4
$\begingroup$

I am trying to understand a relation between the two theorems stated in the title. What I observed so far is that since $H^{2}(\mathbb{R},U(1))=\{e\}$, using Bargmann's theorem, we have that projective unitary representations of the additive group $(\mathbb{R},+)$ are in one-to-one correspondence with unitary representations of the additive group $(\mathbb{R},+)$, which will lead to the Schrödinger equation via Stone's theorem.

On the other hand, we have that $H^2(\mathbb{R}^{2n},U(1))=\mathbb{R}$, so the unitary projective representations of $\mathbb{R}^{2n}$ are not in one-to-one correspondence with its unitary representations, rather in one-to-one correspondence with the unitary representations of the Heisenberg group, a central extension of $\mathbb{R}^{2n}$. This is what we get from Bargmann's theorem.

Finally, the Stone-Von-Neumann theorem says that all the irreducible unitary representations of the Heisenberg group are unitarily equivalent to the Schrödinger representation. I stated these theorems without all their technical details, but my question is:

Can the two theorems be composed to obtain the following theorem, which, let's say, we call (Stone-Von-Neumann-Bargmann):

There exists a unique (up to unitary equivalence) strongly continuous projective representation of $\mathbb{R}^{2n}$ for all $n \in \mathbb{N}$ (including zero) on a separable Hilbert space.

$\endgroup$

1 Answer 1

6
$\begingroup$

Consider the strongly-continuous projective unitary representations of $\mathbb{R}^2$ on $L^2(\mathbb{R}, dx)$ given by $$\mathbb{R}^2\ni (a,b)\mapsto U(a,b) := e^{i(\overline{aX - i b \frac{d}{dx}})}$$ and $$\mathbb{R}^2\ni (a,b)\mapsto V(a,b) := e^{i(\overline{aX - i 2b \frac{d}{dx}})}$$ where the bar denotes the closure of the (essentially selfadjoint) operators initially defined on the Schwartz' space. It holds (please check my computations, I am not sure on coefficients of the phases) $$U(a,b)U(a',b')= e^{-i(ab'-ba')} U(a+a',b+b')\:,$$ whereas $$V(a,b)V(a',b')= e^{-i2(ab'-ba')} V(a+a',b+b')\:.\tag{1}$$ En passant, we observe that the projective nature of the representations is evident above in view of the cocycles (the exponential functions) as factors in the right-hand sides.

If a unitary operator existed $K: L^2(\mathbb{R}, dx)\to L^2(\mathbb{R}, dx)$ such that $$KU(a,b)K^{-1}= V(a,b)\tag{2}$$ (if I correctly interpret your notion of unitary equivalence) we would obtain a contradiction. In fact, we have $$KU(a,b)U(a',b')K^{-1}= e^{-i(ab'-ba')} KU(a+a',b+b')K^{-1}\:,$$ that is $$KU(a,b)K^{-1}KU(a',b')K^{-1}= e^{-i(ab'-ba')} KU(a+a',b+b')K^{-1}\:.$$ From (2), $$V(a,b)V(a',b') = e^{-2i(ab'-ba')} V(a+a',b+b')\:.$$ Comparing with (1) and using the fact that all operators are bijective as they are unitary, we would obtain $$e^{-i(ab'-ba')} =e^{-2i(ab'-ba')} \quad \forall a,a',b,b' \in \mathbb{R}$$ that is evidently false.

Another possibility, in the spirit of projective representations, is to make weaker condition (2) into $$KU(a,b)K^{-1}= \chi(a,b) V(a,b)\tag{3}$$ where $\chi(a,b)$ is a unit complex number. That is equivalent to saying that the two representations are in the same class of projective representations. With some computation the condition above leads to $$e^{-i(ab'-ba')} =e^{-2i(ab'-ba')} \chi(a,b)^{-1} \chi(a',b')^{-1} \chi(a+a',b+b')\:.$$ That is $$e^{i(ab'-ba')} = \chi(a,b)^{-1} \chi(a',b')^{-1} \chi(a+a',b+b')\quad \forall a,a',b,b' \in \mathbb{R}$$ The right-hand side is invariant under swapping $(a,b) \leftrightarrow (a',b')$, whereas the left-hand side changes the sign of the exponent under that transformation. As a consequence $$e^{i2(ab'-ba')} = 1 \quad \forall a,a',b,b' \in \mathbb{R}$$ that is impossible as well.

In summary, the answer is NO: there are infinitely many inequivalent (according to (2) or (3)) projective unitary representations of $\mathbb{R}^2$.

The result easily extends to $\mathbb{R}^{2n}$.

Physically speaking, the root of the problem is here the value of the Planck constant, the central charge which fixes the coefficient in front of the exponent $i(ab'-a'b)$. It is an unfixed parameter when dealing with strongly-continuous projective unitary representations of $\mathbb{R}^{2n}$. However it must be fixed as soon as one sees these representations as properly unitary representations of the Heisenberg group.

Actually $\hbar$ also takes place in the Heisenberg group, but usually is taken to be $1$.

As a final comment, I stress that separability is not necessary in the Stone-von Neumann theorem as it is a consequence of a general result: an irreducible strongly continous unitary representation of a (finite-dimensional) Lie group is ncessarily defined in a separable Hilbert space. (see e.g. Proposition 7.36 of this book of mine.)

$\endgroup$
4
  • $\begingroup$ Thank You Valter, you answered my question. As an aside: is there something we can say about strongly continuous unitary projective representations of $R^{2n}$? Using Bargmann's theorem, they are in one-to-one correspondence to strongly continuous unitary representations of which group? $\endgroup$
    – ProphetX
    Sep 18, 2023 at 21:41
  • 1
    $\begingroup$ I do not know, or better, I am not sure. I have to think about it. I am not completely sure about my previous version of this comment. $\endgroup$ Sep 18, 2023 at 21:47
  • 1
    $\begingroup$ Yes, I think that my previous comment, now cancelled, was correct. The group you are looking for is the trivial generalisation of the Heisenberg group in 2n+1 dimensions. $\endgroup$ Sep 18, 2023 at 21:56
  • $\begingroup$ Yes, I think that is it. Thanks! $\endgroup$
    – ProphetX
    Sep 18, 2023 at 22:01

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.