5.2: Time-dependent Perturbation Theory
( \newcommand{\kernel}{\mathrm{null}\,}\)
Consider the Hamiltonian
\hat{H} = \hat{H}_0 + \hat{V} (t) \nonumber
where the time dependent part is small. We can write the time dependent coefficients c_n
c_n (t) = c_n (0) + \Delta c_n (t) \nonumber
Where c_n (0) is the value of c_n at t=0. We substitute in the equation for \dot{c}_m derived above to give
\dot{c}_m(t) = (i\hbar )^{−1} \sum_n [c_n (0) + \Delta c_n (t)]V_{mn} \text{ exp}(i \omega_{mn} t) \nonumber
We can assume that for a perturbation c_n (0) >> \Delta c_n (t), and ignore the second term. This allows us to obtain the coefficients c_m(t) by integrating the first–order differential equation to give:
c_m(t) = (i\hbar )^{−1} \sum_n c_n (0) \int^t_0 V_{mn} \text{ exp}(i \omega_{mn}t) dt \nonumber
In the special case where the system is known to be in an eigenstate of \hat{H}_0, say |k \rangle, at t = 0, then c_k(0) = 1 and all other c_m(0) = 0, m \neq k, giving
c_m(t) = (i\hbar )^{−1} \int^t_0 V_{mk} \text{ exp}(i \omega_{mk}t) dt \nonumber
Thus a system starting in a known eigenstate of the unperturbed system may transform to a different eigenstate through the action of the perturbing potential. Notice that c_m(t) is an integral over time, if we wait a long time, the transition may become more likely.
The probability of finding the system at a later time, t, in the state |m \rangle where m \neq k is given by
p_m(t) = |c_m(t)|^2 \nonumber
Since we have assumed a small perturbation, this result is only reliable if p_m(t) \ll 1. “Small” here applies to both V_{mk} and its integral over time.