Twin boundary defect engineering improves lithium-ion diffusion for fast-charging spinel cathode materials

Defect engineering on electrode materials is considered an effective approach to improve the electrochemical performance of batteries since the presence of a variety of defects with different dimensions may promote ion diffusion and provide extra storage sites. However, manipulating defects and obtaining an in-depth understanding of their role in electrode materials remain challenging. Here, we deliberately introduce a considerable number of twin boundaries into spinel cathodes by adjusting the synthesis conditions. Through high-resolution scanning transmission electron microscopy and neutron diffraction, the detailed structures of the twin boundary defects are clarified, and the formation of twin boundary defects is attributed to agminated lithium atoms occupying the Mn sites around the twin boundary. In combination with electrochemical experiments and first-principles calculations, we demonstrate that the presence of twin boundaries in the spinel cathode enables fast lithium-ion diffusion, leading to excellent fast charging performance, namely, 75% and 58% capacity retention at 5 C and 10 C, respectively. These findings demonstrate a simple and effective approach for fabricating fast-charging cathodes through the use of defect engineering.


Supplementary Tables
Supplementary Table 1: Joint refinement result of the neutron diffraction and XRD data of LMO.    In the mapping region of 17×10 μm 2 , twinning particles can be easily observed, marked, magnified and exhibited. It is known that three Euler angles Ψ, θ, φ (around the z', x' and z' axes, respectively) can be used to describe the orientation of each grain with respect to the sample coordinates. By adopting matrix multiplication, the rotation can be described as: According to the HAADF STEM images in Here, two typical twinning particles are enlarged and demonstrated in Supplementary Fig. 8. In Supplementary Fig. 8a-c, the [110] direction of the particle is nearly parallel to the normal direction (Z-axis) of the sample plane. Therefore, the {110} pole figure in Supplementary Fig. 8b displays two sets of trace points that have mirror symmetry and an angle difference of ~70°, which corresponds well with the discussion above. From the {111} pole figure in Supplementary Fig. 8c, one shared trace point is readily observed, representing the twin boundary of the grains. No other shared {111} point exists, indicating that there is only one variant in this particle. Similar results are obtained in another twinning structure, as shown in Supplementary  Fig. 8d-e, except that the [110] direction is not perpendicular to the observation plane. At low frequencies, the Li + diffusion coefficient formula can be expressed as follows 1,2 : where R is the gas constant, T is the Kelvin temperature, A is the electrode surface area, n is the number of transferred electrons per mole in the electrode reaction, C is the concentration of lithium in the electrode, and Aw is the Warburg coefficient. In addition, the Warburg coefficient Aw has the following relationship 1,2 : where RS is the solution resistance and Rct is the charge-transfer reaction resistance.
Therefore, the value of the Warburg coefficient can be obtained from the Z'- -1/2 diagram, and the slope fitted from the Z'- -1/2 diagram is the Warburg coefficient Aw. According to equation (1) and the Warburg coefficient Aw, we can obtain the lithiumion diffusion rate DLi+. The Li + diffusion rate calculated from the GITT curve is based on the following formula according to Fick's second law of diffusion 3,4 : where τ is the duration of the current pulse, n is the number of moles, V is the molar volume of the electrode, S is the electrode-electrolyte contact area, and ΔEs and ΔEt are the change in the steady state potential and the total change during the current flux by deducting the IR drop, respectively.

Supplementary Discussion
Systematic Analysis of the compositions, defects and properties. To further validate the positive correlations between the amount of excess Li, contents of defects and lithium-ion diffusion kinetics in spinel LMO material, a batch of Li1+δMn2O4 samples with different contents of excess Li, i.e., δ = 0% (LMO), 2%, 4%, 6%, 8% (LMO-TB), and 10%, were studied. The XRD (Supplementary Fig. 22) patterns show that the samples with a Li excess below 8% are all single phases, whereas a small number of impurities appears in that with a Li excess equal to 10%, indicating that the limit of excess lithium is 8%. Based on the Rietveld refinement results, the lattice constants of all samples are deduced, and the lattice constant of samples with the pure spinel phase decreases in a monotonic manner upon the integration of excess lithium.
This result signals not only a shrinkage in the unit cell but also the successful integration of different amounts of excessive Li + into the structural lattice.
Consequently, additional bright-field TEM experiments were performed on samples with three representative contents of excess Li, i.e., Li-excess δ = 0%, 4%, and 8%, respectively. All the bright-field STEM images of the stoichiometric LMO sample (Supplementary Fig. 23a) show that twin boundaries can hardly be observed, although more than 60 particles are studied and these particles exhibit various morphologies. In contrast, 11 twin boundaries are found in 62 measured particles measured in the LMO-4% sample (Supplementary Fig. 23b), yielding a calculated ratio (numbers of twin boundaries divided by the numbers of particles measured) of approximately 17.7%. As for the LMO-8% sample, i.e., LMO-TB, this ratio further increases to approximately 37.9% (Supplementary Fig. 23c). Clearly, the concentration of twin boundaries in the LMO-8% sample is higher than that in the LMO-4% sample, and quite a few particles possess more than one twin boundary. It should be noted that for randomly dispersed particles, especially with a large amount of particle agglomerations and with tilted twinning boundaries that may not be seen under a fixed incident electron beam, quantitative measurements of the twin boundary ratio are challenging, if not even impossible. Nevertheless, according to the statistics listed above, it can be fairly concluded that the number of planar defects, i.e., twin boundaries, is consistent with the amount of excess Li.
Next, the electrochemical properties of the Li1+δMn2O4 samples were evaluated. First, Supplementary Fig. 24a shows that the specific capacities at all rates (0.2, 0.5, 1, 2, 5 and 10C) increase with the content of TB defects in the range of a Li excess δ = 0% to 8%. Once the amount of excess lithium is further increased up to 10%, the rate capacities decrease due to poor lithium-ion diffusion in the newly formed Li2Mn2O4 impurity 5 . In addition, Supplementary Fig. 24b shows that the appearance of an impurity phase in the LMO-10% sample decreases the cycling life of the battery. Furthermore, EIS of all Li1+δMn2O4 samples under the same battery conditions was performed (Supplementary Fig. 24c). On the basis of the fitting results at low frequency (Supplementary Fig. 24d), the lithium-ion diffusion coefficients of Li1+δMn2O4 (δ = 0% (LMO), 2%, 4%, 6%, and 8% (LMO-TB)) samples are deduced to be 0.59×10 -9 , 0.65×10 -9 , 0.75×10 -9 , 0.84×10 -9 , and 1.02×10 -9 cm 2 s -1 , respectively ( Supplementary Fig. 25), indicating the gradual increase in diffusion kinetics with an increasing amount of excess lithium and an increasing concentration of twin boundaries.
In summary, the LMO-TB sample possesses the best electrochemical performance, which can be attributed to the generation of the largest content of TB defects. The results of LMO-TB can be well explained within the framework of twin boundary defect engineering.