Semiconductor Crystals Tweaked to Realize Superior Properties for Electronics

Scientists from Skoltech, together with their collaborators in the United States and Singapore, have developed a neural network that enables the tweaking of semiconductor crystals in a controlled way to achieve excellent properties for electronics.

Semiconductor Crystals Tweaked to Realize Superior Properties for Electronics

Image Credit: imred

This facilitates a new way of developing next-generation solar cells and chips by leveraging a controllable deformation that could potentially alter the properties of a material on the go. The study was published in the npj Computational Materials journal.

At the nanoscale level, materials are capable of resisting major deformation. In the so-called strained state, they show significant electronic, thermal, optical and other characteristics as a result of variations in the interatomic distances. The inherent properties of a strained material may vary, with the semiconducting silicon, for example, changing into a material that freely conducts the electric current.

By altering the strain level, it is possible to change the properties as required. This concept has led to a whole field of inquiry: elastic strain engineering (ESE). For example, this method can also be utilized to alter the performance of semiconductors, thereby offering a potential workaround for the impending Moore’s law limit, when other options for increasing chip performance are exhausted.

One more potential use is in the field of solar cell development. According to Alexander Shapeev, a study co-author from Skoltech, a solar cell can be designed with tunable properties that can be altered on demand to optimize performance and adapt to external circumstances.

In earlier research, Skoltech PhD graduate Evgenii Tsymbalov, Associate Professor Alexander Shapeev, and their collaborators exploited ESE to convert nanoscale diamond needles from insulating to highly conductive and metal-like substances. Thus, they offered insights into the range of prospective applications of this technology. Currently, the team has come up with a convolutional neural network architecture that can guide ESE measures for semiconductors.

The neural network we have designed takes the strain tensor as an input and predicts the electronic band structure — a physical ‘snapshot’ that describes the electronic properties of a strained material. It may then be used to calculate any properties of interest, including the bandgap, its properties, and electron effective mass tensor.

Alexander Shapeev, Associate Professor, Skolkovo Institute of Science and Technology

The project continues the previous study and expands on it.

We go beyond the previously used approaches by designing and implementing a tailored model based on the convolutional neural network architecture, for the ESE task. We also take the physical properties and symmetries into account to improve the model.

Evgenii Tsymbalov, PhD, Skolkovo Institute of Science and Technology

This method incorporates different data sources, for instance, the computationally economical yet incorrect with the accurate but costly ones to boost the convergence and accuracy of the model.

Another distinct feature is active learning — we allow the model to guess what data may be the most useful to obtain in the next training stage, and use it for training. In the final stage, the network is trained on a set of computationally expensive data from the very accurate GW-based calculations, and this procedure allows us to reduce the number of computations needed.

Evgenii Tsymbalov, PhD, Skolkovo Institute of Science and Technology

The researchers explained that their new neural network is “more versatile, accurate, and efficient in its capacity to facilitate autonomous deep learning of the electronic band structure of crystalline solids” compared to the latest advanced solutions. This increases the speed and enhances accuracy at search and optimization inside the strain space, which results in the optimal strain values for given figures of merit.

In the previous study, the team assessed an earlier iteration of the model in the scenario of a repeating in situ experiment performed with diamond.

Alas, for now, there is no device that can deform the diamond with an arbitrary 6D deformation tensor, yet there are teams and labs pursuing this direction from the experimental point of view,” stated Tsymbalov.

The research is part of a year-long association between Skoltech, the Massachusetts Institute of Technology and Nanyang Technological University, where the Skoltech scientists focused on the computational and machine learning aspect and their collaborators focused on the physical component of the study.

The researchers concluded, “We are currently working on our next paper, which is devoted to the boundaries of admissible elastic strains. It is an important topic since the theoretical limits of safe elastic deformation for ESE are yet to be discovered.”

Journal Reference:

Tsymbalov, E., et al. (2021) Machine learning for deep elastic strain engineering of semiconductor electronic band structure and effective mass. npj Computational Materials.


Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type