Invariant measures and traces on groupoid $\mathrm{C}^\ast$-algebras

This paper establishes sufficient conditions for the existence and uniqueness of traces on the essential and full $\mathrm{C}^\ast$-algebras of (possibly non-Hausdorff) étale groupoids extending invariant measures, particularly linking uniqueness to essential freeness and amenability of isotropy groups, with applications to gauge-invariant algebras of finite-state self-similar groups.

The Gist

Imagine you are an architect trying to build a massive, complex city called C*-Algebras. This city is built using blueprints from three different sources: dynamics (how things move), combinatorics (how things are arranged), and group theory (how things symmetries work).

For a long time, architects had a perfect way to build this city if the blueprints were "clean" and "orderly" (mathematicians call this Hausdorff). But recently, they started getting blueprints that were messy, overlapping, and chaotic (called non-Hausdorff). These messy blueprints come from things like fractals, self-similar patterns, and strange group actions.

The problem? When you try to build the city with these messy blueprints, the standard construction tools break. The walls don't line up, and the "reduced" version of the city (a simplified model) has holes in it that shouldn't be there.

This paper by Alistair Miller and Eduardo Scarparo is a guidebook for fixing these broken cities. They introduce a new, more robust version of the city called the Essential C*-Algebra (the "Essential City") and answer two big questions:

1. Can we put a "weight" (a Trace) on this city that matches the "population density" (Invariant Measure) of the original blueprint?
2. When is this weight unique?

Here is the breakdown using everyday analogies:

1. The Messy City vs. The Essential City

Imagine you are mapping a city where some streets loop back on themselves in confusing ways, or where two different paths lead to the exact same spot, but the map doesn't show it clearly.

• The Reduced City ($C^*_r$): This is a standard map. In a messy city, this map might have "ghost streets" or "phantom buildings" that exist on the map but don't actually exist in the real city.
• The Essential City ($C^*_{ess}$): The authors say, "Let's tear down the ghost streets." They create a new, cleaner version of the city by removing the parts that cause the confusion. This is the Essential C*-Algebra. It's the "true" version of the city, stripped of the mathematical noise.

2. The "Trace" and the "Measure"

• The Measure (The Population): Imagine you have a map of the city's center (the "unit space"). You have a way of counting how many people live in each neighborhood. This is your Invariant Measure. It's a rule that says, "If you move a neighborhood to a new spot, the number of people stays the same."
• The Trace (The City's Weight): A Trace is like a way of weighing the entire city based on that population count. It's a function that tells you the "size" or "value" of any building in the city, respecting the symmetry of the layout.

The Big Problem: When you take your population count from the messy blueprint and try to weigh the "Essential City," the math sometimes breaks. The weight might not fit because the "ghost streets" you removed were carrying some of the weight.

3. The Solution: When Does the Weight Fit?

The authors provide a checklist (Theorem A) to see if your population count can successfully become a weight for the Essential City. You only need one of these conditions to be true:

• Condition 1: The "Amenable" Neighborhoods. Imagine the city has small, local districts (isotropy groups). If these districts are "nice" and "manageable" (mathematically called amenable), the weight fits perfectly.
• Condition 2: The "Essentially Free" City. Imagine the city is mostly empty of loops. Most people move around freely without getting stuck in a circle. If the "loops" (where a path leads back to the exact same spot) are so rare that they have zero population, the city is Essentially Free. In this case, the weight fits perfectly.
• Condition 3: The "Bundle" City. If the city is just a collection of independent islands (a group bundle) and you have a finite population, it works.

The Takeaway: If your city is "Essentially Free" (mostly no loops) or has "Nice" local districts, you can safely transfer your population count to the weight of the Essential City.

4. The "Unique" Weight

The paper also answers: "Is there only one way to weigh this city?"

• The Answer: Yes, IF the city is "Essentially Free."
• The Analogy: If the city has no confusing loops, there is only one logical way to assign weight to the buildings based on the population. If there are confusing loops, you might be able to weigh the city in multiple different ways, leading to confusion.

5. The Real-World Application: Self-Similar Groups

The authors apply this to Finite-State Self-Similar Groups.

• What are these? Think of a fractal, like a snowflake or a fern. No matter how much you zoom in, the pattern repeats itself. These are generated by groups of rules that look the same at every scale.
• The Result: They prove that for these specific, fractal-like mathematical structures, there is exactly one correct way to weigh the city (a unique tracial state). This is a huge deal because it means these complex, fractal cities have a very stable, predictable mathematical structure, even though they look chaotic.

Summary in One Sentence

This paper provides the rules for taking a "population count" from a messy, chaotic mathematical blueprint and successfully turning it into a unique, stable "weight" for a cleaned-up version of that blueprint, proving that for many fractal-like structures, this weight is the only one that makes sense.

Technical Summary

Here is a detailed technical summary of the paper "Invariant Measures and Traces on Groupoid $C^*$-Algebras" by Alistair Miller and Eduardo Scarparo.

1. Problem Statement and Context

The paper addresses the fundamental problem of characterizing the trace spaces of $C^*$-algebras associated with twisted étale groupoids, particularly in the non-Hausdorff setting.

• Background: The classification of $C^*$-algebras (the Elliott Programme) relies heavily on $K$-theoretic and tracial data. While the theory is well-developed for Hausdorff groupoids, many natural examples arising from dynamics (e.g., group actions on boundaries, self-similar groups) and geometry yield non-Hausdorff groupoids.
• The Core Difficulty: In non-Hausdorff groupoids, the convolution algebra $C_c(G)$ contains discontinuous functions. Consequently, the canonical conditional expectation $E: C^*(G) \to C_0(G^0)$ (which maps an element to its restriction on the unit space) does not exist in the standard sense.
• The Essential $C^*$-Algebra: To handle non-Hausdorffness, the essential $C^*$-algebra $C^*_{ess}(G)$ is introduced as a quotient of the reduced $C^*$-algebra $C^*_r(G)$ by an ideal $J$ consisting of elements whose "diagonal" values vanish almost everywhere.
• The Specific Gap: While it is known that invariant Radon measures on the unit space $G^0$ extend to traces on the full and reduced $C^*$-algebras, it is unclear when such a measure extends to a trace on the essential $C^*$-algebra. The canonical extension might not vanish on the ideal $J$, preventing the descent to the quotient. Furthermore, the conditions for the uniqueness of such extensions in the non-Hausdorff, twisted setting were not fully established.

2. Methodology

The authors employ a combination of operator algebraic techniques, measure theory, and groupoid topology:

• Twisted Étale Groupoids: The work is conducted in the generality of twisted groupoids $(G, L)$, where $L$ is a Fell line bundle. This unifies the study of standard groupoid $C^*$-algebras and those arising from Cartan pairs.
• Pedersen Ideals and Traces: The authors define traces on the Pedersen ideal $\text{Ped}(A)$ (the smallest dense ideal of a $C^*$-algebra $A$). This allows for the treatment of unbounded traces corresponding to infinite measures, extending beyond the standard bounded tracial state framework.
• Essential Freeness: A central concept introduced is essential freeness with respect to a measure $\mu$. A groupoid $G$ is essentially free with respect to $\mu$ if the set of points in the unit space where the isotropy groups are non-trivial (excluding the units themselves) has measure zero, specifically regarding the semifinite part of the measure ($\mu_{sf}$).
• Disintegration and Representation Theory: The authors utilize representations induced by isotropy groups and quasi-regular representations to construct traces and analyze their properties. They avoid relying on Renault's disintegration theorem in favor of more direct arguments inspired by Kawamura, Takemoto, and Tomiyama.
• Approximation by Hausdorff Subgroupoids: To handle non-Hausdorff pathologies, the authors often reduce problems to open subgroupoids that can be covered by countably many bisections (which are Hausdorff), utilizing the fact that $C^*_{ess}(G)$ is the direct limit of these subalgebras.

3. Key Contributions and Results

A. Existence of Traces on the Essential $C^*$-Algebra (Theorem A)

The paper provides sufficient conditions under which an invariant Radon measure $\mu$ on $G^0$ extends to a trace on the essential $C^*$-algebra $C^*_{ess}(G, L)$. The conditions are:

1. Amenable Isotropy: Every isotropy group of $G$ is amenable and the twist is trivial.
2. Essential Freeness: $G$ is essentially free with respect to $\mu$.
3. Group Bundles: $G$ is a group bundle and $\mu$ is a probability measure.

In the case of essential freeness, the canonical extension of $\mu$ (defined on the full algebra) descends to the essential algebra.

B. Uniqueness of Trace Extensions (Theorem B)

The authors establish a precise correspondence between essential freeness and the uniqueness of trace extensions:

• Forward Direction: If $G$ is essentially free with respect to $\mu$, then $\mu$ extends to a unique trace on the full $C^*$-algebra $C^*(G, L)$.
• Converse: If the twist is trivial, the converse holds: if the extension is unique, then $G$ is essentially free with respect to $\mu$.
• Twisted Case: In the twisted setting, uniqueness implies essential freeness only in one direction; the converse may fail (e.g., in irrational rotation algebras).

C. Bijection of Trace Spaces (Theorem C)

If $G$ is essentially free with respect to every invariant Radon measure, the space of traces on $C^*_{ess}(G, L)$ is affinely homeomorphic to the space of invariant Radon measures on $G^0$. This generalizes previous results from Hausdorff groupoids to the non-Hausdorff setting.

D. Application to Self-Similar Groups (Section 6)

The theory is applied to finite-state self-similar groups (e.g., Grigorchuk group).

• The authors prove that for a finite-state automorphism $\phi$, the Bernoulli measure $\mu$ satisfies $\mu(\text{Fix}_\phi \setminus \text{int}(\text{Fix}_\phi)) = 0$.
• This implies the associated groupoid of germs is essentially free with respect to the Bernoulli measure.
• Conclusion: The gauge-invariant $C^*$-algebra of a finite-state faithful self-similar group admits a unique tracial state. This result holds for both the full and essential $C^*$-algebras, confirming and generalizing previous results by Yoshida.

4. Technical Nuances and Definitions

• Semifinite Part of Measures: The authors introduce the semifinite part $\mu_{sf}$ of a Radon measure to handle cases where measures might be infinite on certain sets but finite on compact subsets. Essential freeness is defined using $\mu_{sf}$ to ensure robustness against pathologies in non-$\sigma$-finite settings.
• Dangerous Units: The set $D_0$ of "dangerous units" (points where nets converge to non-identity isotropy elements) is analyzed. The authors show that for countably generated groupoids, $D_0$ is meagre, allowing for the construction of states on the essential algebra via limits of point evaluations.
• Non-Hausdorff Pathologies: The paper explicitly addresses why standard conditional expectations fail in non-Hausdorff settings and how the ideal $J$ (measurable functions vanishing on a meagre set) serves as the correct kernel for the essential algebra.

5. Significance

• Unification: The paper unifies the study of traces for Hausdorff and non-Hausdorff groupoids, providing a framework that works for the broad class of Elliott-classifiable $C^*$-algebras.
• Resolution of Open Problems: It resolves the question of whether invariant measures always extend to the essential $C^*$-algebra, providing clear criteria (amenability or essential freeness) for existence.
• Classification Implications: By establishing the bijection between invariant measures and traces under essential freeness, the paper strengthens the link between the dynamical properties of the groupoid and the tracial data required for the Elliott classification program.
• New Examples: The application to self-similar groups provides a rigorous proof of the uniqueness of the tracial state for a wide class of simple, nuclear $C^*$-algebras that were previously difficult to analyze due to their non-Hausdorff groupoid models.

In summary, Miller and Scarparo provide a comprehensive theory of traces on non-Hausdorff twisted groupoid $C^*$-algebras, proving that essential freeness is the key dynamical property ensuring the existence and uniqueness of trace extensions, with significant applications to the structure of self-similar group algebras.