I am reading lecture notes Deriving the Path Integral by A. Wipf (pdf) trying to better understand how to derive the path integral. Specifically, equation (2.27) is:
$$ K(t,q',q)=\int d\omega_1... d\omega_2\prod_{j=0}^{j=n-1}\langle \omega_{j+1} | e^{-itH/\hbar n} | \omega_j \rangle. $$
What I do not like about the author's proof is that midway in the proof he converts $H$ to a non-relativistic particle in a potential, then show that the prod part is equivalent to an integral over $t$ (equation 2.29) for this specific Hamiltonian. Then, simply 'suggests' that it works for a general action (equation 2.30).
How can I show that
$$ \lim_{n\to \infty} \prod_{j=0}^{j=n-1}\langle \omega_{j+1} | e^{-itH/\hbar n} | \omega_j \rangle \to e^{\int L(t,q',q)dt} $$
Without having to use the special case of a non-relativistic particle in a potential. Can it be done in the general case for any Hamiltonian? One would have to consider $H$ as a matrix, and work out the correspondance leaving $H$ as a matrix.
- As a side node, I am not sure why the author changes the notation from $q$ to $\omega$ at 2.26 to 2.27. Could we had simply kept using $q$?