Stacked semiconductor architecture provides local memory for wearables

Researchers from KAIST, UNIST, and POSTECH have developed a vertically stacked semiconductor architecture designed to provide wearable electronics with local memory. This technology allows small devices to process time-dependent signals, such as movement and physiological data, locally rather than relying on external computing.

The system utilizes all-solid ion-gated carbon nanotube transistors controlled by a solid-state ionogel. By adjusting the ion concentration and ionogel thickness, the team created layers with varying response speeds. Faster layers respond to recent signals, while slower layers retain information over longer intervals. This stacking allows the hardware to distinguish temporal patterns by combining immediate and historical data.

In testing, the devices successfully differentiated all 16 possible combinations of four consecutive input signals. Simulations using reservoir computing to classify moving handwritten digits achieved a validation accuracy exceeding 90%. The researchers fabricated the devices on both 4-inch wafers and flexible substrates, noting that the electrical characteristics remained stable for 55 months after fabrication.