Skip to main content

CMOS/magnetoelectronic Integrated Circuits wil Multifunctional Capabilities

Periodic Reporting for period 3 - MAGICAL (CMOS/magnetoelectronic Integrated Circuits wil Multifunctional Capabilities)

Reporting period: 2018-11-01 to 2020-04-30

The microelectronics industry is currently investing heavily on Spin Transfer Torque Magnetic memory (STT-MRAM) with the aim of replacing embedded FLASH memory and later address SRAM and DRAM applications at sub-20nm technology nodes.
In this context, this ERC project MAGICAL “CMOS/Magnetoelectronic Integrated Circuits with Multifunctional Capabilities” intends to realize groundbreaking advances in ultra-low power multifunctional systems based on hybrid CMOS/magnetic technology. With the development of portable electronics and of the Internet of Things (IOT), more and more functions must be embedded on chip: logic/memory, sensing, communication, etc. With existing technologies, the hurdles are power consumption, communication bandwidth, processing/ packaging costs. MAGICAL will demonstrate that these limitations can be largely overcome through hybrid CMOS/magnetic technology.
The project follows three main goals:
- Firstly, to strengthen the STT-MRAM technology by investigating two novel ideas aiming at solving two remaining difficulties in sub-20nm STT-MRAM development: the nanostructuration of magnetic tunnel junctions at small feature size and narrow pitch and the long-term data retention. This will open the path to high density (>Gbit) STT-MRAM. These goals were actually already reached in the first half of the project. A new method for nanopatterning magnetic tunnel junctions by depositing the magnetic stacks on prepatterned metallic pillars was successfully demonstrated. In addition, a new concept of MRAM taking advantage of the shape of the storage layer to extend the memory retention of memory cell of size below 10nm was proposed and experimentally demonstrated. This novel concept of Perpendicular Shape Anisotropy STT-MRAM (PSA-STT-MRAM) should allow to extend the downsize scalability of MRAM to 4nm.
-Secondly, we will demonstrate that Digital, analog (3D orientation sensor), RF communication functions can be realized with the same baseline technology as the one developed for STT-MRAM. As a result, these three types of functions can be homogeneously integrated in a single chip, a major improvement compared to conventional heterogeneous integration. The prime benefits expected from this project are:
ultralow power thanks to STT-MRAM non volatility and on-chip computation capability, greatly improved communication functionalities (cloud as well as intrachip communication), reduced process/packaging costs.
-Thirdly, through various actions, MAGICAL aims at narrowing the cultural gap that still exists between magnetism and microelectronics communities. Along this line, an annual summer school “Introductory Course on MRAM” has been organized twice during the first half of the project and special events (MRAM special Poster session and Forum) gathering experts in magnetism and microelectronics were organized at IEDM 2016, 2017 and 2018.
From scientific and technical point of view, several major achievements were realized to this point in the project.
We proposed and demonstrated a new approach for the nanopatterning of magnetic tunnel junctions for dense memory applications such as DRAM replacement. The conventional approach for the nanopatterning of the magnetic material constituting the memory dot is by ion beam etching at various etching angles to avoid redeposition of the etched species on the sidewalls of the tunnel junctions. However, this approach does not work at narrow pitch because of shadowing effects. Instead, we proposed and demonstrated that we could obtain nanostructured memory dots by directly depositing the magnetic stacks on prepatterned metallic non-magnetic pillars by a Phase Vapor Deposition tyechnique such as sputtering. The material is then naturally nanostructured during the deposition thus not requiring any post-deposition etching. The material in the trenches between the dots can be left there, not causing any drawback and even playing the role of a magnetic flux absorber which reduces the crosstalk between adjacent memory dots.

A second major breakthrough is the demonstration of the concept of Perpendicular Shape Anisotropy STT-MRAM. By dramatically increasing the thickness of the storage layer (thickness comparable to diameter), we have shown that we can make functional memory dots of diameter down to 8nm. The perpendicular shape anisotropy provides a very robust source of bulk anisotropy to the storage layer magnetization which allows to extend the scalability of STT-MRAM down to 4nm. Besides, this anisotropy decreases with temperature much more slowly than in conventional STT-MRAM which is very good for applications operating on a wide range of temperature such as in automotive industry.

Another breakthrough result concerns short distance communication using spin transfer torque oscillators. We demonstrated voice tranmission over tens of meters by using the phase modulation of vortex spin-transfer-torque nano-oscillators. For that, the spin transfer oscillators is synchronised on an external source whose frequency is modulated by the signal to be transmitted. The modulated signal is then amplified and emitted through a RF antenna. It is then detected and demodulated by a RF rectifying device at distances exceeding several tens of meters. A new concept of spectrum analyzer using these high quality spin transfer oscillators also emerged from the project.

Another important result of the project is the realization of memristive function for neuromorphic computing using magnetic tunnel junctions. For the first time, we demonstrated that the angular variation of the conductance as a function of the angle between the storage layer and reference layer magnetization can be used to achieve intermediate levels of resistance which can be accessed by pulses of current flowing through the junction. This allows to realize spintronic memristors which can be used as artificial synapses in artificial intelligence.

Another important aspect of this project is to foster more relationships between magnetism and microelectronics communities. Along this line, several events were organized such as the Introductory Course on MRAM (session in 2017 and 2018) aiming to give to researchers, engineers, students not specialist in the field the basis to understand the physics, working principles, materials involved, fabrication process of this new class of magnetic memories. InMRAM is being organized again in July 2020. In addition, a MRAM special poster session and a MRAM Global Innovation Forum gathering 300 attendees were organized by the PI at IEDM 2017, IEDM 2018 and again at IEDM 2019. These events contribute reducing the cultural gap between these two communities and ease the penetration of this hybrid CMOS/magnetic technology in microelectronics industry.
In addition to the breakthrough mentioned above, several optimization of magnetic tunnel junction stacks were realized in the project aiming to increase the stability of the synthetic antiferromagnetic reference layer and at thinning the stack to ease its etching.

In the second half of the project, efforts will be focused on the use of this hybrid technology for other functionalities: magnetic field sensing, communications using spin transfer torque oscillators and spin diodes as well as memristors for neuromorphic circuits. This will demonstrate the multifunctionality of spintronic devices.
Illustration of Perpendicular Shape Anisotropy STT-MRAM concept and examples of properties