Data-driven prediction of complex crystal structures of dense lithium

Area of Science:

• Materials Science
• Condensed Matter Physics
• Computational Chemistry

Background:

• Lithium (Li) exhibits significant structural and electronic property changes under compression.
• The precise crystalline structures of dense Li, especially near the melting minimum, remain debated.
• Experimental findings suggest the existence of unidentified crystalline phases.

Purpose of the Study:

• To investigate the energy landscape of lithium under compression.
• To identify and predict stable or metastable crystalline structures of dense lithium.
• To resolve the debate surrounding the structure of dense Li and explain experimental observations.

Main Methods:

• Employed an advanced crystal structure search method.
• Integrated a machine-learning approach to accelerate the search.
• Performed extensive exploration of the energy landscape for lithium.

Main Results:

• Predicted four novel, complex lithium crystal structures.
• These structures contain up to 192 atoms per unit cell.
• The predicted structures are energetically competitive with known Li phases.

Conclusions:

• The predicted structures offer a potential explanation for the experimentally observed, yet unidentified, crystalline phases of lithium.
• This study demonstrates the efficacy of global structure search methods combined with machine learning for discovering complex crystal structures.
• Highlights the predictive power of computational approaches in materials science.