The Stone-von-Neumann theorem states (in rough terms) that a pair of operators $\left(\hat{Q}, \hat{P}\right)$ satisfying the exponentiated canonical commutation relation $e^{is\hat{Q}}e^{it\hat{P}} = e^{-ist}e^{it\hat{P}}e^{is\hat{Q}}$ acting irreducibly are unitarily equivalent to the standard position and momentum operators $\hat{q}, \hat{p}$ acting on $L^2(\mathbf{R})$.

I would like to figure how to extend this result. Precisely, given 2 operators $\hat{A}, \hat{B}$ satisfying the commutation relations $\left[\hat{B}, \hat{A}\right] = if(\hat{A})$ (for some Borel function $f$) acting irreducibly, can one say anything about how $\left(\hat{A}, \hat{B}\right)$ relates to $\left(\hat{q}, \frac{1}{2}\left(f(\hat{q})\hat{p} + \hat{p}f(\hat{q})\right)\right)$ (which indeed satisfy the latter commutation relation)?

Loosely speaking, the Stone-von-Neumann theorem corresponds to $f = 1$ (this is not rigorously true as it only applies to operators satisfying the exponentiated canonical commutation relation, which is stronger than the basic one). To start (hopefully) simple, I would now like to look at $f(x) = x$. I have tried to see whether the proof of the Stone-von-Neumann may be adapted there, following the derivation given in Hall, Quantum Theory for Mathematicians, paragraph 14.3, but this does not seem promising. I anticipate though that this case may still be simple enough as the relation $\left[\hat{A}, \hat{B}\right] = i\hat{A}$ gives you a nice $2$-dimensional Lie algebra structure; this would no longer be the case for general $f$, e.g $f(x) = x^2$.

Any idea how to tackle the problem?

P.S.: Also, do not hesitate to tell if you think the question would the better suited for Math StackExchange.

EDIT: This paper may be relevant to the discussion in the comments.

  • 1
    $\begingroup$ I'm not sure why you assume that two arbitrary operators relate to $q$ and $p$ at all. If the operators in question do not have fully continuous spectrum, then certainly neither of them can be mapped to $q$, dooming your ansatz from the start (this is also part of the reason SvN needs the exponentiated relation). Worse, your 2d Lie algebra example does not have any infinite-dimensional irreducible unitary representations, hence there cannot be a SvN analogue for it. It seems to me that you need much stronger hypotheses here to say anything. $\endgroup$
    – ACuriousMind
    Dec 8, 2019 at 13:38
  • $\begingroup$ Thanks for your comment. Concerning the "2d Lie algebra" example you mentioned, do you mean the one involving $\hat{q}$ and $\frac{1}{2}\left(\hat{q}\hat{p} + \hat{p}\hat{q}\right)$? (In this case, $e^{it\hat{B}} = e^{\frac{it}{2}\left(\hat{q}\hat{p} + \hat{p}\hat{q}\right)}$ is a squeeze operator in the position space). Can you expand (maybe pointing to some references) on why in this case there exists no infinite-dimensional irreducible unitary representation? $\endgroup$ Dec 8, 2019 at 13:50
  • 1
    $\begingroup$ I meant your algebra $[A,B] = \mathrm{i}A$ - there is only one 2d non-Abelian Lie algebra, the affine algebra in 1 dimension. All irreps of this algebra are one-dimensional, cf. math.stackexchange.com/q/791386/143136. I.e. your starting point "Two operators $A,B$ acting irreducibly with $[A,B] = \mathrm{i}f(A)$" is inconsistent for the choice $f(A) = A$ (in fact, it is inconsistent for every function of $A$ whose result is a linear combination of $A$ and $B$, because that's always the same two-dimensional algebra, just with different basis). $\endgroup$
    – ACuriousMind
    Dec 8, 2019 at 13:55
  • $\begingroup$ Thanks for the link! I am still confused though (sorry, my background in representation theory is a bit weak) how this applies to the operators $e^{it\hat{q}}, e^{it\left(\hat{q}\hat{p} + \hat{p}\hat{q}\right)}$ acting on $L^2(R)$. Is it that $L^2(R)$ cannot be decomposed in irreps in this case? $\endgroup$ Dec 8, 2019 at 14:02
  • 1
    $\begingroup$ No, I'm saying that 1. Any irreducible action of $[A,B] = \mathrm{i}A$ is on a one-dimensional space. 2. It therefore cannot be unitarily equivalent to any pair of operators on $L^2(\mathbb{R})$, since that space is not one-dimensional but unitary equivalences preserve cardinality. $\endgroup$
    – ACuriousMind
    Dec 8, 2019 at 14:08

1 Answer 1


The only theorem that could be useful, I think, is a theorem by Nelson proving what follows. Suppose we have a base for a representation of a certain Lie algebra in terms of $n$ densely defined (anti)-symmetric operators $A_j$ on a common dense domain $D$ in a Hilbert space $H$. Suppose also that $\sum_j A_j^2$ is essentially selfadjoint, then the operators $A_j$ are the generators of a unitary strongly-continuous rep of the simply connected Lie group $G$ whose Lie algebra is $g$.

Now, if the representation $U$ is irreducible and you know the classes of irreducible equivalent unitary representations of $G$, then $U$ must be unitarily equivalent to one of them. For the Weyl-Heisenberg group there is only one such representation, but this case is very peculiar. For instance you know that for $SU(2)$ there is an infinite number of those representations.

Regarding the case $[B,A] =iA$ that is the Lie algebra of the triangular subgroup of $SL(2,\mathbb{R})$. If $A^2+B^2$ is essentially selfadjoint and the rep is irreducible, your rep must be unitarily equivalent to one if that group (which are many).


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.