstructures originating from Li ordering
the former was a powder neutron diffraction study using bond were identified via powder diffraction combined with electron
valence analysis[32] and the latter was an XRD study on a high- diffraction measurements on single-crystal samples.[73,74] In the
quality single-crystal.[45] The latter structure model was also used future, it is an essential work to establish techniques that utilize
for LGPS-type phases in different chemical systems such as a high-flux neutron source to measure the diffraction from LGPS
Li–Sn–P–S and Li–Si–P–S–O.[28,37] It is also employed in this single crystals that have sufficiently large size and high quality.
review in the following section for a detailed description of Li sites.

3.4. Variation in Tetrahedral Arrangements


3.3. Conduction Pathways Deduced from Li Sites
3.4.1. Na11Sn2PS12 and Li11Si2PS12 Phases
Among all the four Li sites depicted in Figure 8a1, Li1 and Li3
are coordinated by four neighboring sulfur atoms, thereby Single-crystal studies have facilitated the characterization of
forming an edge-shared tetrahedral chain of [Li1]S4-[Li1] Na11Sn2PS12.[42,43] Prior to the single-crystal study, the sodium
S4-[Li3]S4- along the c-axis as depicted in Figure 8a2. On the analogue was believed to be Li10SnP2S12, which has the same
contrary, the Li2 and Li4 sites are relatively distant from any symmetry as the original LGPS. However, the phase was identified
surrounding sulfur atoms. When these Li2 and Li4 sites are as Li11Sn2PS12 with a tetragonal lattice of space group I41/acd.
considered six-coordinated,[57] they consist of [Li2]S6 and [Li4]S6 Figure 8b1,b2 depict the LGPS-type structure units in multiple
octahedral units, respectively, each of which are connect by unit cells for Na11Sn2PS12. According to the literature,[42] there
edge-sharing [Li3]S4 or [Li1]S4, thus forming chains over the are five crystallographic sites for Na, thus forming a 3D con-
ab-plane as shown in Figure 8a3,a4. duction pathway. Among the five sites, Na1, Na3, and Na4 can
The aforementioned chains are candidates for the Li be considered as components of the channel along the c-axis
conduction pathway in LGPS-type lattices. An initial report
­ seen in the LGPS-type cell; whereas Na2 and Na5 correspond to
confirmed that the pathway along the c-axis is the primary the Li4 and Li2 sites in LGPS to form a section of the lithium
one, because the Li1 and Li3 sites demonstrate larger atomic pathway over the ab-plane.
displacement parameters (ADPs) than the Li2 sites.[12] The Figure 8c1–c3 illustrate the arrangements of [M]S4 and [P]S4
large ADP indicates that Li ions located at Li1 and Li3 sites tetrahedral units for LGPS, Li11Si2PS12, and Na11Sn2PS12.[30,42,45]
have s­ignificant thermal vibration or are broadly distributed For all three, the P atoms occupy the pattern-filled tetrahedra at
around these sites, both of which are considered to facilitate Li the center in the top views, which correspond to 2b site in the
migration. Since the anisotropic ADPs for Li1 and Li3 are par- LGPS unit cell. The filled tetrahedra, which correspond to 4d
ticularly large along the c-axis, the pathway along the c-axis via sites in LGPS unit cell, are shared by P and Ge atoms in LGPS,
the Li1 and Li3 sites presumably has a larger contribution to Li while they are occupied by a single species here (Si in Li11Si2PS12
ionic conduction than those associated with Li2 sites over the or Sn in Na11Sn2PS12). To identify species in the 2b and 4d site,
ab-plane. Following the first report,[12] Kwon et al. measured the coordination environment was characterized by 31P- magic
the neutron diffraction at a low temperature of 100 K where angle spinning-nuclear magnetic resonance (MAS-NMR). In
the thermal vibration of Li is quite small, and identified the the case of Li11Si2PS12, where Si and P cannot be distinguished
exact Li positions.[57] Further, the determined ADP is large for by XRD due to their iso-electronic nature, practically only one
the Li1 and Li3 sites at low temperature, thus confirming a chemical state was observed for phosphorous in Li11Si2PS12,
broad Li distribution around these sites, particularly along whereas two states were clearly observed in the LGPS. Since
the c-axis. In addition, the Li4 position was observed, thus the Si/P molar ratio equals two in the composition Li11Si2PS12,
suggesting a migration path via Li4 site based on the atomic it was concluded that Si and P occupy the 4d and 2b sites, respec-
density distribution.[57] tively. Based on this model, the Si/P occupation is assigned as
For pathways along the c-axis or over the ab-plane, identi- [P/Si] and P for the 4d and 2b sites in the structural analysis of
fying the exact spatial distribution of Li plays a key role in quan- other Si-related LGPS-type phases, such as Li10+δSi1+δP2−δS12 and
tifying the pathway’s individual contributions to the overall Li Li9.54Si1.74P1.44S11.7Cl0.3. However, characterization using 31P MAS-
conductivity. However, as revealed by the history of structural NMR for these phases is required to obtain additional details.
modifications of LGPS (Section 3.2), the precise determina-
tion of Li position, occupancy ratio, and ADP is quite difficult.
Two factors hinder the study of Li using powder XRD meas- 3.4.2. Properties Related to Tetrahedral Arrangements
urements: 1) the X-ray scattering intensity of Li is extremely
weak and 2) there are barely any independent reflections in the Arrangements of the tetrahedral units affect the structure sta-
powder samples. The latter is a limitation when measuring the bility of LGPS. A previous ab initio calculation suggests that the

Adv. Energy Mater. 2020, 2002153 2002153 (10 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

structure may be stable at >276 K when considering entropic thio-LISICONs, for example, Li3PS4 and Li4GeS4, the arrays of
effects based on the disordered distributions of [Ge]S4 and [P]S4 sulfur are regarded as hcp arrays.[54] Conversely, sulfur packing
units, as well as the Li atoms.[75] Otherwise, the LGPS should be in LGPS is reported to be more bcc-like rather than hcp.[82]
thermodynamically unstable at the ground state.[76] It is empirically known that materials with bcc anion
The site-sharing [M/P] tetrahedral units can be associated packing exhibit high cation conductivities, such as AgX and
with the mixed polyanion effect proposed for the LISICON CuX (X = Cl, Br, or I).[83] A previous theoretical study suggested
system, where the conductivity is enhanced when numerous that this empirical rule is true for Li conductors with sulfur
types of tetrahedral units exist in the crystal. Conductivity sublattices.[82] Figure 9 depicts the potential Li jumps in the face-
enhancement has actually been observed for the LGPS-type shared polyhedron for thio-LISICONs and LGPS. In γ-Li3PS4
phase containing [Si/Sn/P]S4,[34] which can be expected when or Li4GeS4 with distorted hcp sulfur sublattices, there are three
a suitable M/P ratio is selected for the other LGPS-type phases kinds of possible Li jumps: from tetrahedra(+) to tetrahedra(−)
containing site-shared tetrahedral units, such as [Ge/Si/P]S4,[59] or vice versa [abbreviated as Tetra.(+)–Tetra.(−)], Tetra.(+/−)–
[Ge/Sn/P]S4,[59] or prospectively [Si/Ge/Sn/P]S4. Octa., and Octa.–Octa. The combination of these jump steps
contribute to the Li migration process in the crystal. In the case
of bcc sulfur lattices, Li jumps between face-shared tetrahedra
3.4.3. Tetrahedral Orientation Differences Among LGPS- (indicated by Tetra.1 and Tetra.2 in Figure 9b1), contributing
and Thio-LISICON-Type Phases to the Li migration process. For ideal or undistorted sulfur
sublattices, theoretical calculations indicated that the jump
In addition to the manner in which the tetrahedral units are barrier was lower for Tetra.1–Tetra.2 in bcc than for jumps in
occupied, their orientations are different in the LGPS-type hcp, thereby suggesting that the bcc sublattice is favorable for
phases with P42/nmc symmetry and Na11Sn2PS12 with I41/acd. achieving high Li conductivity. In the case of LGPS, bcc-like
Figure 8c3 shows that in Na11SnP2S12, an apex on one of the [P] arrangement can be readily found around Li1 and Li3 sites.
S4 units differs from those in LGPS and Li11Si2PS12, in accord- These facts are consistent with each other, thus verifying that
ance with their different space groups. the structure of LGPS favors Li migration and also suggesting
The tetrahedral orientations or arrangements provide mate- that the main conduction pathway for LGPS is a chain in one-
rials with variety and help us to comprehend their structural direction along the c-axis via the Li1 and Li3 sites (Figure 8a2).
relationships or classification[77–79] (e.g., LISICON systems[80]).
Similarly, visualizing the tetrahedral arrangements could reveal
the structural differences among LGPS- and thio-LISICON- 3.6. Li Distribution Visualized by MEM
type phases, all of which are similar to each other in compo-
sition. For example, such structural differences occur for thio- From Section 3.5, a low energy barrier between the Li1 and Li3
LSICONs in the Li3PS4–Li4GeS4 pseudo-binary system con- sites in the LGPS crystal (corresponding to a flat energy land-
taining LGPS. In the γ-Li3PS4 crystal structure, the [P]S4 units scape) is expected around these sites, and as a result the Li ions
are ordered with their apexes oriented in the same direction. In could be distributed broadly. Such an expected flat potential
contrast, the tetrahedral units of β-Li3PS4 and Li4GeS4 are situ- is supported by the bond valence sum and one-particle poten-
ated with their apexes alternating in opposite directions, thus tial calculated using results from the structural analysis.[32] In
forming a zigzag arrangement. Similarly, a zigzag arrangement fact, the widely spread lithium distribution has been visualized
is observed in the α-Li3PS4. In contrast, as a unique character- using MEM.[13,32,57,84,85]
istic of the LGPS-type members, the GeS4 and PS4 tetrahedra MEM enables the deduction of nuclear or electronic dis-
arrays are aligned in a more disordered manner, such that tributions without assuming a structure model. The use of
the vertices do not point in the same direction or in a zigzag MEM in conjunction with the neutron data for LGPS-type
arrangement with alternating directions. The change in tet- materials facilitates visualization of the Li distributions in the
rahedral arrangements in the Li4GeS4–Li3PS4 system may be LGPS-type lattices as a negative portion of the nuclear scat-
attributed to the differently sized tetrahedra arising from the tering density distribution. Figure 10 shows the nuclear den-
difference in ionic radii between Ge4+ (0.39 Å for C.N. = 4) and sity distribution of Li atoms in various LGPS-type cells. At a
P5+ (0.17 Å for C.N. = 4).[81] As the Ge content increases, the low temperature (Figure 10a), Li distribution in the original
γ-Li3PS4-type phase with the PS4 tetrahedra ordered in the same LGPS exhibits 1D connections along the c-axis between the Li1
direction changes to the β-Li3PS4-type phase, which possesses and Li3 sites, whereas the other sites (Li2 and Li4) are isolated
more disordered tetrahedral arrangements. This eventually from each other. As the temperature increases, the 1D chain
results in specific arrangements of the tetrahedra in the LGPS- becomes more evident (b and c). Such a wide distribution
type structure at a Ge4+/P5+ molar ratio close to 1/1. Thus, intro- along the c-axis is consistent with the theoretical suggestion
ducing or mixing different-sized tetrahedral units might be a that Li conduction could be enhanced if Li occupies the tetra-
key factor for producing the LGPS-type structure. hedral sites in crystals with bcc-like packing. The Li sites that
are connected along the c-axis even at low temperature imply
that this chain is the primary or more important pathway, as
3.5. Sulfur Sublattice compared to the pathways through Li2 and Li4 sites in the
ab-plane. At RT, the Li2 and Li4 sites are still isolated, while
Due to changes in the tetrahedral arrangements, the packing of the Li4 site is connected to Li1 at high temperature (750 K).
sulfur atoms differs between LGPS and thio-LISICONs. In the Kwon et al. highlighted changes in the conduction mechanism

Adv. Energy Mater. 2020, 2002153 2002153 (11 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 9. Single Li+ jumps between tetrahedral or octahedral sites in the sulfur sublattice with hcp-like packing in thio-LISICONs (a1–a3) and bcc-like
packing in LGPS (b1–b3).[82] a1) Octahedral and tetrahedral positions in the hcp lattice. a2) Li+ jumps in γ-Li3PS4 between tetrahedra(+)–octahedra.
a3) A Li+ jump in Li4GeS4 between octahedra–octahedra and tetrahedra(+)–tetrahedra(−). b1) Tetrahedral positions in hcp lattice. b2,b3) A Li+ jump
from the Li1 (b2) or Li3 (b3) site (tetrahedra 1) to the next face-shared tetrahedral site (tetrahedra 2).

Figure 10. Li distribution for various LGPS-type materials visualized by MEM.[13,57,84,85] For the purpose of comparison, the iso-surface level is set at
−0.06 fm Å−3. Z-MEM and Z-3D software are used for the calculation and visualization of negative nuclear density distributions, respectively. Each unit
cell was sliced at z = 0.75 in fractional coordinates to display the connections in Li distributions around the Li1 and Li4 sites.

Adv. Energy Mater. 2020, 2002153 2002153 (12 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

from 1D-like to 3D-like, and suggested that a 3D Li distribution Table 2. Activation energy (Ea), pre-exponential factor (σ0), and RT
at lower temperature may help improve the conductivity.[57] ­conductivity (σRT) of LGPS-type materials.
Although this has not been achieved for Li9.81Sn0.81P2.19S12
Composition σRT [S cm−1] σ0 [KScm−1] Ea [eV] Temperature
(5 × 10−3 S cm−1 at 298 K) shown in Figure 10d, the intro- range [K]
duction of Si provides Li10.35[Sn0.27Si1.08]P1.65S12[84] with 3D
Li9.54Si1.74P1.44S11.7Cl0.3[13] 2.5 × 10−2 1.87 × 105 0.26 243–298
Li distribution and a higher conductivity of 1.1 × 10−2 S cm−1
[59] −3
(Figure 10e). Moreover, a conductivity of 2.5 × 10−2 S cm−1 has Li10(Si0.5Ge0.5)P2S12 4.2 × 10 1.05 × 10 5 0.29 243–298

been achieved for the Li9.54Si1.74P1.44S11.7Cl0.3 system with 3D Li Li10(Ge0.5Sn0.5)P2S12[59] 6.0 × 10−3 8.27 × 104 0.28 243–298
distribution (Figure 10f). Li10(Si0.5Sn0.5)P2S12[59] 4.3 × 10−3 2.35 × 105 0.31 243–298
Note that information provided by MEM analysis is limited Li10GeP2S11.7O0.3 [38]
1.2 × 10 −2
3.30 × 10 3 0.18 298–398
to the static behavior of Li, and thus the dimensionality of Li
Li9.6P3S12[13] 1.2 × 10−3 5.57 × 105 0.36 173–253
distribution from MEM may not perfectly explain the conduc-
tivity–composition relationship. Nevertheless, the visualization 3.36 × 103 0.24 253–373
of Li distribution using MEM has qualitatively rationalized this Li9P3S9O3[40] 4.3 × 10−5 1.03 × 104 0.35 298–473
relationship in aforementioned LGPS-type phases. A more in- Li10GeP2S12 [12]
1.2 × 10−2 5.65 × 105 0.31 193–298
depth understanding may come from combining this approach
5.73 × 103 0.17 322–673
with future studies on the dynamic behaviors of Li ions.
Li10.35Si1.35P1.65S12[31] 6.7 × 10−3 2.09 × 105 0.30 173–298
6.71 × 102 0.14 322–673

4. Ionic Transportation Li9.81Sn0.81P2.19S12[31] 5.5 × 10−3 1.76 × 105 0.29 173–298


2.65 × 103 0.19 322–573
Among the lithium ionic conductors, Li10GeP2S12 and related
Li9.42Si1.02P2.1S9.96O2.04[37] 1.0 × 10−4 1.44 × 103 0.27 298–523
materials exhibit very high ionic conductivities in the range
of 5 to 20 × 10−3 S cm−1 as listed in Table 2. These conduc- Li10GeP2S12 Single 7.0 × 10−3 1.41 × 106 0.34 243–293
tivity values are comparable to or even higher than those of crystal [110][45]

the organic liquid electrolytes currently used in lithium-ion 1.47 × 108 0.43 200–243
batteries. Figure 11 compares the conductivities over a wide Li10GeP2S12 Single 2.8 × 10 −2
3.76 × 10 6 0.33 243–293
temperature range. A recent single-crystal study reported a con- crystal [001][45]
ductivity of 2.7 × 10−2 S cm−1 at 300 K along [001].[45] Analysis of 1.00 × 108 0.40 200–243
the ionic conductivity of LGPS-related materials is used to facil- [34] −3
Li10GeP2S12 7.6 × 10 9.29 × 10 4 0.27 233–333
itate a fundamental understanding of those ionic conductive
Li10Ge2/3Sn1/3P2S12[34] 6.4 × 10−3 8.42 × 104 0.28 233–333
mechanisms and to foster the future development of superi-
onic conductors. Li10Ge1/3Sn2/3P2S12 [34]
4.8 × 10−3 9.25 × 104 0.28 233–333
The ionic conductivities of LGPS-type materials exhibit Li10SnP2S12[34] 3.8 × 10−3 1.55 × 105 0.30 233–333
Arrhenius-type evolutions (Figure 11) similar to other inorganic Li10GeP2S12 grain[33] 1.0 × 10−2 1.01 × 106 0.32 143–213
superionic conductors. Most of the ionic conductivity data were
Li10GeP2S12 powder[33] 9.5 × 10−3 6.76 × 105 0.32 143–333
obtained by impedance measurements using materials that
were compressed or in a sintered powder state. Therefore, the Li10(Si0.3Sn0.7)P2S12[33] 7.6 × 10−3 3.20 × 105 0.30 143–333
obtained impedance values include both the grain and grain Li10SnP2S12[28] 5.6 × 10−3 4.55 × 105 0.32 143–333
boundary contributions. LGPS-related materials exhibit grain
As Ea and σ0 can vary with the measurement temperature, the corresponding
and boundary capacitances of 35–50 pF cm−2 and 0.1–10 nF cm−2,
t­ emperature range are also listed.
respectively. Theoretically, this difference in capacitance ena-
bles deconvolution of the processes in the grain and grain
boundary. Unfortunately, at RT, it is extremely difficult to obtain N, q, kB, T and Dσ,0 are the number of carriers per unit volume,
the semicircle of the grain process in the Nyquist plot, even at charge of carriers, the Boltzmann constant, temperature and
a frequency of 100 MHz and with a single crystal.[45] To obtain the pre-exponential factor for charge diffusion, respectively.
the separate steps, extremely low temperatures are typically Considering the above equation, σ0 and the activation energy
required, that is, below 180 K. Bron et al. conducted a detailed (Ea) are the key parameters used to describe the ionic conduc-
analysis of the impedance spectra of LGPS-type materials,[33] tivity of materials.
and reported that the grain and grain boundary conductivities The conductivity parameters for the LGPS-type materials
have similar activation energies at low temperatures. However, are listed in Table 2. At high temperatures (T above ≈350 K),
when the temperature exceeds RT, the contribution of the grain a slightly negative curvature than that below RT was observed
boundary becomes minor compared to that of the lithium con- for a majority of the Li–M–P–S (M = Ge, Si, Sn, or their com-
ductive oxides. binations) materials.[13,57,85] Additionally, a higher activation
The conductivity of inorganic superionic conductors (σ) can energy was observed at low temperature from the single-crystal
be described as an Arrhenius-type activated process measurement.[45] Figure 12 is the plot of σ0 as a function of Ea
(Meyer–Neldel plot) for the LGPS-type materials. For the LGPS
 E 
σ T = σ 0 exp − a  (1) structure, Figure 12 only shows the plot below 350 K. There is a
 kBT  general positive correlation between Ea and σ0, which is termed

Adv. Energy Mater. 2020, 2002153 2002153 (13 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 11. Comparison of temperature dependence for LGPS ionic conductivity with those of other Li solid electrolytes. For comparison with LGPS-type
phases, panels (a), (b), and (c), respectively, display data for oxides, sulfides, and others, such as hydrides and halides, all of which show the highest-
class conductivity values among each material group. For reference, data for several polymer and liquid electrolytes, as well as Li2SO4 and RbAg4I5, are
also included. Presented conductivity–temperature relationships were obtained from published works provided in Note S2, Supporting Information.

the Meyer–Neldel rule. This general trend has been observed Among the LGPS-type structures, a majority of the materials in
for different kinds of thermally activated conduction systems, the Li–M–P–S system exhibit Ea and σ0 in the 0.25–0.35 eV and
such as semiconductors[86] and ionic conductors.[87–89] Zeier 105–107 K S cm−1 ranges, respectively, at RT to the low tempera-
et al. highlighted that this trend also occurs for state-of-the-art ture range (below ≈350 K). The single-crystal measurements
superionic conductors, and it is the major challenge toward also yielded similar results for Ea and σ0 as those obtained
achieving both a high prefactor and a low activation energy.[34,90] from the powder samples. At higher temperatures (>350 K),

Figure 12. Mayer–Neldel plot for LGPS,[12,13,28–30,33–35,37,45,57] argyrodite,[90–92] garnet,[93–95] and thio-LISICON.[21,96] The activation energy (Ea) and pref-
actor (σ0) was estimated and the conductivities reported at the temperature range listed in Table 2. The entire material system follows a general
dependence between the prefactor and activation energy (Mayer–Neldel rule). The LGPS exhibits a relatively high prefactor among the superionic
conductors listed herein.

Adv. Energy Mater. 2020, 2002153 2002153 (14 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

the conductivity parameters take on different values from those in Figure 13a.[76,102,103] The tracer-diffusion coefficients of the
at RT and low temperatures. In this situation, the activation LGPS-type structures at RT are in the range of 2 × 10−12 to
energy becomes lower (0.15 eV < Ea < 0.20 eV), which is con- 3 × 10−12 m2 s−1 with an activation energy of ≈0.2–0.25 eV. The
sidered extremely low among typical solid-state ionic conduc- highest diffusion coefficient at RT was obtained for the Si
tors. On the other hand, the prefactor of the high-temperature system (Li11Si2PS12), followed by the Ge (Li10GeP2S12) and Sn
region ranges within ≈103–104 K S cm−1. The c­ombination of (Li10SnP2S12) systems.[30] This is in agreement with an AIMD
low Ea and σ0 results are comparable to the extrapolated con- simulation performed by Ong et al.[104] A diffusion coeffi-
ductivity at RT, σ300K ≈10−2 S cm−1. Interestingly, only the single- cient in the order of 10−12 m2 s−1 at RT is considered very high
crystal measurement indicates the combination of relatively among the superionic conductors. For example, the diffusion
high Ea (0.4 eV) and σ0 (108 K S cm−1) in the low-temperature coefficients of LISICON at RT are in the 10−13 m2 s−1 range. The
region. This combination also results in an extrapolated high diffusion coefficient is the main factor responsible for the
σ300K = 10−2 S cm−1.[33] The difference in the low-temperature high ionic conductivity of LGPS-type materials.
behavior between the powder and the single crystal may be However, if we limit the discussion to materials with a
due to the effect of grain boundary conduction, which typi- conductivity of 10−2 S cm−1 at RT, the unique conduction
cally displays an activation energy of 0.3 eV in the LGPS struc- mechanism of LGPS is revealed. Apart from the LGPS system,
ture.[33] Deviation from Li–M–P–S was observed when more there are very few systems that exhibit such high conductivity
than 16% of the sulfur was replaced by oxygen; Li9P3S9O3 and at RT. Therefore, we selected three different systems—argyro-
Li9.42Si1.02P2.1S9.96O2.04 exhibit decreased σ0 compared to Li–M–P–S dite, LiTi2(PS4)3 (LTPS), and complex hydride system.[91,103,105,106]
systems despite possessing comparable activation energies.[13,37] We could obtain the parameters related to ionic conductivity,
The lattice dynamics of the LGPS structure may also be influ- namely the number of lithium ions per unit volume (N), the
enced by anion substitution. A decrease in the prefactor was charge diffusion coefficient (Dσ), and Dtr, which are listed
also observed in the cation substitution system of Li9.6P3S12. in Table 3. We plot the two diffusion coefficients versus N in
The Meyer–Neldel plot for different types of lithium ionic Figure 13b,c. It is apparent that the N values for LTPS and
conductors are also shown in Figure 12.[21,90–96] Here, the the complex hydride system are an order of magnitude lower
Meyer–Neldel rule also holds for the argyrodite, garnet, and than those of the thiophosphate LGPS and argyrodite systems.
thio-LISICON structures. The activation energies of these supe- The diluted and condensed Li systems exhibit N values on the
rionic conductors are in the 0.2–0.5 eV range, which is sig- order of 10−21 and 10−22 cm−3, respectively. For LGPS, the value
nificantly lower than those of materials discovered in the early N = 2.1 × 10−22 cm−3 corresponds to 35 mol L−1, which is con-
stages of Li conductor research (Ea ranging in 0.65–1.2 eV). siderably higher than those of organic liquid electrolyte systems
Nevertheless, such a low activation energy is a major feature (typically 1–3 mol L−1). Interestingly, the high conductivity of
of modern superionic conductors and a source of their high diluted Li systems is achieved by significantly higher charges
ionic conductivity near RT. LGPS-type materials generally have and tracer-diffusion coefficients than those of LGPS and argy-
a relatively high prefactor (σ0) among the superionic conduc- rodite (Figure 13b). The tracer-diffusion coefficients for LTPS
tors, as shown in Figure 12. This indicates that the prefactor (1.2 × 10−11 m2 s−1) is one order of magnitude higher than that
is also an important parameter for explaining the high con- for LGPS (2.2 × 10−12 m2 s−1). Since the activation energy of dif-
ductivity of LGPS at RT (σ300K). As the prefactors may vary fusion is almost the same, the high diffusion coefficient of the
greatly depending on the extrapolation method, obtaining diluted systems is mainly a result of the high prefactor. The
and discussing reliable data about them are more challenging energy frustration due to the highly distorted local environment
compared with the activation energy. Therefore, the prefactor is proposed as the main contributor to the high prefactor of
is not as well-interpreted as the activation energy. However, in LTPS and hydride complex systems.[103,106] In other words, the
theory, the prefactor provides a significant amount of physical environment of lithium and the conduction mechanisms are
information, and thus understanding it is important for LGPS completely different between LTPS and LGPS.[103]
and other superionic conductors. As a result, the prefactor Based on the observation of thiophosphate systems, LGPS-
has received significant attention and remains under active and argyrodite-type materials possess similar lithium num-
investigation. bers (N) and charge diffusion coefficients. However, interest-
Dtr  ingly, the tracer-diffusion coefficient of LGPS-type materials
By introducing the Haven ratio HR = D  , the conductivity
 σ  is one order of magnitude lower than that of the argyrodites,
can be connected to the thermally activated hopping process. which immediately results in a smaller HR for the former.
Dσ, Dtr are the charge diffusion and the tracer-diffusion coef- HR = 1 is understood as simple diffusion mechanism (no
ficient, respectively. The tracer-diffusion coefficient of a tagged interaction between charge carriers) and the smaller HR indi-
atom is obtained from the mean-squared displacement of an cates the stronger interaction. We listed the values of N, Dσ,
ensemble of such atoms. It can also be directly measured by Dtr, and HR for these four systems in Table 3. HR ≈ 0.25 for
neutron radiography,[97,98] or pulsed-field gradient nuclear LGPS, whereas >0.4 for argyrodites. We note that HR = 0.23 was
magnetic resonance (PFG-NMR).[99–101] Bron et al. and Kuhn calculated for argyrodite[105] by carefully defining the number of
et al. measured the tracer-diffusion coefficients of LGPS mobile lithium ions (4 × 10−21 cm−3). In order to simplify the
structures using PFG-NMR (Figure 13a).[28–30] In addition, ab discussion, all the calculated Dσ values are based on the total
initio molecular dynamics (AIMD) is a powerful technique number of lithium ions in the unit cell, which follows the
to simulate the tracer-diffusion coefficients. The results of IUPAC definition. The smaller HR compensates for the ionic
AIMD obtained by different research groups are also shown conductivity of LGPS (1.0 × 10−2 S cm−1 at RT). Since the HR

Adv. Energy Mater. 2020, 2002153 2002153 (15 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 13. Transport parameters of LGPS. a) Temperature dependence of the tracer-diffusion coefficients for LGPS measured by PFG-NMR.[28–30,76,102,103]
Lithium concentration (N) versus charge diffusion b) and tracer-diffusion c) coefficient of LGPS,[12,28–30,33,34,57] and other superionic conductors.[91,103,105,106]

is interpreted as the extent of cooperative motion of lithium, a of the nuclear density distribution depends on the surface
stronger Li–Li interaction is implied in LGPS than that in argy- level.[85] Therefore, all of the Li nuclear density distributions in
rodite. Therefore, this interaction may be more important for Figure 10 are displayed at the same surface level of −0.06 fm Å−3
understanding the ion transport mechanism in LGPS. to enable a reasonable comparison of distribution-dimension-
ality. In the LGPS-type structure, a continuous nuclear distribu-
tion of lithium along the c-axis is observed, regardless of the
4.1. Framework Structure and Ionic Conduction Path composition. This evidence suggests a strong preference for
lithium diffusion along [001]. Interestingly, the distribution of
As mentioned in Section 3.5, one of the characteristics of the
LGPS-type structure is its unique anion substructure. The S2−
Table 3. Lithium concentration (N), charge diffusion coefficient (Dσ),
anion in LGPS exhibits a more bcc-like sublattice, which is and lithium tracer- diffusion coefficient (Dtr) at RT.
different from the hcp sulfur sublattice that is often observed
in other thiophosphate systems (Figure 9).[54] The ideal bcc S2− Structure Composition a)N [cm−3] Dσ [m2 s−1] Dtr [m2 s−1] HR
substructure provides a continuous face-sharing tetrahedral LGPS Li10GeP2S12 [12]
2.1 × 1022 9.2 × 10−12 d)2.2 × 10−12 0.24
path for lithium with a low activation energy of 0.15 eV.[82]
Li10GeP2S12[57] 2.1 × 1022 9.2 × 10−12 d)2.2 × 10−12 0.24
On the other hand, lithium possesses an octahedral site that
migrates across the unit cell of the hcp and fcc S2− sublattice Li10GeP2S12[34] 2.1 × 1022 5.9 × 10−12 d)2.2 × 10−12 0.38

structures. A relatively high activation barrier (≈0.4 eV) is con- Li10GeP2S12[29] 2.1 × 1022 7.0 × 10−12 2.2 × 10−12 0.32
firmed at the transition state from tetrahedra to octahedra. Li10GeP2S12 [33]
2.1 × 1022 7.7 × 10−12 d)2.2 × 10−12 0.28
Therefore, the bcc S2− structure is essential because it consists Li10SnP2S12[28] 1.7 × 1022 6.6 × 10−12 1.8 × 10−12 0.27
of continuous face-sharing tetrahedra that facilitate in ion trans-
Li10SnP2S12 [30]
2.0 × 1022 3.1 × 10−12 1.4 × 10−12 0.45
port.[82] Di Stefano et al. also conducted a quantitative analysis
of the interstitial sites of LGPS using the concept of tetrahe- Argyrodite Li5.5PS4.5Cl1.5[105] b)
2.3 × 1022 8.3 × 10−12 1.0 × 10−11 <1
dricity[103,107] and revealed that the majority of interstitial sites in 4.2 × 10 21 4.5 × 10−11 1.0 × 10−11 0.22
the LGPS structure consisted of slightly distorted tetrahedra.[103] Li5.75PS4.75Cl1.25[105] b)
2.3 × 1022 2.9 × 10−12 6.9 × 10−12 <1
Another characteristic of the LGPS-type structure is its c)4.2
× 1021 1.6 × 10−11 6.9 × 10−12 0.43
unique arrangement of MS4 tetrahedra. The tetrahedral
arrangement style of LGPS is known to be different from either Li6PS5Cl[105] b)2.3 × 1022 1.8 × 10−12 4.0 × 10−12 <1

the α, β, or γ phases of Li3PS4 or any of the other thiophos- c)4.2 × 1021 9.6 × 10−12 4.0 × 10−12 0.41
phates.[54] As a result, the interstitial structure, which is the Li6.6P0.4Ge0.6S5I[91] b)2.5 × 1022 1.2 × 10−11 6.4 × 10−12 0.54
pathway for lithium, forms a 1D tunnel along the c-axis and an Hydride 0.7Li(CB9H10)– 4.6 × 1021 2.4 × 10−11 5.7 × 10−12 0.24
additional interstitial path in the ab plane. The mobile lithium complex 0.3Li(CB11H12)[106]
is located in the tunnel along the c-axis (16h:Li1 and 16h:Li3
LTPS LiTi2(PS4)3[103] 2.0 × 1021 5.1 × 10−11 1.2 × 10−11 0.24
in Figure 8) and the interstitial octahedral site in the ab-plane
(4c:Li4 in Figure 8). The ADPs of lithium in the tunnel and Dσ was obtained from the Nernst–Einstein relation using the reported ionic con-
in the ab-plane are highly anisotropic along the [001] and [110] ductivities and N. a)The concentration of lithium ion (the number of atoms per unit
directions, respectively (Figure 8).[12,44,45] This indicates that volume) was calculated based on the crystal structure (all the lithium was consid-
ered for the concentration calculation); b)All the lithium was counted (Li number/
lithium migration occurs along [001] and [110]. MEM has ena-
unit cell = 24); c)Reported value; the number was carefully determined from the
bled visualization of the nuclear distribution of lithium in the diffusion mechanism (Li number/unit cell = 4); d)The tracer diffusion coefficient for
LGPS structure (Figure 10).[32,57,85] We note that the appearance Li10GeP2S12 from ref. [26].

Adv. Energy Mater. 2020, 2002153 2002153 (16 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 14. Anisotropic lithium transport properties of LGPS. a) Tracer-diffusion coefficient simulated by AIMD. Reproduced with permission.[76]
Copyright 2012, the American Chemical Society. b) Ionic conductivity of single-crystal LGPS measured at different directions along [110] and [001].
c) Schematic of the possible Li-ion conducting paths in LGPS. Reproduced with permission.[45] Copyright 2018, the American Chemical Society.

lithium along [110] was found to be composition-dependent. also reported in a single crystal experiment (Figure 14b),[45] with
At RT, Li9.81Sn0.81P2.19S12 exhibits a slightly lower ionic conduc- the measured conductivity at RT being 27 and 7 × 10−3 S cm−1
tivity (5 × 10−3 S cm−1) than Li10GeP2S12 (1.2 × 10−2 S cm−1), but along [001] and [110], respectively. Despite the low activation
there is no continuous distribution between lithium atoms energy (0.17 eV) along the [001] direction from theoretical
in the ab-plane and those in the tunnel along [110] even at simulation, interestingly, the experimentally measured values
800 K.[85] In the more conductive Li10GeP2S12, lithium atoms are much higher (0.3 eV near RT and 0.4 eV at low tempera-
in the ab-plane begin to have a continuous nuclear distribu- ture) and comparable to that along [110]. The value of 0.3 eV
tion toward those in the tunnels at elevated temperature.[57,85] is in agreement with that of the [110] path obtained by simula-
A 3D connection of lithium nuclear densities was observed at tion.[76] In addition, by considering the charge plane densities of
300 K for Li9.54Si1.74P1.44S11.7Cl0.3, which has a very high ionic 5.328 nm−2 ([001] path) and 2.594 nm−2 ([100] path), the tracer-
conductivity of 2.5 × 10−2 S cm−1 at RT.[84] The recently reported diffusion coefficient along [001] is expected to be higher than
Li10.35Sn0.27Si1.08P1.65S12 also exhibited a high ionic conductivity that of [110] by only a factor of two or three. This is also not in
above 1.0 × 10−2 S cm−1 and a 3D diffusion path of lithium at agreement with the ab initio calculations. The PFG-NMR exper-
RT.[84] This suggests that the formation of a 3D diffusion path iment performed by Kuhn et al. also did not detect the very low
of lithium through the octahedral interstitial site (4c) has an activation energy predicted for [001], and their obtained value
influence on the ionic conductivity of the material. of 0.25 eV may correspond with [110] diffusion.[29] It indicates
that during the given diffusion time, the majority of lithium in
the LGPS structure undergoes in-plane diffusion. The diffusion
4.2. Properties of [001] and [110] Transport time (Δ ≈ 25 ms) measured by Kuhn et al. corresponds to a dif-
fusion length of ≈ 0.5 µm. Even for the only 0.5 µm displace-
Mo et al. observed both [001] and [110] diffusion using AIMD. ment along [001], the [110] path might be involved as a detour.
The diffusion property is better along [001] than that along [110], Iwasaki et al. proposed a model to explain this weak aniso-
with the respective simulated activation energies of 0.17 and tropic conductivity (Figure 14c). Due to its internal condition,
0.28 eV (Figure 14a).[76] Adams also reported similar activation Li migrates through the [110] path for long-range diffusion. A
energies of 0.19 and 0.3 eV for [001] and [110], respectively.[72] The simple explanation for this is the existence of a possible defect
high conductivity along [001] is consistent with the inference in the [001] channel.[42] The significant impact of a small defect
from structural analysis. The high conductivity along [001] was on 1D conductors has been highlighted by Malik et al. and

Adv. Energy Mater. 2020, 2002153 2002153 (17 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Mo et al.[76,108] The most important finding of these studies is Table 4. Jump rate (t−1) for LGPS-type materials.
that Li transportation, most likely along the [110] path, which
Composition Temperature, Tm [K] Jump rate [t−1/s−1] Method
was considered a secondary contributor to conductivity, is
highly involved in long-range (>0.5 µm) diffusion. The combi- Li10GeP2S12 [29]
330 1× 109 Spin–lattice relaxation R1
nation of the ultrafast lane along [001] and the slower but still 300 5 × 108 Spin–lattice relaxation R1
very conductive path along [110] (7 × 10−3 S cm−1) is the primary 150 1 × 105 Spin–spin relaxation R2
reason why LGPS displays a conductivity comparable to that of
135 1.4 × 104 Line narrowing
liquid electrolytes.
Using the dipolar line width and Li spin-alignment echo Li10GeP2S12[109] 370 1 × 109 Spin–lattice relaxation R1

(SAE) NMR at low temperature, Liang et al. succeeded in 210 3 × 105 Spin–lattice relaxation R1
experimentally detecting a low activation energy of 0.16 eV Li10SnP2S12 [30]
360 9.8 × 108 Spin–lattice relaxation R1
for lithium displacement in Li10GeP2S12, which may be attrib- 145 1.4 × 104 Line narrowing
uted to displacement along [001].[109] The NMR relaxometry [30]
Li11Si2PS12 125 1.5 × 104 Line narrowing
data obtained by Liang et al. (Figure 15) and the results from
Kuhn et al.[29] both indicated a comparable jump rate by spin–lat-
tice relaxation. The jump rate of τ−1 = 1 × 109 s−1 was obtained Nevertheless, a simple interpretation of the result suggests that
by Kuhn et al. and Liang et al. at 330 and 370 K with the acti- a soft-mode lithium jump along [001] is more liquid-like and is
vation energies of 0.22 and 0.26 eV, respectively. The jump rates associated with a shallow energy landscape, that is, the lithium
obtained by the two groups are listed in Table 4. The activation position is entropically stabilized in the [001] tunnels. Again, since
energy of Kuhn et al. strongly correlates to that obtained using the NMR relaxometry produces extremely interesting insights, it
PFG-NMR. Therefore, Kuhn et al. stated that the spin–lattice is anticipated to yield even greater understanding.
relaxometry measures a displacement along [110]. Liang et al.
obtained almost identical results and also mentioned the pos- 4.3. Lattice Dynamics Effect for Ionic Transport
sibility of [110] displacement. Thus, the jump rate obtained by
spin–lattice relaxation can be considered as that for a [110] jump. The important roles of phonon dynamics have been highlighted
On the other hand, the line width data by Liang et al. clearly by Wakamura and Aniya.[111,112] The soft vibration or polarized
indicates Arrhenius-type thermal behavior (0.16 eV), which may anion framework has been recognized as a basic strategy for the
hence be considered as a [001] jump. Using the jump rate and design of superionic conductors.[113] By introducing the concept
activation energy, the attempt frequency (υ0) of each jump may of phonon band center (phonon frequency weighted by phonon
be estimated. The attempt frequency along [110] is in the range of density of states), Muy et al. quantified the lattice vibration and
1012–1013 s−1, which is in agreement with the Debye frequency and summarized the correlation between activation energy (Ea) and
is typically observed in solids.[110] On the other hand, that for the lattice softening.[114] Figure 16a summarizes the relationships of
[001] jump is very low at 109 s−1. We must emphasize, however, the Ea and phonon band center for thio-LISICON-related ionic
that the interpretation of NMR relaxometry is still not conclusive. conductors. The LGPS-type structure has a low phonon band
center (softer vibrational mode) with Ea = 0.25 eV. This result
indicates that the softness of the LGPS lattice also facilitates
high conductivity.
The importance of lattice vibration was also reported by Kahle
et al.[115] They calculated the ionic diffusion properties with a
frozen host lattice (pinball model) that was fixed at equilibrium
positions. Figure 16b shows their calculation results. The lithium
tracer-diffusion in the LGPS framework is noticeably lowered
in the pinball model compared to that with free lattice vibra-
tion. Apparently, lattice vibration has a significant impact on
lithium diffusion properties in the LGPS structure. In another
interesting calculation by Kahle et al.,[115] when the LGPS-type
structure contained the oxygen anion rather than sulfur (namely
Li10GeP2O12), the tracer-diffusion coefficient in the free lattice
and in frozen lattice almost overlapped (Figure 16c). The results
of the sulfur and oxygen models suggest that lattice vibration is
more important in the sulfur system.
The softness of the lattice can be controlled by the frame-
work cation. Zeier et al. conducted a systematic investigation of
the lattice dynamics in Li10Ge1−xSnxP2S12.[34] The lattice became
softer upon introducing the large Sn4+ into the LGPS structure
(Figure 16d). Although one expects a higher ionic conductivity
Figure 15. Temperature dependence of lithium jump rate. Adapted with in the softer Sn system, the conductivity actually decreased as
permission.[29] Copyright 2013, the Royal Society of Chemistry; Adapted the Sn content increased. This trend in ionic conductivity was
with permission.[109] Copyright 2015, the American Physical Society. also reported by Kato et al.[59] Zeier et al. demonstrated that

Adv. Energy Mater. 2020, 2002153 2002153 (18 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 16. Lattice dynamic properties of LGPS. a) Dependence of activation energy and the band center of LGPS structure (Li10SnP2S12) together with
other thio-LISICON materials. Reproduced with permission.[114] Copyright 2018, the Royal Society of Chemistry. Simulated tracer-diffusion coefficient for
Li10GeP2S12 (b) and Li10GeP2O12 (c) with free lattice and frozen lattice. Reproduced with permission.[115] Copyright 2018, the American Physical Society.
d) Measured Debye frequency for Li10(Ge1−xSn)P2S12. Reproduced with permission.[34] Copyright 2018, the American Physical Society.

the lower electronegativity of Sn compared to that of Ge and a scale for the cooperative motion of mobile ions,[118] it is nat-
the longer Sn–S bond result in a greater electron density at S2− ural to expect cooperative lithium transport in the LGPS struc-
and a stronger Coulombic attraction between Li+ and S2−, thus ture, which has HR ≈ 0.23–0.3. Molecular dynamics ­simulation
inhibiting the Li transport.[34] is a powerful tool for investigating this phenomenon. He et al.
The role of rotation or vibration of polyhedra in ionic trans- carefully investigated the nature of ionic motion in the LGPS
port was discussed for several ionic conductors (e.g., hydride structure, and reported that a single hopping event is rare there.
complex system[116] and argyrodite[117]). The interaction of PS4 Instead, multiple lithium ions at tetrahedral positions are typi-
tetrahedra with mobile lithium in LGPS was also discussed both cally cooperatively transported along [001] (Figure 17). This is
experimentally and theoretically. Liang et al. found that there is referred to as concerted migration.[119] The energy barrier for
dipolar interaction between Li(4c) and P(2b)S4.[109] Adams et al. concerted migration (Figure 17a) is ­considerably lower than that
reported a certain degree of rotational disorder in the P(2b)S4 for a classical single-ion hopping model (Figure 17b), and there-
tetrahedra.[72] In their molecular dynamics simulation, a 30° fore lithium is expected to diffuse in a cooperative manner in
flip of P(2b)S4 around [001] was observed. This rotation coin- the LGPS structure. He et al. highlighted that, due to the fewer
cided with a local lithium hop along both [001] and [110], thus low-energy sites among all lithium sites in the LGPS structure,
indicating that the lattice dynamics, particularly the rotational the mobile lithium occupies high-energy sites at tetrahedral
disorder of P(2b)S4, may play a role in ionic transportation. positions along the c-axis, which are near the highest energy
point along the diffusion path (Figure 17a). The lithium atoms
located at high-energy sites are stabilized not by interaction with
4.4. Li–Li Interaction the framework structure, but by Coulombic interaction with
neighboring mobile lithium ions instead. In other words, the
As previously discussed, the HR is an important parameter for displacement of a mobile ion affects the Coulombic repulsion
ionic conductivity in the LGPS-type structure. Since the HR is and therefore the energy landscape of its neighbor ions. Such a

Adv. Energy Mater. 2020, 2002153 2002153 (19 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Figure 17. Cooperative lithium transport properties in LGPS. Energy barriers for concerted migration (a) and single hopping (b). c) The number (n) of
Li hopping from AIMD simulations. Reproduced with permission.[28] Copyright 2017, Nature publishing group. d) Energy barriers at different coopera-
tive motions. Reproduced with permission.[120] Copyright 2016, the American Chemical Society.

unique energy landscape causes concerted migration, in which temperature within the simulation range (518–1195 K). Those
the downhill motion of one lithium in the energy landscape can- computational contributions strongly support the cooperative
cels the energy barrier of the lithium climbing up the energy motion of Li in the LGPS structure, which was also implied by
landscape. As a result, the total energy required for concerted the low HR value obtained experimentally. Nevertheless, studies
migration is significantly reduced. In addition, the possibility of of the cooperative mechanism were mainly led by computation.
cooperative motion of the [001] and [110] jumps was reported by The accumulation of more experimental data in combination
Bhandari et al.[120] Using the nudged elastic band (NEB) method, with computation on diffusion would lead to a more comprehen-
different energy barriers were observed depending on the sive understanding of this material and superionic phenomena.
neighboring situation. When a lithium at 4c attempts to jump
to the neighboring site (16h) that is occupied, then a 4c→16h
and 16h→4c cooperative jump occurs with the barrier of 0.3 eV. 4.5. Other Interesting Aspects of Ion Conduction Mechanism
However, if the 16h is unoccupied, the cooperative jump along
[001] and [110] occurs at 0.17 eV. Their NEB calculation suggests Multiple phenomena in the ionic conduction in LGPS have
that, depending on the situation, the [001] and [110] cooperative been described, and numerous interesting issues were simul-
jump is energetically preferred over the [001] jumps (Figure 17d). taneously clarified. The ionic conduction mechanism has
Marcolongo et al. calculated the HR using AIMD and obtained been elucidated through theoretical and computational efforts,
different HR for [001] (HR[001] = 0.32) and [110] (HR[001] = 0.61) but there are also some deviations from experimental data.
diffusion, which implies that both types of diffusion occur in a For example, the ionic conductivity obtained experimentally
cooperative manner. However, the diffusion is more pronounced changes trend at ≈350 K and transitions to a process with a
along [001], which is the faster conduction path.[102] Interest- very low activation energy (≈0.15 eV). The reason for this pheno­
ingly, the computed HR does not depend on the temperature, menon has not been clearly explained either theoretically or
indicating that the diffusion mechanism does not change with experimentally. One possibility is that the transport-limiting

Adv. Energy Mater. 2020, 2002153 2002153 (20 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

step shifted from [110] to [001]. However, the bending of the volumetric efficiency of the battery system. The expected advan-
Arrhenius plot at high temperatures has also been observed tages exist at all battery package levels. Even at the fundamental
in other ionic conductors. Hence, this may be a universal phe- electrochemical level, the benefits of superionic conductors
nomenon in ionic conductors. Nevertheless, it is an interesting have been proposed and experimentally confirmed. One of
subject that warrants further investigation. the important differences between solids and liquids in charge
Another interesting matter is the ionic conductivity of [001]. It transport is the transference number. In liquid electrolytes,
has been suggested that long-range diffusion occurs via the [110] both anions and cations move toward opposite sites, whereas
pathway, but the actual conductivity in the [001] direction is a in superionic conductors only mobile species (e.g., lithium for
very interesting issue because we have not been able to directly LGPS) carry the charge. This is quite favorable for electrochem-
observe such conduction, even in single-crystal impedance and ical systems because the concentration gradient that causes the
PFG-NMR experiments. In order to experimentally measure diffusion limit observed in liquid electrolytes is not expected.
the conductivity (or diffusion coefficient) in the [001] direction, Since the reaction of all-solid-state batteries is not limited by
it will be necessary to prepare single crystals as small as pos- diffusion (more precisely the chemical diffusion of lithium),
sible without any defects and construct a more precise experi- there is the potential to overcome the limit of the liquid elec-
mental system. Fortunately, the synthesis of single-crystal LGPS trolyte. The diffusion limit usually occurs at either large current
has been established, and recent advances in instrumentation drain or a thick electrode configuration. Needless to say, a large
and methods will make the aforementioned measurements current is required for fast charging in high-power applications,
possible. A number of studies have highlighted the unusual, while a thick electrode is a highly popular strategy for achieving
liquid-like conduction in the [001] direction of LGPS. The exper- high-energy-density battery cells by minimizing the volume
imental understanding of the liquid-like diffusion in solids will ratio of energetically inactive components (separator or current
be an important contribution to the development of this field. collectors). In other words, an all-solid-state battery could fur-
Additional studies on HR may also be useful for compre- ther improve both the energy and power densities.
hending the ion transport mechanism of modern superionic Kato et al. experimentally observed that a larger current den-
conductors. In order to obtain HR, the conductivity and tracer- sity could be applied to an all-solid-state battery than that using
diffusion values are required. While the conductivity of LGPS liquid electrolyte.[13] They carefully prepared solid and liquid
has been measured in a wide temperature range (150–600 K), cells using the same active materials and configurations (sur-
no single method could cover the entire range due to the limi- face area and thickness). The all-solid-state cell equipped with
tation of each method. Meanwhile, the tracer-diffusion coeffi- LGPS exhibited greater rate capabilities than the liquid cell due
cient has been investigated using AIMD and PFG-NMR. We to the absence of diffusion limit in LGPS. In addition, they have
believe that obtaining the HR over a wide temperature range demonstrated a thick electrode (600 µm cathode layer) system
will provide a better understanding of Li–Li interaction, which with LGPS.[51] The thickness of liquid electrolyte cells is usually
is an important conduction mechanism in LGPS. For example, limited to ≈100–150 µm; therefore, a 600-µm-thick electrode is
it would be essential to examine the changing behavior of HR considered extremely thick.
with respect to bending in the Arrhenius plot of ionic conduc- We emphasize that the above-mentioned overcoming of
tivity at high temperatures. liquid diffusion limit in all-solid-state batteries has only been
The grain boundary behavior is also an interesting subject. demonstrated for LGPS with the extremely high ionic con-
Bron et al. reported that the LGPS grain boundary has a com- ductivity. Although these batteries do not suffer from dif-
position-dependent conductivity,[33] which tends to decrease in fusion limit in the electrolyte, ohmic loss due to ion migra-
the order of Si > Ge > Sn. However, the reason for this trend tion in the solid-state electrolyte can be the main cause of
is not yet elucidated. The softening of the lattice may be due to overvoltage.[51,121,122] Therefore, a high ionic conductivity is
the addition of Sn or to the state of the crystal surface, but there essential for harnessing the full potential of all-solid-state bat-
are few micrometer-scale studies comparing the intracrystal- tery at the level of electrochemical reaction (particularly for
line diffusion. This kind of research on interfacial conduction is composite electrodes). For example, we propose that an ionic
very important because minimizing sintering and hot-pressing conductivity of more than 1.0 × 10−2 S cm−1 is preferable for
is desirable when we consider the applications. Since the com- demonstrating thick-film electrodes.[51] Bielefeld et al. similarly
position dependence at the homo interface is observable, the stated that the composite cathode layer would require an ionic
same could also be observed at the heterointerface between conductivity of more than 1.0 × 10−2 S cm−1 to reach moderate
active materials and the LGPS family. performance.[123] On the other hand, LGPS has been reported
to have low stabilities against metallic lithium.[13,60,124] Since
Si4+, Ge4+, and Sn4+ can be reduced to form mixed conductors,
5. Battery Applications most of LGPS materials are considered incompatible with the
metallic lithium anode. However, LGPS-type structures can
The most promising application of superionic conductors will also be formed without such metalloid constituent elements
be in all-solid-state batteries. There are several advantages of (e.g., Li9.6P3S12[13] or halogen doped Li–P–S systems[64]). These
switching organic liquid electrolytes to solid superionic conduc- materials could form a passivation layer at the interface with
tors in the modern lithium-ion batteries. 1) The temperature metallic lithium to enable charge–discharge cycling of Li metal
window for operation is wider than that for liquid electro- battery. ­Unfortunately, their reported conductivities are lower
lytes, which may lead to simplification of the cooling system. than 1.0 × 10−2 S cm−1 at RT. Nevertheless, for the usual con-
2) Bipolar battery stacking could possibly increase the figuration metallic lithium anode (2D interface between solid

Adv. Energy Mater. 2020, 2002153 2002153 (21 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

e­lectrolyte plate and metallic lithium plate), the conductivity temperature range will improve our understanding of the rela-
may not be as important as in the composite electrode system tionship between conductivity and diffusion. Moreover, a uni-
where mass transportation along thickness direction for the fied interpretation of lithium displacement data on different
electrode is required. The various options in the family of scales of time and space is desirable. Clarifying the conduction
LGPS-type materials enable researchers to select the superi- mechanism should provide significant insight into improving
onic conductors suitable for different parts in an energy storage current LGPS-type conductors, as well as discovering additional
device. This concept of applying multiple LGPS chemistries ones with new structures.
in a single cell was already reported.[13,52] Therefore, LGPS is
the most promising candidate material for use in all-solid-state
batteries, thereby extending the capabilities of liquid electrolyte Supporting Information
systems.
Supporting Information is available from the Wiley Online Library or
from the author.

6. Summary
The first report of Li10GeP2S12 in 2011 led to the development of Acknowledgements
a series of LGPS-type phases that are chemically or structurally The authors thank T. Nakamura and A. Kasama for their support in
related to Li10GeP2S12. Yet, the role of doped elements (espe- collecting and organizing data. R.K. acknowledges support by the Grant-
cially halogens) is still unclear. It is necessary to investigate in-Aid for Scientific Research (S, no. 17H06145).
the particle surface or grain-boundary morphologies. For the
­framework cations that substitute Ge or P in the original LGPS,
the candidates are still limited to the relatively redox-inactive Conflict of Interest
elements. In addition, the phase diagrams and techniques to The authors declare no conflict of interest.
grow large single crystal are available only for the original LGPS
and its solid solutions. To examine the diverse crystal chemistry
and further improve superionic conduction, elucidating the
phase–structure–composition relationships in the previously Keywords
unexplored chemical systems should be elucidated. LGPS, Li10GeP2S12, solid electrolytes, superionic conductors
Single-crystal growth techniques have enabled precise anal-
ysis of the LGPS crystal structure. The bcc-like sulfur lattice is Received: June 30, 2020
presumably favorable for a wide distribution of Li ions along Revised: September 4, 2020
Published online:
the c-axis in LGPS, which has been proven by the large ADP
values for Li1 (16h) and Li3 (16h) sites, and visually verified
using the MEM techniques.
Structural investigation of the Li distribution and anisotropic [1] S. Hull, Rep. Prog. Phys. 2004, 67, 1233.
ADPs for Li sites in LGPS as well as NMR relaxometry and [2] K. Funke, Sci. Technol. Adv. Mater. 2013, 14, 043502.
AIMD studies indicate that the primary path for lithium con- [3] O. Yamamoto, Sci. Technol. Adv. Mater. 2017, 18, 504.
duction is along the c-axis ([001] direction), whereas the pathway [4] B. B. Owens, G. R. Argue, Science 1967, 157, 308.
over the ab-plane ([110] direction) is the secondary path at the [5] T. Takahashi, O. Yamamoto, S. Yamada, S. Hayashi, J. Electrochem.
microscopic level around several unit cells. In contrast, long- Soc. 1979, 126, 1654.
range diffusion analysis on the macroscopic scale (PFG-NMR [6] Y. Yung-Fang Yu, J. T. Kummer, J. Inorg. Nucl. Chem. 1967, 29, 2453.
and single-crystal impedance) strongly indicates that the [110] [7] H. Y.-P. Hong, Mater. Res. Bull. 1978, 13, 117.
[8] T. Lapp, S. Skaarup, A. Hooper, Solid State Ionics 1983, 11, 97.
jump is always involved in long-range (>0.5 µm) diffusions.
[9] C. A. Vincent, Prog. Solid State Chem. 1987, 17, 145.
This contrast between microscopic and macroscopic results
[10] A. R. Kulkarni, H. S. Maiti, A. Paul, Bull. Mater. Sci. 1984, 6, 201.
indicates that the [110] pathways plays an important role in the [11] A. Pradel, M. Ribes, Mater. Chem. Phys. 1989, 23, 121.
practical lithium conduction phenomena. Further studies are [12] N. Kamaya, K. Homma, Y. Yamakawa, M. Hirayama, R. Kanno,
desired to distinguish the contribution from all pathways and M. Yonemura, T. Kamiyama, Y. Kato, S. Hama, K. Kawamoto,
thereby rationalize the highly conductive nature of LGPS. A. Mitsui, Nat. Mater. 2011, 10, 682.
Various studies have revealed that the high conductivity of [13] Y. Kato, S. Hori, T. Saito, K. Suzuki, M. Hirayama, A. Mitsui,
LGPS is caused by the combination of complex phenomena M. Yonemura, H. Iba, R. Kanno, Nat. Energy 2016, 1, 16030.
such as lattice vibrations and Li–Li interactions, in addition [14] S. Randau, D. A. Weber, O. Kötz, R. Koerver, P. Braun, A. Weber,
to favorable energy landscape in a unique framework. To E. Ivers-Tiffée, T. Adermann, J. Kulisch, W. G. Zeier, F. H. Richter,
J. Janek, Nat. Energy 2020, 5, 259.
further understand the physics behind superionic conduc-
[15] T. Famprikis, P. Canepa, J. A. Dawson, M. S. Islam, C. Masquelier,
tion, it is preferable to obtain more precise crystal structure
Nat. Mater. 2019, 18, 1278.
information and diffusion parameters. Furthermore, experi- [16] S. Xia, X. Wu, Z. Zhang, Y. Cui, W. Liu, Chem 2019, 5, 753.
mental measurements must not neglect the influence of grain [17] K. H. Park, Q. Bai, D. H. Kim, D. Y. Oh, Y. Zhu, Y. Mo, Y. S. Jung,
boundaries. Therefore, systems that contain no or little grain Adv. Energy Mater. 2018, 8, 1800035.
boundaries, such as single crystals, are necessary. Diffusion [18] Y. Xiao, Y. Wang, S.-H. Bo, J. C. Kim, L. J. Miara, G. Ceder, Nat. Rev.
coefficient measurements and AIMD simulations over a wider Mater. 2020, 5, 105.

Adv. Energy Mater. 2020, 2002153 2002153 (22 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

[19] R. Kanno, T. Hata, Y. Kawamoto, M. Irie, Solid State Ionics 2000, [53] S. T. Kong, Ö. Gün, B. Koch, H. J. Deiseroth, H. Eckert, C. Reiner,
130, 97. Chem. - Eur. J. 2010, 16, 5138.
[20] M. Murayama, R. Kanno, Y. Kawamoto, T. Kamiyama, Solid State [54] K. Homma, M. Yonemura, T. Kobayashi, M. Nagao, M. Hirayama,
Ionics 2002, 154–155, 789. R. Kanno, Solid State Ionics 2011, 182, 53.
[21] M. Murayama, R. Kanno, M. Irie, S. Ito, T. Hata, N. Sonoyama, [55] K. Homma, M. Yonemura, M. Nagao, M. Hirayama, R. Kanno,
Y. Kawamoto, J. Solid State Chem. 2002, 168, 140. J. Phys. Soc. Jpn. 2010, 79, 90.
[22] R. Kanno, M. Murayama, J. Electrochem. Soc. 2001, 148, A742. [56] R. Mercier, J. P. Malugani, B. Fahys, J. Douglande, G. Robert,
[23] M. Murayama, N. Sonoyama, A. Yamada, R. Kanno, Solid State J. Solid State Chem. 1982, 43, 151.
Ionics 2004, 170, 173. [57] O. Kwon, M. Hirayama, K. Suzuki, Y. Kato, T. Saito,
[24] J. H. Kennedy, Y. Yang, J. Electrochem. Soc. 1986, 133, 2437. M. Yonemura, T. Kamiyama, R. Kanno, J. Mater. Chem. A 2015, 3,
[25] R. Mercier, J.-P. Malugani, B. Fahys, G. Robert, Solid State Ionics 438.
1981, 5, 663. [58] A. Sakuda, A. Hayashi, S. Hama, M. Tatsumisago, J. Am. Ceram.
[26] T. Inada, T. Kobayashi, N. Sonoyama, A. Yamada, S. Kondo, Soc. 2010, 93, 765.
M. Nagao, R. Kanno, J. Power Sources 2009, 194, 1085. [59] Y. Kato, R. Saito, M. Sakano, A. Mitsui, M. Hirayama, R. Kanno,
[27] T. Kobayashi, A. Yamada, R. Kanno, Electrochim. Acta 2008, 53, J. Power Sources 2014, 271, 60.
5045. [60] P. Bron, B. Roling, S. Dehnen, J. Power Sources 2017, 352, 127.
[28] P. Bron, S. Johansson, K. Zick, J. Schmedt auf der Günne, [61] X. Feng, P.-H. Chien, Z. Zhu, I.-H. Chu, P. Wang,
S. Dehnen, B. Roling, J. Am. Chem. Soc. 2013, 135, 15694. M. Immediato-Scuotto, H. Arabzadeh, S. P. Ong, Y.-Y. Hu,
[29] A. Kuhn, V. Duppel, B. V. Lotsch, Energ. Environ. Sci. 2013, 6, 3548. Adv. Funct. Mater. 2019, 29, 1807951.
[30] A. Kuhn, O. Gerbig, C. Zhu, F. Falkenberg, J. Maier, B. V. Lotsch, [62] A. Hayashi, S. Hama, T. Minami, M. Tatsumisago, Electrochem.
Phys. Chem. Chem. Phys. 2014, 16, 14669. Commun. 2003, 5, 111.
[31] S. Hori, K. Suzuki, M. Hirayama, Y. Kato, T. Saito, M. Yonemura, [63] Y. Ooura, N. Machida, T. Uehara, S. Kinoshita, M. Naito,
R. Kanno, Faraday Discuss. 2014, 176, 83. T. Shigematsu, S. Kondo, Solid State Ionics 2014, 262, 733.
[32] D. A. Weber, A. Senyshyn, K. S. Weldert, S. Wenzel, W. Zhang, [64] S. Ujiie, A. Hayashi, M. Tatsumisago, J. Solid State Electrochem.
R. Kaiser, S. Berendts, J. Janek, W. G. Zeier, Chem. Mater. 2016, 28, 2013, 17, 675.
5905. [65] K. Suzuki, M. Sakuma, S. Hori, T. Nakazawa, M. Nagao,
[33] P. Bron, S. Dehnen, B. Roling, J. Power Sources 2016, 329, 530. M. Yonemura, M. Hirayama, R. Kanno, Solid State Ionics 2016, 288,
[34] T. Krauskopf, S. P. Culver, W. G. Zeier, Chem. Mater. 2018, 30, 1791. 229.
[35] Y. Sun, K. Suzuki, S. Hori, M. Hirayama, R. Kanno, Chem. Mater. [66] A. Neveu, V. Pelé, C. Jordy, V. Pralong, J. Power Sources 2020, 467,
2017, 29, 5858. 228250.
[36] J. Liang, N. Chen, X. Li, X. Li, K. R. Adair, J. Li, C. Wang, C. Yu, [67] S. Daikuhara, H. Satoshi, K. Suzuki, M. Hirayama, R. Kanno,
M. Norouzi Banis, L. Zhang, S. Zhao, S. Lu, H. Huang, R. Li, Abstract of the 43rd Solid State Ionics Japan, Yamagata, Japan,
Y. Huang, X. Sun, Chem. Mater. 2020, 32, 2664. December 2017.
[37] S. Hori, K. Suzuki, M. Hirayama, Y. Kato, R. Kanno, Front. Energy [68] K. Yang, J. Dong, L. Zhang, Y. Li, L. Wang, J. Am. Ceram. Soc. 2015,
Res. 2016, 4, 38, https://doi.org/10.3389/fenrg.2016.00038. 98, 3831.
[38] Y. Sun, K. Suzuki, K. Hara, S. Hori, T.-A. Yano, M. Hara, [69] H. Yamane, M. Shibata, Y. Shimane, T. Junke, Y. Seino, S. Adams,
M. Hirayama, R. Kanno, J. Power Sources 2016, 324, 798. K. Minami, A. Hayashi, M. Tatsumisago, Solid State Ionics 2007,
[39] K.-H. Kim, S. W. Martin, Chem. Mater. 2019, 31, 3984. 178, 1163.
[40] K. Takada, M. Osada, N. Ohta, T. Inada, A. Kajiyama, H. Sasaki, [70] H.-J. Deiseroth, S.-T. Kong, H. Eckert, J. Vannahme, C. Reiner,
S. Kondo, M. Watanabe, T. Sasaki, Solid State Ionics 2005, 176, 2355. T. Zaiß, M. Schlosser, Angew. Chem., Int. Ed. 2008, 47, 755.
[41] W. D. Richards, T. Tsujimura, L. J. Miara, Y. Wang, J. C. Kim, [71] V. Favre-Nicolin, R. Cerny, J. Appl. Crystallogr. 2002, 35, 734.
S. P. Ong, I. Uechi, N. Suzuki, G. Ceder, Nat. Commun. 2016, 7, [72] S. Adams, R. Prasada Rao, J. Mater. Chem. 2012, 22, 7687.
11009. [73] Y. Deng, C. Eames, J.-N. Chotard, F. Lalère, V. Seznec, S. Emge,
[42] Z. Zhang, E. Ramos, F. Lalère, A. Assoud, K. Kaup, P. Hartman, O. Pecher, C. P. Grey, C. Masquelier, M. S. Islam, J. Am. Chem. Soc.
L. F. Nazar, Energ. Environ. Sci. 2018, 11, 87. 2015, 137, 9136.
[43] M. Duchardt, U. Ruschewitz, S. Adams, S. Dehnen, B. Roling, [74] D. Tranqui, R. D. Shannon, H.-Y. Chen, S. Iijima, W. H. Baur, Acta
Angew. Chem., Int. Ed. 2018, 57, 1351. Crystallogr. B 1979, 35, 2479.
[44] A. Kuhn, J. Köhler, B. V. Lotsch, Phys. Chem. Chem. Phys. 2013, 15, [75] F. Du, X. Ren, J. Yang, J. Liu, W. Zhang, J. Phys. Chem. C 2014, 118,
11620. 10590.
[45] R. Iwasaki, S. Hori, R. Kanno, T. Yajima, D. Hirai, Y. Kato, Z. Hiroi, [76] Y. Mo, S. P. Ong, G. Ceder, Chem. Mater. 2012, 24, 15.
Chem. Mater. 2019, 31, 3694. [77] A. R. West, Z. Kristallogr. 1975, 141, 422.
[46] N. Machida, S. Hashimoto, C. Kinoshita, presented at Spring [78] W. H. Baur, T. J. McLarnan, J. Solid State Chem. 1982, 42, 300.
Meeting of Japan Society of Powder and Powder Metallurgy, Japan, [79] A. R. West, P. G. Bruce, Acta Crystallogr. B 1982, 38, 1891.
June 2019. [80] W. H. Baur, T. Ohta, J. Solid State Chem. 1982, 44, 50.
[47] Y. Wang, Z. Liu, X. Zhu, Y. Tang, F. Huang, J. Power Sources 2013, [81] R. D. Shannon, Acta Crystallogr. A 1976, 32, 751.
224, 225. [82] Y. Wang, W. D. Richards, S. P. Ong, L. J. Miara, J. C. Kim, Y. Mo,
[48] H. Tsukasaki, S. Mori, S. Shiotani, H. Yamamura, H. Iba, J. Power G. Ceder, Nat. Mater. 2015, 14, 1026.
Sources 2017, 369, 57. [83] T. Kudo, K. Hueki, Solid State Ionics, Kodansha, Tokyo, Japan 1990.
[49] S. Hori, M. Kato, K. Suzuki, M. Hirayama, Y. Kato, R. Kanno, [84] M. Inagaki, K. Suzuki, S. Hori, K. Yoshino, N. Matsui,
J. Am. Ceram. Soc. 2015, 98, 3352. M. Yonemura, M. Hirayama, R. Kanno, Chem. Mater. 2019, 31,
[50] Y. Kato, Dissertation, Tokyo Institute of Technology, Tokyo 2014. 3485.
[51] Y. Kato, S. Shiotani, K. Morita, K. Suzuki, M. Hirayama, R. Kanno, [85] S. Hori, S. Taminato, K. Suzuki, M. Hirayama, Y. Kato, R. Kanno,
J. Phys. Chem. Lett. 2018, 9, 607. Acta Crystallogr. B 2015, 71, 727.
[52] B. R. Shin, Y. J. Nam, D. Y. Oh, D. H. Kim, J. W. Kim, Y. S. Jung, [86] R. Metselaar, G. Oversluizen, J. Solid State Chem. 1984, 55, 320.
Electrochim. Acta 2014, 146, 395. [87] A. S. Nowick, W. K. Lee, H. Jain, Solid State Ionics 1988, 28–30, 89.

Adv. Energy Mater. 2020, 2002153 2002153 (23 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

[88] K. L. Ngai, Solid State Ionics 1998, 105, 231. [105] P. Adeli, J. D. Bazak, K. H. Park, I. Kochetkov, A. Huq,
[89] S. Muy, J. C. Bachman, H.-H. Chang, L. Giordano, F. Maglia, G. R. Goward, L. F. Nazar, Angew. Chem., Int. Ed. 2019, 58, 8681.
S. Lupart, P. Lamp, W. G. Zeier, Y. Shao-Horn, Chem. Mater. 2018, [106] S. Kim, H. Oguchi, N. Toyama, T. Sato, S. Takagi, T. Otomo,
30, 5573. D. Arunkumar, N. Kuwata, J. Kawamura, S. Orimo, Nat. Commun.
[90] M. A. Kraft, S. P. Culver, M. Calderon, F. Bocher, T. Krauskopf, 2019, 10, 1081.
A. Senyshyn, C. Dietrich, A. Zevalkink, J. Janek, W. G. Zeier, J. Am. [107] M. Pinsky, D. Avnir, Inorg. Chem. 1998, 37, 5575.
Chem. Soc. 2017, 139, 10909. [108] R. Malik, D. Burch, M. Bazant, G. Ceder, Nano Lett. 2010, 10, 4123.
[91] M. A. Kraft, S. Ohno, T. Zinkevich, R. Koerver, S. P. Culver, [109] X. Liang, L. Wang, Y. Jiang, J. Wang, H. Luo, C. Liu, J. Feng, Chem.
T. Fuchs, A. Senyshyn, S. Indris, B. J. Morgan, W. G. Zeier, J. Am. Mater. 2015, 27, 5503.
Chem. Soc. 2018, 140, 16330. [110] H. Mehrer, Diffusion in Solids, Springer, New York 2016.
[92] L. Zhou, A. Assoud, Q. Zhang, X. Wu, L. F. Nazar, J. Am. Chem. [111] M. Aniya, K. Wakamura, Physica B Condens. Matter. 1996, 219–220,
Soc. 2019, 141, 19002. 463.
[93] S. Ohta, T. Kobayashi, T. Asaoka, J. Power Sources 2011, 196, 3342. [112] K. Wakamura, Phys. Rev. B 1997, 56, 11593.
[94] J. L. Allen, J. Wolfenstine, E. Rangasamy, J. Sakamoto, J. Power [113] J. C. Bachman, S. Muy, A. Grimaud, H.-H. Chang, N. Pour,
Sources 2012, 206, 315. S. F. Lux, O. Paschos, F. Maglia, S. Lupart, P. Lamp, L. Giordano,
[95] D. Rettenwander, G. Redhammer, F. Preishuber-Pflugl, L. Cheng, Y. Shao-Horn, Chem. Rev. 2016, 116, 140.
L. Miara, R. Wagner, A. Welzl, E. Suard, M. M. Doeff, M. Wilkening, [114] S. Muy, J. C. Bachman, L. Giordano, H.-H. Chang, D. L. Abernathy,
J. Fleig, G. Amthauer, Chem. Mater. 2016, 28, 2384. D. Bansal, O. Delaire, S. Hori, R. Kanno, F. Maglia, S. Lupart,
[96] L. Zhou, A. Assoud, A. Shyamsunder, A. Huq, Q. Zhang, P. Lamp, Y. Shao-Horn, Energ. Environ. Sci. 2018, 11, 850.
P. Hartmann, J. Kulisch, L. F. Nazar, Chem. Mater. 2019, 31, [115] L. Kahle, A. Marcolongo, N. Marzari, Phys. Rev. Mater. 2018, 2,
7801. 065405.
[97] S. Takai, T. Mandai, Y. Kawabata, T. Esaka, Solid State Ionics 2005, [116] K. E. Kweon, J. B. Varley, P. Shea, N. Adelstein, P. Mehta, T. W. Heo,
176, 2227. T. J. Udovic, V. Stavila, B. C. Wood, Chem. Mater. 2017, 29, 9142.
[98] S. Takai, K. Kurihara, K. Yoneda, S. Fujine, Y. Kawabata, T. Esaka, [117] I. Hanghofer, B. Gadermaier, H. M. R. Wilkening, Chem. Mater.
Solid State Ionics 2004, 171, 107. 2019, 31, 4591.
[99] K. Hayamizu, Y. Matsuda, M. Matsui, N. Imanishi, Solid State [118] Á. W. Imre, H. Staesche, S. Voss, M. D. Ingram, K. Funke,
Nucl. Magn. Reson. 2015, 70, 21. H. Mehrer, J. Phys. Chem. B 2007, 111, 5301.
[100] M. T. Chowdhury, R. Takekawa, Y. Iwai, N. Kuwata, J. Kawamura, [119] X. He, Y. Zhu, Y. Mo, Nat. Commun. 2017, 8, 15893.
J. Chem. Phys. 2014, 140, 124509. [120] A. Bhandari, J. Bhattacharya, J. Phys. Chem. C 2016, 120,
[101] A. Dorai, N. Kuwata, R. Takekawa, J. Kawamura, K. Kataoka, 29002.
J. Akimoto, Solid State Ionics 2018, 327, 18. [121] P. Braun, C. Uhlmann, M. Weiss, A. Weber, E. Ivers-Tiffée, J. Power
[102] A. Marcolongo, N. Marzari, Phys. Rev. Mater. 2017, 1, 025402. Sources 2018, 393, 119.
[103] D. Di Stefano, A. Miglio, K. Robeyns, Y. Filinchuk, M. Lechartier, [122] Y. Kato, K. Kawamoto, M. Hirayama, R. Kanno, Electrochemistry
A. Senyshyn, H. Ishida, S. Spannenberger, D. Prutsch, 2012, 80, 749.
S. Lunghammer, D. Rettenwander, M. Wilkening, B. Roling, [123] A. Bielefeld, D. A. Weber, J. Janek, ACS Appl. Mater. Interfaces 2020,
Y. Kato, G. Hautier, Chem 2019, 5, 2450. 12, 12821.
[104] S. P. Ong, Y. Mo, W. D. Richards, L. Miara, H. S. Lee, G. Ceder, [124] S. Wenzel, S. Randau, T. Leichtweiß, D. A. Weber, J. Sann,
Energ. Environ. Sci. 2013, 6, 148. W. G. Zeier, J. Janek, Chem. Mater. 2016, 28, 2400.

Yuki Kato received his B.S. degree from Kyushu University in 2008 and the Doctor of Engineering
degree from Tokyo Institute of Technology in 2014. He worked at Toyota Motor Corporation and
Toyota Motor Europe NV/SA from 2008 to 2019, where he researched all-solid-state batteries.
Currently, he works at an energy related company in the United States.

Adv. Energy Mater. 2020, 2002153 2002153 (24 of 25) © 2020 Wiley-VCH GmbH
www.advancedsciencenews.com www.advenergymat.de

Satoshi Hori received his B.S. degree from the University of Tokyo and Ph.D. degree from Tokyo
Institute of Technology (2016). He continued his study on solid-state batteries as a JSPS research
fellow and then as a project assistant professor. He is interested in synthesizing lithium conduc-
tive sulfides for solid-state battery.

Ryoji Kanno is the unit leader at All-Solid-State Battery Unit and a professor at Institute of
Innovative Research, Tokyo Institute of Technology. Since 1980, he has been investigating mate-
rials for electrochemical energy conversion devices, particularly lithium batteries and solid-state
batteries. He has developed electrode materials for lithium batteries and introduced physico-
chemical methods to understand the reaction mechanisms during electrochemical processes. He
is currently developing solid-electrolyte materials and all-solid-state batteries for future systems.

Adv. Energy Mater. 2020, 2002153 2002153 (25 of 25) © 2020 Wiley-VCH GmbH

You might also like