Playing Rough to Get the Light Out

Hank Hogan

Researchers at the Electronics and Telecommunications Research Institute of Daejeon, South Korea, have shown that, when it comes to getting a silicon quantum dot LED to shine, it helps to play rough — at least with the surface. The group enhanced the light-extraction efficiency of the diode 2.8 times by roughing up its surface in micron-scale patterns. The technique could improve the performance of silicon quantum dot LEDs significantly.

Because silicon has an indirect bandgap, devices built with the material are less bright than those constructed from other semiconductors. Silicon, however, has several advantages, including cost, that make it the material of choice for electronic devices; therefore, improving the optical performance of silicon-based devices is important. One possible solution, the researchers found in a previous study, is the use of a silicon quantum dot LED. The challenge has been to boost its external quantum efficiency.

To achieve this goal, they turned to geometry and optics. For most LEDs, external quantum efficiencies are low because the light has trouble getting free. Calculations showed that the situation could be improved by capping the light-emitting layer with a layer of material with a different refractive index, provided that the second layer was patterned in a series of hills and valleys. Unlike some other schemes to boost performance, this approach has the advantage of being easily integrated with standard semiconductor processing.

In a proof of principle of this idea, the scientists fabricated an LED comprising an indium tin oxide top electrode, an N-doped layer of silicon carbide, an active layer of nanocrystalline silicon embedded in amorphous silicon nitride and, finally, a P-type silicon substrate. They capped this with a micron-thick layer of amorphous silicon nitride, which they etched to create bumps of various periodicities.

These bumps increased the escape angle of light emission — and therefore the quantum efficiency of the device — by a factor of 2.8 in the best case. The performance was a function of the ratio between pattern size and period of the bumps, with a ratio of ~0.7 optimum.

Applied Physics Letters, Nov. 6, 2006, 191120.