Ab-initio modeling of bend-losses in silicon nitride waveguides from visible to near-infrared

Abstract

Integrated photonics features applications in high-speed telecommunication, computing, and sensing. These devices are ultimately limited by the optical loss occurring in the waveguide structures. One of its primary sources is surface-roughness-induced scattering and bend-losses. Surface roughness is unavoidably introduced during deposition, mainly during etching and lithographic steps. In photonic integrated circuits, tight bends enable a compact footprint yet increase the mode mismatch loss , radiation loss and scattering loss. Previously, the bend losses were estimated from a parametric model. However, it lacks flexibility w.r.t. the waveguide platform. We apply a recently developed model of the surface-roughness-induced scattering in guided-mode systems to substantiate the dependence of the scattering loss on the bend-radius for waveguides based on a silicon nitride platform. The model incorporates the surface roughness via its autocorrelation. Further, it inherently considers the overlap of the modes with the roughness. As waveguide material, we used both plasma-enhanced CVD silicon nitride as a low-temperature, back-end-compatible process, and low-pressure CVD silicon nitride, as a high-temperature frontend process. As bottom and top cladding, we deposited high-density plasma (HDP) and sputtered silicon oxide, respectively. The latter offers flexibility to adapt the platform for sensing purposes. We evaluate different waveguide widths, bend radii, and wavelengths in the visible and near-infrared ranges. We set the observed propagation losses into context with estimated absorption, scattering, and mode-overlap loss sources and point to their shifting importance at the measured wavelengths. We believe that this model allows to increase our knowledge about the various aspects of loss in guided mode systems and predict the propagation loss based on foregoing absorption and roughness measurements.

Enhancing silicon nitride waveguide performance: optimization of sidewall roughness for low-loss applications through resist reflowing via tree-based modeling of relevant processing parameters

“…The roughness for the undercladding, waveguide top-surface, and sidewalls was assessed using an Anton Paar Tosca 400 Atomic Force Microscope, with sidewall measurements made using a tilting technique, which will be explained in more detail in the consecutive section. 4 The undercladding was either produced via high-density plasma-enhanced chemical vapor deposition (HD-PECVD) at a substrate temperature of 450 • C, or by thermal growth in a wet chemical process at 1200 • C. The Si 3 N 4 waveguide layer was deposited using plasma-enhanced chemical vapor deposition (PECVD) at 400 • C or low-pressure chemical vapor deposition (LPCVD) at 770 • C. Finally, the uppercladding was deposited either using the same HD-PECVD process or through sputtering at 100 • C using polycrystalline Si as the target material.…”

“…3 This study utilizes qualitative comparison of roughness and absorption utilizing atomic force microscopy (AFM) for precise roughness measurements and ellipsometry. [4][5][6] The study delves into techniques aimed at reducing sidewall roughness (SWR) in photonic integrated circuits (PICs), with a particular emphasis on the resist reflow method. A critical aspect of this research is the implementation of the comprehensive design of the experiments (DOE), specifically utilizing a full-factorial approach, to explore the effects and interactions of various variables at the wafer level, which is vital for analyzing propagation and bend excess losses in these complex systems.…”

CART-based optimization of core and cladding layers in silicon nitride photonic integrated circuits towards propagation and bend loss minimization

“…The roughness of the undercladding surface, the waveguide top-surface and the sidewall (measured with a tilting technique) were determined utilizing the Anton Paar Tosca 400 Atomic Force Microscope. 13 As undercladding, either high-density plasma-enhanced chemical vapor deposition (HD-PECVD) SiO 2 , deposited at 450 • C substrate temperature, was utilized or thermal SiO 2 , grown in a wet chemical process at a 1200 • C substrate temperature in a furnace. The Si 3 N 4 waveguide layer was deposited via plasma-enhanced chemical vapour deposition (PECVD) at 400 • C substrate temperature and low-pressure chemical vapour deposition (LPCVD) at 770 • C substrate temperature in a furnace.…”

“…This approach reflects a thorough strategy to tackle AFM data complexities due to misalignment, focusing on surface and sidewall roughness measurements. 13 The data can be seen in Table 2.…”

“…17,18 The reason for the fact that for LPCVD Si 3 N 4 , neither cladding is taken as a node for widths of the waveguide greater than 620 nm can be accounted for by the high confinement of the TE-mode at these widths, which, due to the slightly higher refractive index of LPCVD Si 3 N 4 even increases the confinement. 13 The reason the uppercladding dominates at 350 nmwaveguide width over the undercladding might be because of the non-conformality in combination with the interaction of the mode on the interface between the waveguide and SiO 2 , where there are also air gaps. For PECVD Si 3 N 4 at a width of 350 nm, no further splitting is necessary to lower the error.…”