Documentation
How the workbench works
Every page here draws on one chemistry layer: atoms are typed by their lone pairs, hybrids follow the real bond angles, and π electrons are shared by Hückel theory with heteroatom parameters. These notes explain each model, the numbers behind it, and where it stops being trustworthy.
Foundations
The orbital layerThe chemistry layer under every page: lone-pair typing, Coulson hybrids from the real bond angles, Hückel π with heteroatom parameters, Slater exponents, Löwdin densities, protonation and pKa descriptors.Radial functions: an experimentWhich radial function should an orbital have? Slater's rules and hydrogenic functions measured against Hartree–Fock atoms: forecast first, then the test, then what changed in the app.ArchitectureHow the workbench is put together: one Python process, a shared orbital layer, browser-side field evaluation, and what scales with the number of atoms.pKa of every site, tested against IUPAC dataSite pKa values and site–site charge interaction fitted on 2832 molecules of the IUPAC Digitized pKa Dataset: the forecast, the held-out test, what improved and what did not.pKa of every site: the forecastWhat we expected before comparing a single number with the IUPAC pKa data: six forecasts and the acceptance rule, written down first and left unedited.Protein sites and zinc: the experimentZinc's acceptor orbitals from Hartree–Fock, zinc placed where AlphaFold leaves none, and an orbital layer that scales to protein fragments: thirteen forecasts and what the measurements said.Protein sites: the forecastWhat we expected before computing zinc's orbitals, placing a single zinc or timing a 400-atom fragment, written down first and left unedited.Ligands: from atoms to SMILES, the experimentBond orders, hydrogens and stereo of bound ligands from their coordinates, checked against the PDB's dictionary on 104 ligands and 156 held out: seven forecasts, what failed, and haem c.Ligands: the forecastWhat we expected before the first ligand was read from its coordinates, written down first and left unedited.A lab server: the experimentWorker processes, larger sites, batch surveys and a local PDB dictionary for a 22-core lab server: nine forecasts, what held in a 2-core sandbox, and what is measured on the machine.A lab server: the forecastWhat we expected before writing the worker pool, the batches and the local dictionary, written down first and left unedited.
Models
π density field engineThe browser engine behind the π density and pKa maps: Löwdin densities, interference, column density through any plane, electric potential and the payload it consumes.Solution pH and the map windowHow the pKa maps titrate every ionisable site together at a chosen pH, how the fields of all sites are mixed, why orbitals are not averaged, and how the map window adapts to larger molecules.Phase, charge and colourWhy orbital phase colours are not charges, how lone pairs are marked, and the bounded charge-sensitive radial display used by the 3D viewer.Spherical-harmonic orbital surfacesHow the 3D viewer turns hybrid coefficients into true R(r)·Y(θ,φ) isosurfaces with marching tetrahedra, and the radial phase convention it uses.Inductive and mesomeric effectsThe −I/+I and −M/+M language used by the catalogue, with halogens, nitro groups and Fischer esterification as worked examples.Polarity overlayHow the catalogue draws its qualitative polarity layer inside the RDKit SVG and what each colour and arrow means.
Tools
3D orbital viewerHow the 3D viewer combines the RDKit conformer, the MolBlock and the orbital coefficients from /api/model in one 3Dmol.js scene.Mol* and files in the molecule's frameThe Mol* tab of the 3D viewer and the cube, PLY and MolBlock downloads: one coordinate frame, checked in both engines, so π fields and orbital surfaces stay on the atoms in any viewer.Ligands: SMILES from a structureHow /sites reads a bound ligand: the four routes to its SMILES, links to the protein, protonation, the bound fragment with its pocket, the GL1X ligand index and the links to the other pages.Protein sites and GL1XHow /sites finds a motif, cuts out a capped, protonated fragment with its zinc, computes donor and acceptor orbitals, and draws them on the whole protein; the /api/site endpoint and GL1X deep links.3D scene and overlaysWhy the 3Dmol scene owns the atomic coordinates and every orbital surface or π field is a derived overlay that reads them, with the consistency check the viewer shows.Functional-group catalogueWhat the 63 SMARTS rules of the catalogue record, why matches may overlap, and why a match is not a reaction prediction.Chemical name to SMILESHow systematic names are parsed with OPSIN and validated with RDKit, and how stereodescriptors and optical-rotation signs are treated.Peptide sequence ⇄ SMILESHow a one-letter sequence becomes peptide SMILES and how a peptide SMILES is segmented back into standard residues, or rejected.
API
Using the APIThe JSON API in one page: endpoints, the OpenAPI 3.1 description, Swagger UI, the π-model endpoint and bearer authentication for SVG rendering.SVG render APIRender a standalone SVG of the orbital model from any SMILES with a bearer-protected POST request, with curl and Node.js examples.
Project
Guiding rule. Show local chemistry quickly and honestly, and keep five things apart: observed structure, geometric inference, local orbital construction, delocalised π treatment, and any thermodynamic interpretation.