This technology presents a capacitive multiplication of two analog quantities where one quantity can be stored as a mechanical stress/deformation state in a ferroelectric domain fraction.
The critical invention in this technology is a resistive material as one of the electrodes on the ferroelectric material. The resistive layer allows for a linear variation in the voltage on the electrode when different voltages are applied on the left and right sides of the resistive electrode via electrodes. This linear change in voltage leads to a linear change in the electric field applied across the ferroelectric material. Moreover, two electrodes are used to capacitively measure the gap changes due to the motion induced by the ferroelectric capacitor when operated as a piezoelectric material.
This architecture takes advantage of two principles. The first is the ferroelectric mechanism for storing the weights and deforming a beam. The working principle of storing weights for multiplication to inputs is described in three modes: Mode 0: Resetting of the ferroelectric layer; Mode 1: Storing Weights and Mode 2: MAC (Multiply and Add calculations). This mechanism can also be used to induce a controllable bending moment of a released structure. By modifying the strain level in the structure, it is possible to shift the resonance frequency of the piezoelectric film resonator operating as an FBAR (Thin Film Bulk Acoustic Resonator) or as a CMR (Contour Mode Resonator). This frequency shift can be important to tune a resonator to a specific frequency as the frequency changes due to errors in fabrication and temperature. Second, is the capacitive mechanism for sensing the beam displacement due to the applied bending moments. In contrast with CMOS transistor-based neuromorphic devices that usually consume power even in an idle state, this technology only consumes power while the beam moves from one state to another.
- Energy-efficient operation in MAC (Multiply and Add calculations) mode
- Negligible leakage current value as no physical connection between the electrodes
- Neuromorphic devices
- Frequency Tuning of Linearly Graded Poled Ferroelectric Device
Name: Ryan Luebke