nLab infinitesimal interval object




homotopy theory, (∞,1)-category theory, homotopy type theory

flavors: stable, equivariant, rational, p-adic, proper, geometric, cohesive, directed

models: topological, simplicial, localic, …

see also algebraic topology



Paths and cylinders

Homotopy groups

Basic facts


Synthetic differential geometry

synthetic differential geometry


from point-set topology to differentiable manifolds

geometry of physics: coordinate systems, smooth spaces, manifolds, smooth homotopy types, supergeometry



smooth space


The magic algebraic facts




infinitesimal cohesion

tangent cohesion

differential cohesion

graded differential cohesion

singular cohesion

id id fermionic bosonic bosonic Rh rheonomic reduced infinitesimal infinitesimal & étale cohesive ʃ discrete discrete continuous * \array{ && id &\dashv& id \\ && \vee && \vee \\ &\stackrel{fermionic}{}& \rightrightarrows &\dashv& \rightsquigarrow & \stackrel{bosonic}{} \\ && \bot && \bot \\ &\stackrel{bosonic}{} & \rightsquigarrow &\dashv& \mathrm{R}\!\!\mathrm{h} & \stackrel{rheonomic}{} \\ && \vee && \vee \\ &\stackrel{reduced}{} & \Re &\dashv& \Im & \stackrel{infinitesimal}{} \\ && \bot && \bot \\ &\stackrel{infinitesimal}{}& \Im &\dashv& \& & \stackrel{\text{étale}}{} \\ && \vee && \vee \\ &\stackrel{cohesive}{}& \esh &\dashv& \flat & \stackrel{discrete}{} \\ && \bot && \bot \\ &\stackrel{discrete}{}& \flat &\dashv& \sharp & \stackrel{continuous}{} \\ && \vee && \vee \\ && \emptyset &\dashv& \ast }


Lie theory, ∞-Lie theory

differential equations, variational calculus

Chern-Weil theory, ∞-Chern-Weil theory

Cartan geometry (super, higher)



An infinitesimal interval object is like an interval object that is an infinitesimal space.

Where maps out of an interval object model paths that may arrange themselves into path ∞-groupoids, maps out of infinitesimal interval object model infinitesimal paths that may arrange themselves into infinitesimal path ∞-groupoid.


In any lined topos (𝒯,(R,+,))(\mathcal{T},(R,+,\cdot)) the line object RR is naturally regarded as a cartesian interval object.

R 0 * 1 * * *. \array{ && R \\ & {}^{0_{*}}\nearrow && \nwarrow^{1_{*}} \\ {*} &&&& {*} } \,.

If the lined topos (𝒯,R)(\mathcal{T}, R) is also a smooth topos in that it satisfies the Kock-Lawvere axiom it makes sense to consider the subobject DD of RR defined as the equalizer

D=lim(R0() 2R) D = \lim\left( R \stackrel{\stackrel{(-)^2}{\to}}{\underset{0}{\to}} R \right)

or equivalently expressed in topos logic as

D={xR|x 2=0}. D = \{x \in R | x^2 = 0 \} \,.

and think of DD as the infinitesimal interval object of the smooth topos (𝒯,R)(\mathcal{T},R) inside its finite interval object RR. By the axioms satisfied by a smooth topos it is in particular an infinitesimal object.

Various constructions induced by a finite interval object have their infinitesimal analog for infinitesimal interval objects.

Infinitesimal path \infty-groupoid induced from infinitesimal interval

Urs Schreiber: the following should be checked


Recall the discussion at interval object of how the interval **R{*}\sqcup {*} \stackrel{\to}{\to} R in 𝒯\mathcal{T} alone gave rise to the cosimplicial object

Δ R:Δ𝒯 \Delta_R : \Delta \to \mathcal{T}

of (collared) kk-simplices modeled on RR, and how that induces for each object X𝒯X \in \mathcal{T} the simplicial object

Π(X):=X Δ R . \Pi(X) := X^{\Delta_R^\bullet} \,.

Here as an object in 𝒯\mathcal{T} the kk-simplex is actually a kk-cube Δ R k:=R k\Delta_R^k := R^k, but equipped with face and degeneracy maps that identify the boundary of a kk-simplex inside the kk-cube thus realizing the interior of that boundary as the kk-simplex proper and the exterior as its collar .

We want to mimic that construction with the finite interval RR replaced by the infinitesimal interval object DD, to get a simplicial object Π inf(X)\Pi^{inf}(X) for every object XX.

While the infinitesimal situation is formally very similar to the finite situation, one technical diference is that the infinitesimal interval does not fit into a nontrivial cospan as the finite interval does. This is because DD typically has a unique morphism *D{*} \to D from the terminal object, as a consequence of the fact that all the infinitesimal elements it contains are genuinely generalized elements.

The natural way to encode an infinitesimal path between two elements in an object XX in the smooth topos 𝒯\mathcal{T} is therefore not as an element of X DX^D but of X D×DX^D \times D, which we may think of as the space of pairs consisting of infinitesimal paths in XX and infinitesimal parameter lengths of these paths.

This naturally yields the span

X D×D p TXp 1 ev X X, \array{ && X^D \times D \\ & {}^{\mathllap{p_{T X}\circ p_1}}\swarrow && \searrow^{\mathrlap{ev}} \\ X &&&& X } \,,


  • the left leg is the projection X D×DX DX^D \times D \to X^D followed by the tangent bundle projection p TX=X (*D):X DXp_{T X} = X^{(* \to D)} : X^D \to X

  • the right leg is the evaluation map, i.e. the inner hom-adjunct of Id:X DX DId : X^D \to X^D.

With this setup a pair of (generalized) elements x,yXx,y \in X may be thought of as connected by an infinitesimal path if there is an element (v,ϵ)X D×D(v,\epsilon) \in X^D \times D such that

δ 1(v,ϵ)=v(0)=x \delta_1(v,\epsilon) = v(0) = x


δ 0(v,ϵ)=v(ϵ)=y. \delta_0(v,\epsilon) = v(\epsilon) = y \,.

But not all elements (v,ϵ)(v,\epsilon) define different pairs of infinitesimal neighbour elements this way: specifically in the case that XX is a microlinear space, the tangent bundle object X DX^D is fiberwise RR-linear, and thus for any tRt \in R the elements (v,tϵ)(v,t \epsilon) and (tv,ϵ)(t v, \epsilon) define the same pair of infinitesimal neighbours, x=v(0)x = v(0) and y=(tv)(ϵ)=v(tϵ)y = (t v)(\epsilon) = v(t \epsilon).

We may identify such elements (tv,ϵ)(t v, \epsilon) and (v,tϵ)(v, t \epsilon) by passing to the tensor product X D RDX^D \otimes_R D, i.e. the coequalizer of

X D×R×DId××IdX D×D X^D \times R \times D \stackrel{\stackrel{\cdot \times Id}{\to}}{\underset{Id \times \cdot}{\to}} X^D \times D

where \cdot here denotes the monoid-action of (R,)(R,\cdot) on DD (by the embedding DRD \hookrightarrow R) and on X DX^D (by the fact that XX is assumed to be microlinear).

In this same fashion we can then define infinitesimal analogs of the finite higher path object X Δ R k=X R ×kX^{\Delta_R^k} = X^{R^{\times k}}.

The definition


(infinitesimal path simplicial object)

Let X𝒯X \in \mathcal{T} be a microlinear space in the smooth topos (𝒯,R)(\mathcal{T},R) with infinitesimal interval object DD.

Then define the simplicial object

Π inf(X):Δ op𝒯 \Pi^{inf}(X) : \Delta^{op} \to \mathcal{T}

as follows:

  • in degree nn it assigns the object

    [n]X D(n) R nD(n)(X D RD) × X n [n] \mapsto X^{D(n)} \otimes_{R^n} D(n) \hookrightarrow (X^D \otimes_R D)^{\times_X^n}

    whose generalized elements we write (ϵ iv i) x(\epsilon_i v_i)_x with ϵD(n)\vec \epsilon \in D(n) or (ν i) x(\nu_i)_x for short; where xXx \in X indicates the fiber of X D(n)XX^{D(n)} \to X that the element lives in

  • the face maps d i:(X D× RD) × X n+1(X D× RD) × X nd_i : (X^D \times_R D)^{\times_X^{n+1}} \to (X^D \times_R D)^{\times_X^{n}} are

    • for 0<i<n+10 \lt i \lt n+1

      given on generalized elements by

      d i:(ν i) x(ν 0,,ν i2,ν i1+ν i,ν i+1,,ν m+1) x d_i : (\nu_i)_x \mapsto (\nu_0 , \cdots, \nu_{i-2}, \nu_{i-1} + \nu_i, \nu_{i+1}, \cdots, \nu_{m+1} )_x
    • for i=0i = 0

      given by

      d 0:(ν i)(v 0(ϵ 0)+ν i) d_0 : (\nu_i) \mapsto (v_0(\epsilon_0) + \nu_i)

      where the element on the right denotes the evaluation of the map (ν i):D(n)X(\nu_i) : D(n) \to X in its first argument on ϵ 0\epsilon_0, regarded as an element in the fiber over v 0(ϵ 0)v_0(\epsilon_0).

    • for i=n+1i = n+1

      given by

      d n+1:(ν 0,,ν n+1)(ν 0,,ν n)d_{n+1} : (\nu_0, \cdots, \nu_{n+1}) \mapsto (\nu_0, \cdots, \nu_n)

  • the degeneracy maps σ i\sigma_i act by inserting the 0-vector in position i:

    σ i:( i)(ν 0,,ν i1,0,ν i+1,,ν n)\sigma_i : (\ni_i) \mapsto (\nu_0, \cdots, \nu_{i-1}, 0, \nu_{i+1}, \cdots, \nu_{n}).


These face and degeneracy maps indeed satisfy the simplicial identities.


This is straightforward checking that proceeds entirely analogously as the proof of the simplicial identities for the finite path \infty-groupoid Π(X)\Pi(X) discussed at interval object. See also the following remark.


By thinking of the v i:DXv_i : D \to X as infinitesimal collared curves in XX with source v i(0)v_i(0) and target v i(ϵ i)v_i(\epsilon_i) the above definition is an immediate analog of the definition of the path \infty-groupoid Π(X)\Pi(X) of finite paths as discussed at interval object.

This is made manifest by the following construction that embeds Π inf(X)\Pi^{inf}(X) into Π(X)\Pi(X).

The inclusion of infinitesimal paths into finite paths

Recall the finite path \infty-groupoid Π(X)\Pi(X) induced from the interval object

(0 *,1 *):**R (0_*,1_*) : * \coprod * \to R

as discussed there. On object this assigns

Π(X):[n]X R n. \Pi(X) : [n] \mapsto X^{R^n} \,.

(inclusion of infinitesimal into finite paths)

For nn \in \mathbb{N} define a morphism

X (D(n)) R nD(n)X R n X^{(D(n))} \otimes_{R^n} D(n) \to X^{R^n}

on generalized elements by

ι n:(ϵ iv i)((t 0,,t n1) i=0 n1v i(t iϵ i)). \iota_n : (\epsilon_i v_i) \mapsto ((t_0, \cdots, t_{n-1}) \mapsto \sum_{i=0}^{n-1} v_i(t_i \epsilon_i)) \,.

These morphism (ι n)(\iota_n) respect the face and degeneracy maps on both sides and hence induce an inclusion of simplicial objects

Π inf(X)Π(X) \Pi^{inf}(X) \hookrightarrow \Pi(X)

Straightforward checking on generalized elements.


under construction

Let XX be a microlinear space.

Sketch of Proposition

We want to show that the morphism of simplicial sets

[Δ op,𝒯](U×D,X (Δ inf ))[Δ op,𝒯](U,X (Δ inf )) [\Delta^{op},\mathcal{T}](U \times D, X^{(\Delta_\inf^\bullet)}) \to [\Delta^{op},\mathcal{T}](U , X^{(\Delta_\inf^\bullet)})

induced by pullback along UU×*U×DU \simeq U \times * \to U \times D is a weak homotopy equivalence.

Sketch of proof

First consider the case that UU itself has no infinitesimal directions in that Hom(U,D)=*Hom(U,D) = *. Then we claim that the morphism [Δ op,𝒯](U×D,X (Δ inf ))[Δ op,𝒯](U,X (Δ inf )) [\Delta^{op},\mathcal{T}](U \times D, X^{(\Delta_\inf^\bullet)}) \to [\Delta^{op},\mathcal{T}](U , X^{(\Delta_\inf^\bullet)}) is an acyclic Kan fibration in that all squares

Δ[n] [Δ op,𝒯](U×D,X (Δ inf )) Δ[n] [Δ op,𝒯](U,X (Δ inf )) \array{ \partial \Delta[n] &\to& [\Delta^{op},\mathcal{T}](U \times D, X^{(\Delta_\inf^\bullet)}) \\ \downarrow && \downarrow \\ \Delta[n] &\to& [\Delta^{op},\mathcal{T}](U , X^{(\Delta_\inf^\bullet)}) }

have lifts.

For n=0n = 0 this says that the map is surjective on vertices, which it is, since any UXU \to X is in the image of U×DU×*UXU \times D \to U \times * \simeq U \to X.

For n=1n=1 we need to check that every homotopy UX D×DU \to X^D \times D downstairs with fixed lifts U×DXU \times D \to X over the endpoints may be lifted. But by assumption UDU \to D factors through the point, so that the homotopy UX D×DXU \to X^D \times D \stackrel{\to}{\to} X has the same source as target f:UXf : U \to X. This means the two lifts of its endpoints are morphisms (u,ϵ)(f(u)+ϵν i(u)(u,\epsilon) \mapsto (f(u) + \epsilon \nu_i(u) for i=1,2i=1,2 and ν i\nu_i tangent vectors. A homotopy between these is given by a map U×DX D×DU \times D \to X^D \times D defined by (u,ϵ)(f(u)+()ν 1(u)+(ϵ())ν 2(u),ϵ)(u,\epsilon) \mapsto (f(u)+ (-)\nu_1(u) + (\epsilon-(-))\nu_2(u), \epsilon).

Finally, for n2n \geq 2 we have unique flllers, because by construction every morphism Δ[n]YX (Δ inf )\Delta[n]\cdot Y \to X^{(\Delta^\bullet_{inf})} is uniquely fixed by the restriction Δ[n]Δ[n]YX (Δ inf )\partial \Delta[n] \to \Delta[n]\cdot Y \to X^{(\Delta^\bullet_{inf})} to its boundary.

The case for U×DU \times D replaced with U×D nU \times D^n works analogously.

Last revised on October 13, 2016 at 09:40:01. See the history of this page for a list of all contributions to it.