Figure 1 | Multi-task, large-scale integrated optical vision processor using ultrafast parallel nanofabrication. (IMAGE)
Caption
Figure 1 | Multi-task, large-scale integrated optical vision processor using ultrafast parallel nanofabrication. a, Flowchart of the multi-task integrated ONN processor. b, Illustration of the randomized multi-focus TPL system printing at a rate of 0.267 million neurons per minute for ONN device fabrication. To achieve clearly isolated diffractive pixels with both high spatial resolution and fabrication throughput, we propose a random scanning strategy. In this method, each 10 μm × 10 μm region is patterned with pixels composed of four adjacent voxels arranged in a quadrilateral configuration, spaced 200 nm apart, and exposed simultaneously in a single-shot process. c, Design of the 3D-printed diffractive layer. Each unit is fabricated from photoresist (n =1.520 at 520 nm) on a soda lime glass substrate (n = 1.527 at 520 nm). Five phase gradients with π/5 increments are implemented, facilitated by a height-modulated, untrained design featuring a random and uniform distribution with five distinct levels (heights of 100-500 nm). d, Comparison with previously reported works including both 3D printing, photolithography and EBL (such as neuron number, density and fabrication speed). e, Ultra-low-cost nanoimprinting for mass production. The reusable mold is generated by PDMS soft lithography, and large-scale replication is achieved via UV nanoimprinting with high throughput and high feature fidelity.
Credit
Chaoran Huang et al.
Usage Restrictions
Credit must be given to the creator.
License
CC BY