Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $H$ be a separable Hilbert space. Let $(e_i)_i$ be an orthonormal basis. For any bounded linear map $T$ we write, whenever possible

$$\operatorname{tr} T := \sum_{i}^{\infty} \langle T e_i, e_i \rangle$$

Now let $L(H)$ be the set of bounded operators, $N(H)$ the set of nuclear operators. We want to show

$$i: L(H) \rightarrow N(H)' , \quad T \mapsto ( S \mapsto \operatorname{tr}(TS))$$

is an isometric isomorphism.

By the ideal property of the nuclear operators, it is easy to see that $i$ is well-defined, linear and bounded with $|i| \leq 1$.

Recall that the operators $\langle \cdot, e_i \rangle e_j$ form a basis of $N(H)$. You easily see the operator is injective (if the induced functional is zero, it's preimage must have been zero,too) and surjective (simply build up an operator $T \in L(H)$ which induces a given functional in $N(H)$).

Question: How can I finish the proof with showing that $i$ is an isometry?

share|cite|improve this question
up vote 2 down vote accepted

I would do it as follows:

  1. For $(x,y) \in H \times H$ define an element of $N(H)$ by $x \,\tilde{\otimes}\, y = \langle \cdot , y \rangle x$ and note that $\|x \, \tilde{\otimes}\,y\|_{1} = \|x\|\,\|y\|$.
  2. For a functional $\varphi \in N(H)'$ define a sesquilinear form on $H$ by $B_{\varphi}(x,y) = \varphi(x \,\tilde{\otimes}\, y)$. The map $\varphi \mapsto B_{\varphi}$ is clearly linear and of norm $\leq 1$ since $|B_{\varphi}(x,y)| = |\varphi(x \, \tilde{\otimes}\,y)| \leq \|\varphi\|\,\|x\|\,\|y\|$ by 1.
  3. Use the version of the Riesz representation theorem stating that every bounded sesquilinear form $B$ is of the form $B(x,y) = \langle x,T y \rangle$ for a unique $T$. Moreover, $\|B\| = \|T\|$. In other words, the map $B \mapsto T$ from bounded sesquilinear forms to $L(H)$ is a linear isometry.
  4. Define $j(\varphi)$ by $\varphi \mapsto B_{\varphi} \mapsto T_{\varphi}$ by combining the maps from 2. and 3. Verify that $ji = 1_{L(H)}$ and $ij = 1_{N(H)'}$. Since both $i$ and $j$ have norm $\leq 1$, they must be isometries.
share|cite|improve this answer
+1: I prefer this approach. I somehow didn't notice your answer until I was nearly done writing my answer. We didn't really overlap, but my answer does much less, assuming the rest has been done. – Jonas Meyer Feb 22 '11 at 20:58
@Jonas: Thanks! I think it is good for the OP to have these complementary answers. For once our answers didn't overlap that much... – t.b. Feb 22 '11 at 21:25

I think it helps to think of the commutative analogue of this, which is the fact that $\ell^\infty$ is the dual of $\ell^1$. If $(a_n)_n\in\ell^\infty$, then the corresponding functional is $(x_n)_n\mapsto\sum_n a_nx_n$. To show that the map from $\ell^\infty$ to the dual of $\ell^1$ does not decrease norm, let $\varepsilon\gt0$, take $m$ such that $|a_m|\gt \|(a_n)_n\|_\infty-\varepsilon$, and let $(x_n)_n\in\ell^1$ be such that $x_m=1$ and $x_n=0$ for $n\neq m$. Then $|\sum_n a_nx_n|=|a_m|=|a_m|\|(x_n)_n\|_1 \geq(\|(a_n)_n\|_\infty-\varepsilon)\|(x_n)_n\|_1$.

In the noncommutative case, you have a bounded operator $T$, and given small $\varepsilon$ you can find a unit vector $e\in H$ such that $\|Te\|\geq\|T\|-\varepsilon$, and a unit vector $f\in H$ such that $\langle Te,f\rangle = \|Te\|$. Consider the rank one operator $S=\langle\cdot,f\rangle e$, which has norm $1$ in $N(H)$. If my calculations are correct, the trace of TS is $\langle Te,f\rangle=\|Te\|$, showing that the norm of $i(T)$ is at least $\|T\|-\varepsilon$.

Note that on Hilbert space, nuclear operators are often called "trace class", and that a proof that $L(H)$ is the dual of the trace class operators can be found in standard texts on operator theory.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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