As flat as it gets: the first 2D microprocessor

Roberto Saracco May 23, 2017 Blog 231 Views

The chip is three atoms thick. Overview of the entire chip. AC = Accumulator, internal buffer; PC = Program Counter, points at the next instruction to be executed; IR = Instruction Register, used to buffer data- and instruction-bits received from the external memory; CU = Control Unit, orchestrates the other units according to the instruction to be executed; OR = Output Register, memory used to buffer output-data; ALU = Arithmetic Logic Unit, does the actual calculations. Credit: TU Wien

Chip manufacturers are at work to develop 3D structures that can support ever denser processing and storage capabilities. In February 2017 Toshiba has introduced the first 3D flash chip storage stacking 64 layers, and in April 2017 SK Hynix unveiled a 73 layer 256Gb 3D NAND memory.

Yet, as industry is busy in finding ways to stack more and more layers in a single chip, some researchers are taking a complementary approach: decreasing the thickness of the material used to create transistors.

At Vienna University of Technology, TU Wien, a team of researchers has developed a chip based on a layer composed by just 3 atoms, 3 atoms thick. Theoretically, graphene could work as a transistor, hence, theoretically, one could create a chip that is just one atom thick. In practice this is impossible, at least with today’s technology, from an industrial point of view. Current chips use thousands of atoms to store a bit (in a flash memory), mother Nature is more efficient being able to store a bit in less than 50 atoms (those are the atoms used, on average by the sequences AGCT). Researchers have explored amazing ways to store bits, the current record , dating back to 2009, to my knowledge is storing 35 bits on a single electron! The problem is that to be able to use these few atoms you need a machinery, like scanning electron microscopes, that have trillions and trillions of atoms…

The result obtained at Tu WIen, on the contrary, has an industrial perspective. The chip created has just 115 transistors, nothing if you compare it to a normal chips where transistors are likely to exceed 1 billion. However, 115 transistors are sufficient to process, store and communicate data, what is basically needed for a sensor. Being so thin, 3 atoms thin!, it is very flexible and can be ideal for biosensors. The transistors are not made in silicon but in Molybdenum disulphide a compound that has characteristics similar to graphene.

It is clearly a prototype but it has been designed with an industrial production process in mind. We won’t be seeing it on shelves anytime soon but it shows that we have really just begun!