This is a question about application of group theory to physics.

The starting point is the group $SU(n)$. I have a representation $R$ of $SU(n)$ that takes values on the unitary group on an infinite dimensional separable Hilbert space $H$. This representation can be written as the direct sum of finite-dimensional irreducible representations. Let me write $R(g) = \oplus_j R_j(g)$ for $g \in SU(n)$. The space $H_j$ of each irreducible representation is finite-dimensional. I denote as $P_j$ the projector on $H_j$. For those who are familiar with the subject, what I have in mind is the representation of $SU(n)$ obtained when applying to a set of $n$ bosonic modes the canonical transformations that are linear and preserve the photon-number operator. In this setting, the subspaces $H_j$ are the subspaces with $j$ photons, with $j=0,1,..,\infty$.

My question is the following:

From the representation $R$ given above we can define another representation: $R \otimes R \, : \, g \to R(g) \otimes R(g)$. What is the commutant (also known as centralizer) of the representation $R \otimes R$?

It is easy to see that the commutant of the representation $R$ is given by the projectors $P_j$'s. It is also easy to check that the following operators belong to the commutant of $R \otimes R$:

  • $P_i \otimes P_j$, for $i,j=0,1,...,\infty$

  • $S_{jj}$

where $S_{jj}$ is the "swap" operator in the subspace $H_j \otimes H_j$. I wonder if there are other operators in the commutant. How can I check it?

Thanks a lot and please accept my apologies if my notation is not very clear.

  • 2
    $\begingroup$ Worth noting that apparently "commutant" is another word for "centralizer," for those who were confused like me. Maybe this is a regional thing? $\endgroup$
    – user10851
    Jan 27 '15 at 0:35

There are necessarily further operators: a von Neumann algebra of operators if, as I am assuming, your representations are made of unitary operators.

All those operators can be constructed out of the orthogonal projectors commuting with the representation as complex linear combinations of them, since a complex von Neumann algebra is the closure of the complex span of its orthogonal projectors and here the closure is not necessary since the dimension is finite.

As a matter of fact you therefore should look for the invariant subspaces of each subrepresentation $R_i \otimes R_j$ in the corresponding product space $\cal{H}_i \otimes {\cal H}_j$ and the span of the orthogonal projectors onto these subspaces gives rise to the full space of operators commuting with $R_i \otimes R_j$. In particular, reversing the procedure, with $S_{jj}$ you can construct the subspace of symmetric and the subspace of antisymmetric tensors in ${\cal H}_j \otimes {\cal H}_j$, the projectors are $(I+S_{jj})/2$ and $(I-S_{jj})/2$, all complex combinations of these operators commute with the representation.

The fact that there are however many subspaces is already evident if looking at the case of $SU(2)$. In this case there is a canonical decomposition of $R_i\otimes R_j$ and $\cal{H}_i \otimes {\cal H}_j$ into orthogonal irreducible representations and the complex span of the corresponding projectors define operators in the commutant of $R_i\otimes R_j$ which actually exhaust this commutant (see below).

The said decomposition of $R_i\otimes R_j$ is nothing but the standard Clebsh-Gordan decomposition. In that case the procedure is relatively easy since the irreducible representations of $SU(2)$ are completely fixed by the eigenvalue of the unique Casimir operator $J^2$ and the overall argument reduces to look for the single eigenvalues of $J^2$ admitted for a product $R^{(j)}\otimes R^{(j')}$ of a pair of irreducible representations of $SU(2)$. Each eigenvalue defines an invariant subspace and the complex linear combinations of the corresponding orthogonal projectors are in the commutant of the representation in ${\cal H}^{(j)}\otimes {\cal H}^{(j)'}$.

Here you find the analogous decomposition for $SU(3)$, for $n>3$... I do not know, sorry :). You could try to have a look at Barut-Raczka's book on representations of groups.

An interesting fact is the following.

If the irreducible representations of a compact group are exactly all common eigenspaces of (mutually commuting) Casimir operators and each set of eigenvalues appears at most once, then the operators which commute with a tensor product of two irreducible representations $R_i \otimes R_j$ (a) are mutually commuting and (b) they are exactly the complex combinations of the orthogonal projectors onto the said common eigenspaces.

In fact: the space $\cal{H}_i \otimes {\cal H}_j$ is direct sum of mutually orthogonal irreducible representations by Peter-Weyl theorem which are eigenspaces of the Casimirs. If an operator $A$ commutes with the product representation $R_i \otimes R_j$ it also commute with its Casimir operators and thus it leaves invariant their common eigenspaces, i.e., the spaces of the irreducible representations that decompose $\cal{H}_i \otimes {\cal H}_j= \sum_k {\cal H}^{(i,j)}_k$. Since every such subspace ${\cal H}^{(i,j)}_k$ is irreducible, Schur's lemma entails that $A$ has the form $a_k I_k$ in every irreducible subspace, where $I_k$ is the identity operator on ${\cal H}^{(i,j)}_k$, and $a_k \in \mathbb C$. In summary the von Neumann algebra of the operators commuting with $R_i\otimes R_j$ is just made of the operators of form $\sum_k a_kI_k$, for some compex numbers $a_k$. This algebra is evidently Abelian.


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.