Probe Memory Packs 138 Terabytes per Square Inch

New automated approach could help make STM probe memory commercially viable

4 min read
maple leaf and alphabet
To demonstrate their new automated probe memory technique, Canadian researchers not only fabricated the world's smallest maple leaf, they also encoded the entire alphabet at a density of 138 terabytes, roughly equivalent to writing 350,000 letters across a grain of rice.
Image: Roshan Achal/Nature Communications

Researchers at the University of Alberta in Canada have developed a new approach to rewritable data storage technology by using a scanning tunneling microscope (STM) to remove and replace hydrogen atoms from the surface of a silicon wafer. If this approach realizes its potential, it could lead to a data storage technology capable of storing 1,000 times more data than today’s hard drives, up to 138 terabytes per square inch.

As a bit of background, Gerd Binnig and Heinrich Rohrer developed the first STM in 1986 for which they later received the Nobel Prize in physics. In the over 30 years since an STM first imaged an atom by exploiting a phenomenon known as tunneling—which causes electrons to jump from the surface atoms of a material to the tip of an ultrasharp electrode suspended a few angstroms above—the technology has become the backbone of so-called nanotechnology.

In addition to imaging the world on the atomic scale for the last thirty years, STMs have been experimented with as a potential data storage device. Last year, we reported on how IBM (where Binnig and Rohrer first developed the STM) used an STM in combination with an iron atom to serve as an electron-spin resonance sensor to read the magnetic pole of holmium atoms. The north and south poles of the holmium atoms served as the 0 and 1 of digital logic.

The Canadian researchers have taken a somewhat different approach to making an STM into a data storage device by automating a known technique that uses the ultrasharp tip of the STM to apply a voltage pulse above an atom to remove individual hydrogen atoms from the surface of a silicon wafer. Once the atom has been removed, there is a vacancy on the surface. These vacancies can be patterned on the surface to create devices and memories.

“We have automated the removal procedure so that it is possible to enter a design, and have it created without user intervention,” explained Roshan Achal, a Ph.D. student at the University of Alberta and lead author on the research described in the journal Nature Communications.

A demonstration of how the removing and replacing of hydrogen atoms on silicon surface can create the 0s and 1s of digital logic is shown in the video below.

While Achal explained that this process is highly accurate, it is not perfect.  Sometimes hydrogen atoms can be removed from incorrect locations. This type of error would require another attempt from scratch to create a design.

“We have developed a procedure that allows us to replace individual atoms on the surface, to erase these mistakes, instead of starting from zero each time an error occurs,” he explained.

In order to replace an atom on the surface, a hydrogen atom must be sitting on the surface of the tip. By bringing the tip closer and closer to the surface, it is possible for it to jump from the tip to the surface. “We discovered that there are two unique signatures when a transfer occurs, that can be used to help automate this procedure as well,” he added.

While these techniques could fall into the broad categories of nanoscale fabrication, the major difference with this technique, according to Achal, is the level of accuracy and degree of automation that has been achieved in fabricating structures on this silicon surface. In addition, no other nanofabrication technique had managed to replace in a controlled way several hydrogen atoms on the surface in succession.

Perhaps one of the biggest restrictions of STMs in data storage has been the need for cryogenic temperatures. However, with this latest approach, the hydrogen-terminated silicon system, by its nature, enables the fabrication of structures that are stable well above room temperature.

“This stability comes at the cost of increased difficulty in fabricating structures,” said Achal. “However, with these new techniques we have overcome many of the associated problems, making this system a very interesting candidate for new technological applications.”

This approach certainly addresses the issue of cryogenic temperatures being needed to allow such a device to work, but does it scale? Achal argues that there are no physical limitations preventing the speeds of these processes from reaching practical levels.

Achal and his colleagues are looking into new schemes to improve speeds, but there are also currently available ways to scale these processes that don’t involve changing the procedures significantly. The most accessible solution would be the parallelization of many tips inside an STM, according to Achal. He also noted that there are some tip materials that are known to hold up to a thousand hydrogen atoms simultaneously. If these materials turn out to be viable options for a tip, then they would allow for even faster erasing/rewriting speeds.

In the meantime, the specter of IBM’s Millipede project looms over proposals of massive parallelization. The IBM Millipede essentially used an array of thousands of miniaturized Atomic Force Microscopes (AFMs) as a memory device. Paul Seidler, a researcher at IBM Research in Zurich, said seven years ago that IBM had abandoned the Millipede project as an alternative data storage technology and instead had found its most likely role as probes for lithography.

If this STM approach is to experience a more successful fate than the Millipede project in data storage, Achal and his colleagues will need to take the huge step of parallelization of tips inside a STM.

The Conversation (0)