Obviously the rate at and manner in which the MBPT energy series converges toward the full CI energy (the exact solution within a given basis) are intimately connected to the performance of the method. Basically, two questions are (or should be) of concern when using MBPT methods. First, does the series even converge to the full CI answer? If so, how fast?
Handy, Knowles and Somasundram addressed these very issues.[6] It is rather widely accepted that MBPT energies converge quite rapidly for ``well behaved'' sytems which are can be crudely characterized as follows.
is small (i.e. 
is a good approximation to 
)