UC Berkeley

On-chip optical interconnects promise to drastically reduce energy consumption compared to

electrical interconnects, which dominate power dissipation in modern integrated circuits (ICs).

One key requirement is a low-power, high-efficiency, and high-speed nanoscale light source.

However, existing III-V semiconductor light sources face a high surface recombination velocity

(SRV ~ 10^4 – 10^6 cm/s) that greatly reduces efficiency at nanoscale sizes. An alternative material

system is the monolayer transition metal dichalcogenides (TMDCs), single-molecule-thick

direct-bandgap semiconductors that are being investigated for a variety of applications in

photonics and electronics. In particular, they are intrinsically nanoscale in one dimension and

lack dangling bonds at the surface, leading to high optical efficiency. In addition, they can be

electrically injected, transferred to arbitrary substrates, and processed in a top-down manner

similar to traditional semiconductors.

However, the development of electrically-injected light emitting devices based on TMDCs is still

in an early stage, with relatively few reports of high-efficiency light emission. In this

dissertation, we identify current decay over time as the main limitation preventing stable

operation of lateral-junction TMDC light-emitting diodes (LEDs), particularly in ambient (nonvacuum) conditions. To solve this, we propose operating WSe2 LEDs under pulsed voltage,

which shows much more stable light emission over time than DC operation (hours vs. seconds of

continuous device operation). Electroluminescence (EL) efficiency matches that of

photoluminescence, confirming material-quality-limited efficiency. In addition, we demonstrate

fast rise and fall times of ~15 ns, a record for TMDC LEDs.

For nanoscale light emitters that require high efficiency at low input power, LEDs are preferred

over lasers due to their lack of threshold requirement. The slow speed of LEDs can be greatly

enhanced by coupling the emitter to an optical antenna, extending its optical transition dipole

length and resulting in possible ~100x-1000x enhancement of spontaneous emission rate. We

experimentally demonstrate three electrically-injected antenna-coupled designs: the double-gate

LED coupled to bowtie antennas, the light-emitting capacitor coupled to a slot antenna array, and

the single-gate LED coupled to a nanosquare antenna array. The light-emitting capacitor shows

very high polarization ratios >30x, and the nanosquare array shows >10x improved intensity

over control devices, both indicators of strong antenna enhancement. We discuss the tradeoffs

involved in each design. Finally, we theoretically investigate high-speed, highly scaled TMDC

nanoLED devices and identify the limits to speed and efficiency. We conclude that edge

recombination can be largely overcome with sufficient antenna enhancement, while exciton-exciton annihilation ultimately limits efficiency at high injection levels. Under our model, high

speeds >70 GHz can be achieved at modest quantum efficiencies >10%. However, much further

work remains in improving material properties and devices.