Droplet Computing

June 22, 2015

Nearly everyone has one or two computer diskettes at the bottom of a drawer, somewhere. The more veteran scientists might even have a 5-1/4-inch floppy acting as a bookmark in some notebook. I admit to having a few eight-inch diskettes that I keep as mementos of my earliest computing days. These media have one thing in common, they're magnetic. Another thing they have in common is that they're unreadable by today's computers.

Magnetic bubble memory is an obsolete magnetic media. This was an important memory technology of the late 1970s and early 1980s, and it was my first industrial research topic. The logical bits of magnetic bubble memory were magnetic regions in garnet thin films. They were called "bubbles," since they looked like bubbles through a polarizing microscope.

In actuality, magnetic bubbles aren't bubbles; rather, they are cylindrical magnetic domains. Unlike computer discs, magnetic bubble memory was non-mechanical, so it was far faster than disks and more reliable. In the mid-1980s, semiconductor memory technology advanced in speed and density beyond magnetic bubble memory, so I moved on to other research.

Photograph of an Intel Magnetics magnetic bubble memory module (c. 1982) This module contains the thin film bubble medium, and two orthogonal coils that served to cycle the bubbles in storage tracks. (Photo by the author, via Wikimedia Commons.)

As I explained in an earlier article (Droplet Logic, September 19, 2012), magnetic bubble memories were serial memory devices. The magnetic bubbles were shuttled around the magnetic film using an in-plane rotating magnetic field, provided by orthogonal coils. They were held captive on a pattern of a magnetic alloy. The operation of a T-bar pattern and an asymmetric chevron pattern are shown in the figure.

Two magnetic bubble circuits, the T-bar (left) and asymmetric chevron (right). The deposited permalloy pattern is magnetized in different states by an in-plane rotating magnetic field, and the bubbles are drawn towards the north-most pole. Diagram by author, rendered using Inkscape. (Click for larger image).

Just as bar magnets will repel each other when their magnetic vectors (N-S directions) are aligned, magnetic bubbles will repel each other. A magnetic bubble running along a track can be diverted onto another track by the presence of another bubble. The repulsion of proximate magnetic bubbles can be used to make some simple logic gates.

As we fast-forward three decades, a team of engineers from Stanford University (Stanford, California) has resurrected the bubble memory concept, this time using "real" bubbles. Their bubbles are droplets of a magnetic fluid contained in a non-magnetic liquid, and they are shuttled around similar magnetic patterns. Their research is published in a recent issue of Nature Physics.[1-3]

It's possible to manipulate droplets using electric, optical, and acoustic forces, also. Several years ago, a team of Finnish and Swedish scientists used gravity to propel water droplets along superhydrophobic channels so that they could interact to produce logic functions and a flip-flop.[4] The Stanford magnetic droplets are ferrofluid, and they run on permalloy tracks.[1]

A ferrofluid droplet being guided along a permalloy T-bar track containing a logic pattern. (Still image from a YouTube Video.)[3]

Unlike magnetic bubbles in garnet materials, these ferrofluid droplets in another liquid medium are under the influence of more than just magnetic forces, since hydrodynamics is involved. Manu Prakash, an assistant professor of bioengineering at Stanford, who directed this research project, says that the dissipative forces of hydrodynamics at low Reynolds number are important to the operation of these devices.[5] The research team was able to utilize the interplay of magnetic and hydrodynamic forces between droplets to develop AND, OR, XOR, NOT, and NAND logic gates, a full adder, a flip-flop, and a finite-state machine.[1]

In the laboratory experiments, the droplets were in the range of 0.3 - 1.0 millimeter, with a preferred value of about 0.6 mm, but the droplet size can be smaller.[1-2] The ferrofluid droplets were water-based, and the carrier medium was oil-based. The parallel glass surfaces containing the media were coated with Teflon.[1] Says graduate student and paper co-author, Jim Cybulski,

"We can keep making it smaller and smaller so that it can do more operations per time, so that it can work with smaller droplet sizes and do more number of operations on a chip... That lends itself very well to a variety of applications."[2]

Manu Prakash, left, and graduate students Jim Cybulski and Georgios Katsikis, developers of the droplet computer. (Stanford University photo by Kurt Hickman.)

As shown by the diverse number of logic functions demonstrated, these droplet circuits can do the same operations as a conventional electronic computer, albeit at a much slower rate. As graduate student and first author on the paper, Georgios "Yorgos" Katsikis, explains, "Give us any Boolean logic circuit in the world, and we can build it with these little magnetic droplets moving around."[2] computation, itself, is not the target for this device. As Prakash explains,

"We already have digital computers to process information. Our goal is not to compete with electronic computers or to operate word processors on this... Our goal is to build a completely new class of computers that can precisely control and manipulate physical matter. Imagine if when you run a set of computations that not only information is processed but physical matter is algorithmically manipulated as well. We have just made this possible at the mesoscale."[2]

The orthogonal coils used to generate the rotating magnetic field in the experiments. Just as for the magnetic bubble device in the first figure, these can be shrunk in size. (Still image from a YouTube Video.)[3]

The ability to computationally control the flow of small droplets of reagents might be useful for high-throughput chemistry and medical diagnostics.[2] The Stanford team plans to create a design tool for such droplet circuits to allow other research groups to build on their experience.[2] Says Prakash,

"We are trying to build a new class of computers that not only manipulate information - but also manipulate physical matter simultaneously. Such an approach will make algorithmic assembly at mesoscale a reality."[5]

References:

1. Georgios Katsikis, James S. Cybulski, and Manu Prakash, "Synchronous universal droplet logic and control," Nature Physics (advance online publication, June 8, 2015), doi:10.1038/nphys3341.
2. Bjorn Carey, "Just add water: Stanford engineers develop a computer that operates on water droplets," Stanford Report, June 8, 2015.
3. Stanford engineers build a water-droplet based computer that runs like clockwork, Stanford University YouTube Video by Kurt Hickman, June 5, 2015.
4. Henrikki Mertaniemi, Robert Forchheimer, Olli Ikkala and Robin H. A. Ras, "Rebounding Droplet-Droplet Collisions on Superhydrophobic Surfaces: from the Phenomenon to Droplet Logic," Advanced Materials, vol. 24, no. 35 (2012), DOI: 10.1002/adma.201202980.
5. Manu Prakash, Private Communication, June 10, 2015.