Electronic circuit interconnects in high-speed computing systems are becoming ever more power hungry and the race is on to develop chip-integrated optical communications devices to replace these. Silicon photonics is the architecture of choice today because it will allow components like waveguides, couplers, interferometers and modulators to be directly integrated into existing silicon-based processors. Happily, researchers have recently discovered that a class of 2D materials known as transition metal dichalcogenides (TMDs) could be used to make optical interconnect components that can indeed be integrated with silicon photonics and complementary metal oxide semiconductor (CMOS) processing.

TMDs have the chemical formula MX2, where M = Mo, W and X= S, Se and are promising for a variety of electronics and optoelectronics device applications such as light-emitting diodes and solar cells thanks to the fact that they go from being indirect-bandgap semiconductors in the bulk to direct-bandgap semiconductors when scaled down to monolayers. The electrons in TMDCs also interact exceptionally strongly with light, which means that even though these materials are just a few atoms thick, a large portion of absorbed photons can be used to produce electric current.

p-n junction works as an LED or photodetector

Pablo Jarillo-Herrero and colleagues of the Massachusetts Institute of Technology (MIT), together with co-workers at Columbia University in New York, the Barcelona Institute of Science and Technology and The National Institute of Materials in Tsukuba, have now integrated an electrically-powered light source made from the 2D material MoTe2 on a silicon photonic crystal waveguide for the first time. MoTe2 has a bandgap of around 1eV at room temperature.

The researchers begin by shaving off thin layer flakes of MoTe2 from the bulk crystal using the now famous scotch-tape technique. “With two separate metal gates, we then create a monolayer (or bilayer) MoTe2 p-n junction in which electrons and holes can combine to emit light so that it works as an LED,” explains team member and lead author of the study Ya-Qing Bie. “Electrons and holes can also be separated at this junction to generate current as it also works as a photodetector.”

Process needs to be optimized

Since the wavelength of the light emitted from the MoTe2 lies outside the wavelength range at which silicon absorbs light, losses coming from light absorption by the silicon are dramatically reduced, he tells nanotechweb.org. This makes it very different from its TMD siblings MoS2 or WSe2, recently reported on by many other research groups, including ours. “To make the integrated device, we stack the bilayer MoTe2 p-n junction and a silicon crystal waveguide together. The light emitted by the LED can travel though the waveguide to another location on the structure, and conversely, the p-n junction can detect the light coming from the waveguide as well.”

Although integrating MoTe2 on silicon photonics is a breakthrough, the process we developed will have to be optimized before it can be used to commercially fabricate optical interconnects, insists Bie. “To this end, we could, for example, make high-speed data communication devices by combining the light source with an efficient on-chip modulator or by integrating the electrically pumped TMD gain materials with photonic crystal nanocavities, which should further increase the optical coupling efficiency and allow for applications like wavelength division multiplexing. The good thing is that we are already able to tune the light source’s wavelength by integrating with different layered 2D materials.”

The research is detailed in Nature Nanotechnology.