ImpCarv Carves 67-Nanometer 3D Metastructures, Enabling 532-Nm Optical AI
Updated
Updated · Nature.com · May 13
ImpCarv Carves 67-Nanometer 3D Metastructures, Enabling 532-Nm Optical AI
4 articles · Updated · Nature.com · May 13
Researchers introduced implosion carving, a hydrogel-based fabrication method that photopatterns internal voids and then shrinks them more than 10-fold to build 3D metastructures for visible-light use.
The process uses two-photon activation of sensitizers to generate reactive oxygen species, cleaving selected regions before staged ion treatment and supercritical drying produce final shrinkage factors up to 13.18 and air-filled voids.
Tests showed lateral trench widths of 67 ± 12 nm, axial step heights of 22 ± 2 nm and a refractive-index contrast of about 0.5 after dehydration, giving precise phase control inside the material.
Using that control, the team built a 60-by-60 micrometer diffractive device with two 120-by-120 neuron arrays, 500-nm lateral neurons and operation at 532 nm visible light.
In MNIST digit tests on 1, 5, 6 and 7, the all-optical classifier sent the strongest signal to the correct output region, pointing to denser 3D nanophotonic computing beyond conventional 2D silicon layouts.
As German firms deploy photonic AI chips, can MIT's shrinking tech catch up in the race to commercialize optical computing?
Could this nanoscale 'carving' breakthrough finally make detecting a single cancer cell in a blood test a clinical reality?
MIT’s Implosion Carving Achieves 2,000x Miniaturization for Next-Gen Optical Circuits and Imaging
Overview
MIT researchers have introduced a groundbreaking technique called 'implosion carving,' published in Science in May 2026, that can shrink three-dimensional photonic devices by up to 2,000 times. This method allows for the precise engineering of complex nanoscale structures, opening new possibilities for advanced technologies. Implosion carving is especially significant for optical computing, enabling the creation of highly integrated optical circuits that use light instead of electrons, which could lead to faster and more efficient data processing. It also enhances advanced imaging by allowing the crafting of minute photonic components, improving resolution and imaging capabilities.