I know how to calculate the expectation of < $\Psi$|A|$\Psi$ > where the operator A is the eigenfunction of energy, momentum or position, but I'm not sure how to perform this for a pure frequency.

In other words, what is the expectation of frequency?

Indeed, is there an expectation, and is it solved using a fourier transform from position space into frequency space?

EDIT In response to the posts by lurscher and David Zaslavsky below, I think both may be right. Frequency f is a parameter and can be considered an operator.

a) For the plane wave there exists temporal and spatial frequencies which act as parameters. $|\textbf{u}_k\rangle = e^{-i k x }$

b) there is also an operator derived from the Hamiltonian E = h f where
$\hat{H} =- \frac{\hbar }{i}\frac{\partial }{\partial t}$
Rearranging and inverting $i$
$\hat{f} = \frac{H}{\hbar } = i \frac{\partial }{\partial t}$

Intuitively this appears like an operator, it has the right units and it transforms into the frequency domain, however, it still requires that $\hat{f}$ be shown to be Hermitian and is an operator that is its own adjoint or satisfies a Poisson algebra. Also intuitively, we can observe a frequency, by taking a measurement.

  • $\begingroup$ Be careful here, $i{\partial\over \partial t}$ as an operator is identically zero: for example apply it to $e^{-ikx}$. If you are going to say « operator » you have to have clear in your mind on what space does it operate on? Here, you have, correctly, set up the space to be spanned by the |$u_k$> and on that space it is zero. And if you are going to say « measurement », what physical apparatus is going to perform this measurement? $\endgroup$ – joseph f. johnson Dec 31 '11 at 7:27
  • 1
    $\begingroup$ Furthermore, the Fourier transform doesn't go from position space to frequency space, it goes to momentum space. For what you want, you would need to set up an « time space » representation of the wave functions, and take a Fourier transform to some kind of «frequency space», and this cannot be done. The Fourier transform always takes functions on a space $V$ to functions on the linear dual to $V$. Momentum is a vector which is dotted against position, so it is the « linear dual », and time does this to frequency, but not to energy.... $\endgroup$ – joseph f. johnson Dec 31 '11 at 8:38

I'm getting the impression that a good part of this question (and perhaps also this physics.SE question?) arises from a wrong presumption that time and position should be on equal footing in quantum mechanics. They are not. The position $\hat{\bf r}$ is an operator, while time $t$ is a parameter. (Notation: In the following boldface denotes a vector quantity, and a hat denotes an operator quantity.) In fact, Pauli's Theorem shows, under mild assumptions, that time cannot be an operator in quantum mechanics, cf. Ref. (1). See also this and this physics.SE questions.

In the position space Schroedinger representation of the Schroedinger picture, we have a wave function $\Psi({\bf r},t)=\left<{\bf r}|\Psi(t)\right>$, the position operator $\hat{\bf r}$ becomes multiplication with ${\bf r}$, and the momentum operator $\hat{\bf p}= \frac{\hbar}{i}{\bf\nabla}$ is differentiation wrt. position. Both $\hat{\bf r}$ and $\hat{\bf p}$ are Hermitian operators. The Hamiltonian $\hat{H}=H(\hat{\bf r},\hat{\bf p})$ is composed out of the operators $\hat{\bf r}$ and $\hat{\bf p}$ in such a way that it is Hermitian.

It is wrong to claim that the Hamiltonian $\hat{H}$ is $i\hbar \partial_{t}$. (If it was, for starters, the Schroedinger equation $i\hbar \partial_{t}\Psi({\bf r},t)=\hat{ H}\Psi({\bf r},t)$ would become an empty statement without any content.) This should be compared with the fact that (in the position space Schroedinger representation) the momentum operator $\hat{\bf p}= \frac{\hbar}{i}{\bf\nabla}$ is a differentiation wrt. position. In particular, time and position differentiations are, in this sense, not on equal footing.

Finally, the frequency operator may be defined as $\frac{1}{h}\hat{H}=\frac{1}{h}{H}(\hat{\bf r},\hat{\bf p})$, as David Zaslavsky observes in a comment to lurscher's answer. The corresponding expectation value is

$$\left<\Psi(t) \left|\frac{1}{h}\hat{H} \right| \Psi(t)\right> ~=~ \int\! d^{3}r \ \Psi^{*}({\bf r},t)\ \frac{1}{h}\hat{H}\Psi({\bf r},t). $$


(1) Wolfgang Pauli 1933, in Handbuch der Physik (Encyclopedia of Physics) (ed. S. Fluegge), vol. 5, pp.1-168, Springer Verlag, 1958.

  • $\begingroup$ Concerning the first paragraph in the answer, see also this question physics.stackexchange.com/q/17477/2451 $\endgroup$ – Qmechanic Nov 26 '11 at 20:37
  • $\begingroup$ Technically, $\Psi (r,t)$ is not a wave function. Yes, it is a wave, but qua wave function, i.e., as a quantum state, it is really a path in the Hilbert space of state vectors. $\Psi(r,0)$ is a wave function, $\Psi(r,1)$ is another, etc. My technical correction here actually reinforces your main point, the one in your first paragraph. $\endgroup$ – joseph f. johnson Dec 31 '11 at 7:33

the expectation for time frequency is obtained replacing A with the Hamiltonian operator, for space frequency is obtained replacing A with the momentum operator

EDIT from your comments i think i have a better idea of the source of your confusion

first, frequency is not an operator; it is a parameter (a c-number); for instance, plane waves $ | \textbf{u}_k \rangle = e^{-i k x } $ have a single parameter $k$ called spatial-frequency (the x is not really a parameter of the state $ | \textbf{u}_k \rangle$ is just an artifact that appears only when you write the coefficients of this state in the basis of eigenstates of the position operator )

De Broglie had the insight of associating classical particle quantities like energy and momentum to temporal and spatial frequencies of plane waves. This means that $ | \textbf{u}_k \rangle$ represents a physical state of momentum $p = \hbar k$

if you consider the algebraic entity $ | \textbf{u}_k \rangle \langle \textbf{u}_k | $ and you accept the bra-ket convention, if you multiply this operator with an eigenfunction $ | \textbf{u}_j \rangle$ you will basically have zero if $k \neq j$ or $ | \textbf{u}_j \rangle$ if $k=j$

then, a correct way to write the momentum operator (which is the same for any operator, the only thing that will change is the form of the basis function) is like:

$$\textbf{P} = \sum_{k} { \hbar k | \textbf{u}_k \rangle \langle \textbf{u}_k |}$$

ANY state vector can be written in the following form

$$ | \Psi \rangle = \sum_{j} { \psi_{j} | \textbf{u}_k \rangle }$$

I leave you as an exercise to write now the expression $\langle \Psi | \textbf{P} | \Psi \rangle$ by replacing each term with the above sums

hint 1: what you want is to obtain an expression that looks like an average after all! but you need to identify the weight in the average

hint 2: the fact that $\textbf{P}$ is hermitian is equivalent to saying that 1) the scalar products $\langle \textbf{u}_j | \textbf{u}_k \rangle$ will be always zero if $j \neq k$ and 2) that the eigenvalues $\hbar k$ will always be real

  • $\begingroup$ Actually, I think that gives the expectation for the energy and momentum eigenfunctions. I'm trying to understand "what is the expectation for the frequency of a free particle", so for a particle moving under a potential - then what is the expectation for a particular frequency, not what is the expectation of the energy associated with that frequency. Although I think they are related. $\endgroup$ – metzgeer May 18 '11 at 23:28
  • $\begingroup$ it gives the expectation for the eigenfunctions and all linear combinations of them; since all physical states can be written as some linear combination of eigenfunctions, the expectation formula is always the same that you wrote; the only difference is that when the state is already "collapsed" in some eigenfunction of the operator being-averaged, then your average is just a single term $\endgroup$ – lurscher May 19 '11 at 3:00
  • $\begingroup$ Quoting lurscher "the expectation for time frequency is obtained replacing A with the Hamiltonian operator, for space frequency is obtained replacing A with the momentum operator" and quoting the Tribble' s Princeton Guide to Advanced Physics p192 - "If operator A represents a physically obserevable quantity, then its expectation value, defined by < \Psi | A | \Psi > = \text{$<$A$>$} must always be real." So you can see my confusion, I' m trying to find < \Psi | A | \Psi > or the observable for frequency < f > $\endgroup$ – metzgeer May 19 '11 at 5:37
  • $\begingroup$ So, are you saying the Energy operator is the Frequency. I can almost see that in Natural Units, but my head starts involuting when I try. $\endgroup$ – metzgeer May 19 '11 at 5:37
  • $\begingroup$ Yeah, pretty much. Remember that $E = hf$, or $f = E/h$. If you "quantize" that equation you get $\langle f\rangle = \langle H\rangle/h$: the expectation value of a (time) frequency operator is the expectation value of the energy operator a.k.a. Hamiltonian divided by Planck's constant. Essentially, what $\langle f\rangle$ gets you is a weighted average of the individual frequencies of the Fourier components of the wavefunction. $\endgroup$ – David Z May 19 '11 at 6:12

Frequency and energy are synonyms in quantum mechanics, so the expected value of the frequency is equal to the expected value of the energy. This is true for angular frequency in radian units, in units where $\hbar=1$. The Hamiltonian operator is the frequency operator, and states of definite H are definite frequency, with frequency given by the DeBroglie relation.

There is no ambiguity or difficulty related to the formalism of nonrelativistic quantum mechanics distinguishing time from space.

  • $\begingroup$ In non-rel. QM, say the non-rel. Schroedinger eq., energy is positive but frequency can be negative. So they are not synonyms. In relativistic mechanics, $E=\sqrt{p^2+m^2}$ (sometimes one must choose a branch, of course, which can be done so as to allow negative energies) and from this plus the de Broglie relation it follows that $\frac E h = \sqrt{f^2}$ so they are not synonyms. Energy is measured in one unit, and frequency is measured in inverse seconds, so they are not synonyms. If you could say something concrete about how the many-time formalism addresses the OP, that would be useful. $\endgroup$ – joseph f. johnson Dec 31 '11 at 17:04
  • $\begingroup$ @Joseph f. Johnson: What? No. The frequency is the energy, period. They are exact synonyms in units where $\hbar=1$, and it is senseless to adopt any other convention when doing QM. The relativistic example gives $E=fh$, even in your calculation, as always, because energy and frequency are the same. They only have different units because of bad human conventions, like (relativistic) mass and energy, or GR's mass and (half the) Schwartschild radius, or a million other things that are pointless to keep track of. $\endgroup$ – Ron Maimon Jan 1 '12 at 17:35

Formally, $H$ is an observable since it is a Hermition operator. But physically, I don't quite see that we ever really measure energy directly, what we tend to measure is the momentum of a photon. But the same applies to frequency: we don't measure the frequency of a photon directly, we measure its momentum, deduce its energy, and apply Planck's law to get the frequency. Classically, we measure the frequency of a wave by trying to time many different peaks of the wave. But according to QM, this cannot be done since after trying to find even one peak, one has disturbed the photon and so cannot proceed to find the next peak (it no longer exists).

The Hamiltonian really is not an operator on the same footing as all the other observables.

This can be seen another way: if it were, then $t$ would be its canonical conjugate, but we already have had it explained that $t$ is a parameter, not an observable. This has often been seen as a defect in the relativistic invariance of QM, but I don't think it has ever succcessfully been removed (without moving to Quantum Field Theory instead).

So I would say that there isn't really any frequency operator in the sense of a quantum mechanical observable since $H$ plays such a special role apart from all the other observables.

  • $\begingroup$ The issue "time is not an operator, it is an observable" is super-fake, as is obvious in a 4d path integral. You are just choosing to parametrize time with coordinate time, instead of proper time (as is standard in non-relativistic physics). There is not such issue in Feynman Schwinger formalism. $\endgroup$ – Ron Maimon Dec 29 '11 at 17:39
  • $\begingroup$ What I said is that $H$ is an operator but not an observable. Time is neither an operator nor an observable. I admit that saying $H$ is not an observable is a littel bit of an exaggeration, but it is not an observable on the same footing as all the others, as can be seen by looking at quantum measurements. $\endgroup$ – joseph f. johnson Dec 31 '11 at 8:43
  • $\begingroup$ The issue of whether something is "an observable" is simply the question of whether you chose a formalism in which you do a path integral integrating over it, or one in which you don't. There are many formalisms, and in the conventional QM formalism, time is treated differently. But this is a property of the formalism, not of the physics. $\endgroup$ – Ron Maimon Dec 31 '11 at 11:42
  • $\begingroup$ whether that is true or not, you quoted me incorrectly in your first comment. $\endgroup$ – joseph f. johnson Dec 31 '11 at 16:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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