European Commission logo
polski polski
CORDIS - Wyniki badań wspieranych przez UE
Zawartość zarchiwizowana w dniu 2024-05-28

FerroElectric Multifunctional tunnel junctions for MEmristors and Spintronics

Final Report Summary - FEMMES (FerroElectric Multifunctional tunnel junctions for MEmristors and Spintronics)

The aim of the project FEMMES was to exploit Ferroelectric Tunnel Junctions (FTJs) composed of two electrodes separated by a ferroelectric tunnel barrier and the intrinsic low-power of “ferroelectric writing”, to obtain:
1) memristive FTJs mimicking the plasticity of synapses for an exploitation in neuromorphic analog circuits.
2) a low-power electrical control of spin polarized electron sources for spintronics in FTJs with magnetic electrodes.
During this project, we optimized FTJs to obtain very large, stable and reproducible electroresistance (TER) phenomena, with OFF/ON ratios as high as 10.000 and a writing voltage of 2V, corresponding to write powers as low as ~1x104 A cm-2 at room temperature (Nature Nanotechnology 7, 101 (2012); ACS Nano 7, 5385 (2013); APL 104, 052909 (2014); APL Materials 3, 061101 (2015), Adv. Elec. Mat. 2, 1500245 (2016)). These results qualify FTJs as a new kind of non volatile memory with the advantage, compared to standard ferroelectric memories, of a simpler reading process and based on a purely electronic mechanism that could circumvent the problems faced by other resistive memories based on voltage-induced electromigration. We also studied FTJs deposited on Si (Nature Communications 7, 11502 (2016)).
In these FTJs, the voltage-controlled of the domain configurations of the barrier yield memristive behavior with resistance variations exceeding 4 orders of magnitude and a 10 ns operation speed. Using models of ferroelectric-domain nucleation and growth, we explain the quasi-continuous resistance variations and derive a simple analytical expression for the memristive effect (Nature Materials 11, 860 (2012); ACS Nano 7, 5385 (2013)). During this project we also studied the potential of these FTJs as electronic nanosynapses for neuromorphic computing and demonstrated that spike-timing-dependent plasticity can be harnessed from the intrinsically inhomogeneous ferroelectric polarisation switching. Simulations based on the precise physical model of ferroelectric domain nucleation, showed that arrays of ferroelectric nanosynapses can autonomously learn to recognise patterns in a reliable and predictable way, opening the path towards unsupervised learning in spiking neural networks (Boyn et al.; to be published).
In order to electrically tune a spin polarized electron source we studied FTJs with different ferromagnetic electrodes (Fe, Co, (La,Sr)MnO3, (Ca, Ce)MnO3..) and ferroelectric tunnel barrier (BaTiO3, BiFeO3, PVDF) and demonstrated the possibility to change electrically the spin polarization of Co or Fe in good agreement with ab initio calculations (Nature Mater. 10, 753 (2011), Nanoletters , 12, 376 (2012)). Interestingly we also demonstrated the opportunity to induce a magnetic character in BaTiO3 at the interface with the ferromagnetic material.
Although a significant effort was made to obtain a large TEMR at room temperature we faced the problem of the growth of oxides on transition metals and lithography of BaTiO3. We decided to focus on an alternative route and concentrated mostly on the development of PVDF tunnel barriers by a Langmuir Blodgett method. Promising results have been published in Nat. Comm. 7, 11502 (2016).
In order to electrically tune the magnetic properties we studied heterostructures combining the ferroelectric material BaTiO3 and the transition metal alloy FeRh which presents a metamagnetic transition at equiatomic proportion. In this system we demonstrated the opportunity to tune the antiferromagnetic to ferromagnetic transition by applying a voltage to BaTiO3 (Nature Materials 13, 245 (2014); Scientific reports 3, 2834 (2015)). Interestingly, this system also appeared as an interesting one for magnetocaloric cooling due to the possibility to reduce by 96% the magnetic losses through a multicaloric (electric and magnetic) cycle (Nature Communications 7, 11614 (2016)).