With the goal of finding new lithium solid electrolytes by a combined computational-experimental method, the exploration of the Li-Al-O-S phase field resulted in the discovery of a new sulfide Li3AlS3. The structure of the new phase was determined through an approach combining synchrotron X-ray and neutron diffraction with 6Li and 27Al magic-angle spinning nuclear magnetic resonance spectroscopy and revealed to be a highly ordered cationic polyhedral network within a sulfide anion hcp-type sublattice. The originality of the structure relies on the presence of Al2S6 repeating dimer units consisting of two edge-shared Al tetrahedra. We find that, in this structure type consisting of alternating tetrahedral layers with Li-only polyhedra layers, the formation of these dimers is constrained by the Al/S ratio of 1/3. Moreover, by comparing this structure to similar phases such as Li5AlS4 and Li4.4Al0.2Ge0.3S4 ((Al + Ge)/S = 1/4), we discovered that the AlS4 dimers not only influence atomic displacements and Li polyhedral distortions but also determine the overall Li polyhedral arrangement within the hcp lattice, leading to the presence of highly ordered vacancies in both the tetrahedral and Li-only layer. AC impedance measurements revealed a low lithium mobility, which is strongly impacted by the presence of ordered vacancies. Finally, a composition-structure-property relationship understanding was developed to explain the extent of lithium mobility in this structure type.
With the goal of finding new lithium solid electrolytes by a combined computational-experimental method, the exploration of the Li-Al-O-S phase field resulted in the discovery of a new sulfide Li3AlS3. The structure of the new phase was determined through an approach combining synchrotron X-ray and neutron diffraction with 6Li and 27Al magic-angle spinning nuclear magnetic resonance spectroscopy and revealed to be a highly ordered cationic polyhedral network within a sulfide anion hcp-type sublattice. The originality of the structure relies on the presence of Al2S6 repeating dimer units consisting of two edge-shared Al tetrahedra. We find that, in this structure type consisting of alternating tetrahedral layers with Li-only polyhedra layers, the formation of these dimers is constrained by the Al/S ratio of 1/3. Moreover, by comparing this structure to similar phases such as Li5AlS4 and Li4.4Al0.2Ge0.3S4 ((Al + Ge)/S = 1/4), we discovered that the AlS4 dimers not only influence atomic displacements and Li polyhedral distortions but also determine the overall Li polyhedral arrangement within the hcp lattice, leading to the presence of highly ordered vacancies in both the tetrahedral and Li-only layer. AC impedance measurements revealed a low lithium mobility, which is strongly impacted by the presence of ordered vacancies. Finally, a composition-structure-property relationship understanding was developed to explain the extent of lithium mobility in this structure type.
All solid-state batteries
(ASSBs) are of considerable current interest
because they are a potential route to the use of lithium metal anodes
while avoiding dendrite formation.[1,2] Solid-state
electrolytes (SSEs) offer advantages over liquid electrolytes, such
as large electrochemical stability windows and better thermal stability,[3,4] and have a lithium transport number of unity. Moreover, inorganic
crystalline lithium ion conductors have superior ionic conductivity
compared to organic polymers.[5] Recent reviews
compared the different families of Li electrolytes,[5−9] and sulfides have the highest Li ion conductivity.
For instance, members of the thio-LISICON family[10−14] such as the Li7P3S11 crystalline phase discovered in the Li2S–P2S5 system[15,16] and the Li–argyrodites
Li6PS5X (X = Cl, Br, I)[17,18] have superior ionic conductivities, σ, ∼1 to 20mScm–1 at room
temperature (RT) and low activation energies, Ea, ∼0.20 eV. These values are competitive with those
of liquid electrolytes and make them promising candidates for integration
in ASSBs. The sulfide anion is larger than oxide and more polarizable
and affords lower frequency lattice vibrations, all favoring higher
lithium mobility,[19−21] while superior mechanical properties offer lower
grain boundary resistance[6] and allow easier
processing into dense pellets. While these characteristics indicate
superiority of sulfides over oxides, the usually narrower electrochemical
stability window of the former can lead to parasitic reactions at
the electrodes and considerably limit the performance and life cycle
of ASSBs containing sulfide electrolytes.[5,11,22] Moreover, sulfide materials are often not
stable in air, which makes them more difficult to handle. Further
improvement of SSEs is needed for their commercialization, which implies
finding materials showing good ionic conductivity as well as being
able to form stable and conducting interfaces at the cathode and anode
sides.[8,14,23]Given
the large number of candidate material compositions for potential
new SSEs, computation has been extensively used to guide experimental
work. On the one hand, some studies focus on currently known materials
in order to produce an understanding of the underlying mechanisms
for ion transport thanks to density functional theory (DFT) energy
landscape methods and molecular dynamics.[20,24,25] The need for higher throughput screening
methods led to database-driven approaches where models of ion transport,
such as the development of predictive performance metrics,[26] are used to prioritize existing materials for
evaluation as SSE. This can be done for example through the use of
machine learning algorithms combined with first-principles calculations[27,28] or with bond valence mapping analysis.[18,29−31] On the other hand, the discovery of new materials
with unique compositions and/or structures can be accelerated by predicting
the most stable compounds within a chosen phase field prior to experimentation.
The crystal structure for which the energy is calculated can be either
thought to be isostructural to a known material in which original
substitutions are performed[32−34] or can be determined through
crystal structure prediction techniques.[35−37]A relatively
few mixed anion systems have been studied as SSE.
Such materials can combine the advantages of each individual anion
and may also offer new structure types for ion transport. The good
performance of argyrodite sulfide halide materials demonstrates the
potential of such mixed anion systems.[18,38] Oxysulfide
phases, especially for lithium solid electrolyte applications, however
remain less explored[39−42] and could be a promising approach to yield new materials. In particular,
the oxygen for sulfur substitution in LGPS structures has been an
area of investigation.[40,42] For instance, Kim et al.[42] reported the study of the Li10SiP2S12–O (0 < x < 1.75) solid solution, which
resulted in an optimal conductivity for x = 0.7.
Another interesting example is the theoretical study by Wang et al.,
presenting O doping into β-Li3PS4, which
showed that the oxygen insertion enabled lower activation energy barriers,
improved electrochemical stability, and created a beneficial 3D connection
pathway for Li diffusion.[41] Moreover, among
non-crystalline materials, oxysulfide glasses such as combinations
of Li2S–P2S5–P2O5[43] or Li2S–SiS2–LiBO2[44] have been studied as potential lithium solid electrolytes
and showed promising properties.In this study, we evaluate
oxysulfides of aluminum as candidate
lithium ion conductors due to the earth abundance, low cost and toxicity,
high polarizability, and redox inactivity. The composition LiAlOS
was previously identified as a new potential interesting solid electrolyte
target in a computation-only study.[37] The
present study of the Li–Al–O–S phase diagram
was guided by computational evaluation of the stability of candidate
compositions, which both highlights where new phases are likely to
be found and allows assessment of the amount of experimental effort
to be invested at each composition. Within this phase diagram, the
new sulfide phase Li3AlS3 was identified and
its structure and lithium ion conductivity were characterized.
Experimental Section
Synthesis
Materials
Li2S (99.98%),
LiNO3 (>99%), Li2CO3 (>99%), 7Li2CO3 (99% 7Li), and Al(OH)3 (reagent grade) were purchased from Sigma Aldrich, while
Li2O (99.5%), urea (99.0%), Al2S3 (99%+), and Al(NO3)3·9H2O
(98%) were obtained from Alfa Aesar.
Exploratory Synthesis of the Compounds in
the Li–Al–O–S Phase Field
For all sulfide-containing
materials, precursors and resulting powders were handled in a He-filled
glovebox. Compositions belonging to the Li–Al–O–S
phase diagram were synthesized from stoichiometric mixtures of Li2O, Al2S3, and, when needed, Li2S and pre-synthesized LiAlO2 (according to a procedure
from Gao et al.,[45]cf. the Supporting Information, SI). The
precursors were weighed in the appropriate amount in order to obtain
a total mass of 300 mg. The powders were then mixed and ground in
an agate mortar for 15 min, transferred to an alumina crucible, and
placed in a quartz tube before sealing under vacuum (10–4 mbar). The tube containing the sample was heated to 800 °C
at a ramp rate of 5 °C·min–1, held at
800 °C for 48 h, and then quenched in water. The resulting powder
was then manually ground in order to obtain a fine powder.
Final Synthesis of Li3AlS3
After identifying Li3AlS3 as
a new phase through the described synthesis method above, its synthesis
was slightly modified in order to improve purity. Li2S
(1.4358 g, 31.2 mmol) and Al2S3 (1.5642 g, 10.4
mmol) were weighed according to the stoichiometric 3:1 ratio. The
resulting powders were then mixed and ground in an agate mortar for
15 min, transferred in an alumina crucible, and placed in a quartz
tube before sealing under vacuum (10–4 mbar). The
tube containing the sample was heated at 700 °C for 12 h and
then to 800 °C for 12 h with an intermediate grinding step in
between both firings. The heating ramp rate of the furnace was 5 °C·min–1, and cooling was performed by quenching the tube
in water. Neutron powder diffraction (NPD) experiments were conducted
on 7Li-enriched samples of 7Li3AlS3, using 7Li2S as the precursor material,
which was synthesized according to a method described by Leube et
al.[46] starting from 7Li2CO3. For consistency of the structural analysis,
the 7Li3AlS3 sample was also used
for synchrotron diffraction experiments.The importance of the
quenching step was investigated by either letting the sample cool
down by turning off the furnace or by setting the cooling ramp rate
to 1 °C·min–1. The sample cooled
down by turning the furnace off did not show any extra impurity peaks
in the X-ray diffraction pattern; however, the sample cooled with
the 1 °C·min–1 ramp resulted in the partial
decomposition into Li5AlS4 and LiAlS2. This decomposition was also observed when a formerly prepared Li3AlS3 sample was heated slowly to 700 °C at
1 °C·min-1, maintained at this temperature
for 30 min, and quenched to room temperature. We therefore maintained
the 5 °C·min–1 fast heating ramp rate
and the quenching step for the synthesis of the material.
Probe Structure Generation and Energy Calculations
All energies were computed using periodic DFT with the VASP program.[47] The PBE functional was used[48] with the projector augmented wave approach to treat core
electrons.[49]A probe structure approach
was used to sample compositions in the Li+–Al3+–O2––S2– phase space.[36] Crystal structure prediction
(CSP) was used to generate a probe structure at each composition.
Each probe structure was assumed to have an energy close enough to
the global minimum energy structure to assess the thermodynamic stability
of a hypothetical compound at that composition against the formation
of an assemblage of known phases.The CSP was performed using
the in-house code ChemDASH (Chemically
Directed Atom Swap Hopping). Cells containing hexagonal close-packed
(hcp) and cubic close-packed (ccp) anion lattices hosting O2– and S2– were constructed, and some octahedral and tetrahedral interstitial
sites occupied by Li+ and Al3+ cations. The hcp cells contained eight anions, and the ccp cell contained nine anions, with a correct number of cations to
satisfy charge neutrality at each composition. Structures were initialized
with a random decoration of the anion and cation sublattices, and
their structures were optimized. To generate new structures, the positions
of some anions were swapped on the anion sublattice, or in the interstitial
sites, the positions of some cations were either swapped or moved
to previously vacant interstitial sites. The new structure was then
optimized by relaxation of the atomic positions to the nearest local
minimum. At each step, a Monte Carlo sampling algorithm was used to
accept swaps, which lowered the energy of the system or increased
it by an amount lower than the Monte Carlo energy threshold. The process
was continued until 1000 structures had been generated and the lowest
energy structure at each composition was taken forward for calculating
the stability.One of the features of ChemDASH is to perform
structural optimizations
in a number of stages, which can use different parameters. This was
done when optimizing each of the structures generated during the CSP
process, with each stage using an increasing level of accuracy. In
the first stage of each geometry optimization, Γ point-only
calculations were used with a plane wave cutoff of 400 eV. By the
final stage of each geometry optimization, a 2 × 2 × 2 k-point grid was used with a plane wave cutoff of 600 eV.
The cell vectors and atomic positions were optimized until forces
fell below 0.02 eV·Å–1.Once a probe
structure had been obtained, its energy was recalculated
at a more accurate level, which was also the level of accuracy used
to calculate the energies of previously reported phases. These energies
were used to generate the convex hulls of chemical stability. A plane
wave cutoff of 700 eV was used with a k-point spacing
of 0.15 Å–1. Cell vectors and atomic positions
were optimized until forces fell below 0.001 eV·Å–1. The convex hull of chemically stable compositions was generated
using pymatgen.[50]DFT calculations
were also performed on an ordered analogue of
the experimentally refined crystal structure of Li3AlS3. The split Li sites were merged onto a single high-symmetry
site, and all sites were given full occupancy. The structure was then
optimized in VASP using the more accurate parameters detailed above.
No imaginary frequency modes were found in phonon calculations, showing
that the structure is stable against displacement of ions from their
relaxed positions. The computed phonon frequencies are presented in Table S2 of the SI.
Elemental Analysis
Elemental analysis
of Li3AlS3 was performed by Mikroanalytishes
Labor Pascher at Remagen-Bandorf, Germany, after dissolution in a
HF/HCl solution at elevated temperature and pressure.
Diffraction
X-ray Diffraction
Synchrotron X-ray
diffraction (SXRD) was performed at Diamond Light Source UK, on high-resolution
beamline I11, at λ = 0.82465 Å. The sample was introduced
into a 0.7 mm diameter borosilicate glass capillary to record the
pattern in transmission mode [0 ° < 2θ < 150°, and Δ(2θ) = 0.004°]
using a high-resolution multianalyzer crystal (MAC) detector. The
experiment was performed at room temperature. For the Rietveld refinement,
the Thompson–Cox–Hastings function[54] with spherical harmonic expansion implemented in FullProf
was used to model the peak shape anisotropy.[55] In particular, the (−311), (−402), (−602),
and (331) reflections showed a pronounced anisotropic peak shape broadening,
which could be due to the presence of structural defects such as antiphase
domains and stacking faults.[56] The improvement
of the fit thanks to the use of spherical harmonics is illustrated
in Figure S3.
Neutron Diffraction
Time-of-flight
neutron powder diffraction (NPD) data were collected on Li3AlS3 using a high-resolution powder diffractometer (HRPD)
instrument at ISIS, UK. Experiments were carried out at ambient temperature
on the 7Li-enriched sample sealed in thin-walled vanadium
cans with a diameter of 8 mm, sealed with an indium gasket under 1
atm of helium gas. For the Rietveld refinement, all banks were fitted
simultaneously with the TOF pseudo-Voigt back-to-back exponential
function with spherical harmonic expansion as for the SXRD data refinement.
The structure determination was performed using Jana2006[68] in order to utilize the charge flipping method
implemented in the software to yield an initial model. FullProf[70] then had our preference for complete refinement.
Nuclear Magnetic Resonance (NMR) Spectroscopy
The 6Li magic-angle Spinning (MAS) NMR spectra were
recorded at 9.4 T on a Bruker DSX spectrometer using a 4 mm HXY MAS
probe (in double resonance mode) and at 20 T on a Bruker NEO spectrometer
using a 3.2 mm HXY MAS probe (in triple resonance mode). The 6Li MAS spectra were obtained at 9.4 T with a pulse length
of 3 μs at a radiofrequency (f) field amplitude of ω1/2π = 83 kHz and a MAS rate of ωr/2π
= 10 kHz and at 20 T with a pulse length of 4.5 μs at an rf
field amplitude of ω1/2π = 56 kHz and a MAS
rate of ωr/2π = 20 kHz. The 27Al
MAS NMR data were recorded at 9.4 T on a Bruker Avance III HD under
MAS at a rate of ωr/2π = 12 kHz using a 4 mm
HXY MAS probe (in double resonance mode) and at 20 T on a Bruker NEO
spectrometer using a 3.2 mm HXY MAS probe (in triple resonance mode).
The 27Al spectra were obtained at 9.4 T with a short pulse
angle of 30° of 0.33 μs duration at an rf amplitude of
ω1/2π = 83 kHz and at 20 T with a short pulse
angle of 30° of 0.55 μs duration at an rf amplitude of
ω1/2π = 50 kHz. The 27Al triple
quantum magic-angle spinning (MQMAS)[57] was
obtained at 9.4 T with a z-filtered sequence[58] and using rf field amplitudes of ω1/2π = 83 kHz for the excitation and reconversion pulses
and 4 kHz for the selective 90° pulse. All spectra were collected
at room temperature and obtained under quantitative recycle delays
of more than 5 times longer than the spin–lattice relaxation
times T1, which were measured using the
saturation recovery pulse sequence and fitted with a stretch exponential
function of the form 1 – exp[−(τ/T1)β] (with β ranging from 0.3 to
1). The 6Li and 27Al shifts were referenced
to 10 M LiCl in D2O and 0.1 M Al(NO3)3 in H2O at 0 ppm, respectively.
AC Impedance Spectroscopy
A pellet
of the Li3AlS3 powder was made by uniaxially
pressing ∼30 mg of powder in a 5 mm diameter cylindrical steel
die at a pressure of 125 MPa, followed by sintering in an evacuated
quartz tube at 800 °C for 12 h. A relative density of 80% was
obtained by this method.AC impedance measurements were performed
using an impedance analyzer (Solartron 1296 dielectric interface coupled
with a Solartron 1255B frequency response analyzer) in the frequency
range from 1 MHz to 100 mHz (with an amplitude of 50 mV). Silver paint
(RS silver conducting paint 186-3600), brushed on both sides of the
pellet and dried under vacuum at room temperature, was used as ion
blocking electrodes. Variable temperature conductivity measurements
were carried out under argon (flow rate 50 mL·min–1), using a custom-built sample holder, in the temperature range 25–125
°C. The impedance spectra were fitted with an equivalent circuit
using the ZView2 program.[59]
Results and Discussion
Computational/Experimental Study of the Li–Al–O–S
Phase Field
We explored the Li–Al–O–S
phase field using probe structure-based material discovery.[35,36] This method involves identifying a set of unexplored compositions
on the phase field of interest and computationally generating a probe
structure for each one. By determining the energy of each of the probe
structures compared to the convex hull, we can identify low-energy
regions of the phase diagram and hence target synthetic efforts toward
regions of the phase field where new compounds are more likely to
be found. To ensure fully occupied anion sublattices, the compositions
of unit cells used in calculations were constrained to contain eight
or nine anions with a stoichiometric number of cations to satisfy
charge neutrality. Under these constraints, a range of compositions
were chosen to span the phase field, with more compositions at the
Li rich end (Figure a). We sampled all of the possible compositions for the cells containing
eight anions and then only considered the Li/Al ratios closest to
1:1 for the cells containing nine anions since these were the lowest
energy regions following the initial screen. All possible S/O ratios
were considered. The computed energy of the probe structure at each
computed composition is presented in Figure a.
Figure 1
(a) Calculated energy of different compositions
in the Li–Al–O–S
phase field using cells containing hexagonal close-packed (hcp, black triangles) and cubic close-packed (ccp, black filled circles) anion lattices. Ehull is the energy above the convex hull. Reported oxide and sulfide
phases in the Li–Al–O–S phase field (black rectangles).
(b) First-stage experimentally tested compositions, which resulted
in a mixture of already reported compounds (empty red squares with
black letters), and a mixture of already reported compounds along
with the presence of the new phase (filled red squares with white
letters). Second-stage experimentally tested compositions (numbered
black circles). Composition of points are as follows: A (Li3Al9O2S13), B (LiAlOS), C (LiAlO0.2S1.8), D (LiAlO1.8S0.2),
E (Li7Al2O4S), F (Li5AlO3S), 1 (Li4Al2O2S3), 2 (Li6Al8O10S5), 3
(Li2Al4O4S3), 4 (Li2Al4O5S2), and 5 (Li3AlS3).
(a) Calculated energy of different compositions
in the Li–Al–O–S
phase field using cells containing hexagonal close-packed (hcp, black triangles) and cubic close-packed (ccp, black filled circles) anion lattices. Ehull is the energy above the convex hull. Reported oxide and sulfide
phases in the Li–Al–O–S phase field (black rectangles).
(b) First-stage experimentally tested compositions, which resulted
in a mixture of already reported compounds (empty red squares with
black letters), and a mixture of already reported compounds along
with the presence of the new phase (filled red squares with white
letters). Second-stage experimentally tested compositions (numbered
black circles). Composition of points are as follows: A (Li3Al9O2S13), B (LiAlOS), C (LiAlO0.2S1.8), D (LiAlO1.8S0.2),
E (Li7Al2O4S), F (Li5AlO3S), 1 (Li4Al2O2S3), 2 (Li6Al8O10S5), 3
(Li2Al4O4S3), 4 (Li2Al4O5S2), and 5 (Li3AlS3).A valley of lower energy is observed at a Li/(Li
+ Al) ratio of
0.5 (LiAlS2–LiAlO2 solid solution line),
with a local energy minimum at 42 meV·atom–1 above the convex hull for the previously proposed composition of
LiAlOS.[37] When using the full convex hull
construction, LiAlOS is predicted to decompose into LiAl5O8, LiAlS2, and Li2S. In comparison,
the probe structure in the previous study was also determined to be
thermodynamically unstable, but with respect to LiAlS2 and
LiAlO2, with a calculated formation reaction energy of
46 meV·atom–1 at 0 K.[37] Because these calculated energies are relatively low, LiAlOS could
be a metastable structure that is possibly synthesizable and
was therefore selected as a candidate composition for experiments.Near the sulfur- and oxygen-rich regions of the LiAlS2–LiAlO2 solid solution line, the calculated energy
is interestingly low compared to other regions close to the borders
in the overall phase diagram. For instance, the energy of composition
LiAlS1.8O0.2 is 34 meV·atom–1 above the convex hull. This value is close to the value of the known
phase LiAl5S8 (38 meV·atom–1 above the convex hull), which has been synthesized in the literature.[60] The composition LiAlS1.8O0.2 was therefore selected for experimental synthesis. The symmetrical
composition near the oxide end member, LiAlS0.2O1.8, shows a higher energy (51 meV·atom–1) but
is close to that of LiAlOS and also synthesized as a matter of comparison.Plateaus with energies below 60 meV·atom–1 above the convex hull are observed in the regions close to the terminal
Al2S3 and Li2O. Within the former,
the composition Li3Al9O2S13 was previously suggested by the Materials Project[61] as an interesting analogue candidate of Ga9Tl3O2S13.[62] The
energy calculated for this compound in this work is 46 meV·atom–1 above the convex hull, which is also close to that
calculated for LiAlOS. The plateau close to Li2O, as well
as presenting relatively low energies, attracted our interest due
to the high content of lithium, possibly favorable for attaining higher
conductivities. Two compositions, Li5AlO3S and
Li7Al2O4S, were selected for experimental
synthesis in this region of the phase diagram.Figure b and Table S1 summarize the six compositions (labeled
A to F and represented by red squares) that were chosen, in a first
stage, after analysis of the results from calculations. The experimental
procedures for all samples were the same and were described in the Experimental Section. In particular, we used Li2O and LiAlO2 as oxide precursors whenever the composition
enabled them, in order to improve the reactivity compared to the use
of Al2O3. Moreover, a relatively high temperature
of 800 °C and cooling by a quenching procedure were chosen in
order to both facilitate the access to high-energy phases and to kinetically
trap them and prevent the decomposition into binary or ternary phases
during cooling.Compositions highlighted with the empty red
squares and black letters
in Figure b consisted
of a mixture of only already reported phases after annealing, which
is detailed in Table S1. However, two points
along the LiAlO2–LiAlS2 solid solution
line (Figure b, points
B and C) showed the presence of an unknown phase along with the already
reported phases. Figure S1 shows the XRD
patterns of samples A to F after reaction made in the first stage.
For composition LiAlOS, the unknown phase seemed to be in a relatively
high amount; thus, in a second stage, four new compositions were tested
around this point (black circles in Figure b, composition given in the caption, and Table S1). For composition number 4 in particular
(Li2Al4O5S2), this new
phase was present along with Al2O3 as a single
impurity. Figure S2 shows the XRD patterns
of samples 1 to 4 after the reaction made in the second stage. Li2Al4O5S2 can then be written
as Li2Al4O5S2 = a Al2O3 + Li2Al2/3(1+OS2 (0 ≤ y < 5), highlighting the
fact that the unknown phase must have a Li/S ratio of 1. By considering
the two end members (y = 0 and y = 5), this result led us to investigate the solid solution line x Li2Al4O5S2 + (1 – x) Li2Al2/3S2, on which the composition Li2Al2/3(1+OS2 is located.
The synthesis of the pure sulfide end member (point number 5) gave
the phase pure new compound sought, which was thus revealed to be
a pure sulfide material, rewritten as Li3AlS3. The energy calculated for Li3AlS3 was found
to be 16 meV·atom–1 above the convex hull.
No other calculated energies in the phase diagram, which do not already
correspond to known phases, drop below this threshold. We therefore
concluded that no other oxysulfide phases are likely to form in the
Li–Al–O–S field. The oxide analogue Li3AlO3 could not be stabilized by a similar quenching method
(cf. Supporting Information) and leads
to a calculated energy of 27 meV·atom–1 above
the convex hull. The calculated energy of Li3AlS3 is slightly above the hull, which suggests that the phase is metastable.In the Li–Al–S phase field, phases with compositions
LiAlS2,[63] Li5AlS4,[64,65] and LiAl5S8[60] have previously been reported. Moreover, an
amorphous phase with composition Li3AlS3 was
identified previously as the discharge product of an Al–S battery.[66] Although no experimental structural data was
presented for this model, a local structure of this amorphous phase
was modeled with DFT and displayed a network of isolated 4- and 5-coordinated
aluminum. Both the LiAlS2 and Li5AlS4 structures possess nearly close-packed anion layers arranged in
a hexagonal stacking sequence. LiAl5S8 is dimorphic
and comprises a low-temperature modification with a normal spinel-type
structure and a high-temperature modification related to the ZnIn2S4 structure. Both LiAl5S8 structures have a cubic close-packed arrangement of the anion lattice.
The integrated computation–experiment approach described here
enabled the identification of a new crystalline phase with the composition
Li3AlS3. This phase was isolated by synthesis,
and its structure and lithium transport properties were experimentally
investigated.
Synthesis and Structure of Li3AlS3
Synthesis
Polycrystalline Li3AlS3 was synthesized via a solid-state reaction
of Li2S and Al2S3 (described in the Experimental Section). The powder XRD profile of
the product could be indexed to a phase whose structure does not match
any of the compounds previously reported for Li–Al–S
systems. A small quantity of Li5AlS4 (3.3(5)
mol %) was also identified in the XRD pattern. Elemental analysis
gave an overall composition of Li3.1(1)Al1.1(1)S3.0(1) (Table S3), but the
ICP measurement is not sufficiently precise to distinguish between
the nominal reaction stoichiometry (Li3AlS3),
the measured composition, and the presence of the secondary Li5AlS4 phase, further complicating its interpretation;
consequently, the compound is referred to as Li3AlS3 hereafter.
Structure Determination
The crystal
structure of Li3AlS3 was solved by first indexing
the SXRD pattern using the first 22 reflections using GSAS-II.[67] The unit cell was indexed in the space group C12/c1 with approximate lattice parameters
of 14.3 × 12.0 × 6.6 Å with β ≈ 117°.
The lattice parameters were then refined across the d spacing range 16–0.67 Å (2θ =
[3–75°]) via a Le Bail fit in Jana2006.[68] The structure was solved initially by locating the S and
Al atoms using superflip[69] implemented
in Jana2006. From this solution, a Rietveld model was refined against
the SXRD and NPD patterns. Once converged, the Li atoms were located
using Fourier Difference mapping on the NPD patterns, searching for
peaks in the difference map located at greater than 1 Å from
existing atoms within the model. When no more new sites could be located,
an initial Rietveld model was constructed and refined.Final
Rietveld refinement of the neutron and synchrotron diffraction data
was carried out using the program FullProf.[70] First, the SXRD pattern was fitted on its own. Position, site occupancy
factor (sof), and atomic displacement parameters
(adp) of lithium atoms remained fixed in the refinement
of the synchrotron data. The sof of aluminum refined
to 0.974(1) on its site and significant drops in the conventional
reliability factors were obtained: RBragg decreased from 2.79 to 2.68. Figure a shows the final fit of the SXRD pattern, and the
following results were obtained for the cell parameters: a = 14.31901(5) Å, b = 11.98037(3) Å, c = 6.62700(2) Å, and β = 116.9231(3)°.
The values of the refined cell parameters were then implemented and
fixed in the refinement of neutron data. The final model of the structure
was obtained through the combined refinement of the neutron data coming
from the three neutron datasets. Figure b–d shows the final fits of the patterns.
Refinement details and outcomes as well as the crystallographic data
are summarized in Tables S4 and S5.
Figure 2
Final Rietveld
refinement of (a) the synchrotron X-ray diffraction
pattern of 7Li3AlS3 (Diamond Light
Source, I11 beam line) with fixed Li positions and (b) 7Li3AlS3 against neutron powder diffraction
data (ISIS neutron source, HRPD) from (b) bank 1 (2θ = 168.330°),
(c) bank 2 (2θ = 89.580°), and (d) bank 3 (2θ = 30.000°),
with Iobs (red dots), Icalc (black line), Iobs – Icalc (blue line), and Bragg reflections (red
tick marks for Li3AlS3, black tick marks for
Li5AlS4, and blue tick marks for the vanadium
can).
Final Rietveld
refinement of (a) the synchrotron X-ray diffraction
pattern of 7Li3AlS3 (Diamond Light
Source, I11 beam line) with fixed Li positions and (b) 7Li3AlS3 against neutron powder diffraction
data (ISIS neutron source, HRPD) from (b) bank 1 (2θ = 168.330°),
(c) bank 2 (2θ = 89.580°), and (d) bank 3 (2θ = 30.000°),
with Iobs (red dots), Icalc (black line), Iobs – Icalc (blue line), and Bragg reflections (red
tick marks for Li3AlS3, black tick marks for
Li5AlS4, and blue tick marks for the vanadium
can).The unit cell shows three sulfur sites (S1, S2,
and S3), three
tetrahedral sites on general positions 8f (Table S5), two of them occupied by Li (Li1 and
Li4) and one by Al and Li ions (Al and LiAl), and two occupied
octahedral lithium sites (Li2 and Li3) located on two 4e Wyckoff positions on the 2-fold rotation axis (at the beginning
of the refinement). The isotropic adp (Biso) were refined to large values: 3.3(5) Å2 for Li2 and 3.7(5) Å2 for Li3. In order to improve
the model, the displacement was modeled as anisotropic, which led
to the decrease of χ from 1.74,
6.46, and 1.54 to 1.73, 6.06, and 1.54 for banks 1, 2, and 3, respectively.
The anisotropic ADPs for Li2 and Li3 remained high, and a marked anisotropic
displacement along the a axis was found for Li3 in
particular, as highlighted by the Fourier difference map in the ab plane in Figure S4a, thereby
prompting us to consider site splitting. Li3 was moved from its fully
occupied 4e position (0, y, 0.25)
to a half occupied general 8f position (x, y, z), which then generates a
second Li3 atom of coordinate (−x, y, z) from the other side of the rotation
axis, within the same coordination polyhedra. For Li2, as the anisotropic
displacement was not as straightforwardly along one single direction,
moving the atom to one half occupied 8f position
did not improve the fit. This site was split into two 8f positions (Li2 and Li2b), each allocated first with an occupancy
of 0.25, generating 4 Li positions within the same coordination polyhedron,
in order to model its large displacement. The occupancies of Li2 and
Li2b were then refined by constraining their sum to be equal to 0.5.
In that way, the total occupancy within each polyhedron is equal to
one. The site splitting of Li2 and Li3 led to much smaller isotropic adp of 1.2(7) Å2 for Li2 (and Li2b) and
1.7(4) Å2 for Li3, along with a reduced residual density
in the Fourier difference map around the sites (Figure S4b). Another indication for preferring the modeling
of both sites with multiple atoms was the increase in the bond valence
sum (BVS) from 0.76(1) to 0.83(16) for Li2 and 0.82(4) for Li2b, while
keeping it to 0.80(3) for Li3.Through occupancy refinement,
a lithium antisite defect was identified
and occupied 6.7% of the aluminum site. No other Li or Al antisite
defects were found. Details of the refinement procedure are presented
in the Supporting Information, page 7,
and results of the final refinement of the occupancies are given in Table S4.NMR spectroscopy at various fields
was deployed to further confirm
the overall pattern of site occupancy of the lithium atoms. The 6Li MAS NMR spectra at 9.4 and 20 T for Li3AlS3 are shown in Figure a and display three well-resolved resonances at 1.7, 1.3,
and −0.2 ppm, which fit yield signals of equal integration.
A small shoulder is also observed at 1 ppm (Figure S8) and corresponds to the Li5AlS4 impurity
seen in the diffraction and 27Al NMR data (see below);[65] this signal was found to integrate 3.0(5) mol
% Li3AlS3, in agreement with the 3.3(5) mol
% value from diffraction. Based on the well-established semi-empirical
correlations relating the lithium coordination environment and 6Li NMR shifts,[71] further aided
by calculations of the NMR parameters using the GIPAW approach[52,53] as implemented in CASTEP[51] (cf. Supporting
Information, Table S7), the signal at −0.2
ppm has been attributed to the octahedrally coordinated Li2/Li2b and
Li3 sites while the resonances at 1.3 and 1.7 ppm correspond to Li4
and Li1, respectively. These assignments agree well with the structural
refinement described above, which identified the sum of the contents
of the three octahedrally coordinated sites Li2 (0.8(3)), Li2b (3.2(3)),
and Li3 (4.0(4)) to 8.0(7) Li per unit cell and the contents of the
two tetrahedrally coordinated Li4 and Li1 to 7.8(3) and 8.00 per unit
cell, respectively.
Figure 3
(a) 6Li MAS spectrum of Li3AlS3 obtained at magnetic fields of 9.4 T (black) and 20 T (blue).
The
experimental spectrum (full lines), total fit (dashed lines) spectral
deconvolution (dotted lines), Li5AlS4 impurity
(red dotted lines), and GIPAW-simulated spectrum (green lines) are
shown. (b) 27Al MQMAS NMR spectrum of Li3AlS3 recorded at a magnetic field of 9.4 T and 20 T. The dotted
lines (black for a field of 9.4 T and blue for 20 T) and the red dotted
lines represent the spectral deconvolution of Li3AlS3 and Li5AlS4, respectively. The dashed
lines show the total fit for the sample, and the solid lines show
the anisotropic one-dimensional 27Al spectrum, while the
vertical spectrum shows the non-quantitative isotropic 27Al spectrum. The solid green line shows the GIPAW-simulated spectrum
with an isotropic chemical shift of 117 ppm, a quadrupolar coupling
constant of 5.1 MHz and an asymmetry parameter of 0.44 (Table S7).
(a) 6Li MAS spectrum of Li3AlS3 obtained at magnetic fields of 9.4 T (black) and 20 T (blue).
The
experimental spectrum (full lines), total fit (dashed lines) spectral
deconvolution (dotted lines), Li5AlS4 impurity
(red dotted lines), and GIPAW-simulated spectrum (green lines) are
shown. (b) 27Al MQMAS NMR spectrum of Li3AlS3 recorded at a magnetic field of 9.4 T and 20 T. The dotted
lines (black for a field of 9.4 T and blue for 20 T) and the red dotted
lines represent the spectral deconvolution of Li3AlS3 and Li5AlS4, respectively. The dashed
lines show the total fit for the sample, and the solid lines show
the anisotropic one-dimensional 27Al spectrum, while the
vertical spectrum shows the non-quantitative isotropic 27Al spectrum. The solid green line shows the GIPAW-simulated spectrum
with an isotropic chemical shift of 117 ppm, a quadrupolar coupling
constant of 5.1 MHz and an asymmetry parameter of 0.44 (Table S7).The 27Al one-dimensional MAS spectra
for Li3AlS3 at 9.4 and 20 T are shown in Figure b and reveal a second
order quadrupolar line
shape that resonates at ∼100 ppm (at 9.4 T) and ∼120
ppm (at 20 T), typical of tetrahedrally coordinated Al sites, and
a very sharp signal, which is field-independent, at ∼130 ppm,
corresponding to the small amount of Li5AlS4 impurity (Figure S7). Note that no signal
in the octahedral region (around 0 ppm) of the 27Al MAS
NMR spectrum is present as expected. The z-filtered
triple quantum MAS[53,54] NMR spectrum of Li3AlS3 is also shown in Figure b and demonstrates that the ∼100 ppm
signal corresponds to one resonance only with an isotropic chemical
shift of 117 ppm (cf. the Supporting Information) and a quadrupolar coupling constant (CQ) of 5.8 MHz in close agreement with the computed value of 5.1 MHz
(cf. SI, Table S7).The final model led to the overall refined composition Li3.13(2)Al0.958(4)S3, slightly different from the composition
determined by ICP. The small differences can be explained by the presence
of the Li5AlS4 impurity, which prevents the
ICP measurement from producing an accuracy to the nearest two decimal
places. The refined composition is different from the ideal, which
will be discussed hereafter.
Structure Description
Polyhedral Arrangement
Li3AlS3 adopts a structure related to that of Na3InS3 reported by Eisenmann and Hofmann where In3+ and Na+ cations are replaced by Al3+ and Li+, respectively.[72] The
only other isostructural phases reported are Na2Mn2S3,[73] Na2Mn2Se3,[74] and LiNa1–Mn2S3 (x ≈ 0.7)[75] in which half of the Mn2+ ions are
replaced by aluminum atoms whereas the other half along with the sodium
atoms are replaced by Li+ ions. The structure is constructed
from an hcp arrangement of sulfur atoms with an AB
A*B* stacking of anion layers where B is the equivalent of A through
the c glide plane and 2-fold axis symmetry operations.
A* and B* are the equivalent of A and B through the C centering translation
(Figure a). In the
tetrahedral layer, Li1 and Al atoms occupy 2/3 of the tetrahedral
interstices between a pair of sulfur atom layers (B and A*, equivalent
to the B*A pair). Between the second pair of sulfur layers (A and
B, equivalent to the A*B* pair), Li2 and Li3 occupy octahedral interstices,
whereas Li4 occupies a tetrahedral interstice, generating a mixed
polyhedral (octahedral–tetrahedral) layer. The two different
polyhedral layers are stacked alternately perpendicular to the bc plane (Figure a). Bond distances and angles of the different polyhedral
and BVS calculations performed for all atoms are summarized in Table S6.
Figure 4
(a) Crystal structure of Li3AlS3 showing
the alternating arrangement perpendicular to the bc plane of the tetrahedral layers containing AlS4 and LiS4 tetrahedra and the mixed polyhedral layers containing Li-only
polyhedra. (b) T+ and T– interstices
in the tetrahedral layer, showing the corner-sharing arrangement of
the Li1, Al, and vacant (empty) tetrahedra in each network, as well
as the interconnection (following the yellow arrow) of each T+ (thin lines) and T– (thick lines) network
so that AlS4 dimers are formed. The highlighted yellow
face of the Li1 tetrahedron corresponds to the only face that shares
two edges with two vacant sites. (c) View of both the mixed polyhedral
layer and the tetrahedral layer in the bc plane and
of their interconnection (following the yellow arrow). Polyhedra colors:
blue: Al tetrahedra; orange: Li tetrahedra; red: Li2 octahedra; light
red: Li3 octahedra.
(a) Crystal structure of Li3AlS3 showing
the alternating arrangement perpendicular to the bc plane of the tetrahedral layers containing AlS4 and LiS4 tetrahedra and the mixed polyhedral layers containing Li-only
polyhedra. (b) T+ and T– interstices
in the tetrahedral layer, showing the corner-sharing arrangement of
the Li1, Al, and vacant (empty) tetrahedra in each network, as well
as the interconnection (following the yellow arrow) of each T+ (thin lines) and T– (thick lines) network
so that AlS4 dimers are formed. The highlighted yellow
face of the Li1 tetrahedron corresponds to the only face that shares
two edges with two vacant sites. (c) View of both the mixed polyhedral
layer and the tetrahedral layer in the bc plane and
of their interconnection (following the yellow arrow). Polyhedra colors:
blue: Al tetrahedra; orange: Li tetrahedra; red: Li2 octahedra; light
red: Li3 octahedra.In the tetrahedral layer, one tetrahedral site
in every three is
vacant in an ordered manner (Figure b). Each T+ and T– interstice
forms a network of alternating Al, Li1, and vacancy corner-shared
tetrahedra. The T+ and T– networks interlock
in such a way that each AlS4 unit shares one edge (S3–S3)
with another Al tetrahedron and a second edge with the Li1S4 unit within the layer. These 4-edge-shared tetrahedra (Li1S4(T+)–AlS4(T–)–AlS4(T+)–Li1S4(T+)) form a unit that is connected to other units of this type
by corner-sharing (Figure a, circled part in Figure b). The Li1 tetrahedron has two S atoms that share
corners with Li1 and Al, and there is no corner-sharing of Al tetrahedra.
Figure 5
Coordination
polyhedra of (a) Li1 and Al in the tetrahedral layer,
(b) Li4, (c) Li2 and Li2b, and (d) Li3 in the mixed polyhedral layer.
Coordination
polyhedra of (a) Li1 and Al in the tetrahedral layer,
(b) Li4, (c) Li2 and Li2b, and (d) Li3 in the mixed polyhedral layer.In the mixed polyhedral layer, each of the Li2,
Li2b, and Li3 atoms
is surrounded by six sulfur atoms to form LiS6 octahedra
(Figures c and 5b). Li2 and Li2b polyhedra form an infinite chain
of edge-shared octahedra along the c axis, which
are also connected to three Li3 octahedra also through three shared
edges. Thus, both octahedra form infinite two octahedra-wide chains
along the c axis. Each chain is separated along the b axis by a chain of edge-shared LiS4 T+ and T– tetrahedra occupied by the Li4 atom (Figure c). Li3 octahedra
are linked to two consecutive T+ and T– Li4 tetrahedra, sharing one face with each of them (Figure c). These two Li4 tetrahedra
form a unit (Figure b) and each T+ and T– of the unit shares
one corner with one Li2/Li2b octahedron from the same chain as the
face-shared Li3 octahedra, as well as one corner with one Li3 octahedron
of the chain on the other side of the Li4 chain. Along the Li4 chain,
octahedral interstices are vacant so that only 2/3 of the octahedral
sites are occupied in the layer. As a result of crystallization, the
ordered structure obtained experimentally in this study strongly differs
with the structure modeled for the amorphous discharged product with
the same composition described by Yu et al.[66]Figure c shows
the connection between the polyhedra of both layer types. Each of
the Al and Li1 T+ (T–) tetrahedra is
connected to the layer below (above) by sharing the face at the base
of the tetrahedron with a vacant tetrahedral site of the mixed polyhedral
layer. For the Al tetrahedra, this face shares one edge with the Li2
octahedra and another edge with the Li3 octahedra of the same chain,
while the third edge is shared with the vacant octahedral site. For
the Li1 tetrahedra, this face shares two edges with the Li2 octahedra
and one edge with the Li3 octahedra of the same chain. The S3 atom,
which is at the vertex of the Al T+ (T–) tetrahedra, is shared with one Li2, one Li3, and four Li4 of the
above (below) mixed polyhedral layer, whereas the S1 atom, which is
at the vertex of the Li1 T+ (T–) tetrahedra,
is shared with one Li3 octahedron and four Li4 of the above (below)
mixed polyhedral layer. The chain of vacant T+ (T–) sites along the c axis in the tetrahedral layer
lies above (below) the chain of Li4 tetrahedra of the mixed polyhedral
layer.
Polyhedral Distortions and Atom Displacements
Figure shows the
bonding environment of each of the cations. In the tetrahedral-only
layer, Li1 atoms are strongly off-centered toward the S1–S2–S2
face so that it both does not share any edges with the AlS4 tetrahedra in the same layer and does not belong to the octahedral
layer (Figure ). This
face is also the only one that shares two edges with two vacant sites
of the tetrahedral layer as opposed to one for the other two faces
that do not belong to the octahedral layer (highlighted yellow face
in Figures b and 6). Moving away from the nearby Al cations as well
as from the Li2 and Li3 atoms of the octahedral layer provides the
electrostatic driving force for its displacement (Figure ). The BVS of Li1 is 0.97(2),
which is very close to the ideal value of +1, considering the oxidation
state of lithium.
Figure 6
Crystal structure of Li3AlS3 showing
the
arrangement of octahedral (red) and tetrahedral (orange) lithium and
tetrahedral aluminum (blue). The direction of the displacement of
atoms is symbolized by arrows: blue for Al, orange for Li1 and Li4,
and yellow for S.
Crystal structure of Li3AlS3 showing
the
arrangement of octahedral (red) and tetrahedral (orange) lithium and
tetrahedral aluminum (blue). The direction of the displacement of
atoms is symbolized by arrows: blue for Al, orange for Li1 and Li4,
and yellow for S.The tetrahedral Li4 site in the mixed layer shares
a face with
the vacant tetrahedral site of the tetrahedral layer. The Li4 position
is strongly displaced toward the base of the tetrahedron and the adjacent
sulfur layer, toward this vacant tetrahedral interstice (Figure ). Li4 thus adopts
a pseudo-trigonal bipyramid environment with one short axial Li4–S1
and one long axial Li4–S2 bond. BVS for Li4 is slightly lower
than that of Li1 (0.93(2)), which is consistent with its pseudo-5-coordinated
environment, making it more loosely bound to the S atoms compared
to the 4-coordinated Li1.This is reflected in the NMR resonance
frequency of Li4 (1.3 ppm)
for which the 5-coordinate environment provides this site with an
intermediate chemical shift between the lithium atoms occupying a
tetrahedral site (Li1 at 1.7 ppm) and octahedral sites (Li2/Li2b and
Li3 at −0.2 ppm). The NMR shift calculation also suggests that
the resonances of Li3 and Li2/Li2b should be resolved as the isotropic
chemical shifts differ by 0.3 ppm; however, this is not observed experimentally,
even at high field, perhaps due to the larger full width at half-maximum
observed for this Li3/Li2/Li2b resonance (18 Hz compared to 10 and
11 Hz for Li4 and Li1, respectively). The presence of the occupied
Li4 sites distinguishes both Li3 and Li2/Li2b sites: Li3 shares faces
with Li4, whereas Li2 does not (Figure c). The shift of the Li4 toward the adjacent sulfur
layer pushes the Li4 atom further away from the octahedral Li3 atoms
and therefore reduces the structural difference between Li2 and Li3,
further explaining the similar shifts observed experimentally for
these octahedral sites (Table S7). On the
contrary, the difference in the environment of Li1 and Li4 is more
pronounced (tetrahedral vs trigonal bipyramid), which explains the
resolution of their respective NMR peaks. The consistency between
the experimental NMR data and the computed ones from the described
structure further reinforces the accuracy of the selected structural
model.The geometry of the AlS4 dimer is represented
in Figure a, and among
the
six edges of each AlS4 tetrahedron, four of them are directly
connected to the S3–S3 edge-shared with the other tetrahedron
of the dimer. Consequently, the aluminum position is shifted toward
the S1–S2 edge that does not share a common sulfur atom with
the other edge-shared Al tetrahedron of the dimer. This displacement
is symbolized by blue arrows in Figure . This can be explained by the proximity to the other
Al atom of the edge-shared dimers with which it tends to maximize
its distance. The Al–Al distance is 3.015(6) Å, which
is similar to distances obtained in other sulfide materials (Na6Al2S6, Na3FeS3) presenting these dimers.[76,77] This is also supported
by the 27Al NMR data obtained experimentally and computed
with GIPAW (cf. SITable S7) that yield a clear second-order quadrupolar line shape at both
9.4 and 20 T and from which large quadrupolar coupling constants (CQ = 5.8 and 5.1 MHz for experimental and computed
values, respectively) and distorted asymmetry parameter values (ηQ = 0.56 vs 0.44 for experimental and computed values, respectively)
are obtained. These data clearly demonstrate the non-symmetrical dimeric
Al coordination environment in Li3AlS3 and is
in sharp contrast to the symmetric AlS4 tetrahedra observed
in Li5AlS4 (CQ ≈
0 MHz, Figure S7) as evidenced by the field-independent
NMR narrow line of this phase.The dimerization leads to the
strong repulsive force between both
highly charged Al3+ cations. This in turn brings the S3
atoms inside the dimer toward each other and therefore toward the
interior of the tetrahedral layer, in order to keep the BVS of Al
and S close to their ideal values (symbolized by yellow arrows in Figure b,c). The S3–S3
distance (represented by a thick line in Figure b,c, dS3–S3 = 3.463(14) Å) is indeed the shortest in the structure. The
compression of the S3–S3 edge of the Al2S6 dimer unit in the tetrahedral layer leads to the stretching of the
S3–S3 edge of Li2(Li2b)S6 and Li3S6 octahedra
in the mixed polyhedral layer (Figure b,c).As shown in Figure a, the Li2b (and Li2) octahedron is highly
distorted, and the S3–S3
edge is considerably longer than the other edges (dS3–S3 = 4.40(1) Å whereas the length of the
other edges ranges from 3.66(1) to 3.951(7) Å). Also, the Li3
octahedron is elongated along an axis defined by two S3 atoms (dS3–S3 = 5.729(14) Å), whereas the
equatorial plane defined by two S1 and two S2 atoms is close to a
square (dS2–S2 = 3.893(13) Å, dS1–S1 = 3.713(14) Å, dS2–S1 = 3.682(8) Å). The explanation for the
site splitting of Li2 can be linked to the distribution of the sulfur
vacancies (Supporting Information, page
10). In contrast to the tetrahedral Li1 and Li4, the BVS values for
octahedral Li2, Li2b, and Li3 are below the theoretical value (0.83(16),
0.82(4), and 0.80(3) for Li2, Li2b and Li3, respectively), which reflects
the fact that they are weakly bonded to the sulfur atoms. This observation
has also been made in similar structures.[65]A sulfur deficiency was found on two S sites. This deficiency
accounts
for the charge compensation with the Al defect. The majority of sulfur
vacancies are located on the S3 site, which bridges the two aluminum
tetrahedra of the dimer. As explained above, the S3–S3 distance
is the shortest in the structure; hence, vacancies here would reduce
anion–anion repulsions. Again, the presence of the dimers leading
to the short S3–S3 distance could be the trigger for the deficiency
of the S site. Further, the occupancy of the Al site cation vacancies
by Li creates a negatively charged antisite defect, which could also
drive the localization of the positively charged sulfur vacancy on
the S3 site.The analysis of the Li3AlS3 structure highlights
the importance that the Al2S6 dimers have on
site geometries as well as on site occupancies of both lithium and
sulfur atoms. Moreover, the comparison with the probe structure, which
also shows the presence of these dimers (Supporting Information, page 11, and Figure S5), further suggests that the stability of the structure is connected
to the presence of the Al2S6 dimers.
Comparison with Known Structures
Lithium sulfide materials showing different arrangements of cation
polyhedra in hcp arrays are common, and some interesting
compositions and structures have been described by Lim et al.[65] Among them, one can cite Li2FeS2, which consists of an octahedral-only layer whose interstices
are 100% occupied by lithium atoms, alternating with a tetrahedral
layer, where all the tetrahedral sites are occupied by Li and Fe atoms
randomly distributed in a 50:50 ratio.[78] Li5AlS4 shows a similar structure where only
half of the Fe atoms are replaced by Al3+ ions whereas
the other half is occupied by Li+ ions.[65] The tetrahedral layer consists of ordered LiS4 and AlS4 units in a 3:1 arrangement. The authors note
that these structures can be expressed as [LiFe]T[Li]OS2 and [Li1.5Al0.5]T[Li]OS2, respectively. We added the superscripts
“T” and “O”, which refer to the tetrahedral
and octahedral coordination of the cation in alternating layers. Following
this representation, the material reported in this study, Li3AlS3, could be written as [Li2/3OLi2/3T][Li2/3Al2/3]TS2 where the cation in the same square bracket
belong to the same layer. Recently, Leube et al. reported a family
of compounds Li4.4M0.4M′0.6S4 (M = Al3+, Ga3+, M′ =
Ge4+, Sn4+) whose structure is closely related
to that of Li5AlS4.[46] The highly charged cations share the same site in the tetrahedral
layer, the octahedral layer is made of ordered partially occupied
lithium sites and fully vacant sites in a 3:1 arrangement, and the
remaining lithium atoms share two crystallographically distinct tetrahedral
sites in both the tetrahedral and octahedral layer in a 74:26 ratio.
Following the same convention, this structure can be written as [Li0.66OLi0.38T][Li1.11M0.2M′0.3]TS2.Other alkali metal sulfides with composition A3MS3, where A is an alkali monovalent cation and M is a trivalent
cation, have been reported in the literature. As stated above, Na3InS3 in particular shows a closely related structure
to Li3AlS3 and can be written as [Na2/3□1/3]O[Na]2/3T[Na2/3In2/3□2/3]T′S2 where In and Na occupy the same sites as Al and Li
in Li3AlS3. Moreover, both compounds show identical
distortions of the alkali octahedra and the same direction of displacements
for the tetrahedral alkali cations. This particularity is most probably
coming from the high degree of constraint imposed by the presence
of the M3+S4 dimers, and the
fact that this observation is made in both phases reinforces the validity
of this explanation. Rothenberger et al.[79] reported a structure with the composition (M(AlS2)(GeS2)4 (M = Na, Ag,
Cu) where 20% of the Al atoms are in similar dimeric tetrahedral units).
However, these Al dimers were much less distorted with very similar
Al–S interatomic distances within the tetrahedra (the average
Al–S distance is 2.2117(5) Å and the standard deviation
1.3%). The absence of tetrahedral distortion in these dimers can be
explained by the 3D polyanionic rigid structure imposed by the three
other Al or Ge sites. This lessening of the distortion from tetrahedral
symmetry is also reflected in the 27Al NMR where the
resonance observed shows,on the spectra, smaller values of CQ for (M(AlS2)(GeS2)4 (M = Na, Ag, Cu).The splitting of the Li2 site in Li3AlS3 is
not reported for Na3InS3, nor is the displacement
from the Wyckoff position to a general position of the second octahedral
alkali site. Lithium is smaller than sodium, and the volume occupied
by each atom considering a hard sphere model is 12 and 18% of the
total volume of each octahedron for Li and Na, respectively. The displacements
and splitting modeled for lithium might therefore be key to generating
an appropriate bonding environment as defined by the BVS.Interestingly,
Na3AlS3 along with Na3GaS3 shows a slightly different structure (Figure a,b).[76,80] In between two sulfide
layers, Al2 occupies 1/3 of the tetrahedral
sites and forms dimers between T+ and T– tetrahedra of the same slab. In this layer, Na4 occupies 1/3 of
the distorted octahedral sites forming infinite chains along the c axis. In between the next two sulfide layers, Na1 occupies
one distorted octahedral site, whereas Na2 and Na3 form 5-coordinated
sulfide polyhedra in a trigonal bipyramid configuration and all three
Na sites are in a 1:1:1 arrangement. The next slab consists of a similar
AlS4 and NaS6 polyhedra arrangement to the first
described slab, with a slight tilt of the polyhedra (Figure b). Na5 lies within each of
the sulfide layers forming the Al2S4 slab and is 6-coordinated
to four of the sulfur atoms of the same layer, one in the layer above
and one in the layer below, therefore forming a 2D network of edge-
and corner-shared octahedra. The structure difference between Na3InS3 or Li3AlS3 (for which
Na or Li is 4-coordinated in the tetrahedral layer) and Na3AlS3 or Na3GaS3 (for which Na is
6-coordinated in the tetrahedral layer) could be attributed to the
size of the M3+ cation with respect to
that of the alkali cation. Indeed, the ionic radius of Al3+ (0.39 Å) and Ga3+ (0.47 Å) is considerably
smaller than that of In3+ (0.62 Å),[81] so the size of tetrahedral interstices will decrease and
might not be suitable to host Na+ cations, which would
then prefer to occupy octahedral sites, in contrast with the smaller
Li+ cation.
Figure 7
(a) Crystal structure of Na3AlS3 showing
the alternating arrangement, along a, of the tetrahedral
layers containing AlS4 and NaS4 tetrahedra and
of the mixed polyhedral layers containing Na-only polyhedra. (b) View
of the two consecutive tetrahedral layers of Na3AlS3 in the bc plane. (c) Crystal structure of
Na3FeS3 showing one type of layer along b (d) View of the layer along b of Na3FeS3 showing the fully occupied octahedral sites
by Na atoms and the 1/3 occupied tetrahedral interstices by Fe atoms
in a dimer arrangement.
(a) Crystal structure of Na3AlS3 showing
the alternating arrangement, along a, of the tetrahedral
layers containing AlS4 and NaS4 tetrahedra and
of the mixed polyhedral layers containing Na-only polyhedra. (b) View
of the two consecutive tetrahedral layers of Na3AlS3 in the bc plane. (c) Crystal structure of
Na3FeS3 showing one type of layer along b (d) View of the layer along b of Na3FeS3 showing the fully occupied octahedral sites
by Na atoms and the 1/3 occupied tetrahedral interstices by Fe atoms
in a dimer arrangement.Another known structure of similar composition
showing different
arrangements of metal and alkali cations within the hcp array is that of Na3FeS3,[77] which is also adopted by Na3FeSe3,[82] Na3AlSe3,[83] and Na3GaSe3[84] (Figure c). In those phases, the M3+ cation assembles in tetrahedral
dimers in between each of the sulfide layers, and the sodium atom
occupies all of the octahedral sites in these layers (Figure c,d). Curiously, the absence
of alkali-only layers and the stabilization of a single layer type
containing both Fe or M (M = Al3+, Ga3+, and In3+) and Na or Li polyhedra
seem to occur in all of the selenide compounds as well as both the
iron sulfide and selenide phases. It seems that this could be linked
to the less ionic character of the bonds in those structures. Indeed,
the higher polarizability of Se2– compared to S2– anions generally leads to softer and more covalent
bonds in selenides.[85] It has also been
shown, through magnetic property measurements, that the 3d orbitals of Fe3+ in Na3FeS3 are
more extended than that of ionic Fe3+, which can be attributed
to Fe–S covalency.[86] Because Al3+, Ga3+, and In3+ do not show this effect,
the ionic character of the M–S (M = Al3+, Ga3+ and In3+) bond might
be more pronounced in Na3MS3 than in Na3FeS3. Different layer-type structures
are often found in materials where ions have different chemical properties,
in particular different polarizability, which results in different
ionicity of the cation–anion bond in each layer.[87,88] A more covalent character of the bonds would then attenuate the
difference in polarizability of the cations and favor the single layer-type
structure.
Lithium Conductivity
The ionic conductivity
of Li3AlS3 was assessed by electrochemical impedance
spectroscopy on sintered pellets with an 80 ± 2% relative density.
The Nyquist plot of the sample measured at room temperature under
an argon atmosphere is shown in Figure a. The presence of the two semicircles is characteristic
of two unique time constants and therefore of the dissociation between
different scattering contributions. The plot has therefore been fitted
by a two-component equivalent electrical circuit (inset in Figure a) that models these
two contributions. Each component consists of a resistance associated
in parallel with a constant phase element (CPE, a modified capacitor
taking into account inhomogeneities in the sample, cf. Supporting Information). The values of the capacitance
obtained for the semicircles at high and low frequencies were 8(1)
× 10–12 and 2.6(4) × 10–9 F and are characteristic of the bulk and grain boundary response,
respectively.[89] The high-frequency intercepts
of both semicircles give direct values of the bulk and total resistance
(Rbulk and Rt = Rbulk + RGB, respectively, where RGB is the resistance
resulting from the grain boundary scattering). Bulk and grain boundary
room-temperature conductivities of 1.3(1) × 10–8 and 2.2(2) × 10–9 S·cm–1, respectively, were obtained. The impedance of the pellet was measured
over the temperature range (24–125 °C), and each Nyquist
plot was fitted with the described equivalent circuit. Tables S8 and S9 present the results of the fits
and the values of the different parameters obtained at each temperature.
For each temperature, the conductivity of the bulk, σbulk, was therefore extracted and showed to follow the
Arrhenius law (Figure b) with an activation energy of 0.48(1) eV.
Figure 8
(a) Nyquist plot at 30
°C of Li3AlS3 and (inset) electrical equivalent
circuit showing the two contributions
to the conductivity. (b) Arrhenius plot of the bulk conductivity of
Li3AlS3 measured by AC impedance. Black squares
correspond to the experimental data, and the red line corresponds
to the fits.
(a) Nyquist plot at 30
°C of Li3AlS3 and (inset) electrical equivalent
circuit showing the two contributions
to the conductivity. (b) Arrhenius plot of the bulk conductivity of
Li3AlS3 measured by AC impedance. Black squares
correspond to the experimental data, and the red line corresponds
to the fits.The room-temperature bulk conductivity is of the
same order of
magnitude as that of Li5AlS4 (σ = 9.7 × 10–9 S·cm–1).[65] In Li5AlS4,
all Li sites are fully occupied, whereas in Li3AlS3, there are multiple ordered vacancy sites: one third of the
octahedral interstices in the mixed polyhedral layer and two thirds
of the tetrahedral interstices in the tetrahedral layer are vacant.
Although the lowering of the activation energy (0.48 eV for Li3AlS3 and 0.61 eV for Li5AlS4) is indeed observed, the increase in Li mobility is not, which suggests
that ordered vacancy sites are not sufficient to improve conductivity
in this structure type.In the recently reported related Li4.4M0.4M′0.6S4 compounds,
which shows considerably
higher conductivities σ = 10–5–10–6 S·cm–1, the
presence of multiple disordered partially occupied lithium sites has
been shown to play a major role in the improvement of the conductivity.
In Li3AlS3, the only disordered vacancies can
be found within the Li4 site, which was determined to be 98% occupied
(Table S5), as all other Li sites are completely
occupied. In order to highlight the different types of vacancies within
the structure, Li3AlS3 can be written aswith □ and Δ being the disordered
and ordered vacancies, respectively. The family of quaternary materials,
using the same convention, can be written asOverall, this structure shows fewer
ordered vacant sites, but the
content of disordered vacancies is considerably higher, which underlines
the importance of this feature in order to improve conductivity.It is interesting to note that, in the Na3FeS3 structure, only one type of layer in which all the sodium is located
in octahedral sites is present. Because the mobile species are believed
to be the octahedral lithium in these types of structures,[46] it would be of high interest to stabilize the
Na3FeS3 structure in lithium-containing compounds
while creating a large number of disordered vacancies.
Influence of the Al2S6 Dimers on the Structure and Li Ion Conductivity
The strong
differences between Li5AlS4 or Li4.4M0.4M′0.6S4 (M = Al3+, Ga3+, M′ = Ge4+, Sn4+) structures
and Li3AlS3 come from the ordering of the tetrahedral
cations and the presence of edge-shared AlS4 tetrahedra
pairs, which form Al2S6 dimers. The three structures
have the same anion sublattice with tetrahedral Li and Al layers alternated
with Li only layers (= mixed polyhedral layer). In the three structures,
Al atoms must spread over the tetrahedral interstices of the tetrahedral
layer, and when this is done in an ordered manner, the M/S ratio imposes the arrangement pattern. Indeed, in Li3AlS3, M/S = 1/3 means that there is one
Al atom for three tetrahedral interstices (one tetrahedra contains
four sulfur atoms, and each tetrahedra is connected to four other
tetrahedra in a layer consisting of T+ and T– interstices, cf. Figure ). Therefore, M/S = 1/3 imposes a 1:2 ordering
in the Al tetrahedra by optimizing the distance between each AlS4 unit (Figure , entry 1). In the same way, M/S = 1/4 in Li5AlS4 and Li4.4M0.4M′0.6S4 leads
to a 1:3 arrangement of the M (or M′) tetrahedra in the layer (Figure S6). The interconnection of both T+ and T– networks presenting this arrangement type inevitably leads to the
presence of the Al2S6 dimers in Li3AlS3 on the contrary to the other two phases.
Figure 9
Representation
of the influence of the M/S = 1/3
ratio on the structure and arrangement of Li polyhedra in Li3AlS3 having the “Li5AlS4-type”
structure leading to the presence of ordered vacancies in the tetrahedral
layer.
Representation
of the influence of the M/S = 1/3
ratio on the structure and arrangement of Li polyhedra in Li3AlS3 having the “Li5AlS4-type”
structure leading to the presence of ordered vacancies in the tetrahedral
layer.These dimers have a crucial influence on the displacement
of atoms
within the structure. Because of the strong electronic repulsive forces
between the two Al3+ cations in the dimer, the Al position
is driven off the center of the tetrahedron and the S3–S3 distances
are successively compressed and stretched among the layers (Figure , entry 2). In Li5AlS4 and Li4.4M0.4M′0.6S4, on
the other hand, the S–S edges are not as compressed. This is
reflected in the values of the standard deviations for the S–S
distances: 0.13 (Li5AlS4), 0.20 (Li4.4Al0.4Ge0.6S4), and 0.24 (Li3AlS3) and in the maximum/minimum S–S distances
(in Å): 4.1564/3.6805 (Li5AlS4), 4.3176/3.6016
(Li4.4Al0.4Ge.6S4), and
4.4206/3.4631 (Li3AlS3). This strongly impacts
the geometry and the volume of the polyhedral sites in the adjacent
layer, i.e., the Li-only layer (Figure , entry 3). In Li3AlS3, the volumes
of the three different octahedral interstices are 26.7227, 27.1162,
and 30.4551 Å3. The latter is considerably bigger
than the other two and is not favorable to hosting small Li atoms.
This explains why only 2/3 of the octahedral sites are occupied in
the Li-only layer (Li2, Li2b, and Li3). In the same way, in Li4.4Al0.4Ge0.6S4, 3/4 of the
octahedral sites are occupied, and the volume of the ordered vacant
site is 31.5654 Å3. In Li5AlS4, all of the octahedral sites have the adequate geometry to host
lithium, therefore leading to a full occupation of the octahedral
sites within this layer. In Li3AlS3, it will
be preferable for the remaining Li atoms to occupy a tetrahedral site.
Li4 thus occupies the tetrahedral interstices, which are away from
the already occupied octahedra as well as away from the above (or
below) occupied Al tetrahedra (Figure , entry 3). The positions of the Li4 atoms are driven
toward the above (or below) S–S slab to minimize repulsion
with the other edge-shared Li4 tetrahedra, creating a pseudo-bipyramid
trigonal environment. The displacement of the Li4 atoms is then detrimental
to the occupation by another cation of the above or below tetrahedra
in the tetrahedral layer. The latter would then more preferably be
left vacant. This explains why among the two other tetrahedral sites
not occupied by Al, only one is occupied by Li, and a 1:1:1 arrangement
of Al, Li, and ordered vacancies is stabilized (Figure , entry 4). In Li4.4Al0.4Ge0.6S4, apart from the tetrahedral site that
lies just below the MS4 tetrahedra of
the above tetrahedral layer, all the other tetrahedral sites of the
Li-only layer are equivalent and Li atoms therefore randomly occupy
these positions. The motivation for the ordering of vacancies in the
tetrahedral layer is therefore suppressed, and the remaining Li ions
are delocalized among all tetrahedral sites in this layer. The ordered
vacancy sites, unfavorable to host lithium atoms, are likely to act
as a barrier for Li diffusion, and result in lower lithium ionic conductivity
in Li3AlS3 than in Li4.4M0.4M′0.6S4. A complete study of the Li energy landscape to elucidate Li diffusion
pathways will be undertaken to yield further insight into the role
of these structural features.In this structure type, the M/S ratio imposes
the arrangement pattern of the Al tetrahedra in dimers or isolated
tetrahedra and triggers the stabilization of ordered or disordered
vacancies within both the tetrahedral and the Li-only layer. The lithium
conductivity properties of each of the compounds can then directly
be related to the structure through the amount of disordered Li vacancies,
which itself can be explained by the composition through the M/S ratio. This work illustrates that, in this structure
type, an M/S ratio that is too large causes structural
arrangements that inhibit the diffusion of Li.
Conclusions
We have investigated the
Li–Al–O–S phase diagram
through a probe structure approach, which combined experimental and
computational studies to yield a new phase while ruling out others.
Indeed, this study revealed that no oxysulfide phases could be successfully
obtained, but led to the discovery of the new sulfide Li3AlS3. The structure and properties of this compound
were determined by means of high-resolution X-ray and neutron diffraction,
multinuclear NMR spectroscopy at various fields, and electrochemical
impedance spectroscopy. The stability of the new phase is believed
to rely on the presence of AlS4 dimers, a peculiar feature
not observed before in other Li ion conducting phases. The structure
was described by comparing the cation polyhedral arrangements with
those of other related phases with similar compositions, such as Na3MCh3 (M = Al,
Ga, In; Ch = S, Se), Li2FeS2, and Li5AlS4. The study of this new compound
in comparison with other similar sodium and lithium chalcogenide phases
widens the spectra of possibilities to explore new interesting structures
in related phase fields.
Authors: Matthew S Dyer; Christopher Collins; Darren Hodgeman; Philip A Chater; Antoine Demont; Simon Romani; Ruth Sayers; Michael F Thomas; John B Claridge; George R Darling; Matthew J Rosseinsky Journal: Science Date: 2013-04-11 Impact factor: 47.728
Authors: Philipp Bron; Sebastian Johansson; Klaus Zick; Jörn Schmedt auf der Günne; Stefanie Dehnen; Bernhard Roling Journal: J Am Chem Soc Date: 2013-10-09 Impact factor: 15.419
Authors: Marvin A Kraft; Sean P Culver; Mario Calderon; Felix Böcher; Thorben Krauskopf; Anatoliy Senyshyn; Christian Dietrich; Alexandra Zevalkink; Jürgen Janek; Wolfgang G Zeier Journal: J Am Chem Soc Date: 2017-07-28 Impact factor: 15.419