Explain the discrepancy between measured and simulated VπL for the CPW-based Design 1 modulator

Determine the physical cause(s) of the larger discrepancy between the measured low-frequency half-wave voltage–length product (VπL) for the hybrid thin-film lithium niobate/silicon nitride Mach–Zehnder modulator Design 1 (which uses coplanar waveguide electrodes with a signal–ground gap G = 6 μm) and the value predicted by electro-optic simulations (measured approximately 4.29 V·cm versus predicted 2.75 V·cm), in contrast to Designs 2 and 3 whose measured values agree with simulations within about 2–3%. Ascertain whether this discrepancy arises from wafer-scale fabrication nonuniformities (layer thickness and feature size variations), differing sensitivity of coplanar waveguide versus slow-wave electrode designs to fabrication tolerances, or other device- or process-level factors.

Background

The paper compares measured low-frequency half-wave voltage–length product (VπL) values for three hybrid thin-film lithium niobate (LN) on silicon nitride (SiN) Mach–Zehnder modulator designs against simulation predictions. Designs 2 and 3 (with slow-wave electrodes and substrate undercut trenches) show good agreement with simulations, whereas Design 1 (with coplanar waveguide electrodes and no substrate trenches) shows a notably larger discrepancy.

The authors suggest potential causes including assumptions in simulated layer thicknesses and feature sizes, site-to-site wafer variations, and differing sensitivity of coplanar waveguide (CPW) versus slow-wave electrode (SWE) designs to fabrication variations. However, they explicitly state they do not have a conclusive explanation for the discrepancy for Design 1, motivating a targeted investigation to reconcile measurement and modeling for this configuration.