High Resolution Image Download MS PowerPoint Slide Solid-state nanopores in ultrathin (<20 nm) membranes enable label-free single-molecule sensing, but their adoption as sensors is limited by the lack of scalable manufacturing methods that deliver nanopores with single-nanometer reproducibility. Self-limiting metal-assisted chemical etching (MACE) in silicon-on-insulator (SOI) membranes offers a parallel wet-chemical route for nanopore fabrication, yet prior demonstrations lacked a predictive design rule and required electron microscopy or electrical tests for confirming presence and number of pores. Here, we convert self-limiting MACE into a manufacturing-oriented workflow with optical process control to obtain and verify the formation of 4 nm nanopores in a scalable fashion. We decouple the deposition of 200 ± 10 nm gold (Au) nanoparticles from etching, enabling independent optimization of the two steps. The nanoparticle size allows for particle-per-membrane counting by dark-field optical microscopy, so that deposition can be repeated when counts are below target. We then map etching behavior across Au nanoparticle diameter d (10–200 nm) and silicon (Si) device-layer thickness t (5–18 nm), finding that d / t ≥ 0.8 ratio predicts self-limiting MACE behavior, where pore diameter becomes independent of particle size. In this regime, 200 ± 10 nm catalysts yield 4 ± 1 nm pores, corresponding to a reduction of ∼50× in pore diameter and ∼10× in pore-diameter variability compared to the catalyst diameter and related variability. Successful through-membrane pore formation produces undercuts in the buried oxide (typically ∼200–300 nm diameter) beneath each pore, which can be characterized for each membrane by bright-field microscopy and used as a proxy for the otherwise optically invisible 4 nm pores. Together, the predictive d / t framework and two-stage optical verification establish a scalable wet-chemical route to fabricate nanopores for biomolecular sensing and related nanofluidic devices.
Ferrari et al. (Thu,) studied this question.