Action

Consider

             

Take the partial with respect to ci

       

Interchange the derivative order – usually O.K.

       

Assume that changes in ci do not effect the boundary conditions, that is assume that

On the boundary  and then integrate by parts to find

      

If A is a minimum the partial in is zero or for a complete set of c’s

                                      

This is Laplace’s equation for the potential inside a conductor. 

This shows that if this potential is written as a function of a vector c, minimizing with respect to the constants in c is equivalent to solving Laplace’s equation.

Action èPoisson’s equation

 

The action that gives rise to Poisson’s equation contains an extra term

 

Again set the derivative with respect to ci equal to zero

Integrate the first term in by parts assuming that  to find

      

This is zero for arbitrary values of if and only if

 

Poisson’s equation can be solved with a sufficiently flexible F(x,c) – satisfying the boundary conditions by simply finding the c for which is a minimum.

Action for Electron Density in a 1d system

            ..\Migma\ADEN.DOC .htm     

                          

Charge neutrality requires

                    

Or

      

          The correct F minimizes an action defined by

 

The forms for F are restricted to those which satisfy the boundary conditions.  The derivative of the last term with respect to ci is

       

Or

            

So that with the usual integration by parts

             

..\Migma\MIGMA1D.htm

Energy

          The ground state energy  is always less than

         

The derivative with respect to c contains a term from the denominator

         

With the usual integration by parts

And in we recognize the Schroedinger equation in the parenthesis that need to be zero.  The integral in is multi-3dimensional but with an intense three d character.

..\WaveFunction\Hartree.htm .doc

Variance

The Schroedinger equation can also be solved by minimizing the error coefficient in the Monte-Carlo estimate of <H>.

This differs from minimizing the action by the presence of w(x)?? and by the fact that the minimum is now zero rather than <H>.  In multi-dimensional situations, using Monte Carlo selection guided by w(x), this becomes

 

In curve fitting the variance is defined as

..\optimization\Extremal.htm