Accelerating TEM sample preparation by machine learning
On April 20, 2021, Shanghai – Thermo Fisher Technology (hereinafter referred to as: Thermo Fisher), the world leader in the field of science services, announced the launch of Helios 5 EXL wafer double beam transmission electron microscope, aiming to meet the increasing sample size and corresponding analysis needs of semiconductor manufacturers with large-scale operation. With its machine learning and advanced automation capabilities, this product can provide accurate sample preparation to support below 5 nm node technology and full surround gate semiconductor process as well as yield improvement.
With the development of semiconductor manufacturing process towards smaller and more complex direction, semiconductor manufacturers need more reproducible and mass transmission electron microscope (TEM) analysis results. The increasing demand for atomic level data brings scalability challenges for busy laboratories to use advanced equipment to achieve ideal results. With the development of full surround gate technology, there are more requirements for interfaces, thin films and measurable cross sections with resolution below nanometer level, which makes it more difficult to upgrade mass TEM analysis.
Through machine learning and closed-loop feedback for pole assignment, Helios 5 EXL can provide accurate cutting, so that the operator can always maintain high-quality output in difficult sample preparation. Compared with existing solutions, the improved automation technology optimizes the ratio of machine to manual operation, aiming at maximizing sample output and technical resource productivity.
“Semiconductor labs are under tremendous pressure to provide TEM analysis data more quickly to support process monitoring and improve the learning curve without increasing costs,” said Glyn Davies, vice president of semiconductor division of semefield. “Helios 5 EXL can meet this challenge through scalable, reproducible and high-precision TEM sample preparation.”
Helios 5 EXL can help semiconductor manufacturers extract more data from each wafer than existing solutions in order to improve the success rate of TEM analysis.