3

I am dealing with a problem from an old exam paper which might be simple but it turns out to be difficult for me to show.


Here is the problem: Suppose that $0\leq A\leq B$ for self-adjoint elements $A,B$ in a $C^\ast$-algebra. Then show that if $0\leq A\leq B$ and both $A$ and $B$ commute, then $A^n\leq B^n$ for every positive integer $n$. More generally, show that if there are positive elements $C_j$, $1\leq j\leq k$ with $0\leq A\leq C_1\leq C_2\leq \cdots \leq C_k\leq B$ so that any two neighbors in this list commute, then $A^n\leq B^n$ for every positive integer $n$.


I don't want the solution for this but maybe a sketch of strategy/sketch of proof so I can work the details out by my self. If you feel you would like to share a example of a solution then feel free. I just want to prepare for my exam in January. Thanks in advance.

NabbKitha
  • 541
  • 1
    if two symmetric matrices commute they are simultaneously diagonalizable through an orthogonal base chang – Exodd Dec 08 '22 at 23:09
  • 1
    I would try induction and $$ A^{n+1} = \sqrt{A} A^n \sqrt{A}. $$ Then I would try to show that if $A,B$ commute, that also $\sqrt{A}, B$ commute. If that does not work, I'd still try to do induction, but $A^{n+2} = A A^n A$ (one would then need to consider the case $n=2$ separately). – Severin Schraven Dec 08 '22 at 23:10

2 Answers2

5

Hint: If $A$ and $B$ commute, $$B^n - A^n = (B-A) (B^{n-1} + B^{n-2} A + \ldots + B A^{n-2} + A^{n-1})$$

Robert Israel
  • 470,583
1

Your claim immediately follows from functional calculus.

Consider the unital $C^*$-algebra $\mathscr{C}$ generated (inside a unitisation of the $C^*$-algebra if the $C^*$-algebra is not-unital) by $A,B$. This $C^*$-algebra is commutative, since $A$ and $B$ commute, so by Gelfand theory, $\mathscr{C}\cong C(X)$ for a compact Hausdorff space $X$. Hence, you may asume that $A,B$ are continuous functions on $X$. But then the claim is trivially true.

If you know basic $C^*$-algebra theory, you can prove this without working. After a while, you can do these kinds of arguments in your head. In this case, this approach is overkill, but in more complicated situations this is the way to go.

J. De Ro
  • 22,064
  • 2
  • 20
  • 49