We coarse-grain a Hamiltonian system with the function mapping microstates to macrostates . The free energy of a macrostate can be calculated from the restricted partition function.
Here is Planks constant, are the unit constant with units of , and is some the unit of time. They make the equations inside the Dirac delta functions unitless so dimensional analysis still works.
We assume a Hamiltonian in a quadratic form with respect to the momentum.
The velocities can be caluclated as below
where
is a symmetric matrix. Substituting for the velocities we get the below.
The potential energy can be split out of the exponent and since it only depends on position it can be moved out of the momentum integral. And the quadratic part can be switched to use because
Let's assume that the free energy is also quadratic, and try to solve for and .
So we have the below equality.
Substitute in the Fourier definition for the Dirac delta.
Do change of variables on the 's to factor out the . So now has units of
Expand all the products out. Also let .
A product of exponents becomes an exponent of sums.
Combine the exponents.
Swap the order of integration.
The part the doesn't depend on the 's can be moved out.
Do complete the squares on the exponent
And assume is positive definite then do complete the squares
Then substitute it back in to get
The part that doesn't depend on momentum can be moved out.
for an arbitrary multidimensional gaussian and symmetric positive definite matrix .
You can show it works even if the offset is an complex vector.
Now the expression is independent of the fine grain momentum. We need to do complete the squares again.
Rearrange a bit.
Completing the squares yields the below.
Note this requires that is positive definite. Substitute it back in.
Then do the same thing again with the Gaussian.
This almost gives a solution for seperating the free energy, but not quite there yet.
This gives a final, exact, expression of (with being whatever units make the rightmost sqrt dimensionless)
Seperating with additional asumptions
Let . If can be written solely as a function of instead of , so it only depends on the coarse grain positions, then we can move it out of the integral. We can do that because it is practically constant in the integral since any time it differs the Dirac delta function removes it's contributions
Then a simple solution as , and in the exponent (or the transpose since symmetric.
And the momentum contribution can be divided out on both sides.
Then just rearrange and basic algebra.