Hiring Institution
École Centrale de Lyon (EC Lyon)

PhD-Awarding Institutions
École Centrale de Lyon (EC Lyon)
RMIT University (RMIT)

Download the full Position Description

Option 1

Realization of ferroelectric artificial synapses for hardware implementation of neuromorphic networks
After more than 40 years of continuous evolution, our computing systems are reaching their limits. Indeed, the architecture of Von-Neumann, on which our computers are based, physically dissociates the hearts of calculations from the memory. The sequential processing of information is thus confronted with a bottleneck, more commonly known as “Memory Bottleneck”. One solution is to draw inspiration from the natural mathematical paradigms of the human brain, in which the data are massively parallel processed with high energy efficiency, realizing the hardware implementation of neuromorphic networks. The latter make it possible to bring the information storage sites (synapses) closer to the treatment sites (neurons). The major challenge of this bio-inspired approach is the realization of dense networks of artificial synapses to implement synaptic plasticity mechanisms. There are few man-made ferroelectric artificial synapses in the microelectronics industry to date. The aim of this project is to achieve the fabrication of such a device from transferable materials and processes in the semiconductor industry.
• The deposition of the ferroelectric HfZrO2 must respect the CMOS processes, in particular at low thermal budget (<450 ° C).
• Since HfZrO2 is a recent ferroelectric material, the influence of its thickness, stoichiometry, interfaces on ferroelectric properties is still poorly known today. This knowledge is essential in order to optimize the operation of the targeted devices.

Expected original contributions :
• Development and demonstration of artificial synapses based on ferroelectric HfZrO2 using CMOS compatible methods
• Identification of the optimal deposition conditions (i.e. thickness, stoichiometry, temperature, etc.) of the functional oxide for the targeted application
• Realization of integrated synaptic matrices of “Crossbar” type

This project will be realized in the framework of common labs between INL and CEA-LETI and STMicroelectronics.

Option 2

Operando characterizations of ferroelectric artificial synapses

After more than 40 years of continuous evolution, our computing systems are reaching their limits. Indeed, the architecture of Von-Neumann, on which our computers are based, physically dissociates the hearts of calculations from the memory. The sequential processing of information is thus confronted with a bottleneck, more commonly known as “Memory Bottleneck”. One solution is to draw inspiration from the natural mathematical paradigms of the human brain, in which the data are massively parallel processed with high energy efficiency, realizing the hardware implementation of neuromorphic networks. The latter make it possible to bring the information storage sites (synapses) closer to the treatment sites (neurons). The major challenge of this bio-inspired approach is the realization of dense networks of artificial synapses to implement synaptic plasticity mechanisms. There are few man-made ferroelectric artificial synapses in the microelectronics industry to date. This means that lot of effort still need to be done in order to understand and control the different factors that can help stabilize and enhance the properties of ferroelectric films. Among them, a proper integration of the films in functioning devices raises the question of how we can use a proper film/electrode interface to better understand the response of the ferroelectric film to an applied voltage. Switching the polarization in such films requires a metallic contact, raising fundamental issues on the behavior of the interface between the ferroelectric layer and the electrode. Indeed, there are several sources of mutual interactions between electrode, ferroelectric film, and substrate, such as:• Charge related effects: The free charges of an electrode or substrate can help screening the polarization-induced surface charges which are detrimental to maintaining exploitable polarization in the ferroelectric thin film. With metallic oxide electrodes, an ionic displacement at the electrode/ferroelectric interface will also screen this depolarizing field.
• Depletion areas due to chemical and electronic reconstruction or defects at interfaces (semiconductor-insulating areas).
• Strain related effects that can rise from thermal expansion mismatch or lattice mismatch.
• Thickness dependent properties (space limiting effects for tunneling, ionic mobility).

Expected original contributions :
• to identify the best routes to optimize the properties in ferroelectric HfZrO2 thin films in order to obtain ferroelectric tunnel junctions answering the industrial requirements;
• to characterize the phase, local domain structure and chemistry, defect related electronic structure, the role of the interface and internal fields will be evaluated as a function of cycling, specially to obtain a unique material/structure chemical and physical characterizations evidencing the role of oxygen vacancies in wake-up, imprint and endurance;
• to establish straightforward relationships between the ferroelectric stack composition and electrical performances and provide guidance rules toward stacks optimization thanks to both physical and compact model of Electrode-Ferroelectric-Electrode structures. Based on the chemical and electrical analyses, a physical model of Electrode-Ferroelectric-Electrode stacks, including interface layers and oxygen vacancies profiles will be realized.

This project will be realized in the framework of common lab between INL and STMicroelectronics.

Option 3

Prototyping and testing the learning capabilities of ferroelectric synaptic network

After more than 40 years of continuous evolution, our computing systems are reaching their limits. Indeed, the architecture of Von-Neumann, on which our computers are based, physically dissociates the hearts of calculations from the memory. The sequential processing of information is thus confronted with a bottleneck, more commonly known as “Memory Bottleneck”. One solution is to draw inspiration from the natural mathematical paradigms of the human brain, in which the data are massively parallel processed with high energy efficiency, realizing the hardware implementation of neuromorphic networks. The latter make it possible to bring the information storage sites (synapses) closer to the treatment sites (neurons). The major challenge of this bio-inspired approach is the realization of dense networks of artificial synapses to implement synaptic plasticity mechanisms. There are few man-made ferroelectric artificial synapses in the microelectronics industry to date. The purpose is to analyse the physical mechanisms that govern the dynamic switching behaviours and highlight how these properties can be utilized to efficiently implement synaptic and neuronal functions. Prototype system that can be used in machine learning and brain-inspired network implementations will be fabricated and simulated. The final challenge is to propose large scale implementations and opportunities for building bio-inspired, highly complex computing systems.

Expected original contributions :
• Theoretical and practical design requirements of HfZrO2 based ferroelectric artificial synapses
• Modelling of HfZrO2 based ferroelectric artificial synapses
• Characterization of integrated synaptic matrices of “Crossbar” type
• Prototyping and testing the learning capabilities of the synaptic network

This project will be realized in the framework of common labs between INL and CEA-LETI and STMicroelectronics.

Bertrand Vilquin
Sharath Sriram

Condensed matter physics, Nanoelectronics, Biomimetism