Metawaveguide for Asymmetric Interferometric Light-Light Switching

Light-light switching typically requires strong nonlinearity where intense laser fields route and direct data flows of weak power, leading to a high power consumption that limits its practical use. Here we report an experimental demonstration of a metawaveguide that operates exactly in the opposite way in a linear regime, where an intense laser field is interferometrically manipulated on demand by a weak control beam with a modulation extinction ratio up to approximately 60 dB. This asymmetric control results from operating near an exceptional point of the scattering matrix, which gives rise to intrinsic asymmetric reflections of the metawaveguide through delicate interplay between index and absorption. The designed metawaveguide promises low-power interferometric light-light switching for the next generation of optical devices and networks.

Figure 1

Asymmetric interferometric light-light switching: (a) Schematic of a metawaveguide with asymmetric reflection, i.e., $|r_L|\ne |r_R|$. The intrinsic reflection asymmetry in the vicinity of the quasi-PT exceptional point facilitates asymmetric interferometric light-light switching of a strong source signal (forward input) by a weak control field (backward input). (b) Electric field distributions of interferometrically controlled asymmetric light-light switching, where the power ratio of the weak control to the strong source signal is set to $1:3$. If the incidenct phase of the control is $\phi =\pi /2$ (upper panel), the CPA mode is achieved with no scattering, validated by the vanishing of interference patterns between inputs and outputs in both ports outside the modulation region, whereas if the incident phase of the control is $\phi =-\pi /2$ (lower panel), the degenerated mode of less absorption is demonstrated, leading to strong output scatterings in both ports that can be seen by the interference patterns outside the modulation region.

Figure 2

Metawaveguide for asymmetric interferometric light-light switching: (a) Design of the non-Hermitian metawaveguide to create a spatial modulation equivalent to that in Eq. (2): The real index modulations of are shifted $2\pi /q$ along the length of the waveguide and emulated using sidewall modulations with a transverse modulation depth cosine-varying from $+71$ to $-48.5\mathrm{nm}$; the imaginary absorption modulations remain unchanged at their original locations, mimicked using bilayer sinusoidal shaped combo structures of 9.8 nm chrome $(\mathrm{Cr})/8\mathrm{nm}$ germanium (Ge) deposited on top of the Si waveguide. (b) SEM picture of the device consisting of 38 periods for strong signal light switching by a weak control. (c) Zoom-in picture of the metawaveguide. The profile of the waveguide sidewall along with $\mathrm{Ge}/\mathrm{Cr}$ combo structures on top, respectively, realizes the designed real and imaginary modulation.

Figure 3

Characterization of asymmetric interferometric light-light switching: (a) Configuration of the experiment setup. (b) Spectra of maximum (red) and minimum (blue) output scattering coefficients as a function of wavelength detuning ($\mathrm{\Delta }$) to the resonant waveglength. Insets: microscope snapshots of the scattering (top) and CPA (bottom) modes at the resonance under an exposure time of 15 ms. While the designd resonance is located at 1550 nm, the measurement results show a resonant wavelength of approximately 1540 nm. The overall detection loss due to the light splitting at the directional coupler is estimated to be 6 dB at 1540 nm. Nevertheless, the experimental results (triangles) of asymmetric interferometric light-light switching remain the same with respect to the wavelength detuning, showing an extinction ratio of approxmiately 60 dB at the resonance.

Figure 4

Phase responses of outputs in interferometric light-light switching. (a) When operated at the resonant wavelength, two outputs oscillate in phase and reach their minimum simultaneously at $\phi =\pi /2$, where almost no light is scattered from the two grating couplers of outputs. As the phase difference is flipped to $\phi =-\pi /2$, peak output scattering is obtained from both grating couplers. (b) When operated at off-resonance wavelength $\mathrm{\Delta }=-4.2\mathrm{nm}$, due to extra phase shift, the two output oscillations move forward. Since the output $O_2$ accumulates more phase than $O_1$, the responses are no more in phase and thus the two outputs reach minimum or maximum asynchronously. (c) When operated at off-resonance wavelength $\mathrm{\Delta }=4.6\mathrm{nm}$, the extra phase shift changes sign and results in a shift of output oscillations in the opposite direction. Note that neither of the output powers can be completely eliminated at $\mathrm{\Delta }\ne 0$ regardless of the phase tuning. The output power is normalized to the total incident power $I_1+I_2$ in the plots.