Is the approximate point spectrum simply the union of the essential and point spectra?

160 Views Asked by At

I'm refreshing functional analysis from "Hilbert space operators in quantum physics" by Blank, Exner and Havlíček (Springer, 2008). They define essential spectrum for any closed operator on a Hilbert space as

[consisting] of all $λ ∈ \mathbb{C}$ to which there is a sequence of unit vectors $x_n ∈ D_T$ which has no convergent subsequence and satisfies $(T-λ)x_n → 0$.

I'm not sure where this definition comes from but it suits me well, other definitions I've found are valid for self-adjoint operators and for general bounded operators but either is in one way or another too restrictive. (Of course, for bounded self-adjoint operators all the definitions reduce to exactly the same thing.)

Then there is in the same source the proposition 4.7.13:

Let $A ∈ \cal{L}_{cs}(H)$. A complex number $λ$ belongs to $σ_\text{ess}(A)$ iff at least one of the following conditions is valid:

(i) $λ$ is an eigenvalue of infinite multiplicity.

(ii) The operator $(A-λ)^{-1}$ is unbounded.

On the other hand, we know that

$\sigma_\text{ap}(A) = \left\{ λ ∈ \mathbb{C} \mid (A - λ)^{-1}: Ran(A-λ) → D_A\ \text{does not exist or is not bounded} \right\}$,

which would be almost the same condition, just including also eigenvalues of finite multiplicities. This requires a Hilbert space, no mention is made of the conditions of $\cal{L}_{cs}(H)$, i.e., densely defined, closed, symmetric operators, at least not in here.

I'm happy to accept the restriction to densely defined and closed operators. All I am wondering is why the 4.7.13 is formulated only for symmetrical operators, and whether the claim holds if this condition is removed. I can't see why it should't, but I also checked the original (Czech) version of the book and there (labelled 8.4.2) they do some more magic involving the reduction of the operator $A$ to the orthogonal complement of $Ker(A-λ)$, which is defined in the English version but not really used in the proposition. The Czech version has $A_λ$ in the (ii) part, this may actually be a typo in the translation. Most importantly, if it's not the same, then also the similarity between the conditions for $σ_\text{ess}$ and $σ_\text{ap}$ breaks.

Unfortunately the book in question does not make any mention of the approximate point spectrum, so I can't compare the appropriate definitions and claims in the same language, so to say. Any clarification on the topic, or a different source where these concepts are discussed using an equivalent definition of the essential spectrum, would be most appreciated.