Filtered and Intersection Homology Jon Woolf, work in progress with - - PowerPoint PPT Presentation

filtered and intersection homology
SMART_READER_LITE
LIVE PREVIEW

Filtered and Intersection Homology Jon Woolf, work in progress with - - PowerPoint PPT Presentation

Filtered and Intersection Homology Jon Woolf, work in progress with Ryan Wissett April, 2016 Part I Review of intersection homology Singular intersection homology Perversities A perversity on a topologically stratified space X is a function p


slide-1
SLIDE 1

Filtered and Intersection Homology

Jon Woolf, work in progress with Ryan Wissett April, 2016

slide-2
SLIDE 2

Part I Review of intersection homology

slide-3
SLIDE 3

Singular intersection homology

Perversities

A perversity on a topologically stratified space X is a function p : {strata of X} → Z. If

  • 1. p(S) = p(codim S) for some p : N → Z
  • 2. p(k) = 0 for k ≤ 2
  • 3. p(k + 1) = p(k) or p(k) + 1.

then it is a Goresky–MacPherson (GM) perversity.

slide-4
SLIDE 4

Singular intersection homology

Perversities

A perversity on a topologically stratified space X is a function p : {strata of X} → Z. If

  • 1. p(S) = p(codim S) for some p : N → Z
  • 2. p(k) = 0 for k ≤ 2
  • 3. p(k + 1) = p(k) or p(k) + 1.

then it is a Goresky–MacPherson (GM) perversity.

Examples

◮ the zero perversity 0(k) = 0 ◮ the top perversity t(k) = max{k − 2, 0} ◮ the lower middle perversity m(k) = max{⌊(k − 2)/2⌋, 0} ◮ the upper middle perversity n(k) = max{⌈(k − 2)/2⌉, 0}

GM perversities p and q are complementary if p + q = t.

slide-5
SLIDE 5

Intersection homology and Poincar´ e duality

Intersection homology

A perversity picks out a subcomplex of intersection chains in S∗X: ∆i

σ

− → X p-allowable ⇐ ⇒ σ−1S ⊂ (i − codim S + p(S)) -skeleton c ∈ SiX p-allowable ⇐ ⇒ all simplices in c are p-allowable Let I pS∗X = {c | c, ∂c are p-allowable} and I pH∗X its homology.

slide-6
SLIDE 6

Intersection homology and Poincar´ e duality

Intersection homology

A perversity picks out a subcomplex of intersection chains in S∗X: ∆i

σ

− → X p-allowable ⇐ ⇒ σ−1S ⊂ (i − codim S + p(S)) -skeleton c ∈ SiX p-allowable ⇐ ⇒ all simplices in c are p-allowable Let I pS∗X = {c | c, ∂c are p-allowable} and I pH∗X its homology.

Theorem (Goresky–MacPherson ’80)

X pseudomfld, p GM perversity = ⇒ I pH∗X topological invariant.

slide-7
SLIDE 7

Intersection homology and Poincar´ e duality

Intersection homology

A perversity picks out a subcomplex of intersection chains in S∗X: ∆i

σ

− → X p-allowable ⇐ ⇒ σ−1S ⊂ (i − codim S + p(S)) -skeleton c ∈ SiX p-allowable ⇐ ⇒ all simplices in c are p-allowable Let I pS∗X = {c | c, ∂c are p-allowable} and I pH∗X its homology.

Theorem (Goresky–MacPherson ’80)

X pseudomfld, p GM perversity = ⇒ I pH∗X topological invariant.

Theorem (Goresky–MacPherson ’80)

X compact, oriented n-dim pseudomfld, p, q complementary GM perversities = ⇒ ∃ intersection pairing I pHiX × I qHn−iX → Z which is non-degenerate over Q.

slide-8
SLIDE 8

Part II Filtered homology

slide-9
SLIDE 9

Filtered spaces and depth functions

Filtered spaces

A filtered space Xα is a topological space with a filtration ∅ = X−1 ⊂ X0 ⊂ X1 ⊂ X2 ⊂ · · · ⊂ X∞ = X. A filtered map f : Xα → Yβ is a map with f (Xk) ⊂ Yk ∀ k ∈ N.

slide-10
SLIDE 10

Filtered spaces and depth functions

Filtered spaces

A filtered space Xα is a topological space with a filtration ∅ = X−1 ⊂ X0 ⊂ X1 ⊂ X2 ⊂ · · · ⊂ X∞ = X. A filtered map f : Xα → Yβ is a map with f (Xk) ⊂ Yk ∀ k ∈ N.

Depth functions

The filtration on Xα is encoded in the depth function α: X → N∞ where α(x) = k ⇐ ⇒ x ∈ Xk − Xk−1 so Xk = α−1{0, . . . , k} and f : Xα → Yβ filtered ⇐ ⇒ α ≥ β ◦ f .

slide-11
SLIDE 11

Examples of filtered spaces

  • 1. A filtered space of depth ≤ 1 is a pair X0 ⊂ X1 = X; a filtered

map of such is a map of pairs.

slide-12
SLIDE 12

Examples of filtered spaces

  • 1. A filtered space of depth ≤ 1 is a pair X0 ⊂ X1 = X; a filtered

map of such is a map of pairs.

  • 2. Filtering a CW complex by its skeleta fully faithfully embeds

CW complexes and cellular maps into filtered spaces.

slide-13
SLIDE 13

Examples of filtered spaces

  • 1. A filtered space of depth ≤ 1 is a pair X0 ⊂ X1 = X; a filtered

map of such is a map of pairs.

  • 2. Filtering a CW complex by its skeleta fully faithfully embeds

CW complexes and cellular maps into filtered spaces.

  • 3. Let ∆n

δ be the standard simplex filtered by depth function

δ(t0, . . . , tn) = #{i | ti = 0}, e.g. 2 2 2 1 1 1 The face maps ∆i−1

δ+1 ֒

→ ∆i

δ are filtered.

slide-14
SLIDE 14

Filtered homology

For filtered Xα define SiXα = Z{∆i

δ → Xα}. Note

∂ : SiXα → Si−1Xα−1 where (α − 1)(x) = max{α(x) − 1, 0}.

slide-15
SLIDE 15

Filtered homology

For filtered Xα define SiXα = Z{∆i

δ → Xα}. Note

∂ : SiXα → Si−1Xα−1 where (α − 1)(x) = max{α(x) − 1, 0}.

Definition

The filtered i-chains on Xα are FSiXα = {c ∈ SiXα | ∂c ∈ Si−1Xα}. The filtered homology FH∗Xα is the homology of FS∗Xα.

slide-16
SLIDE 16

Properties of filtered homology

Functoriality

Filtered f : Xα → Yβ induces a chain map FS∗Xα → FS∗Yβ and f∗ : FH∗Xα → FH∗Yβ.

slide-17
SLIDE 17

Properties of filtered homology

Functoriality

Filtered f : Xα → Yβ induces a chain map FS∗Xα → FS∗Yβ and f∗ : FH∗Xα → FH∗Yβ.

Filtered homotopy invariance

If f and g are filtered homotopic then f∗ = g∗ : FH∗Xα → FH∗Yβ.

slide-18
SLIDE 18

Properties of filtered homology

Functoriality

Filtered f : Xα → Yβ induces a chain map FS∗Xα → FS∗Yβ and f∗ : FH∗Xα → FH∗Yβ.

Filtered homotopy invariance

If f and g are filtered homotopic then f∗ = g∗ : FH∗Xα → FH∗Yβ.

Relative long exact sequence

For filtered f : Xα → Yβ where the underlying map is an inclusion we define FHi(Yβ, Xα) = Hi

  • FS∗Yβ
  • FS∗Xα
  • . There is a LES

· · · → FH∗Xα → FH∗Yβ → FH∗(Yβ, Xα) → FH∗−1Xα → · · ·

slide-19
SLIDE 19

Properties of filtered homology

Functoriality

Filtered f : Xα → Yβ induces a chain map FS∗Xα → FS∗Yβ and f∗ : FH∗Xα → FH∗Yβ.

Filtered homotopy invariance

If f and g are filtered homotopic then f∗ = g∗ : FH∗Xα → FH∗Yβ.

Relative long exact sequence

For filtered f : Xα → Yβ where the underlying map is an inclusion we define FHi(Yβ, Xα) = Hi

  • FS∗Yβ
  • FS∗Xα
  • . There is a LES

· · · → FH∗Xα → FH∗Yβ → FH∗(Yβ, Xα) → FH∗−1Xα → · · ·

Excision

For Zα ⊂ Yα ⊂ Xα with Z ⊂ Y o there are isomorphisms FH∗(Xα − Zα, Yα − Zα) ∼ = FH∗(Xα, Yα).

slide-20
SLIDE 20

Simple examples of filtered homology

Cones

For [x, t] ∈ CX, the cone on X, and d > 1 have β[x, t] =

  • α(x)

t > 0 d t = 0 = ⇒ FHiCXβ ∼ =

  • FHiXα

i < d − 1 i ≥ d − 1. When d ≤ 1 obtain homology of a point.

slide-21
SLIDE 21

Simple examples of filtered homology

Cones

For [x, t] ∈ CX, the cone on X, and d > 1 have β[x, t] =

  • α(x)

t > 0 d t = 0 = ⇒ FHiCXβ ∼ =

  • FHiXα

i < d − 1 i ≥ d − 1. When d ≤ 1 obtain homology of a point.

Suspended torus

Let Xα = ΣT 2 where α(x) = 2 at suspension points and 0

  • elsewhere. Then

FHiXα =            Z i = 0 i = 1 Z2 i = 2 Z i = 3.

slide-22
SLIDE 22

Simple examples of filtered homology

Cones

For [x, t] ∈ CX, the cone on X, and d > 1 have β[x, t] =

  • α(x)

t > 0 d t = 0 = ⇒ FHiCXβ ∼ =

  • FHiXα

i < d − 1 i ≥ d − 1. When d ≤ 1 obtain homology of a point.

Suspended torus

Let Xα = ΣT 2 where α(x) = 3 at suspension points and 0

  • elsewhere. Then

FHiXα =            Z i = 0 Z2 i = 1 i = 2 Z i = 3.

slide-23
SLIDE 23

Perversities and filtrations

Given stratified X and perversity p define a depth function ˆ p(x) = codim S − p(S) for x ∈ S. The identity Xˆ

p → Xˆ q is filtered ⇐

⇒ p ≤ q.

slide-24
SLIDE 24

Perversities and filtrations

Given stratified X and perversity p define a depth function ˆ p(x) = codim S − p(S) for x ∈ S. The identity Xˆ

p → Xˆ q is filtered ⇐

⇒ p ≤ q. Setting X k =

codim S≤k S gives ◮ Xˆ 0 =

  • X 0 ⊂ X 1 ⊂ · · · ⊂ X k ⊂ · · · ⊂ X
slide-25
SLIDE 25

Perversities and filtrations

Given stratified X and perversity p define a depth function ˆ p(x) = codim S − p(S) for x ∈ S. The identity Xˆ

p → Xˆ q is filtered ⇐

⇒ p ≤ q. Setting X k =

codim S≤k S gives ◮ Xˆ 0 =

  • X 0 ⊂ X 1 ⊂ · · · ⊂ X k ⊂ · · · ⊂ X
  • ◮ Xˆ

t =

  • X 0 ⊂ X 1 ⊂ X
slide-26
SLIDE 26

Perversities and filtrations

Given stratified X and perversity p define a depth function ˆ p(x) = codim S − p(S) for x ∈ S. The identity Xˆ

p → Xˆ q is filtered ⇐

⇒ p ≤ q. Setting X k =

codim S≤k S gives ◮ Xˆ 0 =

  • X 0 ⊂ X 1 ⊂ · · · ⊂ X k ⊂ · · · ⊂ X
  • ◮ Xˆ

t =

  • X 0 ⊂ X 1 ⊂ X
  • ◮ X ˆ

m =

  • X 0 ⊂ X 1 ⊂ X 2 ⊂ X 4 ⊂ · · · ⊂ X
slide-27
SLIDE 27

Perversities and filtrations

Given stratified X and perversity p define a depth function ˆ p(x) = codim S − p(S) for x ∈ S. The identity Xˆ

p → Xˆ q is filtered ⇐

⇒ p ≤ q. Setting X k =

codim S≤k S gives ◮ Xˆ 0 =

  • X 0 ⊂ X 1 ⊂ · · · ⊂ X k ⊂ · · · ⊂ X
  • ◮ Xˆ

t =

  • X 0 ⊂ X 1 ⊂ X
  • ◮ X ˆ

m =

  • X 0 ⊂ X 1 ⊂ X 2 ⊂ X 4 ⊂ · · · ⊂ X
  • ◮ Xˆ

n =

  • X 0 ⊂ X 1 ⊂ X 3 ⊂ X 5 ⊂ · · · ⊂ X
slide-28
SLIDE 28

Perversities and filtrations

Given stratified X and perversity p define a depth function ˆ p(x) = codim S − p(S) for x ∈ S. The identity Xˆ

p → Xˆ q is filtered ⇐

⇒ p ≤ q. Setting X k =

codim S≤k S gives ◮ Xˆ 0 =

  • X 0 ⊂ X 1 ⊂ · · · ⊂ X k ⊂ · · · ⊂ X
  • ◮ Xˆ

t =

  • X 0 ⊂ X 1 ⊂ X
  • ◮ X ˆ

m =

  • X 0 ⊂ X 1 ⊂ X 2 ⊂ X 4 ⊂ · · · ⊂ X
  • ◮ Xˆ

n =

  • X 0 ⊂ X 1 ⊂ X 3 ⊂ X 5 ⊂ · · · ⊂ X
  • ◮ p is a Goresky–Macpherson perversity ⇐

⇒ Xˆ

p is filtration by

those X k with p(k) = p(k + 1)

slide-29
SLIDE 29

Perversities and filtrations

Given stratified X and perversity p define a depth function ˆ p(x) = codim S − p(S) for x ∈ S. The identity Xˆ

p → Xˆ q is filtered ⇐

⇒ p ≤ q. Setting X k =

codim S≤k S gives ◮ Xˆ 0 =

  • X 0 ⊂ X 1 ⊂ · · · ⊂ X k ⊂ · · · ⊂ X
  • ◮ Xˆ

t =

  • X 0 ⊂ X 1 ⊂ X
  • ◮ X ˆ

m =

  • X 0 ⊂ X 1 ⊂ X 2 ⊂ X 4 ⊂ · · · ⊂ X
  • ◮ Xˆ

n =

  • X 0 ⊂ X 1 ⊂ X 3 ⊂ X 5 ⊂ · · · ⊂ X
  • ◮ p is a Goresky–Macpherson perversity ⇐

⇒ Xˆ

p is filtration by

those X k with p(k) = p(k + 1)

◮ Complementary perversities p and q give ‘complementary’

filtrations: X k with k ≥ 2 appears in either Xˆ

p or Xˆ q.

slide-30
SLIDE 30

Intersection homology is filtered homology

An elementary calculation gives ∆i

δ σ

− → Xˆ

p filtered ⇐

⇒ σ−1S ⊂ (i − codim S + p(S)) -skeleton ⇐ ⇒ σ p-allowable

slide-31
SLIDE 31

Intersection homology is filtered homology

An elementary calculation gives ∆i

δ σ

− → Xˆ

p filtered ⇐

⇒ σ−1S ⊂ (i − codim S + p(S)) -skeleton ⇐ ⇒ σ p-allowable

Corollary

FS∗Xˆ

p = I pS∗X and FH∗Xˆ p = I pH∗X.

slide-32
SLIDE 32

Intersection homology is filtered homology

An elementary calculation gives ∆i

δ σ

− → Xˆ

p filtered ⇐

⇒ σ−1S ⊂ (i − codim S + p(S)) -skeleton ⇐ ⇒ σ p-allowable

Corollary

FS∗Xˆ

p = I pS∗X and FH∗Xˆ p = I pH∗X.

Remarks

◮ Functoriality of FH∗ =

⇒ known functoriality of IH∗

slide-33
SLIDE 33

Intersection homology is filtered homology

An elementary calculation gives ∆i

δ σ

− → Xˆ

p filtered ⇐

⇒ σ−1S ⊂ (i − codim S + p(S)) -skeleton ⇐ ⇒ σ p-allowable

Corollary

FS∗Xˆ

p = I pS∗X and FH∗Xˆ p = I pH∗X.

Remarks

◮ Functoriality of FH∗ =

⇒ known functoriality of IH∗

◮ Intersection homology is a filtered homotopy invariant

slide-34
SLIDE 34

Intersection homology is filtered homology

An elementary calculation gives ∆i

δ σ

− → Xˆ

p filtered ⇐

⇒ σ−1S ⊂ (i − codim S + p(S)) -skeleton ⇐ ⇒ σ p-allowable

Corollary

FS∗Xˆ

p = I pS∗X and FH∗Xˆ p = I pH∗X.

Remarks

◮ Functoriality of FH∗ =

⇒ known functoriality of IH∗

◮ Intersection homology is a filtered homotopy invariant ◮ Filtered homology LES gives relative LES for IH∗, and

  • bstruction sequence for change of perversities.
slide-35
SLIDE 35

Part III Spectral sequence of a filtered space

slide-36
SLIDE 36

The spectral sequence

For filtered Xα the singular complex S∗X has natural filtration 0 ֒ → S∗Xα ֒ → S∗Xα−1 ֒ → · · · ֒ → S∗X

slide-37
SLIDE 37

The spectral sequence

For filtered Xα the singular complex S∗X has natural filtration 0 ֒ → S∗Xα ֒ → S∗Xα−1 ֒ → · · · ֒ → S∗X yielding a spectral sequence with E 0-page S0Xα S1Xα S2Xα

S0Xα−1 S0Xα S1Xα−1 S1Xα S0Xα−2 S0Xα−1

converging to Gr•H∗X where GriHjX = {[c] ∈ HjX | c ∈ SjXα−i} {[c] ∈ HjX | c ∈ SjXα−i+1}.

slide-38
SLIDE 38

The spectral sequence

The singular complex S∗X of filtered Xα has natural filtration 0 ֒ → S∗Xα ֒ → S∗Xα−1 ֒ → · · · ֒ → S∗X yielding a spectral sequence with E 1-page FS0Xα FS1Xα FS2Xα

FS0Xα−1 S0Xα+∂S1Xα FS1Xα−1 S1Xα+∂S2Xα FS0Xα−2 S0Xα−1+∂S1Xα−1

converging to Gr•H∗X where GriHjX = {[c] ∈ HjX | c ∈ SjXα−i} {[c] ∈ HjX | c ∈ SjXα−i−1}.

slide-39
SLIDE 39

The spectral sequence

The singular complex S∗X of filtered Xα has natural filtration 0 ֒ → S∗Xα ֒ → S∗Xα−1 ֒ → · · · ֒ → S∗X yielding a spectral sequence with E 2-page FH0Xα FH1Xα FH2Xα FH0(Xα−1, Xα) FH1(Xα−1, Xα) FH0(Xα−2, Xα−1) converging to Gr•H∗X where GriHjX = {[c] ∈ HjX | c ∈ SjXα−i} {[c] ∈ HjX | c ∈ SjXα−i−1}.

slide-40
SLIDE 40

The spectral sequence

The singular complex S∗X of filtered Xα has natural filtration 0 ֒ → S∗Xα ֒ → S∗Xα−1 ֒ → · · · ֒ → S∗X yielding a spectral sequence with E ∞-page Gr0H0X Gr0H1X Gr0H2X Gr1H0X Gr1H1X Gr2H0X converging to Gr•H∗X where GriHjX = {[c] ∈ HjX | c ∈ SjXα−i} {[c] ∈ HjX | c ∈ SjXα−i−1}.

slide-41
SLIDE 41

Examples of the spectral sequence

Xα CW-complex with skeletal filtration

E 2-page is cellular chain complex: FH0Xα FH1(Xα−1, Xα)

slide-42
SLIDE 42

Examples of the spectral sequence

Xα CW-complex with skeletal filtration

E 3 = E ∞-page is cellular homology: Hcell

0 X

Hcell

1 X

slide-43
SLIDE 43

Examples of the spectral sequence

Xα = ΣT 2 with α(suspension points) = 3

E 2-page: Z Z2 Z Z4

slide-44
SLIDE 44

Examples of the spectral sequence

Xα = ΣT 2 with α(suspension points) = 3

E 3 = E ∞-page: Z Z Z2

slide-45
SLIDE 45

Part IV Cap products and Poincar´ e Duality?

slide-46
SLIDE 46

Alternative filtration for simplices

Let ∆n

δ′ denote the n-simplex with filtration

δ′(t0, . . . , tn) = min{i | tn−i = 0} and FS′

∗Xα the associated complex of filtered chains.

2 1

Proposition

There is a homotopy equivalence FS∗Xα ≃ FS′

∗Xα provided by

composition with id : ∆n

δ → ∆n δ′ and barycentric subdivision. So

filtered homology can be computed using either complex.

slide-47
SLIDE 47

Cap products

Filtered homology as a module

The inclusions of the ‘back’ faces of ∆n

δ′ are filtered. The usual

cap product formula restricts to SiX ⊗ S′

jXα → S′ j−iXα inducing

HiX ⊗ FHjXα → FHj−iXα, so that FH∗Xα is an H∗X-module.

slide-48
SLIDE 48

Cap products

Filtered homology as a module

The inclusions of the ‘back’ faces of ∆n

δ′ are filtered. The usual

cap product formula restricts to SiX ⊗ S′

jXα → S′ j−iXα inducing

HiX ⊗ FHjXα → FHj−iXα, so that FH∗Xα is an H∗X-module.

Generalised Poincar´ e duality?

A more refined approach should yield a cap product FHiXˆ

p ⊗ FHjXˆ 0 → FHj−iXˆ q−1

where p + q = t.

slide-49
SLIDE 49

Cap products

Filtered homology as a module

The inclusions of the ‘back’ faces of ∆n

δ′ are filtered. The usual

cap product formula restricts to SiX ⊗ S′

jXα → S′ j−iXα inducing

HiX ⊗ FHjXα → FHj−iXα, so that FH∗Xα is an H∗X-module.

Generalised Poincar´ e duality?

A more refined approach should yield a cap product FHiXˆ

p ⊗ FHjXˆ 0 → FHj−iXˆ q−1

where p + q = t. If we can improve this to FHiXˆ

p ⊗ FHjXˆ 0 → FHj−iXˆ q

then generalised Poincar´ e duality would follow.