Using machine learning to create materials that enable energy-efficient electronics – USC Viterbi

Vortex polarization in ferroelectric materials

USC Viterbi researchers have created a new machine learning model to study how light polarizes materials like lead titanate into a vortex-like polarization pattern that dramatically improves the material’s energy-efficient properties.

Since prehistoric times, humans have experimented with altering the properties of materials to make them more useful and desirable; using techniques such as heating and stretching materials to convert them into a more practical form – such as melting and then rapidly cooling quartz to turn it into glass.

As science progressed, scientists investigated how to harness electricity, magnetic fields, and even light to control the properties of a material. In modern electronics, the use of light to control materials offers interesting prospects for the development of new generation devices that are more energy efficient and perform better. Because light interacts with material on extremely rapid timescales, it can discover properties of material that heating or stretching cannot. These properties are often referred to as hidden “quantum phases” due to the interaction of light with the material at the level of quantum or atomic mechanics. However, this makes modeling extremely difficult, hampering our ability to research and design next-generation materials.

Materials researchers at the USC Viterbi School of Engineering have developed a new machine learning framework to study on an unprecedented scale how light can control materials. Typical simulations for understanding the light control of materials can usually only simulate a few hundred atoms, even with state-of-the-art computing resources, which seriously limits their applications. By harnessing the power of machine learning, USC Viterbi researchers were able to perform light-control simulations of materials containing more than a billion atoms, 10 million times more than conventional methods.

Researcher Thomas Linker said the small blue box is the size of the system that can currently be simulated using standard quantum simulation methods, while the entire box is roughly the scale of what can be simulated with the research team's neural network.

Researcher Thomas Linker said the small blue box is the size of the system that can currently be simulated using standard quantum simulation methods, while the entire box is roughly the scale of what can be simulated with the research team’s neural network.

The research team used their machine learning model to perform a large-scale simulation of the light control of lead titanate, a special class of material, called ferroelectric material, which has inherent electronic polarization. Polarization can be thought of as a pattern of arrows in the material that can be controlled by stretching, heating, and electricity, making the material ideal for use in sensors, energy storage, and memory. The research, from Associate Professor of Chemical Engineering and Materials Science Ken-ichi Nomura, Professor of Chemical Engineering and Materials Science Priya Vashishta, Professor of Computer Science Aiichiro Nakano, Professor of Physics Rajiv Kalia and Ph.D. student Thomas Linker and their co-authors was recently published in Scientific advances.

The researchers examined how the electronic polarization of the lead titanate material would react to light. Recently, this material has gained popularity because it allows researchers to create intricate vortex-like patterns in its electronic polarization. When we think of a vortex, we may imagine a chaotic, swirling mass of matter or energy; however, these types of structures have proven to be very stable and efficient in these materials, which is why they are currently being investigated for next-generation memory and energy storage applications. The USC Viterbi researchers wanted to understand if these complex patterns could be controlled by light.

“We wanted to see these large-scale structures with very precise simulation methods that use things like quantum mechanics,” Linker said. “But it’s really difficult and very expensive, so we developed a multi-scale framework where we train a machine learning model to learn a simpler representation of the light-matter interaction. So we can simulate a lot faster.

“Without machine learning, it would have been impossible to design this type of simulation,” Nomura said. “By training the machine learning model to learn how the material behaves in response to a strong laser, we can run our simulation on supercomputers.”

With their framework, the researchers found a new type of phase induced by light-matter interaction in lead titanate. “If we shine the laser (light), we can create a warp pattern in the polarization that is topologically different from the original vortex pattern,” Nomura said.

Overall, the research team said their machine learning framework offers an exciting new avenue for exploring light control of materials that was not possible before.

Machine learning is increasingly becoming an essential tool in the development of new high-performance quantum materials. The Mork family’s Department of Chemical Engineering and Materials Science recently launched a master’s program to train graduates in harnessing machine learning to create advanced materials.

Posted on March 28, 2022

Last updated March 28, 2022

Leave a Comment