Materials for Neuromorphic Circuits

Projektinformationen

MANIC

ID Finanzhilfevereinbarung: 861153

DOI

10.3030/861153

Projekt abgeschlossen

EK-Unterschriftsdatum 16 August 2019

Startdatum 1 November 2019

Enddatum 30 April 2024

Finanziert unter

• EXCELLENT SCIENCE - Marie Skłodowska-Curie Actions

Gesamtkosten

€ 4 128 380,67

EU-Beitrag

€ 4 128 380,67

Koordiniert durch

RIJKSUNIVERSITEIT GRONINGEN
Netherlands

Periodic Reporting for period 2 - MANIC (Materials for Neuromorphic Circuits)

Berichtszeitraum: 2021-11-01 bis 2024-04-30

Zusammenfassung vom Kontext und den Gesamtzielen des Projekts

Nowadays, it is impossible to envision a major industrial, scientific or societal step forward without the role of advanced computing technology, including supercomputers. However, computer technology is demanding an increasing amount of energy. For example, today’s supercomputers consume more than 10 MW each which is equivalent to the electricity consumption of 30,000 households. Moreover, this number will only grow with the increasing developments of the Internet-of-things which requires a massive increase in computing device to enable front-end sensory systems to extract information from its collected data.

Therefore, the European Union, its international competitors (USA, Asia) and high-tech companies, like IBM, are investing considerable efforts into developing computing platforms, so-called neuromorphic computing, that will be able to emulate the human brain. Our brain can make more calculations per second than the fastest of supercomputers because of parallel processing: networks of neurons and synapses that are working simultaneously. And all this at a much lower power consumption of only 20 W, which is a fraction of the power needed by a supercomputer. To this end, there are efforts to emulate brain functions through novel architectures based on CMOS technology (silicon chips). The success of this approach has been made clear with the recent development of neuromorphic processors (TrueNorth by IBM, SpiNNaker and BrainScaleS by the HBP consortium or Loihi by Intel). The approach of MANIC is to look into alternatives to silicon in order to develop basic device units that are more fitting to the needs of cognitive-type processing than current transistors: the so-called memristive devices based on novel resistive switching materials. In the way to CMOS-free architectures, it is also crucial to be able to combine CMOS circuits with memristive devices. These new materials will become the key elements in a new generation of electronic devices that will be able to adapt to software and enable efficient learning.

Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse

The first part of the Action was mostly focused on setting up the management structure, the recruitment of the fellows and starting up the joint research and training programme. In that respect, the following results towards the objective of the Network have been achieved:
‒ A successful set-up of the management bodies and the management structure, with regular Management Team meetings and organization of Action Events, such as the Kick-Off Meeting and Network Wide Events (NWE).
‒ An open, transparent, merit-based, impartial and equitable recruitment procedure in line the principles of the ‘Code of Conduct for the Recruitment of Researchers’.
‒ MANIC successfully set up Recruitment Committees (RCs) for the recruitment of the ESRs.
‒ Two Calls were published on Euraxess and the MANIC website, and were disseminated further via international paid/unpaid channels. Candidates were able to submit their application via the Virtual Portal.
‒ MANIC successfully selected the desired number of 15 talented and motivated ESRs, who are a good match with the scientific programme of MANIC. Due to health issues one of the ESRs needed to stop but was replaced without disturbing the developments of the Action.
‒ The background of the selected ESRs is truly international, with 9 different nationalities represented, ranging from Germany and UK, to Philippines and China.
‒ Considering the strong competition in the recruitment of female candidates in natural sciences, we are pleased to have recruited 2 female ESRs.
‒ All ESRs have enrolled in the PhD programmes at their respective host institutions or partner institutions and have completed their 1st year Personal Career Development Plans.
‒ Several communication materials and channels have been developed, such as a project logo, project website and project social media accounts. Part of these have also been actively used in the recruitment procedure.
‒ The MANIC network has successfully planned for ethics management regarding the ethical issues identified for the MANIC project.
‒ Given the restrictions imposed through COVID-19 related measures, the progress of the Action is beyond our expectation and all ESR projects have started to our satisfaction which already resulted in significant progress on the scientific objectives set out at the start of the project.

Regarding the scientific progress, MANIC has made significant breakthroughs towards the development of materials for neuromorphic circuits. The MANIC community, with its 15 PhDs and supervisors, and a dynamical group of industrial partners, has made a significant difference in the European scientific landscape. WIthin MANIC we have developed interactions between material scientists and circuit designers that did not exist so far and, by doing so, we have brought these two communities together. We are considered important players in this new field and we have served as example to many.

The Secondments of the ESRs have run smoothly and the young researchers have gained valuable experiences working in industrial environments (or academic environments, for those ESRs working at IBM). Some of the ESRs have already defended their thesis and most of them are in the process of writing, ready to defend in the coming months. The joint work on similar issues using competing approaches has led to rather advanced scientific discussions among ESRs; the team of PIs of this project regards this as a confirmation on the original project goals. In task 3.1 ‘Single Synapses’ focusses on the fabrication and characterization of materials suitable for synaptic and neuronal functionality. Ferroic and ferroelectric material were made and characterized, including Barium Ferrite, Strontium Titanate, and Sr2Mg(BO3)2, Materials with metal-insulator-transition: NdNiO3 and VO2 and RRAM materials (e.g. TaOx/HfO2). Some of these materials were transferred toTask 3.2 ‘Self-assembled networks. In this task these materials were characterised in their circuit network context. The materials involved in Task 3.2 are Bismuth Ferrite, WO3, PTO, PGO and PZT. Finally, Task 3.3 ‘Circuit integration and learning’ deals with circuit design aspects with input from T3.1 and T3.2.

Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)

MANIC is training 15 early stage researchers to experiment with new materials, bringing in materials scientists, physicists and chemists to work on the quest for neuromorphic circuits. They will work on materials that will accelerate technological advances towards computing platforms that work efficiently and flexibly without the current high energy demands, just like the human brain. Within the network, universities and world-leading companies are working together, thereby opening the possibility to bring innovation from the scientific community to the industrial world and vice versa in order to have direct impact in the society.
Novel results and breakthrough include: VO2-based relaxation oscillators, discovery of hopfions in nanoferroics, domain walls in LNO with synaptic behaviour, networks of domain walls as memristive networks, functional properties of domain walls in WO3, ferroelectric tunnel junctions for non-volatile memory, oxide-based memristors and ferroelectric memristors as well as their integration in CMOS circuits.

MANIC logo