Interlinked OptoElectronic Spinapses – Spinapse

Heterostructures are grown using either sputtering or pulsed laser deposition. The structural, electronic and magnetic properties are then characterized using, for example, atomic force microscopy, photoluminescence and magnetometHeterostructures are grown using either sputtering or pulsed laser deposition. The structural, electronic and magnetic properties are then characterized using, for example, atomic force microscopy, photoluminescence and magnetometry techniques. We aim to control the nature and density of defects within the tunnel barrier.

Samples are then processed into devices using clean room techniques (UV photolithography, dry Ar etching, dielectric encapsulation, reactive ion etching, electrical contact deposition). Once devices are bonded onto a chip, and the chip installed on the cryostat, multifunctional measurements can begin.

Due to the open-ended nature of this research, there no single, established measurement protocol. We explore, from 10ns-1ms, how the state (resistance, magnetoresistance) of the magnetic tunnel junction device is altered by bias and light pulses of ~ns width.

Ab-initio theory reveals that double oxygen vacancies can enhance the spintronic performance of magnetic tunnel junctions. This counterintuitive result reflects a matching of electronic properties with those that govern the high spintronic performance of these devices, and a propensity to maintain coherent tunnelling transport.

Compared to sputtering MgO in a Ar atmosphere, we have experimentally shown that oxydizing pure Mg to form the MgO barrier generates a different proportion of single and double oxygen vacancies. Furthermore, this proportion can be tuned through the post-deposition annealing temperature. This change in the nature of the defect that defines the tunnel barrier height is accompanied by improvements in spintronic performance.

These results help pinpoint the maximum lateral size of a nanojunction to less than 2nm. They also herald a new research direction toward quantum computing based on the electronic properties of these oxygen vacancies within already industrialized devices, the operation of which already relies on quantum mechanics.

Beata Taudul and et al., “Tunnelling spintronics across MgO driven by double oxygen vacancies,” submitted., 2016.

Oxide electronics as a field of research has come to a crucial crossroads as two schools of thought have effected its development. On one hand, research has focused utilizing nominally intrinsic properties of oxide dielectrics to create so-called multifunctional devices that exhibit an array of functionalities and associated responses. For instance, the ferroelectric property may be used to craft a device whose resistive state is electrically controlled. This is called a memristance and behaves as an artificial synapse.

Yet, within future nanoscale multifunctional devices, the impact of structural defects (eg an oxygen vacancy in the dielectric lattice) can no longer be overlooked. Research has thus also focused on harnessing this extrinsic property, notably to devise magnetic tunnel junctions (MTJs) that combine memristance and spintronics, and mimick synapses. Through ANR JCJC 137 (SpinMarvel, 10/2009-9/2013), some of us have explored these combined effects. A defect defines an energetical potential well located in the dielectric band gap that contains neutral/charged ground states and corresponding excited states. In project SpinMarvel, using bias-dependent spectroscopy (see Fig. 1), some of us have unexpectedly unraveled how to statically address the electronic states of a precisely identified structural defect with electronic symmetry contrast.

This groundbreaking discovery opens exciting new prospects in the arena of neuromorphic computing beyond the scientific scope of ANR SpinMarvel. In the resubmitted (well-ranked in 2013) project Spinapse, we propose a novel, robust and homogeneous synaptic paradigm that considers the electrically driven temporal conversion of a structural defect between its several electronic states. We will implement a time-resolved electrical spectroscopic approach to rationally craft the defect’s electronic state through trapping/detrapping and thus shape the synaptic response. Goal 1: we wish to link the relevant rates of electrically driven, spin-polarized defect state conversion processes to electrical (+spin) synaptic activity in our MTJs.

The optical activity of dielectric defects reflects how light of appropriate wavelength may alter the charge/excitation state of a specific defect. As a proof-of-concept, some of us demonstrated in project SpinMarvel how to tune defect-mediated tunneling magnetoresistance (TMR) by statically light onto the MTJ (see Fig. 4). Furthermore, as a new, paradigm-enhancing lever that we discovered in Dec. 2013, we may optically manipulate the electronic symmetry-resolved transport channels that define TMR. Our static experiments suggest that the conversion rates between the ground and excited electronic levels of a defect state may be optically controlled down to 10-8s temporal range and spintronically measured. Goal 2: we wish to design an electrically and/or optically addressable device with spin synapse-mimicking optical output that can substantially build on light-induced three-factor non-Hebbian learning memristances.

Thus, our proposed paradigm, using bias pulse amplitude and photon energy in up to three time domains down to 10-8s, implies the spectrotemporal crafting of the electronic states of a defect within a magnetic tunnel junction with electrical- or optical spin synapse-mimicking response. To fully express the neuromorphic potential behind our synapse paradigm, we will tailor the spatial distribution of two distinct defect species within the nominal tunnel barrier, each of which can be addressed/probed separately through electrical and optical means. The states of these defect species are interlinked through hopping tunneling transmission. By exploring neuromorphic concepts associated with not just one but two interlinked spinapses, we wish to more broadly disrupt the spintronics and neuromorphic computing research communities thanks to our new paradigm with electrical & optical inputs and electrical (+spin), optical (+spin) outputs.