The eigenvalues of an effectively determined self-adjoint operator are computable, but the sequence of eigenvalues is not (Q1821332): Difference between revisions

This paper explores the computability structure of the eigenvalues and spectrum for bounded and unbounded linear operators on a Hilbert space. As is well-known, these operators - especially the unbounded ones - are important in theoretical physics. The operators considered in this paper are those which are effectively determined. It is easy to verify that the standard operators of analysis and physics satisfy this condition. We now give our main theorem. It implies, as stated in the title, that the individual eigenvalues are computable, although the sequence of eigenvalues need not be. Theorem. Let \(T: H\to H\) be a (bounded or unbounded) self-adjoint operator on a Hilbert space H, and let T be effectively determined. Then there exists a computable sequence of real numbers \(\{\lambda_ n\}\) and a recursively enumerable set A of natural numbers such that: (i) The set of eigenvalues of T coincides with the set \(\{\lambda_ n:\) \(n\in {\mathbb{N}}-A\}\). In particular, each eigenvalue of T is computable. (ii) Each \(\lambda_ n\in spectrum (T)\), and the spectrum of T coincides with the closure of \(\{\lambda_ n\}.\) (iii) Conversely, every set \(\{\lambda_ n: n\in {\mathbb{N}}-A\}\) as in (i) above occurs as the set of eigenvalues of some effectively determined self-adjoint operator. Likewise, every set which is the closure of \(\{\lambda_ n\}\) as in (ii) above occurs as the spectrum of an effectively determined self-adjoint operator. The paper contains a number of additional results. In particular, an example is given showing that if the operator is not normal, then a computable eigenvalue can occur.