Comparison Map in tt-Category Theory

Updated 7 July 2026

The comparison map of a tt-category is a spectral map that connects the Balmer spectrum with the prime spectrum of an endomorphism ring.
Higher comparison maps use tensor-balanced morphisms and iterative cone constructions to refine support and capture intricate tt-geometry.
Functoriality and localization techniques ensure these maps translate complex tensor-triangular structures into accessible algebraic frameworks.

In tensor-triangular geometry, a comparison map is a spectral map that relates the Balmer spectrum of a tt-category to the prime spectrum of a commutative ring built from endomorphisms. In its classical form, it sends the spectrum of prime thick tensor ideals to the Zariski spectrum of the endomorphism ring of the tensor unit; in its higher forms, it is defined on supports of objects, on tensor-multiplicative families, and on arbitrary closed subsets of the Balmer spectrum. The basic purpose of the construction is to replace parts of the typically intricate space $\Spc(\mathcal K)$ by affine schemes $\Spec(R)$ that are more accessible ring-theoretically, while retaining enough functorial and support-theoretic information to analyze $\Spc(\mathcal K)$ recursively (Sanders, 2013).

1. Classical setting and the unit comparison map

Let $\mathcal K$ be an essentially small tensor triangulated category with exact symmetric monoidal structure

$(\mathcal K,\otimes,\mathbb 1).$

Its Balmer spectrum $\Spc(\mathcal K)$ consists of prime thick tensor ideals, equipped with the Balmer topology. For an object $a\in\mathcal K$ , the support is

$\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$

a closed subset of $\Spc(\mathcal K)$. In this framework, Thomason closed subsets are exactly the supports of objects: a closed subset $\mathcal Z\subseteq \Spc(\mathcal K)$ is Thomason if and only if $\Spec(R)$0 for some $\Spec(R)$1 (Sanders, 2013).

The classical comparison map is attached to the tensor unit. Since $\Spec(R)$2 is commutative, one has Balmer’s map

$\Spec(R)$3

and in graded form

$\Spec(R)$4

This map is the basic bridge between tt-geometry and commutative algebra: it compares prime tensor ideals with prime ideals in a ring of endomorphisms of the unit. The later theory of higher comparison maps is an extension of this construction rather than a replacement of it (Sanders, 2013).

The phrase comparison map of a tt-category is therefore used in two closely related senses. In a narrow sense, it denotes the unit map $\Spec(R)$5. In a broader sense, it denotes the family of maps obtained by replacing the unit by arbitrary tensor-multiplicative data, so that supports, fibers, and closed subsets can themselves be analyzed by further comparison maps (Sanders, 2013).

2. Higher comparison maps and canonical commutative rings

The decisive extension is due to Sanders. For any object $\Spec(R)$6, one constructs a commutative ring $\Spec(R)$7 and a map

$\Spec(R)$8

together with a graded analogue

$\Spec(R)$9

More generally, if $\Spc(\mathcal K)$0 is a nonempty tensor-multiplicative subset, one defines

$\Spc(\mathcal K)$1

and a comparison map

$\Spc(\mathcal K)$2

The object map is the special case $\Spc(\mathcal K)$3, while for a closed subset $\Spc(\mathcal K)$4 one may take

$\Spc(\mathcal K)$5

which yields a canonical map $\Spc(\mathcal K)$6 (Sanders, 2013).

The construction begins with tensor-balanced endomorphisms. An endomorphism $\Spc(\mathcal K)$7 is tensor-balanced if

$\Spc(\mathcal K)$8

as endomorphisms of $\Spc(\mathcal K)$9. The ring of balanced endomorphisms is

$\mathcal K$ 0

Its graded analogue is

$\mathcal K$ 1

A key lemma asserts that if $\mathcal K$ 2 is tensor-balanced, then

$\mathcal K$ 3

This is the cone-killing mechanism underlying locality and nilpotence arguments in the theory (Sanders, 2013).

Because $\mathcal K$ 4 need not be commutative, the comparison target cannot simply be $\mathcal K$ 5 itself. Sanders first defines “unnatural” comparison maps from a chosen commutative ring $\mathcal K$ 6 equipped with a homomorphism $\mathcal K$ 7, and then replaces this ad hoc choice by a canonical commutative ring $\mathcal K$ 8. Concretely,

$\mathcal K$ 9

where $(\mathcal K,\otimes,\mathbb 1).$ 0 is generated by the relations

$(\mathcal K,\otimes,\mathbb 1).$ 1

for all $(\mathcal K,\otimes,\mathbb 1).$ 2. Writing $(\mathcal K,\otimes,\mathbb 1).$ 3 for the class of $(\mathcal K,\otimes,\mathbb 1).$ 4, one obtains a commutative ring with operations

$(\mathcal K,\otimes,\mathbb 1).$ 5

$(\mathcal K,\otimes,\mathbb 1).$ 6

For an object $(\mathcal K,\otimes,\mathbb 1).$ 7, the ring $(\mathcal K,\otimes,\mathbb 1).$ 8 is identified with the colimit

$(\mathcal K,\otimes,\mathbb 1).$ 9

This canonicalization is one of the main conceptual steps in the theory: the target ring is extracted from balanced endomorphisms without imposing commutativity externally (Sanders, 2013).

3. Definition by cones, supports, and spectrality

The defining formula for a comparison map is expressed in terms of cones. For an external commutative ring $\Spc(\mathcal K)$0 with $\Spc(\mathcal K)$1, the associated map is

$\Spc(\mathcal K)$2

For the canonical ring $\Spc(\mathcal K)$3, the natural comparison map is

$\Spc(\mathcal K)$4

The graded version is

$\Spc(\mathcal K)$5

Thus a prime tensor ideal is sent to the set of parameters whose associated cones remain outside that prime. This definition is inclusion-reversing, reflecting the specialization order in the Balmer spectrum (Sanders, 2013).

Continuity is encoded by explicit pullback formulas. For the unnatural map,

$\Spc(\mathcal K)$6

If $\Spc(\mathcal K)$7 is Thomason, then

$\Spc(\mathcal K)$8

Similarly, for the canonical object map,

$\Spc(\mathcal K)$9

These formulas are the practical reason comparison maps are useful: closed subsets in the affine target pull back to supports of explicit objects in the tt-category (Sanders, 2013).

The higher comparison maps are spectral maps. In particular, they are continuous and preserve quasi-compact opens under inverse image. Several topological consequences follow. If $a\in\mathcal K$ 0 is connected, then $a\in\mathcal K$ 1 is connected; if $a\in\mathcal K$ 2 is rigid and $a\in\mathcal K$ 3 is Thomason, the converse also holds. Proper closed subsets of the target pull back to proper closed subsets of $a\in\mathcal K$ 4, and therefore $a\in\mathcal K$ 5 and $a\in\mathcal K$ 6 have dense image. The density statement is explicitly noted as new even for Balmer’s original unit comparison map (Sanders, 2013).

4. Iteration, fibers, and maps attached to closed subsets

The adjective higher refers not merely to generality but to recursion. One starts with the unit comparison map

$a\in\mathcal K$ 7

chooses a Thomason closed subset $a\in\mathcal K$ 8 in the target, and studies its preimage. If $a\in\mathcal K$ 9, then the preimage is the support of a tensor product of cones. To retain a canonical object on that support, Sanders takes

$\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$0

so that $\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$1 is precisely the desired closed subset and carries a new comparison map

$\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$2

The process can then be repeated. This yields a descending chain

$\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$3

a filtration obtained by successive affine approximations (Sanders, 2013).

A particularly important variant is the map attached directly to a closed subset. For

$\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$4

one has

$\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$5

hence a canonical comparison map

$\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$6

This avoids choosing a generator $\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$7 with $\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$8. If $\supp(a)=\{\mathcal P\in \Spc(\mathcal K)\mid a\notin \mathcal P\},$9 is Thomason, the resulting ring agrees canonically with the ring obtained from the class of objects with support exactly $\Spc(\mathcal K)$0. The construction is therefore generator-independent and works even when a single object representing $\Spc(\mathcal K)$1 is unavailable (Sanders, 2013).

This closed-subset formalism supports a deterministic strategy for analyzing $\Spc(\mathcal K)$2: choose a filtration of an affine target by closed subsets, pull it back along $\Spc(\mathcal K)$3, and iterate on the resulting closed subsets. A key result states that if $\Spc(\mathcal K)$4 is closed and

$\Spc(\mathcal K)$5

then $\Spc(\mathcal K)$6 is the whole homogeneous spectrum. Hence nontrivial closed subsets in the target always induce nontrivial refinements on the source. This makes higher comparison maps a genuine mechanism for extracting new geometric information rather than merely repackaging existing support data (Sanders, 2013).

5. Functoriality, invariance, and localization

Comparison maps are functorial with respect to tt-functors. If $\Spc(\mathcal K)$7 is a tt-functor and $\Spc(\mathcal K)$8, there is a ring homomorphism

$\Spc(\mathcal K)$9

and a commutative diagram relating the induced maps on spectra. Special cases include naturality for object comparison maps and for closed-subset maps. This makes the construction compatible with the standard morphisms of tt-geometry (Sanders, 2013).

At the object level, the map depends only on the thick-tensor information carried by the object, not on a specific presentation. If $\mathcal Z\subseteq \Spc(\mathcal K)$0, then $\mathcal Z\subseteq \Spc(\mathcal K)$1 and $\mathcal Z\subseteq \Spc(\mathcal K)$2. In a rigid category one also has

$\mathcal Z\subseteq \Spc(\mathcal K)$3

Likewise,

$\mathcal Z\subseteq \Spc(\mathcal K)$4

and more generally the same holds when $\mathcal Z\subseteq \Spc(\mathcal K)$5 is a finite direct sum of suspensions of $\mathcal Z\subseteq \Spc(\mathcal K)$6. Tensoring with another object $\mathcal Z\subseteq \Spc(\mathcal K)$7 induces a map $\mathcal Z\subseteq \Spc(\mathcal K)$8; if $\mathcal Z\subseteq \Spc(\mathcal K)$9, its kernel consists entirely of nilpotents, and for $\Spec(R)$00 there is an isomorphism

$\Spec(R)$01

under which the corresponding comparison maps agree (Sanders, 2013).

A major technical theorem is triangular localization. If $\Spec(R)$02 is a multiplicative set of even homogeneous elements and

$\Spec(R)$03

then for the Verdier quotient $\Spec(R)$04 one has

$\Spec(R)$05

together with a cartesian diagram on homogeneous spectra. Equivalently,

$\Spec(R)$06

This generalizes Balmer’s central localization for the unit and is indispensable for fiberwise and local analyses (Sanders, 2013).

The later literature places these structural properties into broader topological and geometric frameworks. Rowe develops the Balmer spectrum as a locally ringed space and studies associated sheaf functors; in the affine case, the comparison map identifies $\Spec(R)$07 with $\Spec(R)$08 and the sheaves $\Spec(R)$09 become the usual quasi-coherent sheaves attached to $\Spec(R)$10, while the schematic case is obtained by local affine comparison (Rowe, 2021). In a different direction, Shaul shows that for connective rigidly-compactly generated tt-categories the ungraded comparison map is a strong spectral quotient map with connected fibers, and in the unigenic connective case each fiber is local with a canonical relatively closed point detected by a weight complex functor (Sanders, 4 Aug 2025).

The principal worked example in the foundational paper is the stable homotopy category of finite spectra. For finite spectra,

$\Spec(R)$11

since $\Spec(R)$12. This map is far from injective: over each prime $\Spec(R)$13 lies the chromatic chain of primes $\Spec(R)$14. After $\Spec(R)$15-localization, the spectrum is

$\Spec(R)$16

Because the full rings $\Spec(R)$17 are not computed there, Sanders uses the graded comparison maps attached to

$\Spec(R)$18

For $\Spec(R)$19, $\Spec(R)$20 has exactly two points: a generic prime of homogeneous nilpotents and a closed point, the ideal of homogeneous nonunits. If $\Spec(R)$21 is a tensor-balanced $\Spec(R)$22-selfmap, the closed point is $\Spec(R)$23, and

$\Spec(R)$24

Iterating this construction produces the filtration

$\Spec(R)$25

removing one chromatic point at each stage. This is the prototype of higher comparison-map analysis (Sanders, 2013).

The same example also exhibits a limitation. The point $\Spec(R)$26 is not Thomason, so one should not expect every irreducible closed subset of a tt-spectrum to arise directly from noetherian affine targets. The theory nonetheless captures all Thomason closed subsets, and those form a basis of closed sets. A plausible implication is that comparison maps are most naturally adapted to the Thomason geometry intrinsic to compactly generated tt-categories, rather than to arbitrary sobriety phenomena (Sanders, 2013).

Later work shows that the ordinary unit comparison map can be too small even more drastically. In the tt-category of compact permutation-derived objects for a finite group, the tensor unit satisfies

$\Spec(R)$27

so the ungraded comparison map would collapse to a single point; the actual tt-spectrum is instead stratified by modular fixed-points and cohomological spectra of Weyl groups (Balmer et al., 2022). Part II replaces the inadequate target $\Spec(R)$28 by a twisted cohomology ring

$\Spec(R)$29

built from invertible twists $\Spec(R)$30, and defines a new comparison map

$\Spec(R)$31

For elementary abelian groups this map is an open immersion globally, and on a natural open cover it becomes a homeomorphism onto the homogeneous spectrum of a local graded endomorphism ring (Balmer et al., 2023).

This later behavior clarifies the conceptual scope of the notion. The classical comparison map packages information visible from the tensor unit; higher comparison maps enlarge the source and refine the target by using arbitrary objects or closed subsets; twisted comparison maps enlarge the target further by allowing invertible twists when $\Spec(R)$32 is too small. In a complementary topological direction, patch-density results imply that any spectral map out of $\Spec(R)$33 is determined by its restriction to a patch-dense subset; since comparison maps are spectral, this yields a uniqueness principle for reconstructing them from sufficiently rich families of probes (Balmer et al., 19 Mar 2025).

The comparison map of a tt-category is therefore best understood not as a single isolated morphism but as a hierarchy of spectral approximations to the Balmer spectrum. At one end lies the unit map

$\Spec(R)$34

at another the family of higher maps

$\Spec(R)$35

and in settings where the unit is ring-theoretically too poor, twisted or localized variants. Across these forms, the unifying principle is constant: prime tensor ideals are studied by sending them to prime ideals generated by those endomorphism parameters whose cones remain invisible at the given point, thereby translating tt-geometric structure into affine algebraic geometry in a way that can be iterated, localized, and functorially compared (Sanders, 2013).