How Can Spintronic Devices Be Built to Improve Computing Capacity?
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Author | ||
| License | CC Attribution 3.0 Unported: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
| Identifiers | 10.21036/LTPUB10342 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
|
00:00
Data structureAreaMaterialization (paranormal)Multiplication signDirection (geometry)Atomic numberDimensional analysisComputerComputerType theoryData storage deviceArray data structurePoint (geometry)SurfaceInterface (computing)Energy conversion efficiencyCategory of beingPredictabilityPhysical systemSemiconductor memoryElement (mathematics)Vector potentialCartesian coordinate systemProgramming paradigmBuildingPhysicalismLimit (category theory)View (database)Inheritance (object-oriented programming)Field (computer science)NumberSolid geometryState of matterMoment (mathematics)Order (biology)Key (cryptography)Data storage deviceInformationStatic random-access memoryElectric dipole momentOperator (mathematics)CodecKälteerzeugungInteractive televisionBitDistanceChannel capacityThree-dimensional spaceMoore's lawReading (process)Different (Kate Ryan album)Population densityVertex (graph theory)SupercomputerFood energy1 (number)Power (physics)Distribution (mathematics)Object (grammar)Cylinder (geometry)Boundary value problemScaling (geometry)Arithmetic meanPulse (signal processing)Projective planeNetwork topologyTrailMultikollinearitätParallel portCausalityWeightLecture/ConferenceMeeting/Interview
Transcript: English(auto-generated)
00:00
The big question that concerns my research is how can we go beyond today's paradigm of storing information and computing information and for 40 years We've had a paradigm whereby all the devices we're using are essentially two-dimensional in character And after 40 years we've reached the limitations of the technologies where we store information
00:23
For example in magnetic disk drives and we compute information in silicon-based Computing systems today and the reason for the end if you like of the these two technologies Is that we've been scaling both those technologies in two dimensions So the big question is how do we go beyond that paradigm and into the third dimension?
00:42
thereby allowing us to improve for example the computing capacity of memory elements or storage elements or the computing power of computing systems At the heart of the approach that we are using is a new field of research called spintronics so all of conventional Technologies today memory and computing technologies use electrons
01:03
They use currents of electrons or you store numbers of electrons But the electron has a property called spin and we can create currents of electrons that are spin polarized and in certain types of materials We can control the flow of these currents of spin polarized electrons enabling us to create new devices So an important device that actually I invented a few years ago is called a spin valve device and it detects tiny
01:26
magnetic fields and this allowed us for example to improve the storage capacity of a conventional magnetic disk drive by a thousand times because of the exquisite ability of this spin valve device to detect tiny Magnetic fields or tiny magnetic regions in which information is stored
01:43
But we want to go beyond that because that technology has reached the end of its Two-dimensional roadmap, we want to go into the third dimension and with some very interesting new spintronic phenomena and materials We could create a new type of storage device with entirely different principles Which is innately three-dimensional and would consist of billions of what we call race tracks
02:04
Which are essentially vertical columns of magnetic material in which we store tiny magnetic regions representing zeros and ones and we manipulate them using an entirely spintronic phenomena a current of spin polarized Electrons that can move this information up and down these race tracks and we've recently demonstrated at speeds
02:24
Exceeding a kilometer per second this will enable us to build a completely solid-state device With about a hundred times the capacity of today's solid-state drives for example because of the innate three-dimensional nature of this new concept entirely derived from new physics of
02:41
Spintronics that we've evolved and discovered in fact just in the last two or three years The key findings are that we have in the last three or four years discovered four distinct new physical phenomena Related to spin orbit coupling that enables us to manipulate the main walls in these race tracks Extremely efficiently with current pulses much more so than we had anticipated just two or three years ago
03:06
So these key findings are firstly if you pass current into a regular metal you can create a current of purely Spin which is purely a spin current no charge current at all and this means for example Let us take a platinum wire you pass a current along this wire
03:22
You generate a spin current around the circumference of the wire in a chiral fashion It will be rotating in a clockwise or an anticlockwise direction depending upon the direction of the charge current Then the spin current that we create in this way can diffuse into the racetrack the magnetic material in which we're going to store
03:42
This information and that creates talks on the domain walls in this racetrack and then can cause them to move But in order that it can do this a second important finding is that these Racetracks consists of just a nanometer thick layer of magnetic material in which by design the magnetization points
04:02
perpendicular to the racetrack and then as you go along the racetrack the magnetic Moments in one region will be pointing up in the next will be pointing down and then up again But in fact as they go from an up to a down magnetized region the moments will be Rotating in a chiral direction. So as you move from an up to a down direction
04:23
The moments will be let's say going along the wire in this direction But when they rotate from up to down up down to up, they'll be rotating opposite direction So you actually have a chiral magnetization along this wire and this is due to a very interesting exchange Interaction between the magnetic material and for example platinum is called the Dirichinsky Maria exchange
04:45
Which causes magnetic moments not to be parallel as in a ferromagnet or anti parallel as in anti ferromagnet But to be perpendicular to each other in a chiral fashion So we have two chiral physical phenomenon operating in tandem which caused these domain walls to move and this is in essence the physics of
05:05
Manipulating domain walls at these very high speeds but beyond that a very important finding of ours was that We can actually make all the magnetic moments in the racetrack disappear because magnetic moments Interact with each other. So you're familiar with bar magnets on a refrigerator
05:23
They will either want to be parallel or anti-parallel depending on how you position them And this is a major problem on the nanoscale Nanoscopic magnetic elements will interact with each other through these so-called dipole fringing fields and the interactions are enormous They increase inversely with the size of the device
05:42
So we have to make those disappear and just last year We discovered that we could build a racetrack which has two racetracks one on top of each other Which exactly coupled anti-parallel to one another so in aggregate each of these magnetic domains has no net magnetization and Amazingly, we discovered that in this racetrack with no net magnetization
06:04
These magnetic regions will actually move ten times faster. And so this enables us to move these Magnetic bits extremely efficiently ten times more efficiently than was possible just a couple of years ago Last but not least is another physical phenomenon It's called proximity induced magnetization and what happens is the magnetic racetrack
06:25
Causes platinum to become magnetic. It's very significant So that perhaps 50% of the magnetization in a single racetrack will be will be evolved from the platinum itself And in my own mind this proximity induced moment is extremely important in the physics of the motion of these domain walls
06:44
So we have essentially four Intertwined physical phenomenon that amazingly have to all work in tandem to make these domain walls Move when we pass a current into this racetrack We typically use very short current pulses so we can move the race the domain walls the distance that is needed
07:04
The findings have two relevance is one is technological and one is perhaps scientific So technologically what we've done is we've demonstrated that and completely new concept for a technology Innately three-dimensional that could enable us to replace for example the old-fashioned mechanical
07:21
Rotating disk drives with an entirely solid-state device Which would be much more reliable and would be much more energy efficient at the same time we've discovered because of this new physics that this racetrack has the possibility of Extremely fast reading and writing so we could actually use this racetrack memory technology to replace for example the fastest
07:43
Memory today, which is called static random access memory in principle We could devise a racetrack technology that we were in which we could trade off speed and perform speed and Speed and performance and density by having racetracks in which we would allow different numbers of domain walls
08:01
So the same racetrack we could have one domain wall be super fast Hundred domain walls which would be slower still fast, but would be much denser. This is unique to this technology so technologically We've demonstrated after this time that racetrack in principle could work scientifically we've actually discovered entirely new ways of manipulating magnetization more and very interestingly
08:24
I mentioned that we convert charge current to spin current This is something called a spin Hall effect and in metals just of three or four years ago It was thought to be very inefficient. Well, we and others have demonstrated is remarkably that we have very large
08:40
efficiencies of converting charge current into spin currents And so this is what the maybe the technological and scientific impact is. I Think the outlook is extremely exciting the field of spintronics has continues to evolve a lot of new discoveries and new materials So one example is that we've discovered these very interesting chiral domain walls
09:03
Which are innately have non collinear magnetization distributions, and it turns out there are a lot of other objects Magnetically which have even more interesting non collinear magnetic distributions One example is something called a scumion the scumion is a tiny if you like circular magnetic object or cylindrical object in which the boundary is innately chiral and
09:24
Again, these can be manipulated with current in different ways. These seem to be very interesting So the future could be going beyond conventional Magnetic materials in which moments point parallel or anti-parallel to much more interesting Concepts is one example one possible outlook other outlooks is that we want to go
09:44
Of course, we have to be able to build this magnetic racetrack and today it's very interesting that it involves Magnet film layers which are atomic layer dimensions So to create this racetrack, it's about seven atomic layers of magnetic material on a few atomic layers of another material
10:02
So we have to devise ways of implementing it by building these racetracks vertically this is a real challenge, but I think the outlook is For many devices going beyond two dimensions to three dimensions and we have to learn how to build devices in Three-dimensional space and this turns out to be non-trivial many approaches
10:21
but there are many more interesting ways of doing this and I think this is a Hugely interesting area of research how to build structures which are with atomic dimensions on in a three-dimensional world I think spintronics is a really interesting area of research because it is involved with the concept of building Artificial materials and we build these one atomic layer at a time and this we can do this using techniques
10:47
Which are manual manufacturable compatible, but we can build entirely new materials this way with properties not found in any of these materials Very interestingly if you take material a and material B with certain properties the interface between the materials A and B can be entirely distinct and this again means it's difficult to make predictions even but which have a very interesting
11:07
properties so for example you can take two insulating materials and the interface can be a metal and This is another area of which I think is super exciting How can we imagine building computing systems which are innately much more energy efficient than today's computing systems?
11:24
So this goes beyond just memory and storage It goes beyond how can we maybe use those similar concepts to carry out some type of computation? But we want to do this much with much less energy than today's silicon based computers and by way of inspiration We look at our own brain our own brain uses a few watts of power to carry out
11:45
Computations that are equivalent to one of today's supercomputers. So it uses about a million times less energy per computing operation than a silicon based computing system and we have to imagine ways in which we could do that in Not biological material but in conventional materials
12:02
But which will be innately three-dimensional in where they're configured because innately the brain has computing devices the neurons Which are connected by maybe 10,000 wires to each other and we don't know how to do this today Devices we make today are basically two-dimensional arrays of very simple devices connected in very simple ways
12:22
So we have to imagine ways of doing this entirely differently. I think this is a super interesting Project for which we have no obvious answers indeed So the the concept is that we can create artificial materials by building them one atomic layer at a time But another very interesting area of research that has evolved just over the last five or six years
12:43
Is that the surface of a material can have properties entirely distinct from the interior? These are called topological insulators means you can take an insulating material where the surface will have be conducting But it's conducting in very interesting ways So for example electrons traveling in one direction will have their spin oriented in let's say an up direction
13:04
But electrons traveling in the opposite direction will automatically have their spin in the other direction and this is a very interesting Topological phenomenon, but you can imagine of course this could have very important device implications because in this way We can innately create currents of electrons with spin in one direction and we could then use these to manipulate other
13:27
Materials in this way So this is this what I think the point is that today the concept of surfaces and interfaces in two dimensions or even in three dimensions have extraordinarily interesting properties that the individual materials may not have and some of these are simply
13:43
Not found in any other material and it's these topological properties in particular Which means that some aspects of the properties are protected from disturbances Which means again from an application point of view they may be very very relevant So I think it's this intersection of
14:00
extraordinarily interesting physics and materials which we create in engineer to have certain properties coupled with their potential usage for Interesting technological applications that is a super interesting area of research. That's my own research