X-ray Pair Distribution Function Analysis, Electrical and Electrochemical Properties of Cerium Doped Li 5 La 3 Nb 2 O 12 Garnet Solid-State Electrolyte

Garnet solid state electrolytes have been considered as potential candidates to enable next generation all solid state batteries (ASSBs). To facilitate the practical application of ASSBs, a high room temperature ionic conductivity and a low interfacial resistance between solid state electrolyte and electrodes are essential. In this work, we report a study of cerium doped Li 5 La 3 Nb 2 O 12 through X-ray pair distribution function analysis, impedance spectroscopy and electrochemical testing. The successful cerium incorporation was confirmed by both X-ray diffraction refinement and X-ray pair distribution function analysis, showing the formation of an extensive solid solution. The local bond distances for Ce and Nb on the octahedral site were determined using X-ray pair distribution function analysis, illustrating the longer bond distances around Ce. This Ce doping strategy was shown to give a significant enhancement in conductivity (1.4 x 10 −4 S cm −1 for Li 5.75 La 3 Nb 1.25 Ce 0.75 O 12 , which represents one of the highest conductivities for a garnet with less than 6 Li) as well as a dramatically deceased in interfacial resistance (488 cm for Li 5.75 La 3 Nb 1.25 Ce 0.75 O 12 ). In order to demonstrate the potential of this doped system for use in ASSBs, the long term cycling of a Li//garnet//Li symmetric cell over 380 h has been demonstrated.


Introduction
State-of-the-art lithium ion batteries (LIBs) have dominated the energy storage market for more than 2 decades because of their long cycle life and high energy and power density over other battery systems [1][2][3] . Nowadays, the increasing demand in developing electrical vehicles (EVs) not only need higher energy density, but more importantly, require improved safety properties 4 . The flammable organic solvent in LIBs can trigger safety issues, and so makes it a non-ideal system for EVs. Instead, all-solid-state batteries (ASSBs) using a non-flammable inorganic solid-state electrolyte are considered to be a promising candidate to address the safety issues for future EVs 5 .
Garnet lithium ion conductors have attracted increasing interest in the last decade due to their relatively good electrochemical stability against Li metal/cathode materials, and high lithium ionic conductivity at room temperature compared to other solid state electrolyte materials 6-8 . The ideal garnet framework has the chemical formula A 3 B 2 C 3 O 12 , where A, B and C ions are located at eight, six and four oxygen coordinated sites, respectively. La 3 M 2 Li 5 O 12 (M = Nb, Ta), was the first fast Li + ion conducting garnet reported by Thangadurai et al. in 2003 9 . Detailed studies have revealed that the lithium content and the distribution of lithium in the structure are key to the diffusion pathway and resulting lithium ion conductivity in garnet materials 10, 11 . Li + ions occupy both 24d tetragonal and 96h/48g octahedral Wyckoff sites in cubic Li 5 La 3 Nb 2 O 12 , with lithium ions shifting from 24d tetragonal sites to 96h/48g octahedral sites as a function of increasing lithium content to reduce the lithiumlithium interaction strain 6, 7 . The maximum lithium content in the structure was found to be 7, leading to compositions such as Li 7 La 3 Zr 2 O 12 , which adopts tetragonal symmetry with lithium ions ordering in three fully occupied tetrahedral 8a and octahedral 16f and 32g sites; this Li ordering reduces its ionic conductivity 12-14 . The highly conductive cubic phase can be stabilised through the creation of lithium vacancies, and the optimum conductivity is found for lithium contents of 6.4 -6.6 per garnet formula unit [15][16][17][18][19][20][21][22][23][24][25] .
In addition, to facilitate the practical application of ASSBs, issues such as high interfacial resistances between electrode and electrolyte, and the lithium dendrite penetration problem within solid state electrolytes have attracted much attention in recent years [26][27][28][29] . The interfacial impedance between the garnet and electrode mainly comes from the poor contact in association with microscopic voids and grain boundaries of garnet, as well as an insulating Li 2 CO 3 surface layer formed in air initiated by the proton/lithium exchange at the surface 30-33 .
The ionic transport was reported to be limited by the grain boundaries, which are heavily affected by the segregation of dopants and Li 34 . Mechanical polishing, carbon annealing or acid treatment were reported to remove this unwanted surface layer 35,36 . Besides, metal or non-metal coating on garnet pellets and consequently lithium alloying effects have also been used to minimise the interfacial resistance between lithium metal and garnet, and to reduce the dendrite growth [37][38][39][40] . Despite the potential for short-circuiting from electronic conduction, a mixed ion/electron-conductive interface was demonstrated to beneficial for Li dendrite suppression 41, 42 . Another common strategy in this field is to utilise polymer -ceramic composite electrolytes, which combine good wetting properties of polymer electrolyte with the high ionic conductivity of ceramic electrolytes, forming a uniform interfacial contact with decreased interfacial resistance 43-45 .
Ce 3+ or Ce 4+ incorporation on La 3+ site in Li 7 La 3 Zr(Hf) 2 O 12 has been previously studied albeit showing limited solid solution range for this site substitution, before the detection of impurity phases 46,47 . In contrast, we showed in a previous study, that there was a greater degree of Ce 4+ substitution possible on the Zr 4+ site in Li 7 La 3 Zr 2 O 12 leading to a reduction in the tetragonal distortion, which consequently increased the ionic conductivity, although due to the stoichiometric Li content in this system, the room temperature conductivity was just below the value of 10 -4 Scm -1 , which is considered the minimum for applications 48 . In this paper, the possibility to replace Ce 4+ on the Nb 5+ site in Li 5 La 3 Nb 2 O 12 with the creation of excess Li + as the charge compensation mechanism has been examined for the first time, aiming to enhance the overall conductivity for these Ce doped garnets, in addition to providing the beneficial decrease in the interfacial impedance previously reported for Ce-doped prepared from intimately ground stoichiometric amounts of starting reagents which were heated initially to 650 °C for 12 hours at a rate of 5 °C min −1 . 10 -15% excess Li 2 CO 3 was then added to the precursor and the powder milled for 30 minutes using a Pulverisette 5 planetary ball mill. The mixture was pressed into a pellet and heated to 950 -1000 °C for 12 hours at a rate of 5 °C min −1 to form the final product.

Characterisation
A Bruker D8 X-ray diffractometer (XRD) with a CuKα radiation and linear position sensitive detector was used to collect X-ray diffraction data. Patterns were recorded over the 2θ range 15° to 80° with a 0.02° step size. Structural refinement was carried out with the GSAS suite of Rietveld refinement software using the XRD data 49 .
Scanning electron microscopy (SEM, HITACHI TM4000plus) was employed to assess the microstructure. Bulk samples were polished and thermally etched at 90% of the sintering temperature for 0.5 h. The distribution of elements was probed with an energy dispersive X-ray spectroscopy (EDX) detector.
Pellets (9.8 mm diameter) were pressed and sintered at 1000 -1050 °C for 4 hours (ramp rate of 5 o Cmin −1 ) in a dry room to limit H + /Li + exchange and prevent the decomposition of samples for impedance measurements. Mother powder was used to cover the pellets to prevent the Li loss and reduce reaction with the Al 2 O 3 crucible. Au paste was painted on both sides of pellet and heated at 850 °C for 1 hour in air. Impedance data were collected with a HP 4192 analyser over the frequency range from 1 to 10 7 Hz with 100 mV ac amplitude.

Cell assembly and electrochemical test
For the cell tests, a Li 5.75 La 3 Zr 1.25 Ce 0.75 O 12 pellet (2 mm thickness) was sintered at 1000 °C for 12h in a dry room. The Li//Li 5.75 La 3 Zr 1.25 Ce 0.75 O 12 //Li symmetric cell was hot pressed at 175 °C for 1 hour and assembled using a Swagelok cell in an Argon filled glove box. Electrochemical impedance spectroscopy (EIS) was performed over the frequency range from 0.1 to 10 7 Hz (100 mV ac applied potential) with a Solatron 1260 analyser. The Li plating/stripping performance was evaluated using a Bio-logic SP50 cell tester.

Pair Distribution Function data collection
Total scattering data were collected at the I15-1 XPDF beamline at Diamond Light Source, UK. Powdered samples were loaded into borosilicate capillaries (1.5 mm OD, 1.17 mm ID) and spun perpendicular to the beam during data collection to improve powder averaging. Scattering data were collected at an X-ray energy of 76.69 keV using a Perkin Elmer XRD 4343 CT area detector placed ~200 mm from the sample. The 2-D data were corrected for polarization and flat-field, then integrated to 1-D using the DAWN package prior to processed the scattering range 0.7 Å −1 ≤ Q ≤ 25 Å −1 into PDFs using the GudrunX package 50,51 . A modified Lorch Function (∆ 1 = 0.05 Å) was applied to suppress spurious low-r features 52 .

Results and discussion
Phase formation Stoichiometric Li 5 La 3 Nb 2 O 12 (LLNO) can be indexed with an Ia−3d cubic cell. In Kroger -Vink notation, the relationship for substitution of Nb 5+ by Ce 4+ is as follows: As shown in figure 1a, XRD patterns of Ce-doped Li 5 La 3 Nb 2 O 12 with formulae Li 5+x La 3 Nb 2−x Ce x O 12 show peak shifts to lower 2θ angle with increasing Ce content, without the detection of any impurity phases up to x = 0.75. The cell parameter was shown to obey Vegard's law, increasing linearly as a function of x (figure 1b), which further confirms the successful incorporation of Ce into the structure due to the larger radius of Ce 4+ to that of Nb 5+ (0.87 and 0.64 ionic radius for Ce 4+ and Nb 5+ respectively in 6 oxygen coordinated octahedral sites).   figure 4. The peaks in the low-r region of the PDF were assigned based on simulations of the partial PDFs of the undoped structure using the computer program PDFgui (figure S1) 53 . The   To fit the circuit, a resistor R1 and a constant phase element (CPE) CPE1 in parallel were used to simulate the high frequency semicircle. Another CPE2 which is in series with the R1/CPE1 in the circuit was utilised to interpret the low frequency spike.
The corresponding spectroscopic C' plot (figure 7b) shows a higher frequency plateau with a capacitance of 8.2 pF cm −1 with an associated permittivity of 91 (calculated from ε ∞ ' = C/ε 0 , where ε 0 is the permittivity of free space with a value of 8.854 x 10 -14 F cm −1 ), which is a typical value for the bulk response 56 . In addition, a low frequency plateau with a capacitance of 3 μF cm −1 was observed, which is due to the double layer effect at the sample -electrode interface; hence consistent with Li + ion conduction.