Machine-Learning-Assisted Development and Theoretical Consideration for the Al₂Fe₃Si₃ Thermoelectric Material

Area of Science:

• Materials Science
• Thermoelectrics
• Computational Materials Science

Background:

• Chemical composition tuning is key to optimizing thermoelectric materials for waste heat to electricity conversion.
• The Al2Fe3Si3 intermetallic compound shows promise for low-cost, non-toxic thermoelectric devices due to its high power factor (~700 μW m⁻¹ K⁻² at 400 K).

Purpose of the Study:

• To accelerate the exploration of Al2Fe3Si3 thermoelectric properties in the mid-temperature range.
• To enhance the power factor of Al2Fe3Si3 by optimizing the Al/Si ratio using a machine-learning approach.
• To investigate the underlying mechanisms for enhanced thermoelectric performance and thermal conductivity.

Main Methods:

• Employed a machine-learning method to guide the synthesis of off-stoichiometric Al23.5+xFe36.5Si40-x samples.
• Tuned the Aluminum (Al)/Silicon (Si) ratio to identify optimal compositions for high power factor.
• Utilized ab initio density functional theory (DFT) for phonon dispersion calculations and Slack model for thermal conductivity estimation.

Main Results:

• Identified an optimal Al/Si ratio (x = 0.9) that increased the power factor by approximately 40% at ~510 K compared to the stoichiometric sample (x = 0.0).
• Observed precipitations of metallic secondary phases in off-stoichiometric samples, potentially explaining the enhanced power factor.
• Estimated a maximum achievable thermal conductivity of ~10 W m⁻¹ K⁻¹ for Al2Fe3Si3 and found evidence suggesting low thermal conductivity due to avoided-crossing behavior in phonon dispersion.

Conclusions:

• Machine learning efficiently identified an optimal Al/Si ratio for enhancing the mid-temperature power factor of Al2Fe3Si3.
• The study provides insights into the role of secondary phase precipitation and defect structures (Al-Si antisite defects) in tuning thermoelectric properties.
• Al2Fe3Si3 demonstrates potential for low thermal conductivity, making it a promising candidate for efficient thermoelectric applications.