We show that induced representations for a pair of \textit{diffeological Lie groups} exist, in the form of an indexed colimit in the category of diffeological spaces.
Let S^{G^{\infty}}(M) be the regular diffeological Lie group of super symplectic diffeomorphisms of a supermanifold M. S^{G^{\infty}}(M)is a Lie subgroup of the group of super diffeomorphisms D^{G^{\infty}}(M); further, there exist induced representations for Hilbert space representations for these two regular Lie groups.
Let X be a set. An n-parametrization of X is defined as a map f : U \rightarrow X, where U is an open set in \mathbb{R}^{n}, n\ge 0. A constant parametrization is a map of the form f : U\rightarrow X such that f(u) = x for all u \in U.
A diffeology \mathcal{D} on X is a collection of parametrizations of X satisfying the following axioms:
(D1) \mathcal{D} contains the constant parametrizations.
(D2) If p: U \rightarrow X is a parametrization such that \forall u \in U, there exists a neighborhood V of u such that p|_{V} is in \mathcal{D}, then p itself is in \mathcal{D}. In other words, the collection \mathcal{D} satisfies the axioms of a set-valued sheaf.
(D3) Given a parametrization p:U \rightarrow X in \mathcal{D}, if f: V \rightarrow U is a smooth map, then p \circ f is a parametrization in \mathcal{D}.
Let X be a set and \mathcal{F} be a collection of parametrizations on X. The diffeology on X induced by \mathcal{F} is the intersection of all diffeologies on X containing \mathcal{F}.
Let (X,\mathcal {F}) and (Y,\mathcal{G}) be diffeological spaces. A map f:X\rightarrow Y is smooth if f takes plots of the diffeology on X to plots of the diffeology on Y.
A difeological space (S,F) is lattice type (or L-type) when given any two plots f:M_1 \rightarrow S, g:M_2 \rightarrow S at a point f(x) = g(y), there exists a third plot h through which the germs of f at x and g at y factor, that is there exists a map h:N \rightarrow S with f = h\circ \varphi, g=h\circ \gamma where \varphi: \tilde{M_1} \rightarrow N and \gamma: \tilde{M_2} \rightarrow N are plots such that \tilde{M_i} \subset M_i, i=1,2 are neigborhoods of x and y respectively and \varphi(x) = \gamma(y).
A diffeological group G is called a diffeological Lie group when T_{e}G is a diffeological vector space such that
(DL1) For any nonzero vector \alpha \in T_{e}G, there exists a smooth real-valued map T:T_{e}G \rightarrow \mathbb{R} such that T(\alpha) \ne 0:
(DL2) Every plot of T_{e}G factors smoothly through a smooth linear map of a compact Hausdorff topological vector space into T_{e}G.
Let V and W be topological graded modules over \Gamma_{0}. A continuous map f : V \times \cdots \times V \rightarrow W is said to be an n-multimorphism if
(M1) f is n-multilinear with respect to the ground field \mathbb{R},
(M2) f(v_{1}, \dots v_{i}\gamma,v_{i+1}, \dots v_{n}) = f(v_{1}, \dots, v_{i}, \gamma v_{i+1}, \dots, v_{n}), where v_{i} \in V, 1\le i \le n and \gamma \in \Gamma_{0},
(M3) f(v_{1}\dots v_{n}\gamma) = f(v_{1},\dots, v_{n})\gamma, v_{i}\in V, 1 \le i \le n, \gamma \in \Gamma_{0}.
Let V and W be topological graded modules over \Gamma_{0} and let U\subseteq V be open in V. A function f: U \rightarrow W will be called supersmoooth or G^{\infty} when for every n\ge 1, there exist continuous maps which are regular k-multimorphisms in the k terminal variables for each fixed x \in U, denoted by D^{k}f(x;\cdots): U \times V \times \cdots \times V \rightarrow W, where k \le n, such that the map
satisfies the property that
is continuous at (0,h) in V.
\mathfrak{D} is closed under coproducts. For a collection \{X_{i}|i\in I\} of diffeological spaces, the coproduct of the collection is called the sum of the diffeological spaces X_{i} and is denoted \displaystyle \sqcup_{i\in I}X_{i} (Iglesias [Igl], 1.39).
\mathfrak{D} is closed under products. Just as in the above result, given an arbitrary collection \{X_{i}|i\in I\}, the product object is denoted \displaystyle \prod_{i\in I}X_{i} and can be given a canonical diffeology (Iglesias [Igl], 1.55).
\mathfrak{D} is a symmetric monoidal category.
The product map is derived from the definition of the product object. In addition, since \mathfrak{D} is essentially a concrete category, associativity of the product map holds. We can take any one-point set as a unit object, since all one-point sets are unique up to isomorphism of diffeological spaces (all plots are constant). The coherence axioms are clearly satisfied. The symmetry of \mathfrak{D} is given by the canonical map \gamma : X \times Y \rightarrow Y \times X, (x,y)\mapsto (y,x).
\mathfrak{D} is complete with respect to (all small) limits and colimits.
From Mac Lane ([Mac], Corollary V.2.2), we know that if a category \mathcal{C} has equalizers of all pairs of arrows and all small products, then \mathcal{C} is small-complete. In \mathfrak{D}, the equalizer of pairs of smooth maps f,g: X \rightarrow Y is defined to be the subspace Z of X such that f(x) = g(x) for x\in Z. The universal arrow is the canonical inclusion map. The coequalizer is defined for pairs f,g : X \rightarrow Y by considering the smallest equivalence relation E \subset Y \times Y containing all pairs (f(x), g(x)), for x \in X. The coequalizer is the quotient space Z = Y/E, with the universal arrow given by the canonical projection map. From the existence of equalizers, coequalizers, coproducts and products, we conclude that \mathfrak{D} is complete.
\mathfrak{D} is Cartesian closed.
The map \phi :\mathfrak{D}(X\times Y, Z) \rightarrow \mathfrak{D}(X, \mathfrak{D}(Y,Z)) given by (\phi(f)(x))(y) = f(x,y) for f: X\times Y \rightarrow Z is a bijection, and shows that the functor \mathfrak{D}(Y,-): \mathfrak{D} \rightarrow \mathfrak{D} is the right adjoint to the functor - \times Y:\mathfrak{D} \rightarrow \mathfrak{D}.
The properties of \mathfrak{D} as a category make it suitable as an enriching category. This gives us the ability to use the results of Kelly [Kel] on indexed limits, indexed colimits and Kan extensions in the case of enriched categories. These ideas are discussed in the following section.
D^{G^{\infty}}(M) is defined to be the group of super diffeomorphisms of M, where M is a supermanifold. D^{G^{\infty}}(M) can be made into a diffeological space by applying the function space diffeology: a map f: U \subseteq D^{G^{\infty}}(M) is a plot if the induced map F : U \times M \rightarrow M given by F(u,m) = (f(u))(m) is smooth for all u \in U. The smallest diffeology generated by such plots is called the function space diffeology on D^{G^{\infty}}(M).
([Twu] Definition 5.1.1) A Diffeological Category \mathcal{D} is a locally small category such that for any objects x,y in \mathcal{D}, the hom set \mathcal{D}(x,y) is a diffeological space which satisfies the following axiom: for all objects x,y and z in \mathcal{D} and elements g in \mathcal{D}(x,y) and f in \mathcal{D}(y,z), the composition f\circ g in \mathcal{D}(x,z) defines a smooth map:
where \mathcal{D}(y,z) \times \mathcal{D}(x,y) has the product diffeology.
Let F:\mathcal{J} \rightarrow \mathfrak{D} and G:\mathcal{J} \rightarrow \mathcal{C} be smooth functors of diffeological categories. Fix an object c of \mathcal{C}. An (F,c)-cylinder over G is a smooth natural transformation \alpha: F \dot{\rightarrow}\mathcal{C}(c,G-). It can be illustrated with the commutative diagram:
Let F:\mathcal{J}^{op} \rightarrow \mathfrak{D} and G: \mathcal{J} \rightarrow \mathcal{C} be smooth functors of diffeological categories. An (F,c)-cylinder under G is a natural transformation of smooth functors \alpha: F \dot{\rightarrow} \mathcal{C}(G-,c). Again, in the case F=\Delta 1, \alpha: \Delta1 \dot{\rightarrow} \mathcal{C}(G-,c) can be identified with a cone from the base G to c. The colimit of G indexed by F is the unique (up to isomorphism) object of \mathcal{C} such that every (F,c)- cylinder under G factors uniquely through the universal cylinder \mu: F\dot{\rightarrow} \mathcal{C}(G-,F*G). The commutative diagram illustrating the definition is given below.
where t: F\ast G \rightarrow c is the unique map. Hence \alpha_{i} = \mathcal{C}(1,t)\mu_{i}.
Let F,G : \mathcal{J} \rightarrow \mathfrak{D} be smooth functors. Then \{ F,G\} \simeq Nat(F,G), the diffeological space of natural transformations between the functors F and G.
We use the fact that the space of natural transformations between F and G can be expressed as an end: Nat(F,G) \simeq \int_c \mathfrak{D}(Fc, Gc). Using the language of ends, we obtain the following isomorphisms:
Therefore \{F,G\} is isomorphic to Nat(F,G).
Let F: \mathcal{J} \rightarrow \mathfrak{D} and G: \mathcal{J} \rightarrow \mathcal{C} be smooth functors, where \mathcal{C} is a diffeological category. For any object c of \mathcal{C}, we have the following isomorphism of diffeological spaces:
\mathcal{C}(c,G-) is a smooth functor from \mathcal{J} to \mathfrak{D}, so applying the previous proposition (4.4) and the definition of indexed limit (?), we have
Let F \mathcal{J}^{op} \rightarrow \mathfrak{D} and G: \mathcal{J} \rightarrow \mathcal{C} be smooth functors, where \mathcal{C} is a diffeological category. For any object c of \mathcal{C}, we have the following isomorphism of diffeological spaces:
\mathcal{C}(G-,c):\mathcal{J}^{op} \rightarrow \mathfrak{D} is a smooth functor, so applying the definition of indexed colimits and Proposition 4.4 we get
Let F: \mathcal{C} \rightarrow \mathfrak{D} and G: \mathcal{C} \rightarrow \mathfrak{D} be smooth functors. Then Nat(F,G) \simeq \text{Lim\text{ }}(elF\xrightarrow{Q}\mathcal{C} \xrightarrow{G} \mathfrak{D}) as diffeological spaces.
The limiting cone for the composite functor elF\xrightarrow{Q}\mathcal{C} \xrightarrow{G} \mathfrak{D} consists of objects \alpha_{(c,x)}: \star \rightarrow Gc, where \star denotes a 1-point set. For a map f^{*}: (c,x) \rightarrow (c^{\prime},x^{\prime}), Gf \alpha_{(c,x)} = \alpha_{(c^{\prime},x^{\prime})}, since \alpha is a cone. Define a natural transformation \beta: F\dot{\rightarrow}G by \beta_{c}(x) := \alpha_{(c,x)}. For f^{*}: (c,x) \rightarrow (c^{\prime},x^{\prime}), we have:
giving the following commutative diagram:
Hence Nat(F,G) \simeq \text{Lim\text{ }}(elF\xrightarrow{Q}\mathcal{C} \xrightarrow{G} \mathfrak{D}) \simeq \{F,G\}.
Let G: \mathcal{C} \rightarrow \mathcal{B} be a smooth functor of diffeological categories. Then given a smooth index functor F: \mathcal{C} \rightarrow \mathfrak{D},
Equation (?) gives F \simeq \text{Colim\text{ }} \left(elF \xrightarrow{Q} \mathcal{C} \xrightarrow{y^{-1}} \mathfrak{D}^{\mathcal{C}}\right) \simeq \text{Colim\text{ }} \left(C(c,-): el F \rightarrow \mathfrak{D}^{\mathcal{C}}\right). Therefore:
where the canonical limit is indexed by the category el F of elements of F. On the other hand, \text{Lim\text{ }}(elF \xrightarrow{Q} \mathcal{C} \xrightarrow{G} \mathcal{B}) = \text{Lim\text{ }}_{(c \in el F)}{Gc}. This establishes the isomorphism.
As before, Theorem (?) gives a representation of F, this time as F \simeq \text{Colim\text{ }} \mathcal{C}(-,c) since F is a contravariant functor. Therefore
The other side of the isomorphism is \text{Colim\text{ }}\left((elF)^{op} \xrightarrow{Q^{op}} \mathcal{C} \xrightarrow{G} \mathcal{B} \right) = \text{Colim\text{ }}_{(c \in elF)}Gc. The isomorphism is now established.
Corollary (4.8) gives \{F,T\} \simeq \text{Lim\text{ }}\left[el F \xlongrightarrow{Q} \mathcal{M} \xlongrightarrow{T} \mathcal{A} \right]. Defining F as \mathcal{C}(c,K-) gives el F = (c\downarrow K) as shown above. Hence the result holds.
The right Kan extension of the smooth functor T: \mathcal{M} \rightarrow \mathcal{A} along K: \mathcal{M} \rightarrow \mathcal{C} is the functor R: \mathcal{C} \rightarrow \mathcal{A}, given by Rc = \{\mathcal{C}(c,K-),T \}, whenever it exists.
We first construct a smooth natural transformation \mu: RK \dot{\longrightarrow} T. Since Rc = \{\mathcal{C}(c,K-),T\} is the limit of T indexed by \mathcal{C}(c,K-), equation (?) from definition (4.2) gives a universal arrow \lambda_{c,-}: \mathcal{C}(c,K-) \dot{\longrightarrow} \mathcal{A}(Rc,T-). Suppose c = Km for some object m of \mathcal{M}. This gives the transformation \lambda_{Km,-}:\mathcal{C}(Km,K-) \dot{\longrightarrow} \mathcal{A}(RKm,T-). Choosing the second component of \lambda to be m gives
Now define \varepsilon_{m}: RKm \rightarrow Tm as \varepsilon_{m} = \lambda_{Km,m}(1_{Km}), where 1_{Km} is the identity map in \mathcal{C}(Km, Km). We need to show that \varepsilon_{m} forms the component of a smooth natural transformation \varepsilon: RK \dot{\longrightarrow} T. Let g: m \rightarrow n be an arrow in \mathcal{M}. We have the following commutative diagram:
from which we deduce clearly that 1_{Kn}\circ Kg = Kg \circ 1_{Km}.
Now consider the following commutative diagrams
Both the upper and lower squares commute, since \lambda is a universal arrow.
The result 1_{Kn}\circ Kg = Kg \circ 1_{Km} enables us to conclude from the commutative diagrams that Tg \circ \varepsilon_{m}= RKg \circ \varepsilon_{n}, establishing the natural transformation \varepsilon: RK \dot{\longrightarrow}T.
Now suppose S:\mathcal{C} \longrightarrow \mathcal{A} is a smooth functor with \sigma: SK \dot{\longrightarrow} T a smooth natural transformation. We show that \sigma uniquely factors through \varepsilon by means of the adjunction in equation (?). We shall use Kelly’s approach by employing the calculus of ends and coends. We have:
Applying equation (?) in the case where F =\mathcal{C}(c,K-) and G = T, we have
Writing the above equation as an end, we get
Applying the Fubini Theorem,
The isomorphism Nat(S,R) \simeq Nat(SK,T) gives by Yoneda’s Proposition ([Mac], Proposition III.2.1) that the natural transformation \varepsilon :RK \dot{\longrightarrow} T is universal. Hence (R, \varepsilon) is a right Kan extension of the smooth functor T.
From proposition (4.9), we have for F = \mathcal{C}(K-,c) and G=T: \mathcal{M} \longrightarrow \mathcal{A} the result
Since (el F) = (K \downarrow c), the result holds.
Let K: \mathcal{M} \longrightarrow \mathcal{C} and T: \mathcal{M} \longrightarrow \mathcal{A} be smooth functors of diffeological categories. The left Kan extension of T along K is given by the functor L: \mathcal{C}\rightarrow \mathcal{A}, given by Lc = \mathcal{C}(K-,c) * T, whenever it exists.
From definition (4.2), there exits a universal arrow \lambda_{-,c}: \mathcal{C}(K-,c) \dot{\longrightarrow} \mathcal{A}(T-,Lc). Suppose c = Km for some object m of \mathcal{M}. We obtain the transformation \lambda_{-,Km}: \mathcal{C}(K-,Km) \dot{\longrightarrow} \mathcal{A}(T-,LKm). Choosing the first component of \lambda to be m, we get
Define \varepsilon: T \dot{\longrightarrow} LK by \varepsilon_{m}= \lambda_{m,Km}(1_{Km}): Tm \longrightarrow LKm. Just like in the proof for the case of the right Kan extension, we show that this definition gives a universal natural transformation. Let g:m\rightarrow n be an arrow in \mathcal{M}. We have the equality 1_{Kn} \circ Kg = Kg \circ 1_{Km}. Consider the following commutative diagrams:
Both the upper and lower squares commute, since \lambda is a universal arrow. Splitting the diagram into two components and evaluating the identity map for each component gives
From the result 1_{Kn}\circ Kg = Kg \circ 1_{Km} we get Tg \circ \varepsilon_{n}= LKg \circ \varepsilon_{m}, establishing the natural transformation \varepsilon: T \dot{\longrightarrow}LK. This gives the diagram
Finally, we show that \varepsilon: T \dot{\longrightarrow} LK is unique up to natural isomorphism. Let S: \mathcal{C} \longrightarrow \mathcal{A} be a smooth functor with \sigma: T \dot{\longrightarrow} SK a natural transformation. We have:
by the definition of Lc. Applying equation (4.9), we have
Applying the Fubini theorem, we get
Hence by Yoneda’s Proposition ([Mac], Proposition III.2.1), the isomorphism Nat(L,S) \simeq Nat(T, SK) gives the result that the natural transformation \varepsilon: T \dot{\longrightarrow} LK is universal, establishing that (L,\varepsilon) is a left Kan extension of T along K.
The orthogonal complement of R, R^{\bot} = F \ast G.
By the definition of R, the orthogonal projection of {F\alpha (f_{j})\cdot g_{i}} - {f_{j}\cdot G\alpha( g_{i})} gives the zero vector in R^{\bot}.
We define the universal cylinder \lambda: F \dot{\longrightarrow} \mathcal{C}(G-, R^{\bot}) by means of the projection Fi \times Gi \hookrightarrow V \xlongrightarrow{\pi} R^{\bot} , where the projection is given by the composition
where the sum is taken over all i in \mathcal{J} and g_{i} in G_{i}. This gives \lambda_{i}: Fi \rightarrow \mathcal{C}(Gi, R^{\bot}) as follows: fix an element f_{i} of Fi. We have the map
We show that \lambda_{i}(f_{i}) is an arrow of \mathcal{C}. The norm of \lambda_{i}(f_{i}) is given by
since \pi is an orthogonal projection and therefore has norm ||\pi|| \le 1 and ||f_{i}\cdot g_{i}|| = ||g_{i}||.
The naturality of \lambda is given by the following diagram:
All the maps in the diagram are smooth maps of diffeological spaces.
We show that the diagram commutes. Let f_{j} be an element of the diffeological space Fj. We have:
Therefore the diagram commutes and \lambda: F \dot{\longrightarrow} \mathcal{C}(G-,R^{\bot}) is a cylinder. To show that \lambda: F\dot{\longrightarrow} \mathcal{C}(G-, R^{\bot}) is a universal cylinder, suppose we have another (F,c)-cylinder \beta: F \dot{\longrightarrow} (G-,c). This gives the commutative diagram illustrated below.
Therefore for each f_{j} in Fj, \beta_{i}(F\alpha (f_{j}))(g_{i})=\beta_{i}(f_{j})(G\alpha (g_{i})).
The unique map T: R^{\bot} \rightarrow c is defined as T\left[\pi\left(f_{i}\cdot g_{i}\right) \right] = \left(\beta_{i}(f_{i})\right)(g_{i}). We show that T is well defined by proving that the generators of R under the action of T give the zero vector in c. Pick a generator F\alpha (f_{j})\cdot g_{i} - f_{j}\cdot G\alpha( g_{i}) in R. Then \pi(F\alpha (f_{j})\cdot g_{i} - f_{j}\cdot G\alpha( g_{i})) = \pi(F\alpha (f_{j})\cdot g_{i}) - \pi({f_{j}\cdot G\alpha( g_{i})}) = 0 in R^{\bot}. We have
by the commutative diagram (?). Hence for all pairs (f_{i},g_{i}) such that (by means of the inclusion map) f_{i}\cdot g_{i} is in R, (\beta_{i}f_{i})= 0 in c.
Now we compute the norm of T. We have the following commutative diagrams:
We have
Fix a pair (f_{i},g_{i}) in Fi \times Gi. We have that ||\pi(f_{i}\cdot g_{i})|| \le ||f_{i}\cdot g_{i}||. If ||\pi(f_{i}\cdot g_{i})|| = ||f_{i}\cdot g_{i}||, then we have
Now if ||\pi(f_{i}\cdot g_{i})|| < ||f_{i}\cdot g_{i}||, then we can decompose f_{i}\cdot g_{i} in V into the components in R and R^{\bot} to get f_{i}\cdot g_{i} = f_{i}\cdot g_{i}^{\prime} + f_{i}\cdot g_{i}^{\prime\prime}, where f_{i}\cdot g_{i}^{\prime} is in R and f_{i}\cdot g_{i}^{\prime\prime} is in R^{\bot}. We can express f_{i}\cdot g_{i}^{\prime} as f_{i}\cdot g_{i}^{\prime} = F\alpha (h_{j})\cdot g_{i}^{\prime} - h_{j} \cdot G\alpha(g_{i}^{\prime}), where h_{j} is in F_{j} and \alpha is an arrow in \mathcal{J}. Since \beta is a (F,c)-cylinder under G, \beta_{i}(f_{i})g_{i} = \beta_{i}\left[F\alpha (h_{j})\cdot g_{i}^{\prime} - h_{j} \cdot G\alpha(g_{i}^{\prime}) \right]=0. Hence we have the following results:
Therefore, in both cases we have
Hence T is an arrow in \mathcal{C} and \lambda, as defined is a universal cylinder. We conclude that R^{\bot} = F \ast G.
Given a representation T: \mathcal{H} \rightarrow \mathcal{C}, define the inclusion functor I : \mathcal{H} \rightarrow \mathcal{G}. The functor T : \mathcal{H} \rightarrow \mathcal{C} is a representation of \mathcal{H}. The left Kan extension L : \mathcal{G} \rightarrow C takes the form L(\mathcal{G}) = \textrm{Hom}(I-,G) \ast T), which exists by the proof of the previous theorem, and the completeness of the category \mathfrak{D}. Given a representation S: \mathcal{G} \rightarrow \mathcal{C}, the composite SI is the restriction of the representation to \mathcal{H}.
The adjunction can therefore be expressed Nat(L,S) \simeq Nat(T, SI). Let V = T(\mathcal{H}) be an \mathcal{H}-module, and W a \mathcal{G}-module. Denoting L(G) by I\cdot V, we obtain the isomorphism (of diffeological spaces)
establishing that I\cdot V is the (co) induced representation of \mathcal{G}.
The definition of diffeological category raises a question; are there examples of categories with the objects not necessarily diffeological spaces, such that the hom-sets have the structure of a diffeological space? It seems that the underlying structures would need to have the structure of a diffeological space, so we can use the function space diffeology. This means that the category \mathfrak{D} is its own higher-category: for G, H diffeological spaces, the morphism \mathfrak{D}(G,H) is also a diffeological space. We can extend this to 2-morphisms and so on.
Milnor discusses the issue of regularity of (infinite dimensional) Lie groups in [Mil]. Essentially, a Lie group is regular if there exists an exponential map from the Lie algebra to the Lie group, relating representations of both the algebra and the group. In this paper, we have discussed diffeological Lie groups, and it is natural to ask whether they are regular. J-P. Magnot [Mag] has shown that this is not the case, so not all diffeological Lie groups are regular. This leads to the question whether all regular diffeological Lie groups can be classified.
The text on this page is from arXiv:2108.00412v2 by Joshua A. Leslie and Ralph A. Twum, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/2108/2108.00412.jsonl.