Research

Metamaterials made of nonlinear materials can unlock more advanced and exotic responses as compared to their linear counterparts. For example, by combining nonlinear responses with geometrical asymmetries, it is possible to realize optical devices which break Lorentz Reciprocity, allowing electromagnetic waves to propagate with different efficiency along opposite directions, and without the need of any applied biases.¹ We have demonestrated the implementation of this principle in integrated silicon photonics² and in free space metasurfaces made of amorphous silicon operating in the near-infrared,³ achieving nonreciprocal contrasts larger than 10 dB and insertion loss lower than 3 dB.

One of beauties of the field of metamaterials is that similar ideas and design principles can be applied to vastly different material platforms and spectral ranges. For example, artificial nonlinearities — intersubband transitions in multi-quantum-wells — combined with metallic patch antennas can be used to realize highly nonlinear metasurfaces operating in the mid-IR. ^4,5, with appealing applications in spectroscopy, materials processing, sensing, security, and satellite technologies. By leveraging the ultrafast Kerr nonlinearity of these materials, we were able to demonstrate thin metasurfaces acting as saturable absorbers and power limiters. In particular, we demonstrated experimentally saturable metasurfaces where the reflection level changes from more than 60% to less than 10% over a timescale shorter than 2 ps. ^4,5

Nonlinear materials also offer the possibility to extend analog optical computation to the realm of nonlinear mathematical operations, which are essential in applications such as neural networks. Remarkably, nontrivial image processing tasks can be achieved even with unpatterned materials. In this context, we have experimentally shown⁶ that thin unpatterned films of materials with strongly anisotropic χ⁽²⁾ tensors, such as Gallium Arsenide, can impart nonlinear mathematical operations on impinging images, encoded in the produced second-harmonic image. Since this approach does not rely on any optical resonance, it is inherently broadband and isotropic, and it features very large numerical apertures.