New method gets better performance out of atomically thin transistors

Jan 30, 2024

John Timmer - Mar 21, 2023 3:55 pm UTC

Atomically thin materials like graphene are single molecules in which all the chemical bonds are oriented so that the resulting molecule forms a sheet. These often have distinctive electronic properties that can potentially enable the production of electronics with incredibly small features only a couple of atoms thick. And there have been a number of examples of functional hardware being built from these two-dimensional materials.

But almost all the examples so far have used bespoke construction, sometimes involving researchers manipulating individual flakes of material by hand. So we're not at the point where we can mass-manufacture complicated electronics out of these materials. But a paper released today describes a method of doing wafer-scale production of transistors based on two-dimensional materials. And the resulting transistors perform more consistently than those made with more traditional manufacturing approaches.

Most of the efforts made toward easing the production of electronics based on atomically thin materials have involved integrating these materials into traditional semiconductor manufacturing techniques. That makes sense because these techniques allow us to perform incredibly fine-scale manipulations of materials at high volumes. Typically, this has meant that much of the metal wiring needed for the electronics is put in place by traditional manufacturing. The 2D material is then layered on top of the metal, and additional processing is done to form functional transistors.

Often, that "additional processing" involves layering metal on top of the 2D material. This method, the researchers behind the work argue, probably isn't the best way of doing things. Depositing the metal can damage the 2D material, and some individual metal atoms can potentially diffuse into the 2D material, creating small short circuits within the larger feature. All of that degrades the performance of any circuitry built using the technique.

So the team figured out a way to form all the individual parts of the circuitry separately and brought them together under gentle conditions. The simplest part was forming the gates of the transistors, which were simply patterned on a solid substrate and then coated with aluminum oxide.

Separately, the team formed a uniform sheet of an atomically thin material (molybdenum disulfide) on top of a silicon dioxide surface by chemical vapor deposition. That sheet was then lifted off and transferred on top of the aluminum oxide, resulting in an atomically thin layer of semiconductor sitting on top of the gate. To form a transistor, the researchers were just missing source and drain electrodes.

Those were made completely separately by forming all of the wiring on top of a solid surface. The wiring was then embedded in a polymer, and the whole thing was peeled off of the surface, creating a sheet of polymer with the wires embedded on its lower surface. On its own, this polymer is flexible enough that it might stretch or distort, and thus the wiring wouldn't line up with the gates, as is needed to form functional circuits. To limit these distortions, the researchers linked the polymer to a sheet of quartz before stamping it down on the wafer covered with gate electrodes. This deposited the wiring directly on top of the molybdenum disulfide, completing the formation of functional transistors.

Once everything was in place, the polymer could be removed under mild conditions, and any excess material could be cut away using plasma etching. The result was a collection of transistors where the semiconductor's connection to the source and drain electrodes simply formed by the materials being physically placed next to each other. This limits the possibility of damage to the atomically thin semiconducting material.

While all of the processing needed here is much gentler than typical semiconductor manufacturing, that manufacturing simplifies matters by forming all the features where they're ultimately needed. For this approach to work, the source and drain electrodes are made separately from the gate and have to be dropped into place afterward. For circuitry with small features, that requires an incredibly precise alignment.

That... didn't always work out. There were a number of cases in which the entire collection of electrodes ended up out of alignment, typically because of a slight twist as they were dropped into place. This is something that can potentially be improved, but it's likely to remain a challenge.

The good news is that when it worked, it worked very well; the devices performed much more consistently than those produced using more typical techniques. And by most measures, they performed significantly better. The voltage in the on- and off-states differed by nine orders of magnitude. The leakage in the off-state was also very low.

More generally, the approach worked. The researchers were able to build functional circuitry across the entirety of a 2-inch wafer, including half-adder units, an essential component of computational hardware. So while this is clearly still in the demonstration phase, the demonstration is of hardware that could potentially be put to use.

That doesn't mean molybdenum disulfide is on the fast track to replace silicon. Decades of experience have made it possible to do some incredibly sophisticated things with silicon circuitry. But it does mean people are starting to develop the toolkits that might someday make 2D materials a viable competitor to silicon.

Nature Nanotechnology, 2023. DOI: 10.1038/s41565-023-01342-1