Molecular Nanomanufacturing
Assembling atoms into arbitrary structures has been shown only for single atoms in the lab, not for products.
Open in the interactive tree →Richard Feynman sketched manufacturing atom by atom in 1959; K. Eric Drexler described molecular assemblers in 1986. In 1989 IBM pushed 35 xenon atoms with a scanning tunnelling microscope to spell 'IBM'. Biology shows it can be done (ribosomes build proteins exactly), but technical systems have neither the speed nor the error-free precision for products.
As of October 2026
The building blocks exist: molecular machines (Chemistry Nobel 2016), DNA-origami structures, atom positioning with scanning probes and AI design of proteins (Chemistry Nobel 2024). Production with atomic precision across billions of atoms does not exist; whether and how reactions can be steered mechanically is not settled among chemists.
What is missing
- Reliably positioning and bonding single atoms and molecules at room temperature without errors
- Throughput: scanning probes place atoms one at a time, while billions per second (massively parallel) would be needed
- Design and simulation software for atom-precise reaction pathways
- A path to scale-up: first tools must be able to build themselves and further tools
- Clarifying which reactions can be mechanically controlled at all
Becomes possible once solved
- Materials with tailor-made properties (catalysts, superconductors, lightweight structures)
- Medical nanomachines
- Atom-precise chips beyond today's lithography
- Compact factories that build products from simple raw materials
Open steps
- Fast, error-free atom placement Medium AI leveragePlace and bond single atoms and molecules reliably at room temperature, with error detection and repair.
- Simulating mechanically driven reactions High AI leveragePredict which bond-forming reactions a tool tip can drive at room temperature, and with what error rate.
- Self-assembling nanostructure design High AI leverageDesign proteins and DNA structures that self-assemble into target shapes and carry functional parts, as stepping stones to atomic precision.
- Massively parallel assembly Low AI leverageRun millions of probes or assembly sites in parallel with shared control and error checking to reach useful throughput.
Where AI could help
Medium AI leverage. AI helps with atomic-scale simulation, molecular design and automated probes, but throughput and mechanically guided reactions are unproven physics.
- Automate scanning-probe manipulation with learned controllers to raise speed and cut error rates
- Run machine-learned interatomic potentials to simulate reaction pathways at near-quantum accuracy
- Design proteins and DNA nanostructures that self-assemble into target shapes
- Plan assembly sequences and error correction for atomically precise structures
Shown so far
- In December 2022 (Nature Communications) deep reinforcement learning agents positioned single silver atoms with a scanning tunneling microscope with high precision, a step toward autonomous atom assembly. source
- In October 2024 the Chemistry Nobel Prize went to computational protein design (David Baker) and AI protein-structure prediction (Demis Hassabis, John Jumper). source
Prerequisites
- Quantum Mechanics1925
- Scanning Tunnelling Microscope1981The STM was the first tool to place single atoms, spelling IBM with xenon in 1989
- 3D Printing (Additive Mfg)1984
- AI protein structure prediction2020