Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Product rule of derivative of expectation values

  1. Jun 11, 2013 #1
    Hello, first post here.

    I am preparing for my Introductory Quantum Mechanics course, and in the exam questions, we are asked to use Ehrenfest's theorem to show that
    [tex] \frac{d}{dt}\langle \vec{r}\cdot \vec{p} \rangle = \langle 2T-\vec{r}\cdot \nabla V \rangle [/tex]

    Now, from other results:
    [tex] \frac{d}{dt}\langle \vec{r} \rangle = \frac{1}{m} \langle \vec{p} \rangle [/tex]
    [tex] \frac{d}{dt}\langle \vec{p} \rangle = \langle -\nabla V \rangle [/tex]

    it can be solved using a product rule of differentiation:
    [tex] \frac{d}{dt}\langle \vec{r}\cdot \vec{p} \rangle =\frac{d}{dt}\langle \vec{r} \rangle \cdot \vec{p} + \vec{r} \cdot \frac{d}{dt}\langle \vec{p} \rangle [/tex]
    But I severely doubt this is the right way, or even mathematically sound. So my question is, is the above equation sound, or should I be calculating the commutators with the Hamiltonian, as per Ehrenfest's?
    Last edited: Jun 11, 2013
  2. jcsd
  3. Jun 11, 2013 #2


    User Avatar
    Science Advisor

    That equation's fine, as long as you keep the operators in their original order, which you did.
  4. Jun 11, 2013 #3
    Thanks. Google failed me, so I couldn't find any documentation. Happy to know my instincts weren't all off.
  5. Jun 12, 2013 #4

    Jano L.

    User Avatar
    Gold Member

    I think the formula that follows the above words is not correct. On the left-hand side, there is a number, while on the right-hand side, there is an operator (##\mathbf{p} = -i\hbar \nabla##).

    Your intuition is right, you can derive the virial theorem by using the time-dependent Schroedinger equation and the commutation relations. Here is how:

    \frac{d}{dt}\langle \mathbf r\cdot\mathbf p\rangle = \int \dot \psi^* \mathbf r\cdot\mathbf p \psi + \psi^* \mathbf r\cdot\mathbf p \dot \psi\,dV

    Now use Schroedinger's equation

    \dot \psi = \frac{1}{i\hbar} {H} \psi

    to get rid of time derivatives. The integrand contains the commutator

    [\mathbf r\cdot\mathbf p, H].

    Replace ##H## by ##T+V## and use the commutation relations for position and momentum operator and the relation
    [A, BC] = B[A,C] + [A,B]C.
    After some manipulations, you should get ##\langle 2T - \mathbf r\cdot \nabla V\rangle##.
  6. Jun 12, 2013 #5


    User Avatar
    Science Advisor
    Gold Member
    2017 Award

    You cannot use the product rule so easily, because the expectation value of a product is not the product of the expectation value. However, Ehrenfest's theorem tells you that
    [tex]\frac{\mathrm{d}}{\mathrm{d} t} \langle A \rangle=\left \langle \frac{\mathrm{D} \hat{A}}{\mathrm{D} t} \right \rangle.[/tex]
    For a not explicitly time-dependent operator the covariant time derivative is given by
    [tex]\frac{\mathrm{D} \hat{A}}{\mathrm{D} t}=\frac{1}{\mathrm{i} \hbar} [\hat{A},\hat{H}].[/tex]
    Since, mathematically the commutator is a derivation, here the product rule holds.
    [tex]\frac{\mathrm{D} \hat{\vec{r}} \cdot \hat{\vec{p}}}{\mathrm{D} t} = \frac{\mathrm{D}\hat{\vec{r}}}{\mathrm{D} t} \cdot \hat{\vec{p}} + \hat{\vec{r}} \cdot \frac{\mathrm{D}\hat{\vec{p}}}{\mathrm{D} t}.[/tex]
    Now, for a usual non-relativistic Hamiltonian,
    [tex]\hat{H} = \frac{\hat{\vec{p}}^2}{2m} + V(\hat{\vec{r}}),[/tex]
    you get
    [tex]\frac{\mathrm{D} \hat{\vec{r}}}{\mathrm{D} t}=\frac{\hat{\vec{p}}}{m}, \quad \frac{\mathrm{D} \hat{\vec{p}}}{\mathrm{D} t} = -\vec{\nabla} V(\hat{\vec{r}}).[/tex]
    for the kinetic energy, this gives the virial theorem.

    Note, however, that this makes only sense as a property of physical observables, when you apply this to the self-adjoint operator,
    [tex]\frac{1}{2}(\hat{\vec{r}} \cdot \hat{\vec{p}}+\hat{\vec{p}} \cdot \hat{\vec{r}}).[/tex]
  7. Jun 13, 2013 #6


    User Avatar
    Science Advisor
    Homework Helper

    The problem is mathematically ill-posed, since Ehrenfest's theorem is formulated in terms of self-adjoint operators, while the operator [itex] \mathbb{r}\cdot\mathbb{p} [/itex] is not even symmetric, hence, if the equality is correct, it can't be proven using Ehrenfest's theorem.
    Last edited: Jun 13, 2013
  8. Jun 14, 2013 #7
    Thanks for the answers. I ended up using Ehrenfest's theorem, and straight-forward calculation of the commutator with the Hamiltonian.
    @dextercioby: So you're saying Ehrenfest doesn't even apply here? Is it only true in general for self-adjoint operators? Do you have an example of a case where there is a clear discrepancy between the answer from Ehrenfest and the correct answer?

    EDIT: I see now vanhees mentioned the same at the end of his post. Is this a standard method for making non-hermitian operators self-adjoint?
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook