Neuromorphic Computing

Increasing digitization is constantly driving the demands on electronic hardware. Speed, performance, miniaturization and energy efficiency are becoming increasingly important when it comes to enabling Big Data and Artificial Intelligence (AI) applications.

A promising solution approach is offered by so-called neuromorphic computing, which aims to emulate the self-organizing and self-learning nature of the brain. Fraunhofer IPMS develops materials, technologies and complete hardware solutions with high energy efficiency, especially for edge applications.

The technological developments are pursued in different stages of expansion. The so-called "deep neural networks" (DNN) have already arrived in the application with the help of classical technologies (e.g. SRAM or flash-based) and initially emulate the parallelism and efficiency of the brain. Further miniaturization and reduction of power consumption for edge applications is possible using new, innovative technologies. The subsequent generation of so-called "Spiking Neural Networks" (SNN) attempts to additionally physically replicate the temporal component of the functionality of neurons and synapses, which enables even higher energy efficiency and plasticity. Again, innovative technology concepts show promise over classical technologies.

For both generations of neuromorphic hardware, Fraunhofer IPMS is exploring crossbar architectures based on non-volatile memories, the ferroelectric field effect transistors. This is done within various European (TEMPO, ANDANTE, STORAIGE) and Fraunhofer internally funded projects. Particularly innovative materials research for future SNNs using Li-based systems is being conducted within the Saxon project MEMION.