Regularity results for classes of Hilbert C*-modules with respect to special bounded modular functionals

This paper investigates the conditions under which the zero functional on a submodule $M$ of a Hilbert C*-module $N$ admits a unique bounded $A$-linear extension, proving that such uniqueness holds for W*-algebras, monotone complete C*-algebras, and compact C*-algebras, while establishing a new perspective linking the existence of non-trivial separating functionals to the presence of bounded non-adjointable operators with non-biorthogonally closed kernels.

The Gist

The Big Picture: The "Ghost" Problem

Imagine you are a detective trying to solve a mystery in a vast, complex city called Hilbert C-Module Land*.

In this city, there are buildings (mathematical structures called modules) and rules for measuring distance and angles (called inner products). Usually, if you have a small room inside a big building, and you can't find any "empty space" or "ghosts" (orthogonal elements) separating them, you assume the small room fills the big building completely.

The Mystery:
Recently, two mathematicians (Kaad and Skeide) found a weird case where this assumption failed. They found a small room $M$ inside a big building $N$ where:

1. There is no "empty space" between them (the orthogonal complement is zero).
2. But, there is a "ghost" (a special function) that can see the difference between the small room and the big building. It can tell you, "I am looking at the big building, but I see nothing in the small room."

This is like having a security camera that sees the whole warehouse but claims the small office inside it is invisible, even though the office is right there. This breaks the rules of how we thought the city worked.

The Goal of the Paper

Michael Frank (the author) wants to know: "When does this 'ghost' problem happen, and when is it impossible?"

He wants to prove that for certain types of "cities" (specific mathematical algebras), if you have a small room inside a big one with no empty space between them, there are no ghosts. The only way to look at the big room and see "nothing" in the small room is if you are actually looking at nothing at all.

The Three "Safe Zones"

Frank proves that in three specific types of mathematical cities, the "ghost" problem never happens. If you are in these zones, the small room and the big room are effectively the same thing.

1. The W-Algebras (The "Perfectly Organized" Cities):*
   • Analogy: Imagine a city where every building is perfectly structured, like a crystal. There are no loose ends. In these cities, if a small room has no empty space around it, it must be the whole building. You can't have a "ghost" function hiding the difference.
2. The Monotone Complete C-Algebras (The "Infinite Ladder" Cities):*
   • Analogy: Imagine a city where you can always climb higher on a ladder, and the ladder never breaks. Even if the city is huge and complex, the rules of "climbing" (order convergence) are so strict that you can't sneak a ghost in. If the small room fills the space, the big room is just the small room expanded.
3. The Compact C-Algebras (The "Finite Pixel" Cities):*
   • Analogy: Imagine a city made of finite, distinct pixels (like a low-resolution image). Because the pieces are so small and distinct, if a small group of pixels fills the space, there is literally no room for a ghost to hide. The small group is the whole image.

The "Ghost" vs. The "Broken Mirror"

The paper connects this "ghost" problem to another weird phenomenon: Broken Mirrors.

• The Normal World: In standard math, if you have a machine (an operator) that takes things in and spits them out, the "trash can" (the kernel) is usually a perfect, clean shape. You can always find a mirror image of that trash can that fits perfectly back into the machine.
• The Broken World: Kaad and Skeide found machines where the "trash can" is messy and jagged. It doesn't have a clean mirror image.

Frank's Discovery:
He proves a powerful link:

• If a "ghost" function exists (a function that sees the big room but ignores the small one), then there must be a "broken mirror" machine (an operator with a messy trash can).
• Conversely, if you are in one of the "Safe Zones" (W*-algebras, etc.), you can never have a broken mirror. Therefore, you can never have a ghost.

Why This Matters

For a long time, mathematicians assumed that Hilbert C*-modules behaved a lot like standard Hilbert spaces (the math used in quantum physics). They thought, "If there's no empty space between two things, they are the same."

This paper says: "Hold on. That's not always true."

• The Bad News: In some weird, complex mathematical cities, the old rules break. You can have things that look the same but act differently.
• The Good News: Frank identified the specific "Safe Zones" where the old rules still work. If you are working with these specific types of algebras (like the ones used in quantum mechanics or signal processing), you can relax. You don't have to worry about ghosts or broken mirrors. The math is stable and predictable.

A Correction to the Past

The paper also mentions a famous old proof (Lemma 2.4 from a 2002 paper) that everyone thought was correct. Frank shows that this proof was actually wrong for some weird cities, but correct for the "Safe Zones" he identified. It's like finding out a map was wrong for a jungle, but perfectly accurate for a city park.

Summary in One Sentence

This paper proves that in the most well-behaved mathematical worlds (W*-algebras, monotone complete algebras, and compact algebras), you cannot have a "ghost" function that ignores a small part of a space while seeing the whole space; if the space is filled, it is truly filled, and the math behaves nicely.

Technical Summary

1. Problem Statement and Motivation

The paper addresses a fundamental problem in the theory of Hilbert C*-modules concerning the uniqueness of extensions of the zero functional.

• The Core Question: Let $A$ be a C*-algebra and $M \subseteq N$ be two Hilbert $A$-modules such that the orthogonal complement of $M$ in $N$ is trivial (i.e., $M^\perp = \{0\}$). Does there exist a non-trivial bounded $A$-linear functional $r_0: N \to A$ that vanishes on $M$?
• Context: In standard Hilbert space theory, if a subspace has a trivial orthogonal complement, it is dense, and the zero functional has a unique extension (the zero functional). However, Hilbert C*-modules exhibit "pathological" behaviors due to the lack of self-duality and the non-commutative nature of the coefficient algebra.
• The Trigger: The author investigates the deeper reasons behind a remarkable counterexample provided by J. Kaad and M. Skeide (2023), where such non-trivial extensions do exist for certain pairs of modules. The paper aims to identify classes of C*-algebras and modules where this pathology does not occur, restoring the "regularity" found in Hilbert spaces.
• Related Problem: The existence of such a functional is tightly linked to the existence of bounded $A$-linear operators $T_0: N \to N$ whose kernels are not biorthogonally closed (i.e., $\text{Ker}(T_0) \neq \text{Ker}(T_0)^{\perp\perp}$).

2. Methodology

The author employs a combination of structural analysis, duality theory, and operator algebra techniques, focusing on specific "well-behaved" classes of C*-algebras:

1. Self-Duality and Order Convergence: The paper utilizes the concept of self-dual Hilbert C-modules. For monotone complete C-algebras (including W*-algebras), the author employs $\tau^o_2$-convergence (order convergence) to characterize self-duality. This allows for the construction of canonical embeddings of modules into their duals ($M \hookrightarrow M'$).
2. Biorthogonal Complements: A central technique involves analyzing the relationship between a submodule $M$ and its biorthogonal complement $M^{\perp\perp}$ within the dual module. The proof strategy often relies on showing that if $M^\perp = \{0\}$, then $M$ must be "dense" in a way that forces $M' = N'$.
3. Multiplier Algebras and Compact Operators: The paper investigates the C*-algebras of "compact" module operators ($K_A(M)$) and their multiplier algebras ($End^*_A(M)$). It establishes isometric *-isomorphisms between these algebras for submodules with trivial orthogonal complements.
4. Counter-Example Analysis: The author explicitly constructs examples (e.g., using the Dixmier algebra) to show where previous proofs (specifically Lemma 2.4 in Frank, 2002) failed, and then provides corrected proofs for specific algebraic classes.

3. Key Contributions and Results

The paper establishes that the "pathological" behavior (existence of non-trivial zero extensions) is excluded for the following classes of C*-algebras and modules:

A. Monotone Complete C*-Algebras (including W*-algebras)

• Theorem 3.8 (Main Result): If $A$ is a monotone complete C*-algebra and $M \subseteq N$ are Hilbert $A$-modules with $M^\perp = \{0\}$, then the self-dual duals coincide: $M' = N'$.
• Consequence: The zero functional on $M$ has a unique extension to $N$, which is the zero functional. No non-trivial bounded $A$-linear functional vanishing on $M$ exists.
• Corollary 3.10: For any $n \in N \setminus M$, the norm in $N$ equals the norm in the dual $M'$, implying a strong structural rigidity.
• Theorem 4.4: The kernel of any bounded $A$-linear operator between Hilbert modules over a monotone complete C*-algebra is biorthogonally complemented. This corrects and validates a previous claim (Lemma 2.4 in [15]) for this specific class, disproving it for general C*-algebras.

B. Compact C*-Algebras

• Theorem 5.1: If $A$ is a compact C*-algebra (admitting a faithful representation in compact operators), and $M \subseteq N$ with $M^\perp = \{0\}$, then $M = N$.
• Reasoning: Hilbert modules over compact C*-algebras possess orthogonal bases and are direct summands. A submodule with a trivial orthogonal complement must be the whole module. Consequently, the zero extension is unique, and all operator kernels are biorthogonally complemented.

C. One-Sided Maximal Modular Ideals

• Theorem 6.1: For any C*-algebra $A$, if $D$ is a right modular maximal ideal, the zero functional on $D$ (viewed as a submodule of $A$) has a unique extension to $A$ (the zero functional).
• Mechanism: The proof utilizes the correspondence between pure states and modular maximal ideals, analyzing the atomic projections in the enveloping von Neumann algebra ($A^{**}$).

D. Commutative C*-Algebras

• Proposition 6.3: For commutative C*-algebras, if $I \subseteq J$ are essential norm-closed ideals, any bounded $A$-linear functional on $J$ vanishing on $I$ must be zero on $J$.

4. Technical Insights and Corrections

• Correction of Lemma 2.4 (Frank, 2002): The paper provides a rigorous proof that the kernel of a bounded operator is biorthogonally complemented for monotone complete and compact C*-algebras. It explicitly states that this lemma is false for general C*-algebras, citing the Kaad-Skeide counterexample.
• Inner Product Non-Uniqueness: Lemma 3.3 demonstrates that if a non-trivial zero extension $r_0$ existed, the module $N$ would admit two distinct $A$-valued inner products inducing equivalent norms but not related by a unitary or adjointable operator. This highlights how the existence of such functionals destabilizes the geometric structure of the module.
• Duality of Ideals: The paper clarifies the relationship between ideals and their annihilators in AW*-algebras, showing that for order-dense ideals in monotone complete algebras, the left annihilator projection forces the duals to coincide.

5. Significance

• Resolution of a Categorical Separation Problem: The paper resolves the "categorical separation problem" for specific, important classes of Hilbert C*-modules. It clarifies that the "bad" behavior observed by Kaad and Skeide is an artifact of general C*-algebras and does not persist in the more structured settings of W*-algebras, monotone complete algebras, or compact algebras.
• Stability of Operator Theory: By proving that operator kernels are biorthogonally complemented in these classes, the paper restores many "Hilbert space-like" properties (such as the existence of orthogonal projections onto kernels) that are generally lost in Hilbert C*-module theory.
• Foundational Impact: The work forces a re-evaluation of earlier results in the field (specifically regarding Lemma 2.4 of [15]), distinguishing between general C*-theory and the regular theory of monotone complete/compact algebras.
• Future Directions: The paper suggests that the study of non-adjointable module operators (left multiplier operators) is a critical area for further research, as these operators are the primary source of the non-regularity in general cases.

In summary, Michael Frank's paper delineates the boundary between "regular" and "pathological" behavior in Hilbert C*-modules, proving that for monotone complete, compact, and maximal modular ideal contexts, the zero functional extension problem behaves as expected in classical Hilbert space theory, thereby excluding the existence of the counterexamples found in the general case.