Metallic Conductivity Screening: ML-Driven Study of 2.8 Million Compounds

npj Computational Materials

Overview

Researchers have developed a computational framework combining machine learning with ab initio calculations to screen over 2.8 million compounds for high electrical conductivity. The study identified candidates comparable to aluminum (36.59 × 10^6 S/m) and highlighted LiBePt2 as a high-performance conductor. This approach enables the rational design of metallic materials by bypassing the computational costs of traditional density functional theory.

In Depth

Key Findings

The study identified several metallic candidates with predicted electrical conductivities comparable to aluminum, which is 36.59 × 10^6 S/m. A notable finding is that compounds such as LiBePt2 can achieve high performance by utilizing valence electrons from light elements to shift high-scattering d-states beneath the Fermi level. While noble metals (Ag, Au, Cu) remain difficult to surpass due to low scattering and unique electronic structures, the framework successfully identified novel high-performance conductors.

Technical Details

The researchers implemented a multi-stage workflow integrating machine learning (ML) with high-throughput ab initio calculations. The technical process includes:

• Computational Framework: Integration of density functional theory (DFT) and the Boltzmann transport equation (BTE), specifically using an iterative solution (iBTE).
• Descriptors: The use of the Eliashberg parameter λ, originally used in superconductivity studies, as an additional descriptor to accelerate screening.
• Validation: Full electron–phonon coupling calculations were performed for top-performing materials. The calculated conductivities showed relative errors below 30% for the majority of systems when compared to experimental data.

Background & Context

Predicting electrical conductivity is computationally intensive because traditional DFT and BTE methods require detailed calculations of scattering rates and electronic band structures. While databases like the Materials Project, AFLOW, OQMD, and Alexandria have cataloged thousands of materials, transport properties have remained underexplored due to these complexities. More rigorous formulations, such as those based on the Kubo formula, exist but incur substantially higher computational costs.

Significance & Outlook

By using ML to learn structure-property relationships, the researchers reduced the computational cost of property predictions by several orders of magnitude. This allows for the exploration of vast chemical spaces that were previously prohibitively expensive. The ability to optimize transport properties from first principles supports the rational design of materials for microelectronics, power transmission, energy storage, and quantum technologies.

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.

• Your email and selected fields are used only to deliver the newsletter.
• We never share your information with third parties.
• You can unsubscribe anytime via the link in each email.

See our Privacy Policy for details.

Agree & Continue ▶

Takes about a minute · Unsubscribe anytime