Skip to content

Latest commit

 

History

History
72 lines (53 loc) · 3.2 KB

SpecComponents.md

File metadata and controls

72 lines (53 loc) · 3.2 KB
Raw

Specification Components

The JSON format is a general container, and the current work focuses primarily aimed at developing a standard on top of the JSON (and BSON) format for chemical data.

Purpose

The purpose here is to document the format, provide an open specification, establish what is required or optional, and to provide a living specification as we extend the format. This could reuse some of the previous work done in the CML format for XML, and Chemical JSON and NWChem-JSON, and other JSON/XML formats that have been used by codes.

Input Components

Topology

The closest representation to the real physical nature of the system. In practical terms, for molecular sciences, this is the coordinates (in some form) and the elements/Z-number at that coordinate. For both QM and MM, this is your molecule. This may include bonding information and unit cell, lattice parameters, etc, as well.

This is the foundation upon which you build the model basis of your calculation.

Driver

What are you looking to calculate: energy, trajectory, some property, ...? We do need to think about our calculations, which are often workflows with multiple steps in one run. Some codes, especially those with Python front-ends can be used to set up complex workflows.

Model

The overall mathematical model we are using for our calculation. Another way to think about this is the largest superset that still obtains roughly the same result. For example, Direct and Disk-based Hartree-Fock at different Schwarz thresholds could be the same "method". However, density-fitted, LinK, or Cholesky-based Hartree-Fock should be separate methods.

In QM, this is the Hamiltonian (HF, DFT, …) combined with the overall basis of the calculation. An example in QM would be HF/STO-3G or B3LYP/6-311G**. Custom basis sets can be handled with custom keywords.

DGAS Note: Some pushback in combining the model and basis together.

Parameters:

Various tunable parameters for the calculation. These vary widely, depending on the basis and model chemistry. This may come in three tiers of importance:

  • Ancillary - Specific to a given program, does not effect the results
  • Tuning Parameters - Slightly adjust the method and could change the output value slightly (algorithm tweaks, convergence settings, etc)
  • Elemental Parameters - Parameters that greatly effect the energy (orbital spaces in CAS, ?)

Output Components

  • Repeat of input components
  • Driver return - Return of the requested driver (energy/gradient/etc)
  • Ancillary return - Other properties/values constructed as by products of the computation)
  • Provenance - Code, computer, user information, actual settings used by the code (lots of defaults when not specified, that do change over code versions, and are different between codes).
  • Raw Output - If requested, the canonical domain specific ASCII output
  • Skipped Input Fields - If the input allows pass through of other fields print the skipped ones
  • Timing information
  • Errors - If the computation failed the raised error should go here.