Curso de Economia Politica I, Prof. Doutor Rui Teixeira Santos (ULHT, 2011/12)
Rui Loja Fernandes (based on joint work with Ioan Marcut˘2020/04/09 · Local models around...
Transcript of Rui Loja Fernandes (based on joint work with Ioan Marcut˘2020/04/09 · Local models around...
Local models around Poisson submanifolds
Rui Loja Fernandes...
(based on joint work with Ioan Marcut,)
Department of MathematicsUniversity of Illinois at Urbana-Champaign, USA
April 9, 2020
Overview - Main ProblemsNotation:
• (M,π): Poisson manifold
• S ⊂M: closed Poisson submanifold
Local Model Problem:
I Is there a 1st order local model around S?
Linearization/Normal Form Problem:
I When is (M,π) locally isomorphic to the 1st order localmodel around S?
Neighborhood Equivalence Problem:
I If π ′,π are Poisson structures with same 1st order dataalong S, are they locally isomorphic?
Overview - Main ProblemsNotation:
• (M,π): Poisson manifold
• S ⊂M: closed Poisson submanifold
Local Model Problem:
I Is there a 1st order local model around S?
Linearization/Normal Form Problem:
I When is (M,π) locally isomorphic to the 1st order localmodel around S?
Neighborhood Equivalence Problem:
I If π ′,π are Poisson structures with same 1st order dataalong S, are they locally isomorphic?
Overview - Main ProblemsNotation:
• (M,π): Poisson manifold
• S ⊂M: closed Poisson submanifold
Local Model Problem:
I Is there a 1st order local model around S?
Linearization/Normal Form Problem:
I When is (M,π) locally isomorphic to the 1st order localmodel around S?
Neighborhood Equivalence Problem:
I If π ′,π are Poisson structures with same 1st order dataalong S, are they locally isomorphic?
Overview - Main ProblemsNotation:
• (M,π): Poisson manifold
• S ⊂M: closed Poisson submanifold
Local Model Problem:
I Is there a 1st order local model around S?
Linearization/Normal Form Problem:
I When is (M,π) locally isomorphic to the 1st order localmodel around S?
Neighborhood Equivalence Problem:
I If π ′,π are Poisson structures with same 1st order dataalong S, are they locally isomorphic?
Overview - Main Results
Theorem (Existence of local models)The class of first order jets satisfying the partial splittingcondition admits a 1st order local model.
Theorem (Linearization)If (M,π) has restriction T ∗SM integrable by a compact, Liegroupoid, whose s-fibers have trivial 2nd de Rhamcohomology, then it is linearizable at S.
Corollary (Marcut 2014)Let T ∗SM be integrable by a compact, Lie groupoid, whoses-fibers have trivial 2nd de Rham cohomology. If J1
Sπ ′ = J1Sπ,
then π ′ is locally isomorphic to π around S.
Overview - Main Results
Theorem (Existence of local models)The class of first order jets satisfying the partial splittingcondition admits a 1st order local model.
Theorem (Linearization)If (M,π) has restriction T ∗SM integrable by a compact, Liegroupoid, whose s-fibers have trivial 2nd de Rhamcohomology, then it is linearizable at S.
Corollary (Marcut 2014)Let T ∗SM be integrable by a compact, Lie groupoid, whoses-fibers have trivial 2nd de Rham cohomology. If J1
Sπ ′ = J1Sπ,
then π ′ is locally isomorphic to π around S.
Overview - Main Results
Theorem (Existence of local models)The class of first order jets satisfying the partial splittingcondition admits a 1st order local model.
Theorem (Linearization)If (M,π) has restriction T ∗SM integrable by a compact, Liegroupoid, whose s-fibers have trivial 2nd de Rhamcohomology, then it is linearizable at S.
Corollary (Marcut 2014)Let T ∗SM be integrable by a compact, Lie groupoid, whoses-fibers have trivial 2nd de Rham cohomology. If J1
Sπ ′ = J1Sπ,
then π ′ is locally isomorphic to π around S.
Overview - Problems - Global Version
Notation:
• (G,ω)⇒M: symplectic groupoid: ω closed, multiplicative,ker ω = 0.
• (GS ,ωS)⇒ S: over-symplectic groupoid: ωS closed,multiplicative, ker ωS ⊂ ker ds∩ker dt.
Overview - Problems - Global Version
Notation:
• (G,ω)⇒M: symplectic groupoid: ω closed, multiplicative,ker ω = 0.
• (GS ,ωS)⇒ S: over-symplectic groupoid: ωS closed,multiplicative, ker ωS ⊂ ker ds∩ker dt.
Facts:
• Restriction of (G,ω) to saturated submanifold S ⊂M is anover-symplectic groupoid (G|S ,ω|S), and
• Inclusion i : (G|S ,ω|S) → (G,ω) is a coisotropic embedding.
Overview - Problems - Global Version
Notation:
• (G,ω)⇒M: symplectic groupoid: ω closed, multiplicative,ker ω = 0.
• (GS ,ωS)⇒ S: over-symplectic groupoid: ωS closed,multiplicative, ker ωS ⊂ ker ds∩ker dt.
Facts:
• Restriction of (G,ω) to saturated submanifold S ⊂M is anover-symplectic groupoid (G|S ,ω|S), and
• Inclusion i : (G|S ,ω|S) → (G,ω) is a coisotropic embedding.
Overview - Problems - Global Version
Notation:
• (G,ω)⇒M: symplectic groupoid: ω closed, multiplicative,ker ω = 0.
• (GS ,ωS)⇒ S: over-symplectic groupoid: ωS closed,multiplicative, ker ωS ⊂ ker ds∩ker dt.
Groupoid Coisotropic Embedding Problem:
I Given an over-symplectic groupoid (GS ,ωS), is there agroupoid coisotropic embedding i : (GS ,ωS) → (G,ω) intosome symplectic groupoid?
I Given groupoid coisotropic embeddings (k = 1,2)ik : (GS ,ωS) → (Gk ,ωk ), is there a local symplecticgroupoid isomorphism Φ : (G1,ω1)→ (G2,ω2), defined ina neighborhood of i1(GS), and such that Φ i1 = i2?
Overview - Main Results - Global Version
Theorem (Existence of coisotropic embeddings)If (GS ,ωS) is partially split then (ker ωS)∗⇒ (ker ωS |S)∗ is asymplectic groupoid and the zero section i : GS → (ker ωS)∗ isa coisotropic embedding.
Proper over-symplectic groupoids are partially split.
Theorem (Normal form for coisotropic embeddings)A groupoid coisotropic embedding i : (GS ,ωS) → (G,ω) into as-proper symplectic groupoid is locally isomorphic to thecoisotropic embedding i : GS → (ker ωS)∗.
Together these imply:
Groupoid Coisotropic Neighboorhood Theorem
Overview - Main Results - Global Version
Theorem (Existence of coisotropic embeddings)If (GS ,ωS) is partially split then (ker ωS)∗⇒ (ker ωS |S)∗ is asymplectic groupoid and the zero section i : GS → (ker ωS)∗ isa coisotropic embedding.
Proper over-symplectic groupoids are partially split.
Theorem (Normal form for coisotropic embeddings)A groupoid coisotropic embedding i : (GS ,ωS) → (G,ω) into as-proper symplectic groupoid is locally isomorphic to thecoisotropic embedding i : GS → (ker ωS)∗.
Together these imply:
Groupoid Coisotropic Neighboorhood Theorem
Overview - Main Results - Global Version
Theorem (Existence of coisotropic embeddings)If (GS ,ωS) is partially split then (ker ωS)∗⇒ (ker ωS |S)∗ is asymplectic groupoid and the zero section i : GS → (ker ωS)∗ isa coisotropic embedding.
Proper over-symplectic groupoids are partially split.
Theorem (Normal form for coisotropic embeddings)A groupoid coisotropic embedding i : (GS ,ωS) → (G,ω) into as-proper symplectic groupoid is locally isomorphic to thecoisotropic embedding i : GS → (ker ωS)∗.
Together these imply:
Groupoid Coisotropic Neighboorhood Theorem
Overview - Main Results - Global Version
Theorem (Existence of coisotropic embeddings)If (GS ,ωS) is partially split then (ker ωS)∗⇒ (ker ωS |S)∗ is asymplectic groupoid and the zero section i : GS → (ker ωS)∗ isa coisotropic embedding.
Proper over-symplectic groupoids are partially split.
Theorem (Normal form for coisotropic embeddings)A groupoid coisotropic embedding i : (GS ,ωS) → (G,ω) into as-proper symplectic groupoid is locally isomorphic to thecoisotropic embedding i : GS → (ker ωS)∗.
Together these imply:
Groupoid Coisotropic Neighboorhood Theorem
Overview - Brief history of smooth linearization
For a fixed point S = x0:
• Weinstein 83: Linearization problem, local form, conjectures onlinearization
• Conn 85: Linearization for g compact semi-simple (Nash-Moser)
• Monnier & Zung 04: Linearization for g⊕R, when g is compactsemi-simple (Nash-Moser)
• Crainic & RLF 11: Geometric proof of Conn’s Theorem
For a symplectic leaf S:
• Vorbjev 01, 05: First order local model and some partial results onneighborhood equivalence;
• Crainic & Marcut 12: Linearization theorem for symplectic leaves(geometric proof extending Crainic & RLF 11)
For a general Poisson submanifold S:
• Marcut 14: Rigidity (Nash-Moser)
Our Linearization Theorem includes all these results, but. . .
Overview - Brief history of smooth linearization
For a fixed point S = x0:
• Weinstein 83: Linearization problem, local form, conjectures onlinearization
• Conn 85: Linearization for g compact semi-simple (Nash-Moser)
• Monnier & Zung 04: Linearization for g⊕R, when g is compactsemi-simple (Nash-Moser)
• Crainic & RLF 11: Geometric proof of Conn’s Theorem
For a symplectic leaf S:
• Vorbjev 01, 05: First order local model and some partial results onneighborhood equivalence;
• Crainic & Marcut 12: Linearization theorem for symplectic leaves(geometric proof extending Crainic & RLF 11)
For a general Poisson submanifold S:
• Marcut 14: Rigidity (Nash-Moser)
Our Linearization Theorem includes all these results, but. . .
Overview - Brief history of smooth linearization
For a fixed point S = x0:
• Weinstein 83: Linearization problem, local form, conjectures onlinearization
• Conn 85: Linearization for g compact semi-simple (Nash-Moser)
• Monnier & Zung 04: Linearization for g⊕R, when g is compactsemi-simple (Nash-Moser)
• Crainic & RLF 11: Geometric proof of Conn’s Theorem
For a symplectic leaf S:
• Vorbjev 01, 05: First order local model and some partial results onneighborhood equivalence;
• Crainic & Marcut 12: Linearization theorem for symplectic leaves(geometric proof extending Crainic & RLF 11)
For a general Poisson submanifold S:
• Marcut 14: Rigidity (Nash-Moser)
Our Linearization Theorem includes all these results, but. . .
Overview - Brief history of smooth linearization
For a fixed point S = x0:
• Weinstein 83: Linearization problem, local form, conjectures onlinearization
• Conn 85: Linearization for g compact semi-simple (Nash-Moser)
• Monnier & Zung 04: Linearization for g⊕R, when g is compactsemi-simple (Nash-Moser)
• Crainic & RLF 11: Geometric proof of Conn’s Theorem
For a symplectic leaf S:
• Vorbjev 01, 05: First order local model and some partial results onneighborhood equivalence;
• Crainic & Marcut 12: Linearization theorem for symplectic leaves(geometric proof extending Crainic & RLF 11)
For a general Poisson submanifold S:
• Marcut 14: Rigidity (Nash-Moser)
Our Linearization Theorem includes all these results, but. . .
Plan of the talk
I. First order data: jets of Poisson structures
II. Global theory: groupoid coisotropic embeddings
III. Local theory: existence of local models and linearization
I. First order data
Notation:
• (T ∗M, [·, ·]π ,π]): cotangent Lie algebroid of (M,π).
Observation: This is really about jets of bivector fields!
I. First order data
Notation:
• (T ∗M, [·, ·]π ,π]): cotangent Lie algebroid of (M,π).
Proposition (Coste-Dazord-Weinstein, 87)
Poisson structures
π ∈ X2(M)
←→
Lie algebroids (T ∗M, [·, ·],ρ) withρ : T ∗M → TM skew-symmetric
[Ω1cl(M),Ω1
cl(M)]⊂ Ω1cl(M)
Observation: This is really about jets of bivector fields!
I. First order data
Notation:
• (T ∗M, [·, ·]π ,π]): cotangent Lie algebroid of (M,π).
Proposition (Modern formulation)Poisson structures
π ∈ X2(M)
←→
Lie algebroids (T ∗M, [·, ·],ρ) with
id : T ∗M → T ∗M a closed IM 2-form
Observation: This is really about jets of bivector fields!
I. First order data
Notation:
• (T ∗M, [·, ·]π ,π]): cotangent Lie algebroid of (M,π).
Proposition (Modern formulation)Poisson structures
π ∈ X2(M)
←→
Lie algebroids (T ∗M, [·, ·],ρ) with
id : T ∗M → T ∗M a closed IM 2-form
A closed IM 2-form on a Lie algebroid (A, [·, ·],ρ) is a bundle mapµ : A→ T ∗M such that:
〈µ(α),ρ(β )〉=−〈µ(β ),ρ(α)〉µ([α,β ]) = Lρ(α)µ(β )− iρ(β )dµ(α)
Observation: This is really about jets of bivector fields!
I. First order data
Notation:
• (T ∗M, [·, ·]π ,π]): cotangent Lie algebroid of (M,π).
Proposition (Modern formulation)Poisson structures
π ∈ X2(M)
←→
Lie algebroids (T ∗M, [·, ·],ρ) with
id : T ∗M → T ∗M a closed IM 2-form
Observation: This is really about jets of bivector fields!
Notation:
• J1E →M: first jet bundle of a bundle E →M.
• The Schouten bracket induces a graded Lie bracket:
[ , ] : Γ(J1(∧k+1TM))×Γ(J1(∧l+1TM))→ Γ(J1(∧k+l+1TM)),
characterized by
[J1ϑ1,J1
ϑ2] := J1[ϑ1,ϑ2].
Proof.i : J1(∧d TM) → Derd (T ∗M),
i(J1ϑ)(α1, . . . ,αd ) := d(ϑ(α1, . . . ,αd ))−
d
∑i=1
(−1)i iϑ ](α1,...,αi ,...,αd )dαi ,
where for ϑ ∈ Xd (M) we denote by ϑ ] : ∧d−1T ∗M → TM the map given by
ϑ](α1, . . . ,αk−1)(α) := ϑ(α1, . . . ,αk−1,α).
This map:
(i) takes the Schouten bracket to the graded bracket of multi-derivations.
(ii) image is multi-derivations D with symbol σD ∈ ∧d T ?M ⊂ ∧d−1TM⊗TM
Notation:
• J1E →M: first jet bundle of a bundle E →M.
• The Schouten bracket induces a graded Lie bracket:
[ , ] : Γ(J1(∧k+1TM))×Γ(J1(∧l+1TM))→ Γ(J1(∧k+l+1TM)),
characterized by
[J1ϑ1,J1
ϑ2] := J1[ϑ1,ϑ2].
Question. What is the geometric meaning of τ ∈ Γ(J1(∧2TM)) with[τ,τ] = 0?
Proof.i : J1(∧d TM) → Derd (T ∗M),
i(J1ϑ)(α1, . . . ,αd ) := d(ϑ(α1, . . . ,αd ))−
d
∑i=1
(−1)i iϑ ](α1,...,αi ,...,αd )dαi ,
where for ϑ ∈ Xd (M) we denote by ϑ ] : ∧d−1T ∗M → TM the map given by
ϑ](α1, . . . ,αk−1)(α) := ϑ(α1, . . . ,αk−1,α).
This map:
(i) takes the Schouten bracket to the graded bracket of multi-derivations.
(ii) image is multi-derivations D with symbol σD ∈ ∧d T ?M ⊂ ∧d−1TM⊗TM
Notation:
• J1E →M: first jet bundle of a bundle E →M.
• The Schouten bracket induces a graded Lie bracket:
[ , ] : Γ(J1(∧k+1TM))×Γ(J1(∧l+1TM))→ Γ(J1(∧k+l+1TM)),
characterized by
[J1ϑ1,J1
ϑ2] := J1[ϑ1,ϑ2].
Propositionτ ∈ Γ(J1(∧2TM)) with
[τ,τ] = 0
←→
Lie algebroids (T ∗M, [·, ·],ρ) withρ : T ∗M → TM skew-symmetric
Holonomic jets (i.e.., Poisson structures) correspond to Liealgebroids for which id : T ∗M → T ∗M is a closed IM 2-form.
Proof.i : J1(∧d TM) → Derd (T ∗M),
i(J1ϑ)(α1, . . . ,αd ) := d(ϑ(α1, . . . ,αd ))−
d
∑i=1
(−1)i iϑ ](α1,...,αi ,...,αd )dαi ,
where for ϑ ∈ Xd (M) we denote by ϑ ] : ∧d−1T ∗M → TM the map given by
ϑ](α1, . . . ,αk−1)(α) := ϑ(α1, . . . ,αk−1,α).
This map:
(i) takes the Schouten bracket to the graded bracket of multi-derivations.
(ii) image is multi-derivations D with symbol σD ∈ ∧d T ?M ⊂ ∧d−1TM⊗TM
Propositionτ ∈ Γ(J1(∧2TM)) with
[τ,τ] = 0
←→
Lie algebroids (T ∗M, [·, ·],ρ) withρ : T ∗M → TM skew-symmetric
Holonomic jets (i.e.., Poisson structures) correspond to Liealgebroids for which id : T ∗M → T ∗M is a closed IM 2-form.
Proof.i : J1(∧d TM) → Derd (T ∗M),
i(J1ϑ)(α1, . . . ,αd ) := d(ϑ(α1, . . . ,αd ))−
d
∑i=1
(−1)i iϑ ](α1,...,αi ,...,αd )dαi ,
where for ϑ ∈ Xd (M) we denote by ϑ ] : ∧d−1T ∗M → TM the map given by
ϑ](α1, . . . ,αk−1)(α) := ϑ(α1, . . . ,αk−1,α).
This map:
(i) takes the Schouten bracket to the graded bracket of multi-derivations.
(ii) image is multi-derivations D with symbol σD ∈ ∧d T ?M ⊂ ∧d−1TM⊗TM
Propositionτ ∈ Γ(J1(∧2TM)) with
[τ,τ] = 0
←→
Lie algebroids (T ∗M, [·, ·],ρ) withρ : T ∗M → TM skew-symmetric
Holonomic jets (i.e.., Poisson structures) correspond to Liealgebroids for which id : T ∗M → T ∗M is a closed IM 2-form.
Proof.i : J1(∧d TM) → Derd (T ∗M),
i(J1ϑ)(α1, . . . ,αd ) := d(ϑ(α1, . . . ,αd ))−
d
∑i=1
(−1)i iϑ ](α1,...,αi ,...,αd )dαi ,
where for ϑ ∈ Xd (M) we denote by ϑ ] : ∧d−1T ∗M → TM the map given by
ϑ](α1, . . . ,αk−1)(α) := ϑ(α1, . . . ,αk−1,α).
This map:
(i) takes the Schouten bracket to the graded bracket of multi-derivations.
(ii) image is multi-derivations D with symbol σD ∈ ∧d T ?M ⊂ ∧d−1TM⊗TM
Propositionτ ∈ Γ(J1(∧2TM)) with
[τ,τ] = 0
←→
Lie algebroids (T ∗M, [·, ·],ρ) withρ : T ∗M → TM skew-symmetric
Holonomic jets (i.e.., Poisson structures) correspond to Liealgebroids for which id : T ∗M → T ∗M is a closed IM 2-form.
Proof.i : J1(∧d TM) → Derd (T ∗M),
i(J1ϑ)(α1, . . . ,αd ) := d(ϑ(α1, . . . ,αd ))−
d
∑i=1
(−1)i iϑ ](α1,...,αi ,...,αd )dαi ,
where for ϑ ∈ Xd (M) we denote by ϑ ] : ∧d−1T ∗M → TM the map given by
ϑ](α1, . . . ,αk−1)(α) := ϑ(α1, . . . ,αk−1,α).
This map:
(i) takes the Schouten bracket to the graded bracket of multi-derivations.
(ii) image is multi-derivations D with symbol σD ∈ ∧d T ?M ⊂ ∧d−1TM⊗TM
Notation:
• S ⊂M: closed submanifold.
• X•S(M) := ϑ ∈ X•(M) : ϑ |S ∈ Γ(∧•TS)
The restriction J1(∧•TM)|S → S has the sub-bundle:
J1S(∧•TM) := τ ∈ J1(∧•TM)|S : pr∧•TSM(τ) ∈ ∧•TS.
which comes with
I projections:J1
S(∧•TM) //
((
J1(∧•TS)
∧•TS
I a graded Lie bracket on its sections
[ , ] : Γ(J1S(∧k+1TM))×Γ(J1
S(∧l+1TM))→ Γ(J1S(∧k+l+1TM)).
We call τ ∈ Γ(J1S(∧•TM)) holonomic if τ = (J1ϑ)|S for ϑ ∈ X•S(M)
Notation:
• S ⊂M: closed submanifold.
• X•S(M) := ϑ ∈ X•(M) : ϑ |S ∈ Γ(∧•TS)
The restriction J1(∧•TM)|S → S has the sub-bundle:
J1S(∧•TM) := τ ∈ J1(∧•TM)|S : pr∧•TSM(τ) ∈ ∧•TS.
which comes with
I projections:J1
S(∧•TM) //
((
J1(∧•TS)
∧•TS
I a graded Lie bracket on its sections
[ , ] : Γ(J1S(∧k+1TM))×Γ(J1
S(∧l+1TM))→ Γ(J1S(∧k+l+1TM)).
We call τ ∈ Γ(J1S(∧•TM)) holonomic if τ = (J1ϑ)|S for ϑ ∈ X•S(M)
Notation:
• S ⊂M: closed submanifold.
• X•S(M) := ϑ ∈ X•(M) : ϑ |S ∈ Γ(∧•TS)
The restriction J1(∧•TM)|S → S has the sub-bundle:
J1S(∧•TM) := τ ∈ J1(∧•TM)|S : pr∧•TSM(τ) ∈ ∧•TS.
which comes with
I projections:J1
S(∧•TM) //
((
J1(∧•TS)
∧•TS
I a graded Lie bracket on its sections
[ , ] : Γ(J1S(∧k+1TM))×Γ(J1
S(∧l+1TM))→ Γ(J1S(∧k+l+1TM)).
We call τ ∈ Γ(J1S(∧•TM)) holonomic if τ = (J1ϑ)|S for ϑ ∈ X•S(M)
Notation:
• S ⊂M: closed submanifold.
• X•S(M) := ϑ ∈ X•(M) : ϑ |S ∈ Γ(∧•TS)
The restriction J1(∧•TM)|S → S has the sub-bundle:
J1S(∧•TM) := τ ∈ J1(∧•TM)|S : pr∧•TSM(τ) ∈ ∧•TS.
which comes with
I projections:J1
S(∧•TM) //
((
J1(∧•TS)
∧•TS
I a graded Lie bracket on its sections
[ , ] : Γ(J1S(∧k+1TM))×Γ(J1
S(∧l+1TM))→ Γ(J1S(∧k+l+1TM)).
We call τ ∈ Γ(J1S(∧•TM)) holonomic if τ = (J1ϑ)|S for ϑ ∈ X•S(M)
Notation:
• S ⊂M: closed submanifold.
• X•S(M) := ϑ ∈ X•(M) : ϑ |S ∈ Γ(∧•TS)
The restriction J1(∧•TM)|S → S has the sub-bundle:
J1S(∧•TM) := τ ∈ J1(∧•TM)|S : pr∧•TSM(τ) ∈ ∧•TS.
which comes with
I projections:J1
S(∧•TM) //
((
J1(∧•TS)
∧•TS
I a graded Lie bracket on its sections
[ , ] : Γ(J1S(∧k+1TM))×Γ(J1
S(∧l+1TM))→ Γ(J1S(∧k+l+1TM)).
We call τ ∈ Γ(J1S(∧•TM)) holonomic if τ = (J1ϑ)|S for ϑ ∈ X•S(M)
Notation:
• S ⊂M: closed submanifold.
• X•S(M) := ϑ ∈ X•(M) : ϑ |S ∈ Γ(∧•TS)
• J1X•S(M): holonomic sections of J1S(∧•TM)
There is a projection
X•S(M)J1
S //
%%
J1X•S(M)
X•(S)
ϑ 7→ J1Sϑ := (J1
ϑ)|S ,
which gives a canonical identification for the holonomic sections:
J1X•S(M)' X•S(M)/(I2S ·X
•(M)),
where IS ⊂ C∞(M) denotes the vanishing ideal of S.
Notation:
• S ⊂M: closed submanifold.
• X•S(M) := ϑ ∈ X•(M) : ϑ |S ∈ Γ(∧•TS)
• J1X•S(M): holonomic sections of J1S(∧•TM)
There is a projection
X•S(M)J1
S //
%%
J1X•S(M)
X•(S)
ϑ 7→ J1Sϑ := (J1
ϑ)|S ,
which gives a canonical identification for the holonomic sections:
J1X•S(M)' X•S(M)/(I2S ·X
•(M)),
where IS ⊂ C∞(M) denotes the vanishing ideal of S.
Question. What is the geometric meaning of τ ∈ Γ(J1S(∧2TM)) with
[τ,τ] = 0?
DefinitionAn element of the space:
J10X
2S(M) := τ ∈ J1X2
S(M) : [τ,τ] = 0,
is called a 1st order jet of a Poisson structure at S.
PropositionLet µS : T ∗SM → T ∗S, α 7→ α|TS . Then:
τ ∈ Γ(J1S(∧2TM)) with
[τ,τ] = 0
←→
(T ∗SM, [·, ·],ρ) with〈µS(α),ρ(β )〉=−〈µS(β ),ρ(α)〉
Holonomic jets τ ∈ J1X2
S(M) correspond to Lie algebroids forwhich µS is a closed IM 2-form.
DefinitionAn element of the space:
J10X
2S(M) := τ ∈ J1X2
S(M) : [τ,τ] = 0,
is called a 1st order jet of a Poisson structure at S.
PropositionLet µS : T ∗SM → T ∗S, α 7→ α|TS . Then:
τ ∈ Γ(J1S(∧2TM)) with
[τ,τ] = 0
←→
(T ∗SM, [·, ·],ρ) with〈µS(α),ρ(β )〉=−〈µS(β ),ρ(α)〉
Holonomic jets τ ∈ J1X2
S(M) correspond to Lie algebroids forwhich µS is a closed IM 2-form.
Proof.Same as before, where now:
i : J1S(∧
d TM) → Derd (T ∗SM).
DefinitionAn element of the space:
J10X
2S(M) := τ ∈ J1X2
S(M) : [τ,τ] = 0,
is called a 1st order jet of a Poisson structure at S.
PropositionLet µS : T ∗SM → T ∗S, α 7→ α|TS . Then:
τ ∈ Γ(J1S(∧2TM)) with
[τ,τ] = 0
←→
(T ∗SM, [·, ·],ρ) with〈µS(α),ρ(β )〉=−〈µS(β ),ρ(α)〉
Holonomic jets τ ∈ J1X2
S(M) correspond to Lie algebroids forwhich µS is a closed IM 2-form.
DefinitionAn element of the space:
J10X
2S(M) := τ ∈ J1X2
S(M) : [τ,τ] = 0,
is called a 1st order jet of a Poisson structure at S.
Summary of first order dataA first order jet of Poisson structure along a submanfiold S ⊂M, canbe thought as either:
I A class of bivector fields tangent to S:
τ ∈ J10X
2S(M) =
π ∈ X2S(M) : [π,π] ∈ I2
S ·X3S(M)
I2S ·X
2S(M)
I A Lie algebroid τ = (T ∗SM, [·, ·],ρ) for which the restrictionµS : T ∗SM → T ∗S, α 7→ α|TS , is a closed IM 2-form.
There is a commutative diagram:
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
These are the only things you need to remember!
Summary of first order dataA first order jet of Poisson structure along a submanfiold S ⊂M, canbe thought as either:
I A class of bivector fields tangent to S:
τ ∈ J10X
2S(M) =
π ∈ X2S(M) : [π,π] ∈ I2
S ·X3S(M)
I2S ·X
2S(M)
I A Lie algebroid τ = (T ∗SM, [·, ·],ρ) for which the restrictionµS : T ∗SM → T ∗S, α 7→ α|TS , is a closed IM 2-form.
There is a commutative diagram:
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
These are the only things you need to remember!
Example (Marcut 13)
• M = R3 with coordinates (x ,y ,z);
• S = (x ,y ,0) : x ,y ∈ R ⊂ R3 (xy -plane)
Claim 1: The bivector field π ∈ X2S(M) given by:
π = z∂x ∧∂y + xz∂x ∧∂z
defines a 1st order jet τ = J1S(π) ∈ J1
0X2S(M).
Claim 2: The 1st order jet τ is not in the image of J1S :
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
Example (Marcut 13)
• M = R3 with coordinates (x ,y ,z);
• S = (x ,y ,0) : x ,y ∈ R ⊂ R3 (xy -plane)
Claim 1: The bivector field π ∈ X2S(M) given by:
π = z∂x ∧∂y + xz∂x ∧∂z
defines a 1st order jet τ = J1S(π) ∈ J1
0X2S(M).
Proof:[π,π] = z2
∂x ∧∂y ∧∂z ∈ I2S ·X
3(M)
Claim 2: The 1st order jet τ is not in the image of J1S :
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
Example (Marcut 13)
• M = R3 with coordinates (x ,y ,z);
• S = (x ,y ,0) : x ,y ∈ R ⊂ R3 (xy -plane)
Claim 1: The bivector field π ∈ X2S(M) given by:
π = z∂x ∧∂y + xz∂x ∧∂z
defines a 1st order jet τ = J1S(π) ∈ J1
0X2S(M).
Claim 2: The 1st order jet τ is not in the image of J1S :
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
Example (Marcut 13)
• M = R3 with coordinates (x ,y ,z);
• S = (x ,y ,0) : x ,y ∈ R ⊂ R3 (xy -plane)
Claim 1: The bivector field π ∈ X2S(M) given by:
π = z∂x ∧∂y + xz∂x ∧∂z
defines a 1st order jet τ = J1S(π) ∈ J1
0X2S(M).
Claim 2: The 1st order jet τ is not in the image of J1S :
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
Proof: Any π ∈ X2S(M) such that J1(π) = τ takes the form:
π = (z + z2f1)∂x ∧∂y + (xz + z2f2)∂x ∧∂z + z2f3∂y ∧∂z , fi ∈ C∞(R3)
=⇒ [π, π] = z2(1 + O(1))∂x ∧∂y ∧∂z 6= 0
Example (Marcut 13)
• M = R3 with coordinates (x ,y ,z);
• S = (x ,y ,0) : x ,y ∈ R ⊂ R3 (xy -plane)
Claim 1: The bivector field π ∈ X2S(M) given by:
π = z∂x ∧∂y + xz∂x ∧∂z
defines a 1st order jet τ = J1S(π) ∈ J1
0X2S(M).
Claim 2: The 1st order jet τ is not in the image of J1S :
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
Proof: Any π ∈ X2S(M) such that J1(π) = τ takes the form:
π = (z + z2f1)∂x ∧∂y + (xz + z2f2)∂x ∧∂z + z2f3∂y ∧∂z , fi ∈ C∞(R3)
=⇒ [π, π] = z2(1 + O(1))∂x ∧∂y ∧∂z 6= 0
First order local models:
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
DefinitionGiven a class C ⊂ J1
0X2S(M) a first order local model for C is
a splitting σ : C −→ Germs(Poisson(M,S)) of the map J1S .
Theorem (Existence of local models)The class C ⊂ J1
0X2S(M) satisfying the partially split condition
admits a 1st order local model σ : C→ Germs(Poisson(M,S)).
First order local models:
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
DefinitionGiven a class C ⊂ J1
0X2S(M) a first order local model for C is
a splitting σ : C −→ Germs(Poisson(M,S)) of the map J1S .
Splittings give linear models:
• if π ∈ X2S(M) is Poisson and J1
S(π) ∈ C, then σ(J1S(π)) ∈ X2
S(M)is Poisson with the same first order jet at S as π.
Theorem (Existence of local models)The class C ⊂ J1
0X2S(M) satisfying the partially split condition
admits a 1st order local model σ : C→ Germs(Poisson(M,S)).
First order local models:
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
DefinitionGiven a class C ⊂ J1
0X2S(M) a first order local model for C is
a splitting σ : C −→ Germs(Poisson(M,S)) of the map J1S .
Theorem (Existence of local models)The class C ⊂ J1
0X2S(M) satisfying the partially split condition
admits a 1st order local model σ : C→ Germs(Poisson(M,S)).
Normal forms:
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
DefinitionA local model σ : C→ Germs(Poisson(M,S)) is called a localnormal form for C if given τ ∈ C every Poisson structure π
with J1S(π) = τ is locally isomorphic to the linear model σ(τ).
Theorem (Linearization/normal form)Let C ⊂ J1
0X2S(M) be the class for which T ∗SM is integrable by a
compact, Lie groupoid whose s-fibers have trivial 2nd deRham cohomology. It is partially split and admits a localnormal form σ .
Normal forms:
Germs(Poisson(M,S))J1
S //
))
J10X
2S(M)
wwPoisson(S)
DefinitionA local model σ : C→ Germs(Poisson(M,S)) is called a localnormal form for C if given τ ∈ C every Poisson structure π
with J1S(π) = τ is locally isomorphic to the linear model σ(τ).
Theorem (Linearization/normal form)Let C ⊂ J1
0X2S(M) be the class for which T ∗SM is integrable by a
compact, Lie groupoid whose s-fibers have trivial 2nd deRham cohomology. It is partially split and admits a localnormal form σ .
II. Global theory: groupoid coisotropic embeddings
Definition (Bursztyn-Crainic-Weinstein-Zhu 04)An over-symplectic groupoid is a groupoid GS ⇒ S togetherwith a closed multiplicative 2-form ωS ∈ Ω2(GS) such that
ker ωS ⊂ ker dsS ∩ker dtS .
Given GS ⇒ S with a closed multiplicative 2-form ωS , the followingare equivalent:
(i) (GS ,ωS) is over-symplectic;
(ii) ∃ Poisson structure πS ∈ X2(S) such that tS : (GS ,ωS)→ (S,πS)is a forward Dirac map.
If these hold, ker ωS has constant rank = dimGS−2dimS and theorbits of GS coincide with the symplectic leaves of (S,πS).
II. Global theory: groupoid coisotropic embeddings
Definition (Bursztyn-Crainic-Weinstein-Zhu 04)An over-symplectic groupoid is a groupoid GS ⇒ S togetherwith a closed multiplicative 2-form ωS ∈ Ω2(GS) such that
ker ωS ⊂ ker dsS ∩ker dtS .
Given GS ⇒ S with a closed multiplicative 2-form ωS , the followingare equivalent:
(i) (GS ,ωS) is over-symplectic;
(ii) ∃ Poisson structure πS ∈ X2(S) such that tS : (GS ,ωS)→ (S,πS)is a forward Dirac map.
If these hold, ker ωS has constant rank = dimGS−2dimS and theorbits of GS coincide with the symplectic leaves of (S,πS).
• Why is (GS ,ωS) called over-symplectic?
I If ker ωS is a simple foliation then one gets a symplecticgroupoid
(GS/ker ωS ,ωS)⇒ S,
which integrates the Poisson manifold (S,πS).
• Why are over-symplectic groupoids relevant for us?
I Given (G,ω)⇒ (M,π) and Poisson submanifold S ⊂M, then(G|S ,ω|S)⇒ S is an over-symplectic groupoid, coisotropicallyembedded in (G,ω).
I Given a coisotropic embedding of an over-symplectic groupoidi : (GS ,ωS) → (G,ω), the embedding of units (S,πS) → (M,π)is a Poisson map, and
I closed submanifolds S ⊂M correspond to close embeddingsand then S is saturated so GS = G|S .
• Why is (GS ,ωS) called over-symplectic?
I If ker ωS is a simple foliation then one gets a symplecticgroupoid
(GS/ker ωS ,ωS)⇒ S,
which integrates the Poisson manifold (S,πS).
• Why are over-symplectic groupoids relevant for us?
I Given (G,ω)⇒ (M,π) and Poisson submanifold S ⊂M, then(G|S ,ω|S)⇒ S is an over-symplectic groupoid, coisotropicallyembedded in (G,ω).
I Given a coisotropic embedding of an over-symplectic groupoidi : (GS ,ωS) → (G,ω), the embedding of units (S,πS) → (M,π)is a Poisson map, and
I closed submanifolds S ⊂M correspond to close embeddingsand then S is saturated so GS = G|S .
PropositionIf GS ⇒ S is a source 1-connected Lie groupoid with Liealgebroid AS → S, there is a 1:1 correspondence:
over-symplectic structuresω ∈ Ω2(GS)
←→
surjective closed IM 2-forms
µ : AS → T ∗S
Given:
• (GS ,ω)⇒ S: over-symplectic groupoid integrating (AS ,µ)
• coisotropic embedding i : (GS ,ωS) → (G,ω)
We obtain:
(i) a Poisson embedding (S,πS) → (M,πM);
(ii) an isomorphism (AS ,µ)' (T ∗SM,µS);
(iii) a solution (M,πM) of the realization problem for τ = (T ∗SM,µS):
J1SπM = τ.
PropositionIf GS ⇒ S is a source 1-connected Lie groupoid with Liealgebroid AS → S, there is a 1:1 correspondence:
over-symplectic structuresω ∈ Ω2(GS)
←→
surjective closed IM 2-forms
µ : AS → T ∗S
Given:
• (GS ,ω)⇒ S: over-symplectic groupoid integrating (AS ,µ)
• coisotropic embedding i : (GS ,ωS) → (G,ω)
We obtain:
(i) a Poisson embedding (S,πS) → (M,πM);
(ii) an isomorphism (AS ,µ)' (T ∗SM,µS);
(iii) a solution (M,πM) of the realization problem for τ = (T ∗SM,µS):
J1SπM = τ.
We can summarize this discussion by the diagram:
Germs of symplecticgroupoids (G,ω)⇒M
along a saturated S ⊂M
restrictionalong S //
((
over-symplecticgroupoids
(GS ,ωS)⇒ S
xx
coisotropicembedding
nn
symplectic groupoids(GS/kerωS ,ωS)⇒ S
.. and this is the groupoid version of the diagram:
Germs(Poisson(M,S))J1
S //
((
J10X
2S(M)
splittingnn
yyPoisson(S)
This is all you have to remember!
Example (Transitive over-symplectic groupoids)
Given
• principal K -bundle p : P→ S
• symplectic form η ∈ Ω2(S),
One obtains a transitive over-symplectic groupoid:
GS := P×K P⇒ S,
q∗ωS := (p pr2)∗η− (p pr1)∗η
where q : P×P→ P×K P.
I Every transitive over-symplectic groupoid (GS ,ωS) takes thisform.
I It has a coisotropic embedding in the symplectic groupoid:
G := (P×P)×K Lie(K )∗⇒ P×K Lie(K )∗
q∗ω := (p pr2)∗η− (p pr1)∗η + d〈pr∗2 θ , ·〉−d〈pr∗1 θ , ·〉
where q : P×P×Lie(K )∗→ (P×P)×K Lie(K )∗ and θ is aprincipal bundle connection.
Example (Transitive over-symplectic groupoids)
Given
• principal K -bundle p : P→ S
• symplectic form η ∈ Ω2(S),
One obtains a transitive over-symplectic groupoid:
GS := P×K P⇒ S,
q∗ωS := (p pr2)∗η− (p pr1)∗η
where q : P×P→ P×K P.
I Every transitive over-symplectic groupoid (GS ,ωS) takes thisform.
I It has a coisotropic embedding in the symplectic groupoid:
G := (P×P)×K Lie(K )∗⇒ P×K Lie(K )∗
q∗ω := (p pr2)∗η− (p pr1)∗η + d〈pr∗2 θ , ·〉−d〈pr∗1 θ , ·〉
where q : P×P×Lie(K )∗→ (P×P)×K Lie(K )∗ and θ is aprincipal bundle connection.
Example (Transitive over-symplectic groupoids)
Given
• principal K -bundle p : P→ S
• symplectic form η ∈ Ω2(S),
One obtains a transitive over-symplectic groupoid:
GS := P×K P⇒ S,
q∗ωS := (p pr2)∗η− (p pr1)∗η
where q : P×P→ P×K P.
I Every transitive over-symplectic groupoid (GS ,ωS) takes thisform.
I It has a coisotropic embedding in the symplectic groupoid:
G := (P×P)×K Lie(K )∗⇒ P×K Lie(K )∗
q∗ω := (p pr2)∗η− (p pr1)∗η + d〈pr∗2 θ , ·〉−d〈pr∗1 θ , ·〉
where q : P×P×Lie(K )∗→ (P×P)×K Lie(K )∗ and θ is aprincipal bundle connection.
Example (with no coisotropic embeddings)
Integrating the 1st jet τ along S = R2 defined by:
π = z∂x ∧∂y + xz∂x ∧∂z ∈ X2(R3)
one obtains an over-symplectic groupoid (GS ,ωS):
• (GS ,ω)⇒ S is the bundle of groups pr : R5 = R2×R3→ R2 withmultiplication:
(x ,y ,u1,v1,w1,u2,v2,w2) 7→ (u1 +u2,v1 +exu1v2,w1 +w2 +exu1 −1
xv2).
• ωS is the multiplicative closed 2-form:
ωS = dx ∧du + dy ∧dv + w dx ∧dy −x dy ∧dw .
=⇒ (GS ,ωS) does not admit a groupoid coisotropicembedding into some symplectic groupoid!
... but there are many over-symplectic bundles of Lie groups that admit coisotropicembeddings!
Example (with no coisotropic embeddings)
Integrating the 1st jet τ along S = R2 defined by:
π = z∂x ∧∂y + xz∂x ∧∂z ∈ X2(R3)
one obtains an over-symplectic groupoid (GS ,ωS):
• (GS ,ω)⇒ S is the bundle of groups pr : R5 = R2×R3→ R2 withmultiplication:
(x ,y ,u1,v1,w1,u2,v2,w2) 7→ (u1 +u2,v1 +exu1v2,w1 +w2 +exu1 −1
xv2).
• ωS is the multiplicative closed 2-form:
ωS = dx ∧du + dy ∧dv + w dx ∧dy −x dy ∧dw .
=⇒ (GS ,ωS) does not admit a groupoid coisotropicembedding into some symplectic groupoid!
... but there are many over-symplectic bundles of Lie groups that admit coisotropicembeddings!
Example (with no coisotropic embeddings)
Integrating the 1st jet τ along S = R2 defined by:
π = z∂x ∧∂y + xz∂x ∧∂z ∈ X2(R3)
one obtains an over-symplectic groupoid (GS ,ωS):
• (GS ,ω)⇒ S is the bundle of groups pr : R5 = R2×R3→ R2 withmultiplication:
(x ,y ,u1,v1,w1,u2,v2,w2) 7→ (u1 +u2,v1 +exu1v2,w1 +w2 +exu1 −1
xv2).
• ωS is the multiplicative closed 2-form:
ωS = dx ∧du + dy ∧dv + w dx ∧dy −x dy ∧dw .
=⇒ (GS ,ωS) does not admit a groupoid coisotropicembedding into some symplectic groupoid!
... but there are many over-symplectic bundles of Lie groups that admit coisotropicembeddings!
Example (with no coisotropic embeddings)
Integrating the 1st jet τ along S = R2 defined by:
π = z∂x ∧∂y + xz∂x ∧∂z ∈ X2(R3)
one obtains an over-symplectic groupoid (GS ,ωS):
• (GS ,ω)⇒ S is the bundle of groups pr : R5 = R2×R3→ R2 withmultiplication:
(x ,y ,u1,v1,w1,u2,v2,w2) 7→ (u1 +u2,v1 +exu1v2,w1 +w2 +exu1 −1
xv2).
• ωS is the multiplicative closed 2-form:
ωS = dx ∧du + dy ∧dv + w dx ∧dy −x dy ∧dw .
=⇒ (GS ,ωS) does not admit a groupoid coisotropicembedding into some symplectic groupoid!
... but there are many over-symplectic bundles of Lie groups that admit coisotropicembeddings!
Can’t we apply Gotay’s coisotropic embedding thm?
• (C,ωC): presymplectic manifold⇔ ωC closed with constant rank
Gotay’s normal form: A coisotropic embedding i : (C,ωC)→(X ,ω) is locally equivalent to:
(C,ωC) → ((ker ωC)∗,ω0)where:
ω0 := pr∗ωC + j∗ωcan
for some choice of splitting:
T ∗C = (Ker ω)∗⊕E =⇒ j : (ker ωC)∗ → T ∗C
Can’t we apply Gotay’s coisotropic embedding thm?
• (C,ωC): presymplectic manifold⇔ ωC closed with constant rank
Gotay’s normal form: A coisotropic embedding i : (C,ωC)→(X ,ω) is locally equivalent to:
(C,ωC) → ((ker ωC)∗,ω0)where:
ω0 := pr∗ωC + j∗ωcan
for some choice of splitting:
T ∗C = (Ker ω)∗⊕E =⇒ j : (ker ωC)∗ → T ∗C
Can’t we apply Gotay’s coisotropic embedding thm?• (C,ωC): presymplectic manifold⇔ ωC closed with constant rank
Gotay’s normal form: A coisotropic embedding i : (C,ωC)→(X ,ω) is locally equivalent to:
(C,ωC) → ((ker ωC)∗,ω0)where:
ω0 := pr∗ωC + j∗ωcan
for some choice of splitting:
T ∗C = (Ker ω)∗⊕E =⇒ j : (ker ωC)∗ → T ∗C
Idea of proof:T ∗CX
(ω−1)] //
TCX
C
zero
section
// (ker ωC)∗ '// ν(C)
'tubular
neighborhood
// X
Moser method: can choose tubular neighbhd (X ,ω)' ((ker ωC)∗,ω0)
Can’t we apply Gotay’s coisotropic embedding thm?• (C,ωC): presymplectic manifold⇔ ωC closed with constant rank
Gotay’s normal form: A coisotropic embedding i : (C,ωC)→(X ,ω) is locally equivalent to:
(C,ωC) → ((ker ωC)∗,ω0)where:
ω0 := pr∗ωC + j∗ωcan
for some choice of splitting:
T ∗C = (Ker ω)∗⊕E =⇒ j : (ker ωC)∗ → T ∗C
Idea of proof:T ∗CX
(ω−1)] //
TCX
C
zero
section
// (ker ωC)∗ '// ν(C)
'tubular
neighborhood
// X
Moser method: can choose tubular neighbhd (X ,ω)' ((ker ωC)∗,ω0)
Can’t we apply Gotay’s coisotropic embedding thm?
• (C,ωC): presymplectic manifold⇔ ωC closed with constant rank
Gotay’s normal form: A coisotropic embedding i : (C,ωC)→(X ,ω) is locally equivalent to:
(C,ωC) → ((ker ωC)∗,ω0)where:
ω0 := pr∗ωC + j∗ωcan
for some choice of splitting:
T ∗C = (Ker ω)∗⊕E =⇒ j : (ker ωC)∗ → T ∗C
What happens with (ker ωS)∗ for over-symplectic groupoid (GS ,ωS)?
I Is it a groupoid?
I Is it a symplectic groupoid?
LemmaFor an over-symplectic groupoid (GS ,ωS) with (AS ,µS):I k := (ker ωS)|S → S is a subbundle of TSGS contained in
the tangent to the isotropy groups and k = ker µS .
I ker ωS ⇒ S is a subgroupoid of TGS ⇒ TS canonicalisomorphic to the groupoid GS×S k⇒ S, where:
(g,v) · (h,w) = (gh,v + gw).
Important: ker ωS ⇒ S is a VB-groupoid:
GS n kker ωS =
//
$$
TGS
""GS
GS
OSS =
$$
// TS
##S S
LemmaFor an over-symplectic groupoid (GS ,ωS) with (AS ,µS):I k := (ker ωS)|S → S is a subbundle of TSGS contained in
the tangent to the isotropy groups and k = ker µS .
I ker ωS ⇒ S is a subgroupoid of TGS ⇒ TS canonicalisomorphic to the groupoid GS×S k⇒ S, where:
(g,v) · (h,w) = (gh,v + gw).
Important: ker ωS ⇒ S is a VB-groupoid:
GS n kker ωS =
//
$$
TGS
""GS
GS
OSS =
$$
// TS
##S S
ker ωS ⇒ S is a VB-groupoid:
GS n kker ωS =
//
$$
TGS
""GS
GS
OSS =
$$
// TS
##S S
I (ker ωS)∗ has groupoid structureI GS → (ker ωS)∗ is a closed subgroupoidI A coisotropic embedding i : (GS ,ωS) → (G,ω) yields a groupoid
isomorphism:
T ∗GSG
(ω−1)] //
TGSG
(ker ωS)∗ '
// ν(GS)
Dually, we have VB groupoid (ker ωS)∗ = GS n k∗:
GS n k∗(ker ωS)∗ =
$$
22 T ∗GSoooo
##GS
GS
k∗(ker ωS)|∗S =
$$
33oooo A∗S
##S S
AS ≡ Lie algebroid of GS k∗ = (ker µS)∗
I (ker ωS)∗ has groupoid structureI GS → (ker ωS)∗ is a closed subgroupoidI A coisotropic embedding i : (GS ,ωS) → (G,ω) yields a groupoid
isomorphism:
T ∗GSG
(ω−1)] //
TGSG
(ker ωS)∗ '
// ν(GS)
Dually, we have VB groupoid (ker ωS)∗ = GS n k∗:
GS n k∗(ker ωS)∗ =
$$
22 T ∗GSoooo
##GS
GS
k∗(ker ωS)|∗S =
$$
33oooo A∗S
##S S
AS ≡ Lie algebroid of GS k∗ = (ker µS)∗
I (ker ωS)∗ has groupoid structureI GS → (ker ωS)∗ is a closed subgroupoidI A coisotropic embedding i : (GS ,ωS) → (G,ω) yields a groupoid
isomorphism:
T ∗GSG
(ω−1)] //
TGSG
(ker ωS)∗ '
// ν(GS)
What about the symplectic structure on (ker ωS)∗?
It depends on a choice of splitting:
T ∗GS = (ker ωS)∗⊕E
Need to choose VB subgroupoid E ⊂ T ∗GS (it may not be possible!).
DefinitionAn over-symplectic groupoid (GS ,ωS) is called partially split ifthere is a VB groupoid morphism
Θ : (ker ωS)∗→ T ∗GS ,
splitting the projection T ∗GS → (ker ωS)∗. One calls Θ a partialsplitting.
What about the symplectic structure on (ker ωS)∗?
It depends on a choice of splitting:
T ∗GS = (ker ωS)∗⊕E
Need to choose VB subgroupoid E ⊂ T ∗GS (it may not be possible!).
DefinitionAn over-symplectic groupoid (GS ,ωS) is called partially split ifthere is a VB groupoid morphism
Θ : (ker ωS)∗→ T ∗GS ,
splitting the projection T ∗GS → (ker ωS)∗. One calls Θ a partialsplitting.
What about the symplectic structure on (ker ωS)∗?
It depends on a choice of splitting:
T ∗GS = (ker ωS)∗⊕E
Need to choose VB subgroupoid E ⊂ T ∗GS (it may not be possible!).
DefinitionAn over-symplectic groupoid (GS ,ωS) is called partially split ifthere is a VB groupoid morphism
Θ : (ker ωS)∗→ T ∗GS ,
splitting the projection T ∗GS → (ker ωS)∗. One calls Θ a partialsplitting.
What about the symplectic structure on (ker ωS)∗?
It depends on a choice of splitting:
T ∗GS = (ker ωS)∗⊕E
Need to choose VB subgroupoid E ⊂ T ∗GS (it may not be possible!).
DefinitionAn over-symplectic groupoid (GS ,ωS) is called partially split ifthere is a VB groupoid morphism
Θ : (ker ωS)∗→ T ∗GS ,
splitting the projection T ∗GS → (ker ωS)∗. One calls Θ a partialsplitting.
Symplectic groupoid local model of (GS,ωS)
Local Model:(ker ωS)∗ = GS n k∗⇒ k∗, ω0 := pr∗ωS + Θ∗ωcan
for a choice of partial splitting Θ : (ker ωS)∗→ T ∗GS :
Note: ω is multiplicative, closed, and non-degenerate in a groupoid neighborhood ofthe zero section.
Example (Transitive over-symplectic groupoids)
Recall transitive over-symplectic groupoids take the form
GS := P×K P⇒ S,
q∗ωS := (p pr2)∗η− (p pr1)∗η
for a principal K -bundle p : P→ (S,η). They are always partiallysplit and
partial splittingsΘ : (kerωS)∗→ T ∗GS
←→
principal bundleconnections θ
The resulting groupoid local model is exactly the one we saw before.
Symplectic groupoid local model of (GS,ωS)
Local Model:(ker ωS)∗ = GS n k∗⇒ k∗, ω0 := pr∗ωS + Θ∗ωcan
for a choice of partial splitting Θ : (ker ωS)∗→ T ∗GS :
Note: ω is multiplicative, closed, and non-degenerate in a groupoid neighborhood ofthe zero section.
Example (Transitive over-symplectic groupoids)
Recall transitive over-symplectic groupoids take the form
GS := P×K P⇒ S,
q∗ωS := (p pr2)∗η− (p pr1)∗η
for a principal K -bundle p : P→ (S,η).
They are always partiallysplit and
partial splittingsΘ : (kerωS)∗→ T ∗GS
←→
principal bundleconnections θ
The resulting groupoid local model is exactly the one we saw before.
Symplectic groupoid local model of (GS,ωS)
Local Model:(ker ωS)∗ = GS n k∗⇒ k∗, ω0 := pr∗ωS + Θ∗ωcan
for a choice of partial splitting Θ : (ker ωS)∗→ T ∗GS :
Note: ω is multiplicative, closed, and non-degenerate in a groupoid neighborhood ofthe zero section.
Example (Transitive over-symplectic groupoids)
Recall transitive over-symplectic groupoids take the form
GS := P×K P⇒ S,
q∗ωS := (p pr2)∗η− (p pr1)∗η
for a principal K -bundle p : P→ (S,η). They are always partiallysplit and
partial splittingsΘ : (kerωS)∗→ T ∗GS
←→
principal bundleconnections θ
The resulting groupoid local model is exactly the one we saw before.
Symplectic groupoid local model of (GS,ωS)
Local Model:(ker ωS)∗ = GS n k∗⇒ k∗, ω0 := pr∗ωS + Θ∗ωcan
for a choice of partial splitting Θ : (ker ωS)∗→ T ∗GS :
Note: ω is multiplicative, closed, and non-degenerate in a groupoid neighborhood ofthe zero section.
Example (Transitive over-symplectic groupoids)
Recall transitive over-symplectic groupoids take the form
GS := P×K P⇒ S,
q∗ωS := (p pr2)∗η− (p pr1)∗η
for a principal K -bundle p : P→ (S,η). They are always partiallysplit and
partial splittingsΘ : (kerωS)∗→ T ∗GS
←→
principal bundleconnections θ
The resulting groupoid local model is exactly the one we saw before.
Uniqueness of local models
ω0 = pr∗ωS︸ ︷︷ ︸constant
+Θ∗ωcan︸ ︷︷ ︸linear
Where with Mt : GS n k∗→GS n k∗, (g,v) 7→ (g, tv):
• ω constant⇐⇒ M∗t ω = ω, ∀t ;
• ω linear⇐⇒ M∗t ω = tω, ∀t ;
Linear, closed, multiplicative 2-forms are exact with a linearmultiplicative primitive, and this leads to:
PropositionThe local models associated with partial splittings Θ0 and Θ1are isomorphic: there exists a local groupoid automorphismΦ : GS n k∗→GS n k∗ fixing GS , such that Φ∗ω1 = ω0.
Uniqueness of local models
ω0 = pr∗ωS︸ ︷︷ ︸constant
+Θ∗ωcan︸ ︷︷ ︸linear
Where with Mt : GS n k∗→GS n k∗, (g,v) 7→ (g, tv):
• ω constant⇐⇒ M∗t ω = ω, ∀t ;
• ω linear⇐⇒ M∗t ω = tω, ∀t ;
Linear, closed, multiplicative 2-forms are exact with a linearmultiplicative primitive, and this leads to:
PropositionThe local models associated with partial splittings Θ0 and Θ1are isomorphic: there exists a local groupoid automorphismΦ : GS n k∗→GS n k∗ fixing GS , such that Φ∗ω1 = ω0.
Existence of local models
If (GS ,ωS) is over-symplectic with k := ker(ωS)|S = ker µS , tfae:
(i) (GS ,ωS) is partially split;
(ii) ∃ ω lin ∈ Ω2(GS n k∗) linear, closed, multiplicative, with:
ωlin|k×Sk
∗ = ωcan, (k×S k∗ ⊂ T (GS n k∗));
(iii) ∃ α ∈ Ω1(GS ,k) multiplicative with
α|k = id (k⊂ TSGS);
Moreover, these are equivalent to existing ω ∈ Ω2(GS n k∗) closed,multiplicative, non-degenerate at GS with:
i∗ω = ωS (i : GS →GS n k∗).
Existence of local models
If (GS ,ωS) is over-symplectic with k := ker(ωS)|S = ker µS , tfae:
(i) (GS ,ωS) is partially split;
(ii) ∃ ω lin ∈ Ω2(GS n k∗) linear, closed, multiplicative, with:
ωlin|k×Sk
∗ = ωcan, (k×S k∗ ⊂ T (GS n k∗));
(iii) ∃ α ∈ Ω1(GS ,k) multiplicative with
α|k = id (k⊂ TSGS);
Moreover, these are equivalent to existing ω ∈ Ω2(GS n k∗) closed,multiplicative, non-degenerate at GS with:
i∗ω = ωS (i : GS →GS n k∗).
Existence of local models
If (GS ,ωS) is over-symplectic with k := ker(ωS)|S = ker µS , tfae:
(i) (GS ,ωS) is partially split;
(ii) ∃ ω lin ∈ Ω2(GS n k∗) linear, closed, multiplicative, with:
ωlin|k×Sk
∗ = ωcan, (k×S k∗ ⊂ T (GS n k∗));
(iii) ∃ α ∈ Ω1(GS ,k) multiplicative with
α|k = id (k⊂ TSGS);
Moreover, these are equivalent to existing ω ∈ Ω2(GS n k∗) closed,multiplicative, non-degenerate at GS with:
i∗ω = ωS (i : GS →GS n k∗).
Proof: Given such ω, the limit:
ωlin := lim
t→0
1t
M∗t (ω−pr∗ωS),
satisfies (ii).
Existence of local models
If (GS ,ωS) is over-symplectic with k := ker(ωS)|S = ker µS , tfae:
(i) (GS ,ωS) is partially split;
(ii) ∃ ω lin ∈ Ω2(GS n k∗) linear, closed, multiplicative, with:
ωlin|k×Sk
∗ = ωcan, (k×S k∗ ⊂ T (GS n k∗));
(iii) ∃ α ∈ Ω1(GS ,k) multiplicative with
α|k = id (k⊂ TSGS);
Moreover, these are equivalent to existing ω ∈ Ω2(GS n k∗) closed,multiplicative, non-degenerate at GS with:
i∗ω = ωS (i : GS →GS n k∗).
Example (Important!)All s-proper, over-symplectic groupoids are partially split.
Groupoid coisotropic neighborhood theorem
TheoremA symplectic groupoid (G,ω) which is invariantly linearizablearound a saturated S ⊂M is partially split and locallyisomorphic to the symplectic groupoid local model of (GS ,ωS).
Note: No properness assumption!
CorollaryA s-proper symplectic groupoid (G,ω) is locally isomorphicaround a saturated S ⊂M to the symplectic groupoid localmodel of (GS ,ωS).
Groupoid coisotropic neighborhood theorem
TheoremA symplectic groupoid (G,ω) which is invariantly linearizablearound a saturated S ⊂M is partially split and locallyisomorphic to the symplectic groupoid local model of (GS ,ωS).
Note: No properness assumption!
CorollaryA s-proper symplectic groupoid (G,ω) is locally isomorphicaround a saturated S ⊂M to the symplectic groupoid localmodel of (GS ,ωS).
III. Local theory
Philosophical Principle:
Poisson Geometry = Infinitesimal MultiplicativeSymplectic Geometry
This leads to a Dictionary.
To figure out its entries, one only has to recall the correspondences:Lie groupoids
←→
Lie algebroids
multiplicative forms(with coefficients)
←→
IM forms
(with coefficients)
multiplicativevector fields
←→
Lie algebroidderivations
III. Local theory
Philosophical Principle:
Poisson Geometry = Infinitesimal MultiplicativeSymplectic Geometry
This leads to a Dictionary.
To figure out its entries, one only has to recall the correspondences:Lie groupoids
←→
Lie algebroids
multiplicative forms(with coefficients)
←→
IM forms
(with coefficients)
multiplicativevector fields
←→
Lie algebroidderivations
III. Local theory
Philosophical Principle:
Poisson Geometry = Infinitesimal MultiplicativeSymplectic Geometry
This leads to a Dictionary.
To figure out its entries, one only has to recall the correspondences:Lie groupoids
←→
Lie algebroids
multiplicative forms(with coefficients)
←→
IM forms
(with coefficients)
multiplicativevector fields
←→
Lie algebroidderivations
Dictionary
Global side Infinitesimal side
symplectic groupoid(G,ω)⇒M
Poisson manifold (M,π)⇔ algebroid T ∗M such that
id : T ∗M → T ∗M is closed IM-form
saturated submanifoldS ⊂M
Poisson submanifoldS ⊂M
over-symplectic groupoid(GS ,ωS)⇒ S
1st order local data τ ∈ J1SX0(M)
⇔ algebroid T ∗SM such thatµS : T ∗SM → T ∗S is closed IM-form
coisotropic embedding(GS ,ωS) → (G,ω)
S ⊂ (M,π) a realizationof τ = (T ∗SM,µS) with τ = J1
Sπ
partial splittingΘ : GS n k∗→ T ∗GS
partial splitting for T ∗SM
local normal(GS n k∗,ω0)
local model for τ = (T ∗SM,µS)
Local model - algebroid aspects
• τ = (T ∗SM,µS): 1st order local data
• k = ker µS → S: it is a representation of T ∗SM with ∇αs = [α,s]
So we can form the action algebroid:
T ∗SM n k∗→ k∗
PropositionIf τ = J1
Sπ for a Poisson structure (M,π), then k = ν(S)∗ and
T ∗SM n k∗ ' T ∗SM nν(S)
is the linear approximation of the Lie algebroid T ∗M along thesaturated submanifold S ⊂M.
Local model - algebroid aspects
• τ = (T ∗SM,µS): 1st order local data
• k = ker µS → S: it is a representation of T ∗SM with ∇αs = [α,s]
So we can form the action algebroid:
T ∗SM n k∗→ k∗
PropositionIf τ = J1
Sπ for a Poisson structure (M,π), then k = ν(S)∗ and
T ∗SM n k∗ ' T ∗SM nν(S)
is the linear approximation of the Lie algebroid T ∗M along thesaturated submanifold S ⊂M.
Local model - Poisson geometric aspectsNote: In general, T ∗SM n k∗→ k∗ is not the cotangent algebroid of aPoisson structure on k∗.
DefinitionA pair (T ∗SM,µS) is called partially split if there exists a Liealgebroid morphism
θ : T ∗SM n k∗→ T ∗(T ∗SM)
which is a splitting of the natural projection
p : T ∗(T ∗SM)→ T ∗SM n k∗.
We call θ a partial splitting.
Note: For a Lie algebroid A→M we have its cotangent algebroid:
T ∗A→ A∗,
It comes with a canonical, non-degenerate, linear, closed, IM 2-form:
µcan : T ∗A→ T ∗A∗.
This is commonly known as the Tulczyjew isomorphism.
Local model - Poisson geometric aspectsNote: In general, T ∗SM n k∗→ k∗ is not the cotangent algebroid of aPoisson structure on k∗.
DefinitionA pair (T ∗SM,µS) is called partially split if there exists a Liealgebroid morphism
θ : T ∗SM n k∗→ T ∗(T ∗SM)
which is a splitting of the natural projection
p : T ∗(T ∗SM)→ T ∗SM n k∗.
We call θ a partial splitting.
Note: For a Lie algebroid A→M we have its cotangent algebroid:
T ∗A→ A∗,
It comes with a canonical, non-degenerate, linear, closed, IM 2-form:
µcan : T ∗A→ T ∗A∗.
This is commonly known as the Tulczyjew isomorphism.
Local model - Poisson geometric aspectsNote: In general, T ∗SM n k∗→ k∗ is not the cotangent algebroid of aPoisson structure on k∗.
DefinitionA pair (T ∗SM,µS) is called partially split if there exists a Liealgebroid morphism
θ : T ∗SM n k∗→ T ∗(T ∗SM)
which is a splitting of the natural projection
p : T ∗(T ∗SM)→ T ∗SM n k∗.
We call θ a partial splitting.
Note: For a Lie algebroid A→M we have its cotangent algebroid:
T ∗A→ A∗,
It comes with a canonical, non-degenerate, linear, closed, IM 2-form:
µcan : T ∗A→ T ∗A∗.
This is commonly known as the Tulczyjew isomorphism.
Local modelIf (T ∗SM,µS) is partially split, a splitting θ : T ∗SM n k∗→ T ∗(T ∗SM) yieldsa closed IM 2-form:
µ0 = pr∗ µS︸ ︷︷ ︸constant
+θ∗µcan︸ ︷︷ ︸
linear
Definition
(T ∗SM n k∗,µ0).
is the local model of the partially split jet τ = (T ∗SM,µS),associated with the splitting θ .
I The form µ0 is non-degenerate in a neighborhood U of the zerosection S → T ∗SM n k∗.
I In U, we get a local model Poisson structure:
π0 = ρ µ−10 .
I (U,π0) has 1st order jet at S the local data: J1Sπ0 = τ.
Local modelIf (T ∗SM,µS) is partially split, a splitting θ : T ∗SM n k∗→ T ∗(T ∗SM) yieldsa closed IM 2-form:
µ0 = pr∗ µS︸ ︷︷ ︸constant
+θ∗µcan︸ ︷︷ ︸
linear
Definition
(T ∗SM n k∗,µ0).
is the local model of the partially split jet τ = (T ∗SM,µS),associated with the splitting θ .
I The form µ0 is non-degenerate in a neighborhood U of the zerosection S → T ∗SM n k∗.
I In U, we get a local model Poisson structure:
π0 = ρ µ−10 .
I (U,π0) has 1st order jet at S the local data: J1Sπ0 = τ.
Local modelIf (T ∗SM,µS) is partially split, a splitting θ : T ∗SM n k∗→ T ∗(T ∗SM) yieldsa closed IM 2-form:
µ0 = pr∗ µS︸ ︷︷ ︸constant
+θ∗µcan︸ ︷︷ ︸
linear
Definition
(T ∗SM n k∗,µ0).
is the local model of the partially split jet τ = (T ∗SM,µS),associated with the splitting θ .
I The form µ0 is non-degenerate in a neighborhood U of the zerosection S → T ∗SM n k∗.
I In U, we get a local model Poisson structure:
π0 = ρ µ−10 .
I (U,π0) has 1st order jet at S the local data: J1Sπ0 = τ.
If (GS ,ωS) is over-symplectic with k := ker(ωS)|S = ker µ, tfae:
(i) (GS ,ωS) is partially split;
(ii) ∃ ω lin ∈ Ω2(GS n k∗) linear, closed, multiplicative, with:
ωlin|k×Sk
∗ = ωcan, (k×S k∗ ⊂ T (GS n k∗));
(iii) ∃ α ∈ Ω1(GS ,k) multiplicative with
α|k = id (k⊂ TSGS);
Moreover, these are equivalent to existing ω ∈ Ω2(GS n k∗) closed,multiplicative, non-degenerate at GS with:
i∗ω = ωS (i : GS →GS n k∗).
If τ = (T ∗SM,µS) is first order local data, k = ker µS , tfae:
(i) (T ∗SM,µS) is partially split;
(ii) ∃ linear, closed, IM form µ lin ∈ Ω2IM(T ∗SM) such that
µlin|S : T ∗SM → T ∗S⊕ k, (prk µ lin|S)|k = id
(iii) ∃ k-valued, IM form D ∈ Ω1(T ∗SM,k) whose symboll : T ∗SM → k satisfies l |k = id.
Moreover, these are equivalent to existing a closed, IM 2-formµ ∈ Ω2
IM(T ∗SM) satisfying:
µ|S : T ∗S⊕ k→ T ∗S⊕ k, prT ∗S µ|S = µS , (prk µ|S)|k = id
Example (Transitive algebroids are split)
0 // k // T ∗SMρ //
l
hh TS //τ
kk 0
We obtain a linear operator D : Γ(T ∗SM)→Ω1(S,k) satisfying (iii):
D(α)(X) := l([τ(X),α])
The corresponding local model is precisely Vorobjev’s local model.
If τ = (T ∗SM,µS) is first order local data, k = ker µS , tfae:
(i) (T ∗SM,µS) is partially split;
(ii) ∃ linear, closed, IM form µ lin ∈ Ω2IM(T ∗SM) such that
µlin|S : T ∗SM → T ∗S⊕ k, (prk µ lin|S)|k = id
(iii) ∃ k-valued, IM form D ∈ Ω1(T ∗SM,k) whose symboll : T ∗SM → k satisfies l |k = id.
Moreover, these are equivalent to existing a closed, IM 2-formµ ∈ Ω2
IM(T ∗SM) satisfying:
µ|S : T ∗S⊕ k→ T ∗S⊕ k, prT ∗S µ|S = µS , (prk µ|S)|k = id
Example (Transitive algebroids are split)
0 // k // T ∗SMρ //
l
hh TS //τ
kk 0
We obtain a linear operator D : Γ(T ∗SM)→Ω1(S,k) satisfying (iii):
D(α)(X) := l([τ(X),α])
The corresponding local model is precisely Vorobjev’s local model.
If τ = (T ∗SM,µS) is first order local data, k = ker µS , tfae:
(i) (T ∗SM,µS) is partially split;
(ii) ∃ linear, closed, IM form µ lin ∈ Ω2IM(T ∗SM) such that
µlin|S : T ∗SM → T ∗S⊕ k, (prk µ lin|S)|k = id
(iii) ∃ k-valued, IM form D ∈ Ω1(T ∗SM,k) whose symboll : T ∗SM → k satisfies l |k = id.
Moreover, these are equivalent to existing a closed, IM 2-formµ ∈ Ω2
IM(T ∗SM) satisfying:
µ|S : T ∗S⊕ k→ T ∗S⊕ k, prT ∗S µ|S = µS , (prk µ|S)|k = id
Example (Transitive algebroids are split)
0 // k // T ∗SMρ //
l
hh TS //τ
kk 0
We obtain a linear operator D : Γ(T ∗SM)→Ω1(S,k) satisfying (iii):
D(α)(X) := l([τ(X),α])
The corresponding local model is precisely Vorobjev’s local model.
Uniqueness of the local model
PropositionLet (T ∗SM,µS) be a partially split. Given splittings θ0 and θ1,there is an algebroid automorphism Φ : T ∗SM n k∗→ T ∗SM n k∗
defined around S such that:
Φ|T ∗SM = id, Φ∗µ0 = µ1.
Proof: Every linear, closed, IM 2-form is exact. So apply a IM versionof Moser to the path:
µt = µS + tµlin1 + (1− t)µ
lin0 ,
Instead of the flow of a vector field, use the flow of a time-dependentLie algebroid derivation.
Uniqueness of the local model
PropositionLet (T ∗SM,µS) be a partially split. Given splittings θ0 and θ1,there is an algebroid automorphism Φ : T ∗SM n k∗→ T ∗SM n k∗
defined around S such that:
Φ|T ∗SM = id, Φ∗µ0 = µ1.
Proof: Every linear, closed, IM 2-form is exact. So apply a IM versionof Moser to the path:
µt = µS + tµlin1 + (1− t)µ
lin0 ,
Instead of the flow of a vector field, use the flow of a time-dependentLie algebroid derivation.
Linearization Theorem
TheoremLet τ = (T ∗SM,µS) be a first order jet of a Poisson structure. IfT ∗SM is integrable by a compact Lie groupoid whose s-fibershave trivial 2nd de Rham cohomology, then:
(i) (T ∗SM,µS) is partially split, and
(ii) any (M,π) with J1Sπ = τ is locally isomorphic to the local
model.
Assume: ∃ integration (GS ,ωS)⇒ S compact, has 1-connecteds-fibers with trivial 2nd de Rham cohomology.
Geometric Proof: (GS ,ωS) It is partially split, so (i) follows. Theproof of (ii) consists of two steps:
1) Show that a neighborhood of S in (M,π) is integrable by ans-proper symplectic groupoid (G,ω);
2) Apply the groupoid coisotropic neighborhood theorem to theembedding (GS ,ωS) → (G,ω).
Linearization Theorem
TheoremLet τ = (T ∗SM,µS) be a first order jet of a Poisson structure. IfT ∗SM is integrable by a compact Lie groupoid whose s-fibershave trivial 2nd de Rham cohomology, then:
(i) (T ∗SM,µS) is partially split, and
(ii) any (M,π) with J1Sπ = τ is locally isomorphic to the local
model.
Nash-Moser Proof: Use [Marcut 2014] and apply groupoidcoisotropic normal form.
Assume: ∃ integration (GS ,ωS)⇒ S compact, has 1-connecteds-fibers with trivial 2nd de Rham cohomology.
Geometric Proof: (GS ,ωS) It is partially split, so (i) follows. Theproof of (ii) consists of two steps:
1) Show that a neighborhood of S in (M,π) is integrable by ans-proper symplectic groupoid (G,ω);
2) Apply the groupoid coisotropic neighborhood theorem to theembedding (GS ,ωS) → (G,ω).
Linearization Theorem
TheoremLet τ = (T ∗SM,µS) be a first order jet of a Poisson structure. IfT ∗SM is integrable by a compact Lie groupoid whose s-fibershave trivial 2nd de Rham cohomology, then:
(i) (T ∗SM,µS) is partially split, and
(ii) any (M,π) with J1Sπ = τ is locally isomorphic to the local
model.
Assume: ∃ integration (GS ,ωS)⇒ S compact, has 1-connecteds-fibers with trivial 2nd de Rham cohomology.
Geometric Proof: (GS ,ωS) It is partially split, so (i) follows. Theproof of (ii) consists of two steps:
1) Show that a neighborhood of S in (M,π) is integrable by ans-proper symplectic groupoid (G,ω);
2) Apply the groupoid coisotropic neighborhood theorem to theembedding (GS ,ωS) → (G,ω).
Linearization Theorem
TheoremLet τ = (T ∗SM,µS) be a first order jet of a Poisson structure. IfT ∗SM is integrable by a compact Lie groupoid whose s-fibershave trivial 2nd de Rham cohomology, then:
(i) (T ∗SM,µS) is partially split, and
(ii) any (M,π) with J1Sπ = τ is locally isomorphic to the local
model.
Assume: ∃ integration (GS ,ωS)⇒ S compact, has 1-connecteds-fibers with trivial 2nd de Rham cohomology.
Geometric Proof: (GS ,ωS) It is partially split, so (i) follows. Theproof of (ii) consists of two steps:
1) Show that a neighborhood of S in (M,π) is integrable by ans-proper symplectic groupoid (G,ω);
2) Apply the groupoid coisotropic neighborhood theorem to theembedding (GS ,ωS) → (G,ω).
Linearization Thm. - Comments on step 1 of the proof1) Show that a neighborhood of S in (M,π) is integrable by an
s-proper symplectic groupoid (G,ω);
This step uses:
(i) The path space approach to integrability;
(ii) A result from Crainic & RLF, “A geometric approach to Conn’slinearization theorem”.
(iii) The same result was already used by Crainic & Marcut, for theirtheorem of linearization around symplectic leaves.
PropositionLet F be a foliation of finite codimension on a Banach manifold X and letY ⊂ X be a submanifold which is saturated with respect to F (i.e., each leaf ofF which hits Y is contained in Y ). Assume that:
(H0) The holonomy of F at all points in Y is trivial.
(H1) The foliation FY := F |Y is simple, i.e. its leaves are the fibers of asubmersion p : Y → B into a compact manifold B.
(H2) The fibration p : Y → B is locally trivial.
Then one can find:
(i) a transversal TX ⊂ X to the foliation F such that TY := Y ∩TX is acomplete transversal to FY (i.e., intersects each leaf of FY at leastonce).
(ii) a retraction r : TX → TY .
(iii) an action of the holonomy of FY on r : TX → TY along F .
Moreover, the orbit space TX /HolTY(FY ) is a smooth (Hausdorff) manifold.
Linearization Thm. - Comments on step 1 of the proof
PropositionLet F be a foliation of finite codimension on a Banach manifold X and letY ⊂ X be a submanifold which is saturated with respect to F (i.e., each leaf ofF which hits Y is contained in Y ). Assume that:
(H0) The holonomy of F at all points in Y is trivial.
(H1) The foliation FY := F |Y is simple, i.e. its leaves are the fibers of asubmersion p : Y → B into a compact manifold B.
(H2) The fibration p : Y → B is locally trivial.
Then one can find:
(i) a transversal TX ⊂ X to the foliation F such that TY := Y ∩TX is acomplete transversal to FY (i.e., intersects each leaf of FY at leastonce).
(ii) a retraction r : TX → TY .
(iii) an action of the holonomy of FY on r : TX → TY along F .
Moreover, the orbit space TX /HolTY(FY ) is a smooth (Hausdorff) manifold.
Existence of other normal forms
Local normal forms were defined as splittings:
Germs(Poisson(M,S))J1
S //
((
J10X
2S(M)
splittingnn
yyPoisson(S)
I We described a normal form for partially split local data;
I For b-symplectic manifolds, the normal form around the singularlocus is a different kind of normal form!
Are there other kinds of local normal forms?