The Technology
In-memory computing is an emerging strategy that performs computation directly where data is stored, drastically reducing energy-intensive data movement. Electrochemical random-access memory (ECRAM) is a promising foundation for non-von-Neumann in-memory computing using neuromorphic systems. ECRAM is a three-terminal memristor device where the channel’s resistance is modulated through ion intercalation into the channel, which is serving as memory element. Despite its immense potential, the broad deployment of ECRAM is currently hindered by limitations of direct intercalation of ions into the channel material, where it can lead to semiconductor degradation and reduced electronic stability over time.
The novel Floating Gate ECRAM device architecture effectively mitigates the material degradation pathways. This is achieved by confining the ionic movement within the device gate structure and establishing a memory element in which programming occurs via capacitive coupling. This invention allows establishing ECRAM as a major building block for ultra-low power computing systems.
Main Advantages
- Improved durability – no direct ion insertion into the semiconductor, avoiding lattice degradation and electronic instability
- Broader materials choice – enables high-performance electronic channel materials that do not need intrinsic ionic conductivity.
- Training and inference compatibility- meeting wide range of speed, retention and energy requirements within a single device.
- Analog, reversible operation – supports gradual potentiation and depression for multi-level neuromorphic systems.
Applications
- AI accelerators – analog in-memory matrix operations for neural-network workloads.
- Neuromorphic processors – hardware synapses for learning and inference.
- Edge intelligence – low-data-movement computing for sensors, robotics, wearables and autonomous systems.
- Adaptive electronics – multi-level memory and locally trainable circuits.
Business Development Contacts
