Let $SU_q(2)$ be the (polynomial) Hopf algebra introduced by Woronocicz called the quantum special unitary group. For details see


(Note that on the Wikipedia page it is the $C^*$-algebra that is discussed, but this question is about the dense Hopf algebra of the $C^*$-algebra.)

Does $SU_q(2)$ contain any (non-unital) invertible elements?

  • $\begingroup$ This Hopf algebra is unital (because all Hopf algebras are) so... the unit is invertible? $\endgroup$ Feb 23, 2021 at 9:53
  • $\begingroup$ @MathewDaws: I edited to exclude the case of the unit. $\endgroup$ Feb 23, 2021 at 13:03

1 Answer 1


Yes, the nonzero multiples of the identity $1$ are the only invertible elements in this algebra. I am sure that someone with more expertise in Hopf algebras than I, can provide a 'high level' proof of the result. The following is a 'low level' direct argument.

For the proof, I am using some notations and basic results from Woronowicz's original paper [W].

For $\nu \in (0,1) \cup (-1,0)$, we are considering the unital $*$-algebra $A$ generated by the elements $a,b$ subject to the relations $$a^*a + b^* b = 1 \;\; , \;\; aa^* + \nu^2 b^* b = 1 \;\; , \;\; bb^* = b^*b \;\; , \;\; ab = \nu ba \;\; , \;\; ab^* = \nu b^* a \; .$$ In [W] it is proven that the elements $a^k b^n (b^*)^m$ and $(a^*)^k b^n (b^*)^m$ form a vector space basis of $A$. Define for $k \geq 0$, the subspace $A(k) \subset A$ as the linear span of $a^k b^n (b^*)^m$ with $n,m \geq 0$. For $k \leq 0$, define $A(k)$ as the linear span of $(a^*)^k b^n (b^*)^m$ with $n,m \geq 0$. One has $A(k_1) A(k_2) \subset A(k_1 + k_2)$ for all $k_1,k_2 \in \mathbb{Z}$.

Another ingredient is the $*$-representation $\pi$ of $A$ on the Hilbert space $H = \ell^2(\mathbb{N} \times \mathbb{Z})$, also defined in [W], and given by $$\pi(a) e_{i,j} = \sqrt{1-\nu^2} e_{i-1,j} \;\; , \;\; \pi(b) e_{i,j} = \nu^n e_{i,j+1} .$$ Write $B = A(0)$, so that $B$ is the unital and abelian $*$-subalgebra of $A$ generated by $b$. We can view $B$ as the algebra of polynomials in two variables. We will use that $B$ has no zero divisors and that the multiples of $1$ are the only invertible elements in $B$.

For every $k \in \mathbb{N}$, denote by $H_k$ the closed linear span of $e_{i,j}$ with $i \geq k$, $j \in \mathbb{Z}$. We use the following observations: for $k \geq 0$, the range of $\pi(a^k)$ is dense in $H$, while the range of $\pi((a^*)^k)$ is dense in $H_k$. Also, every $H_k$ is an invariant subspace for $\pi(B)$ and the resulting representation of $B$ on $H_k$ is faithful for every $k \geq 0$.

Assume that $x,y \in A$ and $xy = 1$. Write $x = \sum_n x_n$ and $y = \sum_n y_n$ with $x_n,y_n \in A(n)$. Let $n_0$, $m_0$ be the largest integers with $x_{n_0} \neq 0$ and $y_{m_0} \neq 0$. Since $xy = 1$, we must have that $n_0 + m_0 \geq 0$. The component of $xy$ in $A(n_0 + m_0)$ is given by $x_{n_0} y_{m_0}$.

We claim that $n_0 + m_0 = 0$. Assume that $n_0 + m_0 > 0$. Then, $x_{n_0} y_{m_0} = 0$. Using the notation $a_k = a^k$ for $k \geq 0$ and $a_k = (a^*)^{-k}$ for $k \leq 0$, we can uniquely write $x_{n_0} = a_{n_0} P$ and $y_{m_0} = Q a_{m_0}$ with $P,Q \in B$. We get that $$\pi(a_{n_0}) \pi(PQ) \pi(a_{m_0}) = 0 .$$ It follows that $\pi(PQ)$ is zero on $H_k$ for $k$ large enough. Hence, $PQ = 0$. This forces $P = Q = 0$, contradicting our choice of $n_0$ and $m_0$. So the claim is proven.

Similarly define the smallest integers $n_1$, $m_1$ such that $x_{n_1} \neq 0$ and $y_{m_1} \neq 0$. The same reasoning leads to $n_1 + m_1 = 0$. It follows that $n_1 = n_0$ and $m_1 = m_0$. Exchanging the roles of $x$ and $y$ if necessary, we may assume that $m_0 \geq 0$ and $n_0 = -m_0$. This means that $x = (a^*)^{m_0} P$ and $y = Q a^{m_0}$ for some $P,Q \in B$. It follows that $$\pi(a^*)^{m_0} \pi(PQ) \pi(a^{m_0}) = 1 .$$ If $m_0 \geq 1$, the left hand side has a nontrivial kernel. Thus, $m_0 = 0$. It then follows that $PQ = 1$, which implies that $P$ and $Q$ are a multiple of $1$. We have proven that $x$ and $y$ are multiples of $1$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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