Figure 1 | Scientific Reports

Figure 1

From: Ferroelectric capacitors and field-effect transistors as in-memory computing elements for machine learning workloads

Figure 1

Deep-learning (DL) applications, associated challenges, and the need for in-memory computing (IMC) with Non-Volatile Memory (NVM) devices. (a) Holistic view of DL applications, the architecture of a fully connected neural network, and the challenges that allow IMC to complement. (b) IMC and NVMs aimed at this purpose, featuring the characteristics of crossbar architecture characteristics based on memristor devices and capacitive crossbar array for its counterpart. The resistance plot shows the values of the Low Resistance State (LRS) and High Resistance State (HRS) for various NVMs, and the internal resistances of FeCaps obtained from our experimental data (at different HZO thicknesses of 4.5 nm and 9.5 nm). When the width of crossbar wire is scaled from 10 to 5 nm13, there is a notable increase in its resistance. This increase brings it into a range comparable to the LRS of ReRAMs, PCMs, and MRAMs, yet maintaining distinction from the internal resistances of FeCaps. Our key contributions are highlighted.

Back to article page