For researchers

The reverse topological approach

Designing a framework by starting from the net you want, rather than from the chemistry you have.

The usual direction, reversed

The ordinary way to learn a framework's topology is to make it, solve its structure, and work out which net it turned out to be. Useful, and backwards for design: you find out what you got after you got it.

The reverse topological approach runs the other way. Choose the net first — because of its pore geometry, its symmetry, its connectivity — then ask what building blocks would realise it, and assemble the structure computationally before anyone touches a flask.

What the machine actually does

  1. Take the net. Vertices with known coordination numbers, edges between them, a cell.
  2. Match building blocks to vertices. A four-coordinated vertex needs a block offering four connectors, and the connectors must point roughly the way the edges do — a square-planar node does not belong on a tetrahedral vertex even though both are four-connected.
  3. Place and orient. Each block is rotated so its connectors line up with the edges meeting at its vertex.
  4. Scale. The abstract net has no length scale. The real one is set by the blocks: the cell is scaled until the connectors meet where bonds should be.
  5. Relax. The result is a plausible structure with strained geometry. A force field — this is where MOF-FF comes in — relaxes it into something you can compute properties from.

Steps 2 and 3 are where hypothetical frameworks fail. Connectivity is easy to match and geometry is not; a block can be four-connected and still be the wrong shape for the vertex.

What comes out, and what it is worth

A hypothetical framework: a structure that is topologically sound and geometrically relaxed, and that nobody has made. That is genuinely useful — you can compute pore volume, surface area, adsorption isotherms, mechanical stability — and it is not evidence that the material is synthesisable. Nothing in the procedure knows about solubility, kinetics, or whether the linker survives the reaction.

Treat a hypothetical framework as a hypothesis with numbers attached.

Why the data model looks the way it does

This procedure explains the shape of the database. Nets are stored separately from frameworks because the same net serves many frameworks. Building blocks are stored with their connectivity and connection geometry because that is what the matching step consumes. Frameworks point at their net. Structures hang off frameworks, because one framework has several geometries — as-generated, relaxed at one level of theory, relaxed at another.

It is also why force fields are first-class here rather than an appendix. Step 5 is not optional, and a generated structure without a force field that covers its chemistry is a structure you cannot relax.

Doing it

The assembly tooling runs from the client rather than from these pages; the Build section says what state that work is in. See also Assign a force field to a structure.