A new tunable-filter design should allow high performance with low mass-production costs, say researchers at the University of Darmstadt in Germany. The technique involves using the stress created in a Fabry-Perot stack to produce a cavity in a two-chip device without the need for sacrificial or spacer layers. Initial experiments have been successful: the device's tuning works as expected. However, in order for the device to be viable, efforts to transfer the technology to cheaper materials, and to reduce its capacitance will have to progress.
The tunable two-chip microcavity1 consists of two distributed Bragg reflector (DBR) stacks of SiO2/Si3N4, one on glass, and the other on indium phosphide. On top of the lower stack are electrode and passivation layers, etched out in the center to allow the filtered beam to pass through (see figure). The other stack also has an electrode deposited on it: this time a set of four cantilever arms. After the substrate carrying this stack has been thinned, the two stacks can be put together—face to face—and the back of the top layer etched away.
After the top DBR stack has been released in this way, the stress induced during deposition pushes it upwards to form a cavity. By applying a voltage across the top and bottom electrodes, the length of the cavity, and so the wavelength to be filtered out, can be changed.
Though the device worked well and predictably to tune out wavelengths as the voltage varied, the amount of voltage required was a problem. In tuning from approximately 1560 nm down to 1520 nm, 100 V was required. To lower the capacitance of the device, researchers are looking at reducing the thickness and/or raising the permittivity of the passivation layer. Optimizing the membrane design is another possibility.
Other improvements the Darmstadt team is considering include refining the processes used to lay down the DBR layers: the mirrors were not as reflective as they had hoped. By removing impurities and inhomogeneities, they expect to be able to increase the reflectivity from 99% to 99.5%. High losses (6 dB) should be improved through using a fiber lens with the correct beam-waist size, unavailable for the initial experiments. Finally, the group is looking at other suitable materials for the device: GaAs or Si for the substrate, and something like sputtered SiO2-TiO2 for the reflectors.
- M. Aziz et al., IEEE Phot. Tech. Lett. 12 (11) 1522 (November 2000).
Sunny Bains is a scientist and journalist based in London, England.