12.1. Introduction

This section describes some of the general characteristics of RMG’s databases.

12.1.1. Group Definitions

The main section in many of RMG’s databases are the ‘group’ definitions. Groups are adjacency lists that describe structures around the reacting atoms. Between the adjacency list’s index number and atom type, a starred number is inserted if the atom is a reacting atom.

Because groups typically do not describe entire molecules, atoms may appear to be lacking full valency. When this occurs, the omitted bonds are allowed to be anything. An example of a primary carbon group from H-Abstraction is shown below. The adjacency list defined on the left matches any of the three drawn structures on the right (the numbers correspond to the index from the adjacency list).


New atom types are also introduced to describe atoms in group definitions. The table below shows all atoms types in RMG.

Atom Type Chemical Element Bonding
R Any No requirements
R!H Any except hydrogen No requirements
H Hydrogen No requirements
C Carbon No requirements
Cs Carbon No double or triple bonds
Cd Carbon Exactly one double bond a non-oxygen, non-sulfur atom
Cdd Carbon Two double bonds to any atoms
Ct Carbon One triple bond
CO Carbon Exactly one double bond to an oxygen atom
CS Carbon Exactly one double bond to an sulfur atom
Cb Carbon Exactly two benzene bonds
Cbf Carbon Three benzene bonds (Fused aromatics)
N Nitrogen No requirements
N1sc Nitrogen Up to one single bond, 3 lone pairs
N1s Nitrogen Up to one single bond, 2 lone pairs
N1d Nitrogen Exactly one double bond, 2 lone pairs
N2s Nitrogen One ot Two single bonds, 2 lone pairs
N3s Nitrogen Up to 3 single bonds, 1 lone pair
N3d Nitrogen Up to 1 single, exactly one double bond, 1 lone pair
N3t Nitrogen Exactly one triple bond, 1 lone pair
N3b Nitrogen Exactly two benzene bonds, 1 lone pair
N5s Nitrogen Up to 5 single bonds, 0 lone pairs
N5d Nitrogen Up to 3 single, exacly one double bond, 0 lone pairs
N5dd Nitrogen Exactly two double bonds, 0 lone pairs
N5t Nitrogen Up to 1 single, exactly one triple bond, 0 lone pairs
N5b Nitrogen Up to 1 single, exactly two benzene bonds, 0 lone
O Oxygen No requirements
Os Oxygen No double bonds
Od Oxygen One double bond
Oa Oxygen No bonds (Oxygen atom)
S Sulfur No requirements
Ss Sulfur No double bond
Sd Sulfur One double bond
Sa Sulfur No bonds (Sulfur atom)

Additionally, groups can also be defined as unions of other groups. For example,:

group=OR{X_H, Xrad_H},

12.1.2. Forbidden Groups

Forbidden groups can be defined to ban structures globally in RMG or to ban pathways in a specific kinetic family.

Globally forbidden structures will ban all reactions containing either reactants or products that are forbidden. These groups are stored in in the file located at RMG-database/input/forbiddenStructures.py.

To ban certain specific pathways in the kinetics families, a forbidden group must be created, like the following group in the intra_H_migration family

    label = "bridged56_1254",
group =
1 *1 C 1 {2,S} {6,S}
2 *4 C 0 {1,S} {3,S} {7,S}
3    C 0 {2,S} {4,S}
4 *2 C 0 {3,S} {5,S} {8,S}
5 *5 C 0 {4,S} {6,S} {7,S}
6    C 0 {1,S} {5,S}
7    C 0 {2,S} {5,S}
8 *3 H 0 {4,S}
    shortDesc = u"""""",
    longDesc =


Forbidden groups should be placed inside the groups.py file located inside the specific kinetics family’s folder RMG-database/input/kinetics/family_name/ alongside normal group entries. The starred atoms in the forbidden group ban the specified reaction recipe from occurring in matched products and reactants.

12.1.3. Hierarchical Trees

Groups are ordered into the nodes of a hierarchical trees which is written at the end of groups.py. The root node of each tree is the most general group with the reacting atoms required for the family. Descending from the root node are more specific groups. Each child node is a subset of the parent node above it.

A simplified example of the trees for H-abstraction is shown below. The indented text shows the syntax in groups.py and a schematic is given underneath.


Individual groups only describe part of the reaction. To describe an entire reaction we need one group from each tree, which we call node templates or simply templates. (C_pri, O_pri_rad), (H2, O_sec_rad), and (X_H, Y_rad) are all valid examples of templates. Templates can be filled in with kinetic parameters from the training set or rules.