Minutes 9 October 2006

Present: Travis, Karl, Jagir, Shane.

(Audio recording of this meeting not available due to equipment failure.)

Cellml incorporation

• Jagir has designed a class that wraps around code from CellML_DOM_API/CCGS.
• Designed so that the definition of such a class (derived from a base cell model class) can be automatically generated for each cell model with code on top of CCGS code or modification to CCGS code.
• Overhead from instantiating an object for every cell
  • Probably small but need to bear in mind when considering design.
  • Model type is shared (state is separate) because code is shared among objects.
• Operator splitting
  • Jagir's interface provides a method for specifying a current (voltage rate) representing the influence of diffusion and so the cell model can do the updating of V_m at each time step.
  • Travis is thinking of first order operator splitting: first cell integration based only on activity in the cell and then diffusion solve uses the new V_m to update V_m for influence of diffusion. (Diffusion and reaction completely separated.)
  • First implement what is done in cm.
  • Jagir's interface makes either method possible.

Bidomain class

• Need to have separate systems if solving separately.
  • Parabolic system for transmembrane potential.
  • Elliptic system for extracellular potential.
• But want to share information between systems (to save memory) as much as possible.
• May need a bidomain class derived from EquationSystems rather than a bidomain System (unless multiple inheritance is helpful).
• These issues don't come up with monodomain.

Finite volumes or elements

• Jagir is familiar with finite volumes but is being encouraged by everyone else to use finite elements because they are already done.