Diffeological Setting in Differential Geometry

• Diffeological setting is a framework that defines smoothness via plots rather than local charts, unifying analysis of classical, singular, and quotient spaces.
• It provides a complete, cartesian-closed category structure that supports products, coproducts, and internal mapping spaces, linking to Frölicher and differential structures.
• The approach extends differential geometry to handle infinite-dimensional objects and singular spaces through innovative constructions in tangent, bundle, and homotopy theories.

The diffeological setting is a framework for differential geometry in which smoothness is specified directly by parametrizations from Euclidean domains rather than by atlases of local charts. A diffeology on a set $X$ is a family of maps $P:U\to X$, with $U\subset\mathbb R^k$ open, satisfying covering, smooth compatibility, and locality; a map between diffeological spaces is smooth precisely when it sends plots to plots (Batubenge et al., 2017). This replacement of manifold charts by plots makes it possible to treat quotient spaces, singular spaces, orbit spaces, mapping spaces, spaces of sections, loop spaces, and other infinite-dimensional objects inside a single category that is complete, cocomplete, and cartesian-closed (Kihara, 2016).

1. Foundational definition and categorical structure

A diffeological space is a set $X$ equipped with a specified set or family of plots $p:U\to X$ from open subsets $U\subset\mathbb R^n$, for all $n$, such that all constant maps are plots, plots are closed under gluing compatible local families, and plots are closed under pre-composition with smooth maps between Euclidean domains (Kihara, 2016). Smooth maps $f:X\to Y$ are those for which $f\circ p$ is a plot of $Y$ for every plot $P:U\to X$0 of $P:U\to X$1 (Batubenge et al., 2017).

This definition induces a canonical topology, the $P:U\to X$2-topology, characterized as the coarsest topology making all plots continuous; equivalently, $P:U\to X$3 is $P:U\to X$4-open iff $P:U\to X$5 is open in $P:U\to X$6 for every plot $P:U\to X$7 (Batubenge et al., 2017). The category of diffeological spaces and smooth maps is complete, cocomplete, and cartesian-closed, and therefore admits products, coproducts, quotients, subspaces, and exponential objects $P:U\to X$8 (Kihara, 2016). Functional diffeology on $P:U\to X$9 is defined so that evaluation is smooth, which makes mapping spaces internal to the category (Iglesias-Zemmour, 15 Aug 2025).

A recurrent structural point is that manifolds embed fully faithfully into this category as those spaces locally diffeomorphic, in the diffeological sense, to open subsets of $U\subset\mathbb R^k$0 (Iglesias-Zemmour, 15 Aug 2025). At the same time, the diffeological category also contains quotient and singular spaces such as irrational tori, and infinite-dimensional spaces such as diffeomorphism groups and spaces of smooth maps (Christensen et al., 2013). This suggests that the diffeological setting is not a rival local model theory, but a common ambient category in which manifold geometry survives while non-manifold examples remain smooth objects.

2. Relations with differential and Frölicher structures

The diffeological setting is closely linked to differential and Frölicher structures. Given any family $U\subset\mathbb R^k$1 of parametrizations into a set $U\subset\mathbb R^k$2, one defines

$U\subset\mathbb R^k$3

and given any family $U\subset\mathbb R^k$4 of functions $U\subset\mathbb R^k$5, one defines

$U\subset\mathbb R^k$6

These constructions produce reflexive differential structures and reflexive diffeologies, respectively (Batubenge et al., 2017).

A diffeology $U\subset\mathbb R^k$7 is reflexive when $U\subset\mathbb R^k$8, and a differential structure $U\subset\mathbb R^k$9 is reflexive when $X$0 (Batubenge et al., 2017). The assignments $X$1 and $X$2 then give mutually inverse isomorphisms between reflexive diffeological spaces and reflexive differential spaces, while Frölicher spaces are likewise equivalent to reflexive differential spaces (Batubenge et al., 2017). In a related formulation, a Frölicher structure $X$3 induces a canonical “nebulae” diffeology

$X$4

and such diffeologies are called reflexive; on Fréchet manifolds this recovers the usual smooth structure (Magnot, 16 Jul 2025).

Examples show that reflexivity is genuinely restrictive. The spaghetti diffeology on $X$5 is not reflexive; $X$6 with the differential structure of locally $X$7-extendable functions is not reflexive; the irrational torus has trivial quotient differential structure but non-trivial and non-reflexive quotient diffeology (Batubenge et al., 2017). Orbifolds and manifolds-with-corners, by contrast, occur as reflexive differential spaces, though their natural quotient diffeology is often non-reflexive (Batubenge et al., 2017). A plausible implication is that reflexivity isolates the part of diffeological geometry that can be fully recovered from smooth real-valued functions, while non-reflexive examples retain genuinely parametrization-based information.

3. Quotients, singular spaces, and transverse geometry

One of the central uses of the diffeological setting is the treatment of quotient spaces whose ordinary topological quotient is too coarse. If a Lie group $X$8 acts on a manifold $X$9, the quotient diffeology on $p:U\to X$0 is defined by declaring $p:U\to X$1 to be a plot exactly if it locally factors through the projection $p:U\to X$2 (Cinzori, 9 Nov 2025). More generally, quotient diffeology is the final diffeology for the quotient map (Miyamoto, 2023).

For singular foliations this point becomes structural. A Stefan singular foliation $p:U\to X$3 of a manifold $p:U\to X$4 is a partition into connected, weakly embedded submanifolds such that about each point one finds a local chart identifying a neighborhood with $p:U\to X$5 in which each leaf is of the form $p:U\to X$6 for some $p:U\to X$7 (Miyamoto, 2023). Since the leaf space $p:U\to X$8 need not be Hausdorff or even $p:U\to X$9 under the quotient topology, diffeology is used to capture its smooth transverse structure (Miyamoto, 2023). The quotient diffeology on $U\subset\mathbb R^n$0 is defined by declaring $U\subset\mathbb R^n$1 to be a plot if locally it lifts through the projection $U\subset\mathbb R^n$2 to a smooth map into $U\subset\mathbb R^n$3 (Miyamoto, 2023).

The diffeological setting also supplies an intrinsic language for transverse equivalence. For singular foliations $U\subset\mathbb R^n$4 on $U\subset\mathbb R^n$5 and $U\subset\mathbb R^n$6 on $U\subset\mathbb R^n$7, one asks for a third manifold $U\subset\mathbb R^n$8 with a singular foliation $U\subset\mathbb R^n$9 and surjective submersions with connected fibers $n$0 such that $n$1; equivalently, the two foliations admit a common pullback foliation (Miyamoto, 2023). Molino transverse equivalent foliations always have diffeomorphic leaf spaces as diffeological spaces, via

$n$2

but the converse fails in general, even for regular foliations (Miyamoto, 2023).

Proper Lie group actions provide another local model for quotient singularities. If $n$3 is proper and $n$4, the slice theorem and equivariant tube theorem reduce a neighborhood of $n$5 to a quotient $n$6, where $n$7 is a slice through $n$8 (Cinzori, 9 Nov 2025). The internal tangent space at $n$9 is then identified with the fixed subspace

$f:X\to Y$0

which is also the tangent space to the stratum through $f:X\to Y$1 in the orbit-type stratification (Cinzori, 9 Nov 2025). This shows that diffeological tangent data on orbit spaces recovers the usual infinitesimal structure of the corresponding smooth stratum.

4. Tangent structures, Cartan calculus, and non-uniqueness

The diffeological setting supports several tangent constructions, but these do not coincide in general. One classical construction is the internal tangent space $f:X\to Y$2, defined as a colimit over pointed plots: $f:X\to Y$3 where $f:X\to Y$4 is generated by the basic relations $f:X\to Y$5 whenever $f:X\to Y$6 as germs at $f:X\to Y$7 (Cinzori, 9 Nov 2025). For manifolds this recovers the ordinary tangent space, and locality holds for $f:X\to Y$8-open neighborhoods (Cinzori, 9 Nov 2025).

At the categorical level, a more elaborate tangent theory is available on the full subcategory of elastic diffeological spaces. There, the left Kan extension of the standard manifold tangent functor defines an abstract tangent structure in Rosický’s sense, with natural transformations

$f:X\to Y$9

satisfying the axioms needed to define vector fields, differential forms, inner contraction, Lie derivative, and Lie bracket (Blohmann, 2023). On elastic spaces one has a graded algebra $f\circ p$0, the de Rham differential $f\circ p$1, inner contraction $f\circ p$2, Lie derivative $f\circ p$3, and the graded commutation relations

$f\circ p$4

(Blohmann, 2023).

Elastic spaces are closed under arbitrary coproducts, finite products, and retracts, and examples include manifolds with corners and cusps, diffeological groups and diffeological vector spaces with a mild extra condition, mapping spaces between smooth manifolds, and spaces of sections of smooth fiber bundles (Blohmann, 2023). In particular, for sections $f\circ p$5 one has $f\circ p$6, and for mapping spaces $f\circ p$7 one has $f\circ p$8 (Blohmann, 2023).

At the same time, the tangent functor is not unique. Infinitely many non-isomorphic tangent functors on diffeological spaces exist, all agreeing with the classical tangent functor on smooth manifolds (Taho, 22 Nov 2025). New families are obtained by choosing a based test space $f\circ p$9 and defining $Y$0-internal and $Y$1-right tangent functors; if the chosen test space has a nonzero tangent vector at $Y$2, these restrict to the ordinary tangent functor on manifolds (Taho, 22 Nov 2025). Uncountably many pairwise non-isomorphic functors arise from irrational tori, and a countably infinite family arises from orbit spaces $Y$3 (Taho, 22 Nov 2025). A common misconception is therefore that diffeological geometry has a canonical tangent bundle in the same sense as manifold theory; the literature shows that such uniqueness fails without additional universal properties.

5. Homotopy, cohomology, and model structures

The homotopy theory of diffeological spaces is organized by singular complexes built from diffeological simplices. Kihara constructs a compactly generated model structure on the category $Y$4 of diffeological spaces in which weak equivalences are smooth maps inducing weak equivalences of smooth singular simplicial sets, fibrations are maps with the right lifting property against horn inclusions $Y$5, and cofibrations are defined by the left lifting property against trivial fibrations (Kihara, 2016). The generating cofibrations and trivial cofibrations are

$Y$6

and every object is fibrant (Kihara, 2016).

The essential technical point is that the standard simplices $Y$7 are given non-naive diffeologies so that every affine map $Y$8 is smooth, the boundary inclusion $Y$9 is a $P:U\to X$00-embedding, and every horn $P:U\to X$01 is a smooth deformation retract of $P:U\to X$02 (Kihara, 2016). The singular functor $P:U\to X$03 and realization functor $P:U\to X$04 form a Quillen adjunction $P:U\to X$05, and there is a chain of Quillen equivalences linking simplicial sets, diffeological spaces, and arc-generated spaces (Kihara, 2016).

For pointed diffeological spaces, smooth homotopy groups $P:U\to X$06 are defined by $P:U\to X$07-homotopy classes of smooth maps $P:U\to X$08, and there is a natural bijection

$P:U\to X$09

which is a group isomorphism for $P:U\to X$10 (Kihara, 2016). Earlier work already showed that for fibrant diffeological spaces weak equivalences can be detected by smooth homotopy groups, that every smooth manifold without boundary is fibrant, and that free loop spaces $P:U\to X$11 are fibrant (Christensen et al., 2013).

The cohomological side is equally rich. Diffeological Čech cohomology is defined by replacing ordinary open covers with covering generating families of plots, and it is an exact $P:U\to X$12-functor of the section functor for sheaves on a diffeological space (Ahmadi, 2023). Under a partition-of-unity hypothesis on covering generating families, one obtains

$P:U\to X$13

a diffeological de Rham theorem (Ahmadi, 2023). In a different direction, Halperin’s local systems together with Kihara’s model structure yield a framework of rational homotopy theory for diffeological spaces with arbitrary fundamental groups, including an equivalence between the homotopy category of fibrewise rational diffeological spaces and an algebraic category of minimal local systems (Kuribayashi, 2021).

Open problems remain. It is open whether the three versions of diffeological Čech theory discussed in the higher-stack setting agree in all degrees $P:U\to X$14 (Minichiello, 2022). For singular foliations, the quotient map $P:U\to X$15 induces

$P:U\to X$16

whose surjectivity is known under mild hypotheses, while whether $P:U\to X$17 is always an isomorphism remains open (Miyamoto, 2023).

6. Bundles, connections, and infinite-dimensional applications

Diffeological principal bundles admit both geometric and homotopy-theoretic classifications. For a diffeological group $P:U\to X$18, Milnor’s infinite join construction yields diffeological spaces $P:U\to X$19 and $P:U\to X$20, with $P:U\to X$21 a locally trivial principal $P:U\to X$22-bundle (Magnot et al., 2016). On the category of diffeological spaces whose $P:U\to X$23-topology is Hausdorff, second-countable and smoothly paracompact, there is a natural bijection

$P:U\to X$24

between isomorphism classes of $P:U\to X$25-numerable principal $P:U\to X$26-bundles and smooth homotopy classes of smooth maps $P:U\to X$27 (Magnot et al., 2016).

This bundle theory extends to higher and simplicial frameworks. Diffeological spaces can be embedded as discrete simplicial presheaves on the site of cartesian spaces with the coverage of good open covers, and the Čech model structure on simplicial presheaves provides a notion of $P:U\to X$28-stack cohomology (Minichiello, 2022). For a diffeological group $P:U\to X$29, the nerve of the category of diffeological principal $P:U\to X$30-bundles is weak homotopy equivalent to the nerve of the category of $P:U\to X$31-principal $P:U\to X$32-bundles on $P:U\to X$33 (Minichiello, 2022).

Connections also admit intrinsic diffeological formulations. In the Milnor-classifying-space setting, a universal connection $P:U\to X$34-form on $P:U\to X$35 is constructed from the Maurer–Cartan form by

$P:U\to X$36

and this descends to a connection $P:U\to X$37-form on $P:U\to X$38; any principal $P:U\to X$39-bundle admitting a classifying map inherits such a connection $P:U\to X$40-form (Magnot et al., 2016). More recently, Singer’s universal connection has been constructed rigorously in the diffeological setting on the pointed path-bundle $P:U\to X$41, and the resulting holonomy functor yields an equivalence between a holonomy category and the category of diffeological bundle-connection pairs (Mann, 8 May 2026).

The diffeological setting is particularly effective for function spaces and low-regularity analysis. Mapping spaces $P:U\to X$42, Sobolev-type spaces $P:U\to X$43, jet spaces, and spaces of triangulations all fit naturally into diffeological or Frölicher frameworks (Goldammer et al., 2023). In the optimization setting, generalized linearizations $P:U\to X$44 on a diffeological space $P:U\to X$45 make it possible to construct paths

$P:U\to X$46

and a discrete update $P:U\to X$47 without requiring canonical charts or gradients (Magnot, 16 Jul 2025). Under mild topological assumptions the iterates admit cluster points and $P:U\to X$48 is constant on each connected component of the set of cluster points; under a suitable strict convexity hypothesis one gets convergence to the unique minimizer (Magnot, 16 Jul 2025). The method applies to $P:U\to X$49 for very low regularity, including $P:U\to X$50 (Magnot, 16 Jul 2025).

A broader theme is that diffeology carries differential forms, symplectic reduction, moment maps, and prequantization beyond the manifold setting (Iglesias-Zemmour, 15 Aug 2025). It also supports gluing theories for differential forms and Dirac operators on spaces that are not smooth manifolds in any ordinary sense (Pervova, 2016). This suggests that the diffeological setting functions not merely as a language for pathological examples, but as a systematic extension of differential geometry to quotients, singularities, and infinite-dimensional constructions where chart-based smoothness is unavailable or unstable.