Landau-de Gennes numerical simulation of nematic liquid crystals utilizing radial basis functions
Jin-Sheng Wu, Ivan I. Smalyukh · 2026
Read Landau-de Gennes numerical simulation of nematic liquid crystals utilizing radial basis functions by Jin-Sheng Wu, Ivan I. Smalyukh (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-13991.txt.
Introduction
Wu and Smalyukh simulate nematic liquid crystals with radial basis functions under the Landau-de Gennes model. Grid-based codes struggle with complex shapes, while mesh-free radial basis methods adapt with ease. The authors stress gains in geometric flexibility and compute cost over cubic grids. Tests on liquid crystal and colloid systems recover the key topological and energetic features of equilibrium states. An adaptive node refinement step sharpens the fine structure of singular defects.
Catalog record
| CREATOR | Jin-Sheng Wu, Ivan I. Smalyukh |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36817 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.13991
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Landau-de Gennes numerical simulation of nematic liquid crystals utilizing radial basis functions Jin-Sheng Wu, Ivan I. Smalyukh Abstract Numerical simulations based on radial basis functions have been developed for systems with complex geometries and have been successfully applied across various fields, including seismology, coastal hydrodynamics, and biology. However, examples in liquid crystal modeling are limited. In this study, we present a Landau-de Gennes numerical simulation of nematic liquid crystals utilizing radial basis functions, emphasizing its advantages over traditional cubic grid calculations, such as enhanced geometric flexibility and improved computational efficiency. Through simulations of liquid crystal-colloid systems with diverse geometries, we demonstrate that our approach effectively captures the essential topological and energetic features of liquid crystal equilibrium structures. Additionally, we introduce an adaptive node refinement scheme that is crucial for resolving the fine structure of singular defects in nematic liquid crystals. I Introduction Radial basis functions (RBFs) provide a powerful numerical methodology for solving partial differential equations via an elaborate technique that utilizes radially symmetric functions as basis functions instead of delta functions 10 ; 5 ; 21 ; 13 . Therefore, this approach has many advantages as compared to other numerical approaches, especially when applied to large-scale problems. This versatile numerical method has been extensively used in a diverse range of fields, including fluid mechanics, astrophysics, geosciences and mathematical biology, among many other. The Radial Basis Function-generated Finite Difference Method (RBFFD) is essentially a local approximation of RBF, significantly reducing computational costs, often by several orders of magnitude. Specifically, the cost benefit can be estimated as N/n , where N is the total number of nodes—typically ranging from thousands to millions—and n is the number of nearest neighbors considered in the finite difference calculation, usually in the dozens. As a mesh-free numerical approach of the RBFFD method allows for its application to arbitrary geometries since nodes can be placed freely, being unconstrained by grids or geometry of computational boxes 10 . Furthermore, high orders of accuracy, ranging from 4th to 8th order, can be easily achieved within the RBF framework. More importantly, it performs exceptionally well for parallelization on distributed memory computers and GPUs. Finally, yet another advantage is that the local RBF node refinement becomes straightforward without the necessity of mesh construction, regardless of the number of dimensions. Despite of many benefits offered by the RBF-based methodology, its use in soft condensed matter so far is limited to vectorial Frank-Oseen-based modeling of solitonic director structures 35 ; 7 , to the best of our knowledge. At the same time, the more complex tensorial Landau-de Gennes modeling of orientationally ordered media by utilizing the RBF-based methodology has not been demonstrated, which is the goal of this our work. In this study, taking advantage of the geometric flexibility of the RBF-based methodology, we show how RBFFD facilitates tensor-based numerical modeling is utilized for uncovering field configurations in confined liquid crystals (LCs) with embedded colloids of complex shapes or inside droplet-confining volumes with nontrivial topologies. The advantages of using RBFs in LC simulations extend to simpler and faster algorithms for adaptive meshing, which is crucial for performing local refinements near the cores of singular defects and geometric features of confining surfaces. The higher accuracy and potential for parallelization enable the modeling of larger LC systems (as well as LC colloidal dispersions, emulsions and other types of complex composite soft matter). While the application of the RBFFD method in Q -tensor-based liquid crystal simulations has not been documented in the literature to our knowledge, the RBFFD method has been successfully implemented in various range of fields for solving partial differential equations (PDEs) with flexible geometries. These include numerical computation of thermal convection, reaction-diffusion equations, Navier–Stokes equations, and electromagnetics in materials modeling, astrophysics, geophysics, biology, among many others 26 ; 8 ; 28 ; 39 ; 11 ; 4 ; 29 ; 25 ; 14 ; 35 ; 18 ; 15 ; 22 . Thus, the current simulation framework for LC Q -tensor modeling can be significantly enhanced by incorporating the RBFFD method. Irregular node sets of the RBF-based techniques are also beneficial to numerical calculations because they do not impose artificial symmetries which introduce undesired numerical artifacts. Therefore, overall, the RBF-based techniques have the potential to improve the performance of numerical simulations of LCs, as we demonstrate below. II Methods Figure 1: Example superpositions of radial basis functions in one- (a) and two- (b) dimensional systems. II.1 Nematic liquid crystal free energy functional Equilibrium structures of LCs were obtained through numerical minimization of liquid crystal energies. First, the free energies are expressed in terms of the Landau-de Gennes expansion of the bulk free energy densities. These includes the thermotropic terms that govern the isotropic-nematic phase transition f_{\textrm{thermo}}=\frac{A}{2}Q_{ij}Q_{ji}+\frac{B}{3}Q_{ij}Q_{jk}Q_{ki}+\frac{C}{4}({Q_{ij}Q_{ji}})^{2} (1) and one-constant approximation of the elastic energy arising from the deformation to the director alignment f_{\textrm{elastic}}=L\partial_{k}Q_{ij}\partial_{k}Q_{ij} (2) where Q is the tensorial order parameter for the moelcular director; A , B , and C are parameters related to the isotropic-uniaxial phase behavior of CLCs 12 ; 38 ; 1 ; L is an elastic constant determining the associate energy costs of spatial distortions 23 ; 40 ; 16 . Besides, surface boundary conditions at colloidal-molecular interfaces are added to the calculation as conically degenerate surface anchoring described by 42 : \displaystyle f_{\textrm{surf}}=W\left(P_{ik}\tilde{Q}_{kl}P_{lj}-\frac{3}{2}S_{\textrm{eq}}\cos^{2}\theta_{\textrm{e}}P_{ij}\right)^{2} (3) with W the anchoring strength, \textbf{P}=\textbf{v}\otimes\textbf{v} the surface projection tensor, v the surface normal director, and \tilde{\textbf{Q}}=\textbf{Q}+\frac{1}{2}S_{\textrm{eq}}\textbf{I} . Here, f_{\textrm{surf}} is independent of the azimuthal angle of the principal molecular axis with respect to the surface normal vector. If not stated otherwise, planar degenerate anchoring \theta_{\textrm{e}}=90\text{\textdegree} was adopted in our simulations. Finally, we obtained the total energy by the integration of the domain F_{\textrm{total}}=\int d\textbf{r}^{3}~(f_{\textrm{thermo}}+f_{\textrm{elastic}})+\int d\textbf{r}^{2}~f_{\textrm{surface}} (4) and equilibrium structures were found based on the gradient descent method \frac{d\textbf{Q}}{dt}\propto-\frac{\delta F_{\textrm{total}}}{\delta\textbf{Q}} (5) with adaptive Runge-Kutta integration (ARK23) over the imaginary timescale t . Fast inertia relaxation engine (FIRE) was incorporated as well 36 . A typical numerical iteration of energy minimization contains one ARK step and ten FIRE steps. The initial configurations were unidirectional with the molecular alignment along \hat{\textbf{n}}_{0} for LC simulations involving colloidal spheres and handlebodies (Figs. 2 a, 4 , and 5 ). In contrast, the local director field conditions represented within the nematic volume were initially oriented perfectly perpendicular to the center along the circumferences for donut-shaped LC droplets (Fig. 3 ). For a typical computation of LC structure, the following parameter values were used 30 ; 24 ; 40 ; 16 : A=-1.72\times 10^{5}~\textrm{Jm}^{-3} , B=-2.12\times 10^{6}~\textrm{Jm}^{-3} , C=1.73\times 10^{6}~\textrm{Jm}^{-3} , L=20~\textrm{pN} , and W ranges from 2\times 10^{-4} to 10^{-2}~\textrm{Jm}^{-2} . The equilibrium S_{\textrm{eq}}\approx 0.533 . II.2 Radial basis function and finite difference coefficient Similar to the Fourier transformation, we represent position-dependent values, such as the components of the Q tensor, as a linear combination of n independent functions 11 ; 9 ; 35 ; 8 G(\textbf{r})\approx\sum_{i=1}^{n}g_{i}\phi_{i}(\textbf{r})=\sum_{i=1}^{n}g_{i}\phi(\textbf{r}-\textbf{r}_{i}) (6) where \phi_{i} are the radial basis functions and g_{i} are scalar coefficients (Fig. 1 ). In fact, pseudospectral methods, including Fourier transformation, can be viewed as the limit of RBF method 10 , whereas in conventional computations on regular grids, the Dirac delta function is chosen at each grid point \phi_{i}(\textbf{r}-\textbf{r}_{i})=\delta(\textbf{r}-\textbf{r}_{i}) . For RBFFD method, common choices of the