Berlin 2024 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
HL: Fachverband Halbleiterphysik
HL 13: Poster I
HL 13.16: Poster
Montag, 18. März 2024, 15:00–18:00, Poster E
Nanoscopic electrical heaters for mixed-mode in-memory computing with phase change materials — •Nishant Saxena1, Niklas Vollmar1, Ivonne Bente2, Frank Brückerhoff-Plückelmann2, Wolfram Pernice2,3, and Martin Salinga1 — 1Institute of Materials Physics, University of Münster, Wilhelm-Klemm-Str. 10, 48149 Münster, Germany — 2Physical Institute, University of Münster, Heisenbergstraße 11, 48149 Münster, Germany — 3Kirchhoff Institute for Physics, University of Heidelberg, Im Neuenheimer Feld 227, 69120 Heidelberg, Germany
In contrast to von-Neumann processors, in-memory computing systems avoid data movement between the processing unit and memory and are therefore fundamentally more efficient. In particular, matrix-vector multiplications on crossbar arrays of photonic waveguides can be performed with extremely high throughput and low energy consumption. A promising approach uses phase change devices to represent the matrix elements. Depending on the phase of the material, a variable fraction of the incoming light is transmitted. We show the realization of mixed-mode phase change devices for in-memory computing on a silicon-on-insulator platform. Their state is written with heating supplied by electrical PIN diodes, which locally heat up individual pads of phase change material on a waveguide. By switching of separate, nanoscopic phase change pads, this scheme enables a reproducible tuning of transmission states with high resolution. The foundry-fabricated devices allow for large-scale integration in mixed-mode integrated circuits with feedback between the photonic and electronic parts.
Keywords: Phase change materials; PIN diodes; Silicon-on-insulator; Integrated photonics; In-memory computing