Fig. 4.

Cross-talk in e-skid waveguides. a Schematic of coupled e-skid waveguides on an SOI platform. The waveguide height, center-to-center separation between the two waveguides (s), and Λ are 220 nm, 1000 nm, and 120 nm, respectively. A cladding oxide is also added on the top of the waveguides. The number of ridges between the waveguides is dictated by the waveguide core size. b Top view SEM image of the coupled e-skid waveguides. c The experimental setup to measure the cross-talk between the two waveguides at the telecommunication wavelength (λ=1550 nm). Light is in-coupled to the first waveguide through the middle grating coupler. The second waveguide is coupled to the first waveguide for a length of L. In this experiment, the bending radius is 5 μm, hence, we can ignore the bending loss. d The ratio between the measured output powers for strip waveguide and e-skid waveguide vs. L at the telecommunication wavelength. The ratio for the e-skid waveguide is two orders of magnitude lower, indicating that far less power is coupled to the second waveguide. The inset shows the ratio for the waveguides without the top cladding oxide. In this case, the metamaterial cladding can increase the coupling length up to 30 times or reduce the cross-talk -30 dB. See Supplementary Figure 8 for more details. e Comparison of the simulated and measured coupling length for e-skid waveguides and strip waveguides. The coupling length is normalized to the wavelength. Error bars represent the standard deviation of the fitting curves. The optimum match between the simulation and experiment is achieved when ρ = 0.6 for the cladding of e-skid waveguides. The coupling length for e-skid waveguides is an order of magnitude larger in comparison with strip waveguides (shaded region). The coupling length for strip waveguides with larger core size decreases because the overlap between the evanescent tails is increased although more power is confined inside the core (unshaded region)