New simulation method sharpens our view into Earth’s interior

Earth’s magnetic field is vital to sustaining life because it protects the planet from cosmic radiation and solar wind. It is produced by the geodynamo effect. “We know that the Earth’s core is mainly composed of iron,” explains Attila Cangi, head of the Materials Design Machine Learning Department at CASUS. “As you approach the Earth’s core, both temperature and pressure increase. An increase in temperature causes materials to melt, while an increase in pressure causes them to remain solid. Because of the specific temperature and pressure conditions inside the Earth, the outer core is in a molten state, while The inner core remains solid. Electrically charged liquid iron flows around the solid inner core, driven by the Earth’s rotation and convection currents, which generate electrical currents that create the Earth’s magnetic field.

However, important questions about the Earth’s core remain unanswered. For example, what is its core specific structure? What role do other elements (thought to exist along with iron) play? Both factors may profoundly affect the geodynamo effect. The clue came from experiments in which scientists sent seismic waves through the Earth and measured their “echoes” with highly sensitive sensors. “These experiments show that there is more than just iron in the Earth’s core,” said the study’s lead author Svetoslav Nikolov of Sandia National Laboratories. “The measurements are consistent with the assumption of a pure iron core. Computer simulations are inconsistent.”

Simulate shock waves on computer

The research team has now made significant progress by developing and testing new simulation methods. The key innovation of the molecular spin dynamics approach is the integration of two previously independent simulation methods: molecular dynamics, which models atomic motion, and spin dynamics, which explains magnetic properties. CEA physicist Julien Tranchida emphasizes: “By combining these two methods, we were able to study the influence of magnetism on length and time scales under high-pressure and high-temperature conditions.” Specifically, the team simulated 2 million iron atoms and their natural spin behavior to analyze dynamic interactions between mechanics and magnetism. The researchers also employed artificial intelligence (AI), using machine learning to determine force fields (the interactions between atoms) with high accuracy. Developing and training these models requires high-performance computing resources.

Once the model was ready, the researchers carried out practical simulations: a digital model of 2 million iron atoms representing the Earth’s core was affected by the temperature and pressure conditions inside the Earth. This is done by propagating pressure waves through the iron atoms, simulating their heating and compression. When the speed of these shock waves is lower, the iron remains solid and adopts a different crystal structure. When the shock wave is faster, the iron becomes mostly liquid. In particular, the researchers found that magnetic effects can significantly affect the material’s properties. “Our simulations are in good agreement with experimental data, and they show that under certain temperature and pressure conditions, specific phases of iron may be stable and potentially impact Earth’s evolution,” said Mitchell Wood, a materials scientist at Sandia National Laboratories. Motor. This phase, known as the bcc phase, has not been experimentally observed in iron under these conditions and is merely a hypothesis. If confirmed, the results of the molecular spin dynamics method will help resolve the geodynamo effect. a question.

Promoting energy-saving artificial intelligence

In addition to revealing new details about Earth’s interior, the method has the potential to drive innovation in materials science and technology. Cangi plans to use the technology within his department and in external collaborations to simulate neuromorphic computing devices. It’s a new type of hardware inspired by the way the human brain works, and could one day process artificial intelligence algorithms faster and more energy-efficiently. By digitally replicating spin-based neuromorphic systems, new analogies can support the development of innovative and efficient machine learning hardware solutions.

Data storage offers a second compelling avenue for further research: magnetic domains along tiny nanowires could serve as a storage medium that is faster and more energy-efficient than conventional technologies. “Currently there are no accurate simulation methods for either application,” Kanji said. “But I believe that our new method is able to model the required physical processes in such a realistic way that we can significantly accelerate the technological development of these IT innovations.”