If this is a single value (scalar), the energy edges are read from the corresponding ct_edges table in the dbase. To see a description of the codes, look at the file $SSW/hessi/dbase/energy_binning.txt.
If it is a 1D array with M values, it considers each value as the edge of a (contiguous) energy channel (therefore there will be M-1 energy bins).
If it is a 2 x N array, then each 2-element vector is a separate energy channel, and there will be N energy bins.
IM_ENERGY_BINNING is used in conjunction with EB_INDEX. EB_INDEX is the index into the energy bin array (starting at 0) to use. ENERGY_BAND will be set to the energy edges actually used.
The advantage of using IM_ENERGY_BINNING and EB_INDEX instead of ENERGY_BAND is that it will be faster if you are doing multiple energy bins for a time range, since the binned eventlist will only be called once for a given time range.
NOTE: Using ENERGY_BAND still works for setting the energy band. However, once you have set IM_ENERGY_BINNING in a given object, you can no longer use ENERGY_BAND to set the energy edges.