(Ironically, given my recent missive about the importance of condensed matter beyond applications to information technology, here is a post about condensed matter in information technology.)
It may seem like solid-state drives - basically flash memory - have taken over data storage, particularly in phones, tablets, and laptops. However, magnetic storage media, particularly in the form of hard drives, are still the main tools of choice for the cloud. Magnetic storage is very robust and doesn't rely on charge staying put on tiny floating gates to hold your information. Rather, in a hard drive the zeros and ones of your data are encoded as magnetic domains of particular orientation in a specially engineered thin film of material on a disk platter.
The amount of research and engineering development that has gone into hard drives is amazing. The read/write heads "fly" at nanometer separation over incredibly smooth magnetic surfaces. The smaller you can make the magnetic domains and still manipulate and read them in a controlled manner, the higher the density of the information storage. The timing and positioning stability required to store and retrieve terabytes per square inch is amazing. Reading the data requires detecting the magnetic fields produced by the domains. These days that's done using magnetoresistance, some component of the read head with an electrical resistance that changes depending on the local magnetic environment. Giant magnetoresistance went from the laboratory to hard drive read heads to Stockholm for the 2007 Nobel Prize in Physics. Its successor in read heads, tunneling magnetoresistance, now rules the roost.
On the physics side, there are many challenges for continued scaling. Magnetic domains interact with each other. To be reliable for storage, it's important that the magnetization \(\mathbf{M}\) of a domain remain fixed once it is set. The energy scales associated with pinning a domain tend to scale with domain size, meaning that, all other things being equal, tinier domains are easier to flip (for writing, yay, for long term stability, boo). In the extreme limit, thermal fluctuations can provide enough energy to reorient the magnetization, leading to superparamagnetism. A number of years ago, IBM and others came up with ways to increase the coercive fields (and anisotropy energies) of the bits in disk media. Still, as bits get smaller, it's more important to pin them down, but somehow still allow deliberate reorientation of \(\mathbf{M}\) for write operations.
This article spurred me to write a little about this, in part because of what it gets wrong. Hard drive manufacturers have decided that the best plan is to pin down the bits firmly, and then during the write process, locally kick those bits hard enough to allow reorientation of \(\mathbf{M}\).
One method that's been under consideration for a while, and is the favorite of Seagate, is HAMR - heat-assisted magnetic recording. This requires some means of locally heating the disk media right under the write-head so that thermal energy is available to help the bit reorient. The Seagate approach uses an embedded laser source and a plasmonic antenna to drive the local optical heating (this has nothing to do with an electrical discharge, despite the linked article at the start of this paragraph).
Western Digital is pursuing an alternative approach microwave-assisted magnetic recording (MAMR). The idea there (video) is to use a local oscillator (one based on spin torque) to generate microwaves locally at the write head, with the frequency of those microwaves tuned to drive ferromagnetic resonance of the bit to flip it.
Tons of physics in all of this, and an enormous amount of engineering cleverness. Magnetic data storage will be with us for a while yet.
No comments:
Post a Comment