nLab manifold structure of mapping spaces

Contents

Context

Differential geometry

synthetic differential geometry

Introductions

from point-set topology to differentiable manifolds

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

Differentials

V-manifolds

smooth space

Tangency

The magic algebraic facts

Theorems

Axiomatics

cohesion

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 }

Models

Lie theory, ∞-Lie theory

differential equations, variational calculus

Chern-Weil theory, ∞-Chern-Weil theory

Cartan geometry (super, higher)

Mapping space

Contents

Idea

The category of smooth manifolds is not cartesian closed, even when infinite-dimensional manifolds are allowed. However, that does not mean that no mapping spaces between certain smooth manifolds can be given the structure of a smooth manifold. This is true when the source is compact. Thus, in particular, this applies to loop spaces.

The method of proving this depends mostly on the structure of the target and only minimally on that of the source. It is not hard to generalise it to manifolds with boundary (to get, for example, path spaces), or even manifolds with corners. This raises the obvious question as to how general this result can be made. The purpose of this page is to determine the answer. Our conjecture is the following:

Conjecture

Let NN be a Frölicher space whose curvaceous topology is sequentially compact. Let MM be a smooth manifold that admits a local addition. Then the Frölicher space of smooth maps from NN to MM is a smooth manifold.

Construction of smooth manifold structure on mapping space

The following needs attention. For a more recent version see (Stacey).

Background and Remarks

The question discussed here can be viewed as the counterpoint to the oft-heard maxim (attributed to Grothendieck):

It is better to work in a nice category with nasty objects than in a nasty category with nice objects.

Smooth manifolds are an example of “nice objects in a nasty category”; for example, one can rarely take subobjects or quotients. The standard procedure at this point is to embed the nasty category in some larger, nicer category and work there. In the case of smooth manifolds, this has led to all of the categories that are listed at generalized smooth space.

One can now go on to study this enlarged category, and investigate how much of what is known about the original category extends to the larger one. In this line, the original category is viewed mainly as a source of ideas. An alternative approach, and that taken here, is to view the original category as being a subcategory of “special objects” inside the larger one.

One can make an analogy with the real and complex numbers. Many aspects of the study of real numbers become much easier and clearer when extended to the complex numbers. At this point, one has a choice: one can simply study the complex numbers or one can use the complex numbers as a tool to study the real ones.

Thus, to adapt a saying of Hadamard (here), we could introduce our own maxim:

The shortest distance between two truths about nice objects often lies in a nice category.

Having mentioned the plethora of extensions of the category of smooth manifolds, we should comment on our choice of Frölicher spaces. The inclusion of the category of smooth manifolds into each of the extensions factors through the category of Frölicher spaces. Therefore, if we work in, say, the category of diffeological spaces then we can split the question “Is the diffeological space XX a smooth manifold?” into “Is XX a Frölicher space?” and “Is the resulting Frölicher space a manifold?”. Moreover, as we are interested in C (N,M)C^\infty(N,M) with MM a smooth manifold (and thus a Frölicher space), then if we are working with one of the “maps in” approaches, we can replace the NN in C (N,M)C^\infty(N,M) by its “Frölicherification” without changing the set. Thus the key piece of the puzzle is to study C (N,M)C^\infty(N,M) for NN a Frölicher space and the rest will follow by applying “general nonsense”.

Another remark worth saying is that the conjecture stated is not the most general statement that could be considered. It is simple to extend this conjecture to a relative version whereby MM is equipped with a family of submanifolds and NN with a family of subsets and the maps are constrained to take the subsets to the corresponding submanifolds.

Finally, let us note that the main results about the linear model spaces are recorded on the page linear mapping spaces.

Charts

Let MM be a smooth manifold (possibly infinite dimensional) modelled on the convenient vector space VV. Let NN be a sequentially compact Frölicher space. Let {P i:P iM}\{P_i : P_i \subseteq M\} be a family of submanifolds of MM. Let {Q i:Q iN}\{Q_i : Q_i \subseteq N\} be a family of subsets of NN with the same indexing set.

Definition

We write C (N,M;Q i,P i)C^\infty(N,M;Q_i,P_i) for the set of smooth functions NMN \to M which map each Q iQ_i into the corresponding P iP_i.

As a smooth manifold, MM naturally has the structure of a Frölicher space so this mapping space is well-defined.

We assume that the pair (M,{P i})(M,\{P_i\}) admits a local addition. By that, we mean that MM admits a local addition, say η\eta, with the property that it restricts to a local addition on each P iP_i. We shall also assume, for simplicity, that the domain of η\eta is TMT M.

Let g:NMg \colon N \to M be a smooth map with g(Q i)P ig(Q_i) \subseteq P_i. Let E gE_g be the space of sections of g *TMg^* T M with the property that the sections over each Q iQ_i are constrained to lie in the corresponding g *TP ig^* T P_i. In more detail, we define g *TMg^* T M in the usual manner:

g *TM{(x,v)N×TM:g(x)=π(v)} g^* T M \coloneqq \{(x,v) \in N \times T M : g(x) = \pi(v)\}

and then take the space of smooth maps f:Ng *TMf \colon N \to g^* T M with the property that the composition Ng *TMNN \to g^* T M \to N is the identity. Within that space, we further restrict to those ff such that the image of the map Q ig *TMTMQ_i \to g^* T M \to T M lies in TP iT P_i.

Although NN could be quite complicated, because TMMT M \to M is a vector bundle, E gE_g is a vector space. Furthermore, by trivialising g *TMg^* T M using a finite number of trivialisations (possible as NN is sequentially compact), we can embed E pE_p as a closed subspace of C (N,V n)C^\infty(N,V^n) for some nn. This embedding shows that E pE_p is a convenient vector space, in the sense of Kriegl and Michor.

Andrew Stacey This, I think, is the crucial part: that E pE_p is a convenient vector space. I need to expand on this and check that all is as I think it is.

We define a map for Φ:E gC (N,M;{Q i},{P i})\Phi \colon E_g \to C^\infty(N,M;\{Q_i\},\{P_i\}) as follows. Let fE pf \in E_p. Then ff is a section of g *TMg^* T M and so is a map Ng *TMN \to g^* T M. By the definition of g *TMg^* T M, we can think of ff as a map NN×TMN \to N \times T M which projects to the identity on the first factor. By applying the projection to the second factor, we obtain a map f^:NTM\hat{f} \colon N \to T M. Composing with η\eta produces a map ηf^:NM\eta \circ \hat{f} \colon N \to M. As fE gf \in E_g, the restriction of f^\hat{f} to Q iQ_i lands in TP iT P_i, whence ηf^\eta \circ \hat{f} takes Q iQ_i into P iP_i. The map fηf^f \mapsto \eta \circ \hat{f} is what we call Φ\Phi.

Let us identify its image. Let VM×MV \subseteq M \times M be the image of the local addition. Define U gC (N,M;{Q i},{P i})U_g \subseteq C^\infty(N,M;\{Q_i\},\{P_i\}) to be the set of those functions hh such that (g,h):NM×M(g,h) \colon N \to M \times M takes values in VV. We claim that the image of Φ\Phi is U gU_g and that Φ\Phi is a bijection E gU gE_g \to U_g.

Let us start with the image. Let hU gh \in U_g. Then (g,h):NM×M(g,h) \colon N \to M \times M takes values in VV, so we can compose with (π×η) 1(\pi \times \eta)^{-1} to get a map hˇ:NTM\check{h} \colon N \to T M. Together with the identity on NN, we get a map NN×TMN \to N \times T M. By construction, πhˇ=g\pi \check{h} = g and so this map ends up in g *TMg^* T M (which has the subspace structure). Again by construction, the projection of this map to NN is the identity and so it is a section of g *TMg^* T M. That it takes QQ to TPT P follows from the fact that η\eta restricts to a local addition on PP, whence as h(Q)Ph(Q) \subseteq P, hˇ(Q)TP\check{h}(Q) \subseteq T P. Hence Φ\Phi is onto. Moreover, this construction yields the inverse of Φ\Phi and so it is a bijection.

Thus we have charts for C (N,M;Q,P)C^\infty(N,M;Q,P).

Transition functions

The next step is the transition functions. To prove this in full generality, we assume not just two different functions at which to base our charts, but also two different local additions to define them. This will show that our resulting manifold structure is independent of this choice. We could go further than we do, and allow our local additions to be in the most general form given at local addition, but this would crowd the notation with little benefit.

Thus we start with g 1,g 2C (N,M;Q,P)g_1, g_2 \in C^\infty(N,M;Q,P) and two local additions η 1,η 2:TMM\eta_1, \eta_2 \colon T M \to M. Let us write V 1V_1 and V 2V_2 for the images of (π×η 1)(\pi \times \eta_1) and (π×η 2)(\pi \times \eta_2). Let E 1=E g 1E_1 = E_{g_1} and E 2=E g 2E_2 = E_{g_2}.

We define W 12g 1 *TMW_{1 2} \in g_1^* T M as follows. We describe a point in g 1 *TMg_1^* T M by specifying its point in N×TMN \times T M.

W 12{(x,v)g 1 *TM:(g 2(x),η 1(v))V 2} W_{1 2} \coloneqq \{(x,v) \in g_1^* T M : (g_2(x),\eta_1(v)) \in V_2\}

and W 21g 2 *TMW_{2 1} \subseteq g_2^* T M similarly.

Lemma

W 12W 21W_{1 2} \cong W_{2 1}

Proof

The map g 2×η 1:N×TMM×Mg_2 \times \eta_1 \colon N \times T M \to M \times M is smooth and so the preimage of V 2V_2 under this map is open in N×TMN \times T M. Thus W 12W_{1 2} is open in g *TMg^* T M and the map (x,v)(g 2(x),η 1(v))(x, v) \mapsto (g_2(x), \eta_1(v)) is a smooth map W 12V 2W_{1 2} \to V_2. Since (π×η 2):TMV 2(\pi \times \eta_2) \colon T M \to V_2 is a diffeomorphism, we can define a smooth map θ 1:W 12TM\theta_1 \colon W_{1 2} \to T M by

θ 1(x,v)=(π×η 2) 1(g 2(x),η 1(v)) \theta_1(x,v) = (\pi \times \eta_2)^{-1}(g_2(x), \eta_1(v))

Now

πθ 1(x,v)=π(π×η 2) 1(g 2(x),η 1(v))=g 2(x) \pi \theta_1(x,v) = \pi(\pi \times \eta_2)^{-1}(g_2(x),\eta_1(v)) = g_2(x)

so θ 1(x,v)T g 2(x)M\theta_1(x,v) \in T_{g_2(x)} M and thus (x,θ 1(x,v))g 2 *TM(x,\theta_1(x,v)) \in g_2^* T M for all (x,v)W 12(x,v) \in W_{1 2}. Then

η 2θ 1(x,v)=η 2(π×η 2) 1(g 2(x),η 1(v))=η 1(v) \eta_2 \theta_1(x,v) = \eta_2(\pi \times \eta_2)^{-1}(g_2(x),\eta_1(v)) = \eta_1(v)

so (g 1(x),η 2θ 1(x,v))=(g 1(x),η 1(v))(g_1(x),\eta_2\theta_1(x,v)) = (g_1(x),\eta_1(v)) which is in V 1V_1. Hence (x,θ 1(x,v))W 21(x,\theta_1(x,v)) \in W_{2 1}. Thus we have a map

ϕ 21:W 12W 21,ϕ 12(x,v)=(x,θ 1(x,v)) \phi_{2 1} \colon W_{1 2} \to W_{2 1}, \qquad \phi_{1 2}(x,v) = (x, \theta_1(x,v))

Similarly, we have a map ϕ 12\phi_{1 2} in the other direction. Both of these maps are smooth since they are smooth into N×TMN \times T M.

Let us consider ϕ 12ϕ 21(x,v)\phi_{1 2}\phi_{2 1}(x,v). Expanding this out yields:

ϕ 12ϕ 21(x,v) =ϕ 12(x,θ 1(x,v)) =(x,θ 2(x,θ 1(x,v))) =(x,(π×η 1) 1(g 1(x),η 2θ 1(x,v))) =(x,(π×η 1) 1(g 1(x),η 1(v))) =(x,(π×η 1) 1(π(v),η 1(v))) =(x,v) \begin{aligned} \phi_{1 2}\phi_{2 1}(x,v) &= \phi_{1 2}(x, \theta_1(x,v)) \\ &=(x, \theta_2(x, \theta_1(x,v))) \\ &=(x, (\pi \times \eta_1)^{-1}(g_1(x), \eta_2\theta_1(x,v))) \\ &=(x, (\pi \times \eta_1)^{-1}(g_1(x), \eta_1(v))) \\ &=(x, (\pi \times \eta_1)^{-1}(\pi(v),\eta_1(v))) \\ &=(x,v) \end{aligned}

where we have used the fact that (x,v)g 1 *TM(x,v) \in g_1^*T M so π(v)=g 1(x)\pi(v) = g_1(x). Thus ϕ 21\phi_{2 1} and ϕ 12\phi_{1 2} are inverses, whence they are diffeomorphisms.

Lemma

The transition function is fϕ 21ff \mapsto \phi_{2 1} \circ f.

Proof

Let us start with the domain and codomain of the transition function. The domain is {fE 1:Φ 1(f)U 2}\{f \in E_1 : \Phi_1(f) \in U_2\}. The set U 2U_2 consists of those functions h:NMh \colon N \to M such that (g 2,h)(g_2, h) takes values in V 2V_2. Thus Φ 1(f)U 2\Phi_1(f) \in U_2 if and only if (g 2,Φ 1(f))V 2(g_2, \Phi_1(f)) \in V_2. Since Φ 1(f)=η 1f^\Phi_1(f) = \eta_1 \circ \hat{f}, we see that for xNx \in N, vf^(x)TMv \coloneqq \hat{f}(x) \in T M must be such that (g 2(x),η 1(v))V 2(g_2(x), \eta_1(v)) \in V_2. This is precisely the condition that (x,v)(x,v) be in W 12W_{1 2}. Thus the domain of the transition function is the set of sections fE 1f \in E_1 such that f(x)W 12f(x) \in W_{1 2} for each xNx \in N.

The transition function, Ψ 21\Psi_{2 1}, is given by Ψ 21=Φ 2 1Φ 1\Psi_{2 1} = \Phi_2^{-1} \Phi_1. It is therefore completely characterised by the fact that Φ 2Ψ 21=Φ 1\Phi_2 \Psi_{2 1} = \Phi_1.

Let us consider Φ 2\Phi_2 applied to ψ 21f\psi_{2 1} \circ f for fE 1f \in E_1 such that ff takes values in W 12W_{1 2}. Expanding out the definition, we have:

(ψ 21f)(x) =ψ 21(f(x)) =(x,θ 1(x,f^(x))) =(x,(π×η 2) 1(g 2(x),η 1f^(x))) \begin{aligned} (\psi_{2 1} \circ f)(x) &= \psi_{2 1}(f(x)) \\ &= (x, \theta_1(x, \hat{f}(x))) \\ &= (x, (\pi \times \eta_2)^{-1}(g_2(x), \eta_1\hat{f}(x))) \end{aligned}

Now the result applying Φ 2\Phi_2 to hh is η 2h\eta_2 \circ h where h(x)=(x,h^(x))h(x) = (x, \hat{h}(x)). Thus the result of applying Φ 2\Phi_2 to ψ 21f\psi_{2 1} \circ f is the function

xη 2(π×η 2) 1(g 2(x),η 1f^(x))=η 1f^(x) x \mapsto \eta_2 (\pi \times \eta_2)^{-1}(g_2(x), \eta_1\hat{f}(x)) = \eta_1\hat{f}(x)

which is exactly the same function as Φ 1(f)\Phi_1(f). Hence

Ψ 21(f)=ψ 21f \Psi_{2 1}(f) = \psi_{2 1} \circ f

and thus Ψ 21\Psi_{2 1} is a diffeomorphism.

Conclusion

In conclusion we have

Corollary

The mapping space C (N,M;{Q i},{P i})C^\infty(N,M;\{Q_i\},\{P_i\}) of def. equipped with charts as discussed above is a smooth manifold.

Properties

Relation between diffeological and Fréchet manifold structure

Since smooth manifolds form a full subcategory of diffeological spaces, the mapping space C (X,Y)C^\infty(X,Y) between two manifolds always exists canonically as a diffeological space:

Definition

For UU \in CartSp, a smooth plot of C (X,Y) diffC^\infty(X,Y)_{diff} over UU is a smooth function U×XYU \times X \to Y (hence a UU-parameterized smooth collection of smooth functions XYX \to Y).

If XX is a compact manifold then there is also the structure of a Fréchet manifold C (X,Y) FrC^\infty(X,Y)_{Fr} on the mapping space. We discuss that and how these two smooth structures coincide.

Definition

Define a functor

ι:FrechetManifoldsDiffeologicalSpaces \iota \colon FrechetManifolds \to DiffeologicalSpaces

in the evident way by taking for XX a Fréchet manifold for any UU \in CartSp the set of UU-plots of ι(X)\iota(X) to be the set of smooth functions UXU \to X.

Proposition

The functor ι:FrechetManifoldsDiffeologicalSpaces\iota \colon FrechetManifolds \hookrightarrow DiffeologicalSpaces is a full and faithful functor.

This appears as (Losik, theorem 3.1.1).

Since moreover diffeological spaces are fully faithful in smooth sets (them being precisely the concrete objects in smooth sets), this implies in particular that Frechet manifolds are fully faithful in smooth sets

FrechetManifoldsDiffeologicalSpacesSmoothSets. FrechetManifolds \hookrightarrow DiffeologicalSpaces \hookrightarrow SmoothSets \,.
Proposition

Let X,YSMoothManifoldX, Y \in SMoothManifold with XX a compact manifold.

Then under this embedding, the diffeological mapping space structure C (X,Y) diffC^\infty(X,Y)_{diff} on the mapping space coincides with the Fréchet manifold structure C (X,Y) FrC^\infty(X,Y)_{Fr}:

ι(C (X,Y) Fr)C (X,Y) diff. \iota(C^\infty(X,Y)_{Fr}) \simeq C^\infty(X,Y)_{diff} \,.

This appears as (Waldorf, lemma A.1.7).

References

  • M. V. Losik, Categorical Differential Geometry Cah. Topol. Géom. Différ. Catég., 35(4):274–290, 1994.

  • Konrad Waldorf, Transgression to Loop Spaces and its Inverse I (arXiv:0911.3212)

  • Andrew Stacey, Yet more smooth mapping spaces and their smoothly local properties arXiv:1301.5493

Generalising the Lie group structure on the diffeomorphisms of a manifold, the case of non-compact orbifolds is in

The generalization to mapping stacks of differentiable stacks is discussed in

Last revised on September 26, 2023 at 13:11:48. See the history of this page for a list of all contributions to it.