PDF

Description

The invention and development of complementary metal-oxide-semiconductor (CMOS) transistor technology for digital computing have led to a global information technology revolution and economic boom, which have shaped modern society. However, transistor leakage current (IOFF) and limited subthreshold swing set a fundamental limit on the energy efficiency of digital computing today. Micro-electro-mechanical (MEM) switches (relays) previously have been shown to be promising for energy-efficient digital computing applications, due to their abrupt ON/OFF switching characteristics and negligible OFF-state leakage current.

This dissertation focuses on novel applications of MEM relays for facilitating new computer architectures that can be much more efficient than the classic von Neumann architecture. First, MEM relays are demonstrated to operate reliably with millivolt signals at cryogenic temperatures, due to much lower hysteresis voltage and more stable ON-state resistance. A sub-10 mV relay-based inverter circuit is demonstrated at a temperature of 4 K. Our experimental study indicates that MEM relays should be able to operate at temperatures as low as 1.8 K, making them promising candidates for ultra-low power cryogenic digital interface circuits for quantum computing.

Then superconducting MEM relays using niobium (Nb) as the contact material are investigated, in order to further reduce the operation power consumption for quantum computing applications at cryogenic temperatures. The detailed fabrication process flow is shown, and temperature-dependent measurements are conducted to check the superconductivity of the Nb electrodes.

Finally, DC-voltage-driven oscillatory behavior of MEM relays is investigated via experimental study and finite-element-method-based computer simulations. Sub-harmonic injection locking and coupled oscillation behaviors of MEM relays are demonstrated, indicating that MEM relay oscillators are promising for implementing Ising machines, which can solve large-scale combinatorial optimization problems much more efficiently than conventional computer architectures.

Details

Files

Statistics

from
to
Export
Download Full History