The Subgraphs tab contains lists where the possible subgraphs of types in this graph type can be defined. The top half of the tab defines decompositions, whereas the bottom half defines explosions.

The Decompositions group is at the top of the tab. The left-hand list shows all the object types of the modeling language for which a decomposition has been defined. You can add object types to that list from its pop-up menu. When one of the types is selected, the right hand list shows the graph types to which that type can decompose. New decompositions can be added and old ones deleted for the selected type by selecting the appropriate item from the right hand list’s pop-up menu. The Add... menu item presents a list of all available graph types, from which one or several can be selected. Note that selecting a Graph type also allows instances of subtypes of that Graph type to be targets of decomposition links. In some exceptional circumstances it may be desirable to specify the supertype of all graph types, Graph: you can do this by selecting Add... with shift held down, and Graph will then be included in the list of types to select from. There are also menu commands for editing the type selected on the left or its symbol.

All decomposition structures are defined in Graph Tool and they are always dependent both on the object type and graph type. On the instance level, however, things work a little differently: a given instance State always decomposes to the same graph, wherever it is. Thus the type level defines where it is legal to create new decomposition structures: existing decomposition links are part of the object, and can be followed anywhere.

A typical decomposition structure can be found from WatchApplication diagrams, in which a State can decompose into a new WatchApplication model. Note that on the instance level only one decomposition is allowed for an object, and applied in all graphs containing that object. On the type level, each object type in a graph type can have several possible graph types that it could legally decompose to.

The Explosions group works in the same way as the decompositions definition, but the source of the explosion can be an object, relationship or role type. Explosion forms another way to connect elements of a graph to other graphs. The difference from decomposition is that explosion creates a simpler link between an element and a graph whereas decomposition is more often used for a stronger containment. Unlike decomposition, explosion allows users to make links from a design element to several graphs, and different links for the same element used in different graphs, whereas an object can have only one decomposition, which is the same wherever that object is used.

Basically, the explosion structure allows a user to select one instance from a graph and link that to a new graph. An instance can have many explosions in a single graph, and a different set of explosions in each graph where it is used. Similarly, the same graph can be used as the target of multiple instances in multiple graphs. Explosion is often used between graphs of different types. A typical example of explosion can be found from UML, in which a class in a Class Diagram can be exploded into a State Transition Diagram to describe the behavior of the object.