摘要： The confinement of light in three dimensions (3D) is an active research topic in Nanophotonics, since it allows for ultimate control over photons [1]. A powerful tool to this end is a 3D photonic band gap crystal with a tailored defect that acts as a cavity or even a waveguide [2]. When a one-dimensional array of cavities is coupled, an intricate waveguiding system appears, known as a CROW (coupled resonator optical waveguide) [3]. Remarkably, 3D superlattices of coupled cavities that resonate inside a 3D band gap have not been studied to date. Recently, theoretical work has predicted the occurrence of “Cartesian light”, wherein light propagates by hopping only in high symmetry directions in space [4]. This represents the optical analog of the Anderson model for spins or electrons that is relevant for neuromorphic computing and may lead to intricate lasing [5]. To experimentally study the propagation of light in a 3D superlattice of band gap cavities, we have fabricated 3D nanostructures from silicon by reactive ion etching (see Fig. 1). The photonic crystal has the cubic diamond-like inverse woodpile structure that is composed of two perpendicular sets of pores (in the X and Z directions) [6] and that reveals a broad 3D photonic band gap. By intentionally making two proximal perpendicular pores smaller, light is confined at their intersection, thus forming a cavity [7]. By fabricating two arrays of defect pores (see Fig. 1) we realize a 3D cavity superlattice. The challenge we address here is to experimentally identify resonances in this complex system. We measure reflectivity of the crystals with a large aperture (NA = 0.85) for both s and p-polarizations. With our setup we also record lateral scattered light, i.e., light that is input in the Z-direction, scattered, and detected in the X-direction. In presence of the intentional defects, we observe troughs in the stop band in reflectivity. The corresponding lateral scattering spectrum shows peaks in the stopband at the same frequencies as the reflectivity troughs, see Fig. 2. To distinguish cavity resonances from random speckle, we record several spectra at different locations on the crystal. Peaks reproducing at the same frequencies are attributed to light scattered from the cavity resonances. The appearance of the scattering peaks strongly depends on polarization in conformity with numerical calculations [7]. We now extend the numerical work to the range of pore sizes probed experimentally.