The Rainbow Tur an Problem for Even Cycles Shagnik Das University - - PowerPoint PPT Presentation

the rainbow tur an problem for even cycles
SMART_READER_LITE
LIVE PREVIEW

The Rainbow Tur an Problem for Even Cycles Shagnik Das University - - PowerPoint PPT Presentation

The Rainbow Tur an Problem for Even Cycles Shagnik Das University of California, Los Angeles Aug 20, 2012 Joint work with Choongbum Lee and Benny Sudakov Historical Background Rainbow Tur an Problem Our Results Plan Historical


slide-1
SLIDE 1

The Rainbow Tur´ an Problem for Even Cycles

Shagnik Das

University of California, Los Angeles

Aug 20, 2012 Joint work with Choongbum Lee and Benny Sudakov

slide-2
SLIDE 2

Historical Background Rainbow Tur´ an Problem Our Results

Plan

1

Historical Background Tur´ an Problems Colouring Problems

2

Rainbow Tur´ an Problem Definition Motivation Known Results

3

Our Results Summary Warm Up Sketch of Proof

slide-3
SLIDE 3

Historical Background Rainbow Tur´ an Problem Our Results

Tur´ an’s Theorem

Mantel’s theorem: most fundamental in extremal graph theory Theorem (Mantel, 1907) If a graph G on n vertices has no triangle, then G has at most n2

4

edges.

slide-4
SLIDE 4

Historical Background Rainbow Tur´ an Problem Our Results

Tur´ an’s Theorem

Mantel’s theorem: most fundamental in extremal graph theory Theorem (Mantel, 1907) If a graph G on n vertices has no triangle, then G has at most n2

4

edges. Tur´ an generalised to cliques of any order Theorem (Tur´ an, 1941) If a graph G on n vertices has no clique of order r, then G has at most

  • 1 −

1 r−1 + o(1)

  • n2

2 edges.

slide-5
SLIDE 5

Historical Background Rainbow Tur´ an Problem Our Results

Tur´ an’s Theorem: Extended

Can define Tur´ an numbers of general graphs Definition (Tur´ an numbers) Given any graph H, we define the Tur´ an number ex(n, H) to be the maximum number of edges in an H-free graph on n vertices.

slide-6
SLIDE 6

Historical Background Rainbow Tur´ an Problem Our Results

Tur´ an’s Theorem: Extended

Can define Tur´ an numbers of general graphs Definition (Tur´ an numbers) Given any graph H, we define the Tur´ an number ex(n, H) to be the maximum number of edges in an H-free graph on n vertices. Erd˝

  • s and Stone found asymptotics for all non-bipartite graphs

Theorem (Erd˝

  • s-Stone, 1946)

For all graphs H, limn→∞ ex(n, H)/ n

2

  • = 1 − (χ(H) − 1)−1.
slide-7
SLIDE 7

Historical Background Rainbow Tur´ an Problem Our Results

Tur´ an’s Theorem: Open Problems

Tur´ an problem for bipartite graphs generally open

slide-8
SLIDE 8

Historical Background Rainbow Tur´ an Problem Our Results

Tur´ an’s Theorem: Open Problems

Tur´ an problem for bipartite graphs generally open Particularly interesting is the case of even cycles

slide-9
SLIDE 9

Historical Background Rainbow Tur´ an Problem Our Results

Tur´ an’s Theorem: Open Problems

Tur´ an problem for bipartite graphs generally open Particularly interesting is the case of even cycles Conjectured upper bound known Theorem (Bondy-Simonovits, 1974) For all k ≥ 2, ex(n, C2k) = O

  • n1+ 1

k

  • .
slide-10
SLIDE 10

Historical Background Rainbow Tur´ an Problem Our Results

Tur´ an’s Theorem: Open Problems

Tur´ an problem for bipartite graphs generally open Particularly interesting is the case of even cycles Conjectured upper bound known Theorem (Bondy-Simonovits, 1974) For all k ≥ 2, ex(n, C2k) = O

  • n1+ 1

k

  • .

Matching lower bound only known for k = 2, 3, 5

slide-11
SLIDE 11

Historical Background Rainbow Tur´ an Problem Our Results

Colouring Problems: Ramsey Theory

Another central result in extremal combinatorics Theorem (Ramsey, 1930) For any integers k, l ≥ 1, there exists R(k, l) such that any red-blue colouring of KR(k,l) contains either a red Kk or a blue Kl.

slide-12
SLIDE 12

Historical Background Rainbow Tur´ an Problem Our Results

Colouring Problems: Ramsey Theory

Another central result in extremal combinatorics Theorem (Ramsey, 1930) For any integers k, l ≥ 1, there exists R(k, l) such that any red-blue colouring of KR(k,l) contains either a red Kk or a blue Kl. Determining the Ramsey numbers R(k, l) a widely open problem

slide-13
SLIDE 13

Historical Background Rainbow Tur´ an Problem Our Results

Colouring Problems: Ramsey Theory

Another central result in extremal combinatorics Theorem (Ramsey, 1930) For any integers k, l ≥ 1, there exists R(k, l) such that any red-blue colouring of KR(k,l) contains either a red Kk or a blue Kl. Determining the Ramsey numbers R(k, l) a widely open problem

Introduction of the probabilistic method

slide-14
SLIDE 14

Historical Background Rainbow Tur´ an Problem Our Results

Colouring Problems: Extensions

Theorem (Erd˝

  • s-Rado, 1950)

For every t, there is an n such that every edge-colouring of Kn has a copy of Kt with one of the following canonical colourings: 2 1 3 4

constant

2 1 3 4

rainbow

2 1 3 4

minimum

2 1 3 4

maximum

slide-15
SLIDE 15

Historical Background Rainbow Tur´ an Problem Our Results

Colouring Problems: Extensions

Theorem (Erd˝

  • s-Rado, 1950)

For every t, there is an n such that every edge-colouring of Kn has a copy of Kt with one of the following canonical colourings: 2 1 3 4

constant

2 1 3 4

rainbow

2 1 3 4

minimum

2 1 3 4

maximum

In particular, for every t there is an n such that every proper edge-colouring of Kn has a rainbow Kt.

slide-16
SLIDE 16

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem

First introduced by Keevash, Mubayi, Sudakov, Verstra¨ ete

slide-17
SLIDE 17

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem

First introduced by Keevash, Mubayi, Sudakov, Verstra¨ ete Definition (Rainbow Tur´ an Numbers) Given a graph H, we define the rainbow Tur´ an number ex∗(n, H) to be the maximum number of edges in a properly edge-coloured n-vertex graph with no rainbow copy of H.

slide-18
SLIDE 18

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem

First introduced by Keevash, Mubayi, Sudakov, Verstra¨ ete Definition (Rainbow Tur´ an Numbers) Given a graph H, we define the rainbow Tur´ an number ex∗(n, H) to be the maximum number of edges in a properly edge-coloured n-vertex graph with no rainbow copy of H. Trivial bound: ex(n, H) ≤ ex∗(n, H)

slide-19
SLIDE 19

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem

First introduced by Keevash, Mubayi, Sudakov, Verstra¨ ete Definition (Rainbow Tur´ an Numbers) Given a graph H, we define the rainbow Tur´ an number ex∗(n, H) to be the maximum number of edges in a properly edge-coloured n-vertex graph with no rainbow copy of H. Trivial bound: ex(n, H) ≤ ex∗(n, H) Reduction to regular Tur´ an problem

⇒ ex∗(n, H) ≤ ex(n, H) + o(n2)

slide-20
SLIDE 20

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Motivation

Definition (Bk-sets) A subset A of an abelian group G is a Bk-set if every g ∈ G has at most one representation of the form g = a1 + a2 + . . . + ak, ai ∈ A.

slide-21
SLIDE 21

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Motivation

Definition (Bk-sets) A subset A of an abelian group G is a Bk-set if every g ∈ G has at most one representation of the form g = a1 + a2 + . . . + ak, ai ∈ A. Definition (B∗

k-sets)

A subset A of an abelian group G is a B∗

k-set if there are no two

disjoint k-sets {x1, x2, . . . , xk}, {y1, y2, . . . , yk} ⊂ A with the same sum.

slide-22
SLIDE 22

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Motivation

Definition (Bk-sets) A subset A of an abelian group G is a Bk-set if every g ∈ G has at most one representation of the form g = a1 + a2 + . . . + ak, ai ∈ A. Definition (B∗

k-sets)

A subset A of an abelian group G is a B∗

k-set if there are no two

disjoint k-sets {x1, x2, . . . , xk}, {y1, y2, . . . , yk} ⊂ A with the same sum. Example (B∗

k-sets need not be Bk-sets)

Let k = 3, G = Z/6Z, and A = {0, 1, 2, 3, 4, 5}. A is not a B3-set: 0 + 1 + 4 = 0 + 2 + 3 = 0 + 0 + 5. A is a B∗

3-set: the sum of all six elements is odd.

slide-23
SLIDE 23

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Connection to Rainbow Cycles

Bipartite Cayley graph construction: B∗

k-sets → rainbow-C2k-free bipartite graphs

slide-24
SLIDE 24

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Connection to Rainbow Cycles

Bipartite Cayley graph construction: B∗

k-sets → rainbow-C2k-free bipartite graphs

Construct bipartite graph on X ∪ Y , where X, Y = G Edge (x, y) iff y − x = a ∈ A ⊂ G, colour a

slide-25
SLIDE 25

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Connection to Rainbow Cycles

Bipartite Cayley graph construction: B∗

k-sets → rainbow-C2k-free bipartite graphs

Construct bipartite graph on X ∪ Y , where X, Y = G Edge (x, y) iff y − x = a ∈ A ⊂ G, colour a y3 y2 y1 x1 x2 x3 x4 x5 y5 y4

Example with G = Z/5Z, A = {1, 3}

slide-26
SLIDE 26

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Connection to Rainbow Cycles

Bipartite Cayley graph construction: B∗

k-sets → rainbow-C2k-free bipartite graphs

Construct bipartite graph on X ∪ Y , where X, Y = G Edge (x, y) iff y − x = a ∈ A ⊂ G, colour a y3 y2 y1 x1 x2 x3 x4 x5 y5 y4

Example with G = Z/5Z, A = {1, 3}

slide-27
SLIDE 27

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Connection to Rainbow Cycles

Bipartite Cayley graph construction: B∗

k-sets → rainbow-C2k-free bipartite graphs

Construct bipartite graph on X ∪ Y , where X, Y = G Edge (x, y) iff y − x = a ∈ A ⊂ G, colour a y3 y2 y1 x1 x2 x3 x4 x5 y5 y4

Example with G = Z/5Z, A = {1, 3}

slide-28
SLIDE 28

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Connection to Rainbow Cycles

Bipartite Cayley graph construction: B∗

k-sets → rainbow-C2k-free bipartite graphs

Construct bipartite graph on X ∪ Y , where X, Y = G Edge (x, y) iff y − x = a ∈ A ⊂ G, colour a y3 y2 y1 x1 x2 x3 x4 x5 y5 y4

Example with G = Z/5Z, A = {1, 3}

Rainbow C2k x1y1x2y2 . . . xkyk ↔ B = {y1 − x1, . . . , yk − xk} and C = {y1 − x2, y2 − x3, . . . , yk − x1}

slide-29
SLIDE 29

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Known Results

A ⊂ G has |A|+k−1

k

  • different k-sums

⇒ if |G| = n, A a Bk-set, then |A| = O

  • n

1 k

slide-30
SLIDE 30

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Known Results

A ⊂ G has |A|+k−1

k

  • different k-sums

⇒ if |G| = n, A a Bk-set, then |A| = O

  • n

1 k

  • Bose and Chowla: Bk-sets of size Ω
  • n

1 k

  • in G = Z/nZ

⇒ Z/nZ has B∗

k-sets of size Ω

  • n

1 k

slide-31
SLIDE 31

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Known Results

A ⊂ G has |A|+k−1

k

  • different k-sums

⇒ if |G| = n, A a Bk-set, then |A| = O

  • n

1 k

  • Bose and Chowla: Bk-sets of size Ω
  • n

1 k

  • in G = Z/nZ

⇒ Z/nZ has B∗

k-sets of size Ω

  • n

1 k

  • ⇒ ex∗(n, C2k) = Ω
  • n1+ 1

k

slide-32
SLIDE 32

Historical Background Rainbow Tur´ an Problem Our Results

B∗

k-sets: Known Results

A ⊂ G has |A|+k−1

k

  • different k-sums

⇒ if |G| = n, A a Bk-set, then |A| = O

  • n

1 k

  • Bose and Chowla: Bk-sets of size Ω
  • n

1 k

  • in G = Z/nZ

⇒ Z/nZ has B∗

k-sets of size Ω

  • n

1 k

  • ⇒ ex∗(n, C2k) = Ω
  • n1+ 1

k

  • Ruzsa: O
  • n

1 k

  • analytic upper bound for B∗

k-sets

Bound on ex∗(n, C2k) would give a combinatorial proof

slide-33
SLIDE 33

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Bipartite Graphs

If H has maximum degree s in one part, then ex∗(n, H) = O

  • n2− 1

s

slide-34
SLIDE 34

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Bipartite Graphs

If H has maximum degree s in one part, then ex∗(n, H) = O

  • n2− 1

s

  • Gives correct order of magnitude for 4-cycle:

ex∗(n, C4) = O

  • n1+ 1

2

slide-35
SLIDE 35

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Bipartite Graphs

If H has maximum degree s in one part, then ex∗(n, H) = O

  • n2− 1

s

  • Gives correct order of magnitude for 4-cycle:

ex∗(n, C4) = O

  • n1+ 1

2

  • Analysing rainbow paths ⇒ ex∗(n, C6) = O
  • n1+ 1

3

slide-36
SLIDE 36

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Rainbow Acyclic Graphs

How large can an n-vertex graph without rainbow cycles be?

slide-37
SLIDE 37

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Rainbow Acyclic Graphs

How large can an n-vertex graph without rainbow cycles be?

Hypercube construction ⇒ Ω(n log n)

slide-38
SLIDE 38

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Rainbow Acyclic Graphs

How large can an n-vertex graph without rainbow cycles be?

Hypercube construction ⇒ Ω(n log n) 6-cycle bound ⇒ O

  • n1+ 1

3

slide-39
SLIDE 39

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Rainbow Acyclic Graphs

How large can an n-vertex graph without rainbow cycles be?

Hypercube construction ⇒ Ω(n log n) 6-cycle bound ⇒ O

  • n1+ 1

3

  • Two interesting questions:
slide-40
SLIDE 40

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Rainbow Acyclic Graphs

How large can an n-vertex graph without rainbow cycles be?

Hypercube construction ⇒ Ω(n log n) 6-cycle bound ⇒ O

  • n1+ 1

3

  • Two interesting questions:
  • 1. Determine size of rainbow acyclic graphs
slide-41
SLIDE 41

Historical Background Rainbow Tur´ an Problem Our Results

Rainbow Tur´ an Problem: Rainbow Acyclic Graphs

How large can an n-vertex graph without rainbow cycles be?

Hypercube construction ⇒ Ω(n log n) 6-cycle bound ⇒ O

  • n1+ 1

3

  • Two interesting questions:
  • 1. Determine size of rainbow acyclic graphs
  • 2. Determine asymptotics of ex∗ (n, C2k)
slide-42
SLIDE 42

Historical Background Rainbow Tur´ an Problem Our Results

Our Results

slide-43
SLIDE 43

Historical Background Rainbow Tur´ an Problem Our Results

Our Results

Theorem (D.-Lee-Sudakov, 2012+) A rainbow acyclic n-vertex graph can have at most n exp

  • (log n)

1 2 +η

edges for any η > 0.

slide-44
SLIDE 44

Historical Background Rainbow Tur´ an Problem Our Results

Our Results

Theorem (D.-Lee-Sudakov, 2012+) A rainbow acyclic n-vertex graph can have at most n exp

  • (log n)

1 2 +η

edges for any η > 0. Theorem (D.-Lee-Sudakov, 2012+) For every integer k ≥ 2, ex∗ (n, C2k) = O

  • n1+ (1+εk ) ln k

k

  • ,

where εk → 0 as k → ∞.

slide-45
SLIDE 45

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: Warm Up

The black-and-white version of our theorem

slide-46
SLIDE 46

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: Warm Up

The black-and-white version of our theorem Proposition If G on n vertices has more than n1+ε edges, then it has a cycle of length at most 2

ε.

slide-47
SLIDE 47

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: Warm Up

The black-and-white version of our theorem Proposition If G on n vertices has more than n1+ε edges, then it has a cycle of length at most 2

ε.

Proof by induction on n Base case: For n ≤ 2

ε, result follows from existence of cycle

slide-48
SLIDE 48

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: Warm Up

The black-and-white version of our theorem Proposition If G on n vertices has more than n1+ε edges, then it has a cycle of length at most 2

ε.

Proof by induction on n Base case: For n ≤ 2

ε, result follows from existence of cycle

Induction step: May assume minimum degree δ > nε

slide-49
SLIDE 49

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: Warm Up II

v0

Choose an arbitrary vertex v0 ∈ G

slide-50
SLIDE 50

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: Warm Up II

v0 N1 > nε

Expand its neighbourhood N1

slide-51
SLIDE 51

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: Warm Up II

v0 N1 > nε N2 > n2ε

Expand the second neighbourhood N2

slide-52
SLIDE 52

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: Warm Up II

v0 N1 > nε N2 > n2ε ... N 1

ε

> n

If no short cycle, then we will eventually exceed n vertices - contradiction

slide-53
SLIDE 53

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview

Theorem For every ε > 0, there are constants C(ε) and L(ε) such that any n-vertex graph with at least C(ε)n1+ε edges has a rainbow cycle of length at most L(ε).

slide-54
SLIDE 54

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview

Theorem For every ε > 0, there are constants C(ε) and L(ε) such that any n-vertex graph with at least C(ε)n1+ε edges has a rainbow cycle of length at most L(ε). Proof idea: As before, may assume minimum degree δ(G) > C(ε)nε

slide-55
SLIDE 55

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview

Theorem For every ε > 0, there are constants C(ε) and L(ε) such that any n-vertex graph with at least C(ε)n1+ε edges has a rainbow cycle of length at most L(ε). Proof idea: As before, may assume minimum degree δ(G) > C(ε)nε Will grow a subtree T out of an arbitrary vertex v0

slide-56
SLIDE 56

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

v0

Choose an arbitrary vertex v0 ∈ G

slide-57
SLIDE 57

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

v0 L1 nα1

Induction implies the small levels must expand

slide-58
SLIDE 58

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

v0 L1 nα1 L2 nα2

Induction implies the small levels must expand

slide-59
SLIDE 59

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

v0 L1 nα1 L2 nα2 L3 nα3

Induction implies the small levels must expand

slide-60
SLIDE 60

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

v0 L1 nα1 L2 nα2 L3 nα3

We ensure all vertices have a rainbow path back to v0

slide-61
SLIDE 61

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

v0 L1 nα1 L2 nα2 L3 nα3

Let Xi,c ⊂ Li be those with colour c in their paths.

slide-62
SLIDE 62

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

Li nαi

Growing next level: consider unused neighbours of Li

slide-63
SLIDE 63

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

Li nαi Li+1 nαi+1

Restrict Li+1 to those with between d and 2d neighbours in Li

slide-64
SLIDE 64

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

Li nαi Li+1 nαi+1 ≈ nε ≈ d

Must have d ≈ nε+αi−αi+1

slide-65
SLIDE 65

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

Li nαi Li+1 nαi+1

If no short rainbow cycle, then neighbours in Li must share a colour

slide-66
SLIDE 66

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

Li nαi Li+1 nαi+1

If no short rainbow cycle, then neighbours in Li must share a colour

slide-67
SLIDE 67

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

Li nαi Li+1 nαi+1

Let Wc be those in Li+1 with many neighbours in Xi,c

slide-68
SLIDE 68

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

Restrict to G[Xi,c, Wc]

slide-69
SLIDE 69

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

≤ nε ≈ d

If d ≈ nε, short rainbow cycle by induction

slide-70
SLIDE 70

Historical Background Rainbow Tur´ an Problem Our Results

Our Proof: A Sketchy Overview II

≤ nε ≈ d

Thus d ≪ nε ⇒ nαi+1 ≫ nαi - desired expansion

slide-71
SLIDE 71

Historical Background Rainbow Tur´ an Problem Our Results

Open Problems

Question (Rainbow Acyclicity) We know the largest properly edge-coloured n-vertex graph has at least Ω(n log n) edges, and at most n exp

  • (log n)

1 2 +o(1)

edges. Can the gap be narrowed? Question (Rainbow Tur´ an Number for Even Cycles) We have the bounds Ω

  • n1+ 1

k

  • ≤ ex∗(n, C2k) ≤ O
  • n1+ (1+o(1)) ln k

k

  • .

Can this gap be narrowed?

slide-72
SLIDE 72