Introduction
Nematic liquid crystals (NLCs) are characterized by rod-like molecules that exhibit long-range orientational order. Modeling their behavior in confined geometries is crucial for technological applications like displays and sensors. Two primary mathematical frameworks are commonly used: the vector-based Ericksen–Leslie (EL) model and the tensor-based Landau–de Gennes (LdG) model. While the EL model is simpler, it assumes molecular "head-to-tail" asymmetry (orientability), whereas the LdG model accounts for the non-orientable nature of real NLC molecules. This study systematically compares the solution landscapes of a simplified EL (sEL) model and a reduced LdG (rLdG) model in a 2D square domain with tangent boundary conditions.
Methodology
The researchers employed a Solution Landscape Approach to map the full hierarchy of stable and unstable states, including energy minima and saddle points. This involves using numerical algorithms, such as Saddle Dynamics (SD), to identify states based on their Morse Index—the number of directions in which a state is unstable.
To ensure a fair comparison, both models were rescaled into Ginzburg–Landau type functionals. The effective domain sizes were matched using specific parameters: $\eta$ for the sEL model and $\alpha$ for the rLdG model, as shown in the table below.
The study utilized symmetry-exploiting algorithms to account for the rotational and reflective symmetries of the square domain, significantly reducing the computational cost of finding high-index saddle points on a $50 \times 50$ mesh.
Key Findings and Model Comparison
The investigation revealed significant similarities and disparities between the two models across different domain sizes.
Shared Features
Both models agree on fundamental stable states, such as the diagonal and rotated states, particularly in smaller domains. They also exhibit similar bifurcation behaviors—such as pitchfork bifurcations—where stable states split as the domain size increases. A "sub-landscape" correspondence was found for low-index solutions (Morse index $\le 4$), suggesting that the simpler sEL model can serve as a reliable proxy for the rLdG model in certain regimes.
Disparities and "Fake Defects"
A major difference arises from the vector nature of the sEL model. Unlike the tensor-based rLdG model, the sEL model cannot represent half-integer defects where the director rotates by 180 degrees. Consequently, it produces "fake defects"—mathematical artifacts where the vector field abruptly flips direction to minimize energy. Furthermore, the sEL model can stabilize $\pm 1$ vortices ($C\pm$ states) at the center of the square, which are typically unstable in the rLdG model, where they tend to split into pairs of $\pm 1/2$ defects.
Novel Solutions and Pathways
The sEL model also exhibited unique "tessellating" solutions—complex patterns formed by tiling simpler states. These configurations are partly attributed to the "weak" boundary conditions used in the sEL simulations. Additionally, transition pathways in the sEL model were found to be more flexible and varied, suggesting that switching mechanisms in vector-based models might be easier to trigger than in tensor-based ones.
Conclusion
The solution landscape approach provides a comprehensive framework for comparing mathematical models in physics. While the sEL and rLdG models share a core set of solutions, the vector representation in sEL introduces distinct topological features and transition pathways. Future work will explore extending this approach to 3D systems and investigating whether "tessellating" solutions appear in the LdG model under similar weak boundary conditions.
Summary updated