Memristive Devices Toward Smart Technical Systems (SPP 2262);
Termin:
20.12.2019
Fördergeber:
Deutsche Forschungsgemeinschaft (DFG)
A multitude of different devices manufactured in distinct technologies can be classified as memristors or memristive devices. Their most popular application field so far is the non-volatile memory sector. However, in-memory processing, neuromorphic computing, machine learning, are other emerging areas of practical interest. The MemrisTec programme is devoted to seven main research areas, three concerning theoretical and experimental studies, and four focusing on applications. The integration between experimental results and theory is crucial for advancing the understanding of memristive dynamics. The development of innovative mem-computing paradigms and their implementation through novel memristive circuits and systems will enable fast, adaptive, and low-power data sensing, storage and processing, as required in IoT applications. Memristor-based dynamic processors, extending the functionalities of conventional purely-CMOS computing systems, will be prototyped.
The two distinct scientific communities, working on the physical realisation of memristive devices, on one side, and on the theory of memristors, on the other side, have not yet agreed on joining complementary strengths to make effective use of memristors potential on the system level. However, breakthrough advances necessitate a tight connection between these two worlds. Therefore, an essential requirement for a MemrisTec project is a strong interaction between a partner, responsible for the realisation of physical devices, and another one, engaged with theoretical studies, toward a common application-oriented goal. Proposals focused on one of the two pillars only, will not be considered for funding. Physics-based considerations will set the boundary conditions for the system-level design, which, in its turn, will guide the conceptual development of novel signal processing strategies, whose hardware implementation will then require the synthesis of new device concepts and circuit architectures. The focus of a MemrisTec project may stretch over almost any resistive switching technology. However, technologies based upon magnetic effects will not be considered, since they are object of extensive research elsewhere. For the same reason, all non-volatile memory concepts, as well as standard approaches to neuromorphic system design, delivering non-full custom ASICs, fully-digital electronics, or Von-Neumann hardware architectures, are excluded from MemrisTec research.
Theoretical and experimental research: electrical characterisation, modelling and simulation of memristive devices
Studies shall be devoted to develop accurate, numerically-stable, and compact models for the resistive switching dynamics of memristors on the basis of electrical characterisation tests. The availability of circuit simulation-friendly model representations, robust against changed in inputs/initial conditions/parameter space, are particularly beneficial to explore the full potential of memristors in electronics through mixed-signal simulations on both physical and system levels.
System-theoretic model investigation to support circuit design
The models shall be analysed on the basis of concepts from nonlinear system theory, so as to gain a good understanding for the nonlinear dynamics of memristive devices, enabling a conscious design of circuits, which, by harnessing their peculiar capabilities, may process data more efficiently than standard CMOS ones.
Architectural studies: from data processing strategies to circuit and system implementations
Memristors allow a revolution in the way data have been processed in electronics so far. Their multi-level low-power data storage capability, combined with peculiar computing functionalities, enables the introduction of non Von-Neumann processing paradigms, known as near-memory or in-memory computing concepts, for future high-performance embedded processing systems or smart sensors for IoT, Cyber-Physical System, and Edge Computing applications. The architecture of purely-memristive and hybrid memristive/CMOS circuits shall be explored toward the design of innovative electronic systems, allowing to implement such unconventional data processing paradigms, enabling IC performance growth beyond the Moore era.
Applications of the MemrisTec Research Priority Programme
With their rich dynamics and the combined capability to sense, store, process, and encrypt data in a common medium, memristor nano-devices offer the opportunity to realise smart miniaturised technical systems, which, implementing non-standard computing paradigms, may address the IoT industry demanding requirements more efficiently than state-of-the-art hardware. Practical applications under focus include:
design of innovative adaptive analogue circuits
development of mem-computing arrays
synthesis of data encryption systems
realisation of data sensing devices
Within the framework of the Priority Programme, the DFG encourages the submission of projects envisioning a close interaction between theoreticians and experimentalists, toward the development of memristor-centered bio-inspired, reconfigurable, energy-efficient and real-time computing systems.
Proposals must be written in English, and submitted via elan, the DFG portal for project submissions by 15 October 2019. To enter a new project within this Priority Programme, go to Proposal Submission New Project/Draft Proposal Priority Programmes and select SPP 2262 from the current list of calls.
Further information:
https://www.dfg.de/foerderung/info_wissenschaft/info_wissenschaft_19_55/index.html
The two distinct scientific communities, working on the physical realisation of memristive devices, on one side, and on the theory of memristors, on the other side, have not yet agreed on joining complementary strengths to make effective use of memristors potential on the system level. However, breakthrough advances necessitate a tight connection between these two worlds. Therefore, an essential requirement for a MemrisTec project is a strong interaction between a partner, responsible for the realisation of physical devices, and another one, engaged with theoretical studies, toward a common application-oriented goal. Proposals focused on one of the two pillars only, will not be considered for funding. Physics-based considerations will set the boundary conditions for the system-level design, which, in its turn, will guide the conceptual development of novel signal processing strategies, whose hardware implementation will then require the synthesis of new device concepts and circuit architectures. The focus of a MemrisTec project may stretch over almost any resistive switching technology. However, technologies based upon magnetic effects will not be considered, since they are object of extensive research elsewhere. For the same reason, all non-volatile memory concepts, as well as standard approaches to neuromorphic system design, delivering non-full custom ASICs, fully-digital electronics, or Von-Neumann hardware architectures, are excluded from MemrisTec research.
Theoretical and experimental research: electrical characterisation, modelling and simulation of memristive devices
Studies shall be devoted to develop accurate, numerically-stable, and compact models for the resistive switching dynamics of memristors on the basis of electrical characterisation tests. The availability of circuit simulation-friendly model representations, robust against changed in inputs/initial conditions/parameter space, are particularly beneficial to explore the full potential of memristors in electronics through mixed-signal simulations on both physical and system levels.
System-theoretic model investigation to support circuit design
The models shall be analysed on the basis of concepts from nonlinear system theory, so as to gain a good understanding for the nonlinear dynamics of memristive devices, enabling a conscious design of circuits, which, by harnessing their peculiar capabilities, may process data more efficiently than standard CMOS ones.
Architectural studies: from data processing strategies to circuit and system implementations
Memristors allow a revolution in the way data have been processed in electronics so far. Their multi-level low-power data storage capability, combined with peculiar computing functionalities, enables the introduction of non Von-Neumann processing paradigms, known as near-memory or in-memory computing concepts, for future high-performance embedded processing systems or smart sensors for IoT, Cyber-Physical System, and Edge Computing applications. The architecture of purely-memristive and hybrid memristive/CMOS circuits shall be explored toward the design of innovative electronic systems, allowing to implement such unconventional data processing paradigms, enabling IC performance growth beyond the Moore era.
Applications of the MemrisTec Research Priority Programme
With their rich dynamics and the combined capability to sense, store, process, and encrypt data in a common medium, memristor nano-devices offer the opportunity to realise smart miniaturised technical systems, which, implementing non-standard computing paradigms, may address the IoT industry demanding requirements more efficiently than state-of-the-art hardware. Practical applications under focus include:
design of innovative adaptive analogue circuits
development of mem-computing arrays
synthesis of data encryption systems
realisation of data sensing devices
Within the framework of the Priority Programme, the DFG encourages the submission of projects envisioning a close interaction between theoreticians and experimentalists, toward the development of memristor-centered bio-inspired, reconfigurable, energy-efficient and real-time computing systems.
Proposals must be written in English, and submitted via elan, the DFG portal for project submissions by 15 October 2019. To enter a new project within this Priority Programme, go to Proposal Submission New Project/Draft Proposal Priority Programmes and select SPP 2262 from the current list of calls.
Further information:
https://www.dfg.de/foerderung/info_wissenschaft/info_wissenschaft_19_55/index.html