Final Report Summary - M3D (Materials for a Magnetic Memory in Three Dimensions)
Executive Summary:
Until recently, all commercial storage technologies were based on storage at a surface (chip, disk, tape). While the growth of bit density and the associated decrease of price per bit have been following Moore's law for decades, clear physical limitations are currently putting a final end to this growth. M3d pertains to exploratory work on materials science technology, with a view to make possible a change of paradigm in data storage technology. The background idea is the proposal for a so-called 3D race-track memory, put forward by IBM.
The consortium gathered physicists of spintronics and technology, and chemists from the 3D bottom-up synthesis. Two SMEs were associated, one for the synthesis, another one for contributing to making the most relevant choices in terms of magnetic technology. The successful output of the project highlights to the academic sector, industry and society, the potential of innovation arising from interdisciplinarity.
Our actions have been several fold, unlocking several bottlenecks on the material’s side: 1. Synthesis of suitable materials in a wire form 2 Design of a digital media, ie with geometrical notches or segments 3 Evaluation and elimination of cross-talk, specific to magnetic materials in a 3D geometry 4. Outlining through simulation the potential for close-to-ns operation in single wires; preliminary results hint at the practical feasibility 5. Explore routes for reading and writing elements. The former has shown encouraging results, based on the reading a tunnel magneto-resistive element embedded in the wire. The latter has been attempted by heat assistance, however was not successful.
The academic output is very positive, and has triggered a serious attention to magnetic nanowires and nanotubes, and more generally to nanomagnetism and spintronics in a 3D geometry. Several invited reviews and chapters (Nature Communications; Handbook of Magnetic Materials series etc.) have been requested from partners of the consortium. New consortia are already being formed for future projects, of similar or higher TRL. From the point of view of data storage industry, this development is very timely as other technologies for storage are moving to 3D designs, such as flash memories, and those doing so are clearly winning the race of performance. In the future, 3D magnetic technologies promise the potential for high performance combined with extremely long endurance and non-volatility. Thus, the original strategy of M3d is being perfectly met, to highlight the potential to foster future more applied work in the topic.
Direct exploitation is also coming immediately from technical developments carried out during the project. This includes nanoporous templates with a 3D engineering, including modulations of diameter in the depth (application to filtration photonics etc); software driving lithography tools; new atomic layer deposition processes; high added value tips for magnetic force microscopy (high spatial resolution and low moment); a micromagnetic software available open source, and whose support may turn commercial in the mid-term.
Project Context and Objectives:
M3d pertains to exploratory work on materials science, with a view to introduce a change of paradigm in data storage technology. Until recently, all commercial storage technologies were based on storage at a surface (chip, disk, tape). While the growth of bit density and the associated decrease of price per bit have been following Moore's law for decades, clear physical limitations are currently putting a final end to this growth. Multi-level storage are being implemented to sustain this growth rate, however the gains are only incremental and imply an increase of cost and complexity as each level needs to be patterned one atop another.
Our project sets out to break through existing material limitations in order to develop large-scale non-volatile data storage massively relying on the use of the third dimension (3d). We follow the IBM concept of magnetic race-track memory, where series of bits would be shifted along vertical magnetic wires densely packed in arrays, requiring only one single read/write element per wire [S. S. P. Parkin et al., Magnetic Domain-Wall Racetrack Memory, Science 320, 190 (2008)]. This allows us to address the need for media density above 5Tbit/in2 when expressed in the usual terms of area, and reasonable cost per Tbit, whilst retaining the potential for an excellent access speed and energy consumption.
To keep costs moderate the 3D synthesis of a dense array of magnetic nanowires shall be done in a limited number of stages and therefore rely on bottom-up routes. This must be complemented by top-down surface patterning, for eg appending one read/write element per wire. In addition, to minimize risks several strategies shall be explored in parallel for coding bits along the magnetic wires. This includes magnetic solitons in stacked thin disks, solitons in segmented long wires, and magnetic domain walls in continuous wires. In the course of the project, we refocused work on the latter two, based on progress in synthesis and simulations. We scheduled addressing (writing & reading) strategies in the second half of the project.
We gathered partners with the complementary expertise required for these developments in the consortium: Institut NEEL and SPINTEC (CNRS), Cavendish Laboratories and Univ. Hamburg bring in advanced expertise in nanomagnetism and spintronics to implement various routes for coding and addressing bits. Univ. Hamburg, Univ. Erlangen-Nürnberg bring in a cutting-edge expertise in 3D bottom-up design of porous media. Synchrotron SOLEIL provides various instruments for microscopic magnetic investigation of the devices designed.
Through this exploratory work, we aim at pushing Europe ahead of Asia and the USA in 3D storage, an area with a high potential for added value and innovation. With this goal in mind, the SME company SmartMembranes with world-leading expertise in bottom-up products is an active partner in the consortium. Its aim is to boost the efficiency of the academic partners by providing suitable templates for the fabrication of dense arrays of nanowires. Conversely, SmartMembranes shall benefit from the advancement of expertise from those partners, thereby enlarging its portfolio of state-of-the art products after up-scaling the synthesis processes. Second, participation of the data storage industrial partner CROCUS Technology shall provide the consortium with a clear view on relevant current technology and patents, also creating a link that may be strengthened in a future R&D device project.
Project Results:
Early in the project, simulations have been running, to outline features and operating schemes of three possible routes to code information, based on domain walls or solitons. These revealed that only a narrow operating window would exist for transverse solitons, especially when taking into account the consequences of inserting a read element in the middle of a stack. Thus, we focused experimental work on two coding routes only: longitudinal solitons, and domain walls.
On the side of synthesis, we demonstrated the superior quality of electroplated materials, the basis for our magnetic wires. In particular, we demonstrated close-to-bulk electric resistivity, going much beyond what planar physical deposits can achieve. The wires can sustain a dc current above 10^12 A/m^2 without destruction. This is a key achievement for the prospect of shifting bits with electric current. Besides, we developed new processes for the in-depth structuration, as required for a digital media.
On the side of supporting tools, the consortium has devoted effort to develop new capabilities required for the achievement of M3d's objectives. A first series is advanced magnetic imaging (engineered tips for magnetic force microscopy, and PEEM and STXM microscopy synchrotron instruments). A second series concerns magnetic simulation. It covers new micromagnetic simulating codes, macrospin, post-processing to analyze experimental data from microscopies, Monte Carlo codes to consider magnetic interactions in a 3D medium, analytical modeling for bit shifting.
As regards the magnetic functionality of materials, the existence of longitudinal solitons and domain walls has been proven experimentally for diameters in the range 200-20nm, confirming the relevance of the routes chosen. Their internal magnetization texture has been revealed, consistent with theoretical expectations. This is crucial, as the efficient of bit shifting depends on their texture. We also optimized special magnetic alloys, based on the criteria of easing domain-wall motion by removing extrinsic material pinning; we achieved motion at unprecedented 5mT quasistatic magnetic fields. We also demonstrated the quasistatic motion of solitons, and a general scaling law has been outlined by simulation and theory to design at will the propagation field. For domain walls, we sought the digital nature of coding through local modulations of diameter, first outline by simulation and theory, then demonstrated experimentally.
Then, we devoted effort to concepts required to put together these fundamental aspects into a working memory. First, magnetostatic interactions were considered, source of cross-talk in 3D materials. We combined a hardware approach (geometrical design) and a software approach (coding algorithm) to reduce cross-talk well below 1mT. Second, we developed tunnel magnetoresistive elements embedded deep inside each wire, to probe locally the state of one soliton or domain wall at a time. These elements display magnetoresistance, however its magnitude and the resistance of the junction would both need to be increased for a working device. Third, we considered heat-assistance to design a write element at the end of a wire. This proved not efficient, probably due to the intrinsically 3D nature of the wire. Alternative routes should be explored in the future. Fourth, we addressed the dynamics of shifting information. Depending on the soliton and wall texture, we predicted shifting from a few tens of ns down to a few ns. Experimentally, domain walls move with pulses of field down to 2ns. Work remains to be done to clock walls and solitons shifting in a robust manner from one digital site to the next along a wire. Fifth, we have paved the way towards current-induced bit shifting. We demonstrated magnetoresistance in solitons (the reverse effect of spin-transfer torque), showing the suitability of the materials. Solitons could not be shifted even under the highest currents densities. Work shall continue after M3d to focus on shifting walls, with various schemes of spin-transfer torque.
Potential Impact:
We aimed at outlining one or more viable material routes for a magnetic memory, massively exploiting for the first time the third dimension of a magnetic media. Crucial criteria for a data storage solution are the areal density, power consumption, data rate and cost. The demonstrations within this project show the compatibility for a density of 5Tbit/in2, zero standby power consumption thanks to the use of magnetic materials, R/W consumption in the range 10-100pJ/bit, 10GHz data rate and price of 20€/Tbit, with a scalability down to at least 2€/Tbit.
The potential market is clearly in mass data storage with a rather random access (unlike tapes, still used eg for archiving). This applies both for massive storage on the cloud to drastically reduce its energy footprint, and portable devices to boost the capacity currently offered by Flash memory, remaining an essentially 2D technology despite the current sequential stacking strategy.
Since Magnetic Random Access Memories (MRAMs) have been included in the ITRS roadmap in 2010, the world-leading semiconductor industries are on the race to embed magnetic functionalities on chips, with recent demos beyond 1Gbit/chip. Thus, magnetic materials are no more seen as a disruptive technology. While our contribution is centered on materials and thus with low TRL, we clearly unlocked key bottlenecks, and we expected that this will revive the consideration of the concept of a magnetic race track memory in academic laboratories, and raise the awareness of its potential on the side of companies.
While serious obstacles remain in the demonstration of such memories, this proposal is in line with other technologies progressively moving forward a mid-scale use of the third dimension, such as flash and phase-change materials, among others. Magnetic materials offer the specific prospect of the massive use of the third dimension at the cost of low complexity, based on the principle of bits moved along a simple race-track with no 3D interconnections, simply using a spin-polarized current flown along the track. Thus, it is attractive for long-term prospects. The partnership with SMEs in M3d allowed developing technology-relevant concepts, as a sound basis for implementing further actions with higher TRL on the topic in the future. Their European character is an added value to build effective consortia in the future, a boost the competitiveness of Europe in this emerging field. All key partners are present in Europe, from academic laboratories (the two spintronics Nobel prizes for physics in 2007 arise from European labs) to foundries, major joint public / private R&D centers, and SMEs to provide key expertise in emerging concepts.
Apart from direct dissemination at conferences, we have advertised the start of the project on the partners' web sites and brochures, and have set up a dedicated web site (http://mem3d.eu). A dedicated workshop has been held in the middle of the project, in the Spring 2015. It brought together academics and industrial partners and attendees, to disseminate the progress made and build links to explore possibilities for future exploitation. As the achievements of M3d have been made possible through the association of physicists and chemists, the project also contributes to raise the awareness both in research centers and in the broad public, that innovation can be searched through interdisciplinary effort,.
Exploitation has taken place and is still foreseen on technical aspects not specifically related to the objectives of M3d, however which had to be developed beyond the existing state-of-the-art for their sake. This concerns: the upscaling and catalog-ready engineered alumina template designed by the world-leading SMART company; on-demand software-assisted lithography, already included in an industrial product; sharing the latest developments of micromagnetic simulations as open-source, with a model to foster the creation of new consortia on advanced topics, based on the support and handling of the code doe advanced uses. M3d has also contributed to the higher education of several PhDs, who are now working in ICT companies targeting chips and sensors.
List of Websites:
http://mem3d.eu
The web site provides all logos, publicly-available images and dissemination events.
Until recently, all commercial storage technologies were based on storage at a surface (chip, disk, tape). While the growth of bit density and the associated decrease of price per bit have been following Moore's law for decades, clear physical limitations are currently putting a final end to this growth. M3d pertains to exploratory work on materials science technology, with a view to make possible a change of paradigm in data storage technology. The background idea is the proposal for a so-called 3D race-track memory, put forward by IBM.
The consortium gathered physicists of spintronics and technology, and chemists from the 3D bottom-up synthesis. Two SMEs were associated, one for the synthesis, another one for contributing to making the most relevant choices in terms of magnetic technology. The successful output of the project highlights to the academic sector, industry and society, the potential of innovation arising from interdisciplinarity.
Our actions have been several fold, unlocking several bottlenecks on the material’s side: 1. Synthesis of suitable materials in a wire form 2 Design of a digital media, ie with geometrical notches or segments 3 Evaluation and elimination of cross-talk, specific to magnetic materials in a 3D geometry 4. Outlining through simulation the potential for close-to-ns operation in single wires; preliminary results hint at the practical feasibility 5. Explore routes for reading and writing elements. The former has shown encouraging results, based on the reading a tunnel magneto-resistive element embedded in the wire. The latter has been attempted by heat assistance, however was not successful.
The academic output is very positive, and has triggered a serious attention to magnetic nanowires and nanotubes, and more generally to nanomagnetism and spintronics in a 3D geometry. Several invited reviews and chapters (Nature Communications; Handbook of Magnetic Materials series etc.) have been requested from partners of the consortium. New consortia are already being formed for future projects, of similar or higher TRL. From the point of view of data storage industry, this development is very timely as other technologies for storage are moving to 3D designs, such as flash memories, and those doing so are clearly winning the race of performance. In the future, 3D magnetic technologies promise the potential for high performance combined with extremely long endurance and non-volatility. Thus, the original strategy of M3d is being perfectly met, to highlight the potential to foster future more applied work in the topic.
Direct exploitation is also coming immediately from technical developments carried out during the project. This includes nanoporous templates with a 3D engineering, including modulations of diameter in the depth (application to filtration photonics etc); software driving lithography tools; new atomic layer deposition processes; high added value tips for magnetic force microscopy (high spatial resolution and low moment); a micromagnetic software available open source, and whose support may turn commercial in the mid-term.
Project Context and Objectives:
M3d pertains to exploratory work on materials science, with a view to introduce a change of paradigm in data storage technology. Until recently, all commercial storage technologies were based on storage at a surface (chip, disk, tape). While the growth of bit density and the associated decrease of price per bit have been following Moore's law for decades, clear physical limitations are currently putting a final end to this growth. Multi-level storage are being implemented to sustain this growth rate, however the gains are only incremental and imply an increase of cost and complexity as each level needs to be patterned one atop another.
Our project sets out to break through existing material limitations in order to develop large-scale non-volatile data storage massively relying on the use of the third dimension (3d). We follow the IBM concept of magnetic race-track memory, where series of bits would be shifted along vertical magnetic wires densely packed in arrays, requiring only one single read/write element per wire [S. S. P. Parkin et al., Magnetic Domain-Wall Racetrack Memory, Science 320, 190 (2008)]. This allows us to address the need for media density above 5Tbit/in2 when expressed in the usual terms of area, and reasonable cost per Tbit, whilst retaining the potential for an excellent access speed and energy consumption.
To keep costs moderate the 3D synthesis of a dense array of magnetic nanowires shall be done in a limited number of stages and therefore rely on bottom-up routes. This must be complemented by top-down surface patterning, for eg appending one read/write element per wire. In addition, to minimize risks several strategies shall be explored in parallel for coding bits along the magnetic wires. This includes magnetic solitons in stacked thin disks, solitons in segmented long wires, and magnetic domain walls in continuous wires. In the course of the project, we refocused work on the latter two, based on progress in synthesis and simulations. We scheduled addressing (writing & reading) strategies in the second half of the project.
We gathered partners with the complementary expertise required for these developments in the consortium: Institut NEEL and SPINTEC (CNRS), Cavendish Laboratories and Univ. Hamburg bring in advanced expertise in nanomagnetism and spintronics to implement various routes for coding and addressing bits. Univ. Hamburg, Univ. Erlangen-Nürnberg bring in a cutting-edge expertise in 3D bottom-up design of porous media. Synchrotron SOLEIL provides various instruments for microscopic magnetic investigation of the devices designed.
Through this exploratory work, we aim at pushing Europe ahead of Asia and the USA in 3D storage, an area with a high potential for added value and innovation. With this goal in mind, the SME company SmartMembranes with world-leading expertise in bottom-up products is an active partner in the consortium. Its aim is to boost the efficiency of the academic partners by providing suitable templates for the fabrication of dense arrays of nanowires. Conversely, SmartMembranes shall benefit from the advancement of expertise from those partners, thereby enlarging its portfolio of state-of-the art products after up-scaling the synthesis processes. Second, participation of the data storage industrial partner CROCUS Technology shall provide the consortium with a clear view on relevant current technology and patents, also creating a link that may be strengthened in a future R&D device project.
Project Results:
Early in the project, simulations have been running, to outline features and operating schemes of three possible routes to code information, based on domain walls or solitons. These revealed that only a narrow operating window would exist for transverse solitons, especially when taking into account the consequences of inserting a read element in the middle of a stack. Thus, we focused experimental work on two coding routes only: longitudinal solitons, and domain walls.
On the side of synthesis, we demonstrated the superior quality of electroplated materials, the basis for our magnetic wires. In particular, we demonstrated close-to-bulk electric resistivity, going much beyond what planar physical deposits can achieve. The wires can sustain a dc current above 10^12 A/m^2 without destruction. This is a key achievement for the prospect of shifting bits with electric current. Besides, we developed new processes for the in-depth structuration, as required for a digital media.
On the side of supporting tools, the consortium has devoted effort to develop new capabilities required for the achievement of M3d's objectives. A first series is advanced magnetic imaging (engineered tips for magnetic force microscopy, and PEEM and STXM microscopy synchrotron instruments). A second series concerns magnetic simulation. It covers new micromagnetic simulating codes, macrospin, post-processing to analyze experimental data from microscopies, Monte Carlo codes to consider magnetic interactions in a 3D medium, analytical modeling for bit shifting.
As regards the magnetic functionality of materials, the existence of longitudinal solitons and domain walls has been proven experimentally for diameters in the range 200-20nm, confirming the relevance of the routes chosen. Their internal magnetization texture has been revealed, consistent with theoretical expectations. This is crucial, as the efficient of bit shifting depends on their texture. We also optimized special magnetic alloys, based on the criteria of easing domain-wall motion by removing extrinsic material pinning; we achieved motion at unprecedented 5mT quasistatic magnetic fields. We also demonstrated the quasistatic motion of solitons, and a general scaling law has been outlined by simulation and theory to design at will the propagation field. For domain walls, we sought the digital nature of coding through local modulations of diameter, first outline by simulation and theory, then demonstrated experimentally.
Then, we devoted effort to concepts required to put together these fundamental aspects into a working memory. First, magnetostatic interactions were considered, source of cross-talk in 3D materials. We combined a hardware approach (geometrical design) and a software approach (coding algorithm) to reduce cross-talk well below 1mT. Second, we developed tunnel magnetoresistive elements embedded deep inside each wire, to probe locally the state of one soliton or domain wall at a time. These elements display magnetoresistance, however its magnitude and the resistance of the junction would both need to be increased for a working device. Third, we considered heat-assistance to design a write element at the end of a wire. This proved not efficient, probably due to the intrinsically 3D nature of the wire. Alternative routes should be explored in the future. Fourth, we addressed the dynamics of shifting information. Depending on the soliton and wall texture, we predicted shifting from a few tens of ns down to a few ns. Experimentally, domain walls move with pulses of field down to 2ns. Work remains to be done to clock walls and solitons shifting in a robust manner from one digital site to the next along a wire. Fifth, we have paved the way towards current-induced bit shifting. We demonstrated magnetoresistance in solitons (the reverse effect of spin-transfer torque), showing the suitability of the materials. Solitons could not be shifted even under the highest currents densities. Work shall continue after M3d to focus on shifting walls, with various schemes of spin-transfer torque.
Potential Impact:
We aimed at outlining one or more viable material routes for a magnetic memory, massively exploiting for the first time the third dimension of a magnetic media. Crucial criteria for a data storage solution are the areal density, power consumption, data rate and cost. The demonstrations within this project show the compatibility for a density of 5Tbit/in2, zero standby power consumption thanks to the use of magnetic materials, R/W consumption in the range 10-100pJ/bit, 10GHz data rate and price of 20€/Tbit, with a scalability down to at least 2€/Tbit.
The potential market is clearly in mass data storage with a rather random access (unlike tapes, still used eg for archiving). This applies both for massive storage on the cloud to drastically reduce its energy footprint, and portable devices to boost the capacity currently offered by Flash memory, remaining an essentially 2D technology despite the current sequential stacking strategy.
Since Magnetic Random Access Memories (MRAMs) have been included in the ITRS roadmap in 2010, the world-leading semiconductor industries are on the race to embed magnetic functionalities on chips, with recent demos beyond 1Gbit/chip. Thus, magnetic materials are no more seen as a disruptive technology. While our contribution is centered on materials and thus with low TRL, we clearly unlocked key bottlenecks, and we expected that this will revive the consideration of the concept of a magnetic race track memory in academic laboratories, and raise the awareness of its potential on the side of companies.
While serious obstacles remain in the demonstration of such memories, this proposal is in line with other technologies progressively moving forward a mid-scale use of the third dimension, such as flash and phase-change materials, among others. Magnetic materials offer the specific prospect of the massive use of the third dimension at the cost of low complexity, based on the principle of bits moved along a simple race-track with no 3D interconnections, simply using a spin-polarized current flown along the track. Thus, it is attractive for long-term prospects. The partnership with SMEs in M3d allowed developing technology-relevant concepts, as a sound basis for implementing further actions with higher TRL on the topic in the future. Their European character is an added value to build effective consortia in the future, a boost the competitiveness of Europe in this emerging field. All key partners are present in Europe, from academic laboratories (the two spintronics Nobel prizes for physics in 2007 arise from European labs) to foundries, major joint public / private R&D centers, and SMEs to provide key expertise in emerging concepts.
Apart from direct dissemination at conferences, we have advertised the start of the project on the partners' web sites and brochures, and have set up a dedicated web site (http://mem3d.eu). A dedicated workshop has been held in the middle of the project, in the Spring 2015. It brought together academics and industrial partners and attendees, to disseminate the progress made and build links to explore possibilities for future exploitation. As the achievements of M3d have been made possible through the association of physicists and chemists, the project also contributes to raise the awareness both in research centers and in the broad public, that innovation can be searched through interdisciplinary effort,.
Exploitation has taken place and is still foreseen on technical aspects not specifically related to the objectives of M3d, however which had to be developed beyond the existing state-of-the-art for their sake. This concerns: the upscaling and catalog-ready engineered alumina template designed by the world-leading SMART company; on-demand software-assisted lithography, already included in an industrial product; sharing the latest developments of micromagnetic simulations as open-source, with a model to foster the creation of new consortia on advanced topics, based on the support and handling of the code doe advanced uses. M3d has also contributed to the higher education of several PhDs, who are now working in ICT companies targeting chips and sensors.
List of Websites:
http://mem3d.eu
The web site provides all logos, publicly-available images and dissemination events.