Editing ChargeCalculator:FAQ (section)

=Why did different EEM parameter set produce different charges?=

EEM, the empirical method used by '''ACC''' to calculate atomic charges, relies on empirical parameters. Many EEM parameter sets have been developed using different reference data and fitting procedures. 

EEM parameters are generally developed based on reference quantum mechanical (QM) calculations. A QM calculation is characterized by the setup of the wave function calculation (theory level, basis set, environment), and the type of observables that will be calculated and interpreted. Most commonly, QM reference data used for the development of EEM parameters consists of atomic charges, which are derived from the observable electron density according to a specific ''charge definition'', meaning a procedure used to partition the molecular electron density, or to deduce the electrostatic contribution of each atom. Because atomic charges are not physical observables and have only a conceptual character, there is no unique ''charge definition'' that is universally accepted. Rather, a score of ''charge definitions'' have been published and are in use, each with their own strengths and weaknesses. 

We denote as ''approach'' any association of a QM calculation setup and ''charge definition''.

{| class="wikitable"
 |-
 |'''Approach: QM Method, Basis Set, Population Analysis'''
 |-
 |''Description:'' association of a QM calculation setup and ''charge definition''. Gives the nature of the reference QM data used during the development of the EEM parameters. The applicability domain of an EEM parameter set is closely related to the applicability domain of the reference QM data. 
 |-
 |''Possible values:'' 
*''QM Method'' - level of theory used to solve Schr&ouml;dinger's equation - HF, B3LYP, etc.
*''Basis Set'' - set of basis functions used to solve Schr&ouml;dinger's equation - 6-31G*, STO-3G, etc.
*''Population Analysis'' - charge definition used after solving Schr&ouml;dinger's equation to partition the molecular electron density, or to deduce the electrostatic contribution of each atom - MPA (Mulliken population analysis), NPA (Natural population analysis), MK (Merz-Kollman scheme for fitting to electrostatic potentials), etc.
 |}

Therefore, the first major factor which causes differences in charges is the ''charge definition'', or rather the inherently different principles used to define the amount of electron density to be assigned to each atom. The setup of the reference QM calculations may also cause differences, but much smaller than those caused by the ''charge definitions''. Note that such differences are to be expected, and can be quantified and analyzed in the '''Compare''' tab of the '''ACC Specifics page'''.

Further, even when two EEM parameter sets were developed using the exact same ''approach'', they can still produce different sets of charges. The first reason is the size and nature of the training set that was used to produce the reference data. The training set, which is generally a set of molecules, actually defines the accessible parameter space. The second reason has to do with the parameter fitting procedure, which may have identified a completely different point in the parameter space even when the same training set was used.

{| class="wikitable"
 |-
 |'''Training Set Size, Data Source'''
 |-
 |''Description:'' Number and type of molecules used to produce reference data during the development of the EEM parameters.
 |}

Last but not least, the differences might be caused by an error in the parameters. If you detect such an error, please let us know, so that we may correct it.