Throughout history, printing has provided ambitious businessmen the ability to reach out to the public. It began with books, but then advanced to advertising. Soon people were printing ransom notes. In a short time electronics will be added to the list. What’s next, electronic ransom?
Research at the Munich University of Applied Science and at INRS-EMT in Canada could soon open the possibility of printable computer memories. This is seriously no joke. Traditionally electronic memories have needed pricey photolithographic techniques to produce the circuits on silicon. Soon only a specialty inkjet printer will be required.
You should still read the fine print. The technology only has the capacity to produce low performance ReRAM on a flexible composite base. Reliability is also an issue. For those unacquainted, ReRAM is a form of non-volatile random access memory exploiting the variable resistance of a dielectric material. Some people call them Memristors.
Theorist Leon Chua originally speculated about the device back in 1971, while working as a professor at Berkeley. He realized that there is an intrinsic relationship between the flow of charge and magnetic flux. At the time this was simply a mathematical curiosity. Then in 1986 Robert Johnson and Stanford Ovshinsky filed a patent for the production of a 2-terminal configurable resistance switch. It validated the premise of the concept. This was followed in 1994 by the publication of Binary Information Storage At Zero Bias In Quantum-Well Diodes. In the Journal of Applied Physics article Buot and Rajagopal demonstrated the current voltage characteristics of the circuit element. Then in 1993 Katsuhiro Nichogi, Akira Taomoto, Shiro Asakawa, Kunio Yoshida of Matsushita Research Institute wrote a patent for a device of similar properties to a Memristor. Then in 1998, Michael Kozicki and William West submitted a patent for a programmable cell using metal dendrites. Two years later Philip Kuekes, Stanley Williams, and James Heath, of HP Labs presented a patent for a 2-terminal non-linear resistance switch. In 2005 Zhida Lan, Colin Bill, and Michael VanBuskirk filed a patent for a resistance switching memory cell. Wait for it. 3 years later: Dmitri Strukov, Gregory Snider, Duncan Stewart, and Stan Williams, of Hewlett Packard Labs publish an article in Nature claiming to link 2-terminal resistance switching to Leon Chua’s Memristor.
Memristors can operate as digital devices, but there is more to the story. In the Munich study the data is stored as either an on or an off. The dielectric is normally insulating, but can be made conductive by the application of a voltage. The applied voltage leads to the growth of a metallic filament. The filament can be removed by the application of another voltage.
The tools needed are straightforward. You will need an inkjet printer. PET foil sheets, but that’s only a few bucks. Next you’ll need silver ink, which runs about $400 for a 50-millileter cartridge. Roll on roll printing is another option. Don’t forget the printed circuit board design software. Cadsoft Eagle, KiCad, and gEDA project are three low cost options for experimenters. The last and most expensive item is an electronic engineering degree.
Be aware, the resolution of the printing process is limited. ReRAM memory is laid out in arrays of cells with each unit representing a bit. You could write a column of nine cells crossing another nine rows of cells. That would produce a total of eighty-one bits, which is equivalent to about ten bytes or two and a half words. Then there is the issue that the data can only be retained for about four hours. The memory also has a rewrite limit of about a thousand times. Not much, but you get what you pay for.
This should not limit the effectiveness. At these low costs a large number of applications are possible. The authors have envisioned printing smart identification tags for the retail and transportation industry. Wearable electronics is also another possibility. Industry could also consider printing expendable memories for weapons and sensors. At the writing of this story, one of the authors of this research Bernard Huber could not be reached for comment. Huber is a Doctoral student at Munich University.
At the writing of this ransom story, Memristors are still not commercially available. Nine years after HP announced the achievement, no chips have been released. HP once entertained placing the chips into a next generation computer nicknamed the Machine. It never materialized, but more on that later. In October 2015 HP joined with SunDisk to again bring the chip to market. It again evaporated to corporate restructuring.
Today an interested industrialist can still purchase a Memristor. No silver needed, maybe some. One small Santa Fe, New Mexico Company Known is selling trivial amounts of the chips for trial and experimentation. In the blog Alex Nugent said that one of the challenges is that Memristors are much more than switches. Most do not operate as fast binary switches. Instead they exhibit a range of amazing phenomena. Some can be programmed to be act as a capacitor, or a battery. Some will operate as a fast switch, but only probabilistically. Still others can be slowly adjusted for learning applications. One collaborator Dr. Kris Campbell has been able to achieve amazing control of the chips switching speeds, resistance ranges, and threshold voltages. She realized this after recognizing that Memristor behavior mimics the brain. This led her down the path of AHaH computing, which is otherwise known as Machine Learning.
Machine learning is expected to revolutionise industry in this century. In the last decade, it brought us shopping recommendations, streaming music playlists, and movie suggestions. It is still speculative but possible that Memristors will open the door to smart robots. Soon robots could learn the best sequence to assemble a complex product without the lengthy coding normally needed. The robot could also be taught how to check and maintain quality. You could build a product that would learn the preferences of the user. It’s also possible that you could write simple logic circuits directly into your robots memory.
The sky is the limit!