By pouring through datasets of complex computer simulations, an artificial intelligence (AI) model was trained to predict how electronic devices, such as transistors and microchips, will fail.
A team of engineers from the Colorado University Boulder is behind the latest breakthrough, led by physicist and aerospace engineer Sanghamitra Neogi. They present their findings in the article "First-principles prediction of electronic transport in fabricated semiconductor heterostructures via physics-aware machine learning," appearing in the latest NPJ Computational Materials journal.
Monitoring the Behavior of Electronics at the Atomic Level
In the CU Boulder researchers, the engineers mapped out the physics that happens between atoms in electronic devices. They then used a variety of machine learning techniques to have an idea of how larger structures, based on these small atoms that are the building blocks of the electronic devices, will behave. Researchers likened it to examining the strength of a single Lego brick and predicting the strength of an entire castle made from these bricks.
"We're trying to understand the physics of devices with billions of atoms," shares Neogi, who serves as an assistant professor in CU Boulder's Ann and H.J. Smead Department of Aerospace Engineering Sciences, in a news release from the university.
The nature of their AI-powered study has significant implications, especially in the field of electronics manufacturing - the results of which are integral parts of daily life, from smartphones to the latest computers. Neogi additionally notes that engineers could use the methods they developed to identify the weaknesses inherent in electronic devices, even before they get to manufacturing them.
According to the CU Boulder article, the current study is a part of Neogi's larger efforts to understand how the world at the atomic scale could help us build new and more efficient computers, including one patterned from how the human brain works. She also noted that instead of waiting years trying to understand how these electronic devices fail, their novel technique using an AI model could provide knowledge in advance.
Using an AI Model to Project Potential Failure
To create their predictive AI that can foresee failure in electronic devices, researchers developed a computer model that is driven by AI to understand the physical properties of the most common semiconductor materials - silicon and germanium atoms. These two atoms are found in most electronic devices. Specifically, researchers trained their AI model to understand how their subatomic particles and the atoms themselves combine together in forming the semiconductor devices required. The AI model can now extrapolate physical and electrical behavior from these small blocks of atoms and use it to estimate energy distribution in larger groups of these atoms.
"It collects information from each individual unit and combines them to predict the final properties of the collective system, which can be made up of two, three, or more units," Neogi additionally explains.
While the UC Boulder team admits to still having a long way to go before it accurately predicts the weak points in electronic devices the size of an average smartphone, they have reported success in identifying potential failure conditions in a number of silicon and germanium based materials.
Check out more news and information on Electronics in Science Times.
