Abstract:Spatial imaging of magnetic stray fields from magnetic materials is a useful tool for identifying the underlying magnetic configurations of the material. However, transforming the magnetic image into a magnetization image is an ill-poised problem, which can result in artefacts that limit the inferences that can be made on the material under investigation. In this work, we develop a neural network fitting approach that approximates this transformation, reducing these artefacts. Additionally, we demonstrate that this approach allows the inclusion of additional models and bounds that are not possible with traditional reconstruction methods. These advantages allow for the reconstruction of non-trivial magnetization textures with varying magnetization directions in thin-film magnets, which was not possible previously. We demonstrate this new capability by performing magnetization reconstructions on a variety of topological spin textures.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: reconstructing complex, non - trivial magnetization textures from magnetic field images of magnetic materials, especially those textures with different magnetization directions within a single image. Traditional methods, such as Fourier - transform - based methods, encounter serious problems of artifacts and noise amplification when dealing with these complex textures, resulting in the inability to accurately reconstruct magnetization information. Specifically, the main objectives of the paper include: 1. **Reducing artifacts**: Approximate the inverse transformation from magnetic field to magnetization by the neural network (NN) method to reduce artifacts in traditional methods. 2. **Handling complex magnetization textures**: Develop a new method capable of handling non - trivial magnetization textures in thin magnetic films with varying magnetization directions. 3. **Introducing physical constraints**: Utilize neural networks to allow the introduction of additional models and boundary conditions, which are difficult to achieve in traditional methods. 4. **Improving reconstruction accuracy**: Improve the robustness and accuracy of magnetization estimation by using the complete magnetic field vector for reconstruction. 5. **Distinguishing topological spin structures**: Demonstrate how to distinguish different types of topological spin structures, such as Bloch - type and Néel - type Skyrmions, through the neural network reconstruction method, and identify the chirality of Bloch Skyrmions. ### Main challenges - **Ill - posedness of the inverse problem**: Reconstructing a magnetization image from a magnetic field image is an ill - posed problem because there are multiple possible solutions. - **Complex magnetization directions**: Traditional methods can only handle the case of a single magnetization direction and cannot handle complex magnetization textures with multiple directions. - **Artifacts and noise**: Traditional methods are prone to producing artifacts and amplifying noise, affecting the reconstruction quality. ### Solutions - **Neural network reconstruction method**: Use neural networks to approximate the inverse transformation, combined with physical constraints and multi - component fitting, to improve the accuracy and robustness of reconstruction. - **Introducing physical constraints**: Limit the magnetization distribution through weighted masks and other physical constraints to reduce artifacts. - **Multi - component fitting**: Simultaneously use three magnetic field components (Bx, By, Bz) for reconstruction to better separate magnetization components and suppress noise. Through these methods, the paper demonstrates the superior performance of the neural network reconstruction method in handling complex magnetization textures and provides new tools and ideas for future magnetism research.

Reconstruction of non-trivial magnetization textures from magnetic field images using neural networks

Robustly reconstructing magnetic resonance images via structure decomposition.

MagNet: machine learning enhanced three-dimensional magnetic reconstruction

Universal Reconstruction of Complex Magnetic Profiles with Minimum Prior Assumptions

AI-enabled Lorentz microscopy for quantitative imaging of nanoscale magnetic spin textures

Three-dimensional Magnetization Reconstruction from Electron Optical Phase Images with Physical Constraints

Unraveling Nanostructured Spin Textures in Bulk Magnets

Advancements in neural network techniques for electric and magnetic field reconstruction: Application to ion radiography

Quantifying the magnetic interactions governing chiral spin textures using deep neural networks

Holographic vector field electron tomography of three-dimensional nanomagnets

Super-resolution of magnetic systems using deep learning

Machine Learning for Improved Current Density Reconstruction from 2D Vector Magnetic Images

Magnetic Hamiltonian Parameter Estimation Using Deep Learning Techniques

Implication of Localization of Human DNA Repair Enzyme O6-Methylguanine-DNA Methyltransferase at Active Transcription Sites in Transcription-Repair Coupling of the Mutagenic O6-Methylguanine Lesion

Machine learning based classification of vector field configurations

Application of a Generative Adversarial Network in Image Reconstruction of Magnetic Induction Tomography

Image-based prediction and optimization of hysteresis properties of nanocrystalline permanent magnets using deep learning

Deep Magnetic Resonance Image Reconstruction: Inverse Problems Meet Neural Networks

Quantification of Hybrid Topological Spin Textures and Their Nanoscale Fluctuations in Ferrimagnets.