Storing by numbers - Magnetic data storage
It is taken for granted that our ever-larger data files will be stored reliably and retrieved quickly. However, there is a continual development of sophisticated data storage systems to cope with this burgeoning demand. Ferromagnetic materials have long been used for this task, for example in tape or floppy disk storage, but modern magnetic data storage devices are firmly in the realm of nanotechnology. Understanding the precise operation of these complex systems can be difficult to achieve experimentally, but is often essential to developing a reliable device. Fortunately, numerical modelling allows us to simulate behaviour and observe magnetisation processes on a nanoscale.
The atomistic magnetic moments – like tiny compass needles – arrange themselves according to the chemical/physical structure of the magnet. Because the structures of the materials are not regular, they can only be described by equations with computational methods. In this way, predictions about the magnetic behaviour of a material are possible and can be compared with the experimentally observed properties.
How can we write data to objects that are smaller than the line width of semiconductor technology? Will the written bits be stable or will thermal fluctuations erase the information? How can a signal from nanosized magnets be detected and read the data? To address these questions, the computational tools span several orders of magnitude in time and length scale.
In the right direction
For decades, magnetic hard disk storage was longitudinal (horizontal) on a granular media. Grains of Cobalt Chromium Platinum Tantalum (CoCrPtTa) are magnetically decoupled by a Cr-rich grain boundary phase. About 100 grains are used to encode a bit. Within a bit, the magnetisation points in the same direction as shown in the simulated bit pattern. The magnetic field generated a few nanometres above the disk is sensed by the read head. Perpendicular magnetic recording will more than double storage densities from about 200Gbit/inch2 achieved in longitudinal recording, to 500Gbit/inch2.
In perpendicular recording, the bits are arranged vertically. The data layer is placed in the air gap of a magnetic circuit. The write field is higher than in longitudinal recording, which can only make use of the fringing field of the head. If more field (energy) is provided to switch the storage unit, the bit can become more thermally stable (or the bit size can be decreased). The maximum write field is limited by the magnetic polarisation of the material (FeCo) used for the pole tip of the head. Therefore, alternate means of providing energy are being developed.
Modelling played an important role in the advance of assisted writing technologies like heat assisted magnetic recording, microwave assisted recording and exchange spring media. In the latter, a soft magnetic film to reduce the switching field is coupled to the hard magnetic storage layer.
A further boost of area density is expected from bit patterned media. Here, a single physical entity is used to store a single bit. Tiny islands of magnetic materials will be fabricated on a regular array with a pitch of 25nm and below.
A promising route to make bit patterned media is the deposition of magnetic thin films on self-assembled SiO2 nano-spheres, (EU FP6 project) MAFIN – Magnetic Films on Nanospheres. Successful recording at high densities requires the joint optimisation of head and the data layer. A prototype design can be seen above for magnetic recording on bit patterned media optimised numerically for an area density of 1.8Tbit/inch2.
Magnetic Random Access Memory (MRAM) is a development in data storage that offers fast access, no power consumption for data retention, unlimited lifetime, and lower switching energy than flash memory. Commercial MRAM chips were released in 2006 and can rival flash memory and conventional RAM.
The essential parts of an MRAM cell are used to store one data bit. Magnetic Random Access Memory chips comprise many such cells. The heart of the device is a magnetic tunnel junction made from two ferromagnetic layers, typically alloys of Ni, Fe and Co, and a dielectric spacer, usually Al2O¬3 or MgO. One ferromagnetic layer has its magnetisation fixed, or ‘pinned’, by an adjacent anti-ferromagnetic layer. The other ferromagnetic layer’s magnetisation is free to respond to applied magnetic fields. The 1s and 0s are represented by the magnetisation of the free and pinned layers being parallel or anti-parallel. This allows data to be written by rotating the free layer magnetisation using magnetic field generated from nearby current-carrying wires.
Data read-out is performed by measuring the resistance of the MRAM cell, because the rate of electrical tunnelling across the dielectric barrier depends on the relative orientation of the two ferromagnetic layers.
Numerical modelling has been essential to the development of MRAM. In particular, modelling has allowed the magnetisation reversal mechanisms of the free layer to be studied in detail. This has a major impact on the choice of materials, and the shape and dimensions of the MRAM cells.
Nanowire storage technologies
Patterned magnetic Ni-Fe nanowires contain magnetic domains with magnetisation lying in one of two orientations along the wire length, making this a convenient way of storing digital information. Oppositely magnetised domains are separated by mobile ‘domain walls’ that can be positioned using applied magnetic fields and electrical currents passed through the nanowires.
Several teams of scientists are developing shift register memories that use mobile domain walls in magnetic nanowires. The attraction of these devices is their potential operating speed compared with Complementary metal–oxide–semiconductor (CMOS) equivalents and their ability to retain data without power consumption.
Deliberate defects in the nanowires, such as notches, are used to trap domain walls in a controlled way and maintain their separation. The final devices will include positions for data writing, and read-out, potentially by using a current carrying and magnetic tunnel junction respectively.
Numerical modelling has been essential to understanding the nature of domain wall propagation and pinning in nanowires. Even in simple nanowire systems, these processes can become highly complex. This is shown by calculations of a domain wall (green) travelling through a triangular domain wall diode element.
Dr Dan Allwood, Advanced Research Fellow and Professor Thomas Schrefl, Professor of Functional Materials. Department of Engineering Materials, Sir Robert Hadfield Building, University of Sheffield, Sheffield, S1 3JD, UK. Emails: email@example.com and firstname.lastname@example.org.