Related Experiment Video
Updated: Apr 17, 2026
10:28
Compact Lens-less Digital Holographic Microscope for MEMS Inspection and Characterization
Published on: July 5, 2016
10.9K
Model-Driven Deep Learning Enables Speckle-Free Holography for 3D Parallel Nanofabrication
Kexuan Liu1, Wenqi Ouyang2,3, Chuxian Chen2
1Department of Precision Instrument, Tsinghua University, Beijing, 100084, China.
Research (Washington, D.C.)
|April 16, 2026
Summary
We developed SMART HoloTPL, a deep learning framework for speckle-free 3D nanofabrication. This method significantly improves uniformity and resolution in holographic light field fabrication using 2-photon lithography.
Area of Science:
- Nanofabrication
- Optical Engineering
- Deep Learning
Background:
- Holographic light fields enable high-throughput 3D nanofabrication.
- Speckle noise limits fabrication uniformity in current hologram coding methods.
Purpose of the Study:
- To develop a deep learning framework for speckle-free hologram generation with high uniformity for 3D nanofabrication.
- To address limitations in fabrication rate and quality.
Main Methods:
- Developed a model-driven deep learning framework (SMART HoloTPL).
- Established a polymerization model using the broadband angular-spectrum method.
- Utilized self-supervised network training for hologram coding strategies.
- Designed tailored neural network architecture and loss functions.
- Built a 2-photon lithography (TPL) fabrication platform.
Main Results:
- Achieved large-scale, speckle-free 3D nanofabrication.
- Demonstrated high uniformity in hologram generation.
- Reached a throughput of 120,000 voxels/s.
- Attained a resolution of 120 nm.
Conclusions:
- SMART HoloTPL effectively overcomes speckle noise in holographic nanofabrication.
- The approach enables high-throughput, high-resolution, and uniform 3D structure fabrication.
- This work advances holographic 3D nanofabrication capabilities.
