Complete Bornological Convolution Algebras

• Complete bornological convolution algebras are associative structures on complete convex bornological spaces that employ bounded convolution products via fiber integration against Haar systems.
• Their construction leverages completed tensor products and bounded maps to ensure smooth module actions, Morita invariance, and compatibility with homological frameworks.
• These algebras support advanced analytic tools such as cyclic homology and bivariant K-theory, bridging functional analysis with noncommutative geometry.

Searching arXiv for recent and foundational papers on bornological convolution algebras, Lie/groupoid convolution, and related bornological completion frameworks. Complete bornological convolution algebras are associative algebras whose underlying vector spaces are complete convex bornological spaces and whose multiplication is bounded, with the convolution product arising from geometric, algebraic, or quantum-group data. In the recent Lie groupoid setting, they are realized as the completed convolution algebras $A(G)=C^\infty_{\mathrm c}(G_1)$ in the category $\cBorn$ of complete convex bornological vector spaces, with multiplication defined by fiber integration against a Haar system (Aretz et al., 14 Aug 2025). In adjacent frameworks, the same phrase encompasses complete bornological dual convolution algebras of bornological quantum groups (Rivet et al., 2021), as well as completed compact-support algebras that are not formulated bornologically in the original source but admit structurally analogous Banach or $C^*$-completions (Henry, 2016). The subject sits at the intersection of functional analysis, noncommutative geometry, Lie groupoid theory, Hopf-algebraic duality, and homological algebra; its central concern is to retain smooth or test-function-level information while imposing enough completeness to support tensor products, Morita theory, homology, $K$-theory, and analytic completion procedures (Aretz et al., 14 Aug 2025, Miaskiwskyi, 2022, Mukherjee, 2022).

1. Conceptual framework and ambient categories

The modern theory is built in the category $\cBorn$ of complete convex bornological vector spaces, where completeness means that every bounded set is contained in a bounded completant disk, and where the completed tensor product $\Cotimes$ is part of a symmetric monoidal closed structure (Aretz et al., 14 Aug 2025). In the quantum-group setting, the same completeness condition is phrased by requiring every bounded set to lie in a completant disk whose span is Banach under the gauge seminorm; multiplication is then a bounded map $m:A\hat\otimes A\to A$ (Rivet et al., 2021). These formulations agree at the level of emphasis: completeness is not an optional topological refinement, but the condition that makes convolution, multiplier constructions, and derived tensor products workable.

A central structural distinction is between bornological and topological completions. In Fréchet settings, the completed bornological tensor product agrees with the usual projective and inductive completed topological tensor products, so the bornological theory recovers continuous topological homology (Miaskiwskyi, 2022). For Fréchet spaces $V,W$, one has

$$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$

and the completed bornological tensor product $V\hotimes W$ is again Fréchet (Miaskiwskyi, 2022). For nuclear strict LF-spaces, the completed bornological tensor product commutes with strict inductive limits and remains within the same category (Miaskiwskyi, 2022). This is one reason complete bornological methods are well suited to compactly supported smooth convolution spaces, which are typically strict LF-spaces.

The completed tensor product is indispensable because convolution is fundamentally bilinear and often integral-kernel based. In the Lie groupoid theory, the identity

$\cBorn$0

is an explicit starting point for defining completed convolution algebra structures (Aretz et al., 14 Aug 2025). In topological Hopf-algebroid theory, a formally analogous role is played by the completed tensor product $\cBorn$1, defined as the completion of the filtered tensor product, which is then used to formulate complete convolution Hopf algebroids (Kaoutit et al., 2017). Although that paper is not bornological, its tensor-Hom formalism is structurally parallel to the bornological one.

The bornological viewpoint is also tightly linked to module-theoretic nondegeneracy. In bornological quantum groups, algebras and modules are required to be essential, meaning bounded sets in a module arise from bounded sets in $\cBorn$2 under the action map (Rivet et al., 2021). In Lie groupoid convolution theory, non-unitality is handled instead through strong approximate units and smooth/self-induced modules (Aretz et al., 14 Aug 2025). Both approaches replace the algebraic convenience of a unit by bounded or Mackey-convergent approximation data.

2. Lie groupoid convolution as the prototypical complete bornological construction

The most explicit recent theory of complete bornological convolution algebras is the Lie groupoid construction of $\cBorn$3 for a Lie groupoid $\cBorn$4 equipped with a right Haar system (Aretz et al., 14 Aug 2025). The convolution product is

$\cBorn$5

and the involution is

$\cBorn$6

Associativity follows from right invariance and Fubini, and compact support is preserved (Aretz et al., 14 Aug 2025). The algebra is generally non-unital.

The point of the bornological completion is not merely to observe that $\cBorn$7 is already complete as a bornological vector space, but to ensure that tensor products, balanced tensor products, and coequalizers are taken in $\cBorn$8. The completed multiplication is the bounded map

$\cBorn$9

obtained as the composite

$C^*$0

where the last two arrows are pullback to composable pairs and fiber integration along multiplication (Aretz et al., 14 Aug 2025). This formulation clarifies that the algebra structure is an algebra object in $C^*$1, not just an algebraic convolution law on a vector space.

The same paper extends convolution to bibundles. If $C^*$2 is a $C^*$3-$C^*$4 bibundle, then $C^*$5 becomes an $C^*$6-$C^*$7-bimodule, with actions

$C^*$8

and

$C^*$9

These formulas make precise the bornological realization of geometric correspondences as convolution bimodules (Aretz et al., 14 Aug 2025).

The deeper categorical result is that complete bornological convolution is functorial with respect to Morita geometry. The assignments $K$0, $K$1, together with the completed functoriality constraint

$K$2

define a weak monoidal $K$3-functor from the geometric Morita $K$4-category of Lie groupoids and principal bibundles to the Morita $K$5-category of self-induced algebras and smooth bimodules in $K$6 (Aretz et al., 14 Aug 2025). This result makes the complete bornological convolution algebra into a genuine stack-theoretic invariant rather than a presentation-dependent algebraic artifact.

3. Non-unitality, approximate units, and smooth Morita theory

A persistent feature of convolution algebras is non-unitality. For Lie groups, $K$7 is unital only when $K$8 is discrete, and for manifolds $K$9 is unital only when $\cBorn$0 is compact (Aretz et al., 14 Aug 2025). The theory therefore replaces unitality by more refined bornological conditions.

The relevant notion in the Lie groupoid setting is a countable strong approximate right unit: a sequence $\cBorn$1 such that the right multiplication operators $\cBorn$2 Mackey converge to the identity in the functional bornology on $\cBorn$3 (Aretz et al., 14 Aug 2025). This is stronger than pointwise convergence $\cBorn$4, because it gives bounded-control convergence uniformly on bounded subsets. The principal theorem states that the completed bornological convolution algebra $\cBorn$5 has strong approximate right and left units (Aretz et al., 14 Aug 2025).

This operator-bornological convergence has decisive consequences. First, the canonical map $\cBorn$6 is injective when $\cBorn$7 has a strong approximate right unit (Aretz et al., 14 Aug 2025). Second, for a right $\cBorn$8-module $\cBorn$9, smoothness is equivalent to the action map $\Cotimes$0 being a strong epimorphism (Aretz et al., 14 Aug 2025). Third, every convolution bimodule $\Cotimes$1 is smooth, and therefore the algebra $\Cotimes$2 is self-induced: $\Cotimes$3 (Aretz et al., 14 Aug 2025). This is the module-theoretic substitute for unitality that makes Morita bicategory constructions behave correctly.

A further strengthening is projectivity. If a groupoid action is submersive, proper, and transitive on the relevant fibers, then the convolution module $\Cotimes$4 is projective in the bornological module category (Aretz et al., 14 Aug 2025). For Morita bibundles, this implies left and right projectivity, and consequently the convolution algebra of a Lie groupoid is quasi-unital (Aretz et al., 14 Aug 2025). The paper notes that quasi-unitality implies $\Cotimes$5-unitality via known results, making these algebras amenable to homological tools.

Parallel non-unital phenomena appear elsewhere. In the topos-theoretic convolution algebra of an absolutely locally compact topos, the algebra is generally non-unital, and module theory is phrased in terms of non-degenerate modules (Henry, 2016). In bornological quantum groups, essentiality of algebras and modules plays an analogous role, replacing units by nondegenerate multiplier-algebra behavior (Rivet et al., 2021). Across these frameworks, complete bornological convolution algebras are rarely naturally unital; their good behavior rests instead on approximate units, essentiality, or smoothness.

4. Variants and related realizations: quantum groups, toposes, Hopf algebroids, and discrete completions

The phrase “complete bornological convolution algebra” covers several structurally distinct but closely related constructions.

In bornological quantum group theory, the dual object $\Cotimes$6 is transported via the Fourier transform $\Cotimes$7 to a convolution algebra $\Cotimes$8 on the underlying vector space of $\Cotimes$9 (Rivet et al., 2021). Its product is

$m:A\hat\otimes A\to A$0

and its convolution adjoint is

$m:A\hat\otimes A\to A$1

Because the bornology is transferred through a bounded bijection from the complete convex bornological algebra $m:A\hat\otimes A\to A$2, $m:A\hat\otimes A\to A$3 is a complete bornological convolution $m:A\hat\otimes A\to A$4-algebra (Rivet et al., 2021). It embeds densely in $m:A\hat\otimes A\to A$5, while $m:A\hat\otimes A\to A$6 embeds densely in the reduced $m:A\hat\otimes A\to A$7-algebraic locally compact quantum group $m:A\hat\otimes A\to A$8 (Rivet et al., 2021). This realizes bornological convolution algebras as dense analytic cores of quantum groups.

In the topos-theoretic setting, one starts not with bornology but with compactly supported kernels on separating objects of an absolutely locally compact topos. The resulting algebra $m:A\hat\otimes A\to A$9 is an involutive convolution algebra or pseudo-category characterized by an equivalence between its non-degenerate modules and sheaves of $V,W$0-modules over $V,W$1 (Henry, 2016). Composition is by fiberwise summation of kernels, and in the étale groupoid case it reproduces the usual groupoid convolution law

$V,W$2

(Henry, 2016). Although the paper does not formulate a bornology, it constructs norm-based completions $V,W$3, $V,W$4, and $V,W$5, thereby supplying a completion tower adjacent to the bornological one (Henry, 2016).

A filtered/topological variant appears in the theory of complete Hopf algebroids. For a cocommutative right Hopf algebroid $V,W$6, the right $V,W$7-linear dual

$V,W$8

carries convolution

$V,W$9

and, under admissible filtration hypotheses, becomes a complete commutative Hopf algebroid with explicit topological antipode

$$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$0

(Kaoutit et al., 2017). This paper does not use bornological language, but the passage from filtered duals to completed convolution structures mirrors the bornological passage from bounded bilinear operations to complete tensor-algebraic objects.

The nonarchimedean theory over a complete discrete valuation ring $$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$1 furnishes another family of complete bornological convolution-type algebras. The paper on nonarchimedean bivariant $$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$2-theory uses as its canonical stabilizing algebra the completed matrix algebra

$$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$3

with convolution product (Mukherjee, 2022). More generally, length-controlled matrix algebras $$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$4 exhibit coefficient decay governed by a proper length function and strongly resemble weighted convolution or rapid-decay completions (Mukherjee, 2022). The same paper emphasizes complete, torsion-free, semidagger bornological $$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$5-algebras and their dagger completions, which are particularly relevant for discrete or coefficient-valued convolution constructions (Mukherjee, 2022).

A closely related nonarchimedean completion framework is the dagger completion of bornological algebras over a discrete valuation ring. For monoid algebras $$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$6, the dagger completion $$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$7 consists of formal sums

$$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$8

whose coefficients satisfy linear valuation growth in word length,

$$V\otimes_\pi W \cong V\otimes_\iota W \cong V\otimes_\beta W,$$9

and the same applies to twisted monoid algebras and crossed products (Meyer et al., 2018). These are direct models of complete bornological convolution algebras in a discrete or semigroup setting, though the paper emphasizes dagger rather than convolution terminology.

5. Homological, $V\hotimes W$0-theoretic, and functional-analytic structures

One of the main motivations for imposing complete bornological structure is to obtain workable homological and $V\hotimes W$1-theoretic formalisms.

For Fréchet and strict LF algebras, a bornological Hochschild and cyclic theory can be built from the completed bornological tensor product $V\hotimes W$2, and for Fréchet spaces this agrees with continuous Lie algebra homology (Miaskiwskyi, 2022). The bornological Hochschild complex uses terms $V\hotimes W$3, and the continuity of the operators $V\hotimes W$4, $V\hotimes W$5, $V\hotimes W$6, $V\hotimes W$7, $V\hotimes W$8, and $V\hotimes W$9 follows from bounded multiplication together with the completed tensor product formalism (Miaskiwskyi, 2022). The paper proves explicit homology computations for $\cBorn$00, $\cBorn$01, and $\cBorn$02, verifies closed-range conditions for their cyclic differentials, proves bornological $\cBorn$03-unitality in the compact-support case, and establishes a bornological Loday–Quillen–Tsygan theorem in nuclear Fréchet and nonunital $\cBorn$04-unital settings (Miaskiwskyi, 2022). While these algebras use pointwise multiplication rather than convolution, they provide the complete bornological homological framework that convolution algebras would need to satisfy analogous hypotheses.

In the nonarchimedean setting, complete bornological convolution-type algebras fit into bivariant $\cBorn$05-theory $\cBorn$06 for complete, bornologically torsion-free bornological $\cBorn$07-algebras (Mukherjee, 2022). The central ambient conditions are completeness, torsion-freeness, and bounded multiplication; for dagger-homotopy-sensitive results one often also requires semidagger or dagger structure (Mukherjee, 2022). The theory is stabilized using the continuous matrix convolution algebra $\cBorn$08, and $\cBorn$09 is characterized as the universal recipient of functors that are dagger homotopy invariant, $\cBorn$10-stable, and excisive (Mukherjee, 2022). This yields exact triangles, excision, analytic $\cBorn$11-theory, and Chern characters to analytic and periodic cyclic homology (Mukherjee, 2022).

The functional-analytic infrastructure for automatic boundedness is provided by closed graph theorems for bornological spaces over complete non-trivially valued fields. For complete bornological domains and codomains with compatible nets or webs, linear maps with bornologically closed graphs are bounded (Bambozzi, 2015). For bornological algebras, an algebra morphism

$\cBorn$12

is automatically bounded when $\cBorn$13 is complete, $\cBorn$14 is webbed, and $\cBorn$15 admits enough bornologically closed finite-codimensional ideals separating points (Bambozzi, 2015). These theorems do not prove boundedness of the bilinear convolution law itself, but they are relevant to convolution operators, algebra morphisms, and quotient or representation maps between complete bornological convolution algebras.

The algebraic theory of vector-valued integration on measurable bornological sets supplies a more foundational, pre-convolution layer. It constructs a monad $\cBorn$16 of boundedly supported signed measures and interprets $\cBorn$17-algebras as spaces carrying integration operations

$\cBorn$18

compatible with linear structure and Pettis integration (Lucyshyn-Wright, 2011). This theory does not define convolution, but it isolates supportwise measure completion, pushforwards, and flattening operations that would underlie any future categorical theory of bornological convolution on measurable semigroups or groupoids (Lucyshyn-Wright, 2011).

6. Examples, comparisons, and scope

The class of complete bornological convolution algebras includes a wide range of examples, but the exact algebraic meaning of “convolution” varies by context.

The Lie groupoid case includes ordinary Lie group convolution algebras $\cBorn$19, pair-groupoid kernel algebras $\cBorn$20, action-groupoid crossed products $\cBorn$21, Čech groupoid matrix-style convolution algebras, gauge groupoid algebras, and bornological noncommutative tori arising from $\cBorn$22 (Aretz et al., 14 Aug 2025). For pair groupoids, the product takes the integral-kernel form

$\cBorn$23

and for action groupoids it becomes

$\cBorn$24

(Aretz et al., 14 Aug 2025). These examples show that complete bornological convolution algebras subsume both ordinary group convolution and integral-operator algebras.

The topos-theoretic theory recovers étale groupoid convolution algebras, Steinberg algebras, profinite double-coset algebras, and Leavitt path algebras as instances of a general compact-support pseudo-category (Henry, 2016). Its natural equivalence with sheaf categories shows that convolution can be understood as a compact-support presentation of a geometric category, canonical up to Morita equivalence (Henry, 2016).

The bornological quantum-group framework covers classical locally compact groups, where one may take $\cBorn$25, $\cBorn$26, or $\cBorn$27 on the function side and obtain the usual group convolution on the dual side (Rivet et al., 2021). In that sense, complete bornological convolution algebras are presented as “test-function” cores sitting densely inside analytic group or quantum-group algebras.

Not every relevant complete bornological algebra is literally a convolution algebra. The homological paper on $\cBorn$28 and $\cBorn$29 is foundational for complete bornological algebra structures but treats pointwise multiplication, not convolution (Miaskiwskyi, 2022). Its relevance lies in establishing exactness, tensor-product comparison, and cyclic-homological methods for the kinds of Fréchet and LF spaces that frequently underlie smooth convolution algebras (Miaskiwskyi, 2022). Likewise, the topos-theoretic compact-support algebras are not bornological in formulation, but their support control, Morita invariance, and norm completions form a natural comparison class (Henry, 2016).

A common misconception is that complete bornological convolution algebras are merely smooth subalgebras awaiting $\cBorn$30-completion. The recent Lie groupoid work suggests a stronger conclusion: the bornological completion is already the correct environment for Morita $\cBorn$31-functoriality, smooth bimodule theory, and stack-level functoriality, whereas purely algebraic convolution lacks sufficient completion and $\cBorn$32-completion discards smooth structure (Aretz et al., 14 Aug 2025). A plausible implication is that the bornological category occupies an intermediate position: strong enough for analytic and homological constructions, but still fine enough to remember differential-geometric data.

Another misconception is that “complete” means only Banach or norm complete. The literature exhibits several inequivalent but related notions: completant disks in convex bornologies (Rivet et al., 2021), complete convex bornological spaces in $\cBorn$33 (Aretz et al., 14 Aug 2025), completion of bounded $\cBorn$34-adically complete submodules over a DVR (Mukherjee, 2022), and filtered/projective-limit completeness in Hopf algebroid theory (Kaoutit et al., 2017). These are not interchangeable definitions, though they often play analogous roles.

Taken together, the current literature presents complete bornological convolution algebras as a family of algebraic-analytic objects unified less by a single universal definition than by a recurrent pattern: compactly supported or test-function convolution is placed in a complete bornological environment with a completed tensor product, bounded multiplication, and enough approximation or essentiality to support Morita theory, homological algebra, and analytic completion procedures (Aretz et al., 14 Aug 2025, Rivet et al., 2021, Mukherjee, 2022). This pattern now spans Lie groupoids, quantum groups, nonarchimedean analytic algebras, and compact-support geometric categories, suggesting that complete bornological convolution is becoming a common language for “smooth but complete” noncommutative geometry.