Machine Learning System will improve electrical transmission behaviours in semiconductors

Semiconductor refers to a class of crystalline solids, whose electricity conductive behaviour lies in the middle range of conductors and insulators. Semiconductors are used in the development of electronic devices like transistors, and integrated circuits. 

 

Germanium and silicon are semiconductors that are used in the development of transistors. Semiconductors are poor conductors of electricity at room temperature. Electricity transmission properties of semiconductors can be altered through the adjustment of strain level. 

 

Researchers at University of Skoltech in collaboration with U.S. partners have developed a machine learning system that will use convolutional neural networks to  make small modifications in the semiconductor crystals for improving electricity transmission properties. 

 

(Must Check: 5 Common Architectures in Convolution Neural Networks (CNN))

 

The strain tensor is fed in the neural network. Neural networks will predict electronic band structure of materials- physical photographs classify electronic band properties of stressed materials. This network based system will be used in the assessment of bandgap, and  its distinct properties.  Further neural networks are trained using computationally expensive data that is derived from GW based calculations. 

 

Combination of multiple data sources is used for proper training of networks. Active learning will enable the model to categorize beneficial data that will  be used in next stage data training. Convolutional neural networks will improve the performance of Elastic Strain Engineering (ESE). Elastic strain engineering is a new branch that focuses on improvement of semiconductor performance. 

 

This new neural network is accurate and efficient for modification of semiconductors. It simplifies search and optimization operations as a way to achieve best strain values for given merit figures. Nanomaterial properties can be changed by proper adjustment of strain.

According to MARKTECHPOST The Skoltech scientists focused on the computational and machine learning aspect.  They will now be working on the boundaries of admissible elastic strains. It’s a crucial topic because the theoretical limits of safe elastic deformation for ESE are yet unknown and to be discovered.