Data Management and Hybrid Process Modeling for Microelectronics (DaHyME - [dəˈhaɪ̯m])

Hybrid AI process models make microelectronics manufacturing more efficient, robust, and sustainable. We link domain-specific process knowledge from simulations and established empirical models with data-driven machine learning methods – optimized for real factory environments with small, heterogeneous datasets. The goal is to reduce energy and resource usage, shorten development times, and strategically strengthen innovation capacity in Saxony.

Two use cases underpin the project:

• Optimization of plasma etching processes, a core step in chip manufacturing, with measurable savings in energy, climate-harmful process gases, and pilot and control wafers.
• Accelerated development of nanomaterial-based field-effect transistors (e.g., CNT FETs) beyond CMOS for applications such as PUFs, sensing, RF components, 3D logic, and photonic systems.

Technically, we are building a microelectronics-specific data infrastructure, linking process and metadata, and combining physics-based multiscale models with AI methods. A demonstrator at Fraunhofer ENAS and transfer workshops make the results directly tangible for industry partners in Saxony.