Hyperinvariant subspace lattice of isometries

Michał ZającCzech 1987

This 1987 paper by Michal Zajac, published in Mathematica Slovaca, explores a fundamental question in operator theory: how "rich" is the double commutant of an isometry, and is it enough to describe all of its hyperinvariant subspaces?

TL;DR

The paper investigates "Property (L)" for isometries on Hilbert spaces—a condition where the lattice of hyperinvariant subspaces is generated by the kernels and ranges of operators in the double commutant. Zajac proves that while pure unitary operators and pure unilateral shifts satisfy this property, their combination often does not, providing a complete characterization of which isometries possess Property (L).

What Problem Does This Paper Address?

In the study of linear operators on a Hilbert space $\mathfrak{H}$, researchers are often interested in invariant subspaces (subspaces $\mathcal{L}$ such that $T\mathcal{L} \subseteq \mathcal{L}$) and hyperinvariant subspaces (subspaces invariant under every operator that commutes with $T$).

A specific question in the field was whether the lattice of hyperinvariant subspaces, denoted $\text{Hyplat}(T)$, is always generated by the most "natural" operators associated with $T$: those in its double commutant $\{T\}''$. If $\text{Hyplat}(T)$ is the smallest complete lattice containing the kernels and ranges of all operators in $\{T\}''$, the operator is said to have Property (L).

While previous work had shown that many "completely non-unitary" (c.n.u.) contractions have this property, it was unknown if all operators—or even all isometries—shared it. Zajac sets out to test this boundary.

Methods

The author employs several classical tools from operator theory:

  • Wold Decomposition: This fundamental theorem allows any isometry $V$ to be decomposed into an orthogonal sum $V = V_0 \oplus V_1$, where $V_0$ is a unitary operator and $V_1$ is a unilateral shift.
  • Functional Models: The paper uses Hardy spaces ($H^2$) and $L^2$ spaces on the unit circle to represent these operators as multiplication operators. This allows the author to translate abstract operator-theoretic questions into concrete problems involving analytic functions.
  • Commutant Analysis: By explicitly calculating the commutant $\{V\}'$ and double commutant $\{V\}''$ for specific combinations of operators (like the bilateral shift $U$ and unilateral shift $S$), the author identifies gaps between the subspaces generated by $\{V\}''$ and the actual hyperinvariant subspaces.

Datasets

Not discussed in the paper. As a theoretical mathematics paper, it relies on formal proofs and logical derivation rather than empirical data.

Key Insights and Ideas

  • The "Mixing" Failure: The most surprising insight is that Property (L) is not preserved under direct sums. A unitary operator $U$ has Property (L), and a unilateral shift $S$ has Property (L). However, their sum $U \oplus S$ does not.
  • The Role of the Double Commutant: For the mixed operator $U \oplus S$, the double commutant is relatively "small"—it consists only of diagonal multiplication operators by $H^\infty$ functions. This set is too restrictive to generate certain hyperinvariant subspaces that exist in the $L^2$ component of the bilateral shift.
  • Geometric vs. Algebraic Subspaces: The author shows that there are "geometric" hyperinvariant subspaces (defined by support sets on the unit circle) that cannot be reached through the "algebraic" means of kernels and ranges of operators in the double commutant.

Key Results

The paper culminates in a definitive characterization of isometries with Property (L):

  • Theorem 4.4: An isometry $V$ has Property (L) if and only if:
  • $V$ is a pure unitary operator, OR
    1. The absolutely continuous unitary part of $V$ is zero (meaning it is a sum of a singular unitary operator and a unilateral shift).
  • Counterexample: The operator $U \oplus S$ (bilateral shift $\oplus$ unilateral shift) is explicitly shown to lack Property (L), serving as a counterexample to the idea that Property (L) might be universal for all contractions.

Limitations and Future Work

The author notes that this result shows that even subnormal operators (a broad class including isometries) do not necessarily satisfy Property (L).

The paper hints at future work (which was later published) proving that all weak contractions do possess Property (L). This suggests a sharp divide in operator theory: while "weak" or "small" operators might have their hyperinvariant lattices fully described by their double commutants, "larger" operators like certain isometries have a more complex subspace structure than their double commutant can capture.

Summary updated

Abstract

Let § be a complex separable Hilbert space and let B(9)) be the algebra of all bounded linear operators on.§. A subspace (i.e. a closed linear manifold) £ c 9) is called invariant for Te /?(£>) if TQ c £. £ is hyperinvariant for T if it is invariant under ea$h AeB($>) that commutes with T. We denote Lat(7) and
BibTeX
@inproceedings{zajac1987hyperinvariant,
  author = {Michał Zając},
  title = {{Hyperinvariant subspace lattice of isometries}},
  booktitle = {Czech},
  year = {1987},
}

Loading related content...

Figure