New techniques for combining complex oxide thin films promise electrical control of magnetic properties for data storage and computing.
Materials combining the magnetic and electrical properties of different iron oxides, which could underlie new technologies for computer memory and computation, are the focus of intense work in MIT Professor Caroline Ross's Magnetic Materials and Devices Group lab. Some magnetic oxides, such as cobalt iron oxide (CFO), have controllable magnetic behavior, which when combined with an electrically active material such as bismuth iron oxide (BFO), can yield a promising material for future logic or memory devices. Such a material combining the ferroelectric or piezoelectric properties, of BFO with the magnetic properties of CFO, is called multiferroic. The two oxides also have different crystal structures and were grown on a substrate of another oxide – strontium titanate (SrTiO3).
Caroline Ross, Toyota Professor of Materials Science and Engineering at Massachusetts Institute of Technology in Cambridge, Mass., and Associate Head of the Department of Materials Science and Engineering, uses models to demonstrate the crystal structures of complex metal oxides, such as spinel and perovskite phases. Her group studies their structural and magnetic properties as well as creating thin films of the material with potential for use in data storage and computation. On the left, with black, red and silver, atoms are aligned in a perovskite structure, on the right, with blue, red and silver, a spinel structure is modeled. Photo: Denis Paiste, Materials Processing Center
"They grow as regions of one material next to regions of the other, and because of the way the crystal structures match up with the substrate, you can get pillars of one inside a matrix of the other. In our case, we grow pillars of the spinel, which is magnetic, inside a matrix of the perovskite, which is a ferroelectric. So you have a thin film, which is both ferroelectric and ferrimagnetic. Because the two phases are actually touching, if you apply an electric field, you can switch the magnetism, and if you apply a magnetic field, you can switch the ferroelectricity, so you actually get this two-phase material called a multiferroic," said Ross, the Toyota Professor of Materials Science and Engineering. Ross also is Associate Department head for the Department of Materials Science and Engineering in MIT's School of Engineering.
Graduate student Nicolas M. Aimon, working in Ross’s lab, has developed new methods of making of these oxide thin films through a method called combinatorial pulsed laser deposition. In multiple papers, Aimon, Ross and other collaborators have analyzed the structure of the complex metal oxide films, studied their magnetic characteristics and modeled magnetic switching through application of an electric field. (See related article.)
Being able to control the magnetism of the CFO pillars in a BFO matrix means they could potentially be the basis for the zeros and ones in computation and memory. "You can imagine a pillar of this magnetic spinel (CFO) would represent a bit of data. It can be magnetized up or down. You can write the data by making use of the ferroelectric perovskite (BFO), which surrounds it, by applying electric fields," Ross said. The next step will be to develop devices that are based on storage of information in these magnetic pillars.
Integrating into CMOS
While the BFO/CFO thin films were grown on a strontium titanate layer, the lab also grew strontium titanate films (with a variety of metals such as iron and cobalt integrated into the STO) on silicon, which is the primary material for current computer chips, which use complementary metal oxide semiconductor transistors (CMOS). Traditional silicon-based CMOS processing doesn't make use of magnetic properties or ferroelectric properties. "The idea is if we can develop new materials that have these kind of properties, then you can build things like eventually non-volatile memories or logic, or structures where you can use electric fields to switch magnetism or conversely magnetic fields to switch electrical properties, which is less useful, but still possible. Those kinds of cross-functionalities are a very hot topic right now in materials and device research," Ross said. "If you grow them on a silicon substrate, you can introduce those functions into conventional CMOS, which doesn't use anything other than silicon and various kinds of oxides and metals," she said. The advantage of integrating such functionality directly into the CMOS process is the economies of scale the computer chip technology has.
The main advantage of the coupling between electrical and magnetic properties is being able to use an electrical field to control magnetism. "We like magnetism because it retains its memory. That's the whole basis of magnetic data storage, that once you magnetize something, it stays magnetized. But you don't like the fact that in order to magnetize it, you have to apply a magnetic field, because it takes a lot of effort to make magnetic fields.
You need to put a lot of current through a coil to produce a magnetic field. Electromagnets don't give you much field for your current. It's much more convenient to produce electric fields, it uses up much less power because you can apply an electric field by just charging a capacitor, and then the electric field will stay there, and there's no current flowing," Ross said. "If you're a device designer, you don't really want to have to make magnetic fields; you'd much rather make electric fields because it takes less power. It's simpler. So if you can have a device which uses the advantages of magnetism in terms of data storage but you can switch that magnetism using an electric field, then you get the best of both worlds, so that's why there's a big interest right now in multiferroics, because it allows that cross coupling of the magnetic and the electrical properties," Ross said.
No threat to hard disks
Conventional hard disks, the workhorse of computer memory for the past several decades, are made of glass or aluminum with magnetic layers, a head which has a small electromagnet that writes data onto the disk and a magnetoresistive sensor which detects the data on the disk. Ross, who early in her career worked on hard disks as a research scientist at Komag, Inc., said the new devices she envisions will complement rather than replace hard disks. "Whether you use a hard disk or a flash drive or some other manifestation of non-volatile memory, a lot of it's to do with the performance versus the cost. Hard disks have been very good in terms of that metric. Although nowadays, you buy laptops and they don't even have a hard drive in them, so the balance has shifted somewhat. But the hard drive has always represented very good value for money in terms of how much data you can store and how cheap it is per bit and performance parameters like access time. It's really a good technology; it's been evolving for decades," Ross said.
Solid state technologies like flash memory have advantages, such as no moving parts, which make them attractive for small computers, but they are relatively expensive per bit and don't have as long a life because the number of times you can read and write them is less than for hard drives, she said.
"Even if some laptops now no longer have hard drives, it doesn't mean that that the hard drive industry is all of a sudden dead. It's a got a massive amount of inertia and it's going to be the lowest cost solution for all kinds of applications except maybe for these high-end laptops where the size and the weight and the power consumption are the critical parameters, and you'll pay over the odds to get that. It's economics really, but they have a long life ahead of them in terms of applications like larger computers, or back up systems," Ross said.
Faculty Home Page
MIT News Office
Controlling light with magnetism
Besides logic and memory devices, Ross’s group also is studying the potential use of multiferroic complex oxide materials for magneto-optical devices. When a material is magnetized, it interacts with light differently depending whether the light is going parallel to the direction of magnetization or antiparallel to it. Light, for example, usually passes through glass both ways but in a one-way glass, the light can go out, but it can’t come back in. “We’ve been looking at these magnetic oxides, and one of the key devices we’ve been working on is a magneto optical device which is a diode for light. To get that kind of directional behavior, you need a material that itself has some directionality, and magnetism provides that because there’s an arrow – imagine the magnetization – it’s pointing in one direction,” Ross said. In 2011, Ross and Lionel Kimerling, the Thomas Lord Professor of Materials Science and Engineering, along with former students Lei Bi Ph.D. ’11 and Juejun Hu, Ph.D., ’09, demonstrated an optical diode using garnet as a filter to allow light to pass in one direction but divert it from passing through in the opposite direction. “We’re talking about a magnetic garnet, where you have the same crystal structure as garnet gemstone, but you replace the aluminum and silicon with iron, so you have a magnetic garnet. Those are very nice magnetic materials, but it’s quite hard to grow them on non-garnet substrates,” Ross said.
Substituting magnetic oxides for garnet
Magnetic oxides, such as strontium titanate, are a possible alternative to garnet because they are easier to grow on different substrates, which means they are easier to integrate it into a device. “If we can get the same kind of optical properties that you get with garnet, then you can build it into these devices, which currently are based on garnet, and this optical diode is one of these devices. It’s a key device for integrated photonics. Really, you could say that you won’t be able to build a fully integrated optical chip until you have these diodes that can be built onto the chip just like you can’t build an integrated circuit without diodes,” Ross said.
Current methods for making on-chip isolators, or optical diodes are clumsy, because they require bits of garnet to be glued onto the chip to get the effect. “If we can either grow garnet or grow a different material that does the same thing, then you integrate the isolator with everything else,” she said. Other work in the group includes magnetic thin film metallic materials for domain wall devices and block copolymer self-assembly.
Ross has published extensively on block copolymer self-assembly with MIT Professors Karl Berggren, Edwin L. Thomas and Alfredo Alexander-Katz, MIT graduate students Joel Yang and Yeon Sik Jung (Ph.D. ’09, now Associate Professor, Korea Advanced Institute of Science and Technology) and visiting doctoral student Amir Tavakkoli of the National University of Singapore (now a post-doctoral associate at MIT).
In 2012, they demonstrated a method for making three-dimensional nanowires and junctions using self-assembling block copolymer materials. “We use the block copolymer to self-assemble and make these very specific structures on the substrate,” Ross said. Tiny posts on the silicon substrate were key to the three-dimensional work. The self-assembled block copolymers “can be used as masks to pattern whatever is underneath or you can put things on top of it and use it to pattern those. It’s really a technique for making small structures,” Ross said.
Ross also is examining the potential for domain wall devices based on magnetic thin film metallic materials to handle magnetic logic. “We’ve looked at some of the behavior of particular types of domain walls and how they might be integrated into a device, and then how you could move them around and detect them,” Ross said.
Ross joined the MIT faculty in 1997 and was tenured in 2001. A native of the United Kingdom, Ross received her undergraduate and doctoral degrees both from Cambridge University. She is recording her lectures and working on converting her class 3.15, Electrical Optical Magnetic Materials and Devices, to an EdX version for online education.