Final Report Summary - HYMAGINE (Hybrid CMOS/Magnetic components and systems for energy efficient, non-volatile, reprogrammable integrated electronics)
With the steady reduction in size of the microelectronic components, conventional Complementary Metal Oxide Semiconductor (CMOS) technology is facing challenges of increasing difficulty. Among those, the power consumption of electronic circuits and their associated heating is particularly severe. This consumption has both a static origin (associated with the leakage of the transistors) and a dynamic origin (associated with the capacitive loss due to the charging and discharging of interconnects).
Spin-electronics can help circumventing some of these limitations faced by purely CMOS microelectronics. This is the central motivation of this ERC HYMAGINE project. The purpose is not to replace CMOS components with magnetic elements but to combine them in order to get the best of the two worlds. On one hand, CMOS microelectronics is particularly efficient for logic operations thanks to the very low energy and very high speed to switch ON/OFF CMOS transistors. On the other hand, magnetism is very good for memory applications because the information can be easily stored in a reliable and non-volatile way (i.e. without requiring to be electrically powered) through a magnetization orientation. This is why our computers have been using hard disk drives for more than 60 years.
The idea of the HYMAGINE project “Hybrid CMOS/Magnetic components and systems for energy efficient, non-volatile, reprogrammable integrated electronics” was to explore the potentialities of this hybrid CMOS/magnetic integrated technology for microelectronic circuits. For this, we take advantage of the breakthrough discoveries which were made in spin-electronics in the past 15 years in particular the very large tunnel magnetoresistance of MgO based magnetic tunnel junctions (MTJ) and the phenomenon of spin-transfer torque (STT) which is the action that a spin-polarized current can exert on the magnetization of a magnetic nanostructure.
Magnetic tunnel junctions can be easily integrated with CMOS technology with matching impedances and no cross-contamination. They can be deposited and patterned above the CMOS technological levels allowing vertical integration of CMOS logic circuits with magnetic memory elements. This hybrid technology allows bringing non-volatility very close to logic circuits. In other words, memories can be embedded in logic circuits with a very fine granularity. This opens new strategies of power gating, resulting in much reduced static power consumption. The transfer of data between logic and memory can also be made much more efficient (short vertical transfer with large bandwidth instead of long horizontal transfer with low bandwidth) thereby reducing the dynamic power consumption. A new concept of Normally-off/Instant-on electronics has emerged along these lines. Other benefits of this hybrid CMOS/magnetic technology are the possibility of on-fly ultrafast reprogramability (i.e. capability to change the functionality of a given circuit just by modifying the magnetic state of the magnetic tunnel junction) and the resistance to ionizing radiation which is a strong asset for space applications.
Within HYMAGINE, we have investigated routes to improve the reliability of magnetic tunnel junctions and particularly their resistance to dielectric breakdown. Clear routes to increase the write endurance in these junctions have been proposed. New concepts of MTJ stacks were proposed and studied allowing ultrafast magnetic switching (100ps) for SRAM applications. The design tools for conceiving these hybrid CMOS/magnetic circuits were developed and are now fully operational under microelectronics design platform such as CADENCE. Thanks to these developments completed by an ERC PoC, a fabless start-up of design of low-power CMOS/MTJ circuits (in particular microcontrollers for interconnected objects) was launched in 2014 (EVADERIS).
Circuits of increasing complexity have then been designed and built: Magnetic look up table, magnetic FPGA, 1Mbit STT-MRAM chips, numerical filters.
Another goal of this project was to reduce the cultural gap between microelectronics and magnetism communities by reciprocal educational actions (joint symposia, tutorials, summer schools, edition of introductory books…) in order to ease the implementation of this new technology. This was achieved through several actions in particular the launching of an annual “Introductory Course on MRAM” which takes place in Grenoble every year since 2013.
Spin-electronics can help circumventing some of these limitations faced by purely CMOS microelectronics. This is the central motivation of this ERC HYMAGINE project. The purpose is not to replace CMOS components with magnetic elements but to combine them in order to get the best of the two worlds. On one hand, CMOS microelectronics is particularly efficient for logic operations thanks to the very low energy and very high speed to switch ON/OFF CMOS transistors. On the other hand, magnetism is very good for memory applications because the information can be easily stored in a reliable and non-volatile way (i.e. without requiring to be electrically powered) through a magnetization orientation. This is why our computers have been using hard disk drives for more than 60 years.
The idea of the HYMAGINE project “Hybrid CMOS/Magnetic components and systems for energy efficient, non-volatile, reprogrammable integrated electronics” was to explore the potentialities of this hybrid CMOS/magnetic integrated technology for microelectronic circuits. For this, we take advantage of the breakthrough discoveries which were made in spin-electronics in the past 15 years in particular the very large tunnel magnetoresistance of MgO based magnetic tunnel junctions (MTJ) and the phenomenon of spin-transfer torque (STT) which is the action that a spin-polarized current can exert on the magnetization of a magnetic nanostructure.
Magnetic tunnel junctions can be easily integrated with CMOS technology with matching impedances and no cross-contamination. They can be deposited and patterned above the CMOS technological levels allowing vertical integration of CMOS logic circuits with magnetic memory elements. This hybrid technology allows bringing non-volatility very close to logic circuits. In other words, memories can be embedded in logic circuits with a very fine granularity. This opens new strategies of power gating, resulting in much reduced static power consumption. The transfer of data between logic and memory can also be made much more efficient (short vertical transfer with large bandwidth instead of long horizontal transfer with low bandwidth) thereby reducing the dynamic power consumption. A new concept of Normally-off/Instant-on electronics has emerged along these lines. Other benefits of this hybrid CMOS/magnetic technology are the possibility of on-fly ultrafast reprogramability (i.e. capability to change the functionality of a given circuit just by modifying the magnetic state of the magnetic tunnel junction) and the resistance to ionizing radiation which is a strong asset for space applications.
Within HYMAGINE, we have investigated routes to improve the reliability of magnetic tunnel junctions and particularly their resistance to dielectric breakdown. Clear routes to increase the write endurance in these junctions have been proposed. New concepts of MTJ stacks were proposed and studied allowing ultrafast magnetic switching (100ps) for SRAM applications. The design tools for conceiving these hybrid CMOS/magnetic circuits were developed and are now fully operational under microelectronics design platform such as CADENCE. Thanks to these developments completed by an ERC PoC, a fabless start-up of design of low-power CMOS/MTJ circuits (in particular microcontrollers for interconnected objects) was launched in 2014 (EVADERIS).
Circuits of increasing complexity have then been designed and built: Magnetic look up table, magnetic FPGA, 1Mbit STT-MRAM chips, numerical filters.
Another goal of this project was to reduce the cultural gap between microelectronics and magnetism communities by reciprocal educational actions (joint symposia, tutorials, summer schools, edition of introductory books…) in order to ease the implementation of this new technology. This was achieved through several actions in particular the launching of an annual “Introductory Course on MRAM” which takes place in Grenoble every year since 2013.