Li3Y(PS4)2 and Li5PS4Cl2: New Lithium Superionic ConductorsPredicted from Silver Thiophosphates using Efficiently Tiered AbInitio Molecular Dynamics SimulationsZhuoying Zhu, Iek-Heng Chu, and Shyue Ping Ong*

Department of NanoEngineering, University of California San Diego, 9500 Gilman Drive, Mail Code 0448, La Jolla, California92093-0448, United States

*S Supporting Information

ABSTRACT: We report two novel, earth-abundant lithiumsuperionic conductors, Li3Y(PS4)2 and Li5PS4Cl2, that arepredicted to satisfy the necessary combination of good phasestability, high Li+ conductivity, wide band gap and goodelectrochemical stability for solid electrolyte applications in all-solid-state rechargeable lithium-ion batteries. These candidateswere identified from a high-throughput first-principles screen-ing of the Li−P−S ternary and Li−M−P−S (where M is anon-redox-active element) quaternary chemical spaces, includ-ing candidates obtained by replacing Ag with Li in the Ag−P−S and Ag−M−P−S chemical spaces. An efficient tiered screening strategy was developed that combines topological analysis withab initio molecular dynamics simulations to exclude rapidly candidates unlikely to satisfy the stringent conductivity requirementsof lithium superionic conductors. In particular, we find Li3Y(PS4)2 to be an extremely promising candidate exhibiting a room-temperature Li+ conductivity of 2.16 mS/cm, which can be increased multifold to 7.14 and 5.25 mS/cm via aliovalent dopingwith Ca2+ and Zr4+, respectively. More critically, we show that the phase and electrochemical stability of Li3Y(PS4)2 is expected tobe better than current state-of-the-art lithium superionic conductors.

■ INTRODUCTIONAll-solid-state rechargeable lithium-ion batteries (ss-LIBs) are arevolutionary architecture for energy storage that holds thepromise to be both more energy dense as well as safer thantraditional organic-liquid-electrolyte-based lithium-ion bat-teries.1,2 The crucial enabling component in ss-LIBs is thenonflammable lithium superionic conductor solid electrolyte,which must possess a challenging suite of properties, includinghigh Li+ conductivity, excellent phase stability, electrochemicaland mechanical compatibility with the electrodes, and ideally,stability under ambient air.3−8

Thiophosphate lithium superionic conductors have emergedas one of the leading contenders for ss-LIB solid electrolyteapplications. For instance, the Li10GeP2S12 (LGPS) family ofmaterials3,9−12 and the Li7P3S11

13,14 glass-ceramic have ionicconductivities exceeding 10 mS/cm, on par or even exceedingthat of organic solvent electrolytes.15 More recently,Li9.54Si1.74P1.44S11.7Cl0.3, which also has the LGPS structure,was reported to have an ionic conductivity as high as 25 mS/cmat room temperature.16 Besides typically higher ionicconductivities, sulfides have the advantage of being softerthan oxides, allowing low porosity and intimate electrode−electrolyte contact to be achieved via cold-pressing techni-ques.6,17

Despite these advances, the number of known lithiumsuperionic conductors remains relatively few, and existingmaterials all suffer from various limitations. For instance,

Li10GeP2S12 is expensive due to the use of Ge, whereas the Sn10

and Si11,12 substituted analogues tend to form conducting,nonpassivating phases upon contact with the Li anode.9,18,19

Other candidates, such as Li7P3S11, are metastable and tend toform nonpassivating layers with the typical LiCoO2 cathodeused in today’s LIBs.14,20,21 There is thus an urgent need for thedevelopment of novel lithium superionic conductors that canpotentially mitigate some of these limitations.When surveying the space of known lithium thiophosphates,

an interesting observation is that many of them have analoguesin Ag thiophosphates. For example, Li7P3S11 and Li3PS4 bearremarkable structural similarity to Ag7P3S11 and Ag3PS4,respectively.22 The highly interesting Li argyrodite superionicconductors with formula Li6PS5X (X = Cl, Br, I)23 derive theirname from the mineral argyrodite (Ag8GeS6), and showpromising Li+ conductivities of >1 mS/cm for X = Cl and Br.24

Many Ag compounds are also known to exhibit extraordinarilyhigh ionic conductivities; for example, α-AgI is perhaps the bestknown, and one of the first superionic conductors everdiscovered.25−27

Special Issue: Computational Design of Functional Materials

Received: September 22, 2016Revised: December 16, 2016Published: December 20, 2016

Article

pubs.acs.org/cm

© XXXX American Chemical Society A DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

Inspired by this observation, we have performed acomprehensive screening of the ternary Li−P−S andquaternary Li−M−P−S (where M is a non-redox-activeelement) chemical spaces for new lithium superionicconductors using an efficient screening approach based onhigh-throughput density functional theory (DFT) calculations.The scope of this work extends beyond the known Lithiophosphates and includes novel candidates obtained fromLi for Ag substitution of Ag thiophosphates. The screeningyielded two highly promising candidates, Li3Y(PS4)2 andLi5PS4Cl2, which are predicted to satisfy the necessarycombination of excellent phase and electrochemical stability,high Li+ conductivity, and low electronic conductivity. We alsoshow that the conductivity of the more promising Li3Y(PS4)2material can be further enhanced multifold via aliovalentdoping. Finally, we will discuss the relative merits of this newsuperionic conductor compared to current state-of-the-artsuperionic conductors.

■ METHODSInitial Candidate Selection. The initial pool of candidate lithium

superionic conductors was constructed from the following:

1. All known ordered Li−P−S and Li−M−P−S structures fromthe 2015 version of the Inorganic Crystal Structure Database(ICSD).28 Only non-redox-active elements were allowed for M.

2. Substitution of Ag with Li on all known ordered Ag−P−S andAg−M−P−S structures from the ICSD.

Unique structures were identified from the pooled candidates usingan in-house structure matching algorithm implemented in the PythonMaterials Genomics (pymatgen) materials analysis library.29

DFT Calculations. All DFT calculations were performed using theVienna Ab initio Simulation Package (VASP)30 within the projectoraugmented-wave approach.31 The exchange-correlation functional andcalculation parameters were carefully selected to achieve a balancebetween computational accuracy and cost for the different types ofcalculations. We will provide a brief summary of the parameters here,and full details are provided in the Supporting Information forinterested readers.

• Structure Relaxation. Spin-polarized calculations using thePerdew−Burke−Ernzerhof (PBE) generalized-gradient approx-imation (GGA)32 functional were used for all structuralrelaxations. The convergence parameters, e.g., k-point densityof at least 1000/(number of atoms in the unit cell) and energycutoff of 520 eV, were similar to those used in the MaterialsProject (MP), which have been tested extensively over a broadrange of chemistries.33

• Phase Stability. The phase stability of a compound wasestimated by determining its energy above the convex hull Ehullin the relevant Li−P−S and Li−M−P−S phase diagrams.34

Stable compounds have an Ehull of 0, and the higher the value,the more unstable the compound is at 0 K. Apart from thecompounds of primary interest in this work, the energies ofexisting compounds were extracted from MP database33 usingthe Materials Application Programming Interface (API).35 Toaccount for overbinding of sulfur in DFT calculations, anenergy correction of −0.66 eV per S atom for sulfides wasapplied.36

• Electrochemical Stability. The electrochemical stability wasassessed using the lithium grand potential phase diagramapproach.9,14 In this approximation, Li is treated as the mainmobile species and the solid electrolyte/electrode interface canbe modeled as an open system with respect to Li. The relevantthermodynamic potential is therefore the grand potential,which can be approximated as ϕ ≈E−μLiNLi in which E, NLi andμLi are DFT total energy, number of lithium atoms in the opensystem, and lithium chemical potential, respectively. The phaseequilibria at the anode and charged cathode can be

approximated as the lithium superionic conductor compositionat high μLi = μLi

0 and low μLi = μLi0 − 5 eV (μLi

0 is the chemicalpotential of metallic Li), respectively. The limiting value of 5 Vis chosen based on typical cutoff voltages (>4.2 V) in thecurrent LiCoO2 and other next generation cathodes used intoday’s lithium-ion batteries.

• Ab Initio Molecular Dynamics (AIMD) Simulations.Automated non-spin-polarized AIMD simulations were per-formed in an NVT ensemble at elevated temperatures with aNose−Hoover thermostat.37,38 A smaller plane-wave energycutoff of 280 eV, a minimal Γ-centered 1 × 1 × 1 k-point mesh,and a time step of 2 fs were adopted. The simulation supercellsizes were at least 9 Å along each lattice direction. In line withprevious studies,9,39 the simulation cell parameters were fixed atthe fully relaxed cell parameters at 0 K. The Li+ diffusivity wasobtained via a linear fit of the mean square displacement(MSD) with time, and Arrhenius plots were constructed fromsimulations at multiple temperatures to obtain the activationenergy Ea and extrapolated room-temperature self-diffusivityD300 K and conductivity σ300 K.

• Climbing Image Nudged Elastic Band (CI-NEB) Calcu-lations. CI-NEB calculations were performed to determine thevacancy migration barriers for the most promising candidates.Overall charge neutrality was achieved via adding a positivebackground charge. The forces were converged to within 0.05eV/Å.

• Electronic Structure. Band gap calculations were performedusing the Heyd−Scuseria−Ernzerhof (HSE) hybrid func-tional,40,41 due to the well-known underestimation of bandgaps by semilocal functionals.42

■ RESULTSParameterization of Screening Criteria. Figure 1 shows

a schematic of the high-throughput (HT) screening frameworkused in this work, which is tiered based on considerations of therelative importance of each property for lithium solidelectrolyte applications and the computational cost requiredto calculate it.First and foremost, all technologically relevant materials must

be synthesizable, i.e., exhibit good phase stability. In this work,we have adopted a cutoff of Ehull < 30 meV/atom, which isbased on similar cutoffs adopted in previous HT computationalmaterials screening efforts43,44 as well as the fact that Li7P3S11and Li10GeP2S12, both well-known superionic conductors, havebeen predicted to have an Ehull of 21−25 meV/atom.14,39

Second, a lithium superionic conductor must have a high Li+

conductivity at room temperature (σ300 K). Because of the nearunity transference number of lithium superionic conductors,σ300 K exceeding 0.1 mS/cm should suffice for comparableperformance with organic liquid electrolytes, though σ300 K > 1mS/cm is preferred. However, obtaining converged diffusivityand conductivity numbers from AIMD simulations is a highlycomputationally demanding process, usually requiring at leasthundreds of picoseconds of simulation time (∼50 000−100 000time steps) at multiple temperatures. Because we are interestedonly in superionic conductors with extremely high diffusivity,we have adopted the following three-step diffusivity screeningapproach in this work.

Topological Screening Step. The first screening step isbased purely on topological considerations.45,46 Only materialsexhibiting >1D diffusion networks with a minimum bottlenecksize rc of 1.75 Å, are considered as suitable candidates forlithium superionic conductors. This cutoff is slightly smallerthan the channel size for the Li10GeP2S12 superionic conductor(1.84 Å).9 A looser cutoff is used to avoid screening out too

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

B

many candidates in the first screening step. The topologicalevaluation was carried out using the open source software Zeo++.47,48

Quick Estimation Step. Quick estimates of the diffusivityand activation energy were obtained using the mean squaredisplacements (MSDs) obtained from short AIMD simulationsof 50 ps at 800 K (MSD800 K) and 1200 K (MSD1200 K). Figure2 shows a log−log plot of the MSD1200 K versus MSD800 K for awide range of known compounds as well as the materialsscreened in this work. The known superionic conductorsevaluated include an approximate ordered model(Li10Si1.5P1.5S11.5Cl0.5, see Supporting Information for details)for the recently reported Li9.54Si1.74P1.44S11.7Cl0.3 superionicconductor, which has the LGPS structure and an extraordinarilyhigh conductivity of 25 mS/cm.16

We may observe that all known superionic conductors fallwithin the white region bounded by MSD800 K > 5 Å2 andMSD1200 K/MSD800 K < 7, and have therefore used these criteria

in our screening process. The former criterion ensures aminimum baseline diffusivity, whereas the second criterionensures that the activation energy is below ∼400 meV (seeSupporting Information for derivation). The relative trends areconsistent with the known properties of the conductors, withfaster conductors such as Li10Si1.5P1.5S11.5Cl0.5 exhibiting largerMSD800 K (higher diffusivity at 800 K) and lower MSD1200 K/MSD800 K (lower activation barriers) than slightly poorerconductors such as Li10GeP2S12. It should be noted thatthough Li7P3S11 was experimentally reported to have an ionicconductivity of 17 mS/cm,13 recent computational work by theauthors of this work suggests that the intrinsic conductivity ofthis material may be as high as 60 mS/cm,14 which accounts forits high MSD800 K. Also, though Li3OClxBr1−x and Na3PS4 wereinitially reported to be superionic conductors,49,50 the pristinecrystalline phases have been shown to have low room-temperature conductivities, and higher ionic conductivities areobtained only with the introduction of a large number ofdefects/dopants (e.g., doped-Na3PS4) and/or via amorphiza-tion.36,51−53

Converged Screening Step. Finally, longer AIMD simu-lations at six temperatures were performed on the materials thatpass the first two screening steps to obtain convergeddiffusivities (and conductivities) and activation barriers.The above three-step screening process allows us to

eliminate rapidly poor candidates with a minimum amount ofcomputational resources, and devote expensive AIMD simu-lations to obtain converged diffusivity statistics on the mostpromising materials.Besides excellent Li+ conductivity, a solid electrolyte for all-

solid-state rechargeable lithium-ion batteries must also beelectronically insulating and exhibit good electrochemicalstability against the electrodes. An assessment of theseproperties was carried out for the most promising candidates.

Identification of Potential Candidates. Table 1 summa-rizes the phase stability, topological parameters and rapidAIMD screening results of all new Li−P−S and Li−M−P−Scandidates. The rapid AIMD screening results are alsopresented in Figure 2 for comparison with known superionicconductors. We note that though we did perform Li for Agsubstitution of Ag−P−S compounds, the derived compoundsare all well-known compounds such as Li7P3S11 and Li3PS4 thathave already been extensively explored as lithium superionicconductors, and thus will not be considered further here.Among the new quaternary compounds, only Li3Y(PS4)2

(LYPS) and Li5PS4Cl2 (LPSCl) satisfy all the initial screeningcriteria: low Ehull, rc > 1.75 Å, MSD800 K > 5 Å2 and MSD1200 K/MSD800 K < 7. Their MSD800 K are on par with that of theleading LGPS-based candidate, Li10Si1.5P1.5S11.5Cl0.5, but theirMSD1200 K/MSD800 K ratios are slightly higher. ThoughLi15P4S16Cl3, LiZnPS4 and LiAl(PS3)2 are also predicted tohave fairly low Ehull and reasonably high MSD800 K, theirMSD1200 K/MSD800 K are far too high, indicating high activationbarriers. The remaining candidates do not pass either the phasestability criterion or the topological screening. During thepreparation of this paper, it has come to our attention that theLiZnPS4 candidate in Table 1 has been investigated as asuperionic conductor.54 Our screening calculations show thatthe stoichiometric LiZnPS4 compound fails the MSD ratiocutoff by a factor of 2, which is consistent with the highactivation barriers reported for the stoichiometric compoundreported in Richards et al.’s work.54 A more in-depth

Figure 1. Flowchart of the screening procedure for new lithiumsuperionic conductors. M refers to non-redox-active elements. Therelative sizes of the arrows indicate the number of candidatesremaining after the current screening step.

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

C

comparison of our proposed candidates with known superionicconductors is provided in the Discussion section.The crystal structures of LYPS and LPSCl are shown in

Figure 3, and the relaxed lattice parameters are given in Table 2.LYPS and LPSCl belong to the monoclinic C2/c andorthorhombic C2mm space groups, respectively. Both LYPSand LPSCl structures have two symmetrically distinct Li sites,which are labeled as Li1 and Li2 in Figures 6 and 7. Thecrystallographic information files (CIFs) of the DFT-relaxedstructures of the two candidates are provided in the SupportingInformation.Li+ Conductivities and Mechanisms. Long AIMD

simulations of at least 200 ps at multiple temperatures wereperformed on the promising LYPS and LPSCl candidates.Figure 4 shows the Arrhenius plot of the diffusivity versus1000/temperature for the two candidates, and Table 3summarizes the key conductivity properties extracted. Theestimated activation energies Ea for LYPS and LPSCl are 278and 304 meV, respectively. The extrapolated room-temperatureconductivities are 2.16 mS/cm for LYPS and 1.85 mS/cm for

LPSCl, i.e., both candidates are indeed verified to be lithiumsuperionic conductors.To understand further the atomistic diffusion mechanisms

and pathways, we calculated the Li+ probability density function(PDF)36 and the distinct part of van Hove correlation functionfrom the AIMD simulations of the two candidates at 800 K. CI-NEB calculations were then performed to calculate the vacancymigration barriers in the identified pathways. From Figure 5, weobserve that both candidates have 3D diffusion networks.Schematics of the identified diffusion pathways and thecorresponding calculated CI-NEB migration barriers are givenin Figures 6 and 7 for LYPS and LPSCl, respectively. The plotsof the distinct part of van Hove correlation function (see FigureS3 in Supporting Information) indicate that a diffusing Li+ isreplaced by another Li+ within 10 ps, which suggests that theLi+ motions are highly concerted in both materials, similar toother well-known lithium superionic conductors such asLi7P3S11 and Li10GeP2S12.

14,36

For LYPS, there are five symmetrically distinct hops betweenneighboring Li sites, namely, A → B, B → F, B → C, C → Eand C → H (see Figure 6). We may observe that the lowest

Figure 2. Plot of MSD1200 K versus MSD800 K for a wide range of known alkali conductors and the new candidates in this work. A log−log scale isused for better resolution across orders of magnitude differences in diffusivity/conductivity. Square markers indicate known materials, which includewell-established superionic conductors such as Li7P3S11,

13 the LGPS family3,16 (Li10SiP2S12, Li10GeP2S12, Li10SnP2S12, Li10Si1.5P1.5S11.5Cl0.5), anddoped Na3PS4

36,53 as well as relatively poorer conductors such as pristine Na3PS436,50 and Li3OClxBr1−x.

49,51 Circle markers indicate new candidatesscreened, with predicted superionic conductors in green and predicted poor conductors in yellow. All superionic conductors fall into the whitetrapezoid zone, which is bounded by the lines MSD800 K = 5 Å2 and MSD1200 K/MSD800 K = 7.

Table 1. Energy above Hull (Ehull), >1D Channel Size (rc), and Short AIMD (50 ps) Results MSD800 K, MSD1200 K/MSD800 K ofLi Thiophosphates Studied in the Initial Screeninga

Compound Source (ICSD number) Ehull (meV/atom) rc (Å) MSD800 K (Å2) MSDMSD

1200 K

800 KPromising candidates

Li3Y(PS4)2 (C2/c) Ag3Y(PS4)2 (417658) 2 1.88 65.1 4.5Li5PS4Cl2 (C2mm) Ag5PS4Cl2 (416587) 17 1.76 77.9 3.1

Candidates failing at least one screening criteriaLi15P4S16Cl3 (I43d) Ag15P4S16Cl3 (416586) 8 1.76 2.2 145.1LiZnPS4 (I4) ICSD (95785) 0 1.83 24.7 13.9LiAl(PS3)2 (C2/c) ICSD (425979) 0 1.82 27.8 16.5Li2Zn(PS3)2 (C2/c) Ag2Zn(PS3)2 (72719) 33 1.84LiIn(PS3)2 (P31c) AgIn(PS3)2 (202185) 0 0LiZnPS4 (Pna21) AgZnPS4 (48197) 16 0

aMeasures in which various candidates fail one of the screening criteria are bolded.

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

D

barrier percolating pathway in LYPS is A → B→C → D (Li1−Li1−Li2−Li1) with a barrier of 210 meV. To enable >1Ddiffusion, the next lowest barrier pathway is G → C→H (all

Li2) along the c direction with an overall barrier of 268 meV.The E → B→F (all Li1) pathway has a much higher overallbarrier of 406 meV, which suggests it is more likely for avacancy at E to diffuse to B via the lower barrier E → G→C →B path. The estimated effective 3D vacancy migration barrier(268 meV) is therefore in good agreement with the activationenergy from AIMD simulations (278 meV).For LPSCl, we may observe that the crystal structure

comprises layers of Li1 and Li2 (see Figure 7) in the b-c planestacked along the a direction. Within each Li1 plane, thevacancy migration barriers are relatively low (A → B→C → Dwith overall barrier of 168 meV, and B → E with barrier of 217meV). For 3D diffusion, the H → G→F Li1-only pathconnecting different Li1 layers has the lowest overall barrier of321 meV. This is again in reasonably good agreement with theAIMD activation energy of 304 meV. All other paths involvingvacancy hops between Li1 and Li2 sites have significantlyhigher barriers (>380 meV, see Figure S2 in the SupportingInformation).

Electronic Band Gap. Figure 8 shows the calculated HSEdensities of states for LYPS and LPSCl. Both candidates arelarge band gap (>3 eV) insulators. In both candidates, thevalence band maximum is dominated by anion S-p character,whereas the conduction band minimum has contributions fromthe framework cations (Y and P) as well as anion S.The band gap is also an upper limit for the intrinsic stability

of the material against reduction (acceptance of an electron)and oxidation (loss of an electron). Similar to other sulfide-based solid-electrolytes,9,14 the intrinsic electrochemicalstability of the two candidates are limited to ∼3.5 eV.

Electrochemical Stability. Better estimates of the electro-chemical stabilities of LYPS and LPSCl were obtained using thelithium grand potential approach.9,55,56 Table 4 summarizes thepredicted phase equilibria at the solid electrolyte/anode(metallic Li) interface and solid electrolyte/charged 5 V

Figure 3. Crystal structures (conventional cell) of the identified Li superionic conductor candidates, (a) Li3Y(PS4)2 and (b) Li5PS4Cl2. Purpletetrahedra, PS4; dark green octahedra, YS6; green spheres, Li; yellow spheres, S; red spheres, Cl.

Table 2. Relaxed Conventional Unit Cell Lattice Parameters for Li3Y(PS4)2 and Li5PS4Cl2

Compound Atoms/cell a (Å) b (Å) c (Å) α (deg) β (deg) γ (deg)

Li3Y(PS4)2 56 17.122 9.290 9.137 90.0 122.3 90.0Li5PS4Cl2 24 7.212 10.494 6.024 90.0 90.0 90.0

Figure 4. Arrhenius plots of promising lithium superionic conductorcandidates from AIMD simulations. Solid lines indicate the initialLi3Y(PS4)2 and Li5PS4Cl2 candidates. Dashed lines indicate the furtheroptimized candidates obtained via aliovalent doping or substitution ofLi3Y(PS4)2 (Li3La(PS4)2 , Li 3 . 1 2 5Y0 . 8 7 5Ca0 . 1 2 5(PS4)2 andLi2.875Y0.875Zr0.125(PS4)2).

Table 3. Li+ Ionic Conductivity (σ300 K) and Diffusivity(D300 K) at 300 K, Error Range of σ300 K and ActivationEnergy (Ea) for Li3Y(PS4)2 and Li5PS4Cl2 from AIMDSimulations

Formulaσ300 K

(mS/cm)Error range of σ300 K

(mS/cm)Ea

(meV)D300 K(cm2/s)

Li3Y(PS4)2 2.16 [1.46, 3.19] 278 3.56 × 10−8

Li5PS4Cl2 1.85 [1.38, 2.47] 304 1.36 × 10−8

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

E

cathode interface. The dominant product at the anode is Li2S inall cases, which is a good electronic insulator and reasonable Liconductor, especially as an amorphous interphase. The otherproducts at the anode are YP, a semiconductor with band gapof ∼1 eV,57 and Li3P. On the solid electrolyte/charged cathode,P2S5 is always predicted to be one of the products. However,the presence of S2Cl2 and PCl5 at the LPSCl/cathode interfacemay prove problematic in real-world applications as theyundergo hydrolysis readily to form HCl. For reference, thepredicted interfacial equilibria in the 0−5 V vs Li/Li+ for bothcandidates are given in Figure S4 in the SupportingInformation, and we note that the observations made aboveapply for typical cutoff voltages in rechargeable lithium-ionbatteries.For comparison, Table 4 also presents the predicted phase

equilibria for the Li10Si1.5P1.5S11.5Cl0.5 model of the recentlyreported Li9.54Si1.74P1.44S11.7Cl0.3 superionic conductor

16 as wellas Li10GeP2S12

3 and Li7P3S11.13 Similar to the candidates

identified in this work, Li2S is predicted to be the dominantproduct at the anode/electrolyte interface in all instances, withthe small band gap Li3P comprising a relatively small fraction.For Li10Si1.5P1.5S11.5Cl0.5 and Li10GeP2S12, there is an additionalLi21Si5 or Li15Ge4 phase, which also have a small band gap,

consistent with previous experimental studies.18,19 At thecathode/Li10Si1.5P1.5S11.5Cl0.5 interface, S2Cl2 is predicted tobe one of the products, though the proportion is much less incomparison to LPSCl due to the much lower content of Cl.

■ DISCUSSIONFrom the results in the preceding sections, Li3Y(PS4)2 (LYPS)and Li5PS4Cl2 (LPSCl) have emerged as promising novellithium superionic conductors based on a comprehensivescreening of the Li−P−S and Li−M−P−S chemical spaces.Both candidates exhibit good phase stability (low Ehull) andexcellent topological characteristics (>1D large conductionchannels), and are predicted to be electronic insulators withhigh Li+ conductivities (exceeding 1 mS/cm). The Li+

conduction mechanisms and migration barriers were elucidatedusing CI-NEB calculations, and the results further confirm thepredictions from the AIMD simulations. In addition, bothcandidates comprise entirely of earth-abundant elements,making them practical from a cost perspective.Both candidates are derived from the replacement of Ag with

Li in known quaternary Ag thiophosphates in the ICSD. We seethis as a further positive attribute of the two candidates as ionexchange from the known Ag-based compounds is therefore apotential initial synthesis route that can be explored. Forexample, ion exchange has similarly been used to synthesize thewell-known Li7P3S11 superionic conductor from Ag7P3S11.

22 Wespeculate that due to the significantly larger ionic radii of Ag(129 pm) compared to Li (90 pm), Li-substituted Agcompounds may present large percolating voids conducive tofast 3D Li mobility. Such a strategy can certainly be expandedbeyond just the thiophosphate chemistries that are the focus ofthis work. However, we would point out that the large ionicradii difference between Ag+ and Li+ can potentially lead toincompatibility of Li with the Ag-based host framework, whichis why a computational assessment of phase stability is a criticalfirst step to determine the likelihood of synthesis. Also, not allAg compounds have percolating 3D diffusion networks ofsufficient channel size. Here again, the efficient tiered screeningapproach outlined in this work based on inexpensivetopological analysis followed by more computationally intensivefirst-principles calculations can provide useful guidelines.Between the two candidates, it is our belief that LYPS is the

more promising one. Not only is LYPS predicted to have amarginally higher Li+ conductivity than LPSCl in AIMDsimulations, it is also predicted to be significantly more stable(Ehull = 2 meV/atom) and its lack of Cl means that there islikely to be fewer issues with reaction products at highervoltages. We will note that like all sulfide-based materials, airand moisture stability may be a potential area of concern,8

though this limitation has not prevented the development ofprototype all-solid-state rechargeable lithium-ion batteriesutilizing other sulfide solid electrolytes. Like other sulfides,both materials are predicted to be relatively soft (seeSupporting Information), which should make it easier toachieve low porosity using cold-pressing methods.

Further Optimization of LYPS. To explore if furtherenhancement of the conductivity of LYPS is possible, weperformed isolvalent substitutions and aliovalent doping ofLYPS. La3+ was examined as a potential substitute for Y3+ dueto its slightly larger ionic radii (117 pm compared to 104 pmfor Y3+). The computed Ehull for Li3La(PS4)2 is 20 meV,significantly higher than LYPS, and its ionic conductivity is onlyslightly higher at 3.27 mS/cm with a slightly lower activation

Figure 5. Isosurfaces of Li+ probability density distribution P (lightblue) for Li3Y(PS4)2 and Li5PS4Cl2 from AIMD simulations at 800 Kwith P = 0.0001 a0

−3 (a0 is the Bohr radius).

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

F

energy of 263 meV (see Figure 4), which can be attributed tothe ∼5% volume expansion and increase in channel size (1.92Å).Unlike isovalent substitutions, aliovalent doping can have the

additional effect of introducing Li+ vacancies or interstitials.Both Ca2+ or Zr4+ dopants that have comparable ionic radii toY3+ were explored using a 1 × 1 × 2 supercell of LYPS, with theintroduction of Li+ interstitials and vacancies, respectively.Table 5 summarizes the dopant formation energies and room-temperature Li+ conductivities for the doped structures. BothCa2+ and Zr4+ were found to have reasonably low dopantformation energies of 0.63 and 0.26 eV, respectively. FromAIMD simulations, we find that aliovalent doping of LYPS withthe introduction of either vacancies or interstitials can lead tomultifold increases in its ionic conductivity. Substitution of12.5% of Y3+ with Ca2+ and Zr4+ leads to extrapolated roomtemperature conductivities of 7.14 and 5.25 mS/cm,respectively, with corresponding decreases in activationenergies to 231 and 241 meV, respectively (see Figure 4).Due to computational cost considerations, our explorations

of dopant optimization is limited by the size of the supercellaccessible within AIMD simulations. Nevertheless, the dopingresults are a proof of concept that there is significant scope forfurther fine-tuning of dopant and Li concentration in LYPS toachieve even higher conductivities, a claim that we hope will beverified by experimental efforts at synthesizing undoped anddoped LYPS.Comparison with Other State-of-the-Art Superionic

Conductors. In comparison with state-of-the-art sulfidesuperionic conductors such as Li7P3S11,

13 the LGPS family(Li10GeP2S12 and Li9.54Si1.74P1.44S11.7Cl0.3,

3,16 LYPS (undoped or

doped) has slightly lower Li+ conductivity. However, bulk ionicconductivity is no longer the critical factor in all-solid-statebattery performance beyond 1 mS/cm. Indeed, other propertiessuch as interfacial stability play a far more critical role. Forinstance, though the Li9.54Si1.74P1.44S11.7Cl0.3 superionic con-ductor recently reported by Kato et al.16 has an extraordinarilyhigh room temperature ionic conductivity of 25 mS/cm, itsinterfacial stability is much poorer than the Li9.6P3S12composition in the same structure, which has a lowerconductivity of ∼1 mS/cm.16 The result is that Li4Ti5O12,which has a voltage of 1.5 V against Li/Li+, had to be used asthe anode with Li9.54Si1.74P1.44S11.7Cl0.3, lowering achievableenergy densities due to the low overall operating voltage of∼2.5 V. In contrast, full cell performance at a relatively highoperating voltage of up to 4.2 V was demonstrated for Li9.6P3S12with standard graphitic anodes.LYPS compares favorably to these known superionic

conductors in terms of both phase and electrochemical stability.The calculated Ehull of LYPS is only 2 meV/atom, substantiallylower than that of Li7P3S11 (21 meV/atom),14 Li10GeP2S12 (25meV/atom)39 and Li10Si1.5P1.5S11.5Cl0.5 (30 meV/atom).Recently, Richards et al. also reported computational evidenceof extraordinarily high Li+ conductivities exceeding 50 mS/cmin the Li1+2xZn1−xPS4 solid solution,54 a compound that wasalso considered in our screening. However, these highconductivities were obtained only with the introduction of alarge number of defects, requiring high predicted synthesistemperatures exceeding 950 K. In comparison, doped LYPSwith conductivities of up to 7 mS/cm still maintains a relativelylow Ehull and small dopant formation energies.

Figure 6. (a) Investigated Li vacancy diffusion paths in Li3Y(PS4)2 viewed along b direction. Symmetrically distinct Li1 and Li2 are represented bygreen and blue spheres, respectively. Celadon polyhedra and gray tetrahedra indicate YS6 and PS4. (b−d) Calculated CI-NEB migration barriers forselected percolating paths.

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

G

In terms of interfacial stability, there are no reaction productsof major concern at the cathode/LYPS interface, unlikeLi10Si1.5P1.5S11.5Cl0.5 where the presence of Cl is predicted toresult in the formation of S2Cl2. On the anode/LYPS interface,the presence of the semiconducting YP phase may be ofpotential concern, though its band gap (∼1 eV)57 is still higherthan the Li−Si alloys (e.g., band gaps of 0.6 eV for Li12Si7

58 and0.08 eV for Li7Si3

59) predicted to form at the anode/Li10Si1.5P1.5S11.5Cl0.5 interface. An electrically insulating interfaceis desired for passivation to avoid further propagation of thereaction front. Furthermore, Li−Si alloys are also well-knownto undergo significant volume expansion (in excess of 300%) athigh lithiation, which may be detrimental to maintainingintimate electrode/electrolyte contact.6

In summary, we speculate that LYPS may present an overallbetter balance of properties as a lithium superionic conductorsolid electrolyte for all-solid-state battery applications. It hasclearly better predicted phase stability, and likely betterinterfacial stability based on the predicted phase equilibria at

the electrode/electrolyte interface. Its conductivity, thoughsomewhat lower than some of the state-of-the-art candidates, issufficiently high that it is not likely to be a limiting factor, andcan potentially be further improved with the demonstrateddoping strategies.

■ CONCLUSIONS

To conclude, we have performed a comprehensive screening ofthe Li−P−S and Li−M−P−S chemical spaces for novel lithiumsuperionic conductors, including candidates obtained byreplacing Ag with Li in Ag−P−S and Ag−M−P−S compounds.By combining rapid topological evaluation with a short AIMDscreening, a large number of candidates that are unlikely tosatisfy the stringent Li+ conductivity requirements are excludedusing a minimum amount of computational resources.Li3Y(PS4)2 (LYPS) and Li5PS4Cl2 (LPSCl) emerged as themost promising candidates with good phase and electro-chemical stability, high 3D ionic conductivity and lowelectronic conductivity. We performed further optimization

Figure 7. Investigated Li vacancy diffusion paths in Li5PS4Cl2 viewed along (a) a direction; (b) b direction. Green, blue, red spheres represent Li1,Li2 and Cl atoms, respectively. Gray tetrahedra are PS4 tetrahedra. (c−e) Calculated CI-NEB migration barriers for selected paths.

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

H

on LYPS, the more promising of the two candidates, anddemonstrate that the Li+ conductivity can be further increasedfrom 2.16 mS/cm to 7.14 and 5.25 mS/cm via 12.5% doping ofY3+ with Ca2+ and Zr4+, respectively. The LYPS frameworktherefore presents an intriguing candidate ripe for further

optimization as a lithium superionic conductor solid electrolytefor next-generation all-solid-state rechargeable lithium-ionbatteries.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.chemma-ter.6b04049.

Details of computational methods; derivation of meansquare displacement ratio criteria; modeling details ofLi10Si1.5P1.5S11.5Cl0.5; CI-NEB barriers for other paths inLi3Y(PS4)2 and Li5PS4Cl2; the distinct-part of van Hovecorrelation function; Li grand potential phase stabilityplots; elastic constants for Li3Y(PS4)2 and Li5PS4Cl2(PDF)Crystallographic information files (CIFs) of the DFT-relaxed structures of Li3Y(PS4)2 and Li5PS4Cl2 (ZIP)

■ AUTHOR INFORMATIONCorresponding Author*S. P. Ong. E-mail: ongsp@eng.ucsd.edu.ORCIDShyue Ping Ong: 0000-0001-5726-2587NotesThe authors declare the following competing financialinterest(s): Zhuoying Zhu, Iek-Heng Chu and Shyue PingOng have filed a provisional patent on the high- throughputscreening methodology and the new lithium superionicconductors identified in this work.

■ ACKNOWLEDGMENTSThis work was supported by the U.S. Department of Energy,Office of Science, Basic Energy Sciences under Award No. DE-SC0012118. We also acknowledge computational resourcesprovided by Triton Shared Computing Cluster (TSCC) at theUniversity of California, San Diego, the National EnergyResearch Scientific Computing Center (NERSC), and theExtreme Science and Engineering Discovery Environment(XSEDE) supported by National Science Foundation underGrant No. ACI-1053575.

■ REFERENCES(1) Tarascon, J.-M.; Armand, M. Issues and challenges facingrechargeable lithium batteries. Nature 2001, 414, 359−367.(2) Goodenough, J. B.; Park, K.-S. S. The Li-ion rechargeable battery:A perspective. J. Am. Chem. Soc. 2013, 135, 1167−76.(3) Kamaya, N.; Homma, K.; Yamakawa, Y.; Hirayama, M.; Kanno,R.; Yonemura, M.; Kamiyama, T.; Kato, Y.; Hama, S.; Kawamoto, K.;Mitsui, A. A lithium superionic conductor. Nat. Mater. 2011, 10, 682−686.(4) Thangadurai, V.; Weppner, W. Recent progress in solid oxide andlithium ion conducting electrolytes research. Ionics 2006, 12, 81−92.(5) Miara, L. J.; Ong, S. P.; Mo, Y.; Richards, W. D.; Park, Y.; Lee, J.-m.; Lee, H. S.; Ceder, G. Effect of Rb and Ta Doping on the Ionic

Figure 8. Calculated element-projected density of states for (a)Li3Y(PS4)2 (band gap Eg = 3.41 eV) and (b) Li5PS4Cl2 (Eg = 3.57 eV)using the HSE screened hybrid functional.

Table 4. Phase Equilibria for the Li3Y(PS4)2 and Li5PS4Cl2Candidates as well as Other Known Superionic Conductorsat 5 V Cathode and Metallic Li Anodea

ElectrolytePhase equilibria at 5 V

cathode Phase equilibria at anode

μLi = (μLi° − 5) eV μLi = μLi° eV

Promising candidatesLi3Y(PS4)2 YPS4+ 0.5 P2S5 + 1.5 S YP + Li3P + 8 Li2SLi5PS4Cl2 0.067 PCl5 + 0.833 S2Cl2

+ 0.467 P2S5Li3P + 4 Li2S + 2 LiCl

Known superionic conductorsLi10Si1.5P1.5S11.5Cl0.5 0.75 P2S7 + 0.25 S2Cl2 +

1.5 SiS2 + 2.75 S1.5 Li3P + 0.3 Li21Si5 + 0.5LiCl + 11.5 Li2S

Li10GeP2S12 P2S5 + GeS2 + 5 S 2 Li3P + 0.25 Li15Ge4 + 12Li2S

Li7P3S11 1.5 P2S5 + 3.5 S 3 Li3P + 11 Li2SaCoefficients given are per formula unit of solid electrolyte.

Table 5. Calculated Dopant Formation Energy (Ef), Energy above Hull (Ehull), and Room-Temperature Li+ Ionic Conductivity(σ300 K) and Activation Energy (Ea) for Aliovalent Doped Li3Y(PS4)2

Dopant Formula Ef (eV) Ehull (meV/atom) σ300 K (mS/cm) error range of σ300 K (mS/cm) Ea (meV)

Ca Li3.125Y0.875Ca0.125(PS4)2 0.63 6 7.14 [4.67, 10.92] 231Zr Li2.875Y0.875Zr0.125(PS4)2 0.26 4 5.25 [3.77, 7.31] 241

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

I

Conductivity and Stability of the Garnet Li7+2x‑y(La3‑xRbx) (Zr2‑yTay)-O12 (0 ≤ x ≤ 0.375, 0 ≤ y ≤ 1) Superionic Conductor: A FirstPrinciples Investigation. Chem. Mater. 2013, 25, 3048−3055.(6) Deng, Z.; Mo, Y.; Ong, S. P. Computational studies of solid-statealkali conduction in rechargeable alkali-ion batteries. NPG Asia Mater.2016, 8, e254.(7) Wang, Y.; Richards, W. D.; Ong, S. P.; Miara, L. J.; Kim, J. C.;Mo, Y.; Ceder, G. Design principles for solid-state lithium superionicconductors. Nat. Mater. 2015, 14, 1026−1031.(8) Radhakrishnan, B.; Ong, S. P. Aqueous stability of alkalisuperionic conductors from first principles calculations. Front. EnergyRes. 2016, 4, 1−12.(9) Ong, S. P.; Mo, Y.; Richards, W. D.; Miara, L.; Lee, H. S.; Ceder,G. Phase stability, electrochemical stability and ionic conductivity ofthe Li10±1MP2X12 (M = Ge, Si, Sn, Al or P, and X = O, S or Se) familyof superionic conductors. Energy Environ. Sci. 2013, 6, 148−156.(10) Bron, P.; Johansson, S.; Zick, K.; Schmedt auf der Gunne, J. S.;Dehnen, S. S.; Roling, B. Li10SnP2S12 - An Affordable LithiumSuperionic Conductor. J. Am. Chem. Soc. 2013, 135, 15694−15697.(11) Kuhn, A.; Gerbig, O.; Zhu, C.; Falkenberg, F.; Maier, J.; Lotsch,B. V. A new ultrafast superionic Li-conductor: ion dynamics inLi11Si2PS12 and comparison with other tetragonal LGPS-type electro-lytes. Phys. Chem. Chem. Phys. 2014, 16, 14669−74.(12) Whiteley, J. M.; Woo, J. H.; Hu, E.; Nam, K.-W.; Lee, S.-H.Empowering the Lithium Metal Battery through a Silicon-BasedSuperionic Conductor. J. Electrochem. Soc. 2014, 161, A1812−A1817.(13) Seino, Y.; Ota, T.; Takada, K.; Hayashi, A.; Tatsumisago, M. Asulphide lithium super ion conductor is superior to liquid ionconductors for use in rechargeable batteries. Energy Environ. Sci. 2014,7, 627−631.(14) Chu, I.-H.; Nguyen, H.; Hy, S.; Lin, Y.-C.; Wang, Z.; Xu, Z.;Deng, Z.; Meng, Y. S.; Ong, S. P. Insights into the Performance Limitsof the Li7P3S11 Superionic Conductor: A Combined First-Principlesand Experimental Study. ACS Appl. Mater. Interfaces 2016, 8, 7843−7853.(15) Xu, K. Electrolytes and interphases in Li-ion batteries andbeyond. Chem. Rev. 2014, 114, 11503−618.(16) Kato, Y.; Hori, S.; Saito, T.; Suzuki, K.; Hirayama, M.; Mitsui,A.; Yonemura, M.; Iba, H.; Kanno, R. High-power all-solid-statebatteries using sulfide superionic conductors. Nature Energy 2016, 1,16030.(17) Verma, P.; Maire, P.; Novak, P. A review of the features andanalyses of the solid electrolyte interphase in Li-ion batteries.Electrochim. Acta 2010, 55, 6332−6341.(18) Wenzel, S.; Randau, S.; Leichtweiß, T.; Weber, D. A.; Sann, J.;Zeier, W. G.; Janek, J. Direct Observation of the Interfacial Instabilityof the Fast Ionic Conductor Li10GeP2S12 at the Lithium MetalAnode. Chem. Mater. 2016, 28, 2400−2407.(19) Tarhouchi, I.; Viallet, V.; Vinatier, P.; Menetrier, M.Electrochemical characterization of Li10SnP2S12: An electrolyte or anegative electrode for solid state Li-ion batteries? Solid State Ionics2016, 296, 18−25.(20) Sakuda, A.; Hayashi, A.; Tatsumisago, M. InterfacialObservation between LiCoO2 Electrode and Li2S-P2S5 Solid Electro-lytes of All-Solid-State Lithium Secondary Batteries Using Trans-mission Electron Microscopy. Chem. Mater. 2010, 22, 949−956.(21) Mizuno, F.; Iba, H. Research on Electrode-Electrolyte Interfacesof Innovative New Generation Batteries. Meeting Abstracts 2012, 1196.(22) Yamane, H.; Shibata, M.; Shimane, Y.; Junke, T.; Seino, Y.;Adams, S.; Minami, K.; Hayashi, A.; Tatsumisago, M. Crystal structureof a superionic conductor, Li7P3S11. Solid State Ionics 2007, 178, 1163−1167.(23) Boulineau, S.; Courty, M.; Tarascon, J.-M.; Viallet, V.Mechanochemical synthesis of Li-argyrodite Li6PS5X (X = Cl, Br, I)as sulfur-based solid electrolytes for all solid state batteries application.Solid State Ionics 2012, 221, 1−5.(24) Rayavarapu, P. R.; Sharma, N.; Peterson, V. K.; Adams, S.Variation in structure and Li+-ion migration in argyrodite-type Li6PS5X

(X = Cl, Br, I) solid electrolytes. J. Solid State Electrochem. 2012, 16,1807−1813.(25) Parrinello, M.; Rahman, A.; Vashishta, P. Structural transitionsin superionic conductors. Phys. Rev. Lett. 1983, 50, 1073−1076.(26) Hosono, M.; Kawamura, J.; Itoigawa, H. Structure and ionicconductivity of rapidly quenched AgI-Ag2WO4 superionic conductorglasses. J. Non-Cryst. Solids 1999, 244, 81−88.(27) Belin, R.; Zerouale, A.; Pradel, A.; Ribes, M. Ion dynamics in theargyrodite compound Ag7GeSe5I: Non-Arrhenius behavior andcomplete conductivity spectra. Solid State Ionics 2001, 143, 445−455.(28) Bergerhoff, G.; Hundt, R.; Sievers, R.; Brown, I. D. Theinorganic crystal structure data base. J. Chem. Inf. Model. 1983, 23, 66−69.(29) Ong, S. P.; Richards, W. D.; Jain, A.; Hautier, G.; Kocher, M.;Cholia, S.; Gunter, D.; Chevrier, V. L.; Persson, K. a.; Ceder, G.Python Materials Genomics (pymatgen): A robust, open-sourcepython library for materials analysis. Comput. Mater. Sci. 2013, 68,314−319.(30) Kresse, G.; Furthmuller, J. Efficient iterative schemes for abinitio total-energy calculations using a plane-wave basis set. Phys. Rev.B: Condens. Matter Mater. Phys. 1996, 54, 11169−11186.(31) Blochl, P. E. Projector augmented-wave method. Phys. Rev. B:Condens. Matter Mater. Phys. 1994, 50, 17953−17979.(32) Perdew, J. P. J.; Burke, K.; Ernzerhof, M. Generalized GradientApproximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865−3868.(33) Jain, A.; Ong, S. P.; Hautier, G.; Chen, W.; Richards, W. D.;Dacek, S.; Cholia, S.; Gunter, D.; Skinner, D.; Ceder, G.; Persson, K.A. Commentary: The Materials Project: A materials genome approachto accelerating materials innovation. APL Mater. 2013, 1, 011002.(34) Ong, S. P.; Wang, L.; Kang, B.; Ceder, G. Li-Fe-P-O2 phasediagram from first principles calculations. Chem. Mater. 2008, 20,1798−1807.(35) Ong, S. P.; Cholia, S.; Jain, A.; Brafman, M.; Gunter, D.; Ceder,G.; Persson, K. a. The Materials Application Programming Interface(API): A simple, flexible and efficient API for materials data based onREpresentational State Transfer (REST) principles. Comput. Mater.Sci. 2015, 97, 209−215.(36) Zhu, Z.; Chu, I.-H.; Deng, Z.; Ong, S. P. Role of Na+Interstitials and Dopants in Enhancing the Na+ Conductivity of theCubic Na3PS4 Superionic Conductor. Chem. Mater. 2015, 27, 8318−8325.(37) Nose, S. A unified formulation of the constant temperaturemolecular dynamics methods. J. Chem. Phys. 1984, 81, 511.(38) Hoover, W. G. Canonical dynamics: Equilibrium phase-spacedistributions. Phys. Rev. A: At., Mol., Opt. Phys. 1985, 31, 1695−1697.(39) Mo, Y.; Ong, S. P.; Ceder, G. First Principles Study of theLi10GeP2S12 Lithium Super Ionic Conductor Material. Chem. Mater.2012, 24, 15−17.(40) Heyd, J.; Scuseria, G. E.; Ernzerhof, M. Hybrid functionals basedon a screened Coulomb potential. J. Chem. Phys. 2003, 118, 8207.(41) Heyd, J.; Scuseria, G. E.; Ernzerhof, M. Erratum: “Hybridfunctionals based on a screened Coulomb potential” [J. Chem. Phys.118, 8207 (2003)]. J. Chem. Phys. 2006, 124, 219906.(42) Heyd, J.; Scuseria, G. E. Efficient hybrid density functionalcalculations in solids: Assessment of the Heyd−Scuseria−Ernzerhofscreened Coulomb hybrid functional. J. Chem. Phys. 2004, 121, 1187.(43) Hautier, G.; Jain, A.; Ong, S. P.; Kang, B.; Moore, C.; Doe, R.;Ceder, G. ChemInform Abstract: Phosphates as Lithium-Ion BatteryCathodes: An Evaluation Based on High-Throughput ab initioCalculations. Chem. Mater. 2011, 23, 3495−3508.(44) Wang, Z.; Chu, I.-h.; Zhou, F.; Ong, S. P. Electronic StructureDescriptor for the Discovery of Narrow-Band Red-EmittingPhosphors. Chem. Mater. 2016, 28, 4024−4031.(45) Watanabe, T.; Sholl, D. S. Accelerating applications of metal-organic frameworks for gas adsorption and separation by computa-tional screening of materials. Langmuir 2012, 28, 14114−14128.(46) Meutzner, F.; Munchgesang, W.; Kabanova, N. A.; Zschornak,M.; Leisegang, T.; Blatov, V. A.; Meyer, D. C. On the Way to NewPossible Na-Ion Conductors: The Voronoi-Dirichlet Approach, Data

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

J

Mining and Symmetry Considerations in Ternary Na Oxides. Chem. -Eur. J. 2015, 21, 16601−16608.(47) Martin, R. L.; Smit, B.; Haranczyk, M. Addressing Challenges ofIdentifying Geometrically Diverse Sets of Crystalline Porous Materials.J. Chem. Inf. Model. 2012, 52, 308−318.(48) Willems, T. F.; Rycroft, C. H.; Kazi, M.; Meza, J. C.; Haranczyk,M. Algorithms and tools for high-throughput geometry-based analysisof crystalline porous materials. Microporous Mesoporous Mater. 2012,149, 134−141.(49) Zhao, Y.; Daemen, L. L. Superionic Conductivity in Lithium-Rich Anti-Perovskites. J. Am. Chem. Soc. 2012, 134, 15042−15047.(50) Hayashi, A.; Noi, K.; Sakuda, A.; Tatsumisago, M. Superionicglass-ceramic electrolytes for room-temperature rechargeable sodiumbatteries. Nat. Commun. 2012, 3, 856.(51) Deng, Z.; Radhakrishnan, B.; Ong, S. P. Rational CompositionOptimization of the Lithium-Rich Li3OCl1‑xBrx Anti-PerovskiteSuperionic Conductors. Chem. Mater. 2015, 27, 3749−3755.(52) Lu, X.; Wu, G.; Howard, J. W.; Chen, A.; Zhao, Y.; Daemen, L.L.; Jia, Q. Li-rich anti-perovskite Li3OCl films with enhanced ionicconductivity. Chem. Commun. 2014, 50, 11520−2.(53) Chu, I.-H.; Kompella, C. S.; Nguyen, H.; Zhu, Z.; Hy, S.; Deng,Z.; Meng, Y. S.; Ong, S. P. Room-Temperature All-solid-stateRechargeable Sodium-ion Batteries with a Cl-doped Na3PS4Superionic Conductor. Sci. Rep. 2016, 6, 33733.(54) Richards, W. D.; Wang, Y.; Miara, L. J.; Kim, J. C.; Ceder, G.Design of Li1+2xZn1−xPS4, a new lithium ion conductor. Energy Environ.Sci. 2016, 9, 3272−3278.(55) Richards, W. D.; Miara, L. J.; Wang, Y.; Kim, J. C.; Ceder, G.Interface Stability in Solid-State Batteries. Chem. Mater. 2016, 28,266−273.(56) Zhu, Y.; He, X.; Mo, Y. Origin of Outstanding Stability in theLithium Solid Electrolyte Materials: Insights from ThermodynamicAnalyses Based on First-Principles Calculations. ACS Appl. Mater.Interfaces 2015, 7, 23685−23693.(57) Meng, J.; Ren, Y. Studies on the electrical properties of rareearth monophosphides. J. Solid State Chem. 1991, 95, 346−351.(58) Nesper, R.; von Schnering, H. G.; Curda, J. Li12Si7, eineVerbindung mit trigonal-planaren Si4-Clustern und isometrischen Si5-Ringen. Chem. Ber. 1986, 119, 3576−3590.(59) von Schnering, N.; Nesper, R.; Tebbe, K.; Kurda, J. Structureand properties of Li14Si6 (Li2.33Si), the violet phase in the lithium−silicon system. Z. Metallkd. 1980, 71, 357−363.

Chemistry of Materials Article

DOI: 10.1021/acs.chemmater.6b04049Chem. Mater. XXXX, XXX, XXX−XXX

K