> In A. Zee's excellent QFT in a nutshell, he starts off by describing
> the path integral formulation of quantum mechanics. When he gets to
> the part where it's time to calculate
> <q_{j+1}| exp(-i H \delta t) |q_j>
> he uses the free Hamiltonian H=p^2/2m and then sticks in an integral
> over momentum so the time evolution operator is acting on
> eigenvectors. This lets him pull out a scalar to the left. He does
> the integral and with some appropriate definitions, gets for the
> probability amplitude at time T
> \int Dq exp( i \int_0^T dt 1/2 m (dq(t)/dt)^2 ).
> He then leaves it as an exercise to show that had we used an arbitrary
> Hamiltonian H=p^2/2m + V(x), that we would have had the Lagrangian in
> the exponent there:
> \int Dq exp( i \int_0^T dt [1/2 m (dq(t)/dt)^2 - V(q(t))] ).
> Now momentum doesn't commute with energy, so I can't just follow the
> same process. What do I do next?
> ========== Moderator's note =======================================
> Have a look at
> http://theorie.physik.uni-giessen.de/~hees/publ/lect.pdf
Hendrik's notes contain a great introduction to path integrals <q_{j+1}| A(q)B(p) |p_j> = A(q_{j+1})B(p_j) <q_{j+1}|p_j> .
Hope this helps.