Effect
of Metal
d
Band Position
on Anion
Redox
in Alkali-Rich
Sulfides
Published
as part of Chemistry
of Materials
virtual
special
issue “C. N. R. Rao at 90”.
Seong Shik Kim, David N. Agyeman-Budu,
Joshua J. Zak, Jessica L. Andrews,
Jonathan
Li,
Brent C. Melot, Johanna
Nelson Weker, and Kimberly
A. See
*
Cite This:
Chem.
Mater.
2024,
36, 6454−6463
Read Online
ACCESS
Metrics
& More
Article
Recommendations
*
sı
Supporting
Information
ABSTRACT:
New energy
storage
methods
are emerging
to increase
the energy
density
of state-of-the-art
battery
systems
beyond
conventional
intercalation
electrode
materials.
For instance,
employ-
ing anion
redox
can yield
higher
capacities
compared
with transition
metal
redox
alone.
Anion
redox
in sulfides
has been
recognized
since
the early
days
of rechargeable
battery
research.
Here,
we study
the
effect
of
d
−
p
overlap
in controlling
anion
redox
by shifting
the metal
d
band
position
relative
to the S
p
band.
We aim to determine
the
effect
of shifting
the
d
band
position
on the electronic
structure
and,
ultimately,
on charge
compensation.
Two
isostructural
sulfides
LiNaFeS
2
and LiNaCoS
2
are directly
compared
to the hypothesis
that the Co material
should
yield
more
covalent
metal
−
anion
bonds.
LiNaCoS
2
exhibits
a multielectron
capacity
of
≥
1.7
electrons
per
formula
unit, but despite
the lowered
Co
d
band,
the voltage
of anion
redox
is close
to that of LiNaFeS
2
. Interestingly,
the material
suffers
from
rapid
capacity
fade.
Through
a combination
of solid-state
nuclear
magnetic
resonance
spectroscopy,
Co and S X-ray
absorption
spectroscopy,
X-ray
diffraction,
and partial
density
of states
calculations,
we demonstrate
that oxidation
of S nonbonding
p
states
to S
2
2
−
occurs
in early
states
of charge,
which
leads
to an irreversible
phase
transition.
We conclude
that the lower
energy
of
Co
d
bands
increases
their
overlap
with S
p
bands
while
maintaining
S nonbonding
p
states
at the same
higher
energy
level,
thus
causing
no alteration
in the oxidation
potential.
Further,
the higher
crystal
field stabilization
energy
for octahedral
coordination
over
tetrahedral
coordination
is proposed
to cause
the irreversible
phase
transition
in LiNaCoS
2
.
■
INTRODUCTION
Lithium-ion
batteries
(LIBs)
have
revolutionized
the modern
technological
era, with
applications
ranging
from
powering
portable
electronic
devices
to electric
vehicles.
1,2
Charge
compensation
of cathode
materials
in LIBs
predominantly
stems
from
the redox
activity
of
M d
states
in Li
M
O
2
, where
M
is a combination
of transition
metals
that typically
includes
some
combination
of Al, Mn, Co, and Ni. From
the original
commercial
intercalation-type
oxide
LiCoO
2
, partial
and/or
full substitution
of other
metals
such
as Ni, Mn, and Al has
enhanced
the structural
stability,
increased
the operating
voltage
energy
density.
3
−
6
To improve
the energy
density
beyond
the limits
of
conventional
intercalation
chemistry,
there
have
been
increas-
ing efforts
to leverage
both
cation
and anion
redox
in the
lattice
to achieve
multielectron
storage.
7,8
We define
multi-
electron
redox
as redox
exceeding
1 mol of e
−
per 1 mol of
transition
metal.
While
utilizing
anion
redox
in so-called
Li-rich
oxides
is highly
desirable,
many
Li-rich
oxide
materials
suffer
from
undesirable
side reactions
such
as irreversible
capacity
loss,
voltage
fade,
and electrolyte
decomposition
due to the
unstable
electronic
configurations
of oxidized
structures.
8
−
12
The
oxide
literature
is still
grappling
with
conflicting
mechanistic
explanations
for charge
compensation
in apparent
multielectron
materials.
An alternative
approach
to take advantage
of anion
redox
is
to replace
the elemental
O with the more
covalent
elemental
S,
at least
with respect
to the 3d transition
metals,
and develop
Li-rich
sulfide
materials.
Unlike
metal
oxides,
metal
sulfides
have
been
known
to form
stable
S
2
2
−
dimers
without
breaking
the transition
metal
−
S
bond
such
as in TiS
3
(Ti
4+
S
2
−
(S
2
)
2
−
),
pyrite
(
M
2+
(S
2
)
2
−
where
M
= Mn, Fe, Co, Ni, Cu, Zn), or VS
4
(V
4+
(S
2
)
2
−
).
13
−
15
Data
mining
of the Inorganic
Crystal
Received:
February
22, 2024
Revised:
June
9, 2024
Accepted:
June
12, 2024
Published:
June
21,
2024
Article
pubs.acs.org/cm
© 2024
The Authors.
Published
by
American
Chemical
Society
6454
https://doi.org/10.1021/acs.chemmater.4c00490
Chem.
Mater.
2024,
36, 6454
−
6463
This article is licensed under CC-BY-NC-ND 4.0
Structure
Database
(ICSD)
has revealed
that
numerous
thermodynamically
stable
binary
and
ternary
moieties
containing
persulfide
S
2
2
−
exist,
highlighting
the stable
nature
of S
2
2
−
.
12
Also,
anion
redox
in Li-rich
sulfides
occurs
within
the
electrochemical
window
of conventional
carbonate
electro-
lytes,
making
them
excellent
candidates
to study
anion
redox
and also as next-generation
cathode
materials.
In fact, Li
2
FeS
2
,
which
is one of the first, if not the first Li-rich
metal
sulfide,
can
achieve
energy
densities
of around
800 Wh/kg,
despite
the low
operating
voltage
of 2.4 V because
of the multielectron
redox
capabilities.
The
Li-rich
metal
sulfides
largely
show
the formation
of
persulfide
bonds
upon
oxidation.
However,
the making
and
breaking
of these
bonds
are likely
a contributing
factor
to
capacity
fade
and reduced
rate performance
compared
to
conventional
intercalation.
In this work,
we aim to target
a
more
covalent
alkali-rich
metal
sulfide.
We have
a few
hypotheses
as to what
could
happen
to anion
oxidation
with
a more
covalent
metal
−
anion
bond:
•
More
covalent
metal
−
anion
bonds
will stabilize
holes
in
the S band
by distributing
the charge
over
both
the
metal
and the anion.
Since
holes
are stabilized,
the
formation
of persulfide
bonds
will
be prevented,
resulting
in a more
reversible
and faster
anion
redox
process.
•
More
covalent
metal
−
anion
bonds
will shift
the S
p
states
down
in energy
due to hybridization/bonding
with
the metal
d
states
resulting
in a thermodynamic
increase
in the anion
oxidation
potential,
which
would
increase
energy
density.
•
More
covalent
metal
−
anion
bonds
will
shift
the
hybridized
states
to lower
energies
without
affecting
the position
of S nonbonding
p
states
that may
be
revealed
upon
oxidation.
If anion
oxidation
only occurs
from
the nonbonding
states,
then
no shift
in the
oxidation
potential
would
be observed.
A few examples
already
exist in which
the relative
positions
of transition
metal
d
states
and S
p
states
were
tuned
to study
the nature
of anion
redox
in Li-rich
sulfide
materials.
Transition
metal
substitution
of Fe
2+
, Co
2+
, and Ti
3+
for Ti
4+
in redox-inactive
d
0
Li-rich
Li
2
TiS
3
has proven
effective
in
activating
S oxidation
by introducing
a redox-active
transition
metal
whose
d
states
overlap
with those
of S
p
states.
14,16
−
18
However,
the oxidation
mechanism
is unclear
from
these
studies.
Additionally,
greater
d
−
p
overlap
can be achieved
by
shifting
the
p
states.
Substituting
S with even more
covalent
Se
(Li
2
FeS
2
−
y
Se
y
, Li
2
TiS
3
−
x
Se
x
, and (Li
2
Fe)S
1
−
x
Se
x
O) has dem-
onstrated
that
increasing
transition
metal
−
anion
covalency
through
increasing
Se content
results
in systematic
decrease
in
the redox
potential
as a result
of the formation
of hybridized
states
of transition
metal
d
, S
p
, and Se
p
bands.
19
−
21
As for
comparing
different
redox-active
transition
metals
in an
isostructural
phase,
(Li
2
M
)SO
(
M
= Fe
2+
, Mn
2+
, and Co
2+
)
has been
explored
which
revealed
that although
the role of S is
not fully
elucidated,
partial
substitution
of Co improves
the
structural
stability
while
that of Mn does
not.
22
−
24
In this work,
we turn
to alkali-rich
sulfide
LiNaFeS
2
that
displays
transition
metal
and S
2
−
/(S
2
)
2
−
redox
couples
and
achieves
reversible
multielectron
capacity.
25
From
LiNaFeS
2
,
we substitute
Co for Fe in an isostructural
manner
to form
LiNaCoS
2
. The
Co
d
states
are lower
in energy
and thus
overlap
more
with S
p
states
compared
to Fe
2+/3+
.
13,26
The shift
of the
d
states
is shown
in a density
of states
(DOS)
schematic
in Figure
1. We hypothesize
that the greater
overlap
of the
metal
d
band
with
the anion
p
band
will result
in a more
covalent
metal
−
anion
interaction,
allowing
us to determine
how
covalency
might
affect
anion
redox
in two isostructural
materials.
We report
that LiNaCoS
2
can be prepared
via solid-
state
synthesis
and is isostructural
to previously
reported
Li
2
FeS
2
and
LiNaFeS
2
.
25
LiNaCoS
2
exhibits
multielectron
capacity
with anion
redox
in cycle
1, but the material
is plagued
by a rapid
capacity
fade
from
an irreversible
conversion
reaction
that prevents
further
electrochemical
activity.
With
a
multitude
of characterization
techniques,
including
X-ray
diffraction,
solid-state
nuclear
magnetic
resonance
spectrosco-
py, and X-ray
absorption
spectroscopy,
we report
that
S
oxidation
proceeds
with
S
2
2
−
formation
and leads
to the
formation
of thermodynamically
stable
pyrite
CoS
2
and
lithiated
spinel
phases.
■
EXPERIMENTAL
SECTION
Material
Preparation.
All materials
were
stored
and prepared
in
an Ar-filled
glovebox
in which
the levels
of O
2
and H
2
O were
below
1
ppm.
LiNaCoS
2
was prepared
using
solid
state
methods,
from
Li
2
S
(Beantown
Chemical,
≥
99.9%),
Na
2
S (Fisher
Scientific
≥
99%),
Co
(Fisher
Scientific,
≥
99.8%),
and
S
8
(Acros
Organics,
≥
99.5%).
Stoichiometric
quantities
of Li
2
S, Na
2
S, Co, and S
8
were
ground,
pressed
into pellets
of up to 400 mg, and sealed
in a carbon-coated
evacuated
vitreous
silica
ampule.
The sealed
ampules
were
heated
at 2
°
C min
−
1
to 500
°
C, dwelled
for 96 h, and allowed
to cool to room
temperature
in the furnace.
The ampule
was opened
in a glovebox
revealing
a black
pellet
that was ground
into
powder.
LiNaCoS
2
retains
the pellet
morphology
upon
completion
of the reaction.
The
pristine
LiNaCoS
2
pellet
is dark gray/black,
and the color
is retained
upon
grinding.
Material
Characterization.
Powder
X-ray
diffraction
(XRD)
patterns
were
collected
by using
the Rigaku
SmartLab
diffractometer.
The powders
were
placed
on a 6 mm air-free
sample
holder
(Rigaku)
inside
the glovebox
to prevent
air exposure.
XRD
patterns
were
collected
with a Cu K
α
X-ray
source
at 3
°
min
−
1
with 0.04
°
step size.
The diffraction
patterns
were
fit by the Rietveld
method
using
GSAS-
II.
27,28
Visualization
of the crystal
structures
was aided
by VESTA.
29
High-resolution
synchrotron
powder
X-ray
diffraction
(sXRD)
patterns
were
collected
at the Advanced
Photon
Source
at Argonne
National
Laboratory
on beamline
11-BM-B
(
λ
= 0.4597
Å).
30
LiNaCoS
2
powder
was used for the pristine
state,
while
ex situ
charged
(fully
polarized
to 3 V) and discharged
(fully
charged
to 3 V then
discharged
to 1.7 V) samples
were
prepared
with 15 mg of a 60:20:20
wt % pellets
of LiNaCoS
2
, Super
P carbon
(TIMCAL),
and PTFE.
Figure
1.
Schematic
band
structure
showing
the expected
relative
positions
of isolated
(a) Fe
d
band
vs (b) the Co
d
band
vs the S
p
band
before
bonding
in LiNaFeS
2
and LiNaCoS
2
.
Chemistry
of Materials
pubs.acs.org/cm
Article
https://doi.org/10.1021/acs.chemmater.4c00490
Chem.
Mater.
2024,
36, 6454
−
6463
6455
The cathode
composite
was recovered
from
the disassembled
cell,
washed
with
200
μ
L of dimethyl
carbonate,
and vacuum-dried
for
more
than 4 h. The sample
was then sealed
under
vacuum
in 0.7 mm
(o.d.)
glass
capillaries
(Hampton
Research)
to prevent
air exposure
and placed
inside
polyimide
capillaries.
Operando
XRD
data were
collected
using
a Bruker
D8 Advance
diffractometer
in Bragg
−
Brentano
geometry
equipped
with a Cu K
α
source
(
λ
1
= 1.5406
Å,
λ
2
= 1.5444
Å) and a Lynxeye
XE-T
detector.
A custom-made
operando
cell with
a PEEK
body,
stainless
steel
electrical
contacts,
and
an X-ray-transparent
Be window
(SPI
Supplies,
0.25
mm thick)
was used.
The
Be window
served
as a
current
collector
and allowed
for X-ray
penetration
so that diffraction
patterns
could
be collected
while
cycling
galvanostatically.
As with the
ex situ
cells,
pellet
electrodes
composed
of 60% active
material,
20%
carbon
black,
and 20% polymer
binder
were
used
and placed
directly
in the Be window.
The pellet
electrodes
were
cycled
against
Li foil
using
a BioLogic
SP-200
potentiostat
at a C/20
rate with
one
Whatman
glass
fiber
separator
(GF/D)
flooded
with 1 M LiPF
6
in a
1:1:3
EC:PC:DMC
electrolyte.
Patterns
were
continuously
collected
over a range
of 10
°−
45
°
2
θ
approximately
every
20 min throughout
the duration
of the electrochemical
cycling.
Solid-state
nuclear
magnetic
resonance
(ssNMR)
spectroscopy
was
performed
using
a Bruker
Avance
500 MHz
spectrometer
operating
at
194.31
MHz
for
7
Li and 132.29
MHz
for
23
Na. All samples
were
packed
in a 4 mm ZrO
2
HR-MAS
rotor
with a 50
μ
L PTFE
spacer
(Cortecnet).
Pristine
LiNaCoS
2
(30 mg) was packed
into the rotor
in
an Ar-filled
glovebox.
Ex situ
samples
were
prepared
in the same
way
as electrochemical
cells,
with
a total
mass
of 20 mg.
7
Li and
23
Na
magic
angle
spinning
(MAS)
ssNMR
was recorded
at a spinning
rate
of 10.5 kHz.
Single
RF pulses
of 0.5
μ
s
−
π
/14
and 3
μ
s
−
π
/2
were
applied
for
7
Li and
23
Na, respectively.
A Bruker
dual channel
4 mm
wide
variable
temperature
probe
was used.
Different
spinning
rates
were
used
to identify
the isotropic
peaks.
1 M aqueous
solutions
of
LiCl
and NaCl
were
used
as standards
at 0 ppm.
Not washing
the
composite
electrode
resulted
in peaks
from
Li-containing
impurities,
such
as the electrolyte
salts.
Co and S K-edge
X-ray
absorption
spectroscopy
were
conducted
at
beamline
4-3 at the Stanford
Synchrotron
Radiation
Lightsource
at
SLAC
National
Accelerator
Laboratory.
Data
processing
including
calibration
and background
correction
was performed
using
Athena.
31
Co K-edge
X-ray
absorption
spectroscopy
data
were
calibrated
to
collinear
Co foil for each sample.
Ex situ
samples
were
prepared
in the
same
way as electrochemical
cells
with
a total
mass
of 20 mg. All
measurements
were
taken
within
2 weeks
of completing
the
electrochemical
activity.
Samples
were
transported
in Ar-sealed
pouches
prepared
in the glovebox
to prevent
exposure
to air.
Ex
situ
samples
were
placed
onto the sample
holder
by using
Kapton
tape
on each
side inside
an Ar-filled
glovebox.
During
measurement,
the
sample
holder
was placed
in a continuously
He-flushed
chamber
with
minimal
O
2
concentration
(<500
ppm).
S K-edge
XANES
data were
fit using
the pseudo-Voigt
function
to highlight
the emergence
of a
new peak.
Density
functional
theory
(DFT)
calculations
were
performed
using
SCAN
parametrization
with
the Vienna
ab initio
Simulation
Package
(VASP).
Interactions
between
core
and valence
electrons
were
accounted
for with
the projector
augmented
wave
(PAW)
pseudopotential
method.
The PAW
pseudopotentials
used
treat
the
1
s
, 2
s
, and 2
p
orbitals
of Li (Li
sv
); the 3
s
and 2
p
orbitals
of Na (Na
pv
);
the 4
s
and 3
d
orbitals
of Co; the 4
s
and 3
d
orbitals
of Fe; and the 3
s
and 3
p
orbitals
in S as valence
states.
The electronic
structure
was
converged
to a tolerance
of 10
−
5
eV, using
a
Γ
-centered
reciprocal
space
discretization
of 30K points
per Å
−
1
along
each reciprocal
lattice
vector
and a plane-wave
energy
cutoff
of 640 eV. All atomic
positions
were
converged
to a maximum
force
tolerance
of 0.02 eV/Å.
Static
runs were
performed
on the relaxed
structures
using
the tetrahedron
method
with Blochl
corrections.
Calculations
were
spin-polarized
with
magnetic
moments
initialized
in both
ferromagnetic
and antiferro-
magnetic
configurations.
The antiferromagnetic
configurations
yield
states
with
a lower
energy
by >50 meV/formula
unit and are thus
assumed
to be the more
stable
configuration.
Crystal
orbital
Hamilton
population
(COHP)
analysis
was
performed
for the metal
−
S
bond
in the antiferromagnetic
configurations
using
the LOBSTER
package.
32
−
35
Interactions
were
calculated
and summed
for all nearest-neighbor
metal
−
S
bonds
and
normalized
per formula
unit. COHPs
are plotted
between
the metal
d
and S
p
orbitals.
Electrochemical
Characterization.
All cells
were
prepared
inside
Ar-filled
gloveboxes
with
H
2
O and O
2
levels
below
1 ppm.
Powder
of LiNaCoS
2
was mixed
with
Super
P carbon
(TIMCAL,
≥
99%)
and
polytetrafluoroethylene
(PTFE,
Sigma-Aldrich)
at
60:20:20
wt % to prepare
free-standing
composite
electrodes.
Approximately
10 mg of the composite
mixture
was pressed
into
0.25 in. diameter
pellets
with a hand-operated
arbor
press
(19.0
mg/
cm
2
or 5.7 mAh/cm
2
based
on 1.7 e
−
removal
per formula
unit).
Coin
cells
(2032,
MTI)
were
prepared
with
polished
Li metal
counter/
reference
electrodes,
dried
18 mm diameter
glass
fiber
separators
(Whatman
GF/D),
and 11 drops
of electrolyte
(approximately
163
mg).
A 1 M solution
of LiPF
6
(Sigma-Aldrich,
≥
99.99%)
in a 1:1:3
mixture
of ethylene
carbonate
(Sigma-Aldrich,
>99%),
propylene
carbonate
(Sigma-Aldrich,
>99%),
and dimethyl
carbonate
(Sigma-
Aldrich,
>99%)
by volume
was used
as the electrolyte
in all cells.
Potentials
reported
here
are in reference
to the Li metal
electrode
potential,
which
is approximated
to be equal
to that
of Li/Li
+
.
Galvanostatic
cycling
experiments
were
performed
at C/10
based
on
1 e
−
per formula
unit.
The
galvanostatic
intermittent
titration
technique
(GITT)
was conducted
at a rate of C/10
based
on 1 e
−
per
formula
unit for 20 min with 4 h open
circuit
hold
rest periods.
■
RESULTS
DFT to Explore
the Electronic
Structure.
A central
hypothesis
of this paper
is that the covalency
between
the
metal
and the anion
can be tuned
by substituting
Co for Fe in
LiNaFeS
2
. More
covalent
metal
−
anion
bonds
could
help
to
stabilize
holes
in the S
p
band
and/or
lower
the energy
of the S
p
states
and thermodynamically
increase
the anion
oxidation
potential,
making
sulfides
a more
viable
cathode
choice.
By
shifting
from
LiNaFeS
2
to LiNaCoS
2
, we hypothesize
that the
lower
energy
d
band
of Co will overlap
more
with the S
p
band
to yield
a more
covalent
material.
We first
explore
this
hypothesis
through
DFT
calculations
of the electronic
structure.
The
spin-polarized
partial
density
of states
(pDOS)
for antiferromagnetic
ordered
forms
of LiNaFeS
2
and LiNaCoS
2
is shown
in Figure
2. Both
materials
exhibit
a
Figure
2.
Calculated
spin-polarized
partial
density
of states
of
antiferromagnetic
ordered
forms
of (a) LiNaCoS
2
and (b) LiNaFeS
2
.
A greater
degree
of Co
−
S
overlap
in LiNaCoS
2
is observed
near the
Fermi
level compared
to Fe
−
S
overlap
in LiNaFeS
2
. LiNaFeS
2
shows
significant
Fe character
near the Fermi
level.
Chemistry
of Materials
pubs.acs.org/cm
Article
https://doi.org/10.1021/acs.chemmater.4c00490
Chem.
Mater.
2024,
36, 6454
−
6463
6456
small
gap at the Fermi
level,
suggesting
a semiconducting
electronic
ground
state,
though
the band
gap is an estimation
as the calculations
are conducted
only
with
SCAN
para-
metrizations.
However,
here
we are more
interested
in the
relative
positions
of the metal
d
states
vs the S
p
states.
The
states
below
the Fermi
level are composed
of both
S
p
and Co
d
character.
The
high
degree
of overlap
suggests
a more
covalent
Co
−
S
bond
compared
to the Fe
−
S
bond
in
LiNaFeS
2
. LiNaFeS
2
shows
a much
higher
degree
of Fe
character
just below
the Fermi
level compared
to S. To further
evaluate
the bonding
character,
the crystal
orbital
Hamilton
population
(COHP)
analysis
of the metal
−
S
bonds
in
LiNaCoS
2
and
LiNaFeS
2
is shown
in the Supporting
Information.
Structural
Characterization.
Although
this paper
focuses
on the comparison
between
LiNaFeS
2
and LiNaCoS
2
, we
initially
began
our study
targeting
a comparison
of Li
2
FeS
2
and
Li
2
CoS
2
�
a simpler
system
with
only
one alkali
element.
Interestingly,
using
the same
solid-state
reaction
conditions
used
for Li
2
FeS
2
with
Co substitution
does
not result
in the
formation
of Li
2
CoS
2
, indicating
that
Li
2
CoS
2
cannot
be
synthesized
via traditional
solid-state
synthesis
methods.
However,
alkali-rich
LiNaCoS
2
can be prepared
using
similar
reaction
conditions
to that of previously
reported
LiNaFeS
2
.
25
Synchrotron
X-ray
diffraction
(sXRD)
is employed
to
characterize
the structure
of LiNaCoS
2
. sXRD
patterns
and
quantitative
Rietveld
refinement
results
of pristine
LiNaCoS
2
are shown
in Figure
3. LiNaCoS
2
adopts
the
P
3
̅
m
1 space
group
and is isostructural
to previously
reported
LiNaFeS
2
.
25
In
LiNaCoS
2
, layers
of edge-sharing
Na octahedra
are separated
by layers
of edge-sharing
mixed
Li/Co
tetrahedra
(Figure
3).
The Li and Co are disordered
and share
the same
tetrahedral
site at a 1:1 ratio.
The structure
of LiNaCoS
2
was reported
by
Ren et al. and agrees
with their
assignment
of the space
group
as well
as mixed
occupancy
of Li and Co.
36
The
sharp
reflections
indicate
high
crystallinity.
The diffraction
patterns
are well-described
by a two-phase
fit to LiNaCoS
2
with
a
0.0032
phase
fraction
of Co
9
S
8
(<1.6
wt %), suggesting
a very
small
impurity
of Co
9
S
8
. The lattice
parameters
of LiNaCoS
2
are
a
= 3.95618
Å and
c
= 6.73845
Å and agree
well with those
reported
by Ren et al.,
a
= 3.95710
Å and
c
= 6.70108
Å.
36
Electrochemical
Characterization.
To probe
the effect
of transition
metal
anion
covalency
on redox
potential
and
reversibility,
the electrochemical
performance
of LiNaCoS
2
is
examined
with
galvanostatic
cycling
and
compared
to
LiNaFeS
2
. LiNaFeS
2
was first studied
by Hansen
et al.
25
Figure
4a depicts
the first charge
and discharge
curve
for both
materials
at C/10
based
on 1 e
−
per formula
unit (see Figure
S1 for
Q
in mAh/g).
The first charge
of LiNaFeS
2
has been
studied
in detail
previously.
25
Briefly,
the first sloping
region
corresponds
to Fe
2+
oxidation
to mixed
Fe
2+/3+
followed
by a
kink and another
sloping
region
corresponding
to oxidation
of
S
2
−
to S
2
2
−
.
25
The
average
voltage
of the anion
oxidation
region
of the curve
is 2.7 V. The capacity
of LiNaFeS
2
reaches
over 1.7 mol of e
−
per formula
unit.
Next,
we discuss
the first charge
curve
of LiNaCoS
2
and
compare
it with that of LiNaFeS
2
. Perhaps
unsurprisingly,
the
charge
curve
of LiNaCoS
2
bypasses
the low voltage
processes
associated
with
Fe
2+
oxidation
in LiNaFeS
2
. Instead,
we
observe
a single
plateau
with
an average
potential
of 2.6 V.
Despite
the presence
of Co in the structure,
the oxidation
processes
remain
at very
similar
potentials
compared
to
LiNaFeS
2
. In fact, the voltage
is comparable
to potentials
at
which
anion
oxidation
is suggested
in other
Li-rich
sulfide
materials.
14,17,18,20,25
Unlike
the isostructural
material
Li-
NaFeS
2
, no distinct
region
associated
with
transition
metal
oxidation
is observed.
This trend
differs
from
anion-substituted
Li
2
FeS
2
−
y
Se
y
in which
mixing
of the S and Se states
yields
a
systematic
shift
in the voltage
plateau.
19
The
subsequent
discharge
cycle
for LiNaCoS
2
is characterized
mainly
by a
single
plateau
but at much
lower
potentials
compared
to the
charge,
resulting
in large
voltage
hysteresis.
Next,
we compare
the cycling
behavior
of LiNaFeS
2
to that
of LiNaCoS
2
. The
discharge
and charge
profiles
for both
materials
are compared
in Figure
4b. The capacity
of LiNaFeS
2
fades
somewhat
quickly,
largely
due to particle
fracturing
associated
with
the removal
of large
Na
+
.
37
Though
the
capacity
fades,
the cycle
1 and cycle
2 charge
curves
are similar
in shape,
suggesting
that the mechanism
is largely
unchanged.
LiNaCoS
2
, however,
shows
much
more
rapid
capacity
fade
losing
most
of its electrochemical
activity
by cycle
20 and
displaying
significant
changes
in shape
as cycling
continues.
Figure
3.
Synchrotron
powder
X-ray
diffraction
patterns
and
quantitative
Rietveld
refinement
results
of LiNaCoS
2
. The LiNaCoS
2
unit cell is shown
in the inset.
The ticks indicate
the Bragg
reflection
locations
with a two-phase
fit to 99.68%
LiNaCoS
2
and 0.32%
Co
9
S
8
.
Figure
4.
Galvanostatic
cycling
of LiNaFeS
2
compared
with
LiNaCoS
2
for (a) cycle
1 and (b) cycles
2 and 10. The materials
are cycled
at C/10
based
on 1 e
−
per formula
unit.
Chemistry
of Materials
pubs.acs.org/cm
Article
https://doi.org/10.1021/acs.chemmater.4c00490
Chem.
Mater.
2024,
36, 6454
−
6463
6457
The
capacity
as a function
of cycle
number
is shown
in
Supporting
Information
(Figure
S2). The charge
and discharge
voltage
profiles
after
cycle
1 resemble
those
of CoS
2
cycled
between
3 and 1.6 V, suggesting
a similar
mechanism
in
LiNaCoS
2
after
the first charge.
38
When
charged
to 3 V and
then
discharged
to below
1 V, the discharge
curve
resembles
that
of CoS
2
, further
suggesting
that
charge
products
are
similar
to those
of CoS
2
(Figure
S3). The hysteresis
remains
high throughout
the first cycle
at around
0.5 V, which
is much
larger
than that observed
for LiNaFeS
2
(0.27
V at the start of
anion
oxidation,
i.e., approximately
0.6 mol
e
−
25
). The
significant
hysteresis
coupled
with the change
in shape
suggests
that
the oxidation
mechanism
is different
in LiNaCoS
2
compared
with LiNaFeS
2
.
To probe
overpotentials
of LiNaCoS
2
during
cycling
and
determine
the origin
of the hysteresis,
the galvanostatic
intermittent
titration
technique
(GITT)
is performed
on
LiNaCoS
2
. The
cell is polarized
for 20 min
at C/10
per
formula
unit with 4 h open
circuit
hold rest periods.
The near-
equilibrium
potential
at the end of each
relaxation
period
is
approximated
as the equilibrium
potential,
V
eq
, and
the
overpotential,
η
, represents
the potential
difference
between
the beginning
and end of each
rest period.
The
GITT
profiles
for the first
and second
cycles
of
LiNaCoS
2
are plotted
against
the trace
obtained
at C/10
in
Figure
5. The GITT
profile
during
first charge
reveals
high
overpotentials.
When
the voltage
profile
is plotted
as a function
of time (Figure
S4), the potential
overshoots
upon
polarization
suggesting
nucleation
and growth
behavior.
39,40
When
GITT
is
performed
at C/100,
the overpotential
associated
with
the
nucleation
region
decreases
significantly
(Figure
S5), suggest-
ing that the nucleation
and growth
behavior
are caused
by the
sluggish
kinetics.
During
discharge,
an initial
sloping
region
is
followed
by a plateau
(Figure
5a). Also,
the overpotential
spike
is no longer
observed,
and the overpotential
is lower
than
during
charge,
suggesting
an irreversible
phase
transition
at the
end of cycle
1. The large
voltage
hysteresis
between
charge
and
discharge
equilibrium
potentials
further
suggests
that
the
charge
and discharge
pathways
differ
from
each
other.
41
In
cycle
2, the charge
and discharge
GITT
profiles
trace
the
galvanostatic
cycling
data
well
and resemble
that of CoS
2
cycled
between
1.6 and 3 V, further
suggesting
irreversible
phase
transition.
38
Structural
and Spectroscopic
Evolution
during
Cycling.
Next,
we undergo
a series
of studies
to understand
why
LiNaCoS
2
cycles
differently
from
LiNaFeS
2
. First,
we
investigated
the structural
evolution
of LiNaCoS
2
during
cycling.
Solid-state
nuclear
magnetic
resonance
(ssNMR)
spectroscopy
is employed
to explore
the local
Li and Na
environments.
7
Li and
23
Na magic-angle
spinning
(MAS)
ssNMR
spectra
of LiNaCoS
2
at different
states
of charge
collected
at 10.5 kHz are shown
in Figure
6. A single
resonance
is observed
at 87.4 and 652 ppm for
7
Li and
23
Na (marked
∗
),
respectively,
indicating
a single
unique
environment
for both Li
and Na in pristine
LiNaCoS
2
and confirming
the Wyckoff
site
assignments
based
on sXRD.
At 0.2 e
−
charge
(charge
capacity
equivalent
to 0.2 e
−
removal
per formula
unit),
a new resonance
is observed
at
−
1
ppm in
7
Li MAS
ssNMR
(marked
+ ), suggesting
that a new Li
environment
begins
to emerge
at early
states
of charge.
This
new resonance
grows
in intensity
during
charge,
as evidenced
by the higher
intensity
in the spectrum
of the material
oxidized
by 1.0 e
−
, while
the intensity
of the original
resonance
at 87.4
ppm
decreases.
This
result
indicates
that in addition
to Li
removal
during
charge,
the active
material
undergoes
a phase
transition,
resulting
in a new Li environment.
No new peaks
are observed
in the
operando
XRD
(
vide infra
), so the new
environment
could
be associated
with
small
crystallites,
an
amorphous
phase,
or new Li environments
associated
with the
structural
distortions
required
for persulfide
formation,
for
instance
tetrahedral
tilting.
25
At full charge
(1.7
e
−
), both
resonances
disappear
in the
7
Li MAS
ssNMR,
suggesting
that
Li can be removed
from
the new environment.
Meanwhile,
23
Na MAS
ssNMR
spectra
reveal
that the intensity
of the single
resonance
in the pristine
sample
decreases
throughout
charge.
Figure
5.
Near-equilibrium
cycling
curves
of LiNaCoS
2
obtained
by
GITT
compared
with the galvanostatic
cycling
curves
measured
at C/
10 for (a) cycle
1 and (b) cycle
2. GITT
is measured
at C/10
with 4 h
rest periods
every
20 m.
Figure
6.
(a)
7
Li and (b)
23
Na MAS
ssNMR
spectra
of LiNaCoS
2
in
different
states
of charge.
Pristine
LiNaCoS
2
exhibits
a single
resonance
marked
∗
in both
7
Li and
23
Na spectra,
confirming
a
single
environment
for Li and Na. During
charge,
a new feature
(+) is
observed
in
7
Li spectra
while
23
Na spectra
exhibit
a decrease
in
intensity
of the single
resonance.
Chemistry
of Materials
pubs.acs.org/cm
Article
https://doi.org/10.1021/acs.chemmater.4c00490
Chem.
Mater.
2024,
36, 6454
−
6463
6458