I'm starting a more 'formal' course on quantum mechanics, and we're going through momentum and position space, which I'm kind of confused on. I would just like clarification on this since I'm not understanding my teacher or my textbook very well. I'm just going to list what I think is going on and would love it if those who understand more could help me out!

  1. There is a 'big, general' state vector $|s(t)\rangle$ that contains all the information of the system. If I want to find the position wavefunction, that is the description of the system in terms of position, $x$ I compute $\langle x|s(t)\rangle$, where $|x\rangle$ is an eigenfunction of the position operator $\hat{x}$ with eigenvalue $x$. How is this exactly computed? I know the eigenfunction must be of the form $\delta(y-x)$ so do I calculate: $$\int_{-\infty}^{\infty}dy\ \delta(y-x)s(t)$$ but this doesn't seem right.
  2. Similarly, if I wanted to find a description of the system in terms of momentum, $p$ I compute $\langle p|s(t)\rangle$. My professor does the following: He sets up the eigenvalue problem $$P|p\rangle=p|p\rangle$$ and then has $$\langle x|P|p\rangle=p\langle x|p\rangle$$ and then claims $$\langle x|P|p\rangle=-i\hbar\partial_x \langle x|p\rangle$$ Did he just pull the momentum operator out of the bra-ket? Why is he allowed to do this? From what I understand, he seems to be claiming $\langle x|P|p\rangle=P \langle x|p\rangle$? He then goes on to show that $$ \langle x|p\rangle=\dfrac{1}{\sqrt{2\pi\hbar}}e^{(ip/\hbar)x}$$ which I follow, but I'm lost on why he is computing $ \langle x|p\rangle$ in the first place. Why couldn't he just compute $|p\rangle$ instead?

2 Answers 2


Did he just pull the momentum operator out of the bra-ket? Why is he allowed to do this? From what I understand, he seems to be claiming $\langle x|P|p\rangle=P \langle x|p\rangle$?

No, that's not what he did. He's using the correct way to interface between the abstract Hilbert-space momentum operator $\hat P$ and its position-representation action as a derivative; this may well be the first time you're exposed to that.

It might help to first hold off from using Dirac notation, and just to consider an arbitrary wavefunction $\psi\in\mathcal H$ in Hilbert space. The momentum operator $\hat P:\mathcal H \to \mathcal H$ takes our wavefunction and gives us a new one, $\phi = \hat P\psi$, whose value is given by the derivative of the old one: $$ \left(\hat P\psi\right)(x) = \phi(x) = -i\hbar \psi'(x). \tag 1 $$

If you now put the Dirac notation back in, the way to connect the abstract state $|\psi⟩$ to its position wavefunction is to query it with the position functional $⟨x|$, so that e.g. $⟨x|\psi⟩=\psi(x)$. With that in mind, then, $(1)$ tells us that the position wavefunction of $\hat P|\psi⟩$ is just the spatial derivative of $⟨x|\psi⟩$, i.e. $$ ⟨x|\hat P|\psi⟩ = -i\hbar\frac{\mathrm d}{\mathrm dx}⟨x|\psi⟩. \tag 2 $$

And now I'm going to do something crazier still. Consider two the following procedures: (i) take a state $|\psi⟩$ act on it with the abstract $\hat P$, and then take its position state wavefunction at $x$; and (ii) take a state $|\psi⟩$, get its position-state wavefunction, and get ($-i\hbar$ times) the derivative of that at $x$. Both of those describe linear functionals $\mathcal H \to \mathbb C$ because they start with states and give us complex numbers, and $(2)$ tells us that they're equal on all their inputs in $\mathcal H$, which means that they're equal as functions: $$ ⟨x|\hat P = -i\hbar\frac{\mathrm d}{\mathrm dx}⟨x|. \tag 3 $$ In other words, you can essentially "cancel out the $|\psi⟩$" from $(2)$ in a rigorous way.

In the specific example of your professor's calculation, he's using the general property $⟨x|\hat P|\psi⟩ = -i\hbar\frac{\mathrm d}{\mathrm dx}⟨x|\psi⟩$ of the momentum operator for the specific case where $|\psi⟩=|p⟩$ is an eigenstate of momentum, $\hat P|p⟩=p|p⟩$, which gets turned by the general property into a differential equation for the position-space wavefunction of $|p⟩$.

I hope that clears up the handling, at least partially.

  • $\begingroup$ Alright, I think I'm getting a picture of what's going on. Can I put it in this more layman's way: we can't really express $\hat{P}|\psi\rangle$ but we understand how this works in the position basis, $\langle x|\hat{P}|\psi\rangle$, which is just the positional derivative of the position wavefunction times a factor. $\endgroup$ Nov 5, 2017 at 1:55
  • $\begingroup$ Yeah, that is correct. $\endgroup$ Nov 5, 2017 at 8:46

This is because $\langle x\vert p\rangle$ is “known”: it is the spatial wave function for a particle with definite momentum, and this must be proportional to the plane wave solution $e^{i p x/\hbar}$ since plane wave have well-defined $k$-vector and thus well-defined $p$. The factor $\sqrt{2\pi\hbar}$ comes from Fourier analysis.


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.