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 Hilbert space of finite dimension $n \ge 2$. Let $\{v_1,...,v_n\}$ be an orthonormal basis of $H$. Consider the tensor product $V \otimes V$ and the projections $e_1$ onto the subspace $\mathbb{C}(v_1+\cdots+v_n) \otimes V$ of $V \otimes V$, and $e_2$ onto the subspace $\oplus_{j=1}^n \mathbb{C}(v_j \otimes v_j)$ of $V \otimes V$.

I'm beginning to play with tensor products, so that I'd like to see why the following is the case. For all $j,k \in \{1,\ldots,n\}$, we have

$$e_1(v_j \otimes v_k) = \frac{1}{n} \sum_{l=1}^n v_l \otimes v_k,$$

$$e_2(v_j \otimes v_k) = \delta_{jk} (v_j \otimes v_k).$$

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

Let $w = \frac{1}{\sqrt{n}} \sum_{k=1}^n v_k$. Then $\{w\}$ is an orthonormal basis for $E := \mathbb{C}(v_1+\cdots+v_n)$, and hence $\{w \otimes v_1, \dotsc, w \otimes v_k\}$ is an orthonormal basis for $E \otimes V$, so that the orthogonal projection onto $E \otimes V$ is $$ P_{E \otimes V} = \sum_{k=1}^n \left|w \otimes v_k \right\rangle \left\langle w \otimes v_k \right|. $$ Since $\left\langle w \mid v_k \right\rangle = \tfrac{1}{\sqrt{n}}$ for each $k$, it follows that $$ P_{E \otimes V} (v_i \otimes v_j) = \sum_{k=1}^n \left|w \otimes v_k \right\rangle \left\langle w \otimes v_k \right|(v_i \otimes v_j)\\ = \sum_{k=1}^n \langle w \mid v_i \rangle \langle v_k \mid v_j \rangle(w \otimes v_k) = \frac{1}{\sqrt{n}} w \otimes v_j \\ = \left(\frac{1}{n} \sum_{l=1}^n v_l \right) \otimes v_j $$ for any $i$ and $j$, as required.

Now, let $F = \oplus_{k=1}^n \mathbb{C}(v_k \otimes v_k)$. Then $F$ admits the orthonormal basis $\{v_k \otimes v_k\}_{k=}^n$, so that the orthogonal projection onto $F$ is $$ P_F = \sum_{k=1}^n \left| v_k \otimes v_k \right\rangle \left\langle v_k \otimes v_k \right|, $$ so that $$ P_F(v_i \otimes v_j) = \sum_{k=1}^n \left| v_k \otimes v_k \right\rangle \left\langle v_k \otimes v_k \right|(v_i \otimes v_j)\\ = \sum_{k=1}^n \left\langle v_k \mid v_i \right\rangle\left\langle v_k \mid v_j \right\rangle v_k \otimes v_k\\ = \sum_{k=1}^n \delta_{ki}\delta_{kj} v_k \otimes v_k\\ = \delta_{ij} v_j \otimes v_j\\ = \delta_{ij} v_i \otimes v_j. $$

Of course, since $\{v_k \otimes v_k\}_{j=1}^n$ is a subset of the orthonormal basis $\{v_i \otimes v_j\}_{i,j=1}^n$, it follows directly that $v_i \otimes v_j \in F := \operatorname{span}\{v_k \otimes v_k\}_{j=1}^n$ if and only if $v_i \otimes v_j \in \{v_k \otimes v_k\}_{j=1}^n$, if and only if $i=j$, whilst $v_i \otimes v_j \in F^\perp = \{v_k \otimes v_k \mid 1 \leq k \leq n\}^\perp$ if and only if $v_i \otimes v_j \notin \{v_k \otimes v_k\}_{j=1}^n$, if and only if $i \neq j$, which directly yields $P_F (v_i \otimes v_j) = \delta_{ij} v_i \otimes v_j$.

share|cite|improve this answer
What do you mean by $\left|w \otimes v_k \right\rangle$ and $\left\langle w \otimes v_k \right|$? – ragrigg Mar 18 '13 at 7:14
Sorry about that, it's notation borrowed from physics (i.e., bra-ket notation) for lack of anything better. If $V$ is an inner product space, then for $v,w \in V$, $\left|v\right\rangle \left\langle w\right| \in B(V)$ is the rank $1$ operator defined by $\left|v\right\rangle\left\langle w\right|(x) := \langle w \mid x \rangle v$. – Branimir Ćaćić Mar 18 '13 at 7:38
Thanks for the clarification. I've added a second projection to my original question. For $e_2$, do you take $\{ \oplus_{l=1}^n (v_l \otimes v_l) \}$ as your basis? The symbol $\oplus$ is confusing me. – ragrigg Mar 18 '13 at 9:18
No. The $\oplus$ in this case is an internal direct sum, since $\{v_j \otimes v_j\}_{j=1}^n$ is an orthonormal set, so that $$ \mathbb{C} v_1 \otimes v_1 + \cdots + \mathbb{C} v_n \otimes v_n = \oplus_{k=1}^n \mathbb{C} v_k \otimes v_k.$$ In particular, $\{v_j \otimes v_j\}_{j=1}^n$ is therefore an orthonormal basis for $\oplus_{k=1}^n \mathbb{C} v_k \otimes v_k$. – Branimir Ćaćić Mar 18 '13 at 9:21
I see. Thanks, again. – ragrigg Mar 18 '13 at 9:29

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.