You are on page 1of 9

PCCP

View Article Online


PAPER View Journal | View Issue

Enhancing sodium ionic conductivity in


tetragonal-Na3PS4 by halogen doping: a first
Cite this: Phys. Chem. Chem. Phys.,
2018, 20, 20525 principles investigation†
He Huang,b Hong-Hui Wu, c
Xinjiang Wang, b
Baoling Huang *b and
Tong-Yi Zhang*a
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

Tetragonal Na3PS4 (t-Na3PS4) has been demonstrated as a very promising candidate for a solid-state
sodium-ion electrolyte with high Na ionic conductivity at ambient temperature. In this paper, we
systematically investigated the Na ionic conductivity in pristine and halogen (F, Cl, Br, and I) doped
tetragonal-Na3PS4 superionic conductors using first-principles calculations. The Na ionic conductivity of
pristine t-Na3PS4 is calculated to be about 0.01 mS cm1, while much higher Na ionic conductivities
could be achieved by introducing Na ion vacancies via a halogen doping strategy. The calculated Na
ionic conductivity of t-Na3PS4 doped with 1.56% Cl is 1.07 mS cm1 at ambient temperature. Among
Received 14th April 2018, different halogen-doped t-Na3PS4, Br-doped t-Na3PS4 shows the lowest activation energy and the
Accepted 17th July 2018 highest Na ionic conductivity, which reaches 2.37 mS cm1 at 300 K. The low activation energy and high
DOI: 10.1039/c8cp02383b Na ionic conductivity in Br-doped t-Na3PS4 are due to a relatively lower defect binding energy of the
defect pair of halogen substitution and a Na ion vacancy. Our results suggest Br-doped t-Na3PS4 may
rsc.li/pccp serve as a very promising Na-ion solid-state superionic conductor.

Introduction Up until now, there are only a few existing Na-ion solid state
electrolytes with conductivities approaching those of liquid
Na-ion batteries (NIBs) have the potential to compete with electrolytes (1–6 mS cm1 at room temperature).14 Currently
commercially used Li-ion batteries (LIBs) for stationary energy oxide ceramic, sulphide glass–ceramic and borate Na-ion solid
storage applications and innovate the landscape of electro- conductors are the main classes of potential Na-ion SSEs under
chemical energy storage, due to the evenly distributed abun- intensive investigation.9,12,15 NASCON-type oxides9,16,17 (such
dance of the sodium resource and the cost advantages. Recent as Na1+xZr2SixP3xO1217 with 0 r x r 3) and b00 -alumina18,19 are
decades have witnessed the active development of Na-ion well-known Na-ion SSEs exhibiting high room-temperature
batteries.1–5 For both Li-ion and Na-ion batteries, the potential ionic conductivities over 1 mS cm1. However, the harsh
hazards associated with current organic liquid electrolytes (e.g., synthesis conditions (such as high synthesis temperatures over
leakage and flammability) are one of the major hurdles limiting 1473 K), high annealing temperatures (41273 K), and poor
the widespread application of alkali ion batteries in large-scale room-temperature electrolyte–electrode contact make their
electronic devices and engines.6,7 Replacing the organic liquid large-scale applications expensive and complicated. Recently,
electrolytes with inorganic solid-state electrolytes (SSEs), in principle, Udovic et al.20 reported that Na2B12H12, which can be synthesized
could eliminate the safety concerns, offering a promising way easily, shows an ionic conductivity of the order of 0.1 S cm1
forward to extend the application of alkali ion batteries to above 543 K, although its room-temperature conductivity is
industry where safety is of utmost importance.8–13 relatively low. In contrast to previous Na superionic conductors,
sulphide-based Na-ion SSEs21–24 are more promising, owing to
their high ionic conductivities, low synthesis temperatures and
a
Shanghai University Materials Genome Institute and Shanghai Materials Genome low grain boundary resistances. Therefore, they have been
Institute, Shanghai University, 333 Nancheng Road, Shanghai 200444, China. attracting great interest.25–28 Inspired by sulphide-based Li-ion
E-mail: zhangty@shu.edu.cn SSEs, glass–ceramic Na3PS426 was experimentally found to have a
b
Department of Mechanical and Aerospace Engineering, Hong Kong University of
room-temperature conductivity of 0.2 mS cm1, which was further
Science and Technology, Kowloon, Clear Water Bay, Hong Kong, China.
E-mail: mebhuang@ust.hk
increased to 1.46 mS cm1 by cation substitution of P with
c
Department of Chemistry, University of Nebraska-Lincoln, Lincoln, NE 68588, USA As (Na3P0.62As0.38S4)29 and 1.16 mS cm1 by anion substitution
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8cp02383b of S with Se (Na3PSe4).30 Through the aliovalent doping strategy,

This journal is © the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 | 20525
View Article Online

Paper PCCP

the introduction of Na ion defects (interstitial or vacancies) would


greatly improve the ionic conductivities of Na-ion SSEs. In previous
studies, high ionic conductivities induced by the excess of Na
ions via the substitution of P5+ with Si4+, Ge4+ and Sn4+ have been
theoretically investigated.27 Meanwhile, outstanding performance
was achieved by creating Na ion vacancies via halogen anion
doping for S2. Klerk et al.28 have performed an ab initio MD
simulation of c-Na3PS4 doped with 3.2% halogen at 525 K. An ionic
conductivity higher than 1 mS cm1 has been experimentally
reported by Chu et al.31 for the t-Na3PS4 superionic conductor
doped with 1.6% Cl. This makes halogen doped sulphide-based
Na-ion SSEs good candidates for the design of promising Na-ion
all-solid-state batteries. Some other halogen doped superionic
conductors (lithium superionic argyrodites and lithium-rich anti-
perovskites)32–36 have been investigated to achieve better cation
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

ionic conductivities, which are attributed to the optimal anion


polarizability and concerted cation motions. Although chlorine-
doped Na3PS4 has been characterized previously, to the best of our Fig. 1 (a) Unit cell of cubic-Na3PS4. (b) Unit cell of tetragonal-Na3PS4.
knowledge, the influences from different halogen dopants on Na (c) Lowest energy structure of t-Na47P16S63Br determined from DFT
ion transport in t-Na3PS4 and the corresponding Na ion diffusion calculations by replacing one S2 with Br and introducing one Na+
vacancy in the pristine 2  2  2 t-Na3PS4 supercell. Na I in gray, Na II
mechanisms have not yet been systematically explored.
in blue, S in yellow, Br in green and P in purple.
Herein, we performed a first principles based molecular
dynamics calculation of the Na ionic conductivity in pristine
and halogen (F, Cl, Br, and I) doped tetragonal-Na3PS4 super- that the static energy of t-Na3PS4 is 5.51 meV per atom lower
ionic conductors. We found that at room temperature, the Na than that of c-Na3PS4, which is in close agreement with the
ionic conductivity of pristine t-Na3PS4 is only about 0.01 mS cm1, value of 5 meV per atom reported by Ong et al.27 Furthermore,
while much higher Na ionic conductivities could be achieved with it has been experimentally demonstrated that t-Na3PS4 is easier
the introduction of Na ion vacancies by halogen doping. Our to synthesize compared with c-Na3PS4, which could only be
results show that the Na ionic conductivity of t-Na3PS4 doped with successfully synthesized with less than 0.1 g of precursors
1.56% Cl could reach 1.07 mS cm1, which is consistent with the inside either an Al or quartz tube at 50.37 Therefore, we choose
experimental observation.28 Further, a Na ionic conductivity as tetragonal-Na3PS4 as our pristine study object in this paper.
high as 2.37 mS cm1 is predicted for Br-doped t-Na3PS4. Through All DFT calculations in this work were performed using
structural optimization and topology analysis, it is shown that the the Vienna Ab initio Simulation Package (VASP)38 with the
halogen dopants would not affect the overall structure of pristine projected augmented wave approach.39 Generalized gradient
t-Na3PS4, all of which display a 3D Na ion diffusion network. The approximation (GGA) Perdew–Burke–Ernzerhof (PBE) was
binding energy between the halogen substitution and Na ion adopted for the exchange–correlation functional. For sodium,
vacancy defect pair in the halogen doped structure reveals the phosphorus, sulfur, and halogen atoms, we used pseudo-
mechanism of the much lower activation energy and the higher Na potentials with the valence configurations p6s1, s2p3, s2p4,
ionic conductivity in Br-doped t-Na3PS4. A further higher Na ionic and s2p5, respectively. The structural relaxations for pristine
conductivity and lower activation energy could be achieved by and halogen doped t-Na3PS4 (the doping level is 1.56%) were
introducing a higher halogen atom substitution concentration. performed based on 2  2  2 supercells containing 128 atoms
for the pristine structure and 127 atoms for the doped structures.
The Brillouin zone was sampled by a 3  3  3 Monkhorst–Pack
Calculation methods grid and the energy cutoff was set at 520 eV. All geometry
structures were fully relaxed until convergence criteria of forces
Fig. 1 shows the two crystal phases of Na3PS4, i.e., the cubic and energy of 0.01 eV Å1 and 105 eV were reached, respectively.
phase (c-Na3PS4) and the tetragonal phase (t-Na3PS4).27,28 All energy cutoff and k-points have been tested until the energy
c-Na3PS4 is crystalized in the space group I43% m with the lattice reaches its convergence. The fully optimized structures were used
parameters a = b = c = 7.02 Å, while t-Na3PS4 is in the space as the initial configurations for the subsequent calculations.
group P42% 1c with the lattice parameters a = b = 7.01 Å, and Due to the high cost of ab initio molecular dynamics (AIMD)
c = 7.16 Å.27,28,31 In the tetragonal phase, the Na (6b)-site in the methods,40 gamma-point-only sampling of k-space and a plane-
c-Na3PS4 structure will split into two Na ion sites, i.e., the Na I wave energy cutoff of 280 eV were used in the AIMD simulations
(2a) site and the Na II (4d) site, as shown in Fig. 1(c). Although for both the pristine and halogen doped t-Na3PS4 structures,
the two different phases of Na3PS4 are both in stable existence except for fluorine doped cases with a higher energy cutoff
with less than 2% difference of lattice parameters and similar of 400 eV. All energy cutoff values were carefully tested and
Na+ and PS43 polyhedral ion configurations, it is calculated selected to ensure the computation accuracy and convergence

20526 | Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 This journal is © the Owner Societies 2018
View Article Online

PCCP Paper

during the AIMD simulations. The simulation domain was a doped t-Na3PS4 via AIMD simulations. Through the AIMD
supercell of 2  2  2 optimized t-Na3PS4 lattice cells. To simulate simulations, the mean square displacements (MSDs) of Na
the Na ion vacancy effect induced by halogen dopants, one ion diffusion were extracted. According to the calculated MSDs
halogen atom X (X represents F, Cl, Br, or I) was introduced into of Na ions, the tracer diffusivity or self-diffusivity can be
the supercell by replacing one of the 48 S atoms with X while calculated using the Einstein equation:43
introducing one Na ion vacancy inside t-Na47P16S63X, corres- 1 D E
ponding to 2.08% Na ion vacancies via 1.56% halogen doping. D¼ ½rðtÞ2 ; (1)
2dt
Structure optimizations were performed for different configura-
tions containing both halogen substitution and Na ion vacancies. where t is the total simulation time, d the dimensionality of the
The AIMD simulations were performed in NVT ensembles diffusion pathway (3 in this case), and h[r(t)]2i the mean square
with Nosé–Hoover thermostats and the time steps were set to displacement of the diffusion species. By using the Nernst–
2 fs. Target temperatures ranged from 600 K to 1400 K. In each Einstein equation, the ionic conductivity s can be related to the
case, an equilibrium time of 2 ps was applied and the total tracer diffusivity44 through
simulation time was 100 ps. The Na ion trajectories at the ne2 z2 D
equilibrium state were dumped and then were used to calculate s¼ ; (2)
kT
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

the Na ionic conductivities and to determine the activation


energies. Migration barriers for Na ion diffusion in the pristine where n is the Na ion number density, e is the elementary
t-Na3PS4 structure were calculated using the climbing nudged charge, z is the ionic charge, D is the tracer diffusivity, k is
elastic band method (CI-NEB)41 through the VTST code42 linked Boltzmann’s constant, and T is the temperature in Kelvin. It is
with VASP. A 2  2  2 k-point grid and plane wave energy known that the temperature dependency of ionic conductivity
cutoff of 520 eV were used. could be represented by the conventional Arrhenius relationship:45
 
Ea
sT ¼ s0 exp  ; (3)
kT
Results and discussion
where s0 denotes the pre-exponential factor, which depends on
The first principles calculations found that the halogen atom temperature in a much weaker manner than the exponential
prefers to be distributed close to the nearest Na ion vacancy in function in the Arrhenius equation, k the Boltzmann constant,
the 2  2  2 supercell to form a defect pair of halogen T the absolute temperature, and Ea the activation energy of
substitution and a Na ion vacancy XS VNa 0
. Fig. 1(c) typically Na-ion diffusion. It implies that a lower Ea leads to a higher
shows the lowest energy structure of t-Na47P16S63Br with one ionic conductivity at a given temperature. By fitting the MSD of
distinct XS VNa
0
defect pair inside. Na ions at different elevated temperatures during the simula-
Table 1 shows the structural parameters of the relaxed tion period, the diffusivity and conductivity of Na ions in
t-Na3PS4 supercell and the halogen doped structures. It is found pristine and halogen doped Na3PS4 were computed. As shown
that the initial orthorhombic tetragonal crystal becomes slightly in Fig. 2(a), the introduction of Na ion vacancies could signifi-
distorted as the XS VNa0
defect pair exists non-symmetrically cantly increase the Na ion diffusivity and lower the activation
inside the lattice. The volume of the pristine 2  2  2 Na3PS4 energy. Pristine t-Na3PS4 exhibits a very low Na ionic conduc-
supercell is only around 0.1–0.8% smaller than those of the tivity of 0.01 mS cm1 at 300 K with a relatively high activation
corresponding t-Na47P16S63X structures, indicating that the energy of 366.1 meV, consistent with the experiment results
halogen doped structures could maintain the original tetra- from Chu et al.,31 while a 2% Na ion vacancy concentration
gonal lattice even with the existence of the XS VNa
0
defect pair. introduced by halogen doping results in nearly 2 orders of
The small variation in the lattice parameters of t-Na47P16S63X is magnitude larger Na ionic conductivities, especially for Cl
mainly due to the different anionic radii. The effective ionic doping (1.07 mS cm1) and Br doping (2.39 mS cm1). The
radii for S2, F, Cl, Br and I are 184, 133, 181, 196 and extracted activation energies in the doped structures are also
220 pm, respectively, which give a smaller unit cell for the much lower compared with that of the pristine structure,
F-doped structure and larger ones for Br- and I-doped ones. especially for the Br-doped case (only about 216 meV). The
To investigate the Na ionic conduction, we evaluated the Na calculated Na ionic conductivity and energy barrier for
ion diffusion properties by calculating the self-diffusion coeffi- Cl-doped Na3PS4 agree quite well with the experimental values
cient of Na ions in the pristine t-Na3PS4 supercell and halogen reported by Chu et al.31

Table 1 Structural parameters for the optimized pristine Na3PS4 supercell and halogen doped structures

Structure Halogen site Na+ vacancy site a (Å) b (Å) c (Å) a (1) b (1) g (1) Volume (Å3)
t-Na48P16S64 — — 14.018 14.018 14.310 90.00 90.00 90.00 2812
t-Na47P16S63F (8e) (2a) 13.992 14.010 14.383 90.31 90.02 90.11 2819
t-Na47P16S63Cl (8e) (2a) 14.010 14.009 14.392 90.12 89.95 89.98 2825
t-Na47P16S63Br (8e) (2a) 14.023 14.022 14.397 90.03 89.94 89.98 2831
t-Na47P16S63I (8e) (2a) 14.043 14.050 14.399 89.93 89.87 89.91 2841

This journal is © the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 | 20527
View Article Online

Paper PCCP

Fig. 2 (a) Arrhenius plot for pristine Na48P16S64 and halogen doped t-Na47P16S63X from AIMD simulations. The inset expanded the left corner panel for a
clearer version. (b) Isosurface of the Na ion probability distribution network with an isovalue of Pmax/ in t-Na47P16S63Br at 1000 K plotted by the pymatgen
diffusion analysis code.1
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

We used the open source Zeo++ software46,47 to perform a diffuse, which seems like the Na ion diffusivity should be the
topological analysis of the free volume inside which Na ions highest in the I-doped structure. However, it is not the fact in our
could normally flow. Na ions were removed from each structure AIMD simulations.
to study the largest radius of a free sphere that Na ions could In this case, if we consider the Na ion vacancy as a kind of
successfully pass through in the remaining PS43 skeleton, ion of zero mass, the diffusion of Na ions could be regarded as
shown as the channel size in Table 2. All the channel sizes of the diffusion of Na ion vacancies in the opposite direction
the studied systems are very similar with a difference less than inside pristine and halogen doped t-Na3PS4. Owing to the
0.01 Å. That means the introduction of halogen atoms and Na existence of halogen substitution and the nearest Na ion
ion vacancies would not lead to considerable influence on the vacancy in the most stable halogen doped t-Na3PS4, it would
diffusion channel. Additional insights may be obtained by be reasonable to assume that Na ion vacancies induced by
analyzing the trajectories from the AIMD simulations. A plot halogen doping are not free but bound to the halogen dopants
of the probability density function P(r) can provide useful to form defect pairs at 0 K. As the thermal dissociation of defect
information on the low energy (high probability) sites in pairs requires supplementary energy in addition to the energy
crystals, as well as the migration pathways between them. P is for migration of free ion vacancies,48,49 we estimated the
often defined in a spatial 3D uniform grid and can be computed activation energy of Na ion diffusion by dividing it into the
by averaging the number of alkali ions at each grid point within binding energy of the XS VNa 0
defect pair and the migration
Ð
a given time scale. P is normalized such that Pdr ¼ 1. Fig. 2(b) energy of the Na ion in the structure containing charged
typically shows the isosurface of the Na ion probability distribu- vacancies in the absence of halogen substitution, i.e.,50
tion network in Br-doped t-Na3PS4 at 1000 K plotted by the
Eactivation = Ebinding + Emigration, (4)
pymatgen diffusion code.1 It demonstrates that halogen doped
t-Na3PS4 keeps a 3D diffusion pathway composed of the equili- where Ebinding and Emigration are the binding energy of the
brium Na sites interacting with each other, which forms a body- XS VNa
0
defect pair in halogen doped t-Na3PS4 and the migra-
centered cubic diffusion network, consistent with the diffusion tion energy of the Na ion in pristine t-Na3PS4, respectively.
network of t-Na3PS4.27 Meanwhile, we also calculated the channel As the result in Fig. S1 (ESI†) shows little difference between
volume of all the systems by using the whole unit cell volume to the migration energy of the Na ion in t-Na3PS4 and that in the
subtract the volume of the polyhedrons skeleton (PS43 or PXS33), Br-doped pristine structure, it is convenient to study Na vacancy
which is listed as the channel volume in Table 2. It shows that migration by omitting the effect of different kinds of halogen
I-doped t-Na3PS4 has the largest channel volume for the Na ion to elements when the Na ions are remote from those halogen atoms.

Table 2 Channel size, channel volume, extrapolated Na+ diffusivity, conductivity at 300 K, and activation energy extracted from calculations,
in comparison with available experiment results

Our calculation Experiment


Halogen doping Channel Channel D300K
Structures concentration (%) size (Å) volume (Å3) (cm2 s1) s300K (mS cm1) Ea (meV) s303K (mS cm1) Ea (meV)
7
t-Na48P16S64 — 2.11 2739 9.46  10 0.01 366.1 0.05 317
t-Na47P16S63F 1.56 2.22 2748 6.15  105 0.64 260.9 — —
t-Na47P16S63Cl 1.56 2.21 2752 1.04  104 1.07 248.7 1.14 249
t-Na47P16S63Br 1.56 2.22 2757 2.32  104 2.39 215.6 — —
t-Na47P16S63I 1.56 2.21 2767 3.53  105 0.36 287.3 — —

20528 | Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 This journal is © the Owner Societies 2018
View Article Online

PCCP Paper

The migration energy of the Na ion in the vacancy-containing the Na ion hopping possibility along path 2, as the energy barrier
t-Na3PS4 structure was estimated by calculating the diffusion along path 2 is low and only 50 meV larger than that along path
path and diffusion barrier of Na–Na hopping by using the 1. So it is reasonable to observe the 3D cubic diffusion network
climbing nudged elastic band (CI-NEB) method. By checking of Na ions in Fig. 2(b). Table 3 shows that our calculation results
the most stable Na site in the 2  2  2 t-Na3PS4 supercell with for Na3PS4 are in accordance with Yu et al.’s work.29
one Na vacancy, it is found that Na ion diffusion between Na I The binding energy of a defect pair can be calculated using
sites is blocked by P atoms, while Na ions at the Na II site could the separated defect approach50
diffuse to the Na I site through the ab-axis plane or to the next P
Ebinding = Eisolate defects  Edefect pair (5)
Na II site along the c axis. Therefore, we calculated the diffusion
pathway of the Na ion from the Na II site to the Na I site (path 1) where Eisolate defects and Edefect pair are the formation energy of
and from the Na II site to the Na II site (path 2). The illustration separated defects, and the formation energy of the defect pair
of the two diffusion pathways is shown in Fig. 3. Path 1 in the studied system, respectively. In this calculation, the
represents the routine of Na I site hopping to the Na II site, binding energy of the XS VNa 0
defect pair can be specifically
corresponding to Na ion diffusion along the a- or b-axis, while expressed by
path 2 represents the routine of Na II site hopping to the next
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

Na II site, corresponding to Na ion diffusion along the c-axis. Ebinding = E(Na47P16S64, +|e|) + E(Na48P16S63X, |e|)
Apart from path 1 and path 2, no other pathway is observed as  E(Na47P16S63X)  E(Na48P16S64), (6)
the body centered cubic rigid PS43 frame is found to block the
hopping of Na ions through a diagonal like pathway or other where E(Na47P16S64, + |e|), E(Na48P16S63,  |e|), E(Na47P16S63)
distorted pathways. In each diffusion path, 6 transition images and E(Na48P16S64) represent the energy of the 2  2  2 t-Na3PS4
were chosen for the NEB calculations and the charged state of supercell with one V0Na , the energy of the 2  2  2 t-Na3PS4
the system was taken into consideration. Depicted in Fig. 3, the supercell with one XS , the energy of the 2  2  2 t-Na3PS4
Na ion diffusion energy barrier along the Na II site to the Na I supercell with one XS VNa 0
defect pair and the energy of the
site is more than 50 meV lower than that between Na II sites. pristine 2  2  2 t-Na3PS4 supercell. For all calculations,
To demonstrate the preferred pathway, the partial mean square we chose the optimized structure with the lowest energy and
displacements (PMSD) of Na ions along the a-axis or b-axis assumed the same chemical potential of the electron reservoir
(path 1) and c-axis (path 2) are plotted in Fig. S4 (ESI†). It is when the individual charged defects exist.50
found that the diffusion rate of Na ions along the a-axis or the Table 4 shows the binding energy Ebinding of the XS VNa 0

b-axis is larger than that along the c-axis, which means the defect pair, and migration energy Emigration and fitted activation
diffusion preferentially occurs along path 1 (Na II - Na I - energy Eactivation of Na ions in different halogen doped t-Na3PS4.
Na II), which is consistent with the lower barrier from the NEB It reveals that the activation energy of Na ion diffusion in
calculation in Fig. 3. This means the diffusion path of Na ions halogen doped t-Na3PS4 obeys eqn (3), roughly comprising
between the Na I site and the Na II site along the a-axis and b-axis the binding energy of the XS VNa 0
defect pair and the migration
directions are dominant in t-Na3PS4. At elevated temperatures energy of Na ions in the pristine structure. As for the binding
such as 1000 K, the thermodynamic contribution would promote energy of Na vacancies to the halogen atoms compared with
that to sulfur, the calculated binding energies of Na vacancies
and halogen atoms for the F, Cl, Br and I doped cases,
respectively, are 215.1, 184.3, 163.8 and 246.2 meV larger than
that of Na vacancies and sulfur. That means Na vacancies prefer
to site at the equilibrium position nearest to the halogen
substitution initially. This is demonstrated by the Na vacancy
and halogen substitution sites with the nearest XS VNa 0
defect
pair in t-Na47P16S63X by structure optimization. Furthermore,
the energies of the structures with the second nearest Na
vacancy to the halogen atom are calculated 206.6, 173.4,
152.1 and 226.3 meV higher than those of the structures with
the nearest Na vacancy to the halogen atom, for the F, Cl, Br
and I doped cases, respectively. This relatively high energy
difference means a strong binding interaction between the
nearest Na vacancy and the halogen atom, which means the
Fig. 3 Diffusion energy landscape of Na ions in the vacancy containing
t-Na3PS4 supercell along the a- or b-axis direction and along the c-axis
direction. The inset shows diffusion pathways of the Na ions in vacancy Table 3 Migration energy of Na ion diffusion in t-Na47P16S63
containing Na47P16S64. The orange ribbons represent the migration path 1
Diffusion path Path 1 (Na II to Na I) Path 2 (Na II to Na II)
(Na II - Na I - Na II), which is corresponding to Na ion diffusion along the
a-axis or the b-axis. The light green ribbons represent the migration path 2 This work B68 meV 118.8 meV
(Na II - Na II), which is corresponding to Na ion diffusion along the c-axis. Yu et al.29 70 (76) meV 123 meV

This journal is © the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 | 20529
View Article Online

Paper PCCP

Table 4 Binding energy of the defect pairs, migration energy, calculated Our results demonstrate the Na vacancy would circularly escape
activation energy according to eqn (4) and simulated activation energy from the Br substitution and get trapped by durative XS VNa 0
from AIMD calculations
dissociation and association during the whole MD simulation at
Binding Migration Activation Activation energy elevated temperatures, which implies the feasibility of eqn (4).
energy energy energy from MD simulation To explore the influence of halogen atoms on Na ion
Structure (meV) (meV) (meV) (meV)
diffusion in halogen doped t-Na3PS4 structures, we constructed
t-Na47P16S63F 221.6 B68 B289.6 260.9 the charge difference plot by using Bader charge analysis codes.51
t-Na47P16S63Cl 190.7 B68 B258.7 248.7
t-Na47P16S63Br 172.7 B68 B240.7 215.6 The atomic charge difference is calculated by the following:
t-Na47P16S63I 244.1 B68 B312.1 287.3
Dr = rtotal  ratom, (7)

dissociation energy takes an important role in the diffusion of where rtotal and ratom are the total charge density of the system
the Na vacancy. In this case, successful Na vacancy hopping and the superposition of atomic charge densities, respectively.
is achieved by Na vacancy dissociation from the nearest Fig. 4 displays the 3D charge difference of Br-doped t-Na3PS4
halogen atoms and its migration through the PS43 skeleton, and the 2D charge difference slice along the plane passing
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

as described by eqn (4) in the manuscript. However, among the through the Na vacancy, the Br substitution and the nearest
halogen doped structures, the BrS VNa 0
defect pair has the P atom, showing obvious charge transfer between P atoms and
lowest binding energy of 172.7 meV, resulting in the lowest S atoms or Br atoms and little charge difference of these Na
activation energy of Na ions in all halogen doped structures, ions. The 2D charge difference slice particularly indicates a
as shown in our DFT calculations. large number of electrons accumulates between the P–S or P–Br
Considering the feasibility of eqn (4), with the existence of bonds, together maintaining the covalent states of the PS43 or
both Na ion vacancies and halogen dopant substitutions inside PS3Br3 tetrahedrons. It shows that the electron accumulation
t-Na3PS4, it is worth for us to check the dynamic statistic of the between P and Br is smaller than that between P and S in the
Na vacancy distribution. As the total activation energy of Na ion remote positions, indicating weaker interaction of P–Br bonds.
diffusion is ideally divided into two parts, described in eqn (4), Meanwhile, at lower isosurface values (i.e., 0.002 e Bohr3),
i.e., the dissociation energy that the Na vacancy need to escape obvious charge deficiency around Na ions was observed,
from the nearby halogen substitution and the migration energy meaning electron transfer from Na atoms to the PS43 skeleton,
of the Na ion in vacancy containing t-Na3PS4, it is necessary to as also demonstrated by the remaining less electron distribution
track the Na vacancy to see whether it keeps undergoing
association and dissociation processes with the halogen atom
during the whole simulation. Through checking the energy of
different XS VNa 0
orderings, it is found the energy differences
are quite small (o2 meV per atom) in all the halogen doped
cases, providing the possibility of Na vacancy hopping inside
halogen doped t-Na3PS4. This is demonstrated by the even
Na ion probability distribution network shown in Fig. 2(b).
To check whether the Na vacancy becomes re-trapped by the
halides during the simulation, we set 48 equal sized 3.5 Å cubic
regions to take the likelihood of Na vacancies in the
Na47P16S63Br structure and count the distribution of Na vacan-
cies at the central cubic region near the Br atom (B3.1 Å), the
distribution of Na vacancies at the remote cubic region remote
from the Br atom (B9.1 Å), and the distribution of Na vacancies
at the other 46 cubic regions in the AIMD simulation. The
distribution is averaged for each cubic region as the definition
of the frequency ratio of observed Na vacancies inside the cubic
region to the number of simulation steps (2000 steps). Shown
in Table S1 (ESI†), it is found the Na vacancy sites near the
halogen atom have a much larger possibility compared to those
remote from the Br atom at 600 K, while nearly the same
possibilities of Na vacancies near the Br atom and remote from Fig. 4 (a) 3D charge difference plot of the t-Na47P16S63Br structure along
the Br atom are observed at elevated temperatures, which could the (1, 1, 0) view direction. The yellow region indicates electron gain, and
the blue indicates electron loss. The isovalue is 0.01 e Bohr3. The silver
be ascribed to the higher thermal energy increasing the hopping
balls represent Na atoms, orange ones represent S atoms, brown ones
possibility of Na vacancies to overcome the large dissociation represent Br atoms, and purple ones represent P atoms. (b) The corres-
energy from Br substitution, showing even distributions of ponding 2D charge difference slice along the plane passing through the Na
Na vacancies in other cubic regions at elevated temperatures. vacancy, the Br substitution and the nearest P atom.

20530 | Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 This journal is © the Owner Societies 2018
View Article Online

PCCP Paper

Table 5 The electron distribution of the PS3X3 tetrahedrons and corresponding bond lengths of P–X (X represents different halogen dopants) and P–S.
The term ‘‘neighbor’’ means the nearby positions (o3.5 Å) while ‘‘remote’’ means the relatively separated positions (46.5 Å)

Electron distribution (electrons) Bond length (Å)


X Neighbor P Remote P Neighbor S Remote S Neighbor Na Remote Na P–X P–S
F-case 7.816 3.355 3.815 6.817 6.916 0.175 0.169 1.66 2.02
Cl-case 7.564 3.696 3.811 6.809 6.915 0.177 0.169 2.23 2.01
Br-case 7.476 3.790 3.808 6.814 6.915 0.179 0.169 2.42 2.01
I-case 7.316 3.946 3.806 6.810 6.918 0.182 0.169 2.61 2.02

around Na ions in Table 5. Herein, the electron numbers in all Chu et al.31 demonstrated the successful synthesis of a stable
studied cases are calculated by integrating the electron density in Cl-doped Na3PS4 superionic conductor. Due to the large similarity
each Bader volume defined as the boundary of the surface through among these halogen doped t-Na3PS4 structures, it is expected that
which the charge density gradient has a zero flux. The large other halogen doped samples can be synthesized under similar
electron loss indicates the ionic state of these Na atoms inside synthesis conditions.
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

the structures. Furthermore, we investigated the Na ionic conduction with a


Through local charge transfer analysis for different halogen higher concentration of halogen atom substitution. By replacing
doped cases, shown in Table 5, the charge transfer for F, Cl, Br two of the 48 S atoms with Br atoms while introducing two Na ion
and I atoms becomes smaller and smaller, in accordance with vacancies inside t-Na46P16S62Br2, it introduces 4.16% Na ion
the variation of electronegativities of these halogen dopants. vacancies via 3.12% Br doping in the t-Na3PS4 structure. Based
A relatively smaller electronegativity property of the dopants on the energy minimization criteria, it is found that two
may benefit Na ion diffusion because of weaker Na–halogen BrS VNa0
defect pairs prefer to keep away from each other in
interaction. However, the larger bond length between P atoms the supercell. When the concentration of defect pairs doubles
and heavier halogen atoms may cause local tetrahedron exten- (i.e., t-Na46P16S62Br2), the decomposition energy is calculated to
sion and lattice distortion and restrict Na ion fast transport. be 0.010 eV per atom, which is still low and indicates
We conjecture these two factors would affect Na ion diffusion t-Na46P16S62Br2 could be synthesized at elevated temperatures.
in a contrary way, eventually rendering Br substitution most We also conducted ab initio MD simulations for t-Na46P16S62Br2
beneficial for Na ion transport with minor charge difference at different temperatures, and the results of Na ionic conduc-
of adjacent P atoms compared with the remote ones. We also tivities are given in Fig. 5. Compared with the result of 1.56%
plot the electron density of states for different halogen doped Br-doped t-Na3PS4, the 3.12% Br-doped case depicts a
t-Na3PS4, shown in Fig. S2 (ESI†), all of which have band gaps higher Na ionic conductivity, with a value as considerable as
larger than 2 eV as the merit of good solid electrolytes in Na ion 5.42 mS cm1 at room temperature. The certain increase of Na
batteries. ionic conductivity is attributed to the larger change in Na ion
To examine the phase stability of halogen doped t-Na3PS4, hopping provided by more Na ion vacancies and hence
we constructed the corresponding compositional phase diagrams the enhancement of the pre-exponential factor s0 in eqn (3).
at 0 K (see Fig. S3 (ESI†) for each halogen doped case). Apparently, Meanwhile, we noted that when the Na vacancy concentration
the phase diagrams for these halogen doped structures are very doubles with higher Br doping, the calculated Na ionic
similar. At 0 K, the Gibbs energy variations during the formation of conductivity increases more than 100% from 2.37 mS cm1 to
halogen doped t-Na3PS4 (E = 0.002 eV per atom for the F doped
structure, and E = 0.005 eV per atom for the Cl, Br and I doped
structures) are slightly positive, i.e., the t-Na47P16S63X (X represents
F, Cl, Br or I) material tends to decompose into P2S5, Na2S, and
NaX with the following reaction:

Na47P16S63X - P2S5 + Na2S + NaX.


This small energy increase will be compensated for at elevated
temperature due to entropy effects, leading to stable halogen
doped structures. Actually many experimentally synthesized
sulphide solid-state electrolytes are predicted to be unstable
at 0 K by DFT calculations. For example, the fast conductors
Li10GeP2S12 and Na10SnP2S12 have decomposition energies
(energies above hull) of 0.025 eV per atom52 and 0.007 eV per
atom,53 respectively. However, considering the product of the
Boltzmann constant kT as a value of 25.7 meV per atom,
at room temperatures, entropy contributions may reduce the Fig. 5 Arrhenius plot for t-Na47P16S63Br and t-Na46P16S62Br2 from AIMD
Gibbs energy and stabilize these meta-stable solid-state electrolytes. simulations.

This journal is © the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 | 20531
View Article Online

Paper PCCP

5.42 mS cm1. This nonlinear increase may be ascribed to more 6 K. Takada, Acta Mater., 2013, 61, 759–770.
concerted and correlative diffusion events participating in the 7 P. Knauth, Solid State Ionics, 2009, 180, 911–916.
Na ion diffusion between these two XS VNa 0
defect pairs. The 8 Z. Lu and F. Ciucci, Chem. Mater., 2017, 29, 9308–9319.
extracted activation energy for diffusion is 204.9 meV, close to 9 N. Anantharamulu, K. K. Rao, G. Rambabu, B. V. Kumar,
that of 1.56% Br-doped t-Na3PS4. The similar value of the V. Radha and M. Vithal, J. Mater. Sci., 2011, 46, 2821–2837.
activation energy at two different Br doping concentrations 10 B. L. Ellis and L. F. Nazar, Curr. Opin. Solid State Mater. Sci.,
indicates that Na ions would keep their diffusion pathway at 2012, 16, 168–177.
the dilute Na ion vacancy concentration, which demonstrates 11 A. Wang, S. Kadam, H. Li, S. Shi and Y. Qi, npj Comput.
the practicability of eqn (4). Mater., 2018, 4, 15.
12 Z. Lu and F. Ciucci, J. Mater. Chem. A, 2016, 4, 17740–17748.
13 Z. Lu, C. Chen, Z. M. Baiyee, X. Chen, C. Niu and F. Ciucci,
Conclusions Phys. Chem. Chem. Phys., 2015, 17, 32547–32555.
We calculated the Na ionic conductivity in pristine and halogen 14 A. Ponrouch, E. Marchante, M. Courty, J.-M. Tarascon and
(F, Cl, Br, and I) doped tetragonal-Na3PS4 superionic conductors M. R. Palacı́n, Energy Environ. Sci., 2012, 5, 8572–8583.
through ab initio molecular dynamic simulations. We demon- 15 X. Lu, G. Xia, J. P. Lemmon and Z. Yang, J. Power Sources,
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

strated that the Na ionic conductivity of pristine t-Na3PS4 is only 2010, 195, 2431–2442.
about 0.01 mS cm1, while a much higher Na ionic conductivity 16 K. B. Hueso, M. Armand and T. Rojo, Energy Environ. Sci.,
could be achieved with the introduction of Na ion vacancies by 2013, 6, 734–749.
a halogen doping strategy. A Na ionic conductivity as high as 17 H.-P. Hong, Mater. Res. Bull., 1976, 11, 173–182.
2.37 mS cm1 is predicted for Br-doped t-Na3PS4. It was found 18 A. Hooper, J. Phys. D: Appl. Phys., 1977, 10, 1487.
that the introduction of halogen substitution and Na ion vacan- 19 J. T. Kummer, Prog. Solid State Chem., 1972, 7, 141–175.
cies has little effect on the overall structure of pristine t-Na3PS4, 20 T. J. Udovic, M. Matsuo, A. Unemoto, N. Verdal, V. Stavila,
which shows a 3D Na ion diffusion path. The lowest defect A. V. Skripov, J. J. Rush, H. Takamura and S.-I. Orimo, Chem.
binding energy is in the Br-doped t-Na3PS4 structure because of Commun., 2014, 50, 3750–3752.
the relatively weak bond strength of Na–S or Na–Br bonds, which 21 M. Tatsumisago and A. Hayashi, Int. J. Appl. Glass Sci., 2014,
leads to the lowest activation energy of Na ions and the highest Na 5, 226–235.
ionic diffusivity/conductivity in all halogen doped cases. A further 22 I.-H. Chu, H. Nguyen, S. Hy, Y.-C. Lin, Z. Wang, Z. Xu,
higher Na ionic conductivity and lower activation energy could Z. Deng, Y. S. Meng and S. P. Ong, ACS Appl. Mater.
be reached by introducing higher halogen atom substitution Interfaces, 2016, 8, 7843–7853.
concentrations. The AIMD calculations suggest Br-doped t-Na3PS4 23 N. Tanibata, K. Noi, A. Hayashi and M. Tatsumisago, RSC
as a very promising Na-ion solid-state superionic conductor near Adv., 2014, 4, 17120–17123.
ambient temperature. 24 Y. Wang, W. D. Richards, S. P. Ong, L. J. Miara, J. C. Kim,
Y. Mo and G. Ceder, Nat. Mater., 2015, 14, 1026–1031.
25 N. Tanibata, K. Noi, A. Hayashi, N. Kitamura, Y. Idemoto
Conflicts of interest and M. Tatsumisago, ChemElectroChem, 2014, 1, 1130–1132.
26 A. Hayashi, K. Noi, A. Sakuda and M. Tatsumisago, Nat.
There are no conflicts to declare.
Commun., 2012, 3, 856.
27 Z. Zhu, I.-H. Chu, Z. Deng and S. P. Ong, Chem. Mater., 2015,
Acknowledgements 27, 8318–8325.
28 N. J. J. de Klerk and M. Wagemaker, Chem. Mater., 2016, 28,
Mr He Huang acknowledges the financial supports from the 3122–3130.
Joint School of XJTU and HKUST. The project is supported by 29 Z. Yu, S. L. Shang, J. H. Seo, D. Wang, X. Luo, Q. Huang,
the 111 project (no. D16002) from the State Administration of S. Chen, J. Lu, X. Li and Z. K. Liu, Adv. Mater., 2017,
Foreign Experts Affairs and the Ministry of Education, PRC. 29, 1605561.
30 L. Zhang, K. Yang, J. Mi, L. Lu, L. Zhao, L. Wang, Y. Li and
References H. Zeng, Adv. Energy Mater., 2015, 5, 1501294.
31 I.-H. Chu, C. S. Kompella, H. Nguyen, Z. Zhu, S. Hy, Z. Deng,
1 Z. Deng, Z. Zhu, I.-H. Chu and S. P. Ong, Chem. Mater., 2016, Y. S. Meng and S. P. Ong, Sci. Rep., 2016, 6.
29, 281–288. 32 M. A. Kraft, S. P. Culver, M. Calderon, F. Böcher, T. Krauskopf,
2 D. Kundu, E. Talaie, V. Duffort and L. F. Nazar, Angew. A. Senyshyn, C. Dietrich, A. Zevalkink, J. r. Janek and
Chem., Int. Ed., 2015, 54, 3431–3448. W. G. Zeier, J. Am. Chem. Soc., 2017, 139, 10909–10918.
3 A. Manthiram and X. Yu, Small, 2015, 11, 2108–2114. 33 A. Zevgolis, B. C. Wood, Z. Mehmedović, A. T. Hall,
4 R. C. Massé, E. Uchaker and G. Cao, Sci. China Mater., 2015, T. C. Alves and N. Adelstein, APL Mater., 2018, 6, 047903.
58, 715–766. 34 J. Zhu, Y. Wang, S. Li, J. W. Howard, J. r. Neuefeind, Y. Ren,
5 N. Yabuuchi, K. Kubota, M. Dahbi and S. Komaba, Chem. H. Wang, C. Liang, W. Yang and R. Zou, Inorg. Chem., 2016,
Rev., 2014, 114, 11636–11682. 55, 5993–5998.

20532 | Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 This journal is © the Owner Societies 2018
View Article Online

PCCP Paper

35 Y. Zhao and L. L. Daemen, J. Am. Chem. Soc., 2012, 134, 45 S. Arrhenius, Z. Phys. Chem., 1889, 4, 96–116.
15042–15047. 46 T. F. Willems, C. H. Rycroft, M. Kazi, J. C. Meza and
36 Z. Deng, B. Radhakrishnan and S. P. Ong, Chem. Mater., M. Haranczyk, Microporous Mesoporous Mater., 2012, 149,
2015, 27, 3749–3755. 134–141.
37 S.-H. Bo, Y. Wang and G. Ceder, J. Mater. Chem. A, 2016, 4, 47 R. L. Martin, B. Smit and M. Haranczyk, J. Chem. Inf. Model.,
9044–9053. 2011, 52, 308–318.
38 G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter 48 V. I. Parvulescu and C. Tiseanu, Catal. Today, 2015, 253,
Mater. Phys., 1996, 54, 11169. 33–39.
39 P. E. Blöchl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 49 P.-A. Primus, A. Menski, M. P. Yeste, M. A. Cauqui and
50, 17953. M. U. Kumke, J. Phys. Chem. C, 2015, 119, 10682–10692.
40 M. E. Tuckerman, P. J. Ungar, T. Von Rosenvinge and 50 M. Nakayama and M. Martin, Phys. Chem. Chem. Phys., 2009,
M. L. Klein, J. Phys. Chem., 1996, 100, 12878–12887. 11, 3241–3249.
41 G. Henkelman, B. P. Uberuaga and H. Jónsson, J. Chem. 51 W. Tang, E. Sanville and G. Henkelman, J. Phys.: Condens.
Phys., 2000, 113, 9901–9904. Matter, 2009, 21, 084204.
42 G. Henkelman and H. Jónsson, J. Chem. Phys., 2000, 113, 52 Y. Mo, S. P. Ong and G. Ceder, Chem. Mater., 2011, 24,
Published on 18 July 2018. Downloaded on 10/31/2023 10:13:12 AM.

9978–9985. 15–17.
43 A. Van der Ven, G. Ceder, M. Asta and P. D. Tepesch, Phys. 53 W. D. Richards, T. Tsujimura, L. J. Miara, Y. Wang, J. C.
Rev. B: Condens. Matter Mater. Phys., 2001, 64, 184307. Kim, S. P. Ong, I. Uechi, N. Suzuki and G. Ceder, Nat.
44 R. J. Friauf, J. Appl. Phys., 1962, 33, 494–505. Commun., 2016, 7, 11009.

This journal is © the Owner Societies 2018 Phys. Chem. Chem. Phys., 2018, 20, 20525--20533 | 20533

You might also like