The performance of lithium-ion battery cells is intrinsically linked to the complex, three-dimensional microstructure of their porous electrodes. The cathode, in particular, is a composite material where active material particles, conductive carbon additives, and binder form a porous network infiltrated by electrolyte. Understanding and optimizing this microstructure is crucial for enhancing properties like energy density, rate capability, and cycle life. Traditional experimental methods for analyzing and reconstructing these microstructures, such as focused ion beam/scanning electron microscopy (FIB/SEM) or X-ray computed tomography (XCT), are powerful but can be time-consuming and costly for exploring a wide design space. This work presents a framework that combines experimental characterization, numerical reconstruction, and validation to rapidly generate and assess virtual 3D microstructures of a LiNi0.8Co0.1Mn0.1O2 (NCM811) cathode, providing a valuable tool for investigating the impact of microstructural features on transport phenomena within the lithium-ion battery.
The NCM811 cathode sample was fabricated using a standard slurry-casting process. The active material, conductive carbon black (CB), and polyvinylidene fluoride (PVDF) binder were mixed in a weight ratio of 96:2:2 using N-methyl-2-pyrrolidone (NMP) as the solvent. The homogeneous slurry was then coated onto an aluminum foil current collector, dried, and calendared to 80% of its original thickness (20% compression) to achieve a dense, conductive network. This calendaring process is critical for the final microstructure and the volumetric energy density of the lithium-ion battery.
The microstructure of the prepared NCM811 cathode was characterized using both Scanning Electron Microscopy (SEM) and X-ray Computed Tomography (XCT). SEM images, as shown in representative cross-sections, reveal the primary microstructural features: spherical secondary particles of the active material, pores of various sizes, and a filmy substance surrounding the particles which is the mixed phase of carbon black and binder (additive). At higher magnification, it becomes apparent that the secondary particles themselves are agglomerates of smaller primary particles, containing intrinsic micro-pores. XCT scanning provided a full 3D volumetric dataset of the cathode sample. Through threshold segmentation of the reconstructed 3D image stack using visualization software (e.g., Avizo), the volume fractions of the distinct phases were quantified. The composition of the reconstructed cathode sample is summarized in the table below.
| Phase | Active Material (NCM811) | Additive (CB+PVDF) | Pore |
|---|---|---|---|
| Volume Fraction (%) | 65.3 | 7.2 | 27.5 |
Furthermore, a statistical analysis of the active material particle sizes was performed on the experimentally reconstructed 3D model. The particle diameters were found to follow a normal distribution, with the highest frequency occurring around a specific size. This distribution data is essential for guiding the numerical reconstruction algorithm.
| Particle Diameter (μm) | ~2 | ~3 | ~4 | ~5 | ~6 | ~7 | ~8 | ~9 | ~10 |
|---|---|---|---|---|---|---|---|---|---|
| Frequency (%) | 0.7 | 1.0 | 2.3 | 4.3 | 9.8 | 16.4 | 18.3 | 14.5 | 11.0 |
While the experimental reconstruction from XCT provides a “ground truth” model, it is specific to that single sample and process condition. To efficiently explore microstructural variations (e.g., different porosity, thickness, or composition), a numerical reconstruction method was developed. This algorithm, implemented in MATLAB, generates a stochastic 3D microstructure based on key statistical parameters extracted from the experimental data. The reconstruction logic is based on several reasonable assumptions pertinent to the lithium-ion battery cathode:
- Particle Morphology: Active material particles are modeled as perfect spheres. Their high Young’s modulus justifies the assumption that their size and spherical shape remain largely unchanged after calendaring.
- Particle Placement: Sphere centers are randomly and uniformly distributed within the 3D computational domain. Overlap between spheres is permitted to mimic the agglomeration and packing observed in real electrodes.
- Particle Size Distribution: Sphere diameters are not constant but are randomly selected from a normal distribution matching the experimental data.
- Additive Phase: The nano-scale pores within the additive phase are not resolved at this pore-scale. The additive is treated as a solid phase that randomly occupies voxels adjacent to the surfaces of the active material spheres until its target volume fraction is reached.
The algorithm proceeds in two main steps, as illustrated in the workflow. First, active material spheres are sequentially generated until the volume fraction of solid material (active + additive) matches the target porosity. Second, the additive phase is randomly assigned to voxels at the interface of the generated spheres. The inputs to the algorithm are the domain size, the target phase volume fractions (from the table above), and the parameters of the normal distribution for particle diameters. This “two-step” method allows for the rapid generation of cathode microstructures with controlled properties.
To validate the numerical reconstruction algorithm, a quantitative comparison was made between the virtual model and the experimentally reconstructed model. Both models were created with an identical domain size for a fair comparison. The validation focused on three key aspects:
1. Active Material Particle Size Distribution: The diameter distribution of spheres in the numerical model was extracted and compared to that of the experimental model. As shown in the graph below (conceptual representation), both distributions align closely, following a similar normal distribution curve with the peak at the same diameter, confirming that the stochastic algorithm correctly captures the morphological statistics of the lithium-ion battery cathode.
2. Through-Thickness Porosity Profile: The local porosity was calculated for each slice along the thickness direction (Z-axis) for both models. While the experimental model showed slightly higher local fluctuation (range: 0.20-0.31), the numerical model’s porosity varied within a narrower band (0.26-0.29). Critically, the average porosity for both was consistent with the target value of 0.275, and both profiles oscillated around this mean, indicating similar structural heterogeneity.
3. Effective Ionic Transport Property: The most critical validation involves comparing macroscopic transport properties predicted from the 3D microstructures. Pore-Scale Modeling (PSM) was employed to simulate steady-state ion diffusion through the pore phase of both models. In this simulation, transport is driven by diffusion only, governed by Fick’s law:
$$ \nabla \cdot \mathbf{J}_e = \nabla \cdot (-D \nabla c_e) = 0 $$
where $\mathbf{J}_e$ is the ionic flux (mol m-2 s-1), $D$ is the bulk diffusion coefficient, and $c_e$ is the ion concentration. By applying a concentration gradient across the domain in the Z-direction and calculating the total flux, the effective ionic diffusivity $D_{eff}$ can be determined:
$$ D_{eff} = \frac{\mathbf{J}_e \cdot l}{c_2 – c_1} $$
where $l$ is the domain length, and $c_2$, $c_1$ are the boundary concentrations. Multiple samples of both experimental and numerical reconstructions were simulated. The results, summarized below, show that the average effective diffusivity from the numerical models is in excellent agreement with that from the experimental models, with a negligible difference. This confirms that the numerically generated microstructure is not just morphologically similar but also functionally equivalent in terms of ion transport—a key performance metric for the lithium-ion battery.
| Model Type | Average Effective Ionic Diffusivity, $D_{eff}$ (m²/s) | Standard Deviation |
|---|---|---|
| Experimental Reconstruction (10 samples) | D_exp_avg | σ_exp |
| Numerical Reconstruction (10 samples) | D_num_avg | σ_num |
| Difference | ~2 × 10⁻¹⁷ | – |
In conclusion, this study successfully integrates experimental characterization with a stochastic numerical algorithm to reconstruct and validate the 3D pore-scale microstructure of an NCM811 lithium-ion battery cathode. The developed “two-step” numerical method, informed by statistical data from XCT and SEM analysis, can rapidly generate realistic virtual electrodes. The validation through direct morphological comparison (particle size, porosity profile) and, more importantly, through the prediction of effective ionic transport properties demonstrates the algorithm’s accuracy and utility. This framework provides a powerful and efficient computational tool to explore the relationship between manufacturing parameters (e.g., calendaring pressure, slurry composition), the resulting microstructure, and the consequent electrochemical performance. It enables virtual prototyping and optimization of electrode architectures for next-generation lithium-ion battery cells, reducing reliance on purely empirical and costly experimental trials. The methodology is also readily applicable to the microstructural modeling of other porous materials in energy storage and conversion devices.