QUESTION: Is it true that for the algebra of compact operators:

$\mathcal{K}(L^2(\mathbb{R}^m \times \mathbb{R}^n))$ is as a $C^{\ast}$-algebra isomorphic to $\mathcal{K}(L^2(\mathbb{R}^m)) \otimes \mathcal{K}(L^2(\mathbb{R}^n))$?

The latter tensor product is any $C^{\ast}$-tensor product (because the compact operators are nuclear it doesn't matter). On $L^2$ we use the ($\sigma$-finite) Lebesgue measure, but of course the algebra $\mathcal{K}(L^2)$ no longer depends on the measure. Clearly, $L^2(\mathbb{R}^{m} \times \mathbb{R}^n)$ can be identified with $L^2(\mathbb{R}^m) \otimes L^2(\mathbb{R}^n)$ by Fubini's theorem. This makes me think that $L^2$ has a better chance than other Hilbert spaces to make $\mathcal{K}(\mathcal{H}_1 \otimes \mathcal{H}_2) = \mathcal{K}(\mathcal{H}_1) \otimes \mathcal{K}(\mathcal{H}_2)$ hold.

Thanks for your help.

EDIT: Of course $\mathcal{K}(\mathcal{H})$ denotes the compact operators on the Hilbert space $\mathcal{H}$.

  • $\begingroup$ Obvious question: is $\mathcal K$ Kalkin algebra (the quotient of bounded operators by compact ones)? $\endgroup$ – Davide Giraudo Oct 13 '12 at 21:06
  • $\begingroup$ No.. the compact operators. $\endgroup$ – user44568 Oct 13 '12 at 23:27
  • $\begingroup$ @DavideGiraudo it is called "Calkin algebra". $\endgroup$ – user42761 Jan 20 '18 at 19:11


  1. $H$ - some Hilbert space

  2. $H^{cc}$ - complex conjugate Hilbert space, i.e. with multiplication on complex conjugate scalars

  3. $H_1\otimes H_2$ Hilbert tensor product of Hilbert spaces $H_1$ and $H_2$

  4. $\mathcal{F}(H)$ finite rank operators on $H$

  5. $\mathcal{K}(H)$ compact operators on $H$

  6. $\mathcal{B}(H)$ bounded operators on $H$

  7. $x\bigcirc y$ rank one operator on $H$ well defined by $(x\bigcirc y)(z)=\langle z,y\rangle x$ where $x,y,z\in H$

  8. $a\;\dot{\otimes }\;b$ Hilbert tensor product of operators $a\in\mathcal{B}(H_1)$ and $b\in\mathcal{B}(H_2)$ well defined by $(a\;\dot{\otimes }\;b)(x\otimes y)=a(x)\otimes b(y)$


  1. $\mathcal{F}(H)=\operatorname{span}\{ x\bigcirc y:x\in H,\; y\in H\}$

  2. $\mathcal{K}(H)=\operatorname{cl}_{\mathcal{B}(H)}\mathcal{F}(H)$

The proof given below is valid for all Hilbert spaces.

Since $\mathcal{K}(H)$ is a nuclear $C^*$ algebra for any Hilbert space $H$, then we can consider any $C^*$ norm on the algebraic tensor product $\mathcal{K}(H_1)\odot K(H_2)$. We will consider the spatial tensor norm, so $$ \mathcal{K}(H_1)\otimes K(H_2)=\operatorname{cl}_{\mathcal{B(H_1\otimes H_2)}}(\operatorname{span}\{ a\;\dot{\otimes}\; b:a\in\mathcal{K}(H_1),\; b\in\mathcal{K}(H_2)\})\tag{1} $$ where $a\;\dot{\otimes}\; b\in\mathcal{B}(H_1\otimes H_2)$ is well defined by equality $(a\;\dot{\otimes}\; b)(x\otimes y)=a(x)\otimes b(y)$. Denote the closed linear subspace in the right hand side of $(1)$ by $E$.

Lemma 1. $\mathcal{F}(H_1\otimes H_2)\subset E$.

Proof. Since bilinear operator $\bigcirc:H\times H^{cc}\to\mathcal{F}(H)$ is bounded, then $\mathcal{F}(H)=\operatorname{span}\{x\bigcirc y: x,y\in S\}$ for any $S\subset H$ such that $H=\operatorname{cl}_H(\operatorname{span}S)$. For $H=H_1\otimes H_2$ we can take $S=\{x\otimes y:x\in H_1,y\in H_2\}$. Now to prove that $\mathcal{F}(H_1\otimes H_2)\subset E$ it is remains to show that $(x\otimes y)\bigcirc (x'\otimes y')\in E$ for all $x,x'\in H_1$, $y,y'\in H_2$. But this is indeed true because $(x\otimes y)\bigcirc (x'\otimes y')=a'\;\dot{\otimes}\; b'$ for $a'=x\bigcirc x'\in\mathcal{K}(H_1)$ and $b'=y\bigcirc y'\in\mathcal{K}(H_2)$.

Lemma 2. $E\subset \mathcal{K}(H_1\otimes H_2)$.

Proof. Consider $a'=x\bigcirc x'\in\mathcal{F}(H_1)$ and $b'=y\bigcirc y'\in\mathcal{F}(H_2)$ for some $x,x'\in H_1$ and $y,y'\in H_2$. Recall $a'\;\dot{\otimes}\; b'=(x\otimes y)\bigcirc (x'\otimes y')\in\mathcal{F}(H_1\otimes H_2)$. Since $\dot{\otimes}$ is bilinear operator and $\mathcal{F}(H)=\operatorname{span}\{x\bigcirc y:x,y\in H\}$ for any Hilbert space $H$ then $a'\;\dot{\otimes}\; b'\in\mathcal{F}(H_1\otimes H_2)$. In other words $\dot{\otimes}\;(\mathcal{F}(H_1),\mathcal{F}(H_2))\subset \mathcal{F}(H_1\otimes H_2)$. Since the bilinear operator $\dot{\otimes}:\mathcal{B}(H_1)\times \mathcal{B}(H_2)\to\mathcal{B}(H_1\otimes H_2)$ is bounded, then $$ \dot{\otimes}\;(\mathcal{K}(H_1),\mathcal{K}(H_2)) =\dot{\otimes}\;(\operatorname{cl}_{\mathcal{B}(H_1)}\mathcal{F}(H),\operatorname{cl}_{\mathcal{B}(H)}\mathcal{F}(H_2)) \subset\operatorname{cl}_{\mathcal{B}(H_1\otimes H_2)}\left(\dot{\otimes}\;(\mathcal{F}(H_1),\mathcal{F}(H_2))\right) \subset\operatorname{cl}_{\mathcal{B}(H_1\otimes H_2)}\mathcal{F}(H_1\otimes H_2) =\mathcal{K}(H_1\otimes H_2) $$ So $E=\operatorname{cl}_{\mathcal{B}(H_1\otimes H_2)}\dot{\otimes}\;(\mathcal{K}(H_1),\mathcal{K}(H_2))\subset\mathcal{K}(H_1\otimes H_2)$.

Proposition. $E=\mathcal{K}(H_1\otimes H_2)$.

Proof. Since $\mathcal{K}(H)=\operatorname{cl}_{\mathcal{B}(H)}\mathcal{F}(H)$ for any Hilbert space $H$ and $E$ is cloed, then it is enough to show that $\mathcal{F}(H_1\otimes H_2)\subset E\subset\mathcal{K}(H_1\otimes H_2)$. Now the result follows from lemma $1$ and lemma $2$.


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.