2
$\begingroup$

Let $H$ be an infinite-dimensional Hilbert space, and $a_1,b_1,\ldots,a_k,b_k$ be bounded selfadjoint operators on $H$. Can the sum $\sum_{i=1}^k[a_i,b_i]$ be a (pure imaginary) multiple of the identity ?

I know a result of Halmos which says that the identity is the sum of two commutators, but they are of the form $[P,P^\dagger]$ so cannot fit the bill.

$\endgroup$
3
  • 1
    $\begingroup$ No. $[a,b]^*=-[a,b]$ for $a,b$ bounded selfadjoint. $\endgroup$ Commented Jun 9, 2021 at 9:38
  • $\begingroup$ @LiviuNicolaescu Yes, the sum will be anti-selfadjoint, so it could be equal to i times the identity. $\endgroup$ Commented Jun 9, 2021 at 11:59
  • $\begingroup$ Oops! You're right. $\endgroup$ Commented Jun 9, 2021 at 14:55

2 Answers 2

2
$\begingroup$

Yes. Indeed, for self-adjoint $a,b$, the commutator $[a,b]$ is equal to $\frac{i}{2}[p,p^*]$ for $p=a+ib$. So the theorem by Halmos that you refer to (every hermitian operator is a sum $[p_1,p_1^*]+[p_2,p_2^*]$ for bounded operators $p_1,p_2$) implies that every anti-hermitian operator $X$ is a sum of $[a_1,,b_1]+[a_2,b_2]$ for self-adjoint operators.

$\endgroup$
1
  • $\begingroup$ Oh, of course ! Thank you ! $\endgroup$ Commented Jun 10, 2021 at 8:17
1
$\begingroup$

I think that in general, it depends on the algebra of the operators.
If for example your operators span some Lie algebra, then no it cannot happen. Since then, the commutators will be primitive elements (with regards to the hopf structure of the universal enveloping algebra), and thus their sum will also be primitive, which a multiple of identity cannot be.

But for an arbitrary set of operators, i do not know.

$\endgroup$
3
  • $\begingroup$ Very interesting ! As it happens, they belong to a Jordan algebra A. In that case A+[A,A] is a Lie algebra. Does your answer still apply ? $\endgroup$ Commented Jun 10, 2021 at 8:21
  • $\begingroup$ @Fabien, i am not sure. How is A+[A,A] a lie algebra ? Is there some reference to see how this works? $\endgroup$ Commented Jun 11, 2021 at 4:57
  • $\begingroup$ The reason is the formula $[[a,b],x]=[L_a,L_b]x$ where $L_a$ is the Jordan multiplication by $a$. To see that $[[a,b],[c,d]]$ is in $[A,A]$, you first use Jacobi's identity then the above formula. $\endgroup$ Commented Jun 11, 2021 at 15:30

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .