Minimally non-balanced diamond-free graphs Anna Galluccio Istituto - - PowerPoint PPT Presentation

minimally non balanced diamond free graphs
SMART_READER_LITE
LIVE PREVIEW

Minimally non-balanced diamond-free graphs Anna Galluccio Istituto - - PowerPoint PPT Presentation

Minimally non-balanced diamond-free graphs Anna Galluccio Istituto Analisi Sistemi ed Informatica (IASI-CNR)- Italy JOINT WORK WITH: N. Apollonio, Istituto delle Applicazioni del Calcolo (IAC-CNR) Aussois Workshop January 2015 K onig


slide-1
SLIDE 1

Minimally non-balanced diamond-free graphs

Anna Galluccio

Istituto Analisi Sistemi ed Informatica (IASI-CNR)- Italy

JOINT WORK WITH: N. Apollonio, Istituto delle Applicazioni del Calcolo (IAC-CNR)

Aussois Workshop January 2015

slide-2
SLIDE 2

  • nig property

Given a 0, 1-matrix A ν(A) =max # of pairwise non-intersecting rows (matching) τ(A) =min # of columns intersecting all rows (transversal) A satisfies the K¨

  • nig property if and only if ν(A) = τ(A).

C2k+1 =           1 1 · · · 1 1 · · · 1 ... . . . . . . . . . ... 1 1 1 · · · 1          

  • dd cycle matrix has

ν(C2k+1) = k < τ(C2k+1) = k + 1

slide-3
SLIDE 3

  • nig property

Given a 0, 1-matrix A ν(A) =max # of pairwise non-intersecting rows (matching) τ(A) =min # of columns intersecting all rows (transversal) A satisfies the K¨

  • nig property if and only if ν(A) = τ(A).

C2k+1 =           1 1 · · · 1 1 · · · 1 ... . . . . . . . . . ... 1 1 1 · · · 1          

  • dd cycle matrix has

ν(C2k+1) = k < τ(C2k+1) = k + 1

slide-4
SLIDE 4

  • nig property

Given a 0, 1-matrix A ν(A) =max # of pairwise non-intersecting rows (matching) τ(A) =min # of columns intersecting all rows (transversal) A satisfies the K¨

  • nig property if and only if ν(A) = τ(A).

C2k+1 =           1 1 · · · 1 1 · · · 1 ... . . . . . . . . . ... 1 1 1 · · · 1          

  • dd cycle matrix has

ν(C2k+1) = k < τ(C2k+1) = k + 1

slide-5
SLIDE 5

  • nig property

Given a 0, 1-matrix A ν(A) =max # of pairwise non-intersecting rows (matching) τ(A) =min # of columns intersecting all rows (transversal) A satisfies the K¨

  • nig property if and only if ν(A) = τ(A).

C2k+1 =           1 1 · · · 1 1 · · · 1 ... . . . . . . . . . ... 1 1 1 · · · 1          

  • dd cycle matrix has

ν(C2k+1) = k < τ(C2k+1) = k + 1

slide-6
SLIDE 6

Balanced matrices and balanced graphs

  • Def. A is balanced iff ν(A′) = τ(A′) for any A′ submatrix of A.

Theorem (Berge, Las Vergnas 1972) A balanced if and only if A ⊇ C2k+1, k ≥ 1, as a submatrix G graph, AG denote clique-matrix of G (matrix whose rows are the incidence vectors of maximal cliques of G)

  • Def. G is balanced if and only if AG is balanced
  • Def. G minimally non-balanced (MNB) if and only G is not

balanced but each its proper induced subgraphs is balanced.

slide-7
SLIDE 7

Balanced matrices and balanced graphs

  • Def. A is balanced iff ν(A′) = τ(A′) for any A′ submatrix of A.

Theorem (Berge, Las Vergnas 1972) A balanced if and only if A ⊇ C2k+1, k ≥ 1, as a submatrix G graph, AG denote clique-matrix of G (matrix whose rows are the incidence vectors of maximal cliques of G)

  • Def. G is balanced if and only if AG is balanced
  • Def. G minimally non-balanced (MNB) if and only G is not

balanced but each its proper induced subgraphs is balanced.

slide-8
SLIDE 8

Balanced matrices and balanced graphs

  • Def. A is balanced iff ν(A′) = τ(A′) for any A′ submatrix of A.

Theorem (Berge, Las Vergnas 1972) A balanced if and only if A ⊇ C2k+1, k ≥ 1, as a submatrix G graph, AG denote clique-matrix of G (matrix whose rows are the incidence vectors of maximal cliques of G)

  • Def. G is balanced if and only if AG is balanced
  • Def. G minimally non-balanced (MNB) if and only G is not

balanced but each its proper induced subgraphs is balanced.

slide-9
SLIDE 9

Balanced matrices and balanced graphs

  • Def. A is balanced iff ν(A′) = τ(A′) for any A′ submatrix of A.

Theorem (Berge, Las Vergnas 1972) A balanced if and only if A ⊇ C2k+1, k ≥ 1, as a submatrix G graph, AG denote clique-matrix of G (matrix whose rows are the incidence vectors of maximal cliques of G)

  • Def. G is balanced if and only if AG is balanced
  • Def. G minimally non-balanced (MNB) if and only G is not

balanced but each its proper induced subgraphs is balanced.

slide-10
SLIDE 10

Minimally non-balanced graphs

It is natural to ask for characterization of minimally non-balanced graphs but the problem does not appear to be easy. first attempt to attack the general problem is due to Bonomo, Duran, Lin and Swarzficter (2002) with extended

  • dd sun

minimal characterizations exist for special classes of graphs: chordal, line graphs, paw free, complement of line graphs [Bonomo,Chudnovsky,Duran, Safe,Wagler,..] decomposition theorem of Conforti, Cornuejols and Rao (1999) yields a polynomial-time recognition algorithm more recently, a connection with the study of minimally clique-imperfect graphs.

slide-11
SLIDE 11

Minimally non-balanced graphs

It is natural to ask for characterization of minimally non-balanced graphs but the problem does not appear to be easy. first attempt to attack the general problem is due to Bonomo, Duran, Lin and Swarzficter (2002) with extended

  • dd sun

minimal characterizations exist for special classes of graphs: chordal, line graphs, paw free, complement of line graphs [Bonomo,Chudnovsky,Duran, Safe,Wagler,..] decomposition theorem of Conforti, Cornuejols and Rao (1999) yields a polynomial-time recognition algorithm more recently, a connection with the study of minimally clique-imperfect graphs.

slide-12
SLIDE 12

Minimally non-balanced graphs

It is natural to ask for characterization of minimally non-balanced graphs but the problem does not appear to be easy. first attempt to attack the general problem is due to Bonomo, Duran, Lin and Swarzficter (2002) with extended

  • dd sun

minimal characterizations exist for special classes of graphs: chordal, line graphs, paw free, complement of line graphs [Bonomo,Chudnovsky,Duran, Safe,Wagler,..] decomposition theorem of Conforti, Cornuejols and Rao (1999) yields a polynomial-time recognition algorithm more recently, a connection with the study of minimally clique-imperfect graphs.

slide-13
SLIDE 13

Clique-perfection

τc(G) = min # of vertices that meets all the maximal cliques of G. αc(G) = max # of pairwise vertex-disjoint cliques of G. A graph G is clique-perfect if τc(G) = αc(G) for each induced subgraph G′ of G. These graph-invariants were already studied by Tuza and Lehel during ’80s, but recently Bonomo, Chudnovsky and Duran (2007) explicitly asked for complexity of recognizing diamond-free clique-perfect graphs, characterization of minimally clique-imperfect diamond-free graphs.

slide-14
SLIDE 14

Clique-perfection

τc(G) = min # of vertices that meets all the maximal cliques of G. αc(G) = max # of pairwise vertex-disjoint cliques of G. A graph G is clique-perfect if τc(G) = αc(G) for each induced subgraph G′ of G. These graph-invariants were already studied by Tuza and Lehel during ’80s, but recently Bonomo, Chudnovsky and Duran (2007) explicitly asked for complexity of recognizing diamond-free clique-perfect graphs, characterization of minimally clique-imperfect diamond-free graphs.

slide-15
SLIDE 15

Clique-perfection

τc(G) = min # of vertices that meets all the maximal cliques of G. αc(G) = max # of pairwise vertex-disjoint cliques of G. A graph G is clique-perfect if τc(G) = αc(G) for each induced subgraph G′ of G. These graph-invariants were already studied by Tuza and Lehel during ’80s, but recently Bonomo, Chudnovsky and Duran (2007) explicitly asked for complexity of recognizing diamond-free clique-perfect graphs, characterization of minimally clique-imperfect diamond-free graphs.

slide-16
SLIDE 16

Clique-perfection

τc(G) = min # of vertices that meets all the maximal cliques of G. αc(G) = max # of pairwise vertex-disjoint cliques of G. A graph G is clique-perfect if τc(G) = αc(G) for each induced subgraph G′ of G. These graph-invariants were already studied by Tuza and Lehel during ’80s, but recently Bonomo, Chudnovsky and Duran (2007) explicitly asked for complexity of recognizing diamond-free clique-perfect graphs, characterization of minimally clique-imperfect diamond-free graphs.

slide-17
SLIDE 17

Clique-perfection

τc(G) = min # of vertices that meets all the maximal cliques of G. αc(G) = max # of pairwise vertex-disjoint cliques of G. A graph G is clique-perfect if τc(G) = αc(G) for each induced subgraph G′ of G. These graph-invariants were already studied by Tuza and Lehel during ’80s, but recently Bonomo, Chudnovsky and Duran (2007) explicitly asked for complexity of recognizing diamond-free clique-perfect graphs, characterization of minimally clique-imperfect diamond-free graphs.

slide-18
SLIDE 18

Clique-perfection

τc(G) = min # of vertices that meets all the maximal cliques of G. αc(G) = max # of pairwise vertex-disjoint cliques of G. A graph G is clique-perfect if τc(G) = αc(G) for each induced subgraph G′ of G. These graph-invariants were already studied by Tuza and Lehel during ’80s, but recently Bonomo, Chudnovsky and Duran (2007) explicitly asked for complexity of recognizing diamond-free clique-perfect graphs, characterization of minimally clique-imperfect diamond-free graphs.

slide-19
SLIDE 19

Clique-perfection versus balancedness

K3 K4 K5 K7 K0 K1 K2 K3 K4 K5 K6 K7 K8 K9 K1 K2 K6

ν(AG) = ν(C9) = 4 < τ(AG) = 5 = τ(C9) ν(AG) = τ(AG) = 3

Theorem (G. , Apollonio 2013, Safe P.h.D.Thesis 2009) G diamond-free. Then G clique-perfect ⇔ G balanced

slide-20
SLIDE 20

Clique-perfection versus balancedness

K3 K4 K5 K7 K0 K1 K2 K3 K4 K5 K6 K7 K8 K9 K1 K2 K6

ν(AG) = ν(C9) = 4 < τ(AG) = 5 = τ(C9) ν(AG) = τ(AG) = 3

Theorem (G. , Apollonio 2013, Safe P.h.D.Thesis 2009) G diamond-free. Then G clique-perfect ⇔ G balanced

slide-21
SLIDE 21

Non-balanced diamond-free graphs

G diamond-free graph if and only if AG is linear (i.e., AG does not contain 1 1

1 1

  • as a submatrix).

g(A)= min order of an odd cycle submatrix of A B↑ (up-matrix) obtained from B by removing its dominated rows Lemma Let G be diamond-free non-balanced. If g(AG) = n then AG contains an up-matrix Cn

F

  • such that

either F = ∅

  • r the rows of F have ≥ 3

nonzero entries that correspond to stable sets

  • n the odd cycle C

(multisun)

slide-22
SLIDE 22

Non-balanced diamond-free graphs

G diamond-free graph if and only if AG is linear (i.e., AG does not contain 1 1

1 1

  • as a submatrix).

g(A)= min order of an odd cycle submatrix of A B↑ (up-matrix) obtained from B by removing its dominated rows Lemma Let G be diamond-free non-balanced. If g(AG) = n then AG contains an up-matrix Cn

F

  • such that

either F = ∅

  • r the rows of F have ≥ 3

nonzero entries that correspond to stable sets

  • n the odd cycle C

(multisun)

slide-23
SLIDE 23

Non-balanced diamond-free graphs

G diamond-free graph if and only if AG is linear (i.e., AG does not contain 1 1

1 1

  • as a submatrix).

g(A)= min order of an odd cycle submatrix of A B↑ (up-matrix) obtained from B by removing its dominated rows Lemma Let G be diamond-free non-balanced. If g(AG) = n then AG contains an up-matrix Cn

F

  • such that

either F = ∅

  • r the rows of F have ≥ 3

nonzero entries that correspond to stable sets

  • n the odd cycle C

(multisun)

slide-24
SLIDE 24

Non-balanced diamond-free graphs

G diamond-free graph if and only if AG is linear (i.e., AG does not contain 1 1

1 1

  • as a submatrix).

g(A)= min order of an odd cycle submatrix of A B↑ (up-matrix) obtained from B by removing its dominated rows Lemma Let G be diamond-free non-balanced. If g(AG) = n then AG contains an up-matrix Cn

F

  • such that

either F = ∅

  • r the rows of F have ≥ 3

nonzero entries that correspond to stable sets

  • n the odd cycle C

(multisun)

slide-25
SLIDE 25

Theorem G diamond-free with n vertices, n > 3. Then g(AG) = n ⇐ ⇒ either AG = Cn

  • r AG =

Cn

F

  • and for each

F ′ ⊆ F, Cn

F ′

  • does not contain

Cg, g < n, as up-matrix Sub-multisun is obtained from a multisun by deleting the edges of some inscribed clique Definition G multisun is Hereditarily-Odd-Hole-free (HOH-free) iff every sub-multisun of G is odd-hole-free

slide-26
SLIDE 26

Theorem G diamond-free with n vertices, n > 3. Then g(AG) = n ⇐ ⇒ either AG = Cn

  • r AG =

Cn

F

  • and for each

F ′ ⊆ F, Cn

F ′

  • does not contain

Cg, g < n, as up-matrix Sub-multisun is obtained from a multisun by deleting the edges of some inscribed clique Definition G multisun is Hereditarily-Odd-Hole-free (HOH-free) iff every sub-multisun of G is odd-hole-free

slide-27
SLIDE 27

Theorem G diamond-free with n vertices, n > 3. Then g(AG) = n ⇐ ⇒ either AG = Cn

  • r AG =

Cn

F

  • and for each

F ′ ⊆ F, Cn

F ′

  • does not contain

Cg, g < n, as up-matrix Sub-multisun is obtained from a multisun by deleting the edges of some inscribed clique Definition G multisun is Hereditarily-Odd-Hole-free (HOH-free) iff every sub-multisun of G is odd-hole-free

slide-28
SLIDE 28

Theorem G diamond-free with n vertices, n > 3. Then g(AG) = n ⇐ ⇒ either AG = Cn

  • r AG =

Cn

F

  • and for each

F ′ ⊆ F, Cn

F ′

  • does not contain

Cg, g < n, as up-matrix Sub-multisun is obtained from a multisun by deleting the edges of some inscribed clique Definition G multisun is Hereditarily-Odd-Hole-free (HOH-free) iff every sub-multisun of G is odd-hole-free

slide-29
SLIDE 29

Minimally non-balanced graphs

G diamond-free MNB iff G is either an odd hole or a Hereditary Odd Hole Free (HOH-free) multisun

slide-30
SLIDE 30

Minimally non-balanced graphs

G diamond-free MNB iff G is either an odd hole or a Hereditary Odd Hole Free (HOH-free) multisun

slide-31
SLIDE 31

N-conditions

The inscribed cliques intersect in the same vertex and are otherwise vertex disjoint Odd between two cliques Even within vertices of the same cliques whose interior is not in another clique even Cliques are

  • dd
slide-32
SLIDE 32

Canonical labeling

Theorem If G is HOH-free multisun, the G satisfies the N-conditions. Canonical Labeling Let G be a multisun with rim C. Set Σ = {ǫ, σ} ∪ {x | X is an inscribed clique of G}. Let f : V(C) → Σ be defined as follows: f(v) =    σ if v belongs to more than one inscribed clique of G; x if v belongs to the inscribed clique X of G; ǫ if v belongs to no inscribed clique of G. Σ − {ǫ, σ} are called the proper letters.

slide-33
SLIDE 33

Canonical labeling

Theorem If G is HOH-free multisun, the G satisfies the N-conditions. Canonical Labeling Let G be a multisun with rim C. Set Σ = {ǫ, σ} ∪ {x | X is an inscribed clique of G}. Let f : V(C) → Σ be defined as follows: f(v) =    σ if v belongs to more than one inscribed clique of G; x if v belongs to the inscribed clique X of G; ǫ if v belongs to no inscribed clique of G. Σ − {ǫ, σ} are called the proper letters.

slide-34
SLIDE 34

aǫ2aǫ4aǫ2aǫ6aǫ8aǫ2aǫ20 σǫ4aǫ2aǫ6aǫbǫ2bǫ3aǫ2aǫbǫ3cǫ2cǫ3bǫ2bǫ2bǫaǫ2

a a a a a a a ǫ4 ǫ2 ǫ6 ǫ2 ǫ2 σ a a a b b a a b c c b b b a ǫ ǫ4 ǫ2 ǫ6 ǫ2 ǫ2 ǫ2 ǫ ǫ3 ǫ2 ǫ3 ǫ ǫ2 ǫ3 ǫ2 ǫ8 ǫ20 (b) (a)

slide-35
SLIDE 35
  • Being a multisun that satisfies N-conditions is preserved

under taking even subdivision/ contraction of the rim. Two words are pattern-equivalent if one can be transformed into the other by repeatedly applying one of the following

  • perations:

shifting the indices, reversing the order of reading, replace the interval ǫǫ by the empty word φ (even contraction), replace φ (empty) by k times ǫǫ with k ≥ 1 (even subdivision),

slide-36
SLIDE 36
  • Being a multisun that satisfies N-conditions is preserved

under taking even subdivision/ contraction of the rim. Two words are pattern-equivalent if one can be transformed into the other by repeatedly applying one of the following

  • perations:

shifting the indices, reversing the order of reading, replace the interval ǫǫ by the empty word φ (even contraction), replace φ (empty) by k times ǫǫ with k ≥ 1 (even subdivision),

slide-37
SLIDE 37

s-word

⋄ aǫ2aǫ4aǫ2aǫ6aǫ8aǫ2aǫ20 → a7 ⋄ σǫ4aǫ2aǫ6aǫbǫ2bǫ3aǫ2aǫbǫ3cǫ2cǫ3bǫ2bǫ2bǫaǫ2 → σa3ǫb2ǫa2ǫbǫc2ǫb3ǫa A class of pattern-equivalent words

w on the alphabet Σ is

called s-word and it has one of the following forms: i) if Σ = {ǫ, a}, then

w = [aλ], λ odd;

ii) if Σ = {σ, ǫ, a, b, . . .}, then

w = [σxλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is ] with

λ1, . . . , λs positive integers and

  • xih = xih+1 h = 1, 2, . . . , s − 1 and
  • the sum of the exponents of each proper letter is even.
slide-38
SLIDE 38

s-word

⋄ aǫ2aǫ4aǫ2aǫ6aǫ8aǫ2aǫ20 → a7 ⋄ σǫ4aǫ2aǫ6aǫbǫ2bǫ3aǫ2aǫbǫ3cǫ2cǫ3bǫ2bǫ2bǫaǫ2 → σa3ǫb2ǫa2ǫbǫc2ǫb3ǫa A class of pattern-equivalent words

w on the alphabet Σ is

called s-word and it has one of the following forms: i) if Σ = {ǫ, a}, then

w = [aλ], λ odd;

ii) if Σ = {σ, ǫ, a, b, . . .}, then

w = [σxλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is ] with

λ1, . . . , λs positive integers and

  • xih = xih+1 h = 1, 2, . . . , s − 1 and
  • the sum of the exponents of each proper letter is even.
slide-39
SLIDE 39

s-word

⋄ aǫ2aǫ4aǫ2aǫ6aǫ8aǫ2aǫ20 → a7 ⋄ σǫ4aǫ2aǫ6aǫbǫ2bǫ3aǫ2aǫbǫ3cǫ2cǫ3bǫ2bǫ2bǫaǫ2 → σa3ǫb2ǫa2ǫbǫc2ǫb3ǫa A class of pattern-equivalent words

w on the alphabet Σ is

called s-word and it has one of the following forms: i) if Σ = {ǫ, a}, then

w = [aλ], λ odd;

ii) if Σ = {σ, ǫ, a, b, . . .}, then

w = [σxλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is ] with

λ1, . . . , λs positive integers and

  • xih = xih+1 h = 1, 2, . . . , s − 1 and
  • the sum of the exponents of each proper letter is even.
slide-40
SLIDE 40

s-word

⋄ aǫ2aǫ4aǫ2aǫ6aǫ8aǫ2aǫ20 → a7 ⋄ σǫ4aǫ2aǫ6aǫbǫ2bǫ3aǫ2aǫbǫ3cǫ2cǫ3bǫ2bǫ2bǫaǫ2 → σa3ǫb2ǫa2ǫbǫc2ǫb3ǫa A class of pattern-equivalent words

w on the alphabet Σ is

called s-word and it has one of the following forms: i) if Σ = {ǫ, a}, then

w = [aλ], λ odd;

ii) if Σ = {σ, ǫ, a, b, . . .}, then

w = [σxλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is ] with

λ1, . . . , λs positive integers and

  • xih = xih+1 h = 1, 2, . . . , s − 1 and
  • the sum of the exponents of each proper letter is even.
slide-41
SLIDE 41

s-word

⋄ aǫ2aǫ4aǫ2aǫ6aǫ8aǫ2aǫ20 → a7 ⋄ σǫ4aǫ2aǫ6aǫbǫ2bǫ3aǫ2aǫbǫ3cǫ2cǫ3bǫ2bǫ2bǫaǫ2 → σa3ǫb2ǫa2ǫbǫc2ǫb3ǫa A class of pattern-equivalent words

w on the alphabet Σ is

called s-word and it has one of the following forms: i) if Σ = {ǫ, a}, then

w = [aλ], λ odd;

ii) if Σ = {σ, ǫ, a, b, . . .}, then

w = [σxλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is ] with

λ1, . . . , λs positive integers and

  • xih = xih+1 h = 1, 2, . . . , s − 1 and
  • the sum of the exponents of each proper letter is even.
slide-42
SLIDE 42

Sunoids and sunwords

A sunoid is a multisun G that hereditarily satisfies the N-conditions, i.e., each sub-multisun of G satisfies the N-conditions. The class of sunoids is denoted by S∗. If S∗∗ denotes the class of HOH-free multisuns and S denotes the class of multisuns satisfying the N-conditions, then we clearly have S∗∗ ⊆ S∗ ⊆ S. Aim S∗∗ = S∗. Call sunword the s-word of a sunoid Characterizing sunoids among multisuns satisfying the N-conditions is the same as characterizing sunwords among s-words.

slide-43
SLIDE 43

Sunoids and sunwords

A sunoid is a multisun G that hereditarily satisfies the N-conditions, i.e., each sub-multisun of G satisfies the N-conditions. The class of sunoids is denoted by S∗. If S∗∗ denotes the class of HOH-free multisuns and S denotes the class of multisuns satisfying the N-conditions, then we clearly have S∗∗ ⊆ S∗ ⊆ S. Aim S∗∗ = S∗. Call sunword the s-word of a sunoid Characterizing sunoids among multisuns satisfying the N-conditions is the same as characterizing sunwords among s-words.

slide-44
SLIDE 44

Sunoids and sunwords

A sunoid is a multisun G that hereditarily satisfies the N-conditions, i.e., each sub-multisun of G satisfies the N-conditions. The class of sunoids is denoted by S∗. If S∗∗ denotes the class of HOH-free multisuns and S denotes the class of multisuns satisfying the N-conditions, then we clearly have S∗∗ ⊆ S∗ ⊆ S. Aim S∗∗ = S∗. Call sunword the s-word of a sunoid Characterizing sunoids among multisuns satisfying the N-conditions is the same as characterizing sunwords among s-words.

slide-45
SLIDE 45

Sunoids and sunwords

A sunoid is a multisun G that hereditarily satisfies the N-conditions, i.e., each sub-multisun of G satisfies the N-conditions. The class of sunoids is denoted by S∗. If S∗∗ denotes the class of HOH-free multisuns and S denotes the class of multisuns satisfying the N-conditions, then we clearly have S∗∗ ⊆ S∗ ⊆ S. Aim S∗∗ = S∗. Call sunword the s-word of a sunoid Characterizing sunoids among multisuns satisfying the N-conditions is the same as characterizing sunwords among s-words.

slide-46
SLIDE 46

Parity conditions

Two necessary and sufficient conditions:

  • first show that sunwords satisfy certain parity conditions on

the exponents of the proper letters

  • then show that sunwords satisfy a sort of continuity

property with respect to a linear order on the proper letters, Lemma If

w is a sunword on Σ − {ǫ} then it has a representative

σxλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is that satisfies the following parity

conditions: (i) λ1 and λs are both odd; (ii) λh is odd if and only if xih−1 = xih+1, h = 2, . . . , s − 1.

slide-47
SLIDE 47

Parity conditions

Two necessary and sufficient conditions:

  • first show that sunwords satisfy certain parity conditions on

the exponents of the proper letters

  • then show that sunwords satisfy a sort of continuity

property with respect to a linear order on the proper letters, Lemma If

w is a sunword on Σ − {ǫ} then it has a representative

σxλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is that satisfies the following parity

conditions: (i) λ1 and λs are both odd; (ii) λh is odd if and only if xih−1 = xih+1, h = 2, . . . , s − 1.

slide-48
SLIDE 48

a7 σa3ǫb2ǫa2ǫbǫc2ǫb3ǫa

a a a a a a a ǫ4 ǫ2 ǫ6 ǫ2 ǫ2 σ a a a b b a a b c c b b b a ǫ ǫ4 ǫ2 ǫ6 ǫ2 ǫ2 ǫ2 ǫ ǫ3 ǫ2 ǫ3 ǫ ǫ2 ǫ3 ǫ2 ǫ8 ǫ20 (b) (a)

slide-49
SLIDE 49

Jump-freeness

An s-word induces a finite chain σ ≺ a ≺ b ≺ c · · · . Given

w = [xi0xλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is ] s-word with xi0 = σ and

linear order σ ≺ a ≺ b · · · .

  • two letters x and y form a cover pair if x y and

∃z = x, y such that x ≺ z ≺ y.

  • two letters xih and xih+1 , h ≥ 0 are a jump on x and y in
w

if x and y is not a cover pair, and either xih = x, xih+1 = y or xih = y, xih+1 = x (sums mod s + 1). An s-word is jump-free if it contains no jump for any {x, y} = ǫ. Theorem If

w is a sunword then w is jump-free.
slide-50
SLIDE 50

Jump-freeness

An s-word induces a finite chain σ ≺ a ≺ b ≺ c · · · . Given

w = [xi0xλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is ] s-word with xi0 = σ and

linear order σ ≺ a ≺ b · · · .

  • two letters x and y form a cover pair if x y and

∃z = x, y such that x ≺ z ≺ y.

  • two letters xih and xih+1 , h ≥ 0 are a jump on x and y in
w

if x and y is not a cover pair, and either xih = x, xih+1 = y or xih = y, xih+1 = x (sums mod s + 1). An s-word is jump-free if it contains no jump for any {x, y} = ǫ. Theorem If

w is a sunword then w is jump-free.
slide-51
SLIDE 51

Jump-freeness

An s-word induces a finite chain σ ≺ a ≺ b ≺ c · · · . Given

w = [xi0xλ1

i1 ǫxλ2 i2 ǫxλ3 i3 · · · ǫxλs is ] s-word with xi0 = σ and

linear order σ ≺ a ≺ b · · · .

  • two letters x and y form a cover pair if x y and

∃z = x, y such that x ≺ z ≺ y.

  • two letters xih and xih+1 , h ≥ 0 are a jump on x and y in
w

if x and y is not a cover pair, and either xih = x, xih+1 = y or xih = y, xih+1 = x (sums mod s + 1). An s-word is jump-free if it contains no jump for any {x, y} = ǫ. Theorem If

w is a sunword then w is jump-free.
slide-52
SLIDE 52

Theorem An s-word on an alphabet Σ is a sunword if and only if it is jump-free and satisfies the parity conditions.

ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ

ǫ ǫ

ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ ǫ σ σ a a b b b b c c c c a a ǫ σa2ǫb2ǫc2 σaǫbǫc2ǫbǫa

slide-53
SLIDE 53

The characterization

Theorem Let G be a multisun. The following statements are equivalent (1) G is a sunoid (2) The s-word of G is a sunword (3) G is HOH-free Sketch of proof Since HOH-free multisuns satisfies the N-conditions hereditarily, we have that (3)⇒(1). By definition, (1)⇒(2). It remains to show that (2)⇒(3). Parity conditions and jump-freeness are crucial to prove that the sunoid represented by a sunword is HOH-free.

slide-54
SLIDE 54

The characterization

Theorem Let G be a multisun. The following statements are equivalent (1) G is a sunoid (2) The s-word of G is a sunword (3) G is HOH-free Sketch of proof Since HOH-free multisuns satisfies the N-conditions hereditarily, we have that (3)⇒(1). By definition, (1)⇒(2). It remains to show that (2)⇒(3). Parity conditions and jump-freeness are crucial to prove that the sunoid represented by a sunword is HOH-free.

slide-55
SLIDE 55

Characterization of MNB diamond-free graphs

Given a cycle C, the edges of C that form a triangle with another vertex of C are called non-proper. Odd generalized sun is a graph G whose vertex set can be partitioned into two sets: ⋄ a (not necessarily induced) odd cycle C of G with a set of non-proper edges {ej}j∈J (J is allowed to be empty) ⋄ a stable set U = {uj}j∈J such that uj is adjacent only to the endpoints of a non-proper edge of C. Theorem (Bonomo, Chudnovsky, Duran 2007) Let G be a diamond-free graph. Then G is clique-perfect if and only if no induced subgraph of G is an

  • dd generalized sun.
slide-56
SLIDE 56

Characterization of MNB diamond-free graphs

Given a cycle C, the edges of C that form a triangle with another vertex of C are called non-proper. Odd generalized sun is a graph G whose vertex set can be partitioned into two sets: ⋄ a (not necessarily induced) odd cycle C of G with a set of non-proper edges {ej}j∈J (J is allowed to be empty) ⋄ a stable set U = {uj}j∈J such that uj is adjacent only to the endpoints of a non-proper edge of C. Theorem (Bonomo, Chudnovsky, Duran 2007) Let G be a diamond-free graph. Then G is clique-perfect if and only if no induced subgraph of G is an

  • dd generalized sun.
slide-57
SLIDE 57

They noticed explicitly that the above one is far from being a minimal characterization. Indeed, the minimal characterization is: Theorem (G., Apollonio 2014) Let G be a diamond-free graph. Then G is clique-perfect if and only if no induced subgraph of G is an

  • dd hole or a sunoid.

In other words: Theorem (G., Apollonio 2014) Let G be a diamond-free perfect graph. Then G is minimally non-balanced if and only if G is a sunoid.

slide-58
SLIDE 58

They noticed explicitly that the above one is far from being a minimal characterization. Indeed, the minimal characterization is: Theorem (G., Apollonio 2014) Let G be a diamond-free graph. Then G is clique-perfect if and only if no induced subgraph of G is an

  • dd hole or a sunoid.

In other words: Theorem (G., Apollonio 2014) Let G be a diamond-free perfect graph. Then G is minimally non-balanced if and only if G is a sunoid.

slide-59
SLIDE 59

Thank you for your attention!!!

.