When AI Starts Seeing Inside Atoms: What I Learned About Deep Learning From Nuclear Physicists

2026-03-20 · Source: Deep Learning on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Physical Sciences & Chemistry · Depth: Advanced, quick

Summary

Researchers from Jilin University have developed a physics-informed deep neural network capable of predicting charge distribution within atomic nuclei with unprecedented accuracy. This model surpasses traditional theoretical frameworks, such as density functional theory and the Relativistic Continuum Hartree–Bogoliubov (RCHB) framework, which often struggle with predicting properties across a wide range of nuclei. The deep learning approach was trained on experimental data from over 1,000 nuclei, using fundamental nuclear parameters like proton and neutron numbers, proximity to "magic numbers," and nucleon pairing effects as inputs. The network outputs 17 Fourier–Bessel coefficients, which precisely describe the internal charge density distribution.

Key takeaway

For AI Researchers exploring novel applications, this work demonstrates deep learning's capability to solve long-standing problems in fundamental physics. You should consider how physics-informed neural networks, trained on real experimental data, can overcome limitations of purely theoretical models in your domain, potentially leading to more accurate predictions in complex systems.

Key insights

Deep learning can accurately model complex quantum phenomena like atomic nuclear charge distribution.

Principles

• Physics-informed neural networks enhance accuracy.
• Experimental data improves theoretical model limitations.

Method

A deep neural network takes nuclear parameters (proton/neutron count, magic numbers, pairing) and outputs 17 Fourier–Bessel coefficients to describe charge density distribution.

In practice

• Integrate experimental data into theoretical models.
• Apply deep learning to complex physical systems.

Topics

• Physics-informed Deep Learning
• Nuclear Physics
• Atomic Nuclei
• Charge Density Distribution

Best for: AI Researcher, AI Scientist, Research Scientist

Related on AIssential

• When Deep Learning Meets Physics — Integrating physical laws into neural networks is a crucial, under-recognized advancement for AI's real-world utility.
• AI just discovered new physics in the fourth state of matter — AI can discover new physical laws and challenge long-standing assumptions in complex systems.
• Trainable Photonic Measurement for Physics-Informed PDE Learning — Trainable photonic measurement offers a robust, parameter-efficient representation for physics-informed PDE learning, excelling in complex phase regimes.
• Physics-informed convolutional neural networks for fluid flow through porous media — Physics-informed CNNs accurately predict fluid flow in porous media, significantly accelerating traditional LBM simulations.
• Accelerate Clean, Modular, Nuclear Reactor Design with AI Physics — AI-augmented simulations with surrogate models significantly accelerate nuclear reactor design by overcoming computational bottlenecks.
• A better way to model the behavior of metal alloys — MIT researchers developed a motif-based sampling method using information theory to create diverse training datasets for accurate ML material simulations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Deep Learning on Medium.