“We were looking to reach the ultimate density limit of bit-wise storage in a solid, which obviously is coding the information of one bit in one atom,” says team member says Roland Bennewitz. “The atoms must have a certain distance between them so that they do not interact and thereby obscure the information. Using the presence or absence of an atom on a self-organized pattern provides these requirements in an elegant way.”

The scientists created the memory by depositing 0.4 monolayers of gold onto a Si(111) surface at 700°C. Then an annealing treatment at 850°C created the well-known Si(111)5x2-Au structure.

“The memory consists of self-organized gold wires that support the silicon atoms at regular distances,” explains Bennewitz.

The memory’s self-assembled tracks are five rows of atoms – or 1.7 nm – wide. Each bit is encoded by the presence or absence of a silicon atom inside a two-dimensional unit cell of 5 x 4 atoms. The other 19 atoms in the unit cell stop adjacent bits interacting with each other.

Data can be read from or written to the memory using a scanning tunnelling microscope, or STM. The scientists preformatted the memory with ones by the controlled deposition of silicon onto vacant sites. To write zeros, they used the STM tip to remove silicon atoms from the surface.

“It is interesting to note that the system we propose meets the ultimate density predicted by the mastermind of nanotechnology, Richard Feynman,” added Bennewitz. “In his legendary talk he came up with 5 x 5 x 5 atoms to store one bit – in our system we use 5 x 4 atoms.”

Bennewitz says that the system is stable at room temperature, which is a big step forward compared with earlier low-temperature atom-manipulation experiments. What’s more, the atoms are positioned at well-defined distances along tracks, which allows the use of systematic read-out procedures well known from magnetic hard disks.

However, there are two drawbacks: the memory needs to be prepared and preserved in a vacuum, and the writing and reading speeds are relatively slow.

“When comparing storage density and read-out speed of our atom memory with the information density and replication speed in DNA, we end up with comparable numbers,” said Bennewitz.