Let $A \in B(H)$ be a bounded Hilbert space operator. For $z \in \mathbb{C}$ exponential is defined as follows:

$$e^{zA} = \sum_{k=0}^{+\infty}\frac{z^kA^k}{k!}$$

Show that series defined above is convergent.

My attempt:


$$S_n = \sum_{k=0}^{n}\frac{z^kA^k}{k!}$$

I will show that $(S_n)$ is Cauchy in $ B(H)$. Without loss of generality assume that $m > n$.

For every $v \in H$ such that $\lVert v \rVert = 1$ we have:

$$\lVert (S_m - S_n)v \rVert = \lVert \sum_{k=n+1}^{m} \frac{z^kA^kv}{k!} \rVert \leq \sum_{k=n+1}^{m}\lVert \frac{z^kA^kv}{k!}\rVert \leq \sum_{k=n+1}^{m} \frac{(\lvert z \rvert \cdot \lVert A \rVert _{op})^k}{k!} < \varepsilon$$

for large $N$ since $e^x$ is uniformly convergent in $\mathbb{R}$, which is also a complete space. Which leads to convergence of $(S_n)$.

Is it correct so far?

The next step is to show that $$(e^{zA})^* = e^{\bar{z}A^*}$$

where $^*$ denotes Hermitian conjugate.

Again, my attempt:

Let $v, w \in H$ then

$$\langle v, (e^{zA})^*w \rangle =\langle e^{zA}v, w \rangle = \langle \sum_{k=0}^{+\infty}\frac{z^kA^kv}{k!}, w \rangle = \sum_{k=0}^{+\infty}\langle \frac{z^kA^kv}{k!}, w \rangle = \sum_{k=0}^{+\infty} \langle v, \frac{(\bar{z}A^*)^k}{k!}w \rangle = \langle v, e^{\bar{z}A^*}w \rangle$$

Which gives us $$(e^{zA})^* = e^{\bar{z}A^*}$$

Where I used a property $(zA)^* = \bar{z}A^*$ and continuity of inner product. Again, is it correct?

  • $\begingroup$ Note that for $S_n-S_m$ there is no need to apply to a particular $v$, you can just estimate directly. $\endgroup$
    – copper.hat
    May 11 '20 at 18:15

Yes, it is correct. An essential part of the argument is that $B(H)$ is complete. Note that the argument works in any Banach algebra.

For the second part you could just use that taking adjoints is continuous (it is actually an isometry: $\|T^*\|=\|T\|$), so $$ \left(e^{zA}\right)^*=\left(\sum_{k=0}^\infty \frac{z^kA^k}{k!}\right)^* =\sum_{k=0}^\infty\left( \frac{z^kA^k}{k!}\right)^*=\sum_{k=0}^\infty \frac{\overline{z}^k(A^*)^k}{k!}=e^{\overline z A^*}. $$

  • $\begingroup$ Thank you, sir. Can I have one more question? The next point of this exercise is to show that $e^{z_1A}\cdot e^{z_2A} = e^{(z_1+z_2)A}$. Should I use Cauchy product rule for finite partial sums and then just take the limit afterwards? $\endgroup$
    – janusz
    May 11 '20 at 18:13
  • 1
    $\begingroup$ You will have to involve the expansion of the binomial somehow. The proof is exactly the same that for the usual exponential. You can use it to show that $e^{A+B}=e^A \,e^B$ provided that $A$ and $B$ commute. $\endgroup$ May 11 '20 at 18:20
  • $\begingroup$ Actually one more question. If I choose they way of showing exactly that $S_n \rightarrow e^{zA}$ in operator norm I will not have to use completeness of B(H), right? $\endgroup$
    – janusz
    May 11 '20 at 19:14
  • 1
    $\begingroup$ Well, yes and no. You are using the completeness of $B(H)$ to say that $e^{zA}$ exists. $\endgroup$ May 11 '20 at 19:25

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.