When you have a system governed by some Hamiltonian H that starts in some state $\lvert a \rangle$, the way to calculate its time-dependence is to decompose into a linear combination of eigenstates of H, $\lvert a \rangle = \sum_{n}\psi_{n}$ and tack on the time dependence $e^{-iE_{n}t/\hbar}$, such that:

$$\lvert a(t) \rangle = \sum_{n}\psi_{n}e^{-iE_{n}t/\hbar}$$

I would like to prove this. To do I say suppose that at time $t=0$ we have:

$$\Psi(x,0) = \sum_{n}\psi_{n}(x)$$

The Schrödinger equation says:

$$i\hbar \frac{\partial \Psi}{\partial t} = H\Psi \\ \Rightarrow \Psi(x,t) = e^{-iHt/\hbar}\Psi(x,0)$$.

By now inserting the eigenstate decomposition of $\Psi(x,0)$ and using that

$$e^{-iHt/\hbar}\psi_{n} = e^{-iE_{n}t/\hbar}\psi_{n}\quad\text{(since it is an eigenstate)}$$

we arrive at the result. Is this a way to prove it or did I assume something that I am not allowed to. I guess I assumed H is time-independent but are there other problems?


3 Answers 3


Everything in your proof is correct. I will fill in some details for you, in case you were wondering about them.

Consider the operator form of the Schrodinger equation:

\begin{align} \hat{H} \lvert \Psi\rangle =i\hbar\frac{\partial}{\partial t}\lvert \Psi\rangle \end{align}

where I have used the notation $\lvert \Psi \rangle$ to denote $\lvert \Psi(t=0) \rangle$. Now, it is clear that, if the hamiltonian is independant of time, the solution to this equation is that

\begin{align} \lvert \Psi(t)\rangle = e^{-\frac{i}{\hbar}\hat{H}t}\lvert \Psi\rangle \end{align}

where the exponential operator is defined by the taylor series

\begin{align} e^{-\frac{i}{\hbar}\hat{H}t}\equiv\sum_{j=0}^{\infty}\frac{1}{j!}\left(\frac{-i\hat{H}t}{\hbar}\right)^j. \end{align}

Now, we can see that if we insert resolution of unity given by

\begin{align} \hat{\mathbb{1}}=\sum_{n=0}^{\infty}\lvert n\rangle \langle n \rvert \end{align}

,where $\lvert n \rangle$ denotes an energy eigenstate, into the above equation for $\lvert \Psi(t) \rangle$, we obtain that

\begin{align} \lvert \Psi(t)\rangle &= e^{-\frac{i}{\hbar}\hat{H}t}\lvert \Psi\rangle\\ &=\sum_{n=0}^{\infty}e^{-\frac{i}{\hbar}\hat{H}t}\lvert n\rangle\langle n\rvert \Psi\rangle\\ &=\sum_{n=0}^{\infty}e^{-\frac{i}{\hbar}E_nt}\lvert n\rangle\langle n\rvert \Psi\rangle \end{align}

The rigorous reason for why $e^{-\frac{i}{\hbar}\hat{H}t}\rightarrow e^{-\frac{i}{\hbar}E_n t}$ upon the left-action on an energy eigenstate is contained in the taylor series definition of $e^{-\frac{i}{\hbar}\hat{H}t}$. This reduces it to just acting on an energy eigenstate with $\hat{H}^j$, which simply returns $E_n^j$, for each term in the series. In this way, we obtain the familiar taylor expansion of $e^{-\frac{i}{\hbar}E_n t}$. Now, insert the unity resolved in the position basis

\begin{align} \hat{\mathbb{1}}=\int\lvert x\rangle \langle x \rvert\ dx \end{align}

so that

\begin{align} \lvert \Psi(t)\rangle&=\sum_{n=0}^{\infty}e^{-\frac{i}{\hbar}E_nt}\lvert n\rangle\langle n\rvert \Psi\rangle\\ &=\sum_{n=0}^{\infty}e^{-\frac{i}{\hbar}E_nt}\lvert n\rangle \int \langle n\rvert x\rangle \langle x\rvert \Psi\rangle \ dx \end{align}

Now, we define the wave function $\Psi(x)$ to be the state $\lvert \Psi \rangle$ expanded in the position basis. That is, $\Psi(x)\equiv\langle x \lvert \Psi \rangle$. It follows that $\psi_n(x)=\langle x\rvert n \rangle$. Therefore, since we desire $\Psi(x,t)$ we take $\langle x \rvert$ to both sides of the last equation to obtain

\begin{align} \Psi(x,t)=\langle x\rvert \Psi \rangle&=\sum_{n=0}^{\infty}e^{-\frac{i}{\hbar}E_nt}\langle x \lvert n\rangle \int \langle n\rvert x\rangle \langle x\rvert \Psi\rangle \ dx\\ &=\sum_{n=0}^{\infty}e^{-\frac{i}{\hbar}E_nt}\psi_n(x)\int \psi_n^*(x)\Psi(x,0) \ dx. \end{align}

We recognize that this is the familiar expansion of $\Psi(x,t)$ into energy eigenstates

\begin{align} \Psi(x,t)=\sum_{n=0}^{\infty}c_n \psi_n(x) e^{-\frac{i}{\hbar}E_n t} \end{align} where $c_n$ is given by

\begin{align} c_n=\int \psi_n^*(x)\Psi(x,0) \ dx. \end{align}


This is a valid way to prove it, but you will notice that indeed if $\hat H = \hat H(t)$ then the integral is not $\hat H ~ t$ and in fact it is even possible that $\hat H(t + \delta t)$ does not commute with $\hat H(t)$ and therefore you need to be very careful with what this so-called "propagator" $$\hat U(T) = \exp\left[\frac{1}{i\hbar} \int_0^T dt~\hat H(t) \right]$$ even means. If you take $t_n = n T/N$ so $\delta t = T/N$ to approximate this time interval $t:0\to T$ then you want the time-ordered product,$$\hat U(T) = \lim_{N\to\infty} \exp\left[\frac{\delta t}{i\hbar} H(t_{N-1})\right]\exp\left[\frac{\delta t}{i\hbar} H(t_{N-2})\right]\dots\exp\left[\frac{\delta t}{i\hbar} H(t_{0})\right],$$and then once you've defined the propagator that way, you can always state that $$|\Psi(T)\rangle = \hat U(T) ~|\Psi(0)\rangle.$$So yes, you can certainly define this operator $\hat U$ and if the Hamiltonian is time-independent then it does indeed have the limit you've found, $e^{-i~\hat H~t/\hbar}$, and that can indeed be evaluated on $|\psi_n\rangle$ by replacing the $\hat H$ in the exponent by $E_n$. This last fact is just the infinite-dimensional analogue of the statement that the exponential of a diagonal matrix is, $$\exp[\operatorname{diag}(a_1, a_2, \dots)] = \operatorname{diag}(e^{a_1}, e^{a_2}, \dots).$$


Also remember how you solved Schrodinger's equation. You began with an ansatz for the wave function namely that it was separable so $\Psi(x,t)=\psi(x)\phi(t) $. You then broke the PDE into two ODEs for time part you got $ \phi(t) = e^{-iEt} $. Therefore the solution is of the form $ \Psi(x,t)= \psi(x)e^{-iHt} $. So $ \sum_{n} \psi_n e^{-iE_nt} $ follows linear structure of Schrodinger's equation(linear superposition of solutions).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.