Abstract

Silicon nitride has been proposed and investigated recently as an appropriate material for integrated nonlinear signal processing applications due to its ultra-low loss and negligible two-photon absorption (TPA) at Telecom band [1-3]. Performance of the devices using Kerr nonlinearity in silicon nitride waveguides is limited by the accumulated nonlinear phase shift (NLPS). To enhance the Kerr nonlinearity in silicon nitride waveguides, slow light effect can be used [4]. Slow light effect can be enhanced at the band edge of dispersion diagram; however, this enhancement comes at the cost of an increased loss, which limits the accumulated nonlinear phase shift of the whole structure. In order to realize slow light waveguides at 1550 nm using ordinary 130 nm CMOS compatible foundries, corrugated structures are used in this study. It has been shown before that the dispersion diagram of such a slow light waveguides at band edge vicinity can be modelled by a quadratic function [5]. Therefore, the slow light bandwidth can be obtained using the band edge model. The slow light bandwidth is defined by the frequency ranges in which the desired slow down factor has limited (e.g. 5% or 10%) variations. Slow down factor is defined as the ratio of slow light group index to the group index of a waveguide without slow light effect. We show that the slow light bandwidth is inversely proportional to the square of the slow down factor (BW~ 1/S², where S is slow down factor). Bandwidth calculation is obtained also by electromagnetic full wave simulations based on finite element method (FEM) approach for a corrugated waveguide designed to work at 1550 nm (see Fig. 1a). The bandwidth comparison between analytical modelling and simulation is shown in Fig. 1b.

© 2017 IEEE