Question about the structure of the Tensor Product of Hilbert Spaces

37 Views Asked by At

I'm currently trying to understand my error better in my work for problem 3.2.21 from Pederson's Analysis now, which states the following:

"Let $\mathfrak{H},\mathfrak{K}$ be separable Hilbert spaces and form the usual tensor product $\mathfrak{H} \otimes \mathfrak{K}$. Given $S \in \mathcal{B}(\mathfrak{H})$ and $T \in \mathcal{B}(\mathfrak{K})$, prove there exists a unique operator $S \otimes T \in \mathcal{B}(\mathfrak{H} \otimes \mathfrak{K})$ such that: $$(S \otimes T)(e_n \otimes f_m) = \sum_{k=1}^{\infty} \sum_{j=1}^{\infty} \langle Se_n,e_k \rangle_{\mathfrak{H}} \langle Tf_m, f_j \rangle_{\mathfrak{K}} (e_k \otimes f_j)$$ for all $n,m$ and $(e_k)_{k},(f_j)_j$ orthonormal bases of $\mathfrak{H},\mathfrak{K}$ respectively. Moreover prove that $\lVert S \otimes T \rVert = \lVert S \rVert \lVert T \rVert$."

I ended up defining $S \otimes T$ first on $\text{span}\{e_j \otimes f_k \: | \: j,k \in \mathbb{N}\}$ by $(S \otimes T)(e_k \otimes f_j) = S(e_k) \otimes T(f_j)$ and tried to extend linearly. However I needed to first prove that $S \otimes T$ on this span is a bounded operator, which I thought would be simple enough since we could just use: $$\left\lVert (S \otimes T)\left(\sum_{i=1}^n \alpha_{i}(e_{k_i} \otimes f_{j_i}) \right) \right\rVert^2 = \left\lVert \sum_{i=1}^n \alpha_{i}(S \otimes T)(e_{k_i} \otimes f_{j_i}) \right\rVert^2 $$ $$= \left\lVert \sum_{i=1}^n \alpha_{i}(S(e_{k_i}) \otimes T(f_{j_i})) \right\rVert^2 \leq \sum_{i=1}^n |\alpha_{i}|^2 \lVert S(e_{k_i}) \otimes T(f_{j_i}) \rVert^2$$ $$\leq \lVert S \rVert^2 \lVert T \rVert^2 \sum_{i=1}^n |\alpha_{i}|^2 = \lVert S \rVert^2 \lVert T \rVert^2 \left\lVert \sum_{i=1}^n \alpha_i(e_{k_i} \otimes f_{j_i}) \right\rVert^2$$ so $\lVert S \otimes T \rVert \leq \lVert S \rVert \lVert T \rVert$ on $\text{span}\{e_j \otimes f_k \: | \: j,k \in \mathbb{N}\}$ and thus extends to $\overline{\text{span}\{e_j \otimes f_k \: | \: j,k \in \mathbb{N}\}} = \mathfrak{H} \otimes \mathfrak{K}$ with the norm on the completion still bounded above by $\lVert S \rVert \lVert T \rVert$.

However, my professor said this was incorrect, as the elements in the vector space of all formal linear combinations of the form $\sum_{n,m} \alpha_{nm} e_n \otimes f_m$ don't look like $\sum_{i=1}^n \alpha_{i}(e_{k_i} \otimes f_{j_i})$, which confuses me, since I thought the vector space of all such formal linear combinations of this form was just, well, the set of all linear combinations of the tensors $e_k \otimes f_j$, and thus I've only shown boundedness on a smaller subset than the span.

Can someone help me understand whats going on here and what I did wrong? Thank you.