Nitrogen-Doped Lithium Lanthanum Titanate Nanofiber-Polymer Composite Electrolytes for All-Solid-State Lithium Batteries

Currently cation doping is common for improving ionic conductivity of metal oxide-based lithium-ion conductors. In this work, anions (nitrogen ions) have been doped to perovskite Li 3x La 2/3 − x TiO 3 (LLTO) nano ﬁ bers by heat treatment in the ammonia-containing atmosphere, and substituted for oxygen anions partially in the perovskite structure. Density-functional theory (DFT) calculation results reveal that nitrogen doping weakens the bonding of Li ions on the A sites in perovskite ABO 3 structure and allows for larger lattice distortion, reducing the activation energy for Li-ion hopping in both the forward and backward jumping directions. Experimental results have also con ﬁ rmed that nitrogen doping has improved ionic conductivity of LLTO. Nitrogen-doped LLTO nano ﬁ bers have been incorporated with a poly (vinylidene ﬂ uoride)-co-hexa ﬂ uoropropylene (PVDF-HFP) polymer to form a solid-state composite electrolyte, which exhibits ionic conductivity of 3.8 × 10 − 4 S·cm − 1 at room temperature and an electrochemical stability window of up to 4.9 V vs Li ∣ Li + . The all-solid-state Li metal ∣ composite electrolyte ∣ LiFePO 4 lithium batteries, which employ nitrogen-doped LLTO nano ﬁ bers, show better rate capability and cycling stability at room temperature than the counterparts with pristine LLTO nano ﬁ bers.

Lithium-ion batteries are the most popular energy storage devices for portable electronics and electric vehicles. 1 Highly volatile and flammable organic solvent-based liquid electrolytes in commercial lithium-ion batteries have raised serious safety issues. 2 Solid-state electrolytes have better chemical stability and wider electrochemical windows, and are considered as a promising alternative to liquid electrolytes. 3,4 Perovskite-type lithium lanthanum titanates (LLTO) is one of the most attractive solid-state electrolytes due to its high bulk ionic conductivity of ∼10 −3 S·cm −1 at room temperature, good chemical and structural stability in a humid atmosphere, and high electrochemical stability (>8 V). [5][6][7][8] However, the grain-boundary conductivity of LLTO is low (10 −7 ∼ 10 −6 S·cm −1 ), hindering its applications in all-solid-state batteries. 8 Therefore, it is essential to improve both bulk conductivity and grain boundary conductivity of LLTO electrolytes.
The perovskite structure has a general formula of ABO 3 , where A is a rare earth or alkaline-earth metallic element, and B is a transition metallic element. Lithium ions migrate by hopping through A-site vacancies in the perovskite structure. Previous studies show that the ionic conductivity of perovskite-type oxides depends on the size of the second A-site ion, the A-site vacancy concentration, and the nature of the B-O bond. [9][10][11][12][13][14] Hence, cation doping is a common strategy for improving the ionic conductivity of perovskite-type oxides. A-or B-sites have been substituted with other cations with different ion sizes, chemical valences and electronegativities. For example, substitution of La atoms at A-site with larger Sr ions increases the lithium ionic conductivity, which is ascribed to space expansion to increase bottleneck sizes, thus enhancing ion migration. Also, the B-site has been substituted by tetravalent Sn, Zr, Mn and Ge ions, 15 and the results show that the ionic conductivity increases as a decrease in the radius of the doping ions. In addition, the B-site is doped with a narrow concentration range of Al, which strengthens the B(Al)-O bond and weakens the A (Li, La)-O bond, leading to an improvement in ionic conductivity. 16 In addition to perovskite oxides, oxy-sulfides, 17 oxynitrides, 18 and oxy-fluorides 19 with different crystal structures have also received attention. It has been reported that the partial occupation of oxygen sites in perovskite-type oxides could modulate ionic conductivity due to the charge and iconicity differences in the metalanion bonds. 20,21 Because nitrogen ions have higher charge than oxygen ions, nitrogen doping could induce oxygen vacancies to balance charges. 22,23 Also, the electronegativity difference between nitrogen and oxygen (3.50 for O and 3.07 for N) can result in difference in the covalency of the cation-anion bond. Therefore, it is expected that nitrogen doping will modulate ionic conductivity of LLTO. However, the effects of anion doping into LLTO have rarely been studied.
In this work, nitrogen is doped into the perovskite-type Li 0.33 La 0.557 TiO 3 nanofibers by heat-treatment in the ammoniacontaining atmosphere. The effect of nitrogen doping on the ionic conductivity of LLTO is investigated by experiments and densityfunctional theory (DFT) calculation. Especially, first-principles calculations are performed to examine the effects of N doping on the activation energy barrier of Li-ion hopping. The nitrogen-doped LLTO nanofibers are then incorporated into a poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP) matrix to form a composite electrolyte. The ceramic-polymer composite electrolyte is employed to construct a full cell of battery with a lithium foil as anode and LiFePO 4 (LFP) as cathode. The electrochemical performance of battery is investigated at room temperature. lithium loss during preparation. In a typical experiment, 3.3 mmol of lithium nitrate anhydrous and 5.6 mmol of lanthanum (III) nitrate hexahydrate were firstly dissolved into 25 ml of deionized (DI) water. 10 mmol of titanium (IV) isopropoxide was dissolved in a mixture of 15.27 ml of isopropanol and 5.72 ml of acetic acid. The above two solutions were then mixed, followed with addition of an acetic acid/dimethylformamide sol with 10 wt%. polyvinylpyrrolidone (Mw ∼ 1,300,000), giving a volume ratio of 1:1. During electrospinning, 19.0 kV of voltage was applied between the syringe's needle tip and the fiber collector at a distance of 15 cm. The precursor was fed at a rate of 0.3 ml h −1 . The as-spun precursor nanofibers were then sintered in air at 900°C for 2 h to obtain the LLTO nanofibers. 24,25 Doping nitrogen into LLTO nanofibers.-LLTO nanofibers were loaded and heated at a ramp rate of 5°C min −1 , then held at a constant temperature under a flow of NH 3 (100 sccm) for 2 h. The nitrogen-doped LLTO nanofibers were taken out after the furnace was naturally cooled down to room temperature. The LLTO nanofibers treated at 525°C, 550°C, 575°C and 600°C were denoted as N-LLTO 525, N-LLTO 550, N-LLTO 575, and N-LLTO 600, respectively.
Characterization.-The morphology and structure of the samples were observed under JEOL JSM-7600 field-emission scanning electron microscopy (FE-SEM) at an accelerating voltage of 15 kV. The crystal structure was characterized by FEI Titan 80-300 scanning transmission electron microscopy (STEM). The STEM high angle annular dark field (HAADF) was carried out with a probe forming lens aberration corrector and operated at 300 kV. The phase was characterized by X-ray diffractometer (XRD, PANalytical X'pert Pro). The chemical status of elements in samples was measured with X-ray photoelectron spectroscopy (XPS, Physical Electronics PHI 5000 VersaProbe). The binding energies (BE) of XPS spectra were calibrated with adventitious carbon at a reference of 284.80 eV.
Assessment of electrochemical performance.-The ionic conductivity of the electrolyte was tested using electrochemical impedance spectroscopy (EIS). EIS was performed on the electrolyte membrane sandwiched between two stainless-steel electrodes with a Solartron 1260 impedance analyzer in a frequency range of 1 Hz to 1 MHz. The electrochemical window of the electrolyte was measured using linear sweep voltammetry (LSV) mode at a sweep rate of 10 mV s −1 in a range of 0-6 V, in which the lithium metal foils were employed as the counter and reference electrodes, and a stainless-steel plate was used as the working electrode. Symmetric Li|composite electrolyte|Li cells were prepared in an argon filled glove box using Li foil as the two electrodes and tested at a current density of 0.5 mA·cm -2 at room temperature. The half cells were operated for 0.5 h in each cycle.
Evaluation of battery performance.-Li|PVDF-HFP/LiTFSI/N-LLTO|LFP coin cells were assembled in an argon-filled glove box. The LiFePO 4 cathode was prepared by grinding 70% LiFePO 4 (LFP), 20% carbon black and 10% polyvinylidene fluoride (PVDF) in N-methyl-2-pyrrolidone (NMP) into uniform slurry. 26 The prepared slurry was coated on an aluminum foil and dried in vacuum at 80°C for 12 h. The charge and discharge profiles were tested between 2.5 V to 4.2 V at different current rates at room temperature.
Computation.-First-principles DFT calculation was performed to identify the structures of LLTO, N dopant location, and Li + ion distribution. The LLTO unit cell was kept at the experimentally determined tetragonal structure. The super cell has the elemental composition of Li 6 28 and a plane-wave basis set with a kinetic energy cutoff of 400 eV for the valence electrons, while the core electrons were treated by the projected augmented wave (PAW) method. 27 A 3 × 3 × 3 K-point grid was used for sampling of the reciprocal space. Structure optimization was converged below a force threshold of 0.05 eV/Å. The nudged elastic band (NEB) method 29 was subsequently used to connect different Li + configurations to identify the hopping transition states and to compute their activation energies.

Results and Discussion
Morphology and microstructure of N-LLTO nanofibers.- Figure 1a shows the as-spun precursor nanofibers. After calcinated in air for 2 h, the resulting LLTO nanofibers were then treated under NH 3 atmosphere at different temperatures. Figure 1b reveals the morphology of the N-LLTO 550, which was obtained after heattreatment in an ammonia flow at 550°C. The nanofibers displayed an average diameter of 92 nm (Fig. 1c). The STEM image of N-LLTO 550 displayed a lattice spacing of 0.277 nm (102), as shown in Fig. 1d. The XRD patterns in Fig. 1e have confirmed that all the NH 3 -treated samples retained a tetragonal phase for perovskite-type Li 0.33 La 0.56 TiO 3 (Joint Committee on Powder Diffraction Standards card 54-1238).
As shown in the XPS survey scans in Fig. 2a, weak N 1s signal was observed in all the ammonia-treated samples. According to previous study, the N 1s peak at ∼400 eV is typically ascribed to the interstitial nitrogen dopant while the peak at ∼396 eV, which corresponds to the Ti-N bond, is attributed to the substitutional nitrogen dopant. 30,31 The N 1s core-level XPS spectrum of N-LLTO-550 displayed a peak at 395.6 eV (Fig. 2b), which indicated that the doped nitrogen atoms were directly bonded to the titanium atoms in the TiO 6 octahedron. 32,33 No nitrogen element was detected in the pristine LLTO nanofibers (Fig. 2c). As the heat treatment temperature increased, the atomic concentration of nitrogen dopant increased, resulting in 0.0% (below the detection limit), 0.9%, 1.0% and 1.1% for N-LLTO 525, N-LLTO 550, N-LLTO 575 and N-LLTO 600 nanofibers, respectively. The La 3d profile in Fig. 2d displayed the spin-orbit doublet of La 3d5/2 and La 3d3/2 peaks at 834.0 eV and 850.8 eV, respectively. The two satellite peaks at higher binding energies were characteristic of the oxidized status of La due to monopole excitation caused by a sudden change in screening of the valence electrons upon the removal of a core electron. 33 As shown in Fig. 2e, the doublet in the doublet Ti 2p core-level spectra of pristine LLTO exhibited a symmetric Ti 4+ peak at 458.2 eV. In contrast, the Ti 2p 3/2 peak of the N-doped LLTO nanofibers was asymmetric and can be deconvoluted into two components at 458.2 eV (Ti 4+ ) and 456.8 eV (Ti 3+ ). 34 This indicated slight reduction of Ti ions. The O 1s profile of N-LLTO 550 was slightly shifted to a lower energy compared with that of pristine LLTO (Fig. 2f), which was in agreement with the previous conclusion that oxygen vacancies existed in the lattice. 35 Ionic conductivity of LLTO and N-LLTO nanofibers.-The lithium-ion conductivity of the LLTO and N-LLTO nanofibers was measured by EIS with two stainless-steel ion-blocking electrodes. Figure 3a shows the Nyquist plots of LLTO and N-LLTO samples. The depressed semicircle at high-to-intermediate frequencies was attributed to Li + migration in the domain interior and across the domain boundaries. The straight line at low frequencies was ascribed to blocking of Li + at the block electrodes. The ionic conductivity was calculated using the formula: 35 where L is the thickness of the membrane, R is the resistance, and A is the effective electrode area. The resistances of grain (R g ) and grain boundary (R gb ) were obtained through fitting calculation by the ZView program. The total resistance (R t ) is the sum of R g and R gb . Table I lists  The conductivity of the LLTO and N-LLTO nanofibers was comparable with the reported values in literature. 36 After nitrogen doping, the grain boundary conductivity was also improved compared with the pristine LLTO nanofibers. The ion conductivity along the grain boundaries was 0.75 × 10 −6 , 1.52 × 10 −6 , 4.33 × 10 −6 , 2.35 × 10 −6 and 0.77 × 10 −6 S·cm −1 for the LLTO, N-LLTO 525, N-LLTO 550, N-LLTO 575 and N-LLTO 600, respectively. As shown in Table I, among the LLTO and N-LLTO samples, the nitrogen-doped LLTO nanofibers after ammonia-treatment at 550°C achieved the highest total ionic conductivity of 4.28 × 10 −6 S·cm −1 . Both charge mobility and charge carrier concentration contribute to ionic conductivity. The activation energy (E a ) reflects the charge mobility. E a was calculated using the following equation: 35 where T σ ( ) is the ionic conductivity, E a is the activation energy for the ion-hopping conduction process, a is the pre-exponential factor, and R is the universal gas constant (8.314 J mol −1 ). Herein we investigated the influences of nitrogen doping on the ion conductivity and activation energy. Figure 3b presents the Arrhenius plots of the LLTO and N-LLTO samples from room temperature to 90°C. The activation energy for the LLTO nanofibers was 0.38 eV, which was consistent with the reported results. 35,36 After nitrogen doping, the activation energy decreased to 0.29 eV for N-LLTO 550. It can be deduced that the nitrogen doping enhanced mobility of charge carriers in the N-LLTO 550 nanofibers. N-LLTO 575 and N-LLTO 600 exhibited an E a value comparable with LLTO, indicating that the charge mobility in these two samples was close to that in LLTO.
DFT calculation results.-First-principles DFT calculation was performed to investigate the effects of nitrogen doping on the ionic conductivity of LLTO. The equilibrium structure of Li 0.33 La 0.557 TiO 3 can be viewed as a lattice with Ti 4+ ions (B site) at the cube corners while La 3+ (A site) and O 2− ions sit at the body and edge centers, respectively (Fig. 4). The doped N ion was substituted for an oxygen ion in the oxygen plane separating the Larich and La-poor layers. The valency of La 3+ leads to unoccupied A sites (i.e., a stoichiometry of 0.557) and alternating La-rich and Lapoor layers. It is worth noting that Li + ions do not occupy the vacant body centers of the Ti 4+ defined lattice but instead reside at slightly offset four-fold face centers enclosing such unoccupied A sites. This observation is consistent with descriptions from previous calculations of LLTO type perovskites with different stoichiometries. 37,38 Li + migration mechanism is best viewed as hopping from one face center of the Ti 4+ -defined lattice to another face center along a quarter arc, and is accompanied by bending of two corner-sharing TiO 5 N and TiO 6 octahedra due to interaction with the moving Li + ion. This picture is similar to what is described by Kim et al. for a LLTO structure containing a single Li + ion. 38 Our model system contains six non-equivalent Li + ions and a large number of distinct hopping pathways. Fig.5 shows the hopping pathways where a Li + ion starts or ends at a location neighboring the N dopant (shaped region), as Li + migration further away is expected to be minimally affected by the N substitution. The different pathways are indicated by the arrows and letter labels in Fig. 4. In all cases examined, N doping leads to lower or similar barriers in both the forward and the backward jumping directions. Examining the structures of the initial and final states, we have found that structures where two Li + ions occupy neighboring face centers (insets of Figs. 5b-5d, and 5f) are significantly destabilized due to large lattice distortion or off-center Li + positions. This increases the final-state energy ranging from 0.132-0.212 eV for N-LLTO and 0.154-0.215 eV for LLTO relative to the corresponding initial states. When Li + ions sit at the face center involving the N dopant, the Li-O (N) coordination is more weakly bound compared to Li + with four O atoms as the nearest neighbors (e.g., final state in Fig. 5a). These less stable final states also imply that the energy barriers are even lower in N-LLTO for Li + hopping in the reverse directions. The above observations, taken together, suggest that N doping accommodates larger lattice distortion and thus facilitates hopping involving several of the higher-energy states. The largest difference between the N-doped and pristine LLTO is observed in Fig. 5b, where the hopping barrier is 0.264 eV for N-LLTO and 0.377 eV for LLTO (Table II). Longer-range Li + diffusion will inevitably involve more than one hopping modes, for which N-doped materials show consistently comparable and lower barriers.  The computational results have confirmed that the ion conductivity of a solid electrolyte is greatly affected by the distortion of the octahedra [TiO 6 ] structure 39 and the size of the Li + migration bottleneck. 40 A length matching relationship between A-O and B-O bonds in the peroviskite ABO 3 structure has been proposed by Kunz 41 as the criterion for judging octahedral distortion in perovskites. A lattice stress, which is induced by mismatched bond lengths, causes spontaneous distortion of octahedra to relieve such stresses. Our calculation involving a single N dopant used the original LLTO lattice parameters, but at higher nitrogen dopant levels, the larger radius in the oxygen sublattice (r N3 − = 1.5 Å, r O2 − = 1.40 Å) can slightly increase the cell parameter and destroy the length-matching relationship between A-O and B-O bonds. The ionic conductivity is thus reduced when too many nitrogen ions are doped into the lattice due to octahedral distortion that induces narrow bottlenecks surrounded by the [TiO 6 ] octahedra.
Microstructure of composite electrolyte.-As N-LLTO 550 nanofibers exhibited the highest ionic conductivity among the Ndoped samples, N-LLTO 550 was used to form the LLTO-polymer composite electrolyte. Composite electrolytes were prepared by dispersing 30 wt% nanofibers in a PVDF-HFP/LiTFSI salt dimethylformamide (DMF) solution, followed with evaporation of the DMF solvent. Figure 6a shows the cross-section of a 97 μm thick PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte membrane. The free space of 3D porous N-LLTO structure was filled with polymer completely (Fig. 6b). EIS measurement was performed on the PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte, and the Nyquist plot was shown in Fig. 6c. The composite electrolyte with N-doped LLTO nanofibers exhibited ionic conductivity of 3.8 × 10 −4 S·cm −1 , which was higher than that (1.3 × 10 −4 S·cm −1 ) of the PVDF-HFP/LiTFSI/LLTO with pristine LLTO nanofibers. In addition, linear sweep voltammetry (LSV) was carried out to test the electrochemical stability of PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolytes (Fig. 6d). Stainless steel was used as the working electrode, and lithium foils were employed as the counter and reference electrodes for LSV testing. The PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte displayed a stable voltage window up to 4.9 V vs Li|Li + . This indicates that the PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte could be potentially used in highvoltage lithium batteries.
The Li-ion transference numbers (t Li+ ) was measured by chronoamperometry test using symmetric lithium cells at an applied DC voltage of 10 mV. EIS spectra were acquired before and after the DC polarization with a frequency ranging from 1 MHz to 1 Hz. The t Li+ value was calculated by Bruce's equation: Li ss ss ss where ΔV is the polarization voltage, I 0 is the initial current, I ss is the steady state current, R 0 is the initial total resistance, and R ss is the steady state total resistance. The lithium-ion transference number of the polymer electrolyte PVDF-HFP/LiTFSI was 0.17; and it increased with adding the LLTO and N-doped LLTO nanofibers. The lithium transference number of PVDF-HFP/LiTFSI/LLTO and PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolytes were 0.39 and 0.42, respectively. Incorporation of the N-LLTO 550 nanofibers into the polymer can change the local Li + environments due to their stronger synergetic interaction with the polymer electrolytes, which can activate more mobile Li + ions in the polymer and facilitate transport of Li + ions within the polymer. 42 Interaction between ceramic nanofibers and polymer matrix.-The interaction between the inorganic fillers and polymer substrate can suppress crystallinity of polymers. 43 Hence, the effects of the nitrogen doping on the interaction between the ceramic nanofibers and the polymer matrix were investigated. Figure 7a reveals the XRD patterns of the PVDF-HFP, PVDF-HFP/LiTFSI, and  Pristine PVDF-HFP polymer membrane displayed a high degree of crystallinity with strong characteristic diffraction peaks in 2θ range of 18°and 20°with two relatively weak peaks at around 27°and 40°. The characteristic diffraction peaks of PVDF-HFP became weak with addition of lithium salt (LiTFSI). The crystallinity of PVDF-HFP in the composite electrolyte further decreased after the N-LLTO nanofibers were added. It is well known that amorphization of polymer matrix favors conduction of the Li + ions in polymers due to an increase in motion of polymer chain segments and mobility of lithium ions. FTIR and Raman spectra were used to clarify the chemical interaction between the inorganic nanofibers and the polymer. The Raman spectra in Fig. 7b reveal that both the shape and intensity of the vibration modes were altered for the PVDF-HFP polymer after addition of lithium salt and N-LLTO nanofibers. The intensity of the peak at 795 cm −1 , which assigned to the CH 2 rocking vibration of α phase PVDF-HFP, decreased after the lithium salt LiTFSI and N-LLTO nanofibers were added into the PVDF-HFP matrix. The CF 2 stretching vibration mode at 1200 cm −1 of PVDF-HFP also decreased dramatically. A decrease in the CF 2 stretching vibration was ascribed to interaction between the lithium salt and the CF 2 groups in the polymer. 43,44 Reduction of the CH 2 and CF 2 vibration mode intensity indicated deprotonation of the -CH 2 moiety in the polymer chains and reduction of α phase. A new FTIR peak attributed to the C=C stretching vibration modes of polyenes appeared at 1512 cm −1 after the LLTO nanofibers were embedded into the polymer, which suggested dehydrofluorination of the PVDF chains (Fig. 7c). 45 However, the intensity of new peaks at 1510 cm −1 further enhanced after addition of N-LLTO nanofibers, which indicated that deprotonation of the CH 2 groups in PVDF-HFP was more intense with addition of the N-LLTO nanofibers. This could be due to creation of the oxygen vacancies accompanied with nitrogen doping. Our previous studies show that oxygen vacancies in the LLTO nanofibers not only affect the physical properties of ceramic itself, but also enhance the chemical interaction between the LLTO nanofibers and the polymer. 35 The CH 2 bending vibration mode peak at 2980 cm −1 in the FTIR spectra decreased dramatically after addition of the N-LLTO nanofibers (Fig. 7d), which indicated deprotonation of CH 2 in PVDF-HFP. This was in accordance with the Raman spectra in Fig. 7b. Such strong chemical interactions between the inorganic nanofibers, the Li-salt and the polymer can enhance conduction of Li + ion in the ceramic-polymer composite electrolytes. Especially, incorporation of ceramic fillers into the polymer can change the local Li + environment due to their strong synergetic interaction with the polymer electrolytes, which can activate more mobile Li + ions in the polymer and facilitate transport of Li + ions in the polymers. 24  contrast, the Li|PVDF-HFP/LiTFSI/LLTO|Li symmetric cell showed a higher polarization voltage of ±350 mV due to the relatively low ionic conductivity of the electrolyte (Fig. S2 (available online at stacks.iop.org/JES/168/110507/mmedia)).
The Li symmetric cell with the PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte was more stable than the one with the PVDF-HFP/ LiTFSI/LLTO composite electrolyte, which was due to the increased ionic conductivity and Li-ion transference number. An increase in the lithium transference number may reduce the charge concentration gradient and the reversed cell polarization effect. Therefore, it can improve the cycling stability of the electrolyte. The Li symmetric cell with the PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte showed smaller polarization than the one with PVDF-HFP/LiTFSI/LLTO between the discharge and the charge curves at the same current density, which suggested that the PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte can effectively suppress the dead Li formation.
The surface morphology of the Li metal electrode after cycling was characterized with SEM images (Fig. S3). The symmetric cell with PVDF-HFP/LiTFSI/LLTO was used as a control herein. Randomly grown Li dendrites were observed on the Li metal electrode of the symmetric cell after cycling with the PVDF-HFP/ LiTFSI/LLTO composite electrolyte. In contrast, the Li electrode of symmetric cell with the PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte displayed smooth and integrated surface morphology after 500 h of Li plating/stripping cycling testing, which indicated a stable electrode/electrolyte interface.
Full cells of Li-metal batteries with LiFePO 4 (LFP) cathode were used to evaluate the PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte. Fig. 8b shows that a full cell with the PVDF-HFP/ LiTFSI/N-LLTO 550 composite electrolyte was galvanostatically cycled at a current density of 0.2 C between 2.5 V and 4.2 V. Its initial discharge capacity was 143 mAh g −1 and retained 93% after 100 cycles. The Li|PVDF-HFP/LiTFSI/LLTO|LFP full cell, in which pristine LLTO was used, displayed a much lower discharge capacity; and the cell failed after 63 cycles (Fig. S4). In addition, the charge-discharge curves were also tested at different cycles for the LFP|PVDF-HFP/LiTFSI/N-LLTO 550|Li full cell (Fig. 8c). Discharge capacities of 164, 145, 113, 87 and 68 mA h g −1 were observed at varied rates of 0.1, 0.2, 0.5, 1 and 2C, respectively (Fig. 8d). After applying cycles at higher current densities, the discharge capacity returned to as high as 150 mA h g −1 when the current density was reduced to 0.1C. These results confirmed excellent rate performance of the LFP|PVDF-HFP/LiTFSI/N-LLTO 550|Li full cell. The capacity of the all-solid-state Li|PVDF-HFP/LiTFSI/N-LLTO 550|LFP full cell was comparable or higher to that of recently reported Li metal cells with all-solidstate electrolyte. [46][47][48] However, lower rate capacities were found for the Li| PVDF-HFP/LiTFSI/LLTO|LFP cell, showing 135, 113, 78, 33 and 5 mA h g −1 for 0.1, 0.2, 0.5, 1 and 2 C, respectively (Fig. S5).

Conclusions
In conclusion, nitrogen ions were successfully doped into LLTO nanofiber by heat treatment in the ammonia atmosphere, and substituted for the oxygen atoms partially. Nitrogen doping reduced the activation energy of Li + ion hopping and improved the ionic conductivity of the LLTO nanofibers. DFT calculation results revealed that nitrogen doping weakened the anion bonding to Li + cations in the perovskite ABO 3 structure and allowed for larger lattice distortion, thus facilitating Li-ions hopping and leading to overall lower barriers than the undoped LLTO structures. The PVDF-HFP/LiTFSI/N-LLTO 550 exhibited a stable voltage window up to 4.9 V vs Li|Li + . Symmetric Li|PVDF-HFP/ LiTFSI/N-LLTO 550|Li cell showed good stability during repeated lithium plating/stripping at room temperature, and a low overpotential of ∼71 mV under a constant current density of 0.5 mA·cm −2 . A full cell, which was assembled with the PVDF-HFP/LiTFSI/N-LLTO 550 composite electrolyte, exhibited better rate capacity and cycling performance than the counterparts with the pristine LLTO nanofiber-polymer electrolyte.

Acknowledgments
This work was partially supported by the Department of Energy, Office of Energy Efficiency and Renewable Energy (EERE), under Award Number DE-EE0007806. Dr. Bai is grateful to the computation resources at the National Energy Research Scientific Computing Center at DOE Lawrence Berkeley National Laboratory under Contract (DE-AC02-05CH11231) and the Extreme Science and Engineering Discovery Environment through allocation CTS190069 by the National Science Foundation (ACI-1548562).

Disclaimer
Neither the U. S. Government, nor any of its employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy or completeness of any information, product, or process disclosed, or represents that its manufacture or use would not infringe privately owned rights. The views and opinions of the authors stated do not necessarily reflect those of the U. S. Government.