---
title: Glimm Ideals in C*-Algebras
url: https://www.emergentmind.com/topics/glimm-ideals
type: topic
---

# Glimm Ideals in C*-Algebras

For a \(C^*\)-algebra \(A\), a Glimm ideal is the ideal attached to a point of the complete regularization of the primitive ideal space \(\mathrm{Prim}(A)\). Concretely, one identifies primitive ideals \(P,Q\in \mathrm{Prim}(A)\) when every bounded continuous function on \(\mathrm{Prim}(A)\) takes the same value at \(P\) and \(Q\); the resulting quotient is the Glimm space \(\mathrm{Glimm}(A)\), and the point corresponding to the class of \(P\) determines the ideal
\[
G_p=\bigcap \{Q\in \mathrm{Prim}(A):\rho_A(Q)=p\}.
\]
Equivalently, \(G_p=k([P])\), and the corresponding equivalence class is \([P]=\mathrm{hull}(G_p)\). In this sense, Glimm ideals record exactly the primitive-ideal data that survive passage to the “best Hausdorff approximation” of \(\mathrm{Prim}(A)\) [1212.3971].

## 1. Complete regularization and the basic definition

Let \(A\) be a \(C^*\)-algebra. The primitive ideal space \(\mathrm{Prim}(A)\) is equipped with the hull-kernel topology, and the Glimm space is the complete regularization of \(\mathrm{Prim}(A)\). Two primitive ideals \(P_1,P_2\in \mathrm{Prim}(A)\) are equivalent if they cannot be separated by bounded continuous real-valued functions:
\[
P_1 \sim P_2 \quad\Longleftrightarrow\quad f(P_1)=f(P_2)\ \text{for every bounded continuous } f:\mathrm{Prim}(A)\to \mathbb{R}.
\]
The quotient map is denoted \(q_A\) in one formulation and \(\rho_A\) in another, depending on the source [1203.3350].

A Glimm ideal is the ideal corresponding to a point of this quotient. If \(p\in \mathrm{Glimm}(A)\), then the associated ideal is the kernel of the corresponding equivalence class. One useful description is
\[
G_p=\bigcap \{Q\in \mathrm{Prim}(A):\rho_A(Q)=p\},
\]
while another is \(G_p=k([P])\) for any \(P\) with \(\rho_A(P)=p\). This identifies Glimm ideals as intersections of primitive ideals lying in a single complete-regularization fiber [1212.3971].

The topology of \(\mathrm{Glimm}(A)\) is subtle. The literature distinguishes the quotient topology \(T_q\), induced directly by the quotient map, from the completely regular topology \(T_{cr}\), the weakest topology making the descended bounded continuous functions continuous. One always has
\[
T_q \supseteq T_{cr},
\]
and equality is not automatic [1203.3350].

This quotient picture is closely tied to the center of the multiplier algebra. The Dauns–Hofmann theorem is used in this setting to obtain
\[
Z(M(A)) \cong C_b(\operatorname{Glimm}(A)),
\]
so bounded continuous functions on the Glimm space encode central multiplier data. A standard interpretation is therefore that Glimm ideals organize the part of primitive-ideal theory visible to bounded continuous central parameters [1509.06140].

## 2. Relation to primitive, minimal primal, and quasi-standard structures

Glimm ideals sit between primitive ideal theory and more global ideal-theoretic structures. The Glimm space is formed from primitive ideals, but in favorable cases it coincides topologically with the space of minimal primal ideals. A particularly important condition is quasi-standardness. For a \(C^*\)-algebra \(A\), the relation \(\sim\) on \(\mathrm{Prim}(A)\) is defined by
\[
P\sim Q \quad\Longleftrightarrow\quad (\forall C\in \mathrm{Cl}(\mathrm{Prim}(A)))\, \bigl(P\in C \Rightarrow Q\in C \text{ or } P,Q\notin C\bigr),
\]
and \(A\) is quasi-standard if this relation is an open equivalence relation. In that case,
\[
\mathrm{MinPrimal}(A)=\mathrm{Glimm}(A)
\]
as topological spaces [2211.04424].

The group \(C^*\)-algebras of irrational Mautner groups provide a concrete model. For
\[
G_\theta=\mathbb C^2\rtimes_\alpha \mathbb R
\]
with irrational \(\theta\), the paper proves that \(C^*(G_\theta)\) is quasi-standard and that
\[
\mathrm{MinPrimal}(C^*(G_\theta))=\mathrm{Sub}(C^*(G_\theta))=\mathrm{Glimm}(C^*(G_\theta))
\]
as topological spaces. Moreover, all these spaces are homeomorphic to
\[
[0,\infty)^2.
\]
The homeomorphism is induced by
\[
[(z,T)]\longmapsto (|z_1|,|z_2|),
\]
so the Glimm quotient collapses the non-Hausdorff directions in the primitive spectrum exactly down to the two orbit radii [2211.04424].

This example makes the function of Glimm ideals especially transparent. Primitive ideals retain fine phase and stabilizer data, whereas the Glimm ideal remembers only the bounded-continuous invariants of the orbit structure. In quasi-standard settings, the resulting quotient is not merely computable but also aligned with the minimal primal structure.

## 3. Tensor products and product formulas for Glimm ideals

One of the most useful structural facts about Glimm ideals is their behavior under tensor products. For the minimal \(C^*\)-tensor product \(A\otimes_\alpha B\), there is a canonical map on ideals
\[
\Phi(I,J)=\ker(q_I\otimes q_J),
\]
together with the ideal-valued map
\[
\Delta(I,J)=I\otimes_\alpha B + A\otimes_\alpha J.
\]
At the level of Glimm spaces, the main result is a homeomorphism
\[
\psi:\mathrm{Glimm}(A\otimes_\alpha B)\longrightarrow \big(\mathrm{Glimm}(A)\times \mathrm{Glimm}(B),\tau_{cr}\big),
\]
compatible with the regularization maps [1212.3971].

When Glimm spaces are regarded as sets of ideals, the inverse correspondence is implemented explicitly by
\[
(G,H)\longmapsto G\otimes_\alpha B + A\otimes_\alpha H.
\]
More precisely, for Glimm ideals \(G\in \mathrm{Glimm}(A)\) and \(H\in \mathrm{Glimm}(B)\),
\[
\Delta(G,H)=G\otimes_\alpha B + A\otimes_\alpha H
\]
is the inverse of the canonical Glimm-space homeomorphism. This extends earlier work by eliminating the assumption of property (F) [1212.3971].

The topology on the product is again significant. The correct default topology is the complete-regularization topology \(\tau_{cr}\), not necessarily the ordinary product topology \(\tau_p\). The paper gives sufficient conditions ensuring \(\tau_{cr}=\tau_p\): for example, if \(A\) satisfies one of the following—\(\mathrm{Prim}(A)\) is compact, \(\rho_A\) is open, or \(A\) is \(\sigma\)-unital and \(\mathrm{Glimm}(A)\) is locally compact—then for every \(C^*\)-algebra \(B\),
\[
\mathrm{Glimm}(A\otimes_\alpha B)\cong \big(\mathrm{Glimm}(A)\times \mathrm{Glimm}(B),\tau_p\big).
\]
The same paper also identifies
\[
ZM(A\otimes_\alpha B)\cong C\!\left(\beta\big(\mathrm{Glimm}(A)\times \mathrm{Glimm}(B),\tau_{cr}\big)\right),
\]
and gives necessary and sufficient conditions for the inclusion \(ZM(A)\otimes ZM(B)\subseteq ZM(A\otimes_\alpha B)\) to be surjective [1212.3971].

A parallel theory holds for the Haagerup tensor product of a TRO \(V\) with a \(C^*\)-algebra \(B\). In that setting,
\[
\operatorname{Glimm}(V\otimes^h B)=\{\,I\otimes^h B+V\otimes^h J : I\in \operatorname{Glimm}(V),\ J\in \operatorname{Glimm}(B)\,\},
\]
and the canonical map
\[
\Phi(I,J)=I\otimes^h B+V\otimes^h J
\]
is a homeomorphism from \((\operatorname{Glimm}(V)\times \operatorname{Glimm}(B),\tau_{cr})\) onto \(\operatorname{Glimm}(V\otimes^h B)\). The paper further proves that \(V\otimes^h B\) is quasi-standard if and only if both \(V\) and \(B\) are quasi-standard [2508.12367].

## 4. Topological properties of Glimm spaces

Glimm ideals are points of a topological quotient, so the topology of \(\mathrm{Glimm}(A)\) is part of the theory rather than a secondary feature. For \(\sigma\)-unital \(C^*\)-algebras, compact subsets of the Glimm space can be controlled by quotient norms arising from strictly positive elements. If \(A\) is \(\sigma\)-unital and \(a\in A\) is strictly positive, then for every \(T_{cr}\)-compact subset \(K\subset \mathrm{Glimm}(A)\) there exists \(\alpha>0\) such that
\[
K \subset \left\{G\in \mathrm{Glimm}(A)\mid \|a+G\|>\alpha\right\}.
\]
This extends a result of Dauns from special classes, such as quasicentral algebras, to all \(\sigma\)-unital \(C^*\)-algebras [1203.3350].

The same work gives an exact criterion for when a compact set admits such a norm-threshold description. For a \(T_{cr}\)-compact subset \(K\subset \mathrm{Glimm}(A)\), the following are equivalent: there exist \(a\in A\) and \(\alpha>0\) such that
\[
K\subset \{G\in \mathrm{Glimm}(A)\mid \|a+G\|>\alpha\},
\]
and there exists a compact subset \(C\subset \mathrm{Prim}(A)\) such that \(q_A(C)=K\). It follows that any such \(K\) is \(T_q\)-compact. The paper also constructs a \(C^*\)-algebra with a \(T_{cr}\)-compact subset of \(\mathrm{Glimm}(A)\) that is not contained in any set of the above norm-threshold form, showing that the \(\sigma\)-unital hypothesis cannot simply be discarded [1203.3350].

For separable \(C^*\)-algebras, the topology of Glimm spaces admits a precise abstract characterization. A Hausdorff space \(X\) occurs as \(\operatorname{Glimm}(A)\) for some separable \(C^*\)-algebra \(A\) if and only if it is homeomorphic to the quotient of a locally compact Hausdorff second countable space; equivalently, if and only if
\[
X=\bigcup_{n=1}^\infty X_n,
\]
where \(\{X_n\}\) is an increasing sequence of compact metrizable subspaces such that a subset \(F\subseteq X\) is closed iff \(F\cap X_n\) is closed in \(X_n\) for every \(n\). The same theorem holds with “separable \(C^*\)-algebra” replaced by “AF algebra” [1509.06140].

The non-unital case is markedly different from the unital one. If \(A\) is unital, then \(\operatorname{Glimm}(A)\) is compact and homeomorphic to the maximal ideal space of \(Z(A)\). If \(A\) is non-unital, the Glimm space may be much larger than the maximal ideal space of the center; the cited example is \(A=C^*(G)\) for the continuous Heisenberg group, where \(Z(A)=\{0\}\) but \(\operatorname{Glimm}(A)\cong \mathbb R\) [1509.06140].

## 5. Global Glimm phenomena and modern regularity theory

In recent work, “Glimm” has also come to designate a regularity property extending classical Glimm-type halving phenomena from simple algebras to the non-simple setting. The Global Glimm Property is defined by the requirement that for every \(a\in A_+\) and every \(\varepsilon>0\), there exists \(r\in \overline{aAa}\) such that
\[
r^2=0
\]
and \((a-\varepsilon)_+\) lies in the ideal generated by \(r\). Equivalent formulations use \(^*\)-homomorphisms from \(C_0((0,1],M_2)\) or \(C_0((0,1],M_k)\) into \(aAa\) [2204.13059].

This property is not the classical definition of a Glimm ideal, but it is explicitly motivated by Glimm theory. The survey literature describes the historical role of Glimm ideals as part of the primitive-ideal and quotient-space framework from which the Global Glimm Problem emerged. In that perspective, Glimm ideals encode the “best Hausdorff approximation” of the primitive ideal space, while the Global Glimm Property asks whether the absence of local elementary obstructions is already enough to force a strong hereditary halving property [2512.13334].

The Cuntz semigroup formulation is central. One has
\[
A \text{ is nowhere scattered} \iff \mathrm{Cu}(A)\text{ is weakly }(2,\omega)\text{-divisible},
\]
and
\[
A \text{ has the Global Glimm Property} \iff \mathrm{Cu}(A)\text{ is }(2,\omega)\text{-divisible}.
\]
The main structural theorem of one paper states that the following are equivalent: \(A\) has the Global Glimm Property; \(\mathrm{Cu}(A)\) is \((2,\omega)\)-divisible; \(\mathrm{Cu}(A)\) is weakly \((2,\omega)\)-divisible, ideal-filtered, and has property (V); and \(A\) is nowhere scattered and \(\mathrm{Cu}(A)\) is ideal-filtered and has property (V). The same work proves that every \(C^*\)-algebra contains a unique largest ideal with the Global Glimm Property [2204.13059].

A particularly sharp positive result concerns topological dimension zero, meaning that \(\mathrm{Prim}(A)\) has a basis consisting of compact open sets. In that setting,
\[
\boxed{ \text{A \(C^*\)-algebra with topological dimension zero has the Global Glimm Property} \iff \text{it is nowhere scattered.} }
\]
The proof passes through the equivalences
\[
A \text{ nowhere scattered} \iff \Cu(A)\text{ weakly }(2,\omega)\text{-divisible}
\]
and
\[
A \text{ topological dimension zero} \iff \Cu(A)\otimes\{0,\infty\}\text{ is algebraic},
\]
together with a semigroup theorem upgrading weak \((2,\omega)\)-divisibility to full \((2,\omega)\)-divisibility. A corollary is that nowhere scattered \(C^*\)-algebras with finite nuclear dimension and topological dimension zero are pure [2507.16261].

## 6. Glimm-like obstruction theory beyond \(C^*\)-algebras

The term “Glimm” also appears in descriptive set theory through the Glimm–Effros dichotomy. This is not a theory of Glimm ideals in the \(C^*\)-algebraic sense, but it provides a related obstruction-theoretic viewpoint. One recent paper generalizes the Glimm–Effros dichotomy from Polish spaces to quotients by strongly idealistic Borel equivalence relations. If \(X\) is such a quotient and \(E\) is a countable weakly Borel equivalence relation on \(X\), then exactly one of the following holds: either \(X\) is a countable union of Borel transversals of \(E\), or there is a strongly Borel embedding of a canonical quotient \(\Cantorspace/F\) into \(E\), where \(F\) is eventual equality or one of the prime congruence relations [2105.05374].

The same paper proves a finite-basis theorem in finite-index situations: among finite obstructions, prime cyclic relations are the minimal ones. It also generalizes the Lusin–Novikov uniformization theorem in a parallel dichotomic form, again with eventual equality and prime congruence relations as the canonical counterexamples [2105.05374].

This suggests a broader “Glimm-like” pattern: a classification problem either admits a tame decomposition by transversals or uniformizations, or else contains a canonical minimal obstruction. In the operator-algebraic setting, Glimm ideals arise from complete regularization of primitive-ideal spaces; in the descriptive-set-theoretic setting, the analogous role is played by explicit equivalence relations such as eventual equality. The two settings are different, but both organize non-Hausdorff or nonclassifiable behavior by passing to canonical quotients and isolating minimal obstructions.

Source: https://www.emergentmind.com/topics/glimm-ideals