ScalingStacks

8.1. Action on ends of curves

8.1.1. Definition

Let ξ:𝐑>0→Z\xi:{\mathbf{R}}_{>0}\to Z be an injective morphism of curves, where 𝐑>0{\mathbf{R}}_{>0} is viewed as an unoriented curve. Let MM be a subset of Z∖ξ⁡(𝐑≥1)Z\setminus\xi({\mathbf{R}}_{\geq 1}).

We say that ξ\xi is terminal for (Z,M)(Z,M) if the following two conditions hold:

  • •

    given an admissible homotopy class of paths ζ\zeta in ZZ with endpoints in MM, there is an admissible path γ\gamma in ζ\zeta contained in Z∖ξ⁡(𝐑≥1)Z\setminus\xi({\mathbf{R}}_{\geq 1})

  • •

    there is no admissible path in Z∖{ξ⁡(1)}Z\setminus\{\xi(1)\} from a point of MM to ξ⁡(2)\xi(2).

Note that ξ\xi is terminal for (Z,M)(Z,M) if and only if ξ\xi is terminal for (Z⁡(ξ),Z⁡(ξ)∩M)(Z(\xi),Z(\xi)\cap M), where Z⁡(ξ)Z(\xi) is the component of ZZ containing ξ⁡(𝐑>0)\xi({\mathbf{R}}_{>0}).

We say that ξ\xi is outgoing for ZZ if ξ⁡(𝐑≥1)\xi({\mathbf{R}}_{\geq 1}) is closed in ZZ. Note that if ξ\xi is outgoing for ZZ then it is terminal for (Z,M)(Z,M) for any M⊂Z∖ξ⁡(𝐑≥1)M\subset Z\setminus\xi({\mathbf{R}}_{\geq 1}).

0PC5

Remark 8.1.1. Assume ξ\xi is not outgoing for ZZ and let z0∈Zz_{0}\in Z such that ξ⁡(𝐑≥1)¯∖ξ⁡(𝐑≥1)={z0}\overline{\xi({\mathbf{R}}_{\geq 1})}\setminus\xi({\mathbf{R}}_{\geq 1})=\{z_{0}\}. Note that ξ\xi is outgoing for Z∖{z0}Z\setminus\{z_{0}\}. The map ξ\xi is terminal for (Z,M)(Z,M) if and only if z0∉Mz_{0}{\not\in M} and the inclusion induces an isomorphism Hom𝒜∙​(Z∖{z0},1)⁡(m,z)→∼Hom𝒜∙​(Z,1)⁡(m,z)\operatorname{Hom}\nolimits_{{\mathcal{A}}^{\bullet}(Z\setminus\{z_{0}\},1)}(m,z)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}\operatorname{Hom}\nolimits_{{\mathcal{A}}^{\bullet}(Z,1)}(m,z) for all m∈Mm\in M and z∈M∪{ξ⁡(1)}z\in M\cup\{\xi(1)\}.

We assume now that ξ\xi is terminal for (Z,M)(Z,M). Thanks to Lemma 7.4.36, we have a differential pointed functor

L∙=Lξ∙:𝒮M∙​(Z)×𝒮M∙​(Z)opp×𝒰∙→diffL^{\bullet}=L_{\xi}^{\bullet}:{\mathcal{S}}^{\bullet}_{M}(Z)\times{\mathcal{S}}^{\bullet}_{M}(Z)^{\operatorname{opp}\nolimits}\times{\mathcal{U}}^{\bullet}\to\operatorname{diff}\nolimits
L∙​(T,S,en)=Hom𝒮∙​(Z)⁡(S,T⊔{ξ⁡(1),…,ξ⁡(n)})L^{\bullet}(T,S,e^{n})=\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S,T\sqcup\{\xi(1),\ldots,\xi(n)\})
L∙​(β,α,σ)​(f)=(β⊠ξ⁡(σ))⋅f⋅α∈L∙​(T′,S′,n)L^{\bullet}(\beta,\alpha,\sigma)(f)=(\beta\boxtimes\xi(\sigma))\cdot f\cdot\alpha\in L^{\bullet}(T^{\prime},S^{\prime},n)

for α∈Hom𝒮∙​(Z)⁡(S′,S)\alpha\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S^{\prime},S), β∈Hom𝒮∙​(Z)⁡(T,T′)\beta\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(T,T^{\prime}), σ∈End𝒰∙⁡(en)\sigma\in\operatorname{End}\nolimits_{{\mathcal{U}}^{\bullet}}(e^{n}), and f∈L∙​(T,S,n)f\in L^{\bullet}(T,S,n). We have used the strands realization of 𝒰∙{\mathcal{U}}^{\bullet} given by Theorem 7.4.34.

We put L∙​(T,S)=L∙​(T,S,e)L^{\bullet}(T,S)=L^{\bullet}(T,S,e). As usual, we put Lξ=𝐅2​[Lξ∙]L_{\xi}={\mathbf{F}}_{2}[L_{\xi}^{\bullet}].

The naturality in the next lemma is immediate as in Lemma 7.4.36.

0PC6

Lemma 8.1.2. Given S⊂MS\subset M and n≥0n\geq 0, there is an isomorphism of functors 𝒮M∙​(Z)→Sets∙{\mathcal{S}}^{\bullet}_{M}(Z)\to\operatorname{Sets}\nolimits^{\bullet} (forgetting the differential)

⋁S′⊂S|S′|=nHom𝒮∙​(Z)⁡(S′,{ξ⁡(1),…,ξ⁡(n)})∧Hom𝒮∙​(Z)⁡(S∖S′,−)\displaystyle\bigvee_{\begin{subarray}{c}S^{\prime}\subset S\\ |S^{\prime}|=n\end{subarray}}\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S^{\prime},\{\xi(1),\ldots,\xi(n)\})\wedge\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S\setminus S^{\prime},-) →∼L∙​(−,S,en)\displaystyle\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L^{\bullet}(-,S,e^{n})
(α,β)\displaystyle(\alpha,\beta) ↦α⊠β.\displaystyle\mapsto\alpha\boxtimes\beta.

Lemma 8.1.2 shows that there is an isomorphism of functors, functorial in SS and TT

L∙​(T,−,en)∧L∙​(−,S,em)\displaystyle L^{\bullet}(T,-,e^{n})\wedge L^{\bullet}(-,S,e^{m}) →∼L∙​(T,S,en+m)\displaystyle\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L^{\bullet}(T,S,e^{n+m})
(α,β)\displaystyle(\alpha,\beta) ↦(α⊠ξ([r→n+r]1≤r≤m))⋅β.\displaystyle\mapsto(\alpha\boxtimes\xi([r\to n+r]_{1\leq r\leq m}))\cdot\beta.

The functor E=Eξ=L∙​(−,−)E=E_{\xi}=L^{\bullet}(-,-) gives a bimodule 22-representation on 𝒮M∙​(Z){\mathcal{S}}^{\bullet}_{M}(Z). The endomorphism τ\tau of L∙​(−,−,e2)L^{\bullet}(-,-,e^{2}) is given by the non-identity non-zero braid {1,2}→{1,2}\{1,2\}\to\{1,2\}.

We have obtained the following proposition.

0PC7

Proposition 8.1.3. The bimodule EE and the endomorphism τ\tau define a bimodule 22-representation on 𝒮M∙​(Z){\mathcal{S}}^{\bullet}_{M}(Z) and on 𝒮M​(Z){\mathcal{S}}_{M}(Z).

Lemma 8.1.2 shows that Lξ​(−,−)L_{\xi}(-,-) is left finite.

0PC8

Remark 8.1.4. Proposition 8.1.3 generalizes and make more precise a result of Douglas and Manolescu [DouMa, §5.2].

Let (𝒵,𝐚)({\mathcal{Z}},{\mathbf{a}}) be a chord diagram where 𝒵=[0,1]{\mathcal{Z}}=[0,1]. Let Z~′=(0,∞)\tilde{Z}^{\prime}=(0,\infty), viewed as a curve with Z~o′=Z~=(0,1)\tilde{Z}^{\prime}_{o}=\tilde{Z}=(0,1) (with its usual orientation). We extend the equivalence relation from Z~\tilde{Z} to Z~′\tilde{Z}^{\prime} by having all points of [1,∞)[1,\infty) alone in their class. Let Z′=Z~′/∼Z^{\prime}=\tilde{Z}^{\prime}/\!\sim. We have Zo′=ZoZ^{\prime}_{o}=Z_{o}. Let M=Ze​x​c′M=Z^{\prime}_{exc} be the image of 𝐚{\mathbf{a}} in Z′Z^{\prime}. Let ξ:𝐑>0→Z′,x↦x+1\xi:{\mathbf{R}}_{>0}\to Z^{\prime},\ x\mapsto x+1. Note that ξ\xi is outgoing for Z′Z^{\prime}.

The lax 22-representation underlying the 22-representation on 𝒮M​(Z)=𝒮M​(Z′){\mathcal{S}}_{M}(Z)={\mathcal{S}}_{M}(Z^{\prime}) provided by Proposition 8.1.3 is the “bottom algebra module” constructed by Douglas and Manolescu, via the identification of §5.7.

0PC9

Example 8.1.5. The left picture below gives an example where ξ\xi is terminal for (Z,M)(Z,M) but not outgoing for ZZ. The right picture is an example where ξ\xi is outgoing for ZZ.

[Uncaptioned image]

The picture below considers the case of a curve quotient of the disjoint union of an interval and a circle, with an outgoing ξ\xi at an end of the interval. The middle picture describes an element of Lξ∙​(−,−,e2)L^{\bullet}_{\xi}(-,-,e^{2}). The rightmost picture provides a different graphical representation of that element: the interval ξ⁡(𝐑≥1)\xi({\mathbf{R}}_{\geq 1}) has been moved to the bottom horizontal line.

[Uncaptioned image]

The next remark discusses the dependence of Lξ∙L_{\xi}^{\bullet} on ξ\xi.

0PCA

Remark 8.1.6. Assume ξ\xi is terminal for (Z,M)(Z,M). Consider f:Z→∼Zf:Z\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}Z an isomorphism of curves fixing MM. Note that f∘ξf\circ\xi is terminal for (Z,M)(Z,M) and the map ff induces an isomorphism Lξ∙→∼Lf∘ξ∙L_{\xi}^{\bullet}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L_{f\circ\xi}^{\bullet}.

Consider now another injective morphism of curves ξ′:𝐑>0→Z\xi^{\prime}:{\mathbf{R}}_{>0}\to Z such that ξ′\xi^{\prime} is terminal for (Z,M)(Z,M). Assume there is a connected open subset UU of ZuZ_{u} containing ξ⁡(𝐑>0)¯\overline{\xi({\mathbf{R}}_{>0})} and ξ′​(𝐑>0)¯\overline{\xi^{\prime}({\mathbf{R}}_{>0})} and assume the canonical orientations on ξ⁡(𝐑>0)\xi({\mathbf{R}}_{>0}) and ξ′​(𝐑>0)\xi^{\prime}({\mathbf{R}}_{>0}) extend to an orientation of UU. There is an isomorphism of curves f:Z→∼Zf:Z\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}Z fixing Z∖UZ\setminus U such that ξ′=f∘ξ\xi^{\prime}=f\circ\xi. It induces an isomorphism Lξ∙→∼Lξ′∙L_{\xi}^{\bullet}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L_{\xi^{\prime}}^{\bullet}, and that isomorphism does not depend on the choice of ff.

8.1.2. Approximation

Assume ξ−1​(M)\xi^{-1}(M) has no maximum. Fix an increasing sequence m0,m1,…m_{0},m_{1},\ldots of points of (0,1)(0,1) with ξ⁡(mi)∈M\xi(m_{i})\in M for all ii and with limimi>t\lim_{i}m_{i}>t for all t∈ξ−1​(M)t\in\xi^{-1}(M).

Fix n≥0n\geq 0 and define the braid βr:{mr,…,mr+n−1}→{1,…,n}\beta_{r}:\{m_{r},\ldots,m_{r+n-1}\}\to\{1,\ldots,n\} of 𝐑>0{\mathbf{R}}_{>0} by (βr)mr+i=[mr+i→i+1](\beta_{r})_{m_{r+i}}=[m_{r+i}\to i+1].

Let SS and TT be two finite subsets of MM. Consider rr such that mr>ξ−1​(t)m_{r}>\xi^{-1}(t) for all t∈T∩ξ⁡(𝐑>0)t\in T\cap\xi({\mathbf{R}}_{>0}). There is an isomorphism

Hom𝒮∙​(Z)⁡(S,T⊔ξ⁡({mr,…,mr+n−1}))→∼L∙​(T,S,en),α↦(idT⊠ξ⁡(βr))⋅α.\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S,T\sqcup\xi(\{m_{r},\ldots,m_{r+n-1}\}))\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L^{\bullet}(T,S,e^{n}),\ \alpha\mapsto(\operatorname{id}\nolimits_{T}\boxtimes\xi(\beta_{r}))\cdot\alpha.

It follows that there are isomorphisms functorial in SS and TT

(8.1.1) colimr→∞⁡Hom𝒮M∙​(Z)​(S,T⊔ξ⁡({mr,…,mr+n−1}))→∼L∙​(T,S,en).\operatorname{colim}\nolimits_{r\to\infty}\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}_{M}(Z)}(S,T\sqcup\xi(\{m_{r},\ldots,m_{r+n-1}\}))\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L^{\bullet}(T,S,e^{n}).

Here, the colimit is taken over the invertible maps ξ⁡(θr)\xi(\theta_{r}), where θr:{mr,…,mr+n−1}→{mr+1,…,mr+n}\theta_{r}:\{m_{r},\ldots,m_{r+n-1}\}\to\{m_{r+1},\ldots,m_{r+n}\} is the braid in 𝐑>0{\mathbf{R}}_{>0} given by (θr)ms=[ms→ms+1](\theta_{r})_{m_{s}}=[m_{s}\to m_{s+1}].

We deduce that T↦(S↦L∙​(T,S,en))T\mapsto(S\mapsto L^{\bullet}(T,S,e^{n})) is isomorphic to the functor

𝒮M∙​(Z)→𝒮M∙​(Z)​−diff,T↦colimr→∞⁡T⊔ξ⁡({mr,…,mr+n−1}).{\mathcal{S}}^{\bullet}_{M}(Z)\to{\mathcal{S}}^{\bullet}_{M}(Z)\operatorname{\!-diff}\nolimits,\ T\mapsto\operatorname{colim}\nolimits_{r\to\infty}T\sqcup\xi(\{m_{r},\ldots,m_{r+n-1}\}).

8.1.3. 22-representations and morphisms of curves

Let f:Z→Z′f:Z\to Z^{\prime} be a morphism of curves. Assume ξ\xi is terminal for (Z,M)(Z,M) and f∘ξf\circ\xi is terminal for (Z′,f⁡(M))(Z^{\prime},f(M)).

Assume that |f−1​(f​(z))|=1|f^{-1}(f(z))|=1 for all z∈Mz\in M. Let MfM_{f} be the (𝒮M∙​(Z),𝒮f⁡(M)∙​(Z′))({\mathcal{S}}^{\bullet}_{M}(Z),{\mathcal{S}}^{\bullet}_{f(M)}(Z^{\prime}))-bimodule corresponding to ff, i.e. given by Mf​(S,S′)=Hom𝒮∙​(Z′)⁡(S′,f⁡(S))M_{f}(S,S^{\prime})=\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z^{\prime})}(S^{\prime},f(S)). There is a morphism of functors Eξ∧𝒮M∙​(Z)Mf→Mf∧𝒮f⁡(M)∙​(Z′)Ef∘ξE_{\xi}\wedge_{{\mathcal{S}}^{\bullet}_{M}(Z)}M_{f}\to M_{f}\wedge_{{\mathcal{S}}^{\bullet}_{f(M)}(Z^{\prime})}E_{f\circ\xi} defined as making the following diagram commutative

Hom𝒮∙​(Z)⁡(−,T⊔{ξ⁡(1)})∧Hom𝒮∙​(Z′)⁡(S′,f⁡(−))\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(-,T\sqcup\{\xi(1)\})\wedge\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z^{\prime})}(S^{\prime},f(-))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}β∧α′↦f⁡(β)⋅α′\scriptstyle{\beta\wedge\alpha^{\prime}\mapsto f(\beta)\cdot\alpha^{\prime}}Hom𝒮∙​(Z′)⁡(S′,f⁡(T)⊔{f∘ξ⁡(1)})\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z^{\prime})}(S^{\prime},f(T)\sqcup\{f\circ\xi(1)\})}Hom𝒮∙​(Z′)(−,f(T))∧Hom𝒮∙​(Z′)(S′,−⊔{f∘ξ(1)})\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z^{\prime})}(-,f(T))\wedge\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z^{\prime})}(S^{\prime},-\sqcup\{f\circ\xi(1)\})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}β′∧α′↦(β′⊠idf∘ξ⁡(1))⋅α′\scriptstyle{\ \ \ \ \ \ \ \ \ \beta^{\prime}\wedge\alpha^{\prime}\mapsto(\beta^{\prime}\boxtimes\operatorname{id}\nolimits_{f\circ\xi(1)})\cdot\alpha^{\prime}}

The following lemma is a consequence of (8.1.1).

0PCB

Lemma 8.1.7. If ξ−1​(M)\xi^{-1}(M) has no maximum, then the construction above gives an isomorphism

Eξ∧𝒮M∙​(Z)Mf→∼Mf∧𝒮f⁡(M)∙​(Z′)Ef∘ξ,E_{\xi}\wedge_{{\mathcal{S}}^{\bullet}_{M}(Z)}M_{f}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}M_{f}\wedge_{{\mathcal{S}}^{\bullet}_{f(M)}(Z^{\prime})}E_{f\circ\xi},

and ff provides a morphism of bimodule 22-representations Lf∘ξ∙→Lξ∙L^{\bullet}_{f\circ\xi}\to L^{\bullet}_{\xi}.

We consider now an arbitrary MM but we assume that ff is strict. Let Mf#M_{f^{\#}} be the (𝒮f⁡(M)​(Z′),𝒮M​(Z))({\mathcal{S}}_{f(M)}(Z^{\prime}),{\mathcal{S}}_{M}(Z))-bimodule corresponding to f#f^{\#}, i.e. given by

Mf#(S′,S)=⨁p:S′→Z,f∘p=idS′Hom𝒮M​(Z)(S,p(S′)).M_{f^{\#}}(S^{\prime},S)=\bigoplus_{p:S^{\prime}\to Z,\ f\circ p=\operatorname{id}\nolimits_{S^{\prime}}}\operatorname{Hom}\nolimits_{{\mathcal{S}}_{M}(Z)}(S,p(S^{\prime})).

There is a morphism of functors Ef∘ξ⊗𝒮f⁡(M)​(Z′)Mf#→Mf#⊗𝒮M​(Z)EξE_{f\circ\xi}\otimes_{{\mathcal{S}}_{f(M)}(Z^{\prime})}M_{f^{\#}}\to M_{f^{\#}}\otimes_{{\mathcal{S}}_{M}(Z)}E_{\xi} defined as making the following diagram commutative

⨁p:−→Zf∘p=idHom𝒮⁡(Z′)(−,T′⊔{f∘ξ(1)})⊗Hom𝒮⁡(Z)(S,p(−))\textstyle{{\displaystyle\bigoplus_{\begin{subarray}{c}p:-\to Z\\ f\circ p=\operatorname{id}\nolimits\end{subarray}}}\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z^{\prime})}(-,T^{\prime}\sqcup\{f\circ\xi(1)\})\otimes\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S,p(-))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}β′∧α↦f#​(β′)⋅α\scriptstyle{\beta^{\prime}\wedge\alpha\mapsto f^{\#}(\beta^{\prime})\cdot\alpha}⨁p:T′→Zf∘p=idT′Hom𝒮⁡(Z)(S,p(T′)⊔{ξ(1)})\textstyle{{\displaystyle\bigoplus_{\begin{subarray}{c}p:T^{\prime}\to Z\\ f\circ p=\operatorname{id}\nolimits_{T^{\prime}}\end{subarray}}}\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S,p(T^{\prime})\sqcup\{\xi(1)\})}⨁p:T′→Zf∘p=idT′Hom𝒮⁡(Z)(−,p(T′))⊗Hom𝒮⁡(Z)(S,−⊔{ξ(1)})\textstyle{{\displaystyle\bigoplus_{\begin{subarray}{c}p:T^{\prime}\to Z\\ f\circ p=\operatorname{id}\nolimits_{T^{\prime}}\end{subarray}}}\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(-,p(T^{\prime}))\otimes\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S,-\sqcup\{\xi(1)\})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}β∧α↦(β⊠id{ξ⁡(1)})⋅α\scriptstyle{\ \ \ \ \ \ \ \beta\wedge\alpha\mapsto(\beta\boxtimes\operatorname{id}\nolimits_{\{\xi(1)\}})\cdot\alpha}

The following lemma is a consequence of (8.1.1).

0PCC

Lemma 8.1.8. If ξ−1​(M)\xi^{-1}(M) has no maximum, then the construction above gives an isomorphism

Ef∘ξ⊗𝒮f⁡(M)​(Z′)Mf#→∼Mf#⊗𝒮M​(Z)Eξ,E_{f\circ\xi}\otimes_{{\mathcal{S}}_{f(M)}(Z^{\prime})}M_{f^{\#}}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}M_{f^{\#}}\otimes_{{\mathcal{S}}_{M}(Z)}E_{\xi},

and f#f^{\#} provides a morphism of bimodule 22-representations Lξ→Lf∘ξL_{\xi}\to L_{f\circ\xi}.

8.1.4. Twisted object description

We explain how to obtain a version of Lemma 8.1.2 with a differential.

We say that a homotopy class of path in Z⁡(ξ)Z(\xi) is positive if it has the same orientation as ξ([1→2])\xi([1\to 2]).

Fix a finite subset SS of MM and n≥0n\geq 0.

Let S′′S^{\prime\prime} be a subset of SS with nn elements, let s′∈S∖S′′s^{\prime}\in S\setminus S^{\prime\prime} and s′′∈S′′s^{\prime\prime}\in S^{\prime\prime}. Let ζ:s′′→s′\zeta:s^{\prime\prime}\to s^{\prime} be a positive smooth homotopy class of paths in ZZ. We put S′=(S′′∖{s′′})⊔{s′}S^{\prime}=(S^{\prime\prime}\setminus\{s^{\prime\prime}\})\sqcup\{s^{\prime}\}.

We define a map

gS′′,ζ:Hom𝒮∙​(Z)⁡(S′′,{ξ⁡(1),…,ξ⁡(n)})→Hom𝒮∙​(Z)⁡(S′,{ξ⁡(1),…​ξ​(n)})∧Hom𝒮∙​(Z)⁡(S∖S′,S∖S′′).g_{S^{\prime\prime},\zeta}:\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S^{\prime\prime},\{\xi(1),\ldots,\xi(n)\})\to\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S^{\prime},\{\xi(1),\ldots\xi(n)\})\wedge\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S\setminus S^{\prime},S\setminus S^{\prime\prime}).

We put

gS′′,ζ(α)=(α|S′∖{s′}⊠(αs′′∘ζ−1))∧(idS∖(S′′⊔{s′})⊠ζ)g_{S^{\prime\prime},\zeta}(\alpha)=(\alpha_{|S^{\prime}\setminus\{s^{\prime}\}}\boxtimes(\alpha_{s^{\prime\prime}}\circ\zeta^{-1}))\wedge(\operatorname{id}\nolimits_{S\setminus(S^{\prime\prime}\sqcup\{s^{\prime}\})}\boxtimes\zeta)

if

  • •

    αs′′∘ζ−1\alpha_{s^{\prime\prime}}\circ\zeta^{-1} is smooth

  • •

    and given s∈S′′∖{s′′}s\in S^{\prime\prime}\setminus\{s^{\prime\prime}\} and ζ′:s→s′\zeta^{\prime}:s\to s^{\prime} and ζ′′:s′′→s\zeta^{\prime\prime}:s^{\prime\prime}\to s smooth positive with ζ=ζ′∘ζ′′\zeta=\zeta^{\prime}\circ\zeta^{\prime\prime} and with αs′′∘ζ′′−1\alpha_{s^{\prime\prime}}\circ\zeta^{\prime\prime-1} smooth, then αs′′∘ζ′′−1∘αs−1\alpha_{s^{\prime\prime}}\circ\zeta^{\prime\prime-1}\circ\alpha_{s}^{-1} is negative.

We put gS′′,ζ​(α)=0g_{S^{\prime\prime},\zeta}(\alpha)=0 otherwise.

0PCD

Remark 8.1.9. Note that if αs′′∘ζ−1\alpha_{s^{\prime\prime}}\circ\zeta^{-1} is smooth, then the support of ζ\zeta is contained in Z⁡(ξ)Z(\xi).

Given α\alpha non-zero, if ζ\zeta is positive, then both αs′′∘ζ−1\alpha_{s^{\prime\prime}}\circ\zeta^{-1} and ζ\zeta are oriented, since αs′′\alpha_{s^{\prime\prime}} is oriented.

We obtain a map fS′′,ζ:α∧β↦(id∧β)∘gS′′,ζ​(α)f_{S^{\prime\prime},\zeta}:\alpha\wedge\beta\mapsto(\operatorname{id}\nolimits\wedge\beta)\circ g_{S^{\prime\prime},\zeta}(\alpha)

Hom𝒮∙​(Z)⁡(S′′,{ξ⁡(1),…,ξ⁡(n)})∧Hom⁡(S∖S′′,−)→Hom𝒮∙​(Z)⁡(S′,{ξ⁡(1),…​ξ​(n)})∧Hom𝒮∙​(Z)⁡(S∖S′,−).\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S^{\prime\prime},\{\xi(1),\ldots,\xi(n)\})\wedge\operatorname{Hom}\nolimits(S\setminus S^{\prime\prime},-)\to\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S^{\prime},\{\xi(1),\ldots\xi(n)\})\wedge\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S\setminus S^{\prime},-).

Let r⁡(S′′)r(S^{\prime\prime}) be the number of pairs (s′′,s)∈S′′×(S∖S′′)(s^{\prime\prime},s)\in S^{\prime\prime}\times(S\setminus S^{\prime\prime}) such that there exists a positive path s′′→ss^{\prime\prime}\to s.

We define now

Vr=⨁S′⊂S,|S′|=nr⁡(S′)=rHom⁡(S′,{ξ⁡(1),…,ξ⁡(n)})⊗Hom𝒮⁡(Z)⁡(S∖S′,−)∈𝒮M​(Z)​−diff.V_{r}=\bigoplus_{\begin{subarray}{c}S^{\prime}\subset S,\ |S^{\prime}|=n\\ r(S^{\prime})=r\end{subarray}}\operatorname{Hom}\nolimits(S^{\prime},\{\xi(1),\ldots,\xi(n)\})\otimes\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S\setminus S^{\prime},-)\in{\mathcal{S}}_{M}(Z)\operatorname{\!-diff}\nolimits.

Given r′<r′′r^{\prime}<r^{\prime\prime}, define fr′,r′′=∑S′′,ζfS′′,ζf_{r^{\prime},r^{\prime\prime}}=\sum_{S^{\prime\prime},\zeta}f_{S^{\prime\prime},\zeta}, where

  • •

    S′′S^{\prime\prime} is a subset of SS with |S′′|=n|S^{\prime\prime}|=n and r⁡(S′′)=r′′r(S^{\prime\prime})=r^{\prime\prime}

  • •

    ζ\zeta is a positive admissible homotopy class of paths in ZZ with ζ⁡(0)∈S′′\zeta(0)\in S^{\prime\prime} and ζ⁡(1)∈S∖S′′\zeta(1)\in S\setminus S^{\prime\prime}

such that supp⁡(ζ)∩S′′={s′′}\operatorname{supp}\nolimits(\zeta)\cap S^{\prime\prime}=\{s^{\prime\prime}\} and r⁡((S′′∖{ζ⁡(0)})⊔{ζ⁡(1)})=r′r((S^{\prime\prime}\setminus\{\zeta(0)\})\sqcup\{\zeta(1)\})=r^{\prime}.

Let V=Vn​(S)=⨁rVrV=V_{n}(S)=\bigoplus_{r}V_{r} and let dV=∑rdVr+∑r′,r′′fr′,r′′d_{V}=\sum_{r}d_{V_{r}}+\sum_{r^{\prime},r^{\prime\prime}}f_{r^{\prime},r^{\prime\prime}}. We will show below (Proposition 8.1.10) that dV2=0d_{V}^{2}=0, i.e. VV is the object of 𝒮M​(Z)​−diff{\mathcal{S}}_{M}(Z)\operatorname{\!-diff}\nolimits corresponding to the twisted object [⨁Vr,(fr′,r′′)][\bigoplus V_{r},(f_{r^{\prime},r^{\prime\prime}})].

0PCE

Proposition 8.1.10. Given S⊂MS\subset M and n≥0n\geq 0, then dVn​(S)2=0d_{V_{n}(S)}^{2}=0 and the map of Lemma 8.1.2 defines an isomorphism of functors 𝒮M​(Z)→k​−diff{\mathcal{S}}_{M}(Z)\to k\operatorname{\!-diff}\nolimits

Vn​(S)→∼L⁡(−,S,en).V_{n}(S)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L(-,S,e^{n}).
0PCF

Proof. By Remark 8.1.1, we can assume ξ\xi is outgoing for ZZ. We will show that

(8.1.2) the isomorphism of Lemma 8.1.2 is compatible with the differentials.

The proposition will follow immediately from (8.1.2).

Let S′′S^{\prime\prime} be a subset of SS with nn elements, and let TT be a finite subset of MM. Let a:Hom𝒮∙​(Z)⁡(S′′,{ξ⁡(1),…​ξ​(n)})∧Hom𝒮∙​(Z)⁡(S∖S′′,T)→L∙​(T,S,en)a:\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S^{\prime\prime},\{\xi(1),\ldots\xi(n)\})\wedge\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S\setminus S^{\prime\prime},T)\to L^{\bullet}(T,S,e^{n}) be the map of Lemma 8.1.2. Let α∈Hom𝒮∙​(Z)⁡(S′′,{ξ⁡(1),…​ξ​(n)})\alpha\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S^{\prime\prime},\{\xi(1),\ldots\xi(n)\}) and β∈Hom𝒮∙​(Z)⁡(S∖S′′,T)\beta\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S\setminus S^{\prime\prime},T). Let θ=α⊠β=a⁡(α∧β)\theta=\alpha\boxtimes\beta=a(\alpha\wedge\beta). The statement (8.1.2) will follow from the following property:

(8.1.3) a⁡(d⁡(α⊗β))=d⁡(θ).a(d(\alpha\otimes\beta))=d(\theta).

We have

a⁡(d⁡(α⊗β))=d⁡(α)⊠β+α⊠d⁡(β)+∑ζa⁡((id⊗β)⋅gS′′,ζ​(α))a(d(\alpha\otimes\beta))=d(\alpha)\boxtimes\beta+\alpha\boxtimes d(\beta)+\sum_{\zeta}a((\operatorname{id}\nolimits\otimes\beta)\cdot g_{S^{\prime\prime},\zeta}(\alpha))

where ζ\zeta runs over positive admissible homotopy classes of paths starting in S′′S^{\prime\prime} and ending in S∖S′′S\setminus S^{\prime\prime}.

We have

D⁡(θ)/inv=(D⁡(α)/inv)⊔(D⁡(β)/inv)⊔∐(s1,s2)∈S′′×(S∖S′′)I⁡(αs1,βs2)∩D⁡(θ).D(\theta)/\mathrm{inv}=\bigl(D(\alpha)/\mathrm{inv}\bigr)\sqcup\bigl(D(\beta)/\mathrm{inv}\bigr)\sqcup\coprod_{(s_{1},s_{2})\in S^{\prime\prime}\times(S\setminus S^{\prime\prime})}I(\alpha_{s_{1}},\beta_{s_{2}})\cap D(\theta).

Fix (s1,s2)∈S′′×(S∖S′′)(s_{1},s_{2})\in S^{\prime\prime}\times(S\setminus S^{\prime\prime}). Let S′=(S′′∖{s1})⊔{s2}S^{\prime}=(S^{\prime\prime}\setminus\{s_{1}\})\sqcup\{s_{2}\}. Let ζ\zeta be a smooth path s1→s2s_{1}\to s_{2}. Let u′=idS∖(S′⊔{s1})⊠ζu^{\prime}=\operatorname{id}\nolimits_{S\setminus(S^{\prime}\sqcup\{s_{1}\})}\boxtimes\zeta. Write gS′′,ζ​(α)=v∧ug_{S^{\prime\prime},\zeta}(\alpha)=v\wedge u with u:S∖S′→S∖S′′u:S\setminus S^{\prime}\to S\setminus S^{\prime\prime} and v:S′→{ξ⁡(1),…,ξ⁡(n)}v:S^{\prime}\to\{\xi(1),\ldots,\xi(n)\}. We take u=0u=0 and v=0v=0 if gS′′,ζ​(α)=0g_{S^{\prime\prime},\zeta}(\alpha)=0. If gS′′,ζ​(α)≠0g_{S^{\prime\prime},\zeta}(\alpha)\neq 0, then u=u′u=u^{\prime}.

Assume (β⋅u)≠0(\beta\cdot u)\neq 0. Then αs1∘ζ−1\alpha_{s_{1}}\circ\zeta^{-1} and βs2∘ζ\beta_{s_{2}}\circ\zeta are smooth, and ζ\zeta and ζ¯\bar{\zeta} have opposite orientations, since ζ¯\bar{\zeta} is negative (it starts in ξ⁡(𝐙≥1)\xi({\mathbf{Z}}_{\geq 1}) and ends in MM). It follows that ζ∈L⁡(θ)\zeta\in L(\theta).

Assume ζ∈L⁡(θ)\zeta\in L(\theta). Since ζ¯\bar{\zeta} is negative, it follows that ζ\zeta is positive, then (θζ)s=θs(\theta^{\zeta})_{s}=\theta_{s} for s∉{s1,s2}s{\not\in}\{s_{1},s_{2}\}, while (θζ)s1=βs2∘ζ(\theta^{\zeta})_{s_{1}}=\beta_{s_{2}}\circ\zeta and (θζ)s2=αs1∘ζ−1(\theta^{\zeta})_{s_{2}}=\alpha_{s_{1}}\circ\zeta^{-1}. We deduce that (β⋅u)⊠v=θζ(\beta\cdot u)\boxtimes v=\theta^{\zeta} if β⋅u≠0\beta\cdot u\neq 0. So, the assertion (8.1.3) is a consequence of the following:

(8.1.4) given ​ζ∈L⁡(θ)​ positive, we have ​β⋅u≠0​ if and only if ​ζ∈D⁡(θ).\text{given }\zeta\in L(\theta)\text{ positive, we have }\beta\cdot u\neq 0\text{ if and only if }\zeta\in D(\theta).

We will prove that statement by reduction to the non-singular case. Let f:Z^→Zf:\hat{Z}\to Z be a non-singular cover. The morphism ξ:𝐑>0→Z\xi:{\mathbf{R}}_{>0}\to Z lifts uniquely to a morphism of curves ξ^:𝐑>0→Z^\hat{\xi}:{\mathbf{R}}_{>0}\to\hat{Z}. Let M^=f−1​(M)\hat{M}=f^{-1}(M) and let α^:S^′′→{ξ^​(1),…​ξ^​(n)}\hat{\alpha}:\hat{S}^{\prime\prime}\to\{\hat{\xi}(1),\ldots\hat{\xi}(n)\} and ζ^\hat{\zeta} be the unique lifts of α\alpha and ζ\zeta to Z^\hat{Z}. There exist subsets S^,T^\hat{S},\hat{T} of M^\hat{M} and a lift β^:S^∖S^′′→T^\hat{\beta}:\hat{S}\setminus\hat{S}^{\prime\prime}\to\hat{T} of β\beta such that ζ^∈L⁡(θ^)\hat{\zeta}\in L(\hat{\theta}), where θ^=α^⊠β^\hat{\theta}=\hat{\alpha}\boxtimes\hat{\beta} (Lemma 7.4.28). We have ζ∈D⁡(θ)\zeta\in D(\theta) if and only if ζ^∈D⁡(θ^)\hat{\zeta}\in D(\hat{\theta}) (Lemma 7.4.28).

Write gS^′′,ζ^​(α^)=v^∧u^g_{\hat{S}^{\prime\prime},\hat{\zeta}}(\hat{\alpha})=\hat{v}\wedge\hat{u} as above. We have f⁡(u^)=uf(\hat{u})=u and f⁡(v^)=vf(\hat{v})=v. We have gS′′,ζ​(α)≠0g_{S^{\prime\prime},\zeta}(\alpha)\neq 0 if and only if gS^′′,ζ^​(α^)≠0g_{\hat{S}^{\prime\prime},\hat{\zeta}}(\hat{\alpha})\neq 0. Finally, β⋅u≠0\beta\cdot u\neq 0 if and only if β^⋅u^≠0\hat{\beta}\cdot\hat{u}\neq 0. This completes the reduction of (8.1.4) to the case of Z^\hat{Z}.

So, we now prove (8.1.4) assuming ZZ is smooth. Note that Z⁡(ξ)Z(\xi) is isomorphic (as a 11-dimensional space) to an interval of 𝐑{\mathbf{R}}. We consider ζ:s1→s2\zeta:s_{1}\to s_{2} in L⁡(θ)L(\theta) positive with s1∈S′′s_{1}\in S^{\prime\prime} and s2∈S∖S′′s_{2}\in S\setminus S^{\prime\prime}.

Remark 7.4.11 shows that β⋅u′≠0\beta\cdot u^{\prime}\neq 0 if and only if i⁡(βs,βs2∘ζ)=i⁡(βs,βs2)+i⁡(ids,ζ)i(\beta_{s},\beta_{s_{2}}\circ\zeta)=i(\beta_{s},\beta_{s_{2}})+i(\operatorname{id}\nolimits_{s},\zeta) for all s∈S∖(S′⊔{s1})s\in S\setminus(S^{\prime}\sqcup\{s_{1}\}). That equality is always satisfied unless there are ζ′′:s1→s\zeta^{\prime\prime}:s_{1}\to s and ζ′:s→s2\zeta^{\prime}:s\to s_{2} positive. In that case, ζ¯′′\bar{\zeta}^{\prime\prime} is negative and the equality is satisfied if and only if ζ¯′\bar{\zeta}^{\prime} is positive.

We have u≠0u\neq 0 if and only if given ζ′′:s1→s\zeta^{\prime\prime}:s_{1}\to s and ζ′:s→s2\zeta^{\prime}:s\to s_{2} positive with s∈S′∖{s2}s\in S^{\prime}\setminus\{s_{2}\}, then ζ¯′′=αs2∘ζ′′∘αs1−1\bar{\zeta}^{\prime\prime}=\alpha_{s_{2}}\circ\zeta^{\prime\prime}\circ\alpha_{s_{1}}^{-1} is positive.

We deduce that ζ∈D⁡(θ)\zeta\in D(\theta) if and only if β⋅u≠0\beta\cdot u\neq 0. The proposition follows. ∎

0PCG

Example 8.1.11. The picture below gives two examples of description of the map gS′′,ζg_{S^{\prime\prime},\zeta}.

[Uncaptioned image]

8.1.5. Right action

Consider now ξ′:𝐑<0→Z\xi^{\prime}:{\mathbf{R}}_{<0}\to Z an injective morphism of curves, where 𝐑<0{\mathbf{R}}_{<0} is unoriented. Identifying (𝐑<0)opp({\mathbf{R}}_{<0})^{\operatorname{opp}\nolimits} with 𝐑>0{\mathbf{R}}_{>0} by x↦−xx\mapsto-x, we obtain a morphism of curves ξ:𝐑>0→Zopp\xi:{\mathbf{R}}_{>0}\to Z^{\operatorname{opp}\nolimits}. Let MM be a subset of Z∖ξ′​(𝐑≤−1)Z\setminus\xi^{\prime}({\mathbf{R}}_{\leq-1}).

We say that ξ′\xi^{\prime} is initial for (Z,M)(Z,M) if ξ\xi is terminal for (Zopp,M)(Z^{\operatorname{opp}\nolimits},M) and that ξ′\xi^{\prime} is incoming for ZZ if ξ′​(𝐑≤−1)\xi^{\prime}({\mathbf{R}}_{\leq-1}) is closed in ZZ.

Assume ξ′\xi^{\prime} is initial for (Z,M)(Z,M). As in the left action case, we define a differential functor

R∙=Rξ′∙:𝒞×𝒞opp×𝒰\displaystyle R^{\bullet}=R_{\xi^{\prime}}^{\bullet}:{\mathcal{C}}\times{\mathcal{C}}^{\operatorname{opp}\nolimits}\times{\mathcal{U}} →k​−diff\displaystyle\to k\operatorname{\!-diff}\nolimits
R∙​(S,T,en)\displaystyle R^{\bullet}(S,T,e^{n}) =Hom⁡(T⊔{ξ′​(−1),…,ξ′​(−n)},S)\displaystyle=\operatorname{Hom}\nolimits(T\sqcup\{\xi^{\prime}(-1),\ldots,\xi^{\prime}(-n)\},S)
R∙​(β,α,σ)​(f)\displaystyle R^{\bullet}(\beta,\alpha,\sigma)(f) =β⋅f⋅(α⊠ξ′​(σrev​opp))∈R∙​(S′,T′,n)\displaystyle=\beta\cdot f\cdot(\alpha\boxtimes\xi^{\prime}(\sigma^{\mathrm{rev}{\operatorname{opp}\nolimits}}))\in R^{\bullet}(S^{\prime},T^{\prime},n)

for α∈Hom𝒮∙​(Z)⁡(T′,T)\alpha\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(T^{\prime},T), β∈Hom𝒮∙​(Z)⁡(S,S′)\beta\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(S,S^{\prime}) and σ∈End𝒰∙⁡(en)\sigma\in\operatorname{End}\nolimits_{{\mathcal{U}}^{\bullet}}(e^{n}), and f∈R∙​(S,T,n)f\in R^{\bullet}(S,T,n).

We put Rξ∙​(S,T)=Rξ∙​(S,T,e)R_{\xi}^{\bullet}(S,T)=R_{\xi}^{\bullet}(S,T,e) and Rξ=𝐅2​[Rξ∙]R_{\xi}={\mathbf{F}}_{2}[R_{\xi}^{\bullet}].

Recall that the isomorphism (7.4.4) of differential categories 𝒮M∙​(Z)→∼𝒮M∙​(Zopp)opp{\mathcal{S}}^{\bullet}_{M}(Z)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}{\mathcal{S}}^{\bullet}_{M}(Z^{\operatorname{opp}\nolimits})^{\operatorname{opp}\nolimits}. This isomorphism provides an isomorphism Rξ′∙​(S,T,en)→∼Lξ∙​(T,S,en)R^{\bullet}_{\xi^{\prime}}(S,T,e^{n})\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L^{\bullet}_{\xi}(T,S,e^{n}) functorial in SS, TT and ene^{n}.

In particular, R∙R^{\bullet} provides a “right” 22-representation on 𝒮M∙​(Z){\mathcal{S}}^{\bullet}_{M}(Z) and all results of §8.1.1–8.1.4 have counterparts for R∙R^{\bullet}.

Given S⊂MS\subset M and n≥0n\geq 0, there is an isomorphism of functors

⋁S′⊂S|S′|=nHom𝒮∙​(Z)⁡({ξ′​(−1),…,ξ′​(−n)},S′)∧Hom𝒮∙​(Z)⁡(−,S∖S′)→∼R∙​(S,−,en).\bigvee_{\begin{subarray}{c}S^{\prime}\subset S\\ |S^{\prime}|=n\end{subarray}}\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(\{\xi^{\prime}(-1),\ldots,\xi^{\prime}(-n)\},S^{\prime})\wedge\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(-,S\setminus S^{\prime})\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}R^{\bullet}(S,-,e^{n}).

There is an isomorphism of functors, functorial in SS and TT

R∙​(T,−,en)∧R∙​(−,S,em)\displaystyle R^{\bullet}(T,-,e^{n})\wedge R^{\bullet}(-,S,e^{m}) →∼R∙​(T,S,en+m)\displaystyle\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}R^{\bullet}(T,S,e^{n+m})
(α,β)\displaystyle(\alpha,\beta) ↦α⋅(β⊠ξ′([−m−r→−r]1≤r≤n)).\displaystyle\mapsto\alpha\cdot(\beta\boxtimes\xi^{\prime}([-m-r\to-r]_{1\leq r\leq n})).

Assume there is a decreasing sequence m0,m−1,…m_{0},m_{-1},\ldots of points of ξ′−1​(M)\xi^{\prime-1}(M) with limimi<t\lim_{i}m_{i}<t for all t∈ξ′−1​(M)t\in\xi^{\prime-1}(M).

We obtain as in (8.1.1) isomorphisms functorial in SS and TT

(8.1.5) colimr→∞⁡Hom𝒮M∙​(Z)​(T⊔ξ′​({m−r,…,m−r−n+1}),S)→∼R∙​(S,T,en).\operatorname{colim}\nolimits_{r\to\infty}\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}_{M}(Z)}(T\sqcup\xi^{\prime}(\{m_{-r},\ldots,m_{-r-n+1}\}),S)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}R^{\bullet}(S,T,e^{n}).

Let us finally consider functoriality as in §8.1.3. Let f:Z→Z′f:Z\to Z^{\prime} be a morphism of curves and assume f∘ξ′f\circ\xi^{\prime} is initial for (Z′,f⁡(M))(Z^{\prime},f(M)).

The functor f:𝒮f,M∙​(Z)→𝒮f⁡(M)∙​(Z′)f:{\mathcal{S}}_{f,M}^{\bullet}(Z)\to{\mathcal{S}}^{\bullet}_{f(M)}(Z^{\prime}) induces a morphism of bimodule 22-representations Rf∘ξ′∙→Rξ′∙R_{f\circ\xi^{\prime}}^{\bullet}\to R_{\xi^{\prime}}^{\bullet}, when |f−1​(f​(z))|=1|f^{-1}(f(z))|=1 for all z∈Mz\in M.

If ff is strict, then the functor f#:add⁡(𝒮f⁡(M)​(Z′))→add⁡(𝒮M​(Z))f^{\#}:\operatorname{add}\nolimits({\mathcal{S}}_{f(M)}(Z^{\prime}))\to\operatorname{add}\nolimits({\mathcal{S}}_{M}(Z)) induces a morphism of bimodule 22-representations Rξ′→Rf∘ξ′R_{\xi^{\prime}}\to R_{f\circ\xi^{\prime}}.

0PCH

Remark 8.1.12. As in Remark 8.1.4, we recover the construction of “top algebra module” of Douglas and Manolescu by taking the underlying lax 22-representation of Rξ′R_{\xi^{\prime}}.

0PCI

Example 8.1.13. As in Example 8.1.5, we use an alternative graphical description for Rξ′∙R_{\xi^{\prime}}^{\bullet}. This is illustrated in the example of Rξ′∙​(−,−,e2)R_{\xi^{\prime}}^{\bullet}(-,-,e^{2}) below.

[Uncaptioned image]

8.1.6. Duality

Let Z′=𝐑Z^{\prime}={\mathbf{R}} be the smooth curve with Zo′=(−12,12)Z^{\prime}_{o}=(-\frac{1}{2},\frac{1}{2}), with its standard orientation. Consider a morphism of curves ξ~:Z′→Z\tilde{\xi}:Z^{\prime}\to Z such that ξ~​(Z′)\tilde{\xi}(Z^{\prime}) is a component of ZZ.

Fix an increasing homeomorphism α:𝐑>0→∼𝐑>12\alpha:{\mathbf{R}}_{>0}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}{\mathbf{R}}_{>\frac{1}{2}} fixing the positive integers and define α′:𝐑<0→∼𝐑<−12\alpha^{\prime}:{\mathbf{R}}_{<0}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}{\mathbf{R}}_{<-\frac{1}{2}} by α′​(t)=−α⁡(−t)\alpha^{\prime}(t)=-\alpha(-t). Let ξ+=ξ~∘α:𝐑>0→Z\xi^{+}=\tilde{\xi}\circ\alpha:{\mathbf{R}}_{>0}\to Z and ξ−=ξ~∘α′:𝐑<0→Z\xi^{-}=\tilde{\xi}\circ\alpha^{\prime}:{\mathbf{R}}_{<0}\to Z. These are injective morphisms of curves, ξ+\xi^{+} is outgoing for ZZ and ξ−\xi^{-} is incoming for ZZ.

Given n≥0n\geq 0, we denote by θ⁡(n)∈Hom𝒮∙​(Z′)⁡({−n,…,−1},{1,…,n})\theta(n)\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z^{\prime})}(\{-n,\ldots,-1\},\{1,\ldots,n\}) the braid given by θ(n)−i=[−i→i]\theta(n)_{-i}=[-i\to i].

Let TT and T′T^{\prime} two finite subsets of ZZ and I⊂𝐙≥1I\subset{\mathbf{Z}}_{\geq 1} finite. Assume that ξ~​(−I)⊂T\tilde{\xi}(-I)\subset T and that given x∈𝐑x\in{\mathbf{R}} with x<ix<i for all i∈−Ii\in-I, we have ξ~​(x)∉T\tilde{\xi}(x){\not\in}T. Assume also that ξ~​(I)⊂T′\tilde{\xi}(I)\subset T^{\prime} and that given x∈𝐑x\in{\mathbf{R}} with x>ix>i for all i∈Ii\in I, we have ξ~​(x)∉T′\tilde{\xi}(x){\not\in}T^{\prime}.

We consider the pointed map

κI:Hom𝒮∙​(Z)⁡(T,T′)\displaystyle\kappa_{I}:\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(T,T^{\prime}) →Hom𝒮∙​(Z)⁡(T∖(T∩ξ~​(−I)),T′∖(T′∩ξ~​(I)))\displaystyle\to\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(T\setminus(T\cap\tilde{\xi}(-I)),T^{\prime}\setminus(T^{\prime}\cap\tilde{\xi}(I)))
θ\displaystyle\theta ↦{(θt)t∈T∖ξ~​(−I) if ​χ​(θ)​(ξ~​(−i))=ξ~​(i)​ for ​i∈I0 otherwise.\displaystyle\mapsto\begin{cases}(\theta_{t})_{t\in T\setminus\tilde{\xi}(-I)}&\text{ if }\chi(\theta)(\tilde{\xi}(-i))=\tilde{\xi}(i)\text{ for }i\in I\\ 0&\text{ otherwise.}\end{cases}

We put κn=κ{1,…,n}\kappa_{n}=\kappa_{\{1,\ldots,n\}}. Note that κn=κ{n}∘⋯∘κ{2}∘κ{1}\kappa_{n}=\kappa_{\{n\}}\circ\cdots\circ\kappa_{\{2\}}\circ\kappa_{\{1\}}.

Let f:Z→Z¯f:Z\to\bar{Z} be a morphism of curves such that f∘ξ~f\circ\tilde{\xi} is a homeomorphism from Z′Z^{\prime} to a component of Z¯\bar{Z}. Put ξ¯~=f∘ξ~\tilde{\bar{\xi}}=f\circ\tilde{\xi}. Denote by κ¯n\bar{\kappa}_{n} the map defined as above with ZZ replaced by Z¯\bar{Z}.

Let TT and T′T^{\prime} be two finite subsets of ZZ such that |f⁡(T)|=|T||f(T)|=|T| and |f⁡(T′)|=|T′||f(T^{\prime})|=|T^{\prime}|. Put T̊=T∖(T∩ξ~​({−n,…,−1})CLOSE\mathring{T}=T\setminus(T\cap\tilde{\xi}(\{-n,\ldots,-1\}) and T̊′=T′∖(T′∩ξ~​({−n,…,−1})CLOSE\mathring{T}^{\prime}=T^{\prime}\setminus(T^{\prime}\cap\tilde{\xi}(\{-n,\ldots,-1\}). There is a commutative diagram

(8.1.6) Hom𝒮∙​(Z)⁡(T,T′)\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(T,T^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}κn\scriptstyle{\kappa_{n}}f\scriptstyle{f}Hom𝒮∙​(Z)⁡(T̊,T̊′)\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(\mathring{T},\mathring{T}^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f\scriptstyle{f}Hom𝒮∙​(Z¯)⁡(f⁡(T),f⁡(T′))\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(\bar{Z})}(f(T),f(T^{\prime}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}κn\scriptstyle{\kappa_{n}}Hom𝒮∙​(Z¯)⁡(f⁡(T̊),f⁡(T̊′))\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(\bar{Z})}(f(\mathring{T}),f(\mathring{T}^{\prime}))}

Similarly, if ff is strict and UU and U′U^{\prime} are two finite subsets of Z¯\bar{Z}, there is a commutative diagram

(8.1.7) Hom𝒮⁡(Z¯)⁡(U,U′)\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}(\bar{Z})}(U,U^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}κn\scriptstyle{\kappa_{n}}f#\scriptstyle{f^{\#}}Hom𝒮⁡(Z¯)⁡(U∖(U∩ξ¯~​({−n,…,−1})),U′∖(U′∩ξ¯~​({1,…,n}))CLOSE\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}(\bar{Z})}(U\setminus(U\cap\tilde{\bar{\xi}}(\{-n,\ldots,-1\})),U^{\prime}\setminus(U^{\prime}\cap\tilde{\bar{\xi}}(\{1,\ldots,n\}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f#\scriptstyle{f^{\#}}⨁T,T′Hom𝒮⁡(Z)⁡(T,T′)\textstyle{\bigoplus_{T,T^{\prime}}\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(T,T^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}κn\scriptstyle{\kappa_{n}}⨁T,T′Hom𝒮⁡(Z)⁡(T̊,T̊′)\textstyle{\bigoplus_{T,T^{\prime}}\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(\mathring{T},\mathring{T}^{\prime})}

where TT (resp. T′T^{\prime}) runs over finite subsets of ZZ such that f⁡(T)=Uf(T)=U (resp. f⁡(U′)=T′f(U^{\prime})=T^{\prime}).

0PCJ

Lemma 8.1.14. The map κn\kappa_{n} commutes with differentials.

0PCK

Proof. Assume first ξ~\tilde{\xi} is a homeomorphism and Zo=∅Z_{o}=\emptyset. Let TT and T′T^{\prime} be two finite subsets of 𝐑{\mathbf{R}} with same cardinality mm. Let a:{1,…,m}→∼Ta:\{1,\ldots,m\}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}T and a′:{1,…,m}→∼T′a^{\prime}:\{1,\ldots,m\}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}T^{\prime} be the increasing bijections. There is an isomorphism of differential modules (Proposition 7.4.33) ϕ:Hom𝒮⁡(Z)⁡(T,T′)→∼Hm\phi:\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(T,T^{\prime})\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}H_{m}: given θ∈Hom𝒮∙​(Z)⁡(T,T′)\theta\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(T,T^{\prime}) non-zero and given i∈{1,…,m}i\in\{1,\ldots,m\}, we put ϕ⁡(θ)​(i)=a′−1​(θa⁡(i)​(1))\phi(\theta)(i)=a^{\prime-1}(\theta_{a(i)}(1)).

Assume in addition that {−n,…,−1}⊂T\{-n,\ldots,-1\}\subset T and T∖{−n,…,−1}⊂(−1,∞)T\setminus\{-n,\ldots,-1\}\subset(-1,\infty) and {1,…,n}⊂T′\{1,\ldots,n\}\subset T^{\prime} and T′∖{1,…,n}⊂(−∞,1)T^{\prime}\setminus\{1,\ldots,n\}\subset(-\infty,1). There is a commutative diagram

Hom𝒮⁡(Z)⁡(T,T′)\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(T,T^{\prime})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}κn\scriptstyle{\kappa_{n}}ϕ\scriptstyle{\phi}∼\scriptstyle{\sim}Hom𝒮⁡(Z)⁡(T∖{−n,…,−1},T′∖{1,…,n})\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(T\setminus\{-n,\ldots,-1\},T^{\prime}\setminus\{1,\ldots,n\})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϕ\scriptstyle{\phi}∼\scriptstyle{\sim}Hm\textstyle{H_{m}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tm,m−n−\scriptstyle{t^{-}_{m,m-n}}Hm−n\textstyle{H_{m-n}}

The lemma follows now from §6.1.1.

Assume now ZZ is smooth. If Z⁡(ξ+)Z(\xi^{+}) is unoriented, then the lemma holds by the discussion above, using §7.4.10. In general, we consider the morphism of curves f:Z→Z¯f:Z\to\bar{Z} that is an isomorphism outside Z⁡(ξ+)Z(\xi^{+}) and the identity on Z⁡(ξ+)Z(\xi^{+}), with f​(Z⁡(ξ+))o=∅f(Z(\xi^{+}))_{o}=\emptyset. The vertical maps of the commutative diagram (8.1.6) are injective, hence the lemma holds for ZZ since it holds for Z¯\bar{Z}.

Consider now a general ZZ. Let f:Z^→Zf:\hat{Z}\to Z be a non-singular cover. The vertical maps of the commutative diagram (8.1.7) are injective, hence the lemma holds for ZZ since it holds for Z^\hat{Z}. ∎

Let MM be a subset of Z∖ξ~((−∞,−1]∪[1,∞))Z\setminus\tilde{\xi}\bigl((-\infty,-1]\cup[1,\infty)\bigr).

Given SS a finite subset of MM, the pointed map

Lξ+∙​(T,S,en)∧Rξ−∙​(S,T′,en)→Hom𝒮∙​(Z)⁡(T′,T),(θ′,θ)↦κn​(θ′⋅θ)L_{\xi^{+}}^{\bullet}(T,S,e^{n})\wedge R_{\xi^{-}}^{\bullet}(S,T^{\prime},e^{n})\to\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}(Z)}(T^{\prime},T),\ (\theta^{\prime},\theta)\mapsto\kappa_{n}(\theta^{\prime}\cdot\theta)

induces an 𝐅2{\mathbf{F}}_{2}-linear map

κ^​(T,S):Lξ+​(T,S,en)\displaystyle\hat{\kappa}(T,S):L_{\xi^{+}}(T,S,e^{n}) →Hom𝒮​(Z)opp​−diff⁡(Rξ−​(S,−,en),Hom⁡(−,T))\displaystyle\to\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)^{\operatorname{opp}\nolimits}\operatorname{\!-diff}\nolimits}(R_{\xi^{-}}(S,-,e^{n}),\operatorname{Hom}\nolimits(-,T))
θ′\displaystyle\theta^{\prime} ↦((θ∈Rξ−∙​(S,T′,en))↦κn​(θ′⋅θ)).\displaystyle\mapsto\bigl((\theta\in R_{\xi^{-}}^{\bullet}(S,T^{\prime},e^{n}))\mapsto\kappa_{n}(\theta^{\prime}\cdot\theta)\bigr).
0PCL

Proposition 8.1.15. The map κ^\hat{\kappa} induces an isomorphism of differential pointed bimodules Lξ+(−2,−1,en)→∼Rξ−(−1,−2,en)∨L_{\xi^{+}}(-_{2},-_{1},e^{n})\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}R_{\xi^{-}}(-_{1},-_{2},e^{n})^{\vee}.

0PCM

Proof. Lemma 8.1.14 shows that κ^\hat{\kappa} commutes with differentials.

Let SS be a finite subset of MM of cardinality nn.

Assume ξ~\tilde{\xi} is a homeomorphism and Zo=∅Z_{o}=\emptyset. There is a commutative diagram (see the proof of Lemma 8.1.14 with (T,T′)=(S,ξ+​({1,…,n})CLOSE(T,T^{\prime})=(S,\xi^{+}(\{1,\ldots,n\}) and (T,T′)=(ξ−​({−n,…,−1}),S)(T,T^{\prime})=(\xi^{-}(\{-n,\ldots,-1\}),S))

Hom𝒮⁡(Z)⁡(S,ξ+​({1,…,n})CLOSE\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S,\xi^{+}(\{1,\ldots,n\})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}κ^\scriptstyle{\hat{\kappa}}ϕ\scriptstyle{\phi}∼\scriptstyle{\sim}Hom𝒮⁡(Z)⁡(ξ−​({−n,…,−1},S)∗CLOSE\textstyle{\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(\xi^{-}(\{-n,\ldots,-1\},S)^{*}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}(ϕ∗)−1\scriptstyle{(\phi^{*})^{-1}}Hn\textstyle{H_{n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}t^S,∅−\scriptstyle{\hat{t}^{-}_{S,\emptyset}}Hn∗\textstyle{H_{n}^{*}}

The bottom horizontal map is bijective by Corollary 3.1.2, hence κ^​(∅,S)\hat{\kappa}(\emptyset,S) is bijective.

Assume now Z⁡(ξ+)Z(\xi^{+}) is smooth and unoriented. The map κ^​(∅,S)\hat{\kappa}(\emptyset,S) is the same for ZZ and for Z⁡(ξ+)Z(\xi^{+}), so κ^​(∅,S)\hat{\kappa}(\emptyset,S) is still bijective.

Assume Z⁡(ξ+)Z(\xi^{+}) is smooth. There is a morphism of curves f:Z→Z¯f:Z\to\bar{Z} that is an isomorphism outside Z⁡(ξ+)Z(\xi^{+}) and the identity on Z⁡(ξ+)Z(\xi^{+}) with f​(Z⁡(ξ+))o=∅f(Z(\xi^{+}))_{o}=\emptyset. The map κ^​(∅,S)\hat{\kappa}(\emptyset,S) is the same for ZZ and for Z¯\bar{Z}, so κ^​(∅,S)\hat{\kappa}(\emptyset,S) is still bijective.

Consider now a general ZZ and let f:Z^→Zf:\hat{Z}\to Z be a non-singular cover. Let ξ^~:Z′→Z^\tilde{\hat{\xi}}:Z^{\prime}\to\hat{Z} be the morphism of curves such that ξ~=f∘ξ^~\tilde{\xi}=f\circ\tilde{\hat{\xi}}. The functors ff and f#f^{\#} are inverse bijections between Hom𝒮⁡(Z)⁡(S,ξ+​({1,…,n})CLOSE\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S,\xi^{+}(\{1,\ldots,n\}) and ⨁S′Hom𝒮⁡(Z^)⁡(S′,ξ^~​({1,…,n})CLOSE\bigoplus_{S^{\prime}}\operatorname{Hom}\nolimits_{{\mathcal{S}}(\hat{Z})}(S^{\prime},\tilde{\hat{\xi}}(\{1,\ldots,n\}) (resp. Hom𝒮⁡(Z)⁡(ξ−​({−n,…,−1},S)CLOSE\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(\xi^{-}(\{-n,\ldots,-1\},S) and ⨁S′Hom𝒮⁡(Z^)⁡(ξ^~​({−n,…,−1},S′)CLOSE\bigoplus_{S^{\prime}}\operatorname{Hom}\nolimits_{{\mathcal{S}}(\hat{Z})}(\tilde{\hat{\xi}}(\{-n,\ldots,-1\},S^{\prime})), where S′S^{\prime} runs over nn-elements subsets of Z^\hat{Z} such that f⁡(S′)=Sf(S^{\prime})=S. Furthermore, κ^​(∅,S)\hat{\kappa}(\emptyset,S) is compatible with these bijections (see the proof of Lemma 8.1.14). It follows that κ^​(∅,S)\hat{\kappa}(\emptyset,S) is bijective.

We consider now two arbitrary subsets SS and TT of MM. The canonical isomorphisms of Lemma 8.1.2 and of §8.1.5 fit in a commutative diagram of 𝐅2{\mathbf{F}}_{2}-modules

⨁S′Lξ+​(∅,S′,en)⊗Hom𝒮⁡(Z)⁡(S∖S′,T)\textstyle{\bigoplus_{S^{\prime}}L_{\xi^{+}}(\emptyset,S^{\prime},e^{n})\otimes\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S\setminus S^{\prime},T)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}∑S′κ^(∅,S′)⊗id\scriptstyle{\sum_{S^{\prime}}\hat{\kappa}(\emptyset,S^{\prime})\otimes\operatorname{id}\nolimits}Lξ+​(T,S,en)\textstyle{L_{\xi^{+}}(T,S,e^{n})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}κ^​(T,S)\scriptstyle{\hat{\kappa}(T,S)}⨁S′Rξ−​(S′,∅,en)∗⊗Hom𝒮⁡(Z)⁡(S∖S′,T)\textstyle{\bigoplus_{S^{\prime}}R_{\xi^{-}}(S^{\prime},\emptyset,e^{n})^{*}\otimes\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S\setminus S^{\prime},T)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}Hom⁡(Rξ−​(S,−,en),Hom⁡(−,T))\textstyle{\operatorname{Hom}\nolimits(R_{\xi^{-}}(S,-,e^{n}),\operatorname{Hom}\nolimits(-,T))}

where S′S^{\prime} runs over nn elements subsets of SS. The discussion above shows that the left vertical arrow is an isomorphism, hence κ^​(T,S)\hat{\kappa}(T,S) is an isomorphism. ∎

Given x1,x2∈[−1,1]x_{1},x_{2}\in[-1,1], the homotopy class ξ~([x1→x2])\tilde{\xi}([x_{1}\to x_{2}]) is admissible if x1≤x2x_{1}\leq x_{2} or x1≤−12x_{1}\leq-\frac{1}{2} or x2≥12x_{2}\geq\frac{1}{2}. Given x∈[−1,1]x\in[-1,1] and ζ\zeta an admissible class of paths in ZZ with ζ​(1)=ξ~​(x)\zeta(1)=\tilde{\xi}(x) and ξ~([x→1])⋅ζ≠0\tilde{\xi}([x\to 1])\cdot\zeta\neq 0, there is a unique y∈[−1,1]y\in[-1,1] such that ζ=ξ~([y→x])\zeta=\tilde{\xi}([y\to x]).

Let us describe now the unit of the adjunction when n=1n=1.

0PCN

Lemma 8.1.16. The unit of the adjunction (Lξ+(−,−)⊗−,Rξ−(−,−)⊗−)(L_{\xi^{+}}(-,-)\otimes-,R_{\xi^{-}}(-,-)\otimes-) is given by the morphism of bimodules whose evaluation at (T,S)(T,S) is

Hom𝒮⁡(Z)⁡(S,T)\displaystyle\operatorname{Hom}\nolimits_{{\mathcal{S}}(Z)}(S,T) →Rξ−​(T,−)⊗Lξ+​(−,S)\displaystyle\to R_{\xi^{-}}(T,-)\otimes L_{\xi^{+}}(-,S)
γ\displaystyle\gamma ↦∑x∈ξ~−1​(S)(γ|S∖{ξ~(x)}⊠ξ~([−1→x]))⊗(idS∖{ξ~​(x)}⊠ξ~([x→1])).\displaystyle\mapsto\sum_{x\in\tilde{\xi}^{-1}(S)}(\gamma_{|S\setminus\{\tilde{\xi}(x)\}}\boxtimes\tilde{\xi}([-1\to x]))\otimes(\operatorname{id}\nolimits_{S\setminus\{\tilde{\xi}(x)\}}\boxtimes\tilde{\xi}([x\to 1])).
0PCP

Proof. The counit of the adjunction is ε=κ1∘mult\varepsilon=\kappa_{1}\circ\mathrm{mult}. Let γ∈Rξ−∙​(T,S)\gamma\in R_{\xi^{-}}^{\bullet}(T,S). Let η\eta be the map defined in the lemma. We have

η(idT)=∑x∈ξ~−1​(T)(ξ~([−1→x])⊠idT∖{ξ~​(x)})⊗(ξ~([x→1])⊠idT∖{ξ~​(x)}),\eta(\operatorname{id}\nolimits_{T})=\sum_{x\in\tilde{\xi}^{-1}(T)}(\tilde{\xi}([-1\to x])\boxtimes\operatorname{id}\nolimits_{T\setminus\{\tilde{\xi}(x)\}})\otimes(\tilde{\xi}([x\to 1])\boxtimes\operatorname{id}\nolimits_{T\setminus\{\tilde{\xi}(x)\}}),

hence

(id⊗ε)∘(η⊗id)(γ)=∑x∈ξ~−1​(T)(ξ~([−1→x])⊠idT∖{ξ~​(x)})⋅κ1((ξ~([x→1])⊠idT∖{ξ~​(x)})⋅γ)(\operatorname{id}\nolimits\otimes\varepsilon)\circ(\eta\otimes\operatorname{id}\nolimits)(\gamma)=\sum_{x\in\tilde{\xi}^{-1}(T)}(\tilde{\xi}([-1\to x])\boxtimes\operatorname{id}\nolimits_{T\setminus\{\tilde{\xi}(x)\}})\cdot\kappa_{1}\bigl((\tilde{\xi}([x\to 1])\boxtimes\operatorname{id}\nolimits_{T\setminus\{\tilde{\xi}(x)\}})\cdot\gamma\bigr)

Let xx be the unique element of ξ~−1​(χ⁡(γ)​(ξ−​(−1)))\tilde{\xi}^{-1}(\chi(\gamma)(\xi^{-}(-1))). We have γξ−​(−1)=ξ~([−1→x])\gamma_{\xi^{-}(-1)}=\tilde{\xi}([-1\to x]) and κ1((ξ~([x→1])⊠idT∖{ξ~​(x)})⋅γ)=γ|S\kappa_{1}\bigl((\tilde{\xi}([x\to 1])\boxtimes\operatorname{id}\nolimits_{T\setminus\{\tilde{\xi}(x)\}})\cdot\gamma\bigr)=\gamma_{|S}, hence

(id⊗ε)∘(η⊗id)(γ)=(γξ−​(−1)⊠idT∖{χ⁡(γ)​(ξ−​(−1))})⊗γ|S(\operatorname{id}\nolimits\otimes\varepsilon)\circ(\eta\otimes\operatorname{id}\nolimits)(\gamma)=(\gamma_{\xi^{-}(-1)}\boxtimes\operatorname{id}\nolimits_{T\setminus\{\chi(\gamma)(\xi^{-}(-1))\}})\otimes\gamma_{|S}

We deduce that

mult∘(id⊗ε)∘(η⊗id)(γ)=γ\mathrm{mult}\circ(\operatorname{id}\nolimits\otimes\varepsilon)\circ(\eta\otimes\operatorname{id}\nolimits)(\gamma)=\gamma

and the lemma follows. ∎

0PCQ

Remark 8.1.17. There is a bifunctorial injective map

Rξ−∙​(T,−)∧Lξ+∙​(−,S)→Hom⁡(S⊔{ξ−​(−1)},T⊔{ξ+​(1)}),β∧α↦(β⊠idξ+​(1))⋅(α⊠idξ−​(−1)).R_{\xi^{-}}^{\bullet}(T,-)\wedge L_{\xi^{+}}^{\bullet}(-,S)\to\operatorname{Hom}\nolimits(S\sqcup\{\xi^{-}(-1)\},T\sqcup\{\xi^{+}(1)\}),\ \beta\wedge\alpha\mapsto(\beta\boxtimes\operatorname{id}\nolimits_{\xi^{+}(1)})\cdot(\alpha\boxtimes\operatorname{id}\nolimits_{\xi^{-}(-1)}).

The composition of the unit given by Lemma 8.1.16 with this map is the following map

Hom(S,T)→Hom(S⊔{ξ−(−1)},T⊔{ξ+(1)}),γ↦(γ⊗idξ+​(1))⋅d(ξ~([−1→1])⊠idS).\operatorname{Hom}\nolimits(S,T)\to\operatorname{Hom}\nolimits(S\sqcup\{\xi^{-}(-1)\},T\sqcup\{\xi^{+}(1)\}),\ \gamma\mapsto(\gamma\otimes\operatorname{id}\nolimits_{\xi^{+}(1)})\cdot d(\tilde{\xi}([-1\to 1])\boxtimes\operatorname{id}\nolimits_{S}).
0PCR

Example 8.1.18. The first picture below provides an example of description of the unit of the adjunction as in Lemma 8.1.16.

[Uncaptioned image]

The second picture describes a calculation of an image by the counit.

[Uncaptioned image]

8.1.7. Actions for the line

We consider the unoriented curve 𝐑{\mathbf{R}}. Let M={±(1−1n)}n∈𝐙>0M=\{\pm(1-\frac{1}{n})\}_{n\in{\mathbf{Z}}_{>0}}.

Consider S,TS,T two finite subsets of 𝐑{\mathbf{R}} with |S|=|T|=n|S|=|T|=n. Let fS:S→∼{1,…,n}f_{S}:S\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}\{1,\ldots,n\} and fT:T→∼{1,…,n}f_{T}:T\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}\{1,\ldots,n\} be the unique increasing bijections. We define

ϕ⁡(S,T):Hom𝒮∙​(𝐑)⁡(S,T)→∼Hn∙=End𝒰∙⁡(en),θ↦TfT∘χ⁡(θ)∘fS−1.\phi(S,T):\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}({\mathbf{R}})}(S,T)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}H_{n}^{\bullet}=\operatorname{End}\nolimits_{{\mathcal{U}}^{\bullet}}(e^{n}),\ \theta\mapsto T_{f_{T}\circ\chi(\theta)\circ f_{S}^{-1}}.

We define a functor Φ:𝒮M∙​(𝐑)→𝒰∙\Phi:{\mathcal{S}}^{\bullet}_{M}({\mathbf{R}})\to{\mathcal{U}}^{\bullet}. We put Φ⁡(S)=e|S|\Phi(S)=e^{|S|} and Φ​(f)=ϕ​(S,T)​(f)\Phi(f)=\phi(S,T)(f) for f∈Hom𝒮∙​(𝐑)⁡(S,T)f\in\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}({\mathbf{R}})}(S,T).

The next proposition follows from Proposition 7.4.33.

0PCS

Proposition 8.1.19. The functor Φ:𝒮M∙​(𝐑)→𝒰∙\Phi:{\mathcal{S}}^{\bullet}_{M}({\mathbf{R}})\to{\mathcal{U}}^{\bullet} is an equivalence of differential pointed categories.

Consider ξ+:𝐑>0→𝐑\xi_{+}:{\mathbf{R}}_{>0}\to{\mathbf{R}} and ξ−:𝐑<0→𝐑\xi_{-}:{\mathbf{R}}_{<0}\to{\mathbf{R}} the inclusion maps.

We define φ±:Lξ±​(−,−,en)→∼L±​(−,−,en)∘(Φ∧Φ)\varphi_{\pm}:L_{\xi_{\pm}}(-,-,e^{n})\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L^{\pm}(-,-,e^{n})\circ(\Phi\wedge\Phi) by

φ±​(T,S)=ϕ⁡(S,T⊔ξ±​({±1,…,±n})):Hom𝒮∙​(𝐑)⁡(S,T⊔ξ±​({±1,…,±n}))→∼L±​(e|T|,e|S|,n).\varphi_{\pm}(T,S)=\phi(S,T\sqcup\xi_{\pm}(\{\pm 1,\ldots,\pm n\})):\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}({\mathbf{R}})}(S,T\sqcup\xi_{\pm}(\{\pm 1,\ldots,\pm n\}))\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L^{\pm}(e^{|T|},e^{|S|},n).

Similarly, we define φ±′:Rξ±​(−,−,en)→∼R±​(−,−,en)∘(Φ∧Φ)\varphi^{\prime}_{\pm}:R_{\xi_{\pm}}(-,-,e^{n})\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}R^{\pm}(-,-,e^{n})\circ(\Phi\wedge\Phi) by

φ±′​(T,S)=ϕ⁡(T⊔ξ±​({±1,…,±n}),S):Hom𝒮∙​(𝐑)⁡(T⊔ξ±​({±1,…,±n}),S)→∼R±​(e|S|,e|T|,n).\varphi^{\prime}_{\pm}(T,S)=\phi(T\sqcup\xi_{\pm}(\{\pm 1,\ldots,\pm n\}),S):\operatorname{Hom}\nolimits_{{\mathcal{S}}^{\bullet}({\mathbf{R}})}(T\sqcup\xi_{\pm}(\{\pm 1,\ldots,\pm n\}),S)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}R^{\pm}(e^{|S|},e^{|T|},n).
0PCT

Proposition 8.1.20. Together with φ±\varphi_{\pm} (resp. φ±′\varphi_{\pm}^{\prime}), the functor Φ\Phi induces equivalences of bimodule 22-representations between Lξ±L_{\xi_{\pm}} and L±L^{\pm} (resp. Rξ±R_{\xi_{\pm}} and R±R^{\pm}).

8.1.8. Action as functors

We explain here how the 22-representation constructed in §8.1.1 can be described using functors between strand categories of different curves.

Let ZZ be a singular curve and ξ:𝐑>0→Z\xi:{\mathbf{R}}_{>0}\to Z an injective morphism of curves with ξ⁡(𝐑≥1)\xi({\mathbf{R}}_{\geq 1}) closed and contained in Z∖Ze​x​cZ\setminus Z_{exc}.

Let A=⨁I,J⊂Ze​x​cHom𝒜​(Z)opp⁡(J,I)A=\bigoplus_{I,J\subset Z_{exc}}\operatorname{Hom}\nolimits_{{\mathcal{A}}(Z)^{{\operatorname{opp}\nolimits}}}(J,I). We denote by eI∈Ae_{I}\in A the idempotent corresponding to the projection on II, so that eI​A​eJ=Hom𝒜⁡(Z)⁡(I,J)e_{I}Ae_{J}=\operatorname{Hom}\nolimits_{{\mathcal{A}}(Z)}(I,J).

The equivalence A​−diff→∼𝒜​(Z)opp​−diffA\operatorname{\!-diff}\nolimits\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}{\mathcal{A}}(Z)^{{\operatorname{opp}\nolimits}}\operatorname{\!-diff}\nolimits restricts to an equivalence (A¯)i→∼𝒜¯i​(Z)(\bar{A})^{i}\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}\bar{{\mathcal{A}}}^{i}(Z) (cf §2.1.4).

We consider a new singular curve Z^=Z⊔ξ⁡(1)(−1,1)\hat{Z}=Z\sqcup_{\xi(1)}(-1,1) obtained as the quotient of the disjoint union of ZZ and the oriented interval (−1,1)(-1,1) identifying ξ⁡(1)\xi(1) with 00. Note that Z^e​x​c=Ze​x​c∪{ξ⁡(1)}\hat{Z}_{exc}=Z_{exc}\cup\{\xi(1)\}.

We put A^=⨁I,J⊂Z^e​x​cHom𝒜​(Z^)opp⁡(J,I)\hat{A}=\bigoplus_{I,J\subset\hat{Z}_{exc}}\operatorname{Hom}\nolimits_{{\mathcal{A}}(\hat{Z})^{{\operatorname{opp}\nolimits}}}(J,I). As before, we have idempotents eI∈A^e_{I}\in\hat{A} for I∈Z^e​x​cI\in\hat{Z}_{exc}. We put e=∑I⊂Ze​x​ceI⊔{ξ⁡(1)}e=\sum_{I\subset Z_{exc}}e_{I\sqcup\{\xi(1)\}}.

The inclusion i:Z↪Z^i:Z\hookrightarrow\hat{Z} provides a fully faithful functor Ξ:𝒜¯i​(Z)→𝒜¯i​(Z^)\Xi:\bar{{\mathcal{A}}}^{i}(Z)\to\bar{{\mathcal{A}}}^{i}(\hat{Z}). This gives rise to an isomorphism of algebras h:A→∼(1−e)​A^​(1−e)h:A\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}(1-e)\hat{A}(1-e) and we have a commutative diagram

𝒜¯i​(Z)\textstyle{\bar{{\mathcal{A}}}^{i}(Z)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Ξ\scriptstyle{\Xi}𝒜¯i​(Z^)\textstyle{\bar{{\mathcal{A}}}^{i}(\hat{Z})}(A¯)i\textstyle{(\bar{A})^{i}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}can\scriptstyle{{\mathrm{can}}}A^(1−e)⊗A−\scriptstyle{\hat{A}(1-e)\otimes_{A}-}(A^¯)i\textstyle{(\bar{\hat{A}})^{i}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}can\scriptstyle{{\mathrm{can}}}

where AA acts on the right on A^​(1−e)\hat{A}(1-e) by right multiplication preceded by hh.

The inclusion (1−e)​A^​(1−e)↪A^(1-e)\hat{A}(1-e)\hookrightarrow\hat{A} induces a surjective morphism of algebras g:(1−e)​A^​(1−e)↠A^/A^​e​A^g:(1-e)\hat{A}(1-e)\twoheadrightarrow\hat{A}/\hat{A}e\hat{A}. We have e​A^​(1−e)=0e\hat{A}(1-e)=0, hence A^​e​A^∩(1−e)​A^​(1−e)=0\hat{A}e\hat{A}\cap(1-e)\hat{A}(1-e)=0. It follows that gg is an isomorphism.

The right adjoint to A^(1−e)⊗A−\hat{A}(1-e)\otimes_{A}- is HomA^⁡(A^​(1−e),−)\operatorname{Hom}\nolimits_{\hat{A}}(\hat{A}(1-e),-), which is canonically isomorphic to (1−e)A^⊗A^−(1-e)\hat{A}\otimes_{\hat{A}}- and we have a commutative diagram

𝒜¯i​(Z^)\textstyle{\bar{{\mathcal{A}}}^{i}(\hat{Z})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Γ\scriptstyle{\Gamma}𝒜¯i​(Z)\textstyle{\bar{{\mathcal{A}}}^{i}(Z)}(A^¯)i\textstyle{(\bar{\hat{A}})^{i}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}can\scriptstyle{{\mathrm{can}}}(1−e)A^⊗A^−\scriptstyle{(1-e)\hat{A}\otimes_{\hat{A}}-}(A¯)i\textstyle{(\bar{A})^{i}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}can\scriptstyle{{\mathrm{can}}}

where Γ:𝒜¯i​(Z^)→𝒜¯i​(Z)\Gamma:\bar{{\mathcal{A}}}^{i}(\hat{Z})\to\bar{{\mathcal{A}}}^{i}(Z) is the right adjoint of Ξ\Xi.

0PCU

Remark 8.1.21. There is a sequence of four adjoint functors between AA-modules and A^\hat{A}-modules:

(A⊗A^−,A^(1−e)⊗A−,(1−e)A^⊗A^−,HomA((1−e)A^,−)).\bigl(A\otimes_{\hat{A}}-,\hat{A}(1-e)\otimes_{A}-,(1-e)\hat{A}\otimes_{\hat{A}}-,\operatorname{Hom}\nolimits_{A}((1-e)\hat{A},-)\bigr).

The first and fourth functors are not exact in general. Here,

  • •

    A^\hat{A} acts on the right on AA by right multiplication preceded by the composition

    A^→canA^/A^​e​A^→∼g−1(1−e)​A^​(1−e)→∼h−1A\hat{A}\xrightarrow{{\mathrm{can}}}\hat{A}/\hat{A}e\hat{A}\xrightarrow[\sim]{g^{-1}}(1-e)\hat{A}(1-e)\xrightarrow[\sim]{h^{-1}}A
  • •

    (A^(1−e)⊗A−)=((1−e)A^(1−e)⊗A−)→∼h−1(A⊗A−)=HomA(A,−)\bigl(\hat{A}(1-e)\otimes_{A}-\bigr)=\bigl((1-e)\hat{A}(1-e)\otimes_{A}-\bigr)\xrightarrow[\sim]{h^{-1}}\bigl(A\otimes_{A}-\bigr)=\operatorname{Hom}\nolimits_{A}(A,-)

  • •

    ((1−e)A^⊗A^−)→∼can(HomA^(A^(1−e),A^)⊗A^−)→∼canHomA^(A^(1−e),−)\bigl((1-e)\hat{A}\otimes_{\hat{A}}-\bigr)\xrightarrow[\sim]{{\mathrm{can}}}\bigl(\operatorname{Hom}\nolimits_{\hat{A}}(\hat{A}(1-e),\hat{A})\otimes_{\hat{A}}-\bigr)\xrightarrow[\sim]{{\mathrm{can}}}\operatorname{Hom}\nolimits_{\hat{A}}(\hat{A}(1-e),-).

There is also a fully faithful functor

Υ:𝒜¯i​(Z)→𝒜¯i​(Z^),T↦T⊔{ξ⁡(1)}\Upsilon:\bar{{\mathcal{A}}}^{i}(Z)\to\bar{{\mathcal{A}}}^{i}(\hat{Z}),\ T\mapsto T\sqcup\{\xi(1)\}

sending a braid (θt)t∈T(\theta_{t})_{t\in T} to (θt)t∈T⊔(idξ(1)})(\theta_{t})_{t\in T}\sqcup(\mathrm{id}_{\xi(1)\}}). It gives rise to an isomorphism of algebras u:A→∼e​A^​eu:A\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}e\hat{A}e and there is a commutative diagram

𝒜¯i​(Z)\textstyle{\bar{{\mathcal{A}}}^{i}(Z)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Υ\scriptstyle{\Upsilon}𝒜¯i​(Z^)\textstyle{\bar{{\mathcal{A}}}^{i}(\hat{Z})}(A¯)i\textstyle{(\bar{A})^{i}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}can\scriptstyle{{\mathrm{can}}}A^e⊗A−\scriptstyle{\hat{A}e\otimes_{A}-}(A^¯)i\textstyle{(\bar{\hat{A}})^{i}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}∼\scriptstyle{\sim}can\scriptstyle{{\mathrm{can}}}

where the right action of AA on A^​e\hat{A}e is by right multiplication preceded by uu.

We have

L⁡(T,S)=Hom𝒜⁡(Z^)⁡(Ξ⁡(S),Υ⁡(T))→∼Hom𝒜⁡(Z)​(S,Γ​Υ​(T)).L(T,S)=\operatorname{Hom}\nolimits_{{\mathcal{A}}(\hat{Z})}(\Xi(S),\Upsilon(T))\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}\mathrm{Hom}_{{\mathcal{A}}(Z)}(S,\Gamma\Upsilon(T)).

Denote by E=L⊗𝒜¯i​(Z)−E=L\otimes_{\bar{{\mathcal{A}}}^{i}(Z)}- the endofunctor of 𝒜¯i​(Z)\bar{{\mathcal{A}}}^{i}(Z) induced by the bimodule LL. The isomorphism above gives rise to an isomorphism of functors E→∼Γ​ΥE\xrightarrow{\sim}\Gamma\Upsilon.

We put Z^0=Z\hat{Z}_{0}=Z and we define inductively Z^r=Z^r−1⊔ξ⁡(r)(−1,1)\hat{Z}_{r}=\hat{Z}_{r-1}\sqcup_{\xi(r)}(-1,1) for r≥1r\geq 1, where ξ⁡(r)\xi(r) is identified with 00.

We denote by Ξr:𝒜¯i​(Z^r−1)→𝒜¯i​(Z^r),I↦I\Xi_{r}:\bar{{\mathcal{A}}}^{i}(\hat{Z}_{r-1})\to\bar{{\mathcal{A}}}^{i}(\hat{Z}_{r}),\ I\mapsto I the functor associated with the inclusion Z^r−1↪Z^r\hat{Z}_{r-1}\hookrightarrow\hat{Z}_{r}, defined as Ξ\Xi above.

We denote by Υr:𝒜¯i​(Z^r)→𝒜¯i​(Z^r),T↦T⊔{ξ⁡(r)}\Upsilon_{r}:\bar{{\mathcal{A}}}^{i}(\hat{Z}_{r})\to\bar{{\mathcal{A}}}^{i}(\hat{Z}_{r}),\ T\mapsto T\sqcup\{\xi(r)\} the functor Υ\Upsilon for ZZ replaced by Z^r−1\hat{Z}_{r-1}.

Composition with idT⊔{[ξ(1)→ξ(r)]}\operatorname{id}\nolimits_{T}\sqcup\{[\xi(1)\to\xi(r)]\} gives an isomorphism

L(T,S)→∼Hom𝒜⁡(Z^r)(Ξr⋯Ξ1(S),ΥrΞr−1⋯Ξ1(T))=Hom𝒜⁡(Z^r)(S,T⊔{ξ(r)})L(T,S)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}\operatorname{Hom}\nolimits_{{\mathcal{A}}(\hat{Z}_{r})}(\Xi_{r}\cdots\Xi_{1}(S),\Upsilon_{r}\Xi_{r-1}\cdots\Xi_{1}(T))=\operatorname{Hom}\nolimits_{{\mathcal{A}}(\hat{Z}_{r})}(S,T\sqcup\{\xi(r)\})

for r≥1r\geq 1. Similarly, we have an isomorphism

f=(idT⊔{[ξ(1)→ξ(2)],[ξ(2)→ξ(3)]})∘−f=(\operatorname{id}\nolimits_{T}\sqcup\{[\xi(1)\to\xi(2)],[\xi(2)\to\xi(3)]\})\circ-
L⁡(T,S,2)→∼Hom𝒜⁡(Z^3)⁡(Ξ3​Ξ2​Ξ1​(S),Υ3​Υ2​Ξ1​(T))=Hom𝒜⁡(Z^3)⁡(S,T⊔{ξ⁡(2),ξ⁡(3)}).L(T,S,2)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}\operatorname{Hom}\nolimits_{{\mathcal{A}}(\hat{Z}_{3})}(\Xi_{3}\Xi_{2}\Xi_{1}(S),\Upsilon_{3}\Upsilon_{2}\Xi_{1}(T))=\operatorname{Hom}\nolimits_{{\mathcal{A}}(\hat{Z}_{3})}(S,T\sqcup\{\xi(2),\xi(3)\}).

We consider the morphism

g=(idT⊔{ξ⁡(2)}⊔{[ξ(1)→ξ(3)]})∘−g=(\operatorname{id}\nolimits_{T\sqcup\{\xi(2)\}}\sqcup\{[\xi(1)\to\xi(3)]\})\circ-
L⁡(T,S,2)→Hom𝒜⁡(Z^3)⁡(Ξ3​Ξ2​Ξ1​(S),Υ3​Υ2​Ξ1​(T))=Hom𝒜⁡(Z^3)⁡(S,T⊔{ξ⁡(2),ξ⁡(3)}).L(T,S,2)\to\operatorname{Hom}\nolimits_{{\mathcal{A}}(\hat{Z}_{3})}(\Xi_{3}\Xi_{2}\Xi_{1}(S),\Upsilon_{3}\Upsilon_{2}\Xi_{1}(T))=\operatorname{Hom}\nolimits_{{\mathcal{A}}(\hat{Z}_{3})}(S,T\sqcup\{\xi(2),\xi(3)\}).

The composition f−1∘gf^{-1}\circ g is the endomorphism τ\tau of L⁡(T,S,2)L(T,S,2).

[Uncaptioned image]

8.1.9. Action on Fukaya categories

Assume now ZZ is as in §7.4.12, so that we have an associated pair (F,S)(F,S). We sketch a construction of the 22-representation on Fukaya categories of symmetric powers of FF via Auroux’s equivalences (§7.4.12). A rigorous construction would require a general theory of partially wrapped Fukaya categories and Lagrangian correspondences.

The surface associated with the singular curve Z^r\hat{Z}_{r} of §8.1.8 can be identified with FF, with set of stops S^r\hat{S}_{r} obtained from SS by adding points z1,…,zrz_{1},\ldots,z_{r}. We have (Z^r)e​x​c=Ze​x​c⊔{z1,…,zr}(\hat{Z}_{r})_{exc}=Z_{exc}\sqcup\{z_{1},\ldots,z_{r}\} and we put ωi=ωzi\omega_{i}=\omega_{z_{i}}. We denote by z0z_{0} the point of Ze​x​c∩∂FZ_{exc}\cap\partial F such that the interval (z,z1)(z,z_{1}) of ∂F\partial F contains no point of Ze​x​cZ_{exc}.

[Uncaptioned image]

Consider a positive integer rr. We have a fully faithful functor Ξr′:ℱ⁡(Symn​F,S^r−1)→ℱ⁡(Symn​F,S^r)\Xi^{\prime}_{r}:{\mathcal{F}}(\mathrm{Sym}^{n}F,\hat{S}_{r-1})\to{\mathcal{F}}(\mathrm{Sym}^{n}F,\hat{S}_{r}) obtained by moving endpoints of Lagrangians so that they are not on the interval [zr−1,zr][z_{r-1},z_{r}] of ∂F\partial F. There is a commutative diagram where the vertical functors are Auroux’s functors:

𝒜⁡(Z^r−1,n)\textstyle{{\mathcal{A}}(\hat{Z}_{r-1},n)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Ξr\scriptstyle{\Xi_{r}}Φ\scriptstyle{\Phi}𝒜⁡(Z^r,n)\textstyle{{\mathcal{A}}(\hat{Z}_{r},n)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Φ\scriptstyle{\Phi}ℱ⁡(Symn​F,S^r−1)\textstyle{{\mathcal{F}}(\mathrm{Sym}^{n}F,\hat{S}_{r-1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Ξr′\scriptstyle{\Xi^{\prime}_{r}}ℱ⁡(Symn​F,S^r)\textstyle{{\mathcal{F}}(\mathrm{Sym}^{n}F,\hat{S}_{r})}

The Lagrangian correspondence

{({x1,…,xn},{x1,…,xn,y})|x1,…,xn∈F,y∈ωr}⊂−SymnF×Symn+1F\bigl\{(\{x_{1},\ldots,x_{n}\},\{x_{1},\ldots,x_{n},y\})\ |\ x_{1},\ldots,x_{n}\in F,\ y\in\omega_{r}\bigr\}\subset-\mathrm{Sym}^{n}F\times\mathrm{Sym}^{n+1}F

induces a functor

Υr′:ℱ¯i​(Symn​F,S^r−1)→ℱ¯i​(Symn+1​F,S^r),L↦L⊔ωr\Upsilon^{\prime}_{r}:\bar{{\mathcal{F}}}^{i}(\mathrm{Sym}^{n}F,\hat{S}_{r-1})\to\bar{{\mathcal{F}}}^{i}(\mathrm{Sym}^{n+1}F,\hat{S}_{r}),\ L\mapsto L\sqcup\omega_{r}

and there is a commutative diagram

𝒜⁡(Z^r−1,n)\textstyle{{\mathcal{A}}(\hat{Z}_{r-1},n)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Υr\scriptstyle{\Upsilon_{r}}Φ\scriptstyle{\Phi}𝒜⁡(Z^r,n+1)\textstyle{{\mathcal{A}}(\hat{Z}_{r},n+1)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Φ\scriptstyle{\Phi}ℱ¯i​(Symn​F,S^r−1)\textstyle{\bar{{\mathcal{F}}}^{i}(\mathrm{Sym}^{n}F,\hat{S}_{r-1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Υr′\scriptstyle{\Upsilon^{\prime}_{r}}ℱ¯i​(Symn+1​F,S^r)\textstyle{\bar{{\mathcal{F}}}^{i}(\mathrm{Sym}^{n+1}F,\hat{S}_{r})}

We define a bimodule

Lr′=Lr,n′:ℱ⁡(Symn​F,S)⊗ℱ​(Symn+r​F,S)opp\displaystyle L^{\prime}_{r}=L^{\prime}_{r,n}:{\mathcal{F}}(\mathrm{Sym}^{n}F,S)\otimes{\mathcal{F}}(\mathrm{Sym}^{n+r}F,S)^{\operatorname{opp}\nolimits} →k​−diff\displaystyle\to k\operatorname{\!-diff}\nolimits
λ1⊗λ2\displaystyle\lambda_{1}\otimes\lambda_{2} ↦Hom(Ξr′⋯Ξ1′(λ2),Υr′⋯Υ1′(λ1)).\displaystyle\mapsto\operatorname{Hom}\nolimits(\Xi^{\prime}_{r}\cdots\Xi^{\prime}_{1}(\lambda_{2}),\Upsilon^{\prime}_{r}\cdots\Upsilon^{\prime}_{1}(\lambda_{1})).

We put Lr′=⨁n≥0Lr,n′L^{\prime}_{r}=\bigoplus_{n\geq 0}L^{\prime}_{r,n}, a (ℱ⁡(Sym∗​F,S),ℱ⁡(Sym∗​F,S))({\mathcal{F}}(\mathrm{Sym}^{*}F,S),{\mathcal{F}}(\mathrm{Sym}^{*}F,S))-bimodule. We have an isomorphism of bimodules L⁡(−,−,r)→∼Lr′∘(Φ⊗Φ)L(-,-,r)\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}L^{\prime}_{r}\circ(\Phi\otimes\Phi).

Consider t∈Homℱ⁡(F,S^2)⁡(ω1,ω2)t\in\operatorname{Hom}\nolimits_{{\mathcal{F}}(F,\hat{S}_{2})}(\omega_{1},\omega_{2}) corresponding, via Auroux’s equivalence, to [ξ(1)→ξ(2)][\xi(1)\to\xi(2)]. Similarly, we consider the two maps u,v∈Homℱ⁡(Sym2​(F),S^3)⁡(ω1⊔ω2,ω2⊔ω3)u,v\in\operatorname{Hom}\nolimits_{{\mathcal{F}}(\mathrm{Sym}^{2}(F),\hat{S}_{3})}(\omega_{1}\sqcup\omega_{2},\omega_{2}\sqcup\omega_{3}) corresponding, via Auroux’s equivalences, to {[ξ(1)→ξ(2)],[ξ(2)→ξ(3)]}\{[\xi(1)\to\xi(2)],[\xi(2)\to\xi(3)]\} and to {idξ⁡(2),[ξ(1)→ξ(3)]}\{\operatorname{id}\nolimits_{\xi(2)},[\xi(1)\to\xi(3)]\} respectively.

[Uncaptioned image]

Composition with tt induces an isomorphism

ft:L1′​(λ1,λ2)→∼Hom⁡(Ξ2′​Ξ1′​(λ2),Υ2′​Ξ1′​(λ1)).f_{t}:L^{\prime}_{1}(\lambda_{1},\lambda_{2})\mathrel{\mathop{\kern 0.0pt\to}\limits^{\sim}}\operatorname{Hom}\nolimits(\Xi^{\prime}_{2}\Xi^{\prime}_{1}(\lambda_{2}),\Upsilon^{\prime}_{2}\Xi^{\prime}_{1}(\lambda_{1})).

Composition with uu and vv induce morphisms

fu,fv:L2′​(λ1,λ2)→Hom⁡(Ξ3′​Ξ2′​Ξ1′​(λ2),Υ3′​Υ2′​Ξ1′​(λ1)).f_{u},f_{v}:L^{\prime}_{2}(\lambda_{1},\lambda_{2})\to\operatorname{Hom}\nolimits(\Xi^{\prime}_{3}\Xi^{\prime}_{2}\Xi^{\prime}_{1}(\lambda_{2}),\Upsilon^{\prime}_{3}\Upsilon^{\prime}_{2}\Xi^{\prime}_{1}(\lambda_{1})).

The map fuf_{u} is invertible and we put τ=fu−1∘fv\tau=f_{u}^{-1}\circ f_{v}.

The composition map

L1′​(λ1,−)⊗ℱ⁡(Symn​F,S)L1′​(−,λ2)→L2′​(λ1,λ2),x⊗y↦Υ2′​(x)∘ft​(y)L^{\prime}_{1}(\lambda_{1},-)\otimes_{{\mathcal{F}}(\mathrm{Sym}^{n}F,S)}L^{\prime}_{1}(-,\lambda_{2})\to L^{\prime}_{2}(\lambda_{1},\lambda_{2}),\ x\otimes y\mapsto\Upsilon^{\prime}_{2}(x)\circ f_{t}(y)

is an isomorphism. Via this isomorphism, τ\tau defines an endomorphism of (L1′)2(L^{\prime}_{1})^{2}.

The relations (4.1.1) are satisfied because τ\tau arises from a map coming from strand algebras.

0PCV

Remark 8.1.22. Our construction is similar to the sketch provided by Douglas and Manolescu in [DouMa, §2.3].

Original mathematics by the credited authors. Source collection and HTML conversion remain in progress.

Andrew Manion, Raphael Rouquier

Original source: arXiv:2009.09627v2