🔬 materials science

Network of localized magnetic textures revealed using a saddle-point search framework

This paper presents a computational framework that systematically identifies first-order saddle points to map the energy landscape and transition mechanisms of metastable magnetic textures in 2D chiral magnets, revealing a hierarchical network of nucleation, annihilation, and rearrangement processes that preserve or alter topological charge.

Original authors: Hendrik Schrautzer, Tim Drevelow, Hannes Jónsson, Pavel F. Bessarab

Published 2026-09-29

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Hendrik Schrautzer, Tim Drevelow, Hannes Jónsson, Pavel F. Bessarab

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where tiny magnets, arranged in a flat sheet, can form intricate, swirling patterns that behave like distinct, stable objects. These are not just random jumbles of magnetic force; they are organized textures, some resembling tiny whirlpools, others looking like droplets or bags with holes in them. In the realm of physics, these patterns are known as magnetic textures, and they hold a special property called topological charge. Think of this charge as a kind of magnetic fingerprint: a number that counts how many times the magnetic directions wrap around a center. If you try to smoothly change a pattern with one fingerprint into a pattern with a different fingerprint, the magnetic field must tear or break somewhere, creating a sharp discontinuity. However, changing a pattern into another that shares the same fingerprint can be done smoothly, like reshaping a lump of clay without tearing it. Scientists have long been interested in these textures because they could one day power new types of computers or storage devices, but to understand how they work, researchers must know how these shapes transform into one another. Specifically, they need to map out the energy hills and valleys that separate these different states, revealing the exact paths a system takes when it switches from one shape to another.

A team of researchers has now developed a powerful new computational method to explore these energy landscapes, effectively drawing a detailed map of how these magnetic shapes evolve. Instead of guessing how a texture might change, they built a systematic framework that starts with a stable magnetic shape and systematically pushes it in every possible direction to find the lowest energy paths leading to other shapes. The process begins by analyzing the symmetry of the starting shape, identifying its unique geometric features. The researchers then divide the shape into smaller, manageable sections and gently nudge these sections using the natural, low-energy ways the material likes to vibrate. This "nudging" pushes the system out of its comfortable, stable state and onto the edge of an energy hill. Once the system is perched on this edge, a sophisticated mathematical tool takes over, guiding it carefully up the hill to find the very top—the saddle point. This saddle point represents the exact moment of transition, the most difficult part of the journey where the system must overcome an energy barrier to reach a new state. By repeating this process thousands of times, starting from different nudges, the team was able to uncover a complex network of connections between all the possible magnetic shapes.

When they applied this method to a two-dimensional chiral magnet, a material known for hosting a rich variety of these textures, the results revealed a surprisingly organized world beneath the complexity. The researchers found that despite the dizzying array of possible shapes—ranging from simple skyrmions, which are like tiny magnetic bubbles, to more complex structures like skyrmion bags with inner loops or chiral droplets with tails—all the transitions between them could be categorized into just five fundamental mechanisms. Some transitions involve the birth of a "kink," a localized twist in the magnetic line that changes the overall topological charge. Others involve the sudden appearance of a new closed loop or the merging of two existing loops. There are also simpler changes where a texture simply stretches out to form a tail, or where two loops merge while a kink is born at the same time to keep the total charge balanced. The study showed that these five mechanisms act as universal building blocks for change. Whether the starting shape was a simple bubble or a complex bag, the energy required to trigger one of these five specific changes remained remarkably consistent, suggesting that the rules governing these transformations are deeply rooted in the material's nature rather than the specific shape of the texture.

One of the most significant findings concerns the difference between smooth transformations and those that require a break in the magnetic field. The researchers discovered that transitions which preserve the topological charge—smooth reshaping—generally have lower energy barriers than those that change the charge, which require a discontinuity. However, the story is not as simple as "smooth is always easier." As the researchers adjusted the parameters of their simulation to make the material behave more like a continuous sheet rather than a grid of discrete points, they found that the energy barriers for the smooth, charge-preserving changes stayed roughly the same, while the barriers for the charge-changing transitions grew larger. This suggests that in the ideal, continuous limit, the distinction between these two types of paths becomes even more pronounced. Yet, the study also uncovered a subtle and important exception: sometimes, the path of least resistance does not follow a smooth, continuous transformation, even when one exists. In certain conditions, the system finds it energetically cheaper to break the magnetic field, change the topological charge, and then restore it, rather than to squeeze through a smooth, continuous deformation. This means that the most likely path a magnetic texture will take to change its shape is not always the one that looks the most continuous, challenging the intuitive assumption that nature always prefers the smoothest route.

The work provides a comprehensive map of the magnetic landscape, showing how different states are connected and how likely they are to switch under thermal influence. By identifying these specific transition mechanisms and their associated energy costs, the researchers have offered a way to predict how these materials will behave over time. The method they developed is not limited to the specific material they studied; it is a general tool that can be applied to any magnetic system to uncover hidden pathways between states. This capability is crucial for designing future technologies that rely on the stability and controllability of these magnetic textures. The study confirms that while the variety of magnetic shapes is vast, the ways they transform are governed by a small set of universal rules. By understanding these rules, scientists can better predict how to manipulate these tiny magnetic structures, potentially leading to more efficient and robust devices for data storage and processing in the years to come.

Technical Summary: Network of Localized Magnetic Textures Revealed Using a Saddle-Point Search Framework

Problem Statement
The field of topological magnetism has identified various localized magnetic textures (e.g., skyrmions, antiskyrmions, skyrmion bags) in 2D and 3D systems. These systems often possess multiple stable and metastable states, corresponding to local minima on a complex energy surface. Understanding the dynamics of these systems requires characterizing the transitions between these states, which are governed by first-order saddle points (SPs) on the energy landscape.

Existing saddle-point search algorithms for magnetic systems typically address the "double initial point problem," where both initial and final states are known, and the minimum energy path (MEP) is sought. However, for exploring the full energy surface of a system with unknown final states, the "single initial point problem" is more relevant. Current methods for this problem often rely on random displacements or following eigenmodes of the full system. These approaches can be inefficient or lack systematicity, particularly when targeting SPs associated with localized symmetry-breaking deformations (e.g., the nucleation of a kink in a specific region of a texture) or when the system is large. Furthermore, it remains an open question whether continuous transformations (homotopies) preserving topological charge always represent the lowest energy pathways compared to discontinuous transformations that change the topological charge.

Methodology: The Saddle-Point Search Framework (SPSF)
The authors present a computational framework designed for the systematic exploration of energy surfaces in magnetic systems. The framework automates the identification of first-order SPs connecting a given local energy minimum to adjacent states. It consists of four distinct stages:

Preprocessing:
- Texture Recognition: Isolated magnetic textures are extracted from the ferromagnetic background using $m_z=0$ contours and density-based clustering.
- Symmetry Analysis: The symmetry of each texture is identified using spin point groups ( $C_n$ and $D_n$ ). This allows for the determination of the "fundamental domain"—the smallest region containing unique physical information without repetition.
- Subsystem Definition: The fundamental domain is used to define various subsystems (elliptical regions) for excitation.
Escape Stage:
- Instead of random displacements, the system is driven out of the convex region of the initial minimum by exciting low-energy eigenmodes of the defined subsystems.
- A family of ~50 subsystems is sampled by varying the center, semi-axes, and orientation within the fundamental domain.
- Escape trajectories are generated by iteratively displacing magnetic moments in the subsystem along a selected eigenmode, followed by relaxing the rest of the system (L-BFGS).
- An escape is successful if the system reaches a configuration where the lowest Hessian eigenvalue ( $\lambda_1$ ) is negative and sufficiently aligned with the energy gradient, ensuring the system is in an ascending valley leading to an SP.
Convergence Stage:
- Configurations from the escape stage serve as initial points for the Geodesic Minimum Mode Following (GMMF) method to converge to first-order SPs.
- To improve efficiency, a subsystem-constrained GMMF (S-GMMF) is employed first, modifying only the subsystem while relaxing the rest. This is followed by a full-system GMMF refinement.
- Convergence is achieved when the force on magnetic moments is minimized, $\lambda_1 < 0$ , and $\lambda_2 \ge 0$ .
Postprocessing:
- Redundancy Filtering: Identified SPs are clustered based on feature vectors (energy, topological charge, Hessian eigenvalues) to remove duplicates arising from translation or rotation.
- Connectivity: Steepest-descent minimizations from the SP in both directions along the unstable mode determine the adjacent energy minima, constructing a network of states.

Key Contributions and Results
The framework was applied to a 2D chiral magnet described by a lattice Hamiltonian (Heisenberg exchange, Dzyaloshinskii-Moriya interaction, and Zeeman term). The study revealed a complex network of metastable textures and transition mechanisms:

Structural Elements: Metastable textures are composed of three fundamental building blocks: closed contours, chiral kinks, and tails.
Transition Categories: All identified SPs fall into five distinct categories based on the transformation of these elements:
1. Nucleation of a chiral kink (changes topological charge $Q$ by $+1$ ).
2. Formation of a closed contour via Bloch point-like defect injection ( $\Delta Q = \pm 1$ ).
3. Merging of two contours ( $\Delta Q = \pm 1$ ).
4. Formation of a tail (homotopy, $\Delta Q = 0$ ).
5. Merging of contours accompanied by chiral kink nucleation (homotopy, $\Delta Q = 0$ ).
Energy Barriers and Universality: The energy barriers for these mechanisms exhibit weak dependence on the specific host texture (e.g., skyrmion vs. droplet), suggesting a universal character for these SP categories.
Scaling Analysis: A scaling analysis varying the period of chiral modulations ( $L_D$ $L_{D}$ ) toward the continuum limit showed a clear distinction:
- Barriers for homotopic transitions (tail formation, specific contour merging) remain nearly independent of $L_D$ .
- Barriers for non-homotopic transitions (kink nucleation, contour creation) increase with $L_D$ .
Homotopy vs. MEP: The study investigated whether an MEP always exists that represents a homotopy between states of the same topological charge. The results show that while homotopies exist mathematically, they do not always correspond to the MEP. In certain parameter regimes (specific magnetic fields), the MEP involves a discontinuous transformation (changing $Q$ ) that has a lower energy barrier than the continuous homotopic path. Furthermore, even when a homotopic MEP exists, it does not necessarily represent the lowest energy barrier between states.

Significance
The paper claims that the SPSF provides a systematic, automated, and computationally efficient method for exploring the energy landscapes of magnetic systems with multiple localized textures. By utilizing symmetry-informed subsystem excitations, the method overcomes limitations of previous eigenmode-following strategies, particularly in targeting localized, symmetry-breaking transitions.

The findings establish a hierarchy of transition mechanisms in 2D chiral magnets, distinguishing between homotopic and non-homotopic pathways. Crucially, the work demonstrates that the lowest energy transition between two states of the same topological charge is not guaranteed to be a continuous homotopy; it may instead proceed through a discontinuous path involving the temporary creation and annihilation of topological charge. This insight challenges the assumption that topological charge conservation dictates the primary dynamical pathway in these systems. The framework is presented as a tool suitable for integration into global optimization schemes and long-timescale dynamics simulations (e.g., adaptive kinetic Monte Carlo) for magnetic materials.

Technical Summary: Network of Localized Magnetic Textures Revealed Using a Saddle-Point Search Framework

More like this