Scalable quantum dot amplifier based optical switch matrix

A quantum dot active layer is used in a novel, highly scaleable monolithic optical switch matrix architecture. Electronically paired semiconductor optical amplifiers gates are implemented in a four-input four-output configuration to reduce the electrical connections and control complexity. Low power penalty 10Gb/s routing at a wavelength of 1555nm is demonstrated. Increasingly high-capacity data transfer in storage area networking, high performance computing, and server networks is driving research into increasingly elaborate photonic switched interconnect test-beds (1-4). The need for low-latency, high capacity, scalable switch fabrics with low driver complexity, excellent crosstalk and the broad gain bandwidth has lead to a particular focus on semiconductor optical amplifier (SOA) based switches. However, considerable integration is required to remove complex packaging-related restrictions such as the high numbers of fibre pigtails, power consuming cooler circuits, and the complex electronic control circuits. To date, integrated SOA based switch designs have focused on gate arrays in a broadcast and select architecture. The gates and inputs are connected via splitters and combiners utilising fibre (4) and waveguides with hybrid integration (5) and epitaxial regrowth (6). The latter has lead to the smallest circuits with 20mm 2 footprints. The bulk active layer amplifier designs implemented have however lead to low saturation powers. The resulting low distortion threshold necessitates a lower number of photons per bit and thereby impairs signal to noise ratio and data capacity. The broadcast and select architecture requires a high number of waveguide crossings and exhibits a square law scaling in the required electrical control signals with the number of optical inputs. This already leads to 16 independent controls for a four input four output switch. In this work, we present the first implementation of a four input, four output quantum dot based switch matrix. The design is based on a new crossbar element design which both integrates the waveguide crossing within the gate and is implemented with common electrodes to halve the required numbers of electrical connections. The use of complementary electrical signals further halves the number of independent control signals to four. Integration of the crossing within the gate also allows a reduced size for the shuffle network, and in combination with low loss, low distortion, quantum dot epitaxy, this allows an all-active implementation in a reduced chip area of only 3mm 2 .