NOTE: About 70% of this document is written with AI assistance. Tables, and bulleted lists with the first word bolded, are all AI. Paragraphs without bolding are mostly human written.
Used Molecular Renderer for structure generation and visualization. Claude Code made the CLI-centric program more feasible for general workflows, reducing the time to write code from days to minutes. Code is not published.
This is part of a wider holistic investigation into mechanosynthesis. The investigation uses GFN2-xTB, GFN-FF, and MM4, sometimes combining them into 3-layer ONIOM. It also uses DFT: ~100 H100-hours ($400) of cloud compute for the data in this document. PBE0/def2-SVP or M06-2X/TZVP was used for most of the DFT work.
Summary:
- New optimal route for generating Sn-SiH3: Sn-H created in the beaker (not Sn-CH3/CH2X), attach an alkene linker to the cage, treat with SiH3• flux in bulk before any STM work. Many complementary tools would be destroyed by the SiH3• flux, bu