Evidence Grading Legend
Every quantitative and architectural claim below carries one of four grades. This field spans demonstrated experiments to long-range extrapolation; conflating them is the most common failure mode in APM discourse. Grades:
Peer-reviewed, replicated in hardware.
Shown once or at tiny scale; not yet engineered.
DFT/simulation-backed design; no hardware demonstration.
Plausible long-range projection; decades out, contingent.
Hydrogen Depassivation Lithography (HDL)
A STM tip selectively removes single H atoms from a Si(100)-H surface, exposing reactive dangling bonds. Atomic selectivity comes from the tunneling current's extreme distance dependence:
First demonstrated by Lyding's group at Illinois in 1994 (Appl. Phys. Lett. 64, 2010), building on late-1980s STM surface-modification work; atomic-scale pattern fidelity replicated by multiple groups since. Throughput remains ~10³–10⁴ sites/s per tip — this is a patterning tool, not yet manufacturing.
Barrier Lowering vs. Bias
Desorption rate depends on both tip-sample distance and applied voltage. Distance sets how many electrons arrive, bias sets how likely each one is to break the bond. Explore the coupled dependence:
Relative rate = tunnelling current e−2κd × per-electron yield, with the yield rising steeply near the ~6 V electronic-excitation threshold and proceeding by slower vibrational heating below it. Log axis, arbitrary units, positive by construction. The sub-threshold coefficient is chosen for shape rather than fitted to data, so the curve is readable as a trend and not as a number. Real thresholds vary with H–Si bond orientation and tip condition; an intuition pump, not process data.
Isolated Molecular Growth (IM-STM)
At ~5 K under UHV, researchers have used STM tip manipulation to build individual sp-hybridised carbon structures, notably cyclo[18]carbon: a closed ring with polyynic bond alternation rather than an open chain, made by stripping CO from a precursor on bilayer NaCl. Demonstrated one molecule at a time.
- Cryogenic operation required — room-temperature IM-STM is unproven.
- Growth rates measured in minutes per bond, not milliseconds.
- Error correction during growth is an open research problem.
Mechanosynthetic Toolset
Freitas & Merkle's DFT analysis identified a minimal set of radical abstraction/deposition tools (ethynyl, hydrogen donation, carbene insertion, etc.) sufficient in principle to build diamondoid lattices. Key caveat: this is computational chemistry — tool closure has never been demonstrated experimentally, and solvent-free vacuum mechanosynthesis of full workpieces remains theoretical.
dimer tools
donors
deposition
Sources
Grouped by the layer of the stack they speak to, and graded on the same scale used above. Experimental work is separated from design literature throughout, because the gap between the two is the substance of the roadmap argument.
Atomically precise patterning · demonstrated
- Lyding, Shen, Hubacek, Tucker & Abeln (1994). Nanoscale patterning and oxidation of H-passivated Si(100)-2×1 surfaces with an ultrahigh vacuum scanning tunneling microscope. Applied Physics Letters 64, 2010. doi:10.1063/1.111722
The origin of hydrogen depassivation lithography. Everything in the HDL section rests on this. - Randall, Lyding, Schmucker et al. (2009). Atomic precision lithography on Si. J. Vac. Sci. Technol. B 27, 2764. doi:10.1116/1.3237096
- Ballard et al. (2013). Multimode hydrogen depassivation lithography: a method for optimizing atomically precise write times. J. Vac. Sci. Technol. B 31, 06FC01. doi:10.1116/1.4823756
Source of the write-rate and line-width figures. - Fuechsle et al. (2012). A single-atom transistor. Nature Nanotechnology 7, 242. doi:10.1038/nnano.2012.21
The strongest existing demonstration that atomic placement yields a working device. - Roadmap on atomic-scale semiconductor devices (2025). arXiv:2501.04535
Current community view of where the field is, useful as a check on the roadmap section.
Single-molecule synthesis by manipulation · lab-scale
- Kaiser, Scriven, Schulz, Gawel, Gross & Anderson (2019). An sp-hybridized molecular carbon allotrope, cyclo[18]carbon. Science 365, 1299. doi:10.1126/science.aay1914
The cyclocarbon result the IM-STM section refers to. C18 made by eliminating CO from a precursor with atom manipulation, on bilayer NaCl at 5 K. The conditions matter as much as the result. - Sun et al. (2020). Synthesis of cyclo[18]carbon via debromination of C18Br6. PMID 32646214
A second, independent route. Replication matters here, since a single tip-manipulation result is easy to over-read. - Chemistry World, for the non-specialist account. New form of pure carbon made by manipulating atoms
Architecture and self-replication · theory
- Merkle (1997). Convergent assembly. Nanotechnology 8, 18–22. zyvex.com/nanotech/convergent.html
The hierarchical size-spanning argument the Convergent Assembly section is built on. - Freitas & Merkle (2004). Kinematic Self-Replicating Machines. Landes Bioscience. Free full text at molecularassembler.com/KSRM.htm and the Internet Archive.
The 137-dimensional replicator design space, and the source for the containment argument in the safety section. - Phoenix (2003). Design of a primitive nanofactory. J. Evolution and Technology 13. PDF
The most worked-through convergent-assembly factory design, and explicitly a design study rather than a result.
The case against, and the open questions
- Baum (2003). Nanotechnology: Drexler and Smalley make the case for and against molecular assemblers. Chemical & Engineering News 81(48). cen.acs.org
The fat-fingers and sticky-fingers objections in the principals' own words. Any honest version of this page has to survive them. - Jones (2004). Soft Machines: Nanotechnology and Life. Oxford University Press.
The substantive alternative. Biology as evidence that nanoscale machines work by wet, floppy, stochastic means rather than dry mechanical assembly. - Ecker & Tuszynski (2022). Entropy, symmetry, and the difficulty of self-replication. arXiv:2202.02938
A thermodynamic floor on replication cost, which is the constraint the roadmap should be read against. - Jones, Haufe, Sells et al. (2011). RepRap: the replicating rapid prototyper. Robotica 29, 177. doi:10.1017/S026357471000069X
Partial self-replication achieved in hardware, at macro scale and from commodity parts. The only working precedent for the Seed property.