Memristive response and neuromorphic functionality of polycrystalline ferroelectric Ca:HfO$_{2}$-based devices

TL;DR

This study demonstrates that polycrystalline Ca:HfO₂-based memristors exhibit multiple memristive mechanisms, including ferroelectric polarization and oxygen vacancy migration, enabling neuromorphic functionalities such as time-series recognition.

Contribution

It reveals the coexistence of dielectric and ferroelectric memristive responses in Ca:HfO₂ devices and introduces a simple learning algorithm leveraging resistance relaxation.

Findings

Devices are forming-free and support ferroelectric switching.

Multiple memristive mechanisms coexist and compete within the devices.

A new learning algorithm exploits resistance relaxation for time-series tasks.

Abstract

Memristors are considered key building blocks for the development of neuromorphic computing hardware. For ferroelectric memristors with a capacitor-like structure, the polarization direction modulates the height of the Schottky barriers -- present at ferroelectric/metal interfaces -- that control the device resistance. Here, we unveil the coexistence of multiple memristive mechanisms in Pt/Ca:HfO$_2$/Pt devices fabricated on silicon by a simple and effective low-toxicity chemical solution method. Depending on the fabrication conditions, either dielectric or ferroelectric devices are obtained, each one presenting a distinct memristive response. The devices are forming-free and can sustain ferroelectric switching and memristive behavior simultaneously. Aided by numerical simulations, we describe this behavior as a competition of different mechanisms, including the effect of the ferroelectric polarization on Schottky interfaces and oxygen vacancy electromigration. Finally, we propose a simple learning algorithm for time-series recognition, designed to take advantage of the resistance relaxations present in the case of the ferroelectric devices.