Quantum-well semiconductor is an optical buffer

To date, realization of the much-sought-after all-200 optical communications network has been stymied in large part by the lack of a suitable optical buffer with which to delay and store optical ­signals (see figure).

Th 162909

To date, realization of the much-sought-after all-200 optical communications network has been stymied in large part by the lack of a suitable optical buffer with which to delay and store optical ­signals (see figure). Optical-fiber delay lines and waveguide-dispersion structures have been considered as possible optical buffers in the past, but the inability to set random storage times limits the former technique, and the latter is limited by narrow signal bandwidth.1

Th 162909
Contention resolution in an optical switch is one of several potential applications of an optical buffer.
Click here to enlarge image

Results reported by researchers from the University of California (UC) at Berkeley, the University of Oregon (Eugene), and the University of Illinois at Urbana-Champaign, however, indicate that optical buffers without such limitations can be made in semiconductor materials, which also offer a capacity for monolithic integration.2

Based on experimentally measured ­dispersive characteristics of population-oscillation effects in optical signals traveling through semiconductor quantum-well structures, the researchers reported “slow light” group velocities as low as 9600 m/s with a transparency bandwidth window as high as 2-GHz full-width half-maximum (FWHM).

Group-velocity reductions on the order of seven orders of magnitude have been achieved in atomic vapors, both near absolute zero and at 80°C, and light has actually been stopped for periods as long as 0.5 ms in praseodymium-doped crystals at low temperatures (5 K). The UC Berkeley-led team observed the effect in the thin layers of a gallium arsenide/aluminum gallium arsenide (GaAs/AlGaAs) semiconductor quantum-well structure for the first time, according to team leader Connie ­Chang-Hasnain, who is also director of UC Berkeley Center for Optoelectronic Nanostructured Semiconductor ­Technologies.

“Semiconductors have a million to a billion times broader bandwidth capacity than atomic gas or crystal,” she said. “This brings real-world practicality to telecommunications and network applications. Another advantage to using semiconductors is the possibility of cost-effective integration into ­circuitry, specifically photonic integrated circuits.”

Two approaches

Regardless of the medium, slow-light experiments to date have involved coherent induction of a sharp and pronounced absorption dip in an optically thick medium to generate a steep dispersion in the index of refraction over a narrow spectral range using either electromagnetically induced transparency or population oscillation, according to Chang-Hasnain. “The resulting reduction of the group velocity scales inversely with the spectral linewidth of the induced transparency window,” she said.

Electromagnetically induced transparency involves the artificial ­creation of a transparent spectral region based on destructive quantum interference between two transitions in a three-level system. An external light source controls the dispersion characteristic of the medium (measured by changes in refractive index), and thus the corresponding speed reduction in a signal passing through the medium. These effects have proven difficult to produce in semiconductors, because rapid decoherence of optical excitations severely limits the degree of group-velocity reduction. Semiconductor quantum dots may help to solve this problem with discrete electronic energy states caused by their 3-D confinement. But the nonuniform size and ­energy-­resonance distributions of quantum dots present substantial barriers that have yet to be overcome.

Population oscillation involves simultaneously illuminating the sample with pump and less-intense probe beams of slightly different frequencies that create a rhythmic beating pattern of light-slowing interference. Consequently, slowing light via population oscillation depends only on the lifetime of the grating induced by the two coherent beams (on the order of a relatively long 1 ns in a high-quality ­quantum-well structure). The interaction induces a dipole that oscillates at the probe frequency, propagates along the direction of the probe beam, interferes destructively with the dipole induced by the probe alone, and ultimately leads to a narrow dip in the probe-absorption spectrum. This narrow spectral dip can help to overcome the problem of rapid decoherence, and enabled the Berkeley-led research team to produce slow light in a semiconductor.

Each of the 15 semiconductor quantum wells in the experiment consisted of a thin layer of electron-confining GaAs sandwiched between two layers of AlGaAs. They were grown by molecular beam epitaxy (MBE) and mounted on a sapphire disk with the substrate removed. A continuous-wave Ti-sapphire laser and a tunable diode laser were used to provide the pump and probe beams, respectively. The researchers conducted the experiments at a temperature of 10 K and are currently working on electromagnetically induced transparency in semiconductors to produce slow light at room temperature.


1. C.J. Chang-Hasnain et al., Proc. IEEE 91(11) 1884 (November 2003).

2. P-C. Ku et al., Optics Lett. 29(19) 2291 (October 2004).

More in Research