On-demand growth of semiconductor heterostructures guided by physics-informed machine learning

Area of Science:

• Semiconductor Science and Engineering
• Materials Science
• Artificial Intelligence in Materials Synthesis

Background:

• Traditional semiconductor heterostructure fabrication relies on simulation-driven design and iterative optimization, limiting efficiency.
• Developing tailored heterostructures on demand is critical for advancing semiconductor devices.

Purpose of the Study:

• Introduce SemiEpi, a self-driving platform for automated molecular beam epitaxy (MBE) of semiconductor heterostructures.
• Enable precise control over multi-step material growth using real-time feedback.

Main Methods:

• Integration of MBE reactors with physics-informed machine learning (ML) models and in situ Reflection High Energy Electron Diffraction (RHEED) monitoring.
• Development of an on-the-fly feedback control system for optimizing growth parameters.
• Automated design of heterostructures and identification of optimal initial growth conditions.

Main Results:

• Demonstrated optimization of high-density Indium Arsenide (InAs) quantum dot growth for a 1240 nm emission wavelength.
• Achieved a quantum dot density of 5 × 10^10 cm^-2, a 1.6-fold increase in photoluminescence intensity.
• Reduced the full width at half maximum to 29.13 meV through feedback-controlled growth temperatures.

Conclusions:

• SemiEpi enhances process repeatability and stability in semiconductor heterostructure growth.
• The platform's versatility across different MBE reactors highlights its potential for hardware-independent frameworks.
• SemiEpi addresses key challenges in multi-step heterostructure fabrication, paving the way for accelerated materials discovery.