Monolithic 2 x 2 switches use QW intermixing
by Yvonne Carts-Powell
A novel method of creating multiple bandgaps within a single active layer allowed Bo-cang Qiu and others at University of Glasgow (Scotland) and University of Sheffield (England) to fabricate low-loss monolithically integrated 2 x 2 crosspoint switches in InGaAs-InAl-GaAs. Components of the switches include waveguides, electroabsorption (EA) modulators, and semiconductor optical amplifiers (SOAs) created on one chip using sputtered SiO2 quantum-well (QW) intermixing.
Ideally, an integrated optical switch should operate with low crosstalk, low loss, and high switching speeds. However, the different devices that are part of most switches (including SOAs, EA modulators, and passive components) need different semiconductor bandgaps within one epilayer.
Several methods can be used to create different bandgaps in one layer, including regrowth, selective growth, or vertical coupling. The Glasgow group has chosen quantum-well intermixing. This is a postgrowth technique that modifies the bandgap of III-V quantum-well materials and allows cost-effective and flexible monolithic integration of passive and active components.
The material used to fabricate the switches was a laser structure with an emission wavelength of 1.54 μm. It was grown with an active region consisting of six 7.5-nm InGaAs wells with InAlGaAs barriers. The wells were placed in the center of a 250-nm waveguide layer of InAlGaAs. The device was grown between top and bottom cladding layers on InP, and the cladding layers were p and n doped, respectively. An InGaAs contact layer was fabricated atop the top cladding layer.
The quantum-well intermixing technique involves deposition of a thin film of sputtered SiO2, which causes point defects at the surface. Annealing causes the defects to diffuse into the QW regions. This enhances the interdiffusion between the wells and the barriers.
The material was patterned with different thicknesses of SiO2 plus, in some cases, a resist layer, before sputtering and annealing. In areas with thinner layers of SiO2, more point defects are formed by sputtering, and after annealing, the photoluminescence of these areas is shifted towards shorter wavelengths.
The components of the switch require three bandgaps: the amplifiers work at 1.55 μm (to match the optical window of silica fiber), the waveguides need a bandgap shifted to the blue as much as possible (typically 80 nm) to reduce direct bandgap absorption. And the researchers chose a bandgap wavelength of 20 to 50 nm shorter than the operational wavelength (1.55 μm) for the EA modulators, which balances the modulator's extinction ratio and insertion loss.
The amplifier regions were protected by 200-nm-thick SiO2 plus a thick photoresist layer; then they were covered with 100-nm SiO2 only and there was no protection layer on the top of passive regions. After sputtering, the photoresist was removed and the sample was annealed at 650°C for 60 s, resulting in bandgaps in different regions of the device (see figure).
The extinction ratio of the devices is as high as 26 dB, and the total gain of both amplifiers in a particular route is at least 10 to 12 dB, giving the device an on-chip insertion loss of 4 to 5 dB.
1. B. C. Qiu, et al., IEEE Photon. Tech. Lett.. 13(12), 1292 (December 2001).