NEWS

How manganite-based Memristor Behavior Impacts Faster Learning in Hardware Neural Networks  

In a groundbreaking study focused on the emerging field of neuromorphic computing, MELON researchers from CONICET-centre, Argentina, have shed new light on the potential of oxide-based memristor arrays with cross-bar architecture. Neuromorphic computing aims to replicate the intricate and efficient workings of the mammalian brain, bridging the divide between biological and synthetic systems. Central to the study was the exploration of potentiation-depression (P-D) curves on various manganite-based memristive systems. The results indicate that by leveraging specific characteristics of manganite-based memristive systems, one can potentially optimize the design and performance of neuromorphic hardware, leading to faster and more efficient machine learning models.
The results are published in Physica Scripta, 98, 095917 (2023)

Multilevel devices demonstrating switchable polarization enable us to efficiently realize neuromorphic functionalities including synaptic plasticity and neuronal activity. MELON's researchers designed the ferroelectric unit comprising multiple nanodots for implementation of the topologically configurable non-binary logic cell and integrated it into a gate stack of the field-effect transistor as an analog-like device with resistive states.  The devised ferroelectric multilevel devices provide a pathway toward the novel topologically-controlled implementation of discrete synaptic states in neuromorphic computing...
      Neuromorphic Computing and Engineering (2023)

The exponential growth of artificial intelligence has been fueled by advancements in hardware technologies, which have allowed us to tackle a wide variety of scientific and technological problems. One area that has seen significant progress is the development of neuromorphic computing materials, which seeks to mimic the structure and information-processing mechanisms of biological systems. Recently, MELON researchers from Argentina in collaboration with Spain, and France made an important contribution to this field by adding new materials to the list of materials that can realize the behavior of both neurons and synapses on the same device. These new materials are phase-separated manganites, and their ability to function as both neurons and synapses could be a key feature for the development of neuromorphic computing hardware...

Epitaxial ferroelectric memristors integrated with silicon

Neuromorphic computing requires the development of solid-state units able to electrically mimic the behavior of biological neurons and synapses. This can be achieved by developing memristive systems based on ferroelectric oxides. MELON researchers fabricated and characterized high-quality ferroelectric memristors integrated with silicon, demonstrating remanent resistance loops with tunable ON/OFF ratio and asymmetric resistance relaxations. These properties might be harnessed for the development of neuromorphic hardware compatible with existing silicon-based technology. The results are published in the open-science journal Frontiers in Nanotechnology 3:1092177 (2022)

  The ferroelectric FET with negative capacitance

Integrating negative capacitance into the field-effect transistors (FET) promises to break fundamental limits of power dissipation known as Boltzmann tyranny in emergent computing circuits. However, the realization of the stable static negative capacitance remains a daunting task. Here we put forth an ingenious design for the ferroelectric domain-based FET with the stable negative capacitance...  
npj Computational Materials, 8, 52 (2022)

  Multi-mem oxide-based interfaces for data storage and neuromorphic computations

Memristive systems emerge as strong candidates for the implementation of resistive random access memories and neuromorphic computing devices, as they can mimic the electrical analog behavior of biological synapses. In addition, complementary functionalities, such as memcapacitance, could significantly improve the performance of bio-inspired devices in key issues, such as energy consumption. However, the physics of mem systems is not fully understood so far, hampering their large-scale implementation in devices. MELONs' researchers from Buenos Aires, Bariloche, Groningen, and Zaragoza paved the way for the integration of multi-mem interfaces (memristive and memcapacitive) of  oxide-based heterostructures -displaying the largest memcapacitive response reported so far by a factor of 10- in multiple device architectures. They found that current pulses stimulate an optimum memory response of epitaxial (110) La-Sr-Mn-Co oxide films grown on Nb:SrTiO3 substrates. Under these conditions, the system efficiently exchanges oxygen with the environment minimizing, at the same time, self-heating effects that trigger nanostructural and chemical changes that could affect the device integrity and performance.

Polaron formation in Bi-deficient BaBiO3 

In the search of novel interfaces between insulating oxides with 2D metallic behavior, MELON researchers from Buenos Aires and Groningen have explored the electronic and transport properties of the interface between BaBiO3 (BBO) and yttrium-stabilized zirconia (YSZ).  BBO is a charged ordered Peierls-like perovskite well known for its superconducting properties upon K or Pb doping. They demonstrated that a nanometric BBO layer with strong Bi deficiency is stabilized by depositing an YSZ capping layer on top. By combining transport measurements with ab initio calculations it was disclosed a scenario where the Bi vacancies give rise to the formation of polarons and that the electrical transport is dominated by the migration of these polarons trapped at Bi3+ sites. The results show that cation vacancies engineering, hardly explored to date, appears as a promising pathway to tune the electronic and functional properties of perovskites.

The team of researchers, supported by the H2020-MSCA-RISE actions,  for the first time showed the possibility of switching chirality in nanosystems to create the neuromorphic computers that mimic the human brain...

Oxygen vacancies for memristive response

Artificial Intelligence aims to develop computer systems capable of performing tasks complex skills such as voice and pattern recognition or decision-making, mimicking capabilities of the human brain. At present, new materials and devices are intensively researched able to perform these tasks efficiently.  In a joint work of researchers of the MELON, from CNEA-CONICET and RUG, published in it was found that both non-volatile and volatile electrical resistance changes, which can replicate the behavior of brain synapses and neurons, respectively, coexist in ferroelectric capacitors. From experiments and phenomenological modeling, it was evidenced that the physical origin of the resistance changes is based on the combination of an electronic effect, linked to the reversal of ferroelectric polarization, and the electromigration of oxygen vacancies, both modulating the energy barriers present at the interfaces between the ferroelectric layer and the metal electrodes. It was shown that both mechanisms are strongly intertwined and that the observed resistance relaxations are associated with electromigration of oxygen vacancies, dominated by the depolarizing electric field usually present in ferroelectric thin films. The results reported in this work will contribute to the development of neuromorphic computing devices. (Adopted from Gacetilla INN Octobre 2020)