Applying Machine Learning to Elucidate Ultrafast Demagnetization Dynamics in Ni and Ni80Fe20

Aug 4, 2026

Upload by Liam

Understanding the correlation between fast and ultrafast demagnetization processes is crucial for revealing the microscopic mechanisms driving ultrafast demagnetization, a key aspect of spintronics. Machine learning models were trained on limited experimental data using supervised regression and symbolic regression to relate demagnetization time (τM) to the Gilbert damping factor (α). Polynomial regression and K-nearest neighbors best predicted τM. SISSO suggested a direct τM–α correlation for Ni and Ni80Fe20, indicating dominant spin-flip scattering. The models were validated on independent experimental results, while comparisons highlighted the strong role of material properties in ultrafast demagnetization behavior.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Spintronics and ultrafast magnetization dynamics

Theoretical modeling: LLG and three-temperature model

Correlating Gilbert damping (α) with demagnetization time (τM)

Proposed proportional vs inverse dependencies

Applying Machine Learning to Elucidate Ultrafast Demagnetization Dynamics in Ni and Ni80Fe20 Hasan Ahmadian Baghbaderani*, Byoung-Chul Choi Department of Physics and Astronomy, University of Victoria, Victoria, BC V8W 3P6, Canada Abstract Understanding the correlation between fast and ultrafast demagnetization processes is crucial for elucidating the microscopic mechanisms underlying ultrafast demagnetization, which is pivotal for various applications in spintronics. Initial theoretical models attempted to establish this correlation but faced challenges due to the complex interplay of physical phenomena. To address this, we employed a variety of machine learning methods, including supervised learning regression algorithms and symbolic regression, to analyze limited experimental data and derive meaningful mathematical expressions between demagnetization time (τM) and the Gilbert damping factor (α) . The results reveal that polynomial regression and K-nearest neighbors algorithms perform best in predicting τM. Additionally, sure-independence-screening-and-sparsifying-operator (SISSO) as a symbolic regression method suggested a direct correlation between τM and α for Ni and Ni80Fe20, indicating spin-flip scattering predominantly influences the ultrafast demagnetization mechanism. The developed models demonstrate promising predictive capabilities, validated against independent experimental data. Comparative analysis between different materials underscores the significant impact of material properties on ultrafast demagnetization behavior. This study underscores the potential of machine learning in unraveling complex physical phenomena and offers valuable insights for future research in ultrafast magnetism. 1. Introduction Combining spin and charge properties, spintronics offers diverse functionalities essential for industrial applications such as sensing and memory storage [1], [2], while also showing promise in communication and information processing [3], [4] . Materials spin textures can be dynamically excited by various physical parameters: magnetic fields, electrical currents, temperature, and pressure, all of which give rise to different responses at different time scales, ranging from nanoseconds (fast demagnetization) to picoseconds (ultrafast demagnetization) [5], [6] . Recent research has intensified the focus on understanding the rapid dynamics of magnetization [7] . However, the mechanisms underlying ultrafast demagnetization dynamics remain poorly understood. Different theoretical models describe fast and ultrafast magnetization dynamics. For instance, the fast magnetization dynamics can be described through the Landau–Lifshitz–Gilbert (LLG) equation (Eq. 1) [8]: 􀝀􀜯 (􀝎, 􀝐) 1 􀝀􀜯 (􀝎, 􀝐) (1) 􀝀􀝐 = −􀟛 (􀜯 (􀝎, 􀝐) × 􀜪􀯘􀯙􀯙 (􀝎, 􀝐)) + |􀜯 (􀝎, 􀝐) | (􀜯 (􀝎, 􀝐) × 􀟙 􀝀􀝐 ) * Corresponding author E-mail [address: ](address: baghbaderani@uvic.ca)[baghbaderani@uvic.ca](address: baghbaderani@uvic.ca) where 􀜯(􀝎, 􀝐) is the magnetization vector at position 􀝎 and time 􀝐, and 􀟛 is the gyromagnetic ratio that describes the precession of the magnetization around the effective magnetic field, 􀜪􀯘􀯙􀯙 (􀝎, 􀝐) . This effective field is composed of the external field, the magnetic anisotropy field, and the demagnetization field. The Gilbert damping factor, 􀟙 , quantifies the rate at which the magnetization vector relaxes towards the direction of the effective magnetic field. In other words, the first term on the right-hand side represents precession which includes indirect damping, in which 􀟛 describes the precession of the magnetization around the micromagnetic effective field,􀜪􀯘􀯙􀯙 (􀝎, 􀝐) . The second term in Eq. 1 is the damping term (represented by 􀟙) which drives the system toward the direction of 􀜪􀯘􀯙􀯙 [9] . On the other hand, ultrafast dynamics following laser excitation are typically modelled using a phenomenological three-temperature model [10], which accounts for electron, spin, and lattice interactions. In this reg

FAQs

What correlation does the study aim to uncover in Ni and Ni80Fe20?

The study correlates ultrafast demagnetization time (τM) with the Gilbert damping factor (α) to identify the underlying microscopic mechanism.

Which machine learning methods performed best for predicting τM?

Polynomial regression and K-nearest neighbors showed the best predictive performance for τM.

What did the symbolic regression method SISSO indicate about the mechanism?

SISSO suggested a direct τM–α correlation for Ni and Ni80Fe20, pointing to spin-flip scattering as the dominant contributor to ultrafast demagnetization.

Aug 4, 2026

Upload by Liam