Orientational Relationship Between the Solid-Electrolyte Interphase and Li₄Ti₅O₁₂ Electrode in Hybrid Aqueous Electrolytes

2.1 Electrode and Electrolyte preparation

Cathode electrodes comprising LiMn₂O₄ (LMO, L&F Co., Ltd.) were produced by blending the active material, carbon black (Super C, Timcal), and polyvinylidene fluoride (PVDF, Solvay) in a 90:5:5 weight ratio with N-methylpyrrolidinone (NMP, 99.5%, Sigma-Aldrich). The mixture was spread onto 20-μm-thick titanium foil and dried in vacuum at 100°C for 12 h. The same procedure was followed to prepare anode electrodes comprising Li₄Ti₅O₁₂ (LTO, POSCO Future M), except for the use of aluminum foil as the substrate. The hybrid aqueous electrolyte (HT-29) was formulated by combining lithium bis(trifluoromethanesulfonyl)imide (LiTFSI, 99.95%, Sigma-Aldrich), water, and trimethylphosphate (TMP, 99%, Sigma-Aldrich) in a 37:40:23 molar ratio [12].

2.2 Electrolyte characterization

The ionic conductivity of the electrolyte was determined by a conductivity meter (S213-meter, Mettler-Toledo), and the viscosity of the electrolyte was determined using a rotational viscometer (Brookfield DV3TLV). The linear sweep voltammetry (LSV) tests for the electrolyte were performed with a three-electrode system constituting an inactive current collector (titanium foil), activated carbon, and Ag/AgCl as the working, counter, and reference electrodes, respectively. The LSV measurements were conducted on a Solatron 1260 instrument equipped with a 1287 Electrochemical Interface (Solatron Metrology, UK) at a scanning rate of 0.1 mV s⁻¹.

2.3 Cell assembly and Testing

The cathode, anode, aqueous electrolyte, and separator were assembled in a dry room with a controlled relative humidity of <0.2%, as depicted in Fig. 1. A 260-μm-thick Whatman GF/A glass fiber was used as the separator. A standard 2032-coin cell was filled with 100 μL of electrolyte, sealed, and allowed to stand for at least 12 h to ensure thorough impregnation. Charge–discharge testing was conducted in a temperature-controlled environment using a Maccor 4000 battery tester, cycling between 2.8 and 1.5 V at a 0.5C rate (0.5 mA cm⁻²).

Fig. 1

Schematic of the fabrication of the coin cell with a LiMn₂O₄ (LMO) cathode, a Li₄Ti₅O₁₂ (LTO) anode, and a hybrid aqueous electrolyte.

2.4 Microstructural analysis

The post-cycling crystallographic orientation of the SEI and LTO was determined using field-emission transmission electron microscopy (FE-TEM, JEOL, JEM-2100F) at 200 kV. Samples were prepared with a focused ion beam (FIB, JEOL, JIB-4601F). HRTEM images and fast Fourier-transform (FFT) analysis were used to verify the crystal structure of the SEI on the LTO electrode. The chemical composition of the SEI was further analyzed using X-ray photoelectron spectroscopy (XPS, Axis-Supra, Kratos).

2.5 Computational details

Spin-polarized DFT calculations were conducted using the Vienna Ab initio Simulation Package (VASP) [13] using the projector-augmented wave (PAW) method and the Perdew–Burke–Ernzerhof (PBE) [14] exchange-correlation functional within the generalized gradient approximation (GGA). The Grimme D3 dispersion correction was employed to account for van der Waals interactions [15], and calculations were performed with a plane-wave basis set with a cutoff energy of 600 eV for increased accuracy.

For bulk structure optimization, Monkhorst–Pack (6×6×6) k-point meshes for LiF and (3×3×3) for LTO were used. Slab models representing the (200) crystallographic planes of LiF and LTO were derived from these optimized bulk structures, selected based on the results of TEM analysis. The slabs were further optimized with (6×6×1) k-point meshes for LiF and (3×3×1) for LTO. The lowest-energy configurations from various (200) slab models with different surface terminations were chosen for subsequent analyses, as shown in Fig. 2 and 3. The LiF (200)/LTO (200) interface model, constructed from these slabs, was refined using a (6×3×1) k-point meshes, as illustrated in Fig. 4. A 20 Å vacuum layer was introduced perpendicular to the slabs to prevent artificial periodic interactions, and the bottom two layers were fixed to mimic the bulk conditions, allowing the upper layers to relax.

Fig. 2

Optimized structures of Li₄Ti₅O₁₂: (a) bulk, (b) side view, and (c) top view of the (200) plane surface of Li₄Ti₅O₁₂. Green, blue, and red spheres represent Li, Ti, and O atoms, respectively. Li⁺ ions on the 16d sites are encircled; other Li⁺ ions occupy the 8a sites.

Fig. 3

Optimized structures of LiF: (a) bulk, (b) side view, and (c) top view of the (200) plane surface of LiF. Green and gray spheres represent Li and F atoms, respectively.

Fig. 4

Optimized structures of the LiF(200)/Li₄Ti₅O₁₂(200) interface: (a) side view, and (b) top view. Green, blue, red, and gray spheres represent Li, Ti, O, and F atoms, respectively.

The chemical stability and chemical bonding at the LiF/LTO interface were assessed by calculating the formation energy (E_f) as follows:

where E_LiF/LTO, E_LTO, and E_LiF represent the total energies of the combined interface, LTO layer, and LiF layer, respectively.

3.1 Cell performance

Table 1 shows the ion conductivity of the hybrid aqueous electrolyte at 23C and its viscosity at 25C and 40C. The ion conductivity of this hybrid aqueous electrolyte was smaller than that of a non-aqueous electrolyte [16]. Its viscosity was higher than that of a traditional non-aqueous electrolyte, owing to the high concentration of LiTFSI salts in the former.

Table 1

Ion conductivity of the hybrid aqueous electrolyte at 23°C and its viscosity at 25°C and 40°C

The electrochemical stability window (ESW) of the hybrid aqueous electrolyte was determined from the linear sweep voltammetry curves (Fig. 5). The cathodic and anodic limits were 1.76 and 4.66 V, respectively, with an overall ESW of 2.9 V. The redox reaction potentials of LMO and LTO are completely inside the ESW of the hybrid aqueous electrolyte.

Fig. 5

Linear sweep voltammetry curve of the hybrid aqueous electrolyte showing the electrochemical stability window.

An LMO–LTO full cell was constructed to assess the electrochemical performance of the hybrid aqueous electrolyte. The initial voltage profile, recorded at a current density of 0.5C (0.5 mA cm⁻²), is shown in Fig. 6. The first cycle exhibited a charge voltage plateau at ~2.50 V and a discharge plateau at 2.35 V, indicative of Li⁺-ion insertion and removal in the LMO–LTO system. The initial discharge capacity (146.1 mAh g⁻¹) was marginally lower than the theoretical capacity (175 mAh g⁻¹). The coulombic efficiency of the first cycle was 80.1%, suggesting electrolyte consumption for SEI formation and H₂O decomposition at the anode. In the subsequent cycle, the coulombic efficiency improved to above 90%, suggesting minimal interface formation. After 100 cycles at 0.5C, the cell retained a discharge capacity of 51.9 mAh g⁻¹, which is 35.4% of the initial capacity, and a coulombic efficiency of ~97.8%.

Fig. 6

Charge/discharge cycle performance at 0.5C.

The XPS spectra of the SEI on the anode are shown in Fig. 7. The F 1s spectrum of the LTO surface shows peaks of −CF₂ (686.3 eV) of the PVDF binder in the electrode, −CF₃ (688.6 eV) of the TFSI⁻ salt anion, and the inorganic LiF (685.3 eV) formed by the reduction of [Li]²⁺[TFSI]⁻. This suggests that the LiF molecules within the SEI effectively inhibit the hydrogen evolution reaction on the anode by creating a protective film on the LTO surface.

Fig. 7

X-ray photoelectron spectra of the Li₄Ti₅O₁₂ anode cycled 30 times.

Additionally, the C 1s spectrum exhibited peaks of −CF₃ (293 eV) from TFSI⁻ or its reduction byproducts, −CO₃ (290.5 eV) originated from the reaction of CO₂ with H₂O, alkyl carbonates (289 eV), ether (C–O) (286 eV), and other carbonaceous sp³- or sp²-hybridized elemental carbon species below 284 eV. The −CO₃ peak is indicative of the interaction between atmospheric CO₂ and H₂O [17].

3.2 Microstructure of the SEI

Fig. 8 shows a high-resolution image of the specimen after 30 cycles, in which the SEI layer is discernible by the fringe contrast with the LTO electrode, measuring ~10 nm in thickness, as shown in Fig. 8a. Before FIB sample preparation, a Pt layer was applied to the SEI surface to prevent ion-beam damage [18]. The FFT analysis (Fig. 8b) identified the crystal structure of the SEI, as LiF has a face-centered cubic (FCC) lattice (Fm-3m; a = 4.027 nm). The epitaxial growth of the SEI on the LTO electrode is facilitated by the minimal lattice mismatch and the similarity in the FCC crystal structures of the LTO and LiF phases. This allows the LTO to serve as a substrate for the LiF phase. The results confirm that the LiF SEI layer grew epitaxially on the LTO electrode surface, for which the (200) and (111) planes are parallel between the two phases. HR-TEM observations revealed a semi-coherent interface between the LTO and LiF SEI layers, as indicated by the nearly 2:1 ratio of their lattice constants. This small lattice mismatch suggests that the LTO electrode is an effective nucleation site for the LiF SEI layer. The well-formed SEI layers prevent continuous side reactions and repeated SEI formation, leading to reduced resistance and enhanced stability [11]. The crystalline and dense SEI layer fully passivates the electrode surface, maintaining mechanical stability during cycling. The epitaxially grown LiF SEI on the LTO electrode demonstrates uniformity and crystallinity with minimal defects, effectively isolating the electrode from the reactive aqueous electrolyte and providing a robust surface.

Fig. 8

High-resolution transmission electron microscopy images of (a) the LTO electrode/SEI layer interface and (b) the corresponding fast Fourier-transform image obtained with beams parallel to the [011] direction.

3.3 DFT calculations

DFT calculations were conducted to investigate the atomic-scale interactions and stability of the LiF SEI layer on the LTO electrode. The results indicate an energetically favorable LiF-LTO interface structure with a formation energy of −5.26 eV, suggesting strong adhesion and a thermodynamically stable interface. Specifically, upon the formation of the interface, the bond length of Li–F within the LiF(200) surface decreased from 2.04 to 1.97 Å. Additionally, the bond lengths of Ti–O and Li–O within the LTO (200) surface changed from 1.93 to 2.09 Å and 1.97 to 1.92 Å, respectively. Moreover, as a direct consequence of the interface formation, new Li–O and Ti–F bonds with bond lengths of 1.98 and 2.13 Å, respectively, have emerged (Fig. 4). Furthermore, the lattice mismatch between LiF and LTO is only 2.22%, suggesting the epitaxial growth of the SEI layer on the LTO substrate. Significant charge redistribution was observed at the LiF/LTO interface, which is critical for the stability of the SEI (Fig. 9). The charge density difference (Δρ) for LiF/LTO was calculated as follows:

Fig. 9

Charge density difference of the LiF(200)/Li₄Ti₅O₁₂(200) interface, visualized using VESTA [19]. Areas of charge accumulation are indicated in yellow, whereas regions of charge depletion are shown in cyan. Isosurface = 0.0015 e Bohr⁻³. Green, blue, red, and gray spheres represent Li, Ti, O, and F atoms, respectively.

where ρ_(LiF/LTO), ρ_LTO, and ρ_LiF represent the charge densities of the interface, the LTO layer, and the LiF layer, respectively. The results of these calculations indicate electron redistribution and localization because of interfacial interactions. In addition, Bader charge analysis [20] revealed that the LTO layers gain 0.30 electrons, whereas the LiF layers lose 0.30 electrons, which enhanced the ionic character of the bonding and promoted stability at the LiF/LTO interface.