Please, help out of the mind trap. In this prominent paper Kadison and Ringrose prove among other things the following

Corollary 8. Each norm-continuous representation of a connected topological group by automorphisms of a $C^*$-algebra has range consisting of $\pi$-inner automorphisms.

By $\pi$-inner these authors mean permanently weakly inner, see the definition on page 4 of the same paper, for instance.

Question: Now suppose $A$ is a commutative $C^*$-algebra, and let $\pi:A\to\mathcal{B}(H)$ be a faithful $\star$-representation. A commutative $A$ has obviously no non-trivial inner automorphisms. The strong/weak closure $\pi(A)''\subset\mathcal{B}(H)$ is also commutative, so $\pi(A)$ cannot have even non-trivial weakly inner automorphisms. By the above corollary this would mean that $A$ has no norm-continuous group of automorphisms.

But let $A=C[0,1]$ for clarity (acting on $L^2[0,1]$), and let $y(x;t)=x^t$ for all $x\in[0,1]$ and $t\in(0,\infty)$. Now $y(.;t)$ is a homeomorphism of $[0,1]$ for every $t$, and therefore $\alpha_t(f)=y_*f=f[x^t]$ is definitely $\|f\|_{sup}$-preserving automorphism of $C[0,1]$, thus $\|\alpha_t\|_{\mathrm{op}}=1$. More generally, if $A\simeq C_0(M)$ by Gelfand-Naimark then every homeomorphism group $\phi_t\in\mathrm{Homeo}(M)$ gives such an automorphism group $\alpha_{\phi_t}$. In fact, the particular $\alpha_t$ above is not even spatial. The map $U:L^2[0,1]\to L^2[0,1]$ which gives $\alpha_t(f)=U_t f U_t^{-1}$ is also given by $U_t[h](x)=h(x^t)$ for all $h\in L^2[0,1]$, and is not unitary, because Lebesgue measure is not invariant under $y(.;t)$. $U_t$ is even unbounded for $t>1$.

So where does the source of confusion lie? Thank you.


I think the source of confusion is in the definition of "norm-continuous representation". Probably you are interpreting this as "point-norm continuous", i.e., for each $x \in A$ the map $t \mapsto \alpha_t(x)$ is continuous from the group into the norm topology on $A$. The correct definition is that the map $t \mapsto \alpha$ is continuous from the group into the norm topology on ${\rm Aut}(A)$.

Indeed, there are no nontrivial norm-continuous representations of connected topological groups on abelian C*-algebras, because the distance between any two automorphisms of an abelian C*-algebra is $2$. Proof: let $\alpha$ and $\beta$ be distinct automorphisms. Multiplying both by $\beta^{-1}$, we can assume $\beta = {\rm id}_A$. If $A = C(X)$ then $\alpha$ arises by composition with a non-identity self-homeomorphism $\phi$ of $X$, so find $p \in X$ such that $\phi(p) \neq p$. Then find $f \in C(X)$ with $\|f\|_\infty = 1$, $f(p) = 1$, and $f(\phi(p)) = -1$. Evaluating on this function yields $\|\alpha - {\rm id}_A\| \geq 2$. (The reverse inequality is easy because $\|\alpha\| = \|{\rm id}_A\| = 1$.)

  • $\begingroup$ Thank you very much. I was confusing contunuity of $t\mapsto\|\alpha_t\|$ which is identically one with continuity of $t\mapsto\alpha _t$. $\endgroup$ – Bedovlat Apr 21 '17 at 14:46
  • $\begingroup$ Maybe I should add that in my opinion, point-norm continuity is by far the more important notion... $\endgroup$ – Nik Weaver Apr 22 '17 at 14:42

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.