This paper extends the Kricker invariant of knots in homology 3-spheres to a functorial setting and proves splitting formulas for it under null Lagrangian-preserving surgeries, generalizing previous results.
Contribution
It introduces a functorial extension of the Kricker invariant and establishes splitting formulas for this extension under a broader class of surgeries.
Findings
01
Proves splitting formulas for the extended invariant.
02
Generalizes null-move splitting formulas to null Lagrangian-preserving surgeries.
03
Enhances understanding of knot invariants in homology 3-spheres.
Abstract
Kricker defined an invariant of knots in homology 3-spheres which is a rational lift of the Kontsevich integral, and proved with Garoufalidis that this invariant satisfies splitting formulas with respect to a surgery move called null-move. We define a functorial extension of the Kricker invariant and prove splitting formulas for this functorial invariant with respect to null Lagrangian-preserving surgery, a generalization of the null-move. We apply these splitting formulas to the Kricker invariant.
AQ(t)(∅)=Q⟨AS, IHX, LE, Hol, OR⟩Q⟨Q(t)–beaded Jacobi diagrams⟩,
AQ(t)(∅)=Q⟨AS, IHX, LE, Hol, OR⟩Q⟨Q(t)–beaded Jacobi diagrams⟩,
\sum_{I\subset\{1,\dots,n\}}(-1)^{|I|}\tilde{Z}\left((S,\kappa)(\mathbf{C}_{I})\right)\equiv_{n}\left(\textrm{\begin{minipage}{170.71652pt} \begin{center} sum of all ways of gluing all legs of $\mu(\mathbf{C})$ with $\ell_{(S,\kappa)}(\mathbf{C})/2$\end{center}\end{minipage}}\right),
\sum_{I\subset\{1,\dots,n\}}(-1)^{|I|}\tilde{Z}\left((S,\kappa)(\mathbf{C}_{I})\right)\equiv_{n}\left(\textrm{\begin{minipage}{170.71652pt} \begin{center} sum of all ways of gluing all legs of $\mu(\mathbf{C})$ with $\ell_{(S,\kappa)}(\mathbf{C})/2$\end{center}\end{minipage}}\right),
w(\gamma_{i},\gamma_{j})=\left\{\begin{array}[]{l l}\displaystyle\frac{1}{2}\sum_{c}\textrm{sg}(c)t^{\varepsilon_{ij}(c)}&\textrm{ if }i\neq j\\
&\\
\displaystyle\frac{1}{2}\sum_{c}\textrm{sg}(c)(t^{\varepsilon_{ii}(c)}+t^{-\varepsilon_{ii}(c)})&\textrm{ if }i=j\end{array}\right.
w(\gamma_{i},\gamma_{j})=\left\{\begin{array}[]{l l}\displaystyle\frac{1}{2}\sum_{c}\textrm{sg}(c)t^{\varepsilon_{ij}(c)}&\textrm{ if }i\neq j\\
&\\
\displaystyle\frac{1}{2}\sum_{c}\textrm{sg}(c)(t^{\varepsilon_{ii}(c)}+t^{-\varepsilon_{ii}(c)})&\textrm{ if }i=j\end{array}\right.
Wγ=Wη−WηLWL−1WLη.
Wγ=Wη−WηLWL−1WLη.
Wη′=Wη+tPWLη+WηLP+tPWLPandWLη′=WLη+WLP,
Wη′=Wη+tPWLη+WηLP+tPWLPandWLη′=WLη+WLP,
lk(ξ′,ζ′)=lk(ξ,ζ)−Lk(ξ,L).Lk(L)−1.Lk(L,ζ).
lk(ξ′,ζ′)=lk(ξ,ζ)−Lk(ξ,L).Lk(L)−1.Lk(L,ζ).
lke(ζ,ξ)=P(t)1j∈Z∑⟨Σ,τj(ξ)⟩tj∈Q(t).
lke(ζ,ξ)=P(t)1j∈Z∑⟨Σ,τj(ξ)⟩tj∈Q(t).
w(γi,γj)
w(γi,γj)
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Full text
Splitting formulas for the rational lift of the Kontsevich integral
Delphine Moussard
Abstract
Kricker defined an invariant of knots in homology 3-spheres which is a rational lift of the Kontsevich integral
and proved with Garoufalidis that this invariant satisfies splitting formulas with respect to a surgery move called null-move.
We define a functorial extension of the Kricker invariant and prove splitting formulas for this functorial invariant
with respect to null Lagrangian-preserving surgery, a generalization of the null-move. We apply these splitting formulas
to the Kricker invariant.
This paper presents the construction of a functorial extension of the Kricker rational lift of the Kontsevich integral, which aims at expliciting
the properties of this invariant as a series of finite type invariants.
The notion of finite type invariants first appeared in independent works of Goussarov and Vassiliev involving invariants of knots in the 3–dimensional
sphere S3; in this case, finite type invariants are also called Vassiliev invariants.
Finite type invariants of knots in S3 are defined by their polynomial behaviour with respect to crossing changes.
The discovery of the Kontsevich integral, which is a universal invariant among all finite type invariants of knots in S3, revealed that this class
of invariants is very prolific. It is known, for instance, that it dominates all Witten-Reshetikhin-Turaev’s quantum invariants.
A theory of finite type invariants can be defined for any kind of topological objects provided that an elementary move on the set of these objects is fixed;
the finite type invariants are defined by their polynomial behaviour with respect to this elementary move.
For 3–dimensional manifolds, the notion of finite type invariants was introduced by Ohtsuki [Oht96], who constructed the first examples for integral
homology 3–spheres, and it has been widely developed and generalized since then. In particular, Goussarov and Habiro independently developed
a theory which involves any 3–dimensional manifolds —and their knots— and which contains the Ohtsuki theory for Z–spheres [GGP01, Hab00].
In [Kri00], Kricker constructed a rational lift of the Kontsevich integral of knots in integral homology 3–spheres (Z–spheres). In [GK04], he proved with
Garoufalidis that his construction provides an invariant of knots in Z–spheres. They also proved that the Kricker invariant satisfies some splitting
formulas with respect to the so-called null-move. For knots in Z–spheres with trivial Alexander polynomial, these formulas together with work of Garoufalidis
and Rozansky [GR04] imply that the Kricker invariant is a universal finite type invariant with respect to the null-move.
Kricker’s construction easily
generalizes to null-homologous knots in rational homology 3–spheres (Q–spheres); the main goal of this article is to prove splitting formulas for the Kricker
invariant of these knots with respect to null Lagrangian-preserving surgery, a move which generalizes the null-move.
For null-homologous knots in Q–spheres with trivial Alexander polynomial, these formulas are used in [Mou17] to prove that this extended Kricker
invariant is a universal finite type invariant with respect to null Lagrangian-preserving surgeries.
Lescop defined in [Les11] an invariant of null-homologous knots in Q–spheres and proved in [Les13] splitting formulas for this invariant
with respect to null Lagrangian-preserving surgeries, similar to the ones proved in this paper for the Kricker invariant. Lescop conjectured in [Les13] that her invariant
is equivalent to the Kricker invariant. The mentioned results of Garoufalidis, Kricker, Lescop and Rozansky give such an equivalence for knots in Z–spheres
with trivial Alexander polynomial and the results of the present paper allow to generalize this equivalence to null-homologous knots in Q–spheres with
trivial Alexander polynomial [Mou17, Theorem 2.11].
A similar situation arises in the study of finite type invariants of Q–spheres with respect to Lagrangian-preserving surgeries. In this case,
the Kontsevich–Kuperberg–Thurston invariant and the Le–Murakami–Ohtsuki invariant are both universal, up to degree 1 invariants deduced from the cardinality
of the first homology group; this implies an equivalence result for these two invariants, see [Mou12]. For the KKT invariant, splitting formulas with respect
to Lagrangian-preserving surgeries were proved by Lescop [Les04]; for the LMO invariant, similar formulas were proved by Massuyeau [Mas15].
Massuyeau’s proof of his splitting formulas is based on an extension of the LMO invariant of Q–spheres to a functor defined on a category of Lagrangian
cobordisms that he constructed with Cheptea and Habiro [CHM08].
In this paper, we extend the LMO functorial invariant of Cheptea–Habiro–Massuyeau to a category of Lagrangian cobordisms with paths, inserting the Kricker’s
idea in the construction. We obtain a functorial invariant from which the Kricker invariant of null-homologous knots in Q–spheres
is recovered. Following Massuyeau, we use the functoriality to obtain splitting formulas for our invariant and, as a consequence,
for the Kricker invariant.
Notations and conventions.
For K=Z,Q, a K–sphere, (resp. a K–cube) is a 3–manifold, compact and oriented, which has the same homology with coefficients
in K as the standard 3–sphere (resp. 3–cube). The boundary of an oriented manifold with boundary is oriented with the “outward normal first” convention.
Acknowledgments.
I am supported by a Postdoctoral Fellowship of the Japan Society for the Promotion of Science. I am grateful to Tomotada Ohtsuki and the Research Institute for Mathematical Sciences for their support. I also wish to thank Gwénaël Massuyeau for interesting exchanges. Finally, I thank the referee whose useful comments helped to improve the paper.
1.2 Statement of the main result
We first give the definitions we need to state our main result.
Null LP–surgeries.
For g∈N, a genus g rational homology handlebody (Q–handlebody) is a 3–manifold which is compact, oriented,
and which has the same homology with rational coefficients as the standard genus g handlebody.
Such a Q–handlebody is connected, and its boundary is necessarily homeomorphic to the standard genus g surface.
The LagrangianLC of a Q–handlebody C is the kernel of the map i∗:H1(∂C;Q)→H1(C;Q)
induced by the inclusion. The Lagrangian of a Q–handlebody C is indeed a Lagrangian subspace of H1(∂C;Q)
with respect to the intersection form. A Lagrangian-preserving pair, or LP–pair, is a pair C=(CC′)
of Q–handlebodies equipped with a homeomorphism h:∂Cto31.29802pt\rightarrowfill≅∂C′ such that h∗(LC)=LC′.
Given a 3–manifold M, a Lagrangian-preserving surgery, or LP–surgery, on M is a family C=(C1,…,Cn)
of LP–pairs such that the Ci are embedded in M and disjoint. The manifold obtained from M by LP–surgery on C is defined as
[TABLE]
Let M be a 3–manifold such that H1(M;Q)=0 and let K be a disjoint union of knots or paths properly embedded in M. A Q–handlebody null in M∖K is a Q–handlebody C⊂M∖K such that the map i∗:H1(C;Q)→H1(M∖K;Q) induced by the inclusion has a trivial image.
A null LP–surgery on (M,K) is an LP–surgery C=(C1,…,Cn) on M∖K such that each Ci is null in M∖K.
The pair obtained by surgery is denoted by (M,K)(C).
The tensor μ(C).
Given an LP–pair C=(CC′), define the associated total manifoldC=(−C)∪C′ and define
[TABLE]
by associating with a triple of cohomology classes the evaluation of their triple cup product on the fundamental class of C.
For a family C=(C1,…,Cn) of LP–pairs, let C=C1⊔⋯⊔Cn and set:
[TABLE]
The natural identification H1(C;Q)≅⊕i=1nH1(Ci;Q) allows to see ⊗i=1nΛ3H1(Ci;Q) as a subspace of (Λ3H1(C;Q))⊗n. This subspace injects to SnΛ3H1(C;Q)via the canonical surjection (Λ3H1(C;Q))⊗n↠SnΛ3H1(C;Q). Hence we can view μ(C) as an element of SnΛ3H1(C;Q).
The bilinear form ℓ(S,κ)(C).
Let (S,κ) be a QSK–pair, i.e. a pair made of a Q–sphere S and a null-homologous knot κ⊂S.
Let C=(C1,…,Cn) be a null LP–surgery on (S,κ). Let C=C1⊔⋯⊔Cn
be the disjoint union of the associated total manifolds. Fix a lift C~i of each Ci in E~. We will define a hermitian form:
[TABLE]
i.e. a Q–bilinear form such that reversing the order of the arguments changes t to t−1. Let a∈H1(Ci;Q) and b∈H1(Cj;Q)
be homology classes that can be represented by simple closed curves α⊂∂Ci and β⊂∂Cj, disjoint if i=j.
Note that such homology classes generate H1(C;Q) over Q.
Let α~ and β~ be the copies of α and β in C~i and C~j. Set:
[TABLE]
where lke(⋅,⋅) stands for the equivariant linking number (see for instance [Mou17, Section 2.1] for a definition).
We get a well-defined hermitian form ℓ(S,κ)(C) associated with a choice of lifts of the Ci’s. We will keep this choice implicit;
the statement of Theorem 1.1 is valid for any such choice.
Diagrammatic representations.
Let V be a rational vector space. A V–colored Jacobi diagram is a unitrivalent graph whose trivalent vertices are oriented and whose univalent
vertices are labelled by V, where an orientation of a trivalent vertex is a cyclic order of the three edges that meet at this vertex —fixed
as
A symmetric tensor in SnΛ3V can be represented by a Jacobi diagram via the following embedding.
[TABLE]
Now define a Q(t)–beaded Jacobi diagram as a trivalent graph whose vertices are oriented and whose edges are oriented and labelled by Q(t).
Set:
[TABLE]
where the relations are depicted in Figures 1 and 2,
with the IHX relation defined with the central edge labelled by 1.
Define the i–degree, or internal degree, of any Jacobi diagram as its number of trivalent vertices. Given a hermitian form ℓ:V×V→Q(t),
one can glue with ℓ some legs of a V–colored Jacobi diagram as depicted in Figure 3.
If n is even, one can pairwise glue all legs of a V–colored Jacobi diagram of i–degree n in order to
get an element of AQ(t)(∅). This latter space is the target space of the Kricker invariant of QSK–pairs.
We can now state our main result, about the Kricker invariant Z~, proved in Section 6. Note that null LP–surgeries define a move on the set of QSK–pairs.
Theorem 1.1**.**
Let (S,κ) be a QSK–pair. Let C=(C1,…,Cn) be a null LP–surgery on (S,κ). Then:
[TABLE]
where ≡n means “equal up to terms of i–degree at least n+1”.
Example.
Let (S3,O) be the QSK–pair defined by the trivial knot O in the standard 3–sphere. Let C1 and C2 be regular neighborhoods of the graphs Γ1 and Γ2 drawn in Figure 4.
One can define an LP–surgery C=(C1,C2) by associating with each Γi a Borromean surgery, see for instance [Mou17, Section 2.2]. The associated tensor is given by μ(C1)=ζ1∧ζ2∧ζ3 and μ(C2)=ξ1∧ξ2∧ξ3. There are fifteen ways to glue all legs of
[TABLE]
with 21ℓ(S3,O)(C); all associated diagrams but one are trivial by the relation LE since they have a trivially labelled edge. Now lke(ζ1,ξ1)=1, lke(ζ2,ξ2)=1 and lke(ζ3,ξ3)=t−1, so that we finally get:
[TABLE]
where the vanishing of two terms in the left hand side is due to [GGP01, Lemma 2.2].
1.3 Strategy
In this Subsection, we give a rough overview of the strategy developed to prove Theorem 1.1.
The main object of this article is the construction of a functorial LMO invariant defined on a category of Lagrangian cobordisms with paths.
The morphisms of this category are cobordisms between compact surfaces with one boundary component, satisfying a Lagrangian-preserving condition,
with finitely many disjoint paths with fixed extremities which we think of as knots with a fixed part on the boundary.
This category is equivalent to a category of bottom-top tangles in Q–cubes, whose top part has a trivial linking matrix, with paths with fixed extremities.
These bottom-top tangles can be viewed as morphisms in a category of (general) tangles with paths in Q–cubes, with an important difference in the composition law.
Now a tangle with paths in a Q–cube can be expressed as the result of a surgery on a link in a tangle with trivial paths —segment lines— in [−1,1]3.
To sum up, with a Lagrangian cobordism with paths, we associate a tangle with disks —whose boundaries define the paths— in [−1,1]3 with a surgery link.
This is represented in the first line of the scheme in Figure 5. We initiate the construction of the invariant at the “tangle with disks” level.
On the above mentioned categories, we define functorial invariants valued in categories of Jacobi diagrams with beads, i.e. unitrivalent graphs whose
univalent vertices are labelled by some finite set or embedded in some 1–manifold —the skeleton— and whose edges are labelled (beaded) by powers of t,
polynomials in Q[t±1] or rational functions in Q(t). At the first step, we define a functor Z∙ on the category of tangles with disks by applying
the Kontsevich integral and adding a bead t±1 on the skeleton when the corresponding component meets a disk of the tangle. At a second step,
we apply the invariant Z∙ to surgery presentations of tangles with paths in Q–cubes. We use the formal Gaussian integration methods introduced by
Bar-Natan, Garoufalidis, Rozansky and Thurston in [BNGRT02a, BNGRT02b] and adapted to the beaded setting in [Kri00, GK04]. We get a functor Z on the category
of tangles with paths in Q–cubes. At the last step, given a Lagrangian cobordism with paths, we apply Z to the associated bottom-top tangles with paths
and normalize it following [CHM08] to obtain a functor Z~ on the category of Lagrangian cobordisms with paths. Functoriality allows to prove
splitting formulas for this invariant with respect to null Lagrangian-preserving surgeries.
Given a Lagrangian cobordism with one path between genus 0 surfaces, i.e. a Q–cube with one path, one can glue a 3–ball to the boundary to get
a Q–sphere with a knot. In this way, the functor Z~ provides an invariant of QSK–pairs which coincides with the Kricker
invariant for knots in Z–spheres. Splitting formulas for this invariant are deduced from the splitting formulas for the functor Z~.
Plan of the paper.
We define the domain categories of cobordisms and tangles in Section 2. In Section 3, we define the target categories of
Jacobi diagrams and gives the tools of formal Gaussian integration. Section 4 is devoted to the introduction of winding matrices, that will
play the role of the linking matrices in the Cheptea–Habiro–Massuyeau construction of a functorial LMO invariant. The functors Z∙, Z and Z~ are constructed
in Section 5. At the end of this section, from the functor Z~, we deduce our version of the Kricker invariant for QSK–pairs;
the behaviour of this invariant with respect to connected sum is stated. Finally, the splitting formulas are given in Section 6.
2 Domain categories: cobordisms and tangles
2.1 Cobordisms with paths
Given g∈N, we fix a model surface Fg, compact, connected, oriented, of genus g, with one boundary
component represented in Figure 6.
It is equipped with a fixed base point ∗ and a fixed basis (α1,…,αg,β1,…,βg) of π1(Fg,∗).
Denote by Cg−g+ the cube [−1,1]3 with g+ handles on the top
boundary and g− tunnels in the bottom boundary. We have canonical embeddings Fg+↪∂Cg−g+ and Fg−↪∂Cg−g+. For k≥0 and 1≤i≤k, set hi(k)=k+12i−1. A cobordism with paths from Fg+ to Fg− is an equivalence class of triples (M,K,m) where:
•
M is a compact, connected, oriented 3–manifold,
•
m:∂Cg−g+to31.29802pt\rightarrowfill≅∂M is an orientation-preserving homeomorphism,
•
K=⊔i=1kKi⊂M is a union of k≥0 oriented paths Ki from m(0,1,hi(k)) to m(0,−1,hi(k)),
•
K^=⊔i=1kK^i is an oriented boundary link, i.e. the K^i bound disjoint compact oriented surfaces in M, where K^i is the knot defined as the union of Ki with the line segments [(0,−1,hi(k)),(1,−1,hi(k))], [(1,−1,hi(k)),(1,1,hi(k))] and
[(1,1,hi(k)),(0,1,hi(k))].
Two such triples are equivalent if they are related by an orientation-preserving homeomorphism which respects the boundary parametrizations
and identifies the paths. We get embeddings m+:Fg+↪∂M and m−:Fg−↪∂M.
Define a category Cob of cobordisms with paths whose objects are non-negative integers and whose set of morphisms Cob(g+,g−) is the set of
cobordisms with paths from Fg+ to Fg−. The composition of a cobordism (M,K,m) from Fg to Ff with a cobordism (N,J,n) from Fh to Fg
is given by gluing N on the top of M, identifying m+(M) with n−(N) and reparametrizing the new manifold. The identity of g∈N is the cobordism
Fg×[−1,1] with natural boundary parametrization and no path.
Forgetting the datum of the paths in the cobordisms, one gets the category Cob described in [CHM08], which we view as the subcategory of Cob
of cobordisms with no path. For a cobordism (M,m) and a cobordism with paths (N,J,n), define the tensor product (M,m)⊗(N,J,n) by horizontal
juxtaposition in the x direction.
We now define the subcategory of Cob of Lagrangian cobordisms with paths. Set Ag=ker(incl∗:H1(Fg;Q)→H1(C0g;Q))
and Bg=ker(incl∗:H1(Fg;Q)→H1(Cg0;Q)). These are Lagrangian subspaces of H1(Fg;Q) with respect to the intersection form
and Ag (resp. Bg) is generated by the homology classes of the curves αi (resp. βi).
A cobordism with paths (M,K,m) from Fg+ to Fg− is Lagrangian(-preserving) if the following conditions are satisfied:
•
H1(M;Q)=(m−)∗(Ag−)⊕(m+)∗(Bg+),
•
(m+)∗(Ag+)⊂(m−)∗(Ag−) as subspaces of H1(M;Q).
The Lagrangian property is preserved by composition, and we denote by LCob the subcategory of Cob of Lagrangian cobordisms with paths.
The subcategory of Lagrangian cobordisms with no path is the category LCob —denoted by QLCob in [Mas15].
Define categories Cobq, Cobq, LCobq and LCobq of q–cobordisms with objects the non-commutative words in the single letter ∙ and with set of morphisms from a word on g+ letters to a word on g− letters the set of morphisms from g+ to g− in Cob, Cob, LCob and LCob respectively.
Let (M,K,m) be a Lagrangian cobordism with paths from g+ to g−. Since the space H1(M;Q) is non-trivial in general, we have to adapt the definition of null LP–surgeries given in the introduction. Let N(K) be a tubular neighborhood of K in M and set E=M∖Int(N(K)). A standard homological computation gives H2(M;Q)=0, so that the exact sequence in homology associated with the pair (M,E) provides the following short exact sequence:
[TABLE]
The image of H2(N(K),N(K)∩E;Q) in H1(E;Q) is a subspace HK≅Qk generated by meridians of the components of K, where k is the number of these components. Now, the parametrizations m± of the top and bottom boundaries of M can be decomposed into injective maps as follows.
E$$F_{g^{\pm}}$$M$$m_{\pm}$$e_{\pm}
Hence we have a canonical decomposition of H1(E;Q) as
[TABLE]
where (e−)∗(Ag−)⊕(e+)∗(Bg+)≅H1(M;Q)via the inclusion map.
We say that a Q–handlebody C⊂M∖K is null with respect to K if the composed map
[TABLE]
has a trivial image, where the second map is the projection on the first factor in the above decomposition of H1(M∖K;Q)≅H1(E;Q).
A null LP–surgery on (M,K) is an LP–surgery C=(C1,…,Cn) on M∖K such that each Ci is null with respect to K.
2.2 Bottom-top tangles with paths
Let us define the category of bottom-top tangles with paths. For a positive integer g≥0, let (p1,q1),…,(pg,qg) be g pairs of points
uniformly distributed on [−1,1]×{0}⊂[−1,1]2≅F0 as represented in Figure 8.
A bottom-top tangle with paths of type (g+,g−)
is an equivalence class of triples (B,K,γ) where:
•
(B,K)=(B,K,b) is a cobordism with paths form F0 to F0,
•
γ=(γ+,γ−) is a framed oriented tangle in B with g+ components γi+ from b({pi}×{1}) to b({qi}×{1})
and g− components γi− from b({qi}×{−1}) to b({pi}×{−1}),
•
K^ is a boundary link in B∖γ.
Two such triples (B,K,γ) and (B′,K′,γ′) are equivalent if (B,K) and (B′,K′)
are related by an equivalence which identifies γ and γ′.
In order to define the composition, we need the bottom-top tangle ([−1,1]3,∅,Tg) represented in Figure 9.
The composition of a bottom-top tangle
(B,K,γ) of type (g,f) with a bottom-top tangle (C,J,υ) of type (h,g) is given by first making the composition
(B,K)∘([−1,1]3,∅)∘(C,J) in the category Cob and then perfoming the surgery on the 2g components link γ+∪Tg∪υ−.
We get a category btT whose objects are non-negative integers and whose set of morphisms btT(g+,g−) is the set of bottom-top tangles with
paths of type (g+,g−). The identity of g∈N is the bottom-top tangle in [−1,1]3 with no path represented in Figure 10.
Forgetting the datum of the paths, one gets the category btT of bottom-top tangles introduced in [CHM08], that we view as the subcategory of btT
of bottom-top tangles with no path. For a bottom-top tangle (B,γ)
and a bottom-top tangle with paths (C,J,υ), define the tensor product (B,γ)⊗(C,J,υ) by horizontal juxtaposition in the x direction.
Define categories btTq and btTq of bottom-top q–tangles with objects the non-commutative words in the single letter ∙.
The following result is a direct adaptation of [CHM08, Theorem 2.10] which gives an isomorphism D:btT→Cob. The map D is defined by digging
tunnels around the components of the tangle.
Proposition 2.1**.**
There is an isomorphism D:btT→Cob which identifies btT with Cob and preserves the tensor product on btT⊗btT.
Let (B,K,γ) be a bottom-top tangle with paths in a Q–cube. Let γˉ be the link obtained by closing the components of γ
with the line segments [(pi,±1),(qi,±1)]. Define the linking matrix Lk(γ) of γ in B with the linkings of the components
of γˉ. The characterization of the bottom-top tangles sent onto Lagrangian cobordisms by D
given in [CHM08, Lemma 2.12] directly generalizes to:
Lemma 2.2**.**
Given a bottom-top tangle with paths (B,K,γ), the cobordism with paths D(B,K,γ) is Lagrangian if and only if B is a Q–cube
and Lk(γ+) is trivial.
2.3 Tangles with paths and tangles with disks
Given a cobordism (B,b) from F0 to F0, a tangle γ in B is an isotopy (rel. ∂B) class of framed oriented tangles whose boundary
points lie on the top and bottom surfaces and are uniformly distributed along the line segments [−1,1]×{0}×{1} and [−1,1]×{0}×{−1}
in ∂B=b(∂[−1,1]3).
Associate with each boundary point of γ the sign + if γ is oriented downwards at that point and the sign − otherwise. This provides two words
in the letters + and −, one for the top surface and the other for the bottom surface. Lifting these two words into non-associative words wt(γ)
and wb(γ) in the letters (+,−), one gets a q–tangle. A q–tangle γ in a cobordism with paths (B,K,b) defines
a q–tangle with paths(B,K,γ) if K^ is a boundary link in B∖γ.
Define two categories TqCub and TqCub with objects the non-associative words in the letters (+,−) and morphisms the q–tangles in Q–cubes for TqCub
and the q–tangles with paths in Q–cubes for TqCub, up to orientation-preserving homeomorphism respecting the boundary parametrization.
Composition is given by vertical juxtaposition. Given a morphism (C,υ) in TqCub and a morphism (B,K,γ) in TqCub,
define the tensor product (C,υ)⊗(B,K,γ) by horizontal juxtaposition in the x direction.
Lemma 2.3**.**
Let (B,K,γ) be a q–tangle with paths in a Q–cube. There exist a q–tangle with paths ([−1,1]3,Ξ,η) and a framed link
L⊂[−1,1]3∖(Ξ∪η), with Ξ a union of line segments and L null-homotopic in [−1,1]3∖Ξ,
such that (B,K,γ) is obtained from ([−1,1]3,Ξ,η) by surgery on L. Moreover, two such surgery links
are related by the following Kirby moves: the blow-up/blow-down move KI which adds or removes a split trivial component with framing ±1 unknotted with η,
and the handleslide move KII which adds a surgery component to another surgery component or to a component of the tangle η (see Figure 11).
Proof.
Let Σ be a Seifert surface of K^ which is the disjoint union of Seifert surfaces of the K^i. Choose Σ disjoint from γ.
Take a link J⊂B such that surgery on J gives [−1,1]3. Performing isotopies on J if necessary, we can assume that J does not meet Σ.
The handles of (the image in [−1,1]3 of) Σ can be unlinked
by adding surgery components as shown in Figure 12.
In this way, K^ can be turned into a trivial link. This provides a surgery link in B∖(K∪γ), disjoint
from Σ, such that surgery on this link changes (B,K,γ) to a q–tangle with paths ([−1,1]3,Ξ,η)
as required. Let L be an inverse surgery link. It is null-homotopic in [−1,1]3∖Ξ since it is disjoint from Σ.
For the last assertion, apply [HW14, Theorem 3.1] in [−1,1]3∖Ξ. Note that a split trivial component with framing ±1 can always be unknotted
from η using the KII move.
∎
A family (([−1,1]3,Ξ,η),L) satisfying the conditions of the lemma with L oriented is a surgery presentation of (B,K,γ).
When γ (and thus η) is a bottom-top tangle, the components of η can be closed by line segments in the top and bottom surfaces. The obtained
curves are null-homotopic in [−1,1]3∖Ξ since Ξ^ is a boundary link in [−1,1]3∖η.
The notion of a q–tangle in [−1,1]3 with trivial paths, i.e. line segments, is equivalent to the following one.
A q–tangle with disks is an equivalence class of pairs (γ,k),
where γ is a q–tangle in [−1,1]3, k is a non-negative integer understood as the datum of k disks di=[0,1]×[−1,1]×{hi(k)} and the link ⊔i=1k∂di associated with the paths di∂={0}×[−1,1]×{hi(k)} is a boundary link in [−1,1]3∖γ. An example of such a tangle with disks is drawn in Figure 13, projected in the y direction.
Equivalence of such pairs is defined as isotopy relative to (∂[−1,1]3)∪(∪i=1kdi∂).
Define two categories Tq and Tq with objects the non-associative words in the letters (+,−) and morphisms the q–tangles for Tq and the
q–tangles with disks for Tq. Composition is given by vertical juxtaposition. Given a q–tangle γ and a q–tangle with disks (υ,k),
define the tensor product γ⊗(υ,k) in Tq((wt(γ))(wt(υ)),(wb(γ))(wb(υ))) by horizontal juxtaposition in the x direction.
Define similarly two categories T and T of tangles and tangles with disks in [−1,1]3 with objects the associative words
in the letters (+,−).
3 Target categories: Jacobi diagrams with beads
3.1 Diagram spaces
For a compact oriented 1–manifold X and a finite set C, a Jacobi diagram on (X,C) is a unitrivalent graph whose trivalent vertices are oriented
and whose univalent vertices are embedded in X or labelled by C, where an orientation of a trivalent vertex is a cyclic order of the three edges that
meet at this vertex —fixed as
in the pictures. The manifold X is called the skeleton of the diagram. Next, let R be the ring Q[t±1] or Q(t).
An R–beaded Jacobi diagram on (X,C) is a Jacobi diagram on (X,C) whose graph edges are oriented and labelled by R.
Last, an R–winding Jacobi diagram on (X,C) is an R–beaded Jacobi diagram on (X,C) whose skeleton is viewed as a union of edges —defined
by the embedded vertices— that are labelled by powers of t, with the condition that the product of the labels on each component of X is 1.
As defined in the introduction, the i–degree of a trivalent diagram is its number of trivalent vertices.
Set:
[TABLE]
[TABLE]
[TABLE]
with the relations in Figures 1, 2 and 14, where the IHX relation for beaded and winding
diagrams is defined with the central edge labelled by 1.
In the STU relation, the edges corresponding to each other have the same orientation and label.
In the pictures, the skeleton is represented with full lines and the graph with dashed lines. We indeed consider the i–degree completion of these vector spaces,
keeping the same notation.
Remark*.*
For diagrams in ARw(X,∗C), the condition on the labels on the skeleton implies that all labels can be pushed off each component of the skeleton using the Holw relation. When the component is an interval, there is a unique way to do so. Hence, when X contains only intervals, ARw(X,∗C) is isomorphic to AR(X,∗C).
For a finite set S, denote by \leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@lineto0.0pt10.92093pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.00.0pt10.92093pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS (resp. \leavevmodeto9.51pt\vboxto9.51pt\pgfpicture\makeatletter\lower-4.75253ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@moveto4.55254pt0.0pt\pgfsys@curveto4.55254pt2.51433pt2.51433pt4.55254pt0.0pt4.55254pt\pgfsys@curveto-2.51433pt4.55254pt-4.55254pt2.51433pt-4.55254pt0.0pt\pgfsys@curveto-4.55254pt-2.51433pt-2.51433pt-4.55254pt0.0pt-4.55254pt\pgfsys@curveto2.51433pt-4.55254pt4.55254pt-2.51433pt4.55254pt0.0pt\pgfsys@closepath\pgfsys@moveto0.0pt0.0pt\pgfsys@stroke\pgfsys@invoke\pgfsys@moveto4.55254pt0.85355pt\pgfsys@lineto4.55254pt0.67792pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.04.55254pt0.67792pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS) the manifold made of ∣S∣ intervals (resp. circles) indexed by the elements of S.
In the following, Aˉ stands for A, AR or ARw. In [BN95, Theorem 8], Bar-Natan defines a formal PBW isomorphism:
[TABLE]
For a Jacobi diagram D, the image χS(D) is the average of all possible ways to attach the s–colored vertices of D on the corresponding s–indexed
interval in \leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@lineto0.0pt10.92093pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.00.0pt10.92093pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS for each s∈S. The setting of [BN95] is not exactly the same, but the argument adapts directly.
When ∣S∣=1, closing the S–labelled component gives an isomorphism from Aˉ(X∪\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@lineto0.0pt10.92093pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.00.0pt10.92093pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS,∗C) to Aˉ(X∪\leavevmodeto9.51pt\vboxto9.51pt\pgfpicture\makeatletter\lower-4.75253ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@moveto4.55254pt0.0pt\pgfsys@curveto4.55254pt2.51433pt2.51433pt4.55254pt0.0pt4.55254pt\pgfsys@curveto-2.51433pt4.55254pt-4.55254pt2.51433pt-4.55254pt0.0pt\pgfsys@curveto-4.55254pt-2.51433pt-2.51433pt-4.55254pt0.0pt-4.55254pt\pgfsys@curveto2.51433pt-4.55254pt4.55254pt-2.51433pt4.55254pt0.0pt\pgfsys@closepath\pgfsys@moveto0.0pt0.0pt\pgfsys@stroke\pgfsys@invoke\pgfsys@moveto4.55254pt0.85355pt\pgfsys@lineto4.55254pt0.67792pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.04.55254pt0.67792pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS,∗C) [BN95, Lemma 3.1]. However, this isomorphism does not hold for ∣S∣>1. To recover an isomorphism onto Aˉ(X∪\leavevmodeto9.51pt\vboxto9.51pt\pgfpicture\makeatletter\lower-4.75253ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@moveto4.55254pt0.0pt\pgfsys@curveto4.55254pt2.51433pt2.51433pt4.55254pt0.0pt4.55254pt\pgfsys@curveto-2.51433pt4.55254pt-4.55254pt2.51433pt-4.55254pt0.0pt\pgfsys@curveto-4.55254pt-2.51433pt-2.51433pt-4.55254pt0.0pt-4.55254pt\pgfsys@curveto2.51433pt-4.55254pt4.55254pt-2.51433pt4.55254pt0.0pt\pgfsys@closepath\pgfsys@moveto0.0pt0.0pt\pgfsys@stroke\pgfsys@invoke\pgfsys@moveto4.55254pt0.85355pt\pgfsys@lineto4.55254pt0.67792pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.04.55254pt0.67792pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS,∗C), some “link relations” were introduced in [BNGRT02b, Section 5.2].
We recall these relations and introduce additional “winding relations”.
Given a (beaded, winding) Jacobi diagram D on (X,C∪S) and a univalent vertex ∗ of D labelled by s∈S, define the associated link relation
as the vanishing of the sum of all diagrams obtained from D by gluing the vertex ∗ on the edges adjacent to a univalent s–labelled vertex, as follows:
\scriptscriptstyle{\bullet}$$\scriptscriptstyle{\bullet}$$*$$s
,
see Figure 15 (we omit the orientation of the edges when it is not relevant thanks to the OR relation).
Given a winding Jacobi diagram D on (X,C∪S), a label s∈S and an integer k, the associated winding relation identifies D
with the diagram obtained from D by pushing tk at each s–labelled vertex, i.e. by multiplying the label of each edge adjacent
to a univalent s–labelled vertex by tk if the orientation of the edge goes backward the vertex and by t−k otherwise, see Figure 16.
Denote Aˉ(X,∗C,*⃝S) (resp. Aˉ(X,∗C,*◯⃝S)) the quotient of Aˉ(X,∗C∪S) by all link relations (resp. all link and winding relations)
on S–labelled vertices. Note that if X contains no closed component, then the spaces ARw(X,∗C,*◯⃝S) and AR(X,∗C,*◯⃝S) are isomorphic.
When some of the sets X, C, S are empty, we simply drop the corresponding notation, mentionning ∅ only when they are all empty.
Proposition 3.1**.**
The isomorphisms χS:Aˉ(X,∗C∪S)to31.29802pt\rightarrowfill≅Aˉ(X∪\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@lineto0.0pt10.92093pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.00.0pt10.92093pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS,∗C) descend to isomorphisms:
[TABLE]
[TABLE]
[TABLE]
Proof.
In the case Aˉ=A or AR, it is [BNGRT02b, Theorem 3]. We recall briefly their argument in order to add the consideration of the winding relations
when Aˉ=ARw.
The fact that the images by χS of the link relations map to 0 in Aˉ(X∪\leavevmodeto9.51pt\vboxto9.51pt\pgfpicture\makeatletter\lower-4.75253ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@moveto4.55254pt0.0pt\pgfsys@curveto4.55254pt2.51433pt2.51433pt4.55254pt0.0pt4.55254pt\pgfsys@curveto-2.51433pt4.55254pt-4.55254pt2.51433pt-4.55254pt0.0pt\pgfsys@curveto-4.55254pt-2.51433pt-2.51433pt-4.55254pt0.0pt-4.55254pt\pgfsys@curveto2.51433pt-4.55254pt4.55254pt-2.51433pt4.55254pt0.0pt\pgfsys@closepath\pgfsys@moveto0.0pt0.0pt\pgfsys@stroke\pgfsys@invoke\pgfsys@moveto4.55254pt0.85355pt\pgfsys@lineto4.55254pt0.67792pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.04.55254pt0.67792pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS,∗C) follows from the STU relation. For the winding relations, it follows
from the Holw relation applied at each univalent vertex glued on the s–labelled component, where s is the label involved in the relation.
Now, take two diagrams in Aˉ(X∪\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@lineto0.0pt10.92093pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.00.0pt10.92093pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureS,∗C) that are identified when closing an S–labelled component. We have to consider the two situations
depicted in Figures 17 and 18, where the gray zone represents a hidden part of the diagram.
Equalities are obtained by applying STU relations in the first case and Holw relations in the second case.
Application of χ−1 to the right members
provides linear combinations of the same sum with the skeleton component dropped and possibly trees glued. Using the IHX relation in the first case
(resp. the Hol relation in the second case), we obtain link relations (resp. winding relations).
∎
3.2 Product and coproduct
We first define a coproduct on the diagram spaces of the previous subsection. Given a (beaded, winding) Jacobi diagram D on (X,C), denote by D... its graph part, and by D...i, i∈I, the connected components of D.... Set DJ=D∖(⊔i∈I∖JD...i). In the winding case, multiply the labels of the concatenated edges of the skeleton. Define the coproduct of a diagram D by
[TABLE]
Note that the different relations on Jacobi diagrams respect the coproduct. This provides a notion of group-like elements, i.e. elements G such that Δ(G)=G⊗G.
Set Aˉ=A or AR. We will define a Hopf algebra structure on Aˉ(∗C). Define the product of two diagrams as the disjoint union.
The unit ϵ:Q→Aˉ(∗C) is defined by ϵ(1)=∅ and the counit ε:Aˉ(∗C)→Q is given by
ε(D)=0 if D=∅ and ε(∅)=1. The antipode is given by D↦(−1)∣I∣D. We finally have a structure of a
graded Hopf algebra on Aˉ(∗C), where the grading is given by the i–degree. It is known that an element in a graded Hopf algebra is group-like if and only if it is the exponential of a primitive element,
i.e. an element G such that Δ(G)=1⊗G+G⊗1. Here, the primitive elements are the series of connected diagrams.
The isomorphisms χ of the previous subsection are not algebra morphisms, but they preserve the coproduct. For the spaces Aˉ(∗C) with Aˉ=A or AR, we have an exponential map associated with the product. For general (beaded, winding) Jacobi diagrams, we will use the notation exp⊔, namely exponential with respect to the disjoint union, for linear combination of diagrams with no univalent vertex embedded in the skeleton, where the disjoint union applies only to the graph part.
3.3 Formal Gaussian integration
This part aims at defining a formal Gaussian integration along S on ARw(X,∗C∪S).
Definition 3.2**.**
A (beaded, winding) Jacobi diagram on (X,C∪S) is substantial if it has no strut, i.e. no isolated dashed edge.
It is S–substantial if it has no S–strut, i.e. no strut with both vertices labelled in S.
Given two (beaded, winding) Jacobi diagrams D and E on (X,C∪S), one of whose is S–substantial, define ⟨D,E⟩S as the sum
of all diagrams obtained by gluing all s–colored vertices of D with all s–colored vertices of E for all s∈S —if the numbers of s–colored vertices in D and E do not match for some s∈S, then ⟨D,E⟩S=0. In the beaded and winding cases, we must precise the orientation
and label of the created edges. Such an edge is the gluing of two or three edges in the initial diagrams. Fix arbitrarily the orientation of the new edge.
Let P(t) (resp. Q(t)) be the product of the labels of the initial edges whose orientation coincides (resp. does not coincide). Define the label of the new edge
as P(t)Q(t−1), see Figure 19.
We have the following immediate lemma.
Lemma 3.3**.**
If D′ and E′ are obtained from D and E by applying the same winding relation on s–labelled vertices for some s∈S,
then ⟨D,E⟩S=⟨D′,E′⟩S.
This bracketting defines a Q–bilinear operator on the diagram spaces Aˉ(X,∗C∪S) for Aˉ=A,AR, or ARw.
Theorem 3.4** (Jackson, Moffatt, Morales).**
Assume the 1–manifold X is a disjoint union of intervals. If G and H are group-like in Aˉ(X,∗C∪S), then ⟨G,H⟩S is also group-like.
Proof.
When X=∅, it is [JMM06, Theorem 2.4]. The case of a non-empty X follows since the isomorphism χπ0(X):Aˉ(∗π0(X)∪C∪S)to31.29802pt\rightarrowfill≅Aˉ(X,∗C∪S) preserves the coproduct and the bracketting ⟨⋅,⋅⟩S.
∎
Notation*.*
If W=(Wij(t))i,j∈S is an (S,S)–matrix with coefficients in Q(t), we also denote W=∑i,j∈S
An element G∈AQ[t±1]w(X,∗C∪S) is Gaussian if G=exp⊔(21W(t))⊔H where W(t) is an (S,S)–matrix with coefficients in Q[t±1]
and H is S–substantial. If det(W(t))=0, G is non-degenerate and we set:
[TABLE]
Lemma 3.6**.**
Let G=exp⊔(21W(t))⊔H be a non-degenerate Gaussian in AQ[t±1]w(X,∗C∪S).
•
If a non-degenerate Gaussian exp⊔(21W(t))⊔H′ is equal to G in AQ[t±1]w(X,∗C,*⃝S), then
∫S(exp⊔(21W(t))⊔H′)=∫SG.
•
If G′=exp⊔(21W′(t))⊔H′ is obtained from G=exp⊔(21W(t))⊔H by applying a winding relation, then ∫SG′=∫SG.
Proof.
The first point is essentially given by the proof of Bar-Natan and Lawrence [BNL04, Proposition 2.2] in the non-beaded case. Here, multiplication
by
For Aˉ=A,AR, or ARw, define a category Aˉ whose objects are associative words in the letters (+,−) and whose set of morphisms are
Aˉ(v,u)=⊕XAˉ(X), where X runs over all compact oriented 1–manifolds with boundary identified with the set of letters of u and v,
with the following sign convention: for u, a + when the orientation of X goes towards the boundary point and a − when it goes backward,
and the converse for v. Composition is given by vertical juxtaposition, where the label of the created edges in the case of beaded or winding diagrams
is defined with the same rule as in the definition of ⟨D,E⟩. The tensor product given by disjoint union defines a strict
monoidal structure on Aˉ.
We finally define the target category of our extended Kricker invariant.
Notation*.*
Given a positive integer g and a symbol ♮, set ⌊g⌉♮={1♮,…,g♮}. Set ⌊0⌉♮=∅.
Definition 3.7**.**
Fix non-negative integers f and g. An R–beaded Jacobi diagram on (∅,⌊g⌉+∪⌊f⌉−) is top–substantial
if it is ⌊g⌉+–substantial.
Given two such diagrams D and E, define their composition D∘E as the sum of all ways of gluing all i+–labelled vertices of D with all
i−–labelled vertices of E, fixing the orientations and labels of the created edges as in the definition of ⟨D,E⟩S.
We get a category tsA whose objects are non-negative integers, with set of morphisms from g to f the subspace of AQ(t)(∗⌊g⌉+∪⌊f⌉−) generated
by top-substantial diagrams.
The identity of g is \textrm{{exp}}_{\sqcup}(\sum_{i=1}^{g}\raisebox{-9.90276pt}{
\leavevmode\hbox to12.59pt{\vbox to29.37pt{\pgfpicture\makeatletter\hbox{\hskip 0.2pt\lower-6.0719pt\hbox to0.0pt{\pgfsys@beginscope\pgfsys@invoke{ }\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\pgfsys@setlinewidth{0.4pt}\pgfsys@invoke{ }\nullfont\hbox to0.0pt{\pgfsys@beginscope\pgfsys@invoke{ }{{}}
{}{{}}{}
{}{}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@setdash{3.0pt,3.0pt}{0.0pt}\pgfsys@invoke{ }{}\pgfsys@moveto{0.0pt}{0.0pt}\pgfsys@lineto{0.0pt}{17.07182pt}\pgfsys@stroke\pgfsys@invoke{ }
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}{{}}{}{{}}{}{{}}{}{{}}{}\pgfsys@moveto{0.0pt}{0.0pt}\pgfsys@moveto{0.0pt}{17.07182pt}\pgfsys@stroke\pgfsys@invoke{ }\hbox{\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{
{}{}}}{
{}{}}
{{}{{}}}{{}{}}{}{{}{}}
{
}{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{3.533pt}{-2.73889pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{{\scriptstyle{i^{-}}}}
}}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{
{}{}}}{
{}{}}
{{}{{}}}{{}{}}{}{{}{}}
{
}{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{3.533pt}{14.18016pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{{\scriptstyle{i^{+}}}}
}}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}{}{}\hss}\pgfsys@discardpath\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope\hss}}\lxSVG@closescope\endpgfpicture}}})
—the sum is trivial if g=0.
The tensor product defined by disjoint union of diagrams provides tsA a strict monoidal structure.
4 Winding matrices
In this section, we define winding matrices associated with tangles with disks and bottom-top tangles with paths and interpret them as equivariant linking matrices
in the case of bottom-top tangles. They will be useful in expressions of our invariant and splitting formulas.
4.1 First definition
We first define winding matrices in tangles with disks.
Let (γ,k) be a tangle with disks dℓ. Write γ as the disjoint union of components γi for i=1,…,n. Fix a diagram of (γ,k) and
a base point ⋆i for each closed component γi far from the crossings and the disks. Define the associated windingw(γi,γj)∈Z[t±1] of
γi and γj in the following way. For a crossing c between γi and γj, denote εij(c) the algebraic intersection number
of the union of the disks dℓ with the path that goes from ⋆i, or the origin of γi, to c along γi and then from c to ⋆j,
or the end-point of γj, along γj. If i=j, change component at the first occurence of c. Set
[TABLE]
where the sums are over all crossings between γi and γj. Note that w(γj,γi)(t)=w(γi,γj)(t−1).
Now let I and J be two subsets of {1,…,n} and denote by γI and γJ the corresponding subtangles of γ.
The winding matrixWγIγJ associated with the fixed diagram and base points is the matrix whose coefficients are the windings w(γi,γj)
for i∈I and j∈J —denote it WγI when I=J. In this latter case, note that WγI is hermitian.
Lemma 4.1**.**
The winding matrix is invariant by isotopies which do not allow the base points to pass through the disks of the tangle. In particular, when γ
contains no closed components, it is an isotopy invariant.
Proof.
First note that the winding matrix is preserved when a crossing passes through a disk dℓ. It is also preserved when a base point of a closed component passes
through a crossing since the algebraic intersection number of this component with the union of the disks dℓ is trivial. Hence it only remains to check invariance with
respect to framed Reidemeister moves performed far from the base points and the disks, which is direct.
∎
To completely understand the effect of an isotopy on the winding matrix WγI,γJ, we shall describe its modification when a base point passes through
a disk of the tangle. Fix a closed component γi. Fix a diagram of (γ,k) with the base point of γi located “just before” a disk dℓ of the tangle,
as shown in the first part of Figure 20.
Consider another diagram of (γ,k) which differs from the previous one only by the position of the base point ⋆i, which is as shown
on the second part of Figure 20. Let ε=±1 give the sign of the intersection of dℓ and γi which the base point
passes through. It is easily seen that the winding matrix of the latter diagram is obtained from the winding matrix of the previous one by multiplication
on the left by Ti(t−ε) if i∈I and on the right by Ti(tε) if i∈J, where Ti(t) is the diagonal matrix whose
diagonal coefficients are all 1 except a t at the ith position.
We now define winding matrices for bottom-top tangles in Q–cubes.
Let (B,K,γ) be a bottom-top tangle with paths in a Q–cube. Let (([−1,1]3,Ξ,η),L) be a surgery presentation of (B,K,γ).
Denote (η,k) the associated tangle with disks. Fix a diagram of ([−1,1]3,Ξ,η) and a base point ⋆i for each component Li of L.
Define the winding matrix of (B,K,γ), with coefficients in Q(t), as:
[TABLE]
Note that WL(1) is the linking matrix of L, thus WL(1)=0 since B is a homology cube. Hence WL is invertible over Q(t).
Lemma 4.2**.**
The winding matrix Wγ is an isotopy invariant of (B,K,γ).
Proof.
First, when the surgery presentation is fixed, the discussion of the previous subsection implies that the winding matrix does not depend on the choice
of diagram and base points. Then independance with respect to Kirby moves is easily checked. We detail the less direct, which is the case when a surgery
link component is added to a tangle component. Denote η′ the tangle obtained from η by adding the surgery component Lj to ηi.
We have:
[TABLE]
where P is the ∣L∣×∣η∣ matrix whose only non-trivial term is a 1 at the (j,i) position.
∎
4.2 Topological interpretation
Given a bottom-top tangle with paths in a Q–cube, or a tangle with disks defined from a bottom-top tangle with paths in [−1,1]3 with a surgery link,
close the components of the tangle, say γ, by line segments on the top and bottom surfaces to get a link γˉ. This provides well-defined linking numbers for the tangle components.
If there is no path or disk, the winding matrix is the linking matrix. It is clear when working in [−1,1]3. For bottom-top tangles in Q–cubes, apply the following
easy fact, which can be proved by adapting the proof of Proposition 4.5.
Fact 4.3*.*
Let L be an oriented framed link in [−1,1]3 whose linking matrix Lk(L) is non-degenerate. Let ξ and ζ be disjoint oriented knots
in [−1,1]3∖L and denote ξ′,ζ′ the copies of ξ,ζ in the Q–cube obtained from [−1,1]3 by surgery on L. Then:
[TABLE]
More generally, when there are paths or disks, the winding matrix evaluated at t=1 is the linking matrix. We shall give a similar interpretation
for the winding matrix at a generic t.
Let (B,K,γ) be a bottom-top tangle in a Q–cube. Let E be the exterior of K in B and let E~ be its covering associated with the kernel of the map
π1(E)→Z=⟨t⟩ which sends the positive meridians of the components of K to t. The automorphism group of the covering is isomorphic
to Z; let τ be the generator associated with the action of the positive meridians. Let ζ be a knot in E~ such that there are
a rational 2–chain Σ in E~ and P∈Q[t±1] which satisfy ∂Σ=P(τ)(ζ).
Let ξ be another knot in E~ such that the projections of ζ and ξ in E are disjoint. Define the equivariant linking number
of ζ and ξ as:
[TABLE]
The equivariant linking number is well-defined since H2(E~,Q)=0 (see for instance [Mou14, Lemma 2.1])
and satisfies lke(τ(ζ),ξ)=tlke(ζ,ξ).
First consider a tangle with disks (γ,k), with disks dℓ associated to the integer k, defined from a surgery presentation of a bottom-top tangle with paths in a Q–cube, so that the closure
γˉi of each component γi is well-defined. Fix a diagram of (γ,k) and base points ⋆i of
its components. For an interval component, choose the base point to be its origin. Set d=∪ℓ=1kdℓ. Let E be the exterior of
∂d in [−1,1]3 and let E~ be the infinite cyclic covering defined above. Let E~0⊂E~ be a copy of the exterior of d
in [−1,1]3. Define the lift γi~ of γˉi in E~ by lifting ⋆i in E~0. Given two subtangles γI and
γJ of γ, define the equivariant linking matrixLke(γ~I,γ~J) of their lifts with the equivariant linking numbers
of the γi~.
Lemma 4.4**.**
WγIγJ=Lke(γ~I,γ~J)**
Proof.
Set E~ℓ=τℓ(E~0), where τ is the generator of the automorphism group of the covering E~ which corresponds to the action of the
positive meridians of the ∂dℓ.
Fix i∈I and j∈J. Since γˉi is null-homotopic in [−1,1]3∖∂d, it bounds a disk D immersed in [−1,1]3∖∂d.
Let D~ be the lift of D obtained by lifting ⋆i in E~0. Set D~ℓ=D~∩E~ℓ and let Dℓ be the
image of D~ℓ in E. Set cℓ=∂Dℓ and c~ℓ=∂D~ℓ. Similarly, define the cℓ′ and c~ℓ′ from
γˉj. Assume the cℓ′ do not meet the Dℓ along the disks of the tangle.
Thanks to Lemma 4.1, we have w(γi,γj)=∑ℓ,ℓ′∈Zw(cℓ,cℓ′′)tℓ−ℓ′
for any choice of base points of the cℓ and cℓ′′. Since these latter curves do not cross the disks of the tangle, we have
w(cℓ,cℓ′′)=lk(cℓ,cℓ′′)=⟨Dℓ,cℓ′′⟩, thus
w(γi,γj)=∑ℓ,ℓ′∈Z⟨Dℓ,cℓ′′⟩tℓ−ℓ′. Lifting both Dℓ and cℓ′′ in E~ℓ
does not change their algebraic intersection number, hence
[TABLE]
where the third equality holds since τℓ′(c~ℓ−ℓ′′)=τℓ′(γj~)∩E~ℓ.
∎
Now consider a bottom-top tangle with paths in a Q–cube (B,K,γ). Since γ is null-homotopic in B∖K, we have a well-defined
equivariant linking matrixLke(γ~). Here, all components are intervals, so we have a canonical choice of base points.
Proposition 4.5**.**
Let (B,K,γ) be a bottom-top tangle with paths in a Q–cube. Then Wγ=Lke(γ~).
Proof.
Let (([−1,1]3,Ξ,η),L) be a surgery presentation of (B,K,γ). Fix a diagram of ([−1,1]3,Ξ,η∪L) and base points for the components of
L=⊔1≤i≤nLi. Let E~ be the infinite cyclic covering of the exterior of Ξ in [−1,1]3. Let d be the disjoint union of disks in [−1,1]3
bounded by Ξ^. Let L~=⊔1≤i≤nL~i, γ~ and η~ be the lifts of L, γ and η in E~
with all base points in the same copy in E~ of the exterior of d in [−1,1]3.
We have to prove that:
[TABLE]
The infinite cyclic covering E~′ of the exterior of K in B is obtained from E~ by surgery on ∪ℓ∈Zτℓ(L~).
Let η~x and η~y be components of η~, and let γ~x and γ~y be the corresponding components of γ~.
For any knot λ in E~ or E~′, denote m(λ) a positive meridian. For 1≤i≤n, let ci be the parallel of L~i
which bounds a disk after surgery. In the group H1(E~∖∪ℓ∈Z(τℓ(L~)∪τℓ(η~y));Z),
we have
[TABLE]
and
[TABLE]
where multiplication by t is given by the action of τ in homology.
In H1(E~′∖∪ℓ∈Zτℓ(γ~y);Z), this gives
γ~x=lke(η~x,η~y)m(γ~y)−Lke(η~x,L~)Lke(L~)−1Lke(L~,η~y)m(γ~y).
∎
It is easily checked that the null LP–surgery defined in the introduction provides a move on the set of bottom-top tangles with paths in Q–cubes, which corresponds to the move of null LP–surgery on Lagrangian cobordisms defined at the end of Subsection 2.1via the map D of Proposition 2.1. Moreover:
Corollary 4.6**.**
The winding matrix of a bottom-top tangle with paths in a Q–cube is invariant under null LP–surgeries.
Proof.
Let (B,K,γ) be a bottom-top tangle with paths in a Q–cube associated with a Lagrangian cobordism with paths (M,K). Let C be a null LP–surgery
on (M,K) made of a single QSK–pair. Let (M′,K′) be the Lagrangian cobordism with paths obtained by surgery and let (B′,K′,γ′) be the associated
bottom-top tangle with paths in a Q–cube. Let E~ be the infinite cyclic covering of the exterior of K in B. The nullity condition
implies that the preimage of the Q–handlebody C is the disjoint union of Q–handlebodies Cℓ homeomorphic to C. The infinite cyclic covering
E~′ of the exterior of K′ in B′ is obtained from E~ by null LP–surgeries on all the Cℓ. This concludes since LP–surgeries preserve
linking numbers (see for instance [Mou15, Lemma 2.1]).
∎
5 Construction of an LMO functor on LCob
5.1 The functor Z∙:Tq→AQ[t±1]w
The definition of the functor Z∙:Tq→AQ[t±1]w is based on the functor Z:Tq→A of [CHM08], which is a renormalization of the Le–Murakami functor
[LM95, LM96]. We recall in Figure 21 the definition of Z on the elementary q–tangles, where ν∈A(\leavevmodeto9.51pt\vboxto9.51pt\pgfpicture\makeatletter\lower-4.75253ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@moveto4.55254pt0.0pt\pgfsys@curveto4.55254pt2.51433pt2.51433pt4.55254pt0.0pt4.55254pt\pgfsys@curveto-2.51433pt4.55254pt-4.55254pt2.51433pt-4.55254pt0.0pt\pgfsys@curveto-4.55254pt-2.51433pt-2.51433pt-4.55254pt0.0pt-4.55254pt\pgfsys@curveto2.51433pt-4.55254pt4.55254pt-2.51433pt4.55254pt0.0pt\pgfsys@closepath\pgfsys@moveto0.0pt0.0pt\pgfsys@stroke\pgfsys@invoke\pgfsys@moveto4.55254pt0.85355pt\pgfsys@lineto4.55254pt0.67792pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.04.55254pt0.67792pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpicture)≅A(\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.0pt\pgfsys@lineto0.0pt10.92093pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.01.0-1.00.00.0pt10.92093pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpicture) is the value of the
Kontsevich integral on the zero framed unknot, Φ∈A(\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.45999pt\pgfsys@lineto0.0pt11.38092pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.0-1.01.00.00.0pt0.45999pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpicture\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.45999pt\pgfsys@lineto0.0pt11.38092pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.0-1.01.00.00.0pt0.45999pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpicture\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.45999pt\pgfsys@lineto0.0pt11.38092pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.0-1.01.00.00.0pt0.45999pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpicture) is a Drinfeld associator with rational coefficients and
Δu1,u2,u3+++:A(\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.45999pt\pgfsys@lineto0.0pt11.38092pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.0-1.01.00.00.0pt0.45999pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpicture\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.45999pt\pgfsys@lineto0.0pt11.38092pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.0-1.01.00.00.0pt0.45999pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpicture\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.45999pt\pgfsys@lineto0.0pt11.38092pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.0-1.01.00.00.0pt0.45999pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpicture)→A(\leavevmodeto0.4pt\vboxto11.78pt\pgfpicture\makeatletter\lower-0.2ptto0.0pt\pgfsys@beginscope\pgfsys@invoke\definecolorpgfstrokecolorrgb0,0,0\pgfsys@color@rgb@stroke000\pgfsys@invoke\pgfsys@color@rgb@fill000\pgfsys@invoke\pgfsys@setlinewidth0.4pt\pgfsys@invoke\nullfontto0.0pt\pgfsys@beginscope\pgfsys@invoke\pgfsys@moveto0.0pt0.45999pt\pgfsys@lineto0.0pt11.38092pt\pgfsys@stroke\pgfsys@invoke\pgfsys@beginscope\pgfsys@invoke\pgfsys@transformcm0.0-1.01.00.00.0pt0.45999pt\pgfsys@invoke\pgfsys@invoke\lxSVG@closescope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\pgfsys@discardpath\pgfsys@invoke\lxSVG@closescope\pgfsys@endscope\hss\lxSVG@closescope\endpgfpictureu1u2u3) is obtained by applying (∣ui∣−1) times Δ on the i-th factor.
Let (γ,k) be a q–tangle with disks. Assume γ is transverse to [−1,1]2×{hi(k)}
for all i∈{1,…,k}, and write γ as a composition of q–tangles γi by cutting along these levels, see Figure 22.
Write the bottom word of γi as wb(γi)=(vi)(wi), where wi corresponds to the part of the tangle which meets the disk di. Set:
[TABLE]
where Iv is the identity on the word v and Gv is obtained from Iv by adding a label t (resp. t−1) on skeleton components associated with
a − sign (resp. a + sign), see Figure 23.
At the level of objects, Z∙ forgets the parentheses.
Invariance with respect to isotopy and to the cutting of γ is due to invariance of the functor Z and the following observation of Kricker
[Kri00, Lemma 3.2.4].
Lemma 5.1**.**
For a winding Jacobi diagram D∈AQ[t±1]w(w,v), we have Gv∘D=D∘Gw.
Proof.
Apply the relations Hol and Holw at all vertices of the diagram.
∎
Furthermore, Z∙ is a clearly a functor and it preserves the tensor product on Tq⊗Tq since Z is tensor-preserving.
Lemma 5.2**.**
For any q–tangle with disks (γ,k), Z∙(γ,k) is group-like.
Proof.
The fact that Z(γ) is group-like for a q–tangle γ follows from [LM97, Theorem 5.1]. This concludes since the Gv are obviously group-like and the coproduct commutes with the composition.
∎
5.2 The functor Z:TqCub→AQ(t)w
The next step is to evaluate Z∙ on the surgery presentation of a q–tangle with paths in a Q–cube. Let (B,K,γ)∈TqCub(w,v). Let
(([−1,1]3,Ξ,η),L) be a surgery presentation of (B,K,γ).
The trivial link Ξ^ is the union of the boundaries of the disks di=[0,1]×[−1,1]×{hi(k)}, where k is the number of components of Ξ.
Hence we have a q–tangle with disks (η∪L,k) and Z∙(η∪L,k)∈AQ[t±1]w(η∪L). Set:
[TABLE]
where the connected sum means that a copy of ν is summed to each component of L. Note that Z∘((Ξ,η),L) is group-like since Z∙(η∪L,k)
and ν are group-like and χπ0(L) preserves the coproduct.
We want to apply formal Gaussian integration to Z∘((Ξ,η),L). We work with a lift Z∘((Ξ,η),L)∈AQ[t±1]w(η,∗π0(L)).
Fix a diagram of the q–tangle with disks (η∪L,k) transverse to the levels {hi(k)}, and fix base points ⋆i on each component Li
of L. Construct Z∘((Ξ,η),L) following the construction from the beginning of Section 5 for this diagram, with the skeleton components
corresponding to the components of L defined as intervals by cutting each component Li at the base point ⋆i.
Lemma 5.3**.**
The lift Z∘((Ξ,η),L) is group-like, and we have:
[TABLE]
where WL is the winding matrix associated with our choice of diagram and base points and H is π0(L)–substantial.
Proof.
Check as in Lemma 5.2 that Z∘((Ξ,η),L) is group-like. We have to compute the part of Z∘((Ξ,η),L)
made of π0(L)–struts. We work with χπ0(η)−1(Z∘((Ξ,η),L))∈AQ[t±1](∗π0(η)∪π0(L)), which is also group-like, in order to have a Hopf algebra
structure on our diagram space. In particular, the group-like property implies that χπ0(η)−1(Z∘((Ξ,η),L)) is the exponential
of a series of connected diagrams. Since ν and the associator Φ have no terms with exactly two vertices, the only contributions to the π0(L)–struts part
come from the crossings between components of L. For i=j, the definition of Z and the Holw relation show that the contribution
of a crossing c between Li and Lj is \chi_{\pi_{0}(\eta)}^{-1}\raisebox{-3.44444pt}{{{\Huge(}}}\frac{1}{2}\textrm{sg}(c)\raisebox{-28.45274pt}{
\leavevmode\hbox to47.2pt{\vbox to42.4pt{\pgfpicture\makeatletter\hbox{\hskip 6.40398pt\lower-13.74614pt\hbox to0.0pt{\pgfsys@beginscope\pgfsys@invoke{ }\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\pgfsys@setlinewidth{0.4pt}\pgfsys@invoke{ }\nullfont\hbox to0.0pt{\pgfsys@beginscope\pgfsys@invoke{ }{{}}{{}}
{}{{}}{}
{}{}{}{}{}{}{{}}\pgfsys@moveto{0.0pt}{0.0pt}\pgfsys@lineto{0.0pt}{27.99277pt}\pgfsys@stroke\pgfsys@invoke{ }{{}{{}}{}{}{{}}{{{}}{{{}}{\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{0.0}{1.0}{-1.0}{0.0}{0.0pt}{27.99277pt}\pgfsys@invoke{ }\pgfsys@invoke{ \lxSVG@closescope }\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}{{}}}}
{}{{}}{}{{}}{}\pgfsys@moveto{0.0pt}{0.0pt}\pgfsys@stroke\pgfsys@invoke{ }\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{
{}{}}}{
{}{}}
{{}{{}}}{{}{}}{}{{}{}}
{
}{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{-3.07097pt}{-8.31633pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{{\scriptstyle{L_{i}}}}
}}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
{}{{}}{}
{}{}{}{}{}{}{{}}\pgfsys@moveto{34.14365pt}{0.0pt}\pgfsys@lineto{34.14365pt}{27.99277pt}\pgfsys@stroke\pgfsys@invoke{ }{{}{{}}{}{}{{}}{{{}}{{{}}{\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{0.0}{1.0}{-1.0}{0.0}{34.14365pt}{27.99277pt}\pgfsys@invoke{ }\pgfsys@invoke{ \lxSVG@closescope }\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}{{}}}}
{}{{}}{}{{}}{}\pgfsys@moveto{34.14365pt}{0.0pt}\pgfsys@stroke\pgfsys@invoke{ }\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{
{}{}}}{
{}{}}
{{}{{}}}{{}{}}{}{{}{}}
{
}{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{30.82361pt}{-8.31633pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{{\scriptstyle{L_{j}}}}
}}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
{}{{}}{}
{}{}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@setdash{3.0pt,3.0pt}{0.0pt}\pgfsys@invoke{ }{}{}{}{}{{}}\pgfsys@moveto{0.0pt}{14.22638pt}\pgfsys@lineto{16.61183pt}{14.22638pt}\pgfsys@stroke\pgfsys@invoke{ }{{}{{}}{}{}{{}}{{{}}{{{}}{\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{16.61183pt}{14.22638pt}\pgfsys@invoke{ }\pgfsys@invoke{ \lxSVG@closescope }\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}{{}}}}
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}{{}}{}
{}{}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@setdash{3.0pt,3.0pt}{0.0pt}\pgfsys@invoke{ }{}\pgfsys@moveto{17.07182pt}{14.22638pt}\pgfsys@lineto{34.14365pt}{14.22638pt}\pgfsys@stroke\pgfsys@invoke{ }
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}{{}}{}{{}}{}\pgfsys@moveto{17.07182pt}{14.22638pt}\pgfsys@stroke\pgfsys@invoke{ }\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{
{}{}}}{
{}{}}
{{}{{}}}{{}{}}{}{{}{}}
{
}{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{11.15254pt}{17.75938pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{{\scriptstyle{t^{\varepsilon_{ij}(c)}}}}
}}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}
\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}{}{}\hss}\pgfsys@discardpath\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope\hss}}\lxSVG@closescope\endpgfpicture}}}).
Hence the contribution of all crossings between Li and Lj is
\scriptstyle{L_{i}}$$\scriptstyle{L_{j}}$$\scriptstyle{(W_{L})_{ij}}$$=$$\scriptstyle{L_{j}}$$\scriptstyle{L_{i}}$$\scriptstyle{(W_{L})_{ji}}
.
For i=j, the contribution of a self-crossing of Li is:
[TABLE]
Summed over all self-crossings of Li, we get as strut part:
[TABLE]
Hence χπ0(η)−1(Z∘((Ξ,η),L))=exp⊔(21WL)⊔H′ where H′∈AQ[t±1](∗π0(η)∪π0(L))
is π0(L)–substantial. Set H=χπ0(η)(H′).
∎
The matrix WL(1) is the linking matrix of the link L, hence it is the presentation matrix of the first homology group of a Q–cube.
Thus det(WL(1))=0 and Z∘((Ξ,η),L) is a non-degenerate Gaussian. Lemma 3.6 implies:
Lemma 5.4**.**
The formal Gaussian integral ∫π0(L)Z∘((Ξ,η),L) does not depend on the lift
Z∘((Ξ,η),L)∈AQ[t±1]w(η,∗π0(L)) of Z∘((Ξ,η),L)∈AQ[t±1]w(η,*◯⃝π0(L)).
This allows to set:
[TABLE]
Proposition 5.5**.**
Let (B,K,γ) be a q–tangle with paths in a Q–cube. Fix a surgery presentation (([−1,1]3,Ξ,η),L) of (B,K,γ). Then:
We have to check that Z(B,K,γ) does not depend on the surgery presentation. Independance with respect to the orientation of the components of L
follows from the argument of [BNGRT02b, Proposition 3.1]. The normalization term U+−σ+(L)⊔U−−σ−(L) ensures independance with respect
to the KI move as usual. Independance with respect to the KII move mainly follows from [GK04, Section 5.4]. More precisely, the argument of [GK04, Theorem 4]
adapts [LMMO99, Proposition 1] to relate the values of Z∘((Ξ,η),L) for surgery links that differ from each other by
a KII move. Then [GK04, Lemma 5.6] shows that this implies the invariance of the formal Gaussian integral. As noted in [BNGRT02b, Section 5.1],
the argument remains valid when a surgery component is added to a tangle component since [GK04, Lemma 5.6] uses integration along the surgery component.
∎
Restricting the functor Z:TqCub→AQ(t)w to q–tangles in Q–cubes with no path, one recovers the functor Z:TqCub→A of [CHM08, Definition 3.16].
When γ is a bottom-top tangle and K=∅, χπ0(γ)−1(Z(B,∅,γ)) is group-like and
χπ0(γ)−1(Z(B,∅,γ))=exp⊔(Lk(γ))⊔H for some substantial and group-like H [CHM08, Lemma 3.17].
We generalize this in the next result.
Lemma 5.6**.**
For any bottom-top q–tangle with paths (B,K,γ) where B is a Q–cube, χπ0(γ)−1(Z(B,K,γ)) is group-like
and:
[TABLE]
for some substantial and group-like H.
Proof.
The fact that Z(B,K,γ) is group-like follows from the same property for U+, U− and Z∘((Ξ,η),L), and Theorem 3.4.
It implies that χπ0(γ)−1(Z(B,K,γ)) is group-like since χπ0(γ)−1 preserves the coproduct.
The same computation as in the proof of Lemma 5.3 gives:
[TABLE]
where H′ is substantial —note that χπ0(γ) and χπ0(η) are essentially the same before and after surgery on L. Integrate along π0(L):
[TABLE]
∎
5.3 The functor Z~:LCobq→tsA
In this section, we define a functor on Lagrangian q–cobordisms with paths by applying the invariant Z on bottom-top q–tangles with paths in Q–cubes.
The invariant Z is functorial on q–tangles but not on bottom-top q–tangles, due to the different composition laws. To deal with this, we introduce
some specific elements ⊤g∈tsA(∗⌊g⌉+∪⌊g⌉−) following [CHM08, Sec. 4].
Set:
[TABLE]
[TABLE]
[TABLE]
where the bottom-top tangle T1 is drawn in Figure 9.
As proven in [CHM08, Lemma 4.9]:
Lemma 5.7**.**
⊤g* is a group-like element of AQ(t)(∗⌊g⌉+∪⌊g⌉−) and ⊤g=Idg⊔H for some substantial and group-like H. In particular, ⊤g is top-substantial
and ⌊g⌉−–substantial.*
Let (M,K) be a Lagrangian q–cobordism with paths and denote (B,K,γ) the associated bottom-top q–tangle with paths, of type (g,f).
We have Z(B,K,γ)∈AQ(t)w(γ)≅AQ(t)(γ) and we consider χ−1(Z(B,K,γ))∈AQ(t)(∗⌊g⌉+∪⌊f⌉−).
It may not be top-substantial, but since ⊤g is ⌊g⌉−–substantial, we can set:
[TABLE]
At the level of objects, Z~ sends a word on its number of letters.
Direct adaptation of the proof of [CHM08, Lemma 4.10] implies that Z~ preserves the composition and the next result follows, see [CHM08, Theorem 4.13].
Proposition 5.8**.**
Z~:LCobq→tsA* is a functor which preserves the tensor product on LCobq⊗LCobq.*
Restricting the functor Z~:LCobq→tsA to Lagrangian q–cobordisms with no path, one recovers the functor Z~ defined on LCobq in [CHM08, Theorem 4.13].
Let (M,K) be a Lagrangian q–cobordism with paths and let (B,K,γ) be the associated bottom-top q–tangle with paths.
Then Z~(M,K) is group-like and Z~(M,K)=exp⊔(Wγ)⊔H for some substantial and group-like H.
5.4 Application to QSK–pairs
Let (S,κ) be a QSK–pair. Let M be the Q–cube obtained from S by removing the interior of a ball B3 disjoint from κ.
Isotoping κ in M and fixing a boundary parametrization m of M, one can view κ as the knot K^ associated with a Lagrangian cobordism
with one path (M,K). Since the top and bottom words are empty, we get a Lagrangian q–cobordism with one path.
Proposition 5.10**.**
Let (S,κ) be a QSK–pair. Define as above an associated Lagrangian q–cobordism with one path (M,K). Then Z~(S,κ)=Z~(M,K) defines an invariant
of QSK–pairs, which coincides with the Kricker invariant Zrat for knots in Z–spheres.
Proof.
When associating a cobordism with one path with a QSK–pair, we make a choice in the way we isotope the knot to the closure of a path. Once we work
with a surgery presentation of our cobordism, this choice corresponds to the sweeping move represented in Figure 24.
But the right hand side diagram of this figure shows this move is trivial —as noted in [GK04, Lemma 3.26].
Coincidence with Zrat is direct by construction.
∎
Remark*.*
The above proof does not work for a cobordism with more than one path, so we do not get an invariant of boundary links in Q–spheres. One may obtain
such an invariant by quotienting out the target space by suitable relations, see [GK04] for a construction of this kind.
Proposition 5.11**.**
Let (S1,κ1) and (S2,κ2) be QSK–pairs. The invariant Z~ is given on their connected sum by:
[TABLE]
Proof.
As previously, associate Lagrangian q–cobordisms with one path (M1,K1) and (M2,K2) with (S1,κ1) and (S2,κ2) respectively.
Construct a Lagrangian q–cobordism with one path (M,K) associated with (S,κ)=(S1,κ1)♯(S2,κ2) by stacking (M1,K1) and (M2,K2)
together in the y direction. Now (M1,K1) and (M2,K2) are obtained from the cube [−1,1]3
with one disk by surgery on links L1 and L2 respectively. We obtain a surgery diagram for (M,K) by drawing L1 “in front” of L2, or equivalently
“around” L2, see Figure 25. The result follows from this latter diagram since there is no crossing between L1 and L2.
∎
6 Splitting formulas
We first mention useful lemmas, namely [Mas15, Lemma 4.3] and [Mas15, Lemma 4.4]. Recall the tensor μ(C) was defined in the introduction.
Lemma 6.1** (Massuyeau).**
For a Q–handlebody C of genus g, there exists a boundary parametrization c:∂C0g→C such that (C,c)∈LCob(g,0).
Lemma 6.2** (Massuyeau).**
Let C=(CC′) be an LP–pair of genus g. Take boundary parametrizations c:∂C0g→C and c′:∂C0g→C′
compatible with the fixed identification ∂C≅∂C′ such that (C,c)∈LCob(g,0) and (C′,c′)∈LCob(g,0). Then:
[TABLE]
where Z~1 is the i–degree 1 part of Z~ and μ(C) is considered as an element of AQ(t)(∗⌊g⌉+)
via the inclusion Λ3H1(C;Q)↪AQ(t)(∗⌊g⌉+) defined by:
[TABLE]
Let (M,K)∈LCobq(w,v). Let C=(C1,…,Cn) be a null LP–surgery on (M,K). Let ei be the genus of Ci.
For 1≤i≤n, take boundary parametrizations ci:∂C0ei→Ci and ci′:∂C0ei→Ci′ compatible with the fixed identification
∂Ci≅∂Ci′ such that (Ci,ci)∈LCob(ei,0) and (Ci′,ci′)∈LCob(ei,0). Set e=∑i=1nei.
Take a collar neighborhood m−(Ff)×[−1,ε−1]
of the bottom surface m−(Ff). Take pairwise disjoint solid tubes Ti, i=1,…,n, such that Ti connects (ci)−(F0) to a disk in
m−(Ff)×{ε−1} in the complement of the Cj’s, the collar neighborhood and K. This provides a decomposition of the cobordism (M,K) as:
[TABLE]
where f is the number of letters of v (see Figure 26). It is proved in [Mas15, Section 4.4] that N is a Lagrangian cobordism. The nullity condition on the surgery
ensures that J^ is a boundary link. Thus (N,J) is a Lagrangian cobordism with paths.
With the surgery C is associated the tensor μ(C)∈AQ(H1(C;Q)). Let W be a square matrix of size e with coefficients in Q(t).
Interpret W as a hermitian form on H1(C;Q) written in the basis (([(ci)+(βj)])1≤j≤ei)1≤i≤n.
Given an H1(C;Q)–colored Jacobi diagram, one can glue some legs of the diagram with W, see Figure 3. Changing the labels of the univalent vertices via the bijection
[TABLE]
onto {1,…,e}, this provides a diagram in AQ(t)(∗⌊e⌉+).
The following result is a direct adaptation of [Mas15, Section 4.4], with the winding matrices playing the role of the linking matrices.
Proposition 6.3**.**
Let (M,K) be a Lagrangian q–cobordism with paths and let (B,K,γ) be the associated bottom-top q–tangle with paths. Let C=(C1,…,Cn)
be a null LP–surgery on (M,K). Define as above a decomposition of the cobordism (M,K). Choose top and bottom words for (N,J) and the (Ci,∅)
in order to get a decomposition of the Lagrangian q–cobordism (M,K) as (M,K)=((C1,∅)⊗⋯⊗(Cn,∅)⊗Idv)∘(N,J).
Let (D,J,ς) be the bottom-top q–tangle with paths associated with (N,J). Let ςc be the subtangle of ς− corresponding to the Ci’s.
Let e be the number of components of ςc.
Let ρ~c:AQ(t)(∗⌊e⌉+)→AQ(t)(∗⌊g⌉+∪⌊f⌉−) be the linear form which changes the labels
of the univalent vertices as follows:
[TABLE]
Then:
[TABLE]
where CI=((Ci)i∈I) and ≡n means “equal up to i–degree at least n+1 terms”.
For a cobordism with one path, the next result gives a more intrinsic version of these formulas, which does not refer to a decomposition of the cobordism.
A similar result is given by in [Mas15, Lemma 4.1] for a cobordism with no path.
Given a null LP–surgery C=(C1,…,Cn) on a Lagrangian cobordism with paths (M,K), define a hermitian form
ℓ(M,K)(C):H1(C;Q)×H1(C;Q)→Q(t) in the same way as ℓ(S,κ)(C) was defined in the introduction.
Also define a map ρc:AQ(H1(C;Q))→AQ(t)(∗⌊g⌉+∪⌊f⌉−) which changes the labels of the univalent vertices by first sending them in H1(M;Q)viaH1(C;Q)≅⊕i=1nH1(Ci;Q)→H1(M;Q), and then writing them in terms of the [m+(βi)] and [m−(αi)].
A direct adaptation of (the end of) [Mas15, Section 4.4] gives:
Proposition 6.4**.**
Let (M,K)∈LCobq(w,v) be a Lagrangian q–cobordism with one path. Let C=(C1,…,Cn)
be a null LP–surgery on (M,K). Let (B,K,γ) be the bottom-top tangle with one path associated with (M,K). Then:
Use Propositions 5.10 and 6.4. The strut part disappears since we deal with a cobordism from F0 to F0. The map ρc
kills all terms with at least one univalent vertex since a Lagrangian cobordism from F0 to F0 has trivial first homology group over Q.
∎
Bibliography25
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[BN 95] D. Bar-Natan – “On the Vassiliev knot invariants”, Topology 34 (1995), no. 2, p. 423–472.
2[BNGRT 02a] D. Bar-Natan, S. Garoufalidis, L. Rozansky & D. P. Thurston – “The Århus integral of rational homology 3-spheres I: A highly non trivial flat connection on S 3 superscript 𝑆 3 S^{3} ”, Selecta Mathematica 8 (2002), no. 3, p. 315–339.
3[BNGRT 02b] — , “The Århus integral of rational homology 3-spheres II: Invariance and universality”, Selecta Mathematica 8 (2002), no. 3, p. 341–371.
4[BNL 04] D. Bar-Natan & R. Lawrence – “A rational surgery formula for the LMO invariant”, Israel Journal of Mathematics 140 (2004), p. 29–60.
5[CHM 08] D. Cheptea, K. Habiro & G. Massuyeau – “A functorial LMO invariant for Lagrangian cobordisms”, Geometry & Topology 12 (2008), no. 2, p. 1091–1170.
6[GGP 01] S. Garoufalidis, M. Goussarov & M. Polyak – “Calculus of clovers and finite type invariants of 3–manifolds”, Geometry & Topology 5 (2001), p. 75–108.
7[GK 04] S. Garoufalidis & A. Kricker – “A rational noncommutative invariant of boundary links”, Geometry & Topology 8 (2004), p. 115–204.
8[GR 04] S. Garoufalidis & L. Rozansky – “The loop expansion of the Kontsevich integral, the null-move and S 𝑆 S -equivalence”, Topology 43 (2004), no. 5, p. 1183–1210.