Periodic Reporting for period 2 - SKYNOLIMIT (Ultralow power and ultra-wideband spintronics near thermodynamic limits)
Reporting period: 2022-08-01 to 2024-01-31
The issue being addressed: Spintronic logic and memory devices, which might operate with ultra-wideband and ultra-low power near thermodynamic limits, cannot yet benefit microelectronics or public.
Microelectronics industry currently has multiple grand challenges, which are (1) the end of Moore’s Law and transistor scaling, (2) exponentially growing power consumption, (3) memory access and bandwidth. State-of-the-art electronics is based on field-effect transistors, whose dimensions or power consumption could no longer be reduced. Transistors consume ~10^4 kBT energy per switching, which limits virtually all mobile device operations. As external random access memory increasingly takes greater share in the integrated circuit power budget, in-memory computation with minimal power consumption and reduced footprint started becoming essential especially for battery-constrained internet-of-things. Moore’s Law nears the end and fundamental device breakthroughs are needed for power-efficient, non-volatile computational architectures to take over transistors and significantly reduce the energy consumption while meeting the speed, bandwidth and scalability requirements of microelectronics industry. Spintronics emerged as an area of very intense fundamental and device research since it offers orders of magnitude improvements with respect to the state-of-the-art nanoelectronic switching rates and energies with the recently developed magnetic materials. In this ERC project, we create new nanoscale spintronic materials that enable non-volatile ultra-wideband logic and circuits with ultralow power consumption near thermodynamic limits. We demonstrate in modelling and experiments spintronic devices and circuits made of 2D multilayers of topological insulators and transition metal dichalcogenides on insulating magnetic iron garnet films with almost no Joule heating; carry signals on nanoscale spin structures called skyrmions; and I demonstrate their energy and speed advantages in Fourier transformer and neural network circuits.
Importance for the society:
This project has four ground-breaking aspects:
(1) Scientific breakthroughs: By computationally and experimentally investigating the electronic and magnetic transport properties of multilayers of new 2D materials, I have been discovering new topologically protected spintronic materials, device models and device physics. I previously demonstrated such topologically protected transport phenomena in topological insulator/garnet layers experimentally and (magneto)plasmonic excitons (plexcitons) in computational models. Since even the most basic 2D layer graphene can still lead to the highly counterintuitive discovery of superconductivity in two graphene layers stacked on top of each other by 1° misalignment, 2D materials with more complex chemistries investigated in this study could well lead to many more fundamental discoveries and unique devices.
(2) Breakthrough in spin injection/detection efficiencies with 2D materials and magnetic insulator garnets: Giant spin-orbit coupling in 2D materials leads to giant spin Hall angles along 2D interfaces. Spin-orbit coupling is the phenomenon that couples the charge and spin degrees of freedom for electrons and helps convert between electronic charge and spin currents. Spin injection and detection efficiencies could be increased using these 2D materials by a few orders of magnitude compared with traditional spintronic interfaces like YIG (Y3Fe5O12)/Pt. Thus, SKYNOLIMIT project could eliminate this main limitation that has been preventing spintronics from becoming a mainstream low-power chip technology.
(3) Ultralow power logic circuit and memory using nanoscale skyrmions near thermodynamic limits could help reach truly power efficient computation at room temperature. Thus, circuit-scale use of ultralow power skymion signal processors could reduce mobile device power consumption & extend battery lifetimes by a few orders of magnitude. I previously demonstrated that we can generate broadband (7+ octaves) and GHz skyrmions with energy costs (65kBT at 300 K) near thermodynamic limits (kBTln(2)).
(4) By building a compact spintronic device model library and a software for spintronic circuit modelling and layout design, we modularized and accelerated spintronic circuit and chip designs.
Overall objectives of the project:
In SKYNOLIMIT project, I aim to solve the problems by computationally designing and experimentally demonstrating ultra-wideband (20+ GHz), ultralow-power (~aJ/bit) and non-volatile spintronic logic and memory architectures that operate based on nanoscale spins called skyrmions. Skyrmions are nanoscale circular spin structures that allow for room temperature computation and memory near thermodynamic limits while being robust against fabrication imperfections. The project acronym (SKYNOLIMIT) originates from “skyrmion-based non-volatile materials and devices near thermodynamic limits.”
In this project, I first design the novel functional 2D multilayers of topological insulators (i.e. Bi2Se3, Bi2Te3, BiSb) and transition metal dichalcogenides (i.e. MoS2, WTe2, MoSe2) on garnet films (i.e. Tm3Fe5O12, Tb3Fe5O12 and other chemistries) with giant spin-orbit coupling and large Dzyaloshinskii-Moriya effects for all-electric generation/detection of skyrmion currents. In the second step, we design, fabricate and demonstrate the non-volatile skyrmion devices including logic gates and registers using the new materials. Finally, we integrate skyrmion devices in large scale and demonstrate efficient arithmetic & logic units, Fourier transformers and non-volatile neural network circuits. Despite its outstanding ambition, this project is feasible since we published on each method of the project and developed a detailed risk management plan to solve all possible fundamental and technical hurdles.
We refer to topological insulators and transition metal dichalcogenides as “2D materials” since these layers have few monolayers. Topological insulators allow for voltage control of spins of only surface electrons with large spin Hall angles while transition metal dichalcogenides (i.e. MoS2, WTe2, MoSe2) allow for large spin injection efficiencies on a magnetic layer. Insulating rare earth iron garnet films support and carry skyrmions along their interfaces with topological insulators and transition metal dichalcogenides. The insulator garnet layers have low Gilbert damping and negligible Joule heating which allows for low-loss transport of skyrmion spin signals on chip.
Microelectronics industry currently has multiple grand challenges, which are (1) the end of Moore’s Law and transistor scaling, (2) exponentially growing power consumption, (3) memory access and bandwidth. State-of-the-art electronics is based on field-effect transistors, whose dimensions or power consumption could no longer be reduced. Transistors consume ~10^4 kBT energy per switching, which limits virtually all mobile device operations. As external random access memory increasingly takes greater share in the integrated circuit power budget, in-memory computation with minimal power consumption and reduced footprint started becoming essential especially for battery-constrained internet-of-things. Moore’s Law nears the end and fundamental device breakthroughs are needed for power-efficient, non-volatile computational architectures to take over transistors and significantly reduce the energy consumption while meeting the speed, bandwidth and scalability requirements of microelectronics industry. Spintronics emerged as an area of very intense fundamental and device research since it offers orders of magnitude improvements with respect to the state-of-the-art nanoelectronic switching rates and energies with the recently developed magnetic materials. In this ERC project, we create new nanoscale spintronic materials that enable non-volatile ultra-wideband logic and circuits with ultralow power consumption near thermodynamic limits. We demonstrate in modelling and experiments spintronic devices and circuits made of 2D multilayers of topological insulators and transition metal dichalcogenides on insulating magnetic iron garnet films with almost no Joule heating; carry signals on nanoscale spin structures called skyrmions; and I demonstrate their energy and speed advantages in Fourier transformer and neural network circuits.
Importance for the society:
This project has four ground-breaking aspects:
(1) Scientific breakthroughs: By computationally and experimentally investigating the electronic and magnetic transport properties of multilayers of new 2D materials, I have been discovering new topologically protected spintronic materials, device models and device physics. I previously demonstrated such topologically protected transport phenomena in topological insulator/garnet layers experimentally and (magneto)plasmonic excitons (plexcitons) in computational models. Since even the most basic 2D layer graphene can still lead to the highly counterintuitive discovery of superconductivity in two graphene layers stacked on top of each other by 1° misalignment, 2D materials with more complex chemistries investigated in this study could well lead to many more fundamental discoveries and unique devices.
(2) Breakthrough in spin injection/detection efficiencies with 2D materials and magnetic insulator garnets: Giant spin-orbit coupling in 2D materials leads to giant spin Hall angles along 2D interfaces. Spin-orbit coupling is the phenomenon that couples the charge and spin degrees of freedom for electrons and helps convert between electronic charge and spin currents. Spin injection and detection efficiencies could be increased using these 2D materials by a few orders of magnitude compared with traditional spintronic interfaces like YIG (Y3Fe5O12)/Pt. Thus, SKYNOLIMIT project could eliminate this main limitation that has been preventing spintronics from becoming a mainstream low-power chip technology.
(3) Ultralow power logic circuit and memory using nanoscale skyrmions near thermodynamic limits could help reach truly power efficient computation at room temperature. Thus, circuit-scale use of ultralow power skymion signal processors could reduce mobile device power consumption & extend battery lifetimes by a few orders of magnitude. I previously demonstrated that we can generate broadband (7+ octaves) and GHz skyrmions with energy costs (65kBT at 300 K) near thermodynamic limits (kBTln(2)).
(4) By building a compact spintronic device model library and a software for spintronic circuit modelling and layout design, we modularized and accelerated spintronic circuit and chip designs.
Overall objectives of the project:
In SKYNOLIMIT project, I aim to solve the problems by computationally designing and experimentally demonstrating ultra-wideband (20+ GHz), ultralow-power (~aJ/bit) and non-volatile spintronic logic and memory architectures that operate based on nanoscale spins called skyrmions. Skyrmions are nanoscale circular spin structures that allow for room temperature computation and memory near thermodynamic limits while being robust against fabrication imperfections. The project acronym (SKYNOLIMIT) originates from “skyrmion-based non-volatile materials and devices near thermodynamic limits.”
In this project, I first design the novel functional 2D multilayers of topological insulators (i.e. Bi2Se3, Bi2Te3, BiSb) and transition metal dichalcogenides (i.e. MoS2, WTe2, MoSe2) on garnet films (i.e. Tm3Fe5O12, Tb3Fe5O12 and other chemistries) with giant spin-orbit coupling and large Dzyaloshinskii-Moriya effects for all-electric generation/detection of skyrmion currents. In the second step, we design, fabricate and demonstrate the non-volatile skyrmion devices including logic gates and registers using the new materials. Finally, we integrate skyrmion devices in large scale and demonstrate efficient arithmetic & logic units, Fourier transformers and non-volatile neural network circuits. Despite its outstanding ambition, this project is feasible since we published on each method of the project and developed a detailed risk management plan to solve all possible fundamental and technical hurdles.
We refer to topological insulators and transition metal dichalcogenides as “2D materials” since these layers have few monolayers. Topological insulators allow for voltage control of spins of only surface electrons with large spin Hall angles while transition metal dichalcogenides (i.e. MoS2, WTe2, MoSe2) allow for large spin injection efficiencies on a magnetic layer. Insulating rare earth iron garnet films support and carry skyrmions along their interfaces with topological insulators and transition metal dichalcogenides. The insulator garnet layers have low Gilbert damping and negligible Joule heating which allows for low-loss transport of skyrmion spin signals on chip.
This project consists of three parts: (1) modelling materials, devices and circuits, (2) Micro/nanofabrication and characterization of materials and devices, and (3) X-ray and magnetic microscopy and electrical device tests.
In the first half of the project period, we formed the team, bought and set up the major equipment and started giving trainings to our graduate students. In this period, our results can be quantitatively summarized as 6 high-impact journal papers, 1 book chapter, 3 PhD thesis dissertations and about 16 conference/seminar presentations published and filed a patent application. We estimate that these deliverables constitute a 20-25% of all deliverables that we expect to publish from this ERC project.
In the first part, where we have been modelling materials, we already achieved several unexpected breakthrough materials and device designs. In first-principles atomic modelling, we identified previously unknown but stable 2D materials phases that are close to each other but highly different in band gap, topological number, density of states and transport characteristics. These results that are currently in preparation for publication guide epitaxial growth processes where metastable phases or coalescence of these different phases may prevent the experimental realization of the predicted merits of 2D materials. In response to this bottleneck, we have been developing new MBE recipes that achieve single phase growth and overcome multiple phase formation as well as multiphase coalescence.
In the modelling of devices and circuits, we achieved and published several breakthroughs based on micromagnetic modelling of skyrmion-based spintronic circuits. In the first batch of results, we demonstrated a digital skyrmion clock source as a component whose output frequency can be tuned from 100 MHz to 20 GHz using DC current intensity and no external magnetic field. Next, we demonstrated a digital skyrmion-based logic inverter gate that has potential for ultralow energy consumption. The device retains information as a nonvolatile memory element unlike their state-of-the-art microelectronic counterparts (logic NOT gate). We analyzed its full operation using micromagnetic modeling. We also prepared thermal models and found that even all-metallic inverter gates can operate with thermal equilibrium despite large Joule heating. Because of the substrate thermal conductivity, our investigations revealed that the all-metallic inverter gate can function with direct current drive, wide bandwidth, submicron footprint, no or low external magnetic field, and cascadability. By substituting the metallic magnetic layers with magnetic insulators for eliminating Joule heating and lowering the exchange stiffness, we found that energy consumption might be reduced even further by more than four orders of magnitude than all-metallic skyrmion NOT gates. These results suggest that reducing the energy consumption of microelectronic NOT gates by two orders of magnitude using magnetic insulator-based skyrmion logic gates might be feasible.
We demonstrated the entire set of digital skyrmion logic gates to construct full adders, arithmetic and logic units as well as arbitrary combinational or sequential digital logic circuits. We filed a patent application on the subject. We are currently preparing the manuscript of this part of the circuit models for publication as of Fall 2023.
In the second part, where we do growth, micro and nanofabrication of materials and devices, we have started achieving several milestones. The first milestone is the delivery and setup of the molecular beam epitaxy and sputtering systems in our lab. We trained a team of 15 MS students, PhD students and post-docs and they have started doing topological insulator thin film growth as well as sputtering of thin films. The process iterations accelerated the materials trainings of the students as they studied the structure-processing-property relationships further with detailed characterization.
In the micro and nanofabrication part, we trained a PhD student and she developed and refined the lithography process steps needed for coplanar waveguide fabrication and device patterning. We developed several processes that use either maskless optical or e-beam lithography.
In the third part, where we do x-ray and magnetic microscopy and electrical device tests, we have been setting up the device test equipment with vector network analyzer, RF transmission cables, bias T and other components. In addition, we submitted several proposals to European Synchrotron Radiation Facility (ESRF) and organized invited seminars by expert beamline scientists from ESRF and Argonne National Lab. These works are still in progress and with refinement of the materials and the nanofabrication processes, the x-ray microscopy and electric tests are going to be progress faster.
In the first half of the project period, we formed the team, bought and set up the major equipment and started giving trainings to our graduate students. In this period, our results can be quantitatively summarized as 6 high-impact journal papers, 1 book chapter, 3 PhD thesis dissertations and about 16 conference/seminar presentations published and filed a patent application. We estimate that these deliverables constitute a 20-25% of all deliverables that we expect to publish from this ERC project.
In the first part, where we have been modelling materials, we already achieved several unexpected breakthrough materials and device designs. In first-principles atomic modelling, we identified previously unknown but stable 2D materials phases that are close to each other but highly different in band gap, topological number, density of states and transport characteristics. These results that are currently in preparation for publication guide epitaxial growth processes where metastable phases or coalescence of these different phases may prevent the experimental realization of the predicted merits of 2D materials. In response to this bottleneck, we have been developing new MBE recipes that achieve single phase growth and overcome multiple phase formation as well as multiphase coalescence.
In the modelling of devices and circuits, we achieved and published several breakthroughs based on micromagnetic modelling of skyrmion-based spintronic circuits. In the first batch of results, we demonstrated a digital skyrmion clock source as a component whose output frequency can be tuned from 100 MHz to 20 GHz using DC current intensity and no external magnetic field. Next, we demonstrated a digital skyrmion-based logic inverter gate that has potential for ultralow energy consumption. The device retains information as a nonvolatile memory element unlike their state-of-the-art microelectronic counterparts (logic NOT gate). We analyzed its full operation using micromagnetic modeling. We also prepared thermal models and found that even all-metallic inverter gates can operate with thermal equilibrium despite large Joule heating. Because of the substrate thermal conductivity, our investigations revealed that the all-metallic inverter gate can function with direct current drive, wide bandwidth, submicron footprint, no or low external magnetic field, and cascadability. By substituting the metallic magnetic layers with magnetic insulators for eliminating Joule heating and lowering the exchange stiffness, we found that energy consumption might be reduced even further by more than four orders of magnitude than all-metallic skyrmion NOT gates. These results suggest that reducing the energy consumption of microelectronic NOT gates by two orders of magnitude using magnetic insulator-based skyrmion logic gates might be feasible.
We demonstrated the entire set of digital skyrmion logic gates to construct full adders, arithmetic and logic units as well as arbitrary combinational or sequential digital logic circuits. We filed a patent application on the subject. We are currently preparing the manuscript of this part of the circuit models for publication as of Fall 2023.
In the second part, where we do growth, micro and nanofabrication of materials and devices, we have started achieving several milestones. The first milestone is the delivery and setup of the molecular beam epitaxy and sputtering systems in our lab. We trained a team of 15 MS students, PhD students and post-docs and they have started doing topological insulator thin film growth as well as sputtering of thin films. The process iterations accelerated the materials trainings of the students as they studied the structure-processing-property relationships further with detailed characterization.
In the micro and nanofabrication part, we trained a PhD student and she developed and refined the lithography process steps needed for coplanar waveguide fabrication and device patterning. We developed several processes that use either maskless optical or e-beam lithography.
In the third part, where we do x-ray and magnetic microscopy and electrical device tests, we have been setting up the device test equipment with vector network analyzer, RF transmission cables, bias T and other components. In addition, we submitted several proposals to European Synchrotron Radiation Facility (ESRF) and organized invited seminars by expert beamline scientists from ESRF and Argonne National Lab. These works are still in progress and with refinement of the materials and the nanofabrication processes, the x-ray microscopy and electric tests are going to be progress faster.
Overall, we achieved the following in the first half of the project:
- We prepared detailed micromagnetic models, power consumption models, thermal stability models and geometric sensitivity models for skyrmionic devices.
- We established the operation principles for skyrmion-based nonvolatile logic devices that are ultracompact, fast and ultra-low power. These elements are cascadable and could be used for designing arbitary digital sequential or combinational logic circuit.
- We developed an emulator software for modelling and developing skyrmion circuits without initially running any micromagnetic model.
- We are completing the code for translating VHDL/Verilog scripts and their synthesized RTL schematics into skyrmion logic circuits. Thus, the software will do design rule check before fabrication and tapeout. The ultimate output of the software is the mask design and functional, timing and power analysis of device operation resulting from a given floor plan.
- We are starting to share our samples with colleagues across the Europe (i.e. national metrology institutes, beamlines), China and Japan.
- We filed a patent application on the skyrmion logic device library. This patent application is going to be the first of the 5 patent applications related to the project’s skyrmion device designs (drive electronics, 3D integration, garnet/TI and garnet/TMD device types, interfacing with common protocols such as SPI and I2C).
- The next 5 patent applications are going to be related to the new stoichiometries, multilayers and their unique properties related to data storage, sensing and computation.
- We started working with a microelectronics company (Yongatek) to design FPGA chips with in-memory computing. While their designs do not yet use skyrmions, their architectures started bringing more registers within a sub-unit together with the arithmetic and logic units for neural network processing with low latency.
- We advise Koç Holding companies on novel quantum materials and quantum computing.
To be more specific, we have the following 5 novel achievements until now:
1) In Phys. Rev. B 105, 054411 (2022), we demonstrated skyrmion logic inverters and we also demonstrated all of the skyrmion-based logic gates in digital design. We found that these gates consume more than 4 orders of magnitude lower energy than when we substitute metallic magnetic materials with magnetic insulators. The energy consumptions of logic operations can be 2 orders of magnitude lower than those for the state-of-the-art transistors (10000-50000 kT) and near the thermodynamic Landau limit (kTln(2)) as promised in the ERC project.
2) We found a major issue in the field that none of the papers pointed out previously. We found that polycrystallinity in nanowires of magnetic materials can prevent the propagation of skyrmion or even destroy the skyrmions due to the strong near-field total magnetic energy gradients. We discussed this issue in our Physical Review B 105 (5), 054411 (2022) paper. To resolve these issues, the nanowires must be single crystal with uniform magnetic properties to achieve stable skyrmion logic operation.
3) Previous skyrmion publications did not include rigorous energy consumption and temperature stability analyses. We demonstrated that temperature can be stabilized by allowing substrate to conduct heat without requiring any active cooling.
4) In Nanoscale Adv., 5, 4470 (2023), we showed that synthetic antiferromagnetically coupled (SAF) multilayers can be used for eliminating skyrmion Hall effect or topological Hall effect of skyrmions. This is a major step towards long distance propagation of nanoscale skyrmions on chip over many millimeters. These current densities could be reduced to 10^8 A m−2, while 10^11 A m^−2 or above is needed for ferromagnetic skyrmions. Thus, we showed that SAF skyrmions may dissipate as much as 6 orders of magnitude lower energy than ferromagnetic skyrmions, despite still containing metallic magnetic multilayers.
5) We found that the current-driven skyrmion velocities reach ∼200 m s^−1 without skyrmion Hall effect in synthetic antiferromagnetic layers. These velocities are 3–10 times greater than the typical ferromagnetic skyrmion velocities. By reducing the SAF skyrmion drive current by 3 orders, Joule heating is reduced by 6 orders of magnitude. These velocities are sufficient to meet common data transfer protocols in microelectronics, such as SPI and I2C. These results highlight that once the intermediate technical challenges are resolved, the SAF skyrmions may be integrated with microelectronics.
- We prepared detailed micromagnetic models, power consumption models, thermal stability models and geometric sensitivity models for skyrmionic devices.
- We established the operation principles for skyrmion-based nonvolatile logic devices that are ultracompact, fast and ultra-low power. These elements are cascadable and could be used for designing arbitary digital sequential or combinational logic circuit.
- We developed an emulator software for modelling and developing skyrmion circuits without initially running any micromagnetic model.
- We are completing the code for translating VHDL/Verilog scripts and their synthesized RTL schematics into skyrmion logic circuits. Thus, the software will do design rule check before fabrication and tapeout. The ultimate output of the software is the mask design and functional, timing and power analysis of device operation resulting from a given floor plan.
- We are starting to share our samples with colleagues across the Europe (i.e. national metrology institutes, beamlines), China and Japan.
- We filed a patent application on the skyrmion logic device library. This patent application is going to be the first of the 5 patent applications related to the project’s skyrmion device designs (drive electronics, 3D integration, garnet/TI and garnet/TMD device types, interfacing with common protocols such as SPI and I2C).
- The next 5 patent applications are going to be related to the new stoichiometries, multilayers and their unique properties related to data storage, sensing and computation.
- We started working with a microelectronics company (Yongatek) to design FPGA chips with in-memory computing. While their designs do not yet use skyrmions, their architectures started bringing more registers within a sub-unit together with the arithmetic and logic units for neural network processing with low latency.
- We advise Koç Holding companies on novel quantum materials and quantum computing.
To be more specific, we have the following 5 novel achievements until now:
1) In Phys. Rev. B 105, 054411 (2022), we demonstrated skyrmion logic inverters and we also demonstrated all of the skyrmion-based logic gates in digital design. We found that these gates consume more than 4 orders of magnitude lower energy than when we substitute metallic magnetic materials with magnetic insulators. The energy consumptions of logic operations can be 2 orders of magnitude lower than those for the state-of-the-art transistors (10000-50000 kT) and near the thermodynamic Landau limit (kTln(2)) as promised in the ERC project.
2) We found a major issue in the field that none of the papers pointed out previously. We found that polycrystallinity in nanowires of magnetic materials can prevent the propagation of skyrmion or even destroy the skyrmions due to the strong near-field total magnetic energy gradients. We discussed this issue in our Physical Review B 105 (5), 054411 (2022) paper. To resolve these issues, the nanowires must be single crystal with uniform magnetic properties to achieve stable skyrmion logic operation.
3) Previous skyrmion publications did not include rigorous energy consumption and temperature stability analyses. We demonstrated that temperature can be stabilized by allowing substrate to conduct heat without requiring any active cooling.
4) In Nanoscale Adv., 5, 4470 (2023), we showed that synthetic antiferromagnetically coupled (SAF) multilayers can be used for eliminating skyrmion Hall effect or topological Hall effect of skyrmions. This is a major step towards long distance propagation of nanoscale skyrmions on chip over many millimeters. These current densities could be reduced to 10^8 A m−2, while 10^11 A m^−2 or above is needed for ferromagnetic skyrmions. Thus, we showed that SAF skyrmions may dissipate as much as 6 orders of magnitude lower energy than ferromagnetic skyrmions, despite still containing metallic magnetic multilayers.
5) We found that the current-driven skyrmion velocities reach ∼200 m s^−1 without skyrmion Hall effect in synthetic antiferromagnetic layers. These velocities are 3–10 times greater than the typical ferromagnetic skyrmion velocities. By reducing the SAF skyrmion drive current by 3 orders, Joule heating is reduced by 6 orders of magnitude. These velocities are sufficient to meet common data transfer protocols in microelectronics, such as SPI and I2C. These results highlight that once the intermediate technical challenges are resolved, the SAF skyrmions may be integrated with microelectronics.