Development of Highly Homogenous Quantum Dot Micropillar Arrays for Optical Reservoir Computing

Neuromorphic computing has received considerable attention as promising alternatives to classical von Neumann computing architectures. An attractive concept in this field is reservoir computing which is based on coupled non-linear elements to enable for instance ultra-fast pattern recognition. We focus on the development of microlasers in a dense regular array for the implementation of photonic reservoir computing based on the diffractive coupling. The coupling relies on injection locking of microlasers and sets stringent requirements on the spectral homogeneity of the array, which needs to be on the order of the achievable locking range. We realize CaAs/AlCaAs micropillar arrays with InGaAs quantum dots as active medium. To achieve the high spectral homogeneity on the order of 100 mu eV, as determined by injection locking experiments, the emission energy of each individual micropillar is adjusted to compensate for local inhomogeneities of order similar to 1.3 meV in the underlying microcavity structure. The realized micropillar arrays have a spectral inhomogeneity as low as 190 mu eV for an 8 x 8 array and down to 118 mu eV for a 5 x 5 sub-array. The arrays have high potential to enable the implementation of powerful photonic reservoir computing, which can be extended to a reservoir of hundreds of microlasers in the future.