Let $U_s$, $V_t$ be a pair of continuous $n$-parameter groups ($n < \infty$) of unitary operators on a complex Hilbert space $\mathcal{H}$. The Stone-von Neumann Theorem establishes that any such pair forming an irreducible representation of the Weyl relations,

$U_sV_t = e^{is\cdot t}V_tU_s$

is unitarily equivalent to the Schrödinger representation, and hence that all such representations are unitarily equivalent. (Note: the Weyl relations in this context are equivalent to the canonical commutation relations (CCRs) $[Q,P]\psi=i\psi$ for all $\psi$ in the common dense domain of $Q$ and $P$, where $Q$ and $P$ are the generators of $V$ and $U$.)

Question: Is there a known analogue of this result in the context of classical Hamiltonian mechanics?

I don't know of a classical analogue of the Weyl relations. But there is a classical analogue of the CCRs, which is the Poisson bracket $\{q,p\}=1$. So, here's how I imagine a classical analogue of the Stone-von Neumann theorem might look (just a rough attempt, really!).

Let $\mathcal{M}$ be a smooth $2n$-dimensional manifold and $\omega$ a symplectic form on $\mathcal{M}$. Let $\xi = (q,p)$ be any global coordinate system on $\mathcal{M}$, and let $Q:\mathcal{M}\rightarrow\mathbb{R}$ and $P:\mathcal{M}\rightarrow\mathbb{R}$ be the projections onto $q$ and $p$, respectively. Then (conjecture): all such pairs ($Q$, $P)$ satisfying,


where $\{\cdot,\cdot\}$ is the Poisson bracket associated with $(\mathcal{M}, \omega)$, are related by a single canonical transformation.

Does this seem like a reasonable way to formulate the classical analogue? Is the status of this conjecture obvious? Your thoughts are appreciated!

  • 1
    $\begingroup$ Let n=1. What you get is the famous Jacobian conjecture... In classical limit representation correspond to symplectic leaves. R^2n is symplectic - so only 1 symplectic leave with accordance to Stone-vonNeuman... $\endgroup$ Oct 19, 2011 at 13:25

1 Answer 1


Expanding on Chervov's comment: the Jacobian conjecture for two variables conjectures that if a polynomial map $(x,y) \to (X,Y)$ has for its Jacobian $\partial(X,Y)/\partial(x,y)$ a nonzero constant, then this polynomial map has a polynomial inverse. (The chain rule, plus the fact that over ${\mathbb C}$ polynomials have zeros, yields the truth of the converse: if the map has polynomial inverse, then its Jacobian is a nonzero constant.) If $(x,y) = (q,p), (X,Y) = (Q,P)$ then $\{Q, P \}$ is the Jacobian of this transformation. So: an affirmative answer for the Jacobian conjecture would precisely yield a 'yes' to your 'reasonable formulation' in the polynomial category when $n =1$.

Going back to Stone-von-Neumann and the Weyl relations you wrote down suggests adding the hypothesis that the flows of the Hamiltonian vector fields for both $Q$ and $P$ are complete. With that addition, you have a chance of a 'yes' answer when $n=2$, perhaps in the smooth category.

Note: your hypothesis imply that your manifold is ${\mathbb R}^{2n}$.


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.