Available online at www.sciencedirect.com

ScienceDirect

J Materiomics 1 (2015) 325e332www.ceramsoc.com/en/ www.journals.elsevier.com/journal-of-materiomics/

Candidate structures for inorganic lithium solid-state electrolytes identifiedby high-throughput bond-valence calculations

Ruijuan Xiao*, Hong Li, Liquan Chen

Beijing National Laboratory for Condensed Matter Physics, Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China

Received 29 May 2015; revised 8 July 2015; accepted 3 August 2015

Available online 22 August 2015

Abstract

Looking for fast lithium ion conductors as solid state electrolytes is of great significance to achieve better safety for next generation lithiumbatteries. As an important prerequisite for developing all-solid-state lithium secondary batteries, the materials with high lithium ionic con-ductivity and inhibited electronic conductivity must be found. By implementing the bond-valence code and the automation simulation flow, weperform the high-throughput bond-valence calculations to screen fast lithium ion conductors from more than 1000 lithium-contained compoundsin ICSD database. The candidate structures are identified and their kinetic properties as well as electronic structures are analyzed through bond-valence method and density functional theory calculations, respectively. The promising structures are selected to be further optimized in thefuture.© 2015 The Chinese Ceramic Society. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: All-solid-state lithium battery; Solid state electrolyte; High-throughput calculation; Bond-valence; Density functional theory

1. Introduction

Lithium rechargeable batteries are vital technology in to-day's modern society, and they have become widespread day byday in portable electronic devices [1], hybrid and electric ve-hicles [2], and large-scale energy storage systems for wind orsolar energy [3]. Looking for solid state electrolytes with highlithium ionic conductivity is an important precondition fordeveloping all-solid-state lithium secondary batteries [4].Because of the stability and non-flammability of inorganicsolid electrolytes, the all-solid-state batteries are expected toexhibit less side reactions and higher safety, which are prom-ising solutions for problems of leakage, vaporization, decom-position and safety of currently used lithium ionic batteriescontaining liquid electrolytes [5]. One of the most importanttechnology for solid-state batteries is the discovery of solid

* Corresponding author.

E-mail address: rjxiao@iphy.ac.cn (R. Xiao).

Peer review under responsibility of The Chinese Ceramic Society.

http://dx.doi.org/10.1016/j.jmat.2015.08.001

2352-8478/© 2015 The Chinese Ceramic Society. Production and hosting by Elsevi

creativecommons.org/licenses/by-nc-nd/4.0/).

electrolytes with good comprehensive performance, includinghigh total ionic conductivity, wide electrochemical window,stable electrode/electrolyte interface, inhibited electronicconductivity, etc. [6] Among all the conditions, the highlithium ionic conductivity and the high electrical resistance areessential prerequisites because the former reduces the internalresistance of the battery and the later minimizes the self-discharge rate of the system. The experimental search of newlithium ionic conductor has been ongoing nearly half century.The fast lithium ion conduction is mainly found in glassy- andcrystalline-sulfides [7,8], NASICON-, perovskite-, garnet- andg-Li3PO4 type compounds [9e12], as well as Li3N- and LiI-related structures [13,14]. To date, the highest lithium ionicconductivity in solid materials at room temperature is up to12 mS/cm, exhibited in Li10GeP2S12 and comparative to thevalue of liquid organic electrolytes [15], however, the airinstability of sulfides limits the future application of this ma-terial. The experimental search involves numerous combina-tions of elements, compositions and crystal structures make thediscovery of new solid-state electrolytes an arduous task.

er B.V. This is an open access article under the CC BY-NC-ND license (http://

326 R. Xiao et al. / J Materiomics 1 (2015) 325e332

Nowadays, as the development of efficient theoreticalmethods and inexpensive computers, the high-throughputtheoretical calculations have played an increasingly impor-tant role in the discovery of new materials [16]. For the re-quirements needed by solid-state electrolytes, the electricalresistance can be estimated by the electronic structure calcu-lations [17], and the inhibited electronic conductivity is ex-pected to appear in wide bandgap compounds. While thesimulation of migration process of lithium ions is muchdifficult since it is a kinetic process. The lithium ionic con-ductivity is closely related with the diffusion energy barrierwhich measures the minimum energy that must be input to Liþ

to complete a hop from one site to another in the structure. Thecalculations of energy barriers with high accuracy can beobtained from the transition-state method, like the nudgedelastic band (NEB) method, or the molecular dynamic methodbased on density function theory (DFT) [17e19]. Thesemethods are reliable, and besides the energy barrier, they areable to reveal the diffusion mechanism and dominant type ofcarriers [20]. The disadvantages pointed out are the hugecomputational cost, which limit the integration of thesemethods into high-throughput simulation process. Thus, veryfew reports have involved the high-throughput density func-tional theory calculations in the screening or optimization ofsolid state electrolytes. So far the only work is carried out byFujimura et al., who carried out systematic sets of first-principles molecular dynamics simulations on LISICON-typeelectrolytes to optimize the ionic conductivities in this struc-ture [21]. An alternative scheme is to adopt the simulationtechniques in different levels of accuracy at various screeningstages according to the distinctive features of each method.For example, the fast bond-valence (BV) technique is suitablefor high-throughput pre-screening a wide range of compoundssince the trend in the ability of ion motion can be drawn fromthe relative values of the migration energy barriers despite oftheir less accuracy compared with quantum mechanical sim-ulations [22]. While the time-consuming DFT method can beadopted to do more precise calculations only for thosepromising candidates assigned by the BV method. For thederivative structures achieved by substitution or doping theexisting compounds, the DFT computation is a powerful toolto predict exact structures which are important information forperforming BV calculations. By combining the BV methodand the DFT calculations, we proposed a high-throughputscreening and optimization scheme to search fast lithium ionconductors as candidate solid state electrolytes [23]. In thiswork, the candidate structures for fast lithium ion conductorsscreened from the inorganic crystal structure database (ICSD)[24] by high-throughput bond-valence calculations are intro-duced and the electronic insulating properties are evaluated forsome typical structures by DFT band structure calculations.

2. Computational details

To perform the bond-valence calculations, we implementthe computer code, BVpath, to calculate the migration energybarriers and pathways of lithium ions in inorganic crystals

using the BV-based force-field approach proposed by Adamset al. [25]. In this method, the lithium ions are assumed tomove within the framework composed of immobile anions andcations. Thus, the interactions between a dummy lithium ionin the space and the ions around it are empirically evaluated bythe summation of the attraction between Liþ and anions de-scried by Morse-type potential and the repulsion between Liþ

and cations expressed by Coulombic potential as follows [25],

EðLiÞMorse ¼ D0

�ðexp½aðRmin �RÞ� � 1Þ2 � 1� ð1Þ

EðLi�AÞCoulomb ¼qLiqARLi�A

erfc

�RLi�A

rLi�A

�ð2Þ

in which, D0, a, Rmin are Morse potential parameters deter-mined from a large amount of stable compounds, q and Rrefers to the charge and Li-A distance, respectively. Thesummation of above two terms on grids of points with reso-lution of 0.1 Å builds the maps of the total potential energy,E(Li), in a unit cell. Other kinds of empirical potential energyfunctions may selected for this calculations, while only Morse-type and Coulombic potentials are considered here becausetheir extensive use have confirmed their effectiveness in a lotof systems [25]. The regions enclosed by isosurfaces of con-stant E(Li) are considered as the space where lithium ions cango through, and the threshold value of E(Li) to form acontinuous pathway is estimated as the Liþ migration energybarrier. In the calculations, the cutoff radius of 10 Å is takenfor the interaction models. Although this method is limited inrevealing the diffusion mechanisms in materials, it is effectiveto be used in the initial screening because it takes shortcomputational time and it is easy to be extended to broadrange of materials.

The BVpath code is written by Fortran and the calculationtime for a single structure is a few minutes. To carry out thehigh-throughput bond-valence calculations for screening of alarge structure database, we develop an automatic flow con-sisting of a set of scripts written in the Python programminglanguage. The primary role of these scripts is to initiate thecalculations, including generating the input files from thestructures in the database, specifying the necessary parame-ters, submitting the simulation tasks, calling the BVpath code,and detecting the errors and returning the message with min-imal human intervention during the whole process. With thishigh-throughput calculations based on BV methods, more than1000 lithium-contained oxides from ICSD database are stud-ied and their migration energy barriers and pathways are ob-tained. The effectiveness of the high-throughput bond-valencecalculations and the ability to predict reliable tendency of theLiþ migration energy barriers have been confirmed inRef. [23] by comparing with the results from DFTcalculations.

For the compounds with low migration energy barriersscreened from bond-valence calculations, the band structuresare simulated by density functional theory with the Vienna abinitio simulation package (VASP) [26] using the generalizedgradient approximation (GGA) with a parameterized

327R. Xiao et al. / J Materiomics 1 (2015) 325e332

exchange-correlation functional according to PBE [27]. In thefirst step, the structural relaxation is performed with the k-mesh in the density of one point per 0.06 �3. The cutoff forthe wave function is specified as 30% larger than the maximalcutoff value among all elements involved in the compound.Based on the relaxed geometries, the band structures arecalculated along the high-symmetry k-path in the Brillouinzones standardized by Setyawan and Curtarolo according tothe Bravais lattices [28].

3. Results and discussion

Table 1 lists the lithium migration energy barriers, Ea, andthe dimension of pathways for part of the oxides obtained bythe high-throughput bond-valence calculations. The com-pounds listed from line 1 to 10 show the lowest energy barriersamong all the calculated structures, and the lithium migrationpathways of those with Ea less than 0.5 eV are illustrated inFig. 1. The one dimensional channel with very low Ea value of0.25 eV is found in the orthorhombic LiB3O5, which wasconsidered as a nonlinear optical crystal [29] and the pyro-electric, dielectric, and piezoelectric properties of this materialhave been widely investigated [30]. It is hardly to find anyreports about the ionic conductivity of this compound exceptthe work from Radaev et al., in which the atomic structure isidentified by X-ray diffraction and the one-dimensional ionicconductivity is deduced [31]. As shown in Fig. 1(a), the

Table 1

Some candidate structures for fast lithium ion conductors screened from ICSD

database by high-throughput bond-valence calculations.

No. Compound

name

Lattice system Space

group

Ea by

BV [eV]

Path-way ICSD

code

1 LiB3O5 Orthorhombic P n a 21 0.25 1D 415199

2 Li2O Rhombohedral R �3 m:h 0.25 3D 108886

3 Li2Te2O5 Monoclinic P1 21/n 1 0.39 1D 026451

4 LiB(C2O4)2 Orthorhombic P n m a 0.40 1D 281623

5 Li2Si2O5 Orthorhombic P b c n 0.47 1D 067110

6 Li2SO4 Monoclinic P1 21/c 1 0.48 2D 002512

7 LiAlSiO4 Hexagonal P 62 2 2 0.49 1D 032595

8 Li5AlO4 Orthorhombic P m m n 0.49 2D 001037

9 Li6Si2O7 Tetragonal P �4 21 m 0.49 3D 025752

10 Li4CO4 Monoclinic C 1 m 1 0.50 2D 245392

11 LiScP2O7 Monoclinic P 1 21 1 0.55 1D 091496

12 LiInP2O7 Monoclinic P 1 21 1 0.73 1D 060935

13 Li2CaGeO4 Tetragonal I �4 2 m 0.58 1D 019024

14 Li2CaSiO4 Tetragonal I �4 2 m 0.59 1D 019023

15 Li2ZnSiO4 Monoclinic P 1 21/n 1 0.9 3D 008237

16 LiScGeO4 Orthorhombic P n m a 0.85 1D 062481

17 LiInGeO4 Orthorhombic P n m a 0.87 1D 062229

18 LiZnPO4 Orthorhombic P n 21 a 0.87 1D 079352

19 LiMgPO4 Orthorhombic P n m a 0.88 1D 201138

20 LiInSiO4 Orthorhombic P b n m 0.97 1D 281318

21 LiScSiO4 Orthorhombic P b n m 0.99 1D 077544

22 Li2GeO3 Orthorhombic C m c 21 0.55 1D 100403

23 Li2SiO3 Orthorhombic C m c 21 0.58 1D 100402

24 Li2SnO3 Monoclinic C 1 2/c 1 0.76 3D 021032

25 Li3Sc2(PO4)3 Monoclinic P 1 1 21/n 0.76 2D 086457

26 Li3In2(PO4)3 Monoclinic P 1 21/n 1 0.83 2D 060948

27 LiGe2(PO4)3 Triclinic R -3 c:h 0.88 3D 069763

28 LiZr2(PO4)3 Monoclinic P 1 21/n 1 0.89 2D 091112

lithium ions can diffuse along c axis within the space aroundby BeO network. The electronic structure of LiB3O5 has beencomputed from first principles method and a direct bandgap of7.37 eV is characterized by Xu and Ching [32]. The widebandgap is expected to ensure the broad electrochemicalwindow of this material. Thus, it is worth to further study thelithium ionic conductivity of this material in the future. Therhombohedral Li2O also shows low Ea value of 0.25 eV, and itis one of the component units in the fast lithium ion con-ducting glass [33], and may also act as the buffer layer be-tween solid electrolyte and electrodes. One report on ionicconductivity has been found for a-Li2Te2O5 with monocliniclattice and b-Li2Te2O5 with orthorhombic structure (not listedin the table because of the larger Ea value), and the formershows higher ionic conductivity than the later one [34]. Theexperimentally measured energy barrier of a-Li2Te2O5 re-ported in Ref. [34] is higher than the prediction by BVmethod, and no other experimental or theoretical reports canbe found for comparison, therefore, more detailed work isnecessary to understand the kinetic properties of this com-pound. LiB(C2O4)2, abbreviated as LiBOB, is an electrolytesalt already used in the lithium ion batteries [35]. Some worksare focused on the diffusion properties of lithium-silicate glass[36], and a theoretical simulation has been performed oncrystalline Li2Si2O5 [37]. Li2SO4 has been confirmed as both alithium-ion conductor and a proton conductor, and used asaqueous electrolyte in lithium ion batteries [38]. Both theglassy and crystalline state of LiAlSiO4 have been studiedexperimentally and the glassy state shows a little lower energybarrier than that in crystalline phase [39]. b-Li5AlO4 inorthorhombic lattice is known as humidity sensor and CO2

captor [40,41], thus the application of this material in solidstate batteries may be limited because of the complex re-actions between the air and itself. However, the ionic con-ductivity up to 10�3 S/cm at 300 �C in substituted Li5MO4

phase (M ¼ Al, Ga, Fe) can be obtained [42,43], which pro-vide opportunities to optimize the material. Among the isomerof Li5AlO4, a-phase is more stable in low-temperature butshows higher Ea of 0.75 eV in three-dimensional paths, whichis also a candidate structure for fast lithium ion conductors ifsuitable optimization scheme can be found. We carry out theDFT calculations for a-Li5AlO4 and b-Li5AlO4, and the bandstructures are given in Fig. 2. Both of these two structuresbelong to orthorhombic lattice, thus the same high-symmetryk-path is selected for them in the band structure calcula-tions. The calculated bandgap is 4.65 eV and 4.79 eV for a-and b-Li5AlO4, respectively, providing guarantee for inhibitedelectronic conductivity and moderate potential range. There-fore, the orthorhombic Li5MO4 is worth to be thoroughlystudied to optimize the overall performance of this structure.Li6Si2O7 is another lithium silicate may exhibit good kineticproperties. It has been studied in the glassy state ofLi4B2O5eLi6Si2O7eLi4P2O7 system [44], but there is noreport on the ionic conductivity of its crystalline phase. Be-sides the compounds with Ea values less than 0.5 eV, someother structures may behave as good ionic conductors are alsolisted in Table 1. It has been noticed that in most cases the Ea

Fig. 1. The lithium migration pathways reveal by BV calculations for the candidate structures with migration energy barriers Ea less than 0.5 eV.

Fig. 2. The band structure calculated by DFT method for (a) a-Li5AlO4 and (b) b-Li5AlO4. The Fermi energy is referred to zero and the high-symmetry k-path is

selected according to Ref. [28].

328 R. Xiao et al. / J Materiomics 1 (2015) 325e332

329R. Xiao et al. / J Materiomics 1 (2015) 325e332

value by BV method is always larger than that calculated byDFT calculations because the relaxation of the structuralframe during the lithium migration is ignored in the BV-basedsimulations [23]. Accordingly, we choose a relatively largevalue of Ea, up to 1.0 eV, as the threshold for screening fastlithium ion conductors with BV method. Several typicalcandidate structures will be discussed in the next subsections.

3.1. Li4CO4

Fig. 4. The total and partial density of states for monoclinic-Li4CO4. The

Fermi energy is referred to zero.

Carbonates and orthocarbonates are always interesting ob-jects in chemistry. Li2CO3 is an ionic conductor with “knock-off” diffusion mechanism and acts as the main inorganiccomponent of solid electrolyte interphase [45]. Orthocar-bonates, building with [CO4]

4� units, can be considered to beformed by the double hydration of CO2, have been computa-tionally demonstrated stable in S4 symmetry [46], althoughstill not been observed experimentally. The monoclinic-Li4CO4 presented here is predicted to exist at high pressure[47]. According to the Ea value calculated by BV method, theenergy needed to overcome for Liþ migrating in monoclinic-Li4CO4 is 0.50 eV. As shown in Fig. 3, the [CO4]

4� units arearranged in ab plane with the same orientation in this struc-ture. Liþ ions locate within the [CO4]

4� layer or between theneighboring layers. The former is difficult to migrate becauseof the blocking from [CO4]

4� units around Liþ, while thelatter is easy to move along the two-dimensional paths in thelithium layer. The electronic structure calculation indicatesthat the bandgap is about 4 eV, and the highest valence band ismainly composed of O-p states, illustrated by the density ofstates shown in Fig. 4. The hybridization between C and O isobvious at �6 eV and �8 eV, but the interactions between Liand O (or C) is relatively weak. The distortion of the structureand the charge transfer between Liþ and [CO4]

4� units duringLiþ movement is needed to be clarified to further understandthe stability of the compound and figure out a way to syn-thesize the structure. There are some other Li4CO4 isomersalso shown low migration energy barriers, including tetragonalstructure and several monoclinic lattices with different[CO4]

4� orientations. These structures should be furtherstudied and may be used as the start points for the optimizationof properties.

Fig. 3. The two-dimensional migration pathways in monoclinic-Li4CO4

calculated by BV method.

3.2. Li2MXO4

The g-tetrahedral structure, Li2MXO4 (M ¼ Zn, Mg, Caand X ¼ Ge, Si, Ti), is isostructural with g-Li3PO4, whichalready attracts a lot of attention as a solid-state electrolytematerial for batteries [48]. In this structure, the alternatingMO8 dodecahedra and XO4 tetrahedra are sharing edges andlinked to columns. Neighboring columns are connected atcorners to form a three-dimensional network. Lithium ions areenclosed in the channels parallel to c axis, which provides aone-dimensional pathway for Liþ migration. The Ea valuecalculated by BV method is 0.58 and 0.59 eV for Li2CaGeO4

and Li2CaSiO4, respectively, as indicated in line 13 and 14 inTable 1. Previous experiments revealed that tetrahedral-Li2Z-nSiO4 shows dc conductivity of 10�6 S/cm at room tempera-ture [49], implying that there is still much room forimprovement. Fig. 5 is the band structure for Li2CaSiO4 intetrahedral lattice. The energy gap is larger than 5 eV, evincingthe electrochemical stability of this material. As a promisingcandidate structure for solid state electrolytes, it is necessaryto analyze all possible doping strategies in detail forimproving the lithium ionic conductivities.

3.3. LiMXO4, Li2MO3 and LiM2(XO4)3

Besides increasing the lithium ionic conductivity of elec-trolytes, the reduction of interfacial resistance between elec-trolyte and electrodes is another challenge for designing thesolid-state batteries. One hopeful scheme to improve theinterfacial conduction is to construct the interface by twocrystals with similar structures and orientation. According tothis idea, we find some candidate structures to pair with thoseattractive cathode materials, including LiMXO4, Li2MO3 andLiM(PO4)3 as listed in Table 1 from line 16 to 28.

Olivine-type structure LiMXO4 shows same structure withLiFePO4, which is a commercialized cathode with high safety,environmental benignity and low cost. The BV-methodcalculated Ea values for LiScGeO4, LiInGeO4, LiZnPO4,

Fig. 5. The band structure of Li2CaSiO4 calculated by DFT method. The Fermi

level is referred to 0 eV and the high-symmetry k-path is chosen according to

Ref. [28].

Fig. 6. The lithium migration energy barriers and pathways calculated by BV

method for (a) LiScGeO4, (b) Li2SnO3 and (c) LiGe2(PO4)3.

330 R. Xiao et al. / J Materiomics 1 (2015) 325e332

LiMgPO4, LiInSiO4 and LiScSiO4 are about 0.85e1.00 eV.The migration of lithium ions is along the one-dimensionalpath in b axis, shown in Fig. 6(a), as what found in LiFePO4

[50]. Jalem et al. adopted ab initio method to study 66 olivine-type oxides and identified MgeAs, SceGe, IneGe, MgeP aspromising MeX pairs to improve the ionic conductivity in thisstructure [51].

Li2MnO3 plays a key role in Li-rich Mn-based layeredmaterials (mLi2MnO3$nLiMO2, M ¼ Mn, Ni, Co, etc.) forachieving unusually high lithium storage capacity [52]. DFTcalculations revealed that the Liþ hopping can happen withinthe Li plane and between Li and transition-metal layers withactivation energies in the range from 0.51 eV to 0.84 eV [53].By screening with high-throughput bond-valence calculations,Li2SnO3 with same structure and similar diffusion path toLi2MnO3 is detected as shown in Fig. 6(b). The Ea value ofLi2SnO3 calculated by BV method is 0.76 eV, and the latticemismatch with Li2MnO3 is ~7%. Therefore, it is a promisingcandidate to pair with Li-rich cathode in solid-state batteries.The lattice constants of Li2GeO3 and Li2SiO3 are also close tothose of Li2MnO3. However, the variation from MnO6 andSnO6 octahedra to GeO4 and SiO4 tetrahedra changes the ar-rangements of the blocking units. As a result, the one-dimensional migration paths occur in Li2GeO3 and Li2SiO3.

LiM2(XO4)3 in triclinic lattice is NASICON-type phase andalready studied as solid electrolytes in previous investigations[54]. The solid electrolytes based on lithium titanium phos-phate, e.g. Li1.5Al0.5Ti1.5(PO4)3, have been applied in experi-ments to construct all-solid-state lithium batteries [55]. Ourcalculations indicates that the optimization of the NASICON-type LiM2(XO4)3 may lead to new materials with high lithiumionic conductivity.

4. Summary

We report the screening results of high-throughput bond-valence calculations and some candidate structures of fastlithium ion conductors are identified. By implementing thebond-valence simulation code, BVpath, and the automaticflow for high-throughput calculations, more than 1000 com-pounds from the ICSD database are studied. The promisingcompounds with low migration energy barriers are selected asoriginal candidate structures for further optimization. In thecandidates, LiB3O5 and Li2Te2O5 are investigated little on theionic conductivity previously, and worth to be studied indetail. Li5AlO4 and some lithium silicates have alreadynoticed because of their high ionic conductivity. The thoroughstudies to optimize the overall performance of these structuresare necessary. The orthocarbonates have been predicted toexist at high pressure and our calculations indicate the possiblegood ionic conductivity in these meta-structures. The efforts tosynthesize and stabilize these structures maybe carried out inthe future. The g-tetrahedral structure Li2MXO4 also showvery low migration energy barriers. Both the barrier and car-rier concentration can be adjusted through substitution of Mand X sites with atoms in different valence states and ionicradius. Several structures, which may well match the interfacebetween electrolytes and electrodes, are also identified,

331R. Xiao et al. / J Materiomics 1 (2015) 325e332

including LiMXO4, Li2MO3 and LiM2(XO4)3 which showsame structure with typical cathodes LiFePO4, Li2MnO3 andLiTi2(PO4)3, respectively. The well-matched structure mayreduce the interfacial stress and improve the conduction. Theband gaps of the candidate structures are studied using DFTcalculations. However, it is well known that the GGA func-tionals underestimate energy band gaps for many materials,while the meta-GGA functionals yield band gaps with an ac-curacy comparable with hybrid functional but with much lesscomputational cost. Thus, the finer calculations will be carriedout to confirm the band gaps with better accuracy in furtherinvestigations, and we will focus on the optimization of kineticproperties for the candidate structures and their derivatives.

Acknowledgment

We acknowledge the National Natural Science Foundationof China (Grant Nos. 11234013 and 51172274), “863” Project(Grant No. 2015AA034201), and Beijing S&T Project (GrantNo. Z13111000340000) for financial support and the ShanghaiSupercomputer Center for providing computing resources.

References

[1] Tarascon JM, Armand M. Issues and challenges facing rechargeable

lithium batteries. Nature 2001;414:359e67.

[2] Karden E, Ploumen S, Fricke B, Miller T, Snyder K. Energy storage

devices for future hybrid electric vehicles. J Power Sources

2007;168:2e11.

[3] Dunn B, Kamath H, Tarascon JM. Electrical energy storage for the grid: a

battery of choices. Science 2011;334:928e35.

[4] Takada K. Progress and prospective of solid-state lithium batteries. Acta

Mater 2013;6:759e70.

[5] Xu K. Nonaqueous liquid electrolytes for lithium-based rechargeable

batteries. Chem Rev 2004;104:4303e417.[6] Gao J, Chu G, He M, Zhang S, Xiao RJ, Li H, et al. Screening

possible solid electrolytes by calculating the conduction pathways

using bond valence method. Sci China-Phys Mech Astron

2014;57:1526e35.[7] Ribes M, Barrau B, Souquet JL. Sulfide glasses: glass forming region,

structure and ionic conduction of glasses in Na2S-XS2 (X ¼ Si; Ge),

Na2S-P2S5 and Li2S-GeS2 systems. J Non-Cryst Solids

1980;38:271e6.

[8] Kanno R, Hata T, Kawamoto Y, Irie M. Synthesis of a new lithium ionic

conductor, thio-LISICONelithium germanium sulfide system. Solid

State Ion 2000;130:97e104.[9] Winand J-M, Rulmont A, Tarte P. Nouvelles solutions solids LI(MI-

V)2�x(N

IV)x(PO4)3 (L ¼ Li,Na M,N ¼ Ge,Sn,Ti,Zr,Hf) synth�ese et �etude

par diffraction x et conductivit�e ionique. J Solid State Chem

1991;93:341e9.

[10] Belous AG, Novitskaya GN, Polyanetskaya SV, Gornikov YI. Study of

complex oxides with the composition La2/3-xLi3xTiO3. Inorg Mater

1987;23:412.

[11] Thangadurai V, Kaak H, Weppner W. Novel fast lithium ion conduction

in garnet-type Li5La3M2O12 (M ¼ Nb, Ta). J Am Ceram Soc

2003;86:437e40.

[12] Kanehori K, Matsumoto K, Miyauchi K, Kudo T. Thin film solid elec-

trolyte and its application to secondary lithium cell. Solid State Ion

1983;9:1445e8.

[13] Alpen Uv, Rabenau A, Talat GH. Ionic conductivity in Li3N single

crystals. Appl Phys Lett 1977;30:621e2.[14] Stoneham AM, Wade E, Kilner JA. A model for the fast ionic diffusion in

alumina-doped LiI. Mater Res Bull 2012;14:661e6.

[15] Kamaya N, Homma K, Yamakawa Y, Hirayama M, Kanno R,

Yonemura M, et al. A lithium superionic conductor. Nat Mater

2011;10:682e6.

[16] Curtarolo S, Hart GLW, Nardelli MB, Mingo N, Sanvito S, Levy O. The

high-throughput highway to computational materials design. Nat Mater

2013;12:191e201.

[17] Hafner J, Wolverton C, Ceder G. Toward computational materials design:

the impact of density functional theory on materials research. MRS Bull

2006;31:659e68.

[18] Meng YS, Arroyo-de Dompablo ME. First principles computational

materials design for energy storage materials in lithium ion batteries.

Energy Environ Sci 2009;2:589e609.[19] Ouyang CY, Chen LQ. Physics towards next generation Li secondary

batteries materials: a short review from computational materials design

perspective. Sci China-Phys Mech Astron 2013;56:2278e92.

[20] Shi SQ, Lu P, Liu ZY, Qi Y, Hector Jr LG, Li H, et al. Direct calculation

of Li-ion transport in the solid electrolyte interphase. J Am Chem Soc

2012;134:15476e87.

[21] Fujimura K, Seko A, Koyama Y, Kuwabara A, Kishida I, Shitara K, et al.

Accelerated materials design of lithium superionic conductors based on

first-principles calculations and machine learning algorithms. Adv En-

ergy Mater 2013;3:980e5.

[22] Adams S, Rao RP. High power lithium ion battery materials by

computational design. Phys Status Solidi A 2011;208:1746e53.

[23] Xiao RJ, Li H, Chen LQ. High-throughput design and optimization of

fast lithium ion conductors by the combination of bond-valence method

and density functional theory. Sci Rep 2015. http://dx.doi.org/10.1038/

srep14227.

[24] Inorganic Crystal Structure Database. Fachinformationszentrum. Karls-

ruhe: ICSD; 2008.

[25] Adams S. Relationship between bond valence and bond softness of alkali

halides and chalcogenides. Acta Cryst 2001;B57:278e87.

[26] Kresse G, Furthmuller J. Efficiency of ab-initio total energy calculations

for metals and semiconductors using a plane-wave basis set. Comput

Mater Sci 1996;6:15e50.

[27] Perdew JP, Burke K, Ernzerhof M. Generalized gradient approximation

made simple. Phys Rev Lett 1997;78:1396.

[28] Setyawan W, Curtarolo S. High-throughput electronic band structure

calculations: challenges and tools. Comput Mater Sci 2010;49:299e312.

[29] Lin S, Sun Z, Wu B, Chen C. The nonlinear optical characteristics of a

LiB3O5 crystal. J Appl Phys 1990;67:634e8.

[30] Guo R, Markgraf SA, Furukawa Y, Sato M, Bhalla AS. Pyroelectric,

dielectric, and piezoelectric properties of LiB3O5. J Appl Phys

1995;78:7234e9.

[31] Radaev SF, Sorokin NI, Simonov VI. Atomic structure and one-

dimensional ionic conductivity of lithium triborate LiB3O5. Sov Phys

Solid State 1991;33:3597e600.

[32] Xu YN, Ching WY. Electronic structure and optical properties of

LiB3O5. Phys Rev B 1990;41:5471e4.[33] Fu J. Fast Liþ ion conduction in Li2O-Al2O3-TiO2-SiO2-P2O5 glass-ce-

ramics. J Am Ceram Soc 1997;80:1901e3.

[34] Cachau-Herreillat D, Norbert A, Maurin M, Philippot E. Comparative

crystal chemical study and ionic conductivity of two varieties of a- and

b-lithium tellurite (Li2Te2O5). J Solid State Chem 1981;37:352e61.

[35] Xu K, Zhang SS, Lee U, Allen JL, Jow TR. LiBOB: is it an alternative

salt for lithium ion chemistry. J Power Sources 2005;146:79e85.[36] Sen S, George AM, Stebbins JF. Ionic conduction and mixed cation ef-

fect in silicate glasses and liquids: 23Na and 7Li NMR spin-lattice

relaxation and a multiple-barrier model for percolation. J Non-Cryst

Solids 1996;197:53e64.[37] Kennedy KM, Morrison G, Chadwick AV. A computer modelling study

of ionic conductivity in some lithium silicates and aluminosilicates. An

Quimica Int Ed 1998;94:27e30.

[38] Knauth P. Ionic conductor composites: theory and materials. J Electro-

ceram 2000;5:111e25.

[39] Pechenik A, Whitmore DH, Susman S, Ratner MA. Transport in glassy

fast-ion conductors: a study of LiAlSiO4 glass. J Non-Cryst Solids

1988;101:54e64.

332 R. Xiao et al. / J Materiomics 1 (2015) 325e332

[40] Johnson Jr RT, Biefeld RM. Ionic conductivity of Li5AlO4 and Li5GaO4

in moist air environments: potential humidity sensors. Mater Res Bull

1979;14:537e42.

[41] Avalos-Rendon T, Casa-Madrid J, Pfeiffer H. Thermochemical capture of

carbon dioxide on lithium aluminates (LiAlO2 and Li5AlO4): a new

option for the CO2 absorption. J Phys Chem A 2009;113:6919e23.

[42] Esaka T, Greenblatt M. Lithium ion conduction in substituted Li5MO4,

M ¼ Al, Fe. J Solid State Chem 1987;71:164e71.[43] Esaka T, Greenblatt M. Lithium ion conduction in substituted Li5GaO4.

Solid State Ion 1986;21:255e61.

[44] Kowada Y, Tatsumisago M, Minami T. Raman-spectra of rapidly

quenched glasses in the systems Li3BO3-Li4SiO4-Li3PO4 and Li4B2O5-

Li6Si2O7-Li4P2O7. J Phys Chem 1989;93:2147e51.

[45] Shi SQ, Qi Y, Li H, Hector Jr LG, et al. Defect thermodynamics and

diffusion mechanisms in Li2CO3 and implications for the solid electrolyte

interphase in Li-ion batteries. J Phys Chem C 2013;117(17):8579e93.[46] Narasimhamurthy N, Manohar H, Samuelson AG, Chandrasekhar J.

Cumulative anomeric effect: a theoretical and x-ray diffraction study of

orthocarbonates. J Am Chem Soc 1990;112:2937e41.[47] �Can�carevi�c �ZP, Sch€on JC, Jansen M. Possible existence of alkali metal

orthocarbonates at high pressure. Chem A Eur J 2007;13:7330e48.

[48] Du YA, Holzwarth NAW. Li ion diffusion mechanisms in the crystalline

electrolyte g-Li3PO4. J Electrochem Soc 2007;154:A999e1004.[49] Adnan SBRS, Nohamed NS. Conductivity and dielectric studies of

Li2ZnSiO4 ceramic electrolyte synthesized via citrate sol gel method. Int

J Electrochem Sci 2012;7:9844e58.

[50] Sun Y, Lu X, Xiao RJ, Li H, Huang XJ. Kinetically controlled lithium-

staging in delithiated LiFePO4 driven by the Fe center mediated inter-

layer Li-Li interactions. Chem Mater 2012;24:4693e703.

[51] Jalem R, Aoyama T, Nakayama M, Nogami M. Multivariate method-

assisted ab initio study of olivine-type LiMXO4(Main group M2þ-X5þ

and M3þ-X4þ) compositions as potential solid electrolytes. Chem Mater

2012;24:1357e64.

[52] Thackeray MM, Johnson CS, Vaughey JT, Li N, Hackney SA. Advances

in manganese-oxide ‘composite’ electrodes for lithium-ion batteries. J

Mater Chem 2005;15:2257e67.[53] Xiao RJ, Li H, Chen LQ. Density functional investigation on Li2MnO3.

Chem Mater 2012;24:4242e51.

[54] Nuspl G, Takeuchi T, Weiss A, Kageyama H, Yoshizawa K, Yamabe T.

Lithium ion migration pathways in LiTi2(PO4)3. J Appl Phys

1999;86:5484e91.

[55] Kotobuki M, Isshiki Y, Munakata H, Kanamura K. All-solid-state lithium

battery with a three-dimensionally ordered Li1.5Al0.5Ti1.5>(PO4)3 elec-

trode. Electrochim Acta 2010;55:6892e6.

Associate Professor Ruijuan Xiao, Institute of

Physics, Chinese Academy of Sciences Email:

rjxiao@iphy.ac.cn. Dr. Ruijuan Xiao is an Associate

Professor in Key Laboratory For Renewable Energy,

Institute of Physics, Chinese Academy of Sciences,

Beijing, China. Before joining the faculty of Insti-

tute of Physics in 2010, she had conducted Post

Doc. studies in Germany since 2007. Her recent

research focuses on theoretical simulations on

lithium-ion battery materials, including the investi-

gations on the (de)lithiation mechanisms, kinetic

properties, electronic conductivities of battery materials, and the relation-

ships between the basic properties of materials and the performance of the

batteries.