First upgrade on conjecture 1 of Sudakov

zxm

Abstract

1 Conjecture is correct if it correct when n is small

Conjecture 1. If G is a graph with n vertices and m edges, such that m = tk (n) + 1, then the
maximum chordal subgraph of G denoted as F has at least kn − k1 n + 2 − k+1


2 for any n and k.

In this part, we want to proof that:

Theorem 2. For any fixed k̂, if conjecture 1 is established when n ≤ 2k̂ 3 − 2k̂ 2 − 12k̂ for any k ′ ≥ 2,
then the conjecture is established.

Before proof, we have the follow lemma:

Lemma 3. By contradiction, if n > 3k̂ 2 + k̂, then for any k ≤ k̂, if m = tk (n) + 1, then there exist
X ⊂ V (G) and |X| = n2 with e(G[X]) ≤ tk+1 ( n2 ).

Proof. If there is k ≤ k̂, and a graph G with n vertices and m edges, but for all the X ∈ S = {S|S ⊂
V (G), |S| = n2 }, is all has e(X) := e(G[X]) > tk+1 ( n2 ). Then we traverse all X ∈ S and sum the edges
in X in two ways.
First way is to sum all the edges of X directly, the other way is when V (G) is equally divided
into X and Y , then of course X, y ∈ S. So e(X) + e(Y ) = e(G) − e(X, Y ). So we can traverse all the
equally division T and sum the right part.
Then we have:
X n X X
tk+1 ( ) < e(X) = e(G) − e(X, Y ) (1)
2
X∈S X∈S (X,Y )∈T

Since any edge e appears in E(X, Y ) when the endpoints of e and divided into different part. For the
symmetry of the process, we know that its exactly half of the T. So
n n



X n
1 n
e(G) − e(X, Y ) = 2 e(G) − 2
e(G) (2)
2 2 2
(X,Y )∈T

1
Then the inequality can be written as
n n n
 



n n
2
1 n
2
1 n
2
n tk+1 (n) < e(G) − (G) = (tk (n) + 1) (3)
2 2 2 e 2 2

By simplifying we have that: p

n≤ 2k 2 + 2k (4)

Which is a contradiction about the assume that n > 3k̂ 2 + k̂.

Now we can begin the proof of theorem 2

Proof. By contradiction, if the conjecture isn’t established for k ≤ k̂, we may assume that k ≥ 4,
because the theorem is established when k ≤ 3. Then suppose that some n is the smallest number
such that there exist a graph G with n vertices and m = tk (n) + 1 edges and its maximum chordal
subgraph denoted as F with e(F ) ≤ kn − k1 n + 2 − k+1


2 .
If n ≤ 3k̂ + k̂, then it’s a contradiction for the assumption. Then is n ≥ 3k̂ 2 + k̂, for the lemma 3
2

we can know that there exist a equally division (X, Y ) with X ∪ Y = V (G) and |X| = |Y | = n2 that
e(X) ≤ tk+1 ( n2 ).
We may assume x := e(X) satisfying ts−1 ( n2 ) + 1 ≤ x ≤ ts ( n2 ), with s ≤ k + 1.
So from the assumption, n2 satisfying the theorem, i.e. if Fx is the maximum chordal subgraph of
) 2 + 2 − 2s . So if y := e(y), we have:
1 n


G[X] then e(Fx ) ≥ (s − 1 − s−1

y ≥ tk (5)

k−1 2
Lemma 4. There exist a equal division (X, Y ) such that e(X, Y ) ≤ 4k n and either X or Y (for
example X) satisfying e(X) ≤ k−1 2
8k n + n + 1.

k−1 2
Proof. First we proof that there mush be (X, Y ) such that e(X, Y ) ≤ 4k n , otherwise we have:
n n



n
k − 1 X 1 n
2
( n2 ) ≤ = 2
e(G) (6)
2 4k 2 2
(X,Y )∈T

i.e.
k−1 2 k−1 2
2( n ) ≤ e(G) = n (7)
4k 2k
which is a contradiction. So we choose (X1 , Y1 ) is the division with the maximum cue edges e(X1 , Y1 )
satisfying e(X1 , y1 ) leq k−1 2
4k n .
If e(X1 , y1 ) ≥ k−1 2 k−1 2
4k n − 2n, then e(X) + e(Y ) ≤ e(G) − e(X1 , y1 ) ≤ 4k n + 2n + 1. So we may
assume e(X) < e(Y ) and e(X) ≤ 12 ( k−1 2 k−1 2
4k n + 2n + 1) ≤ 8k n + n + 1.
If e(X1 , Y1 ) < k−1 2
4k n − 2n, then if there is a vertex x ∈ X1 with dX1 (x) > dY1 (x) and a vertex
y ∈ Y1 with dX1 (y) < dY1 (y), then we can change x and y then have new (X2 , Y2 ), of course e(X2 , Y2 ) >

2
e(X1 , y1 ), and e(X2 , Y2 ) − e(X1 , Y1 ) ≤ 2n. From the definition of (X1 , Y1 ) we know that e(X2 , Y2 ) ≥
k−1 2 k−1 2
4k n + 1 and e(X1 , Y1 ) < 4k n − 2n, which is a contradiction!
So either X or Y (for example X) satisfying dx1 (x) ≤ dY1 (x) for all x ∈ X1 , so sum all the vertices’
degree of X1 we have that:
k−1 2
2e(X) ≤ e(X1 , Y1 ) < n − 2n (8)
4k
k−1 2 k−1 2
So e(X) ≤ 8k n −n≤ 8k n +n+1

Now we proof the theorem 2.

Proof. By contradiction, n is the minimum integer that doesn’t fit the conjecture which means there
is a integer k and a graph G, F is it’s maximum chordal subgraph of G and e(F ) < (k− k1 )n+2− k+1


2 .
From the lemma, we can find the equal division that e(X, Y ) ≤ k−1 2 k−1 2
4k n , and |X| ≤ 8k n + n + 1.
Now we assume that ts−1 ( n2 ) + 1 ≤ x ≤ ts ( n2 ).
Case 1. s ≤ k:
s−1 n2
y ≥ tk (n) + 1 − ts ( n2 ) − e(X, Y ) ≥ k−1 2
4k n + 1 − 2s 4 .
Next we have to proof y ≥ t2k+2−s ( n2 ) + 1, i.e. to proof:

k−1 2 s − 1 n2 n
n +1− ≥ t2k+2−s ( ) + 1 (9)
4k 2s 4 2
By simplify, it’s equal to proof that:k(k + 1) ≥ s(2k + 2 − s) is established for all s ≤ k, which is right.
Then we may assume that Fx , Fy is the maximum chordal subgraph of G[X] and G[Y ], and F is the
maximum chordal subgraph of G. Then e(F ) − 1 ≥ e(Fx ) + e(Fy ).
From the induction hypothesis, we have that:
 
1 n s
e(Fx ) ≥ (s − 1 − ) +2− (10)
s−1 2 2
 
1 n 2k + 2 − s
e(Fy ) ≥ (2k + 2 − s − ) +2− (11)
2k + 2 − s 2 2
So we have that:
     
1 n s 1 n 2k + 2 − s 1 k+1
(s−1− ) +2− +(2k+2−s− ) +2− ≤ (k− )n+2− (12)
s−1 2 2 2k + 2 − s 2 2 k 2
By simplify, we can find that:
3n
≤ s(s − 1) + (2k + 3 − s)(2k + 2 − s) − k(k + 1) − 6 (13)
k2
From above equation, we have that: n ≤ 2k 3 − 14 2 28 3 2
3 k − 3 k ≤ 2k − 2k − 12k, which is a contradiction
about the assumption of for any k, when n < 2k 3 − 2k 2 − 12k, the conjuecture established.
Case 2. s = k + 1: Then x ≤ tk ( n2 ) + n + 1, and

k−1 2 k − 1 n2 k−1 2 1k−1 2 n

y≥ n +1− −n−1− n = n − n ≥ tk−1 ( ) + 1 (14)
2k 2k 4 4k 8 k 2

3
 k+1 k
By hypothesis, Fx ≥ (k − k1 ) n2 + 2 − 1 n



2 , Fy ≥ (k − 1 − k−1 ) 2 +2− 2 . Then we can find that:

n ≤ 2k 3 − 2k 2 − 12k (15)

which is also a contradiction.

So to proof the conjecture, we only need to proof it’s established for any fixed k when n <
2k 3 − 2k 2 − 12k.

2 Some other exporitions

Theorem 5. C is a clique with k vertices, and G is a graph includes C, if e(C, G − C) ≤ 2k − 3, then
there exist F is the maximum chordal subgraph of G, and C ⊆ G.

Proof. Assume that all the components of F [C] are C1 , C2 , . . . , Ct , if t ≥ 2, then we consider if there
exist a component Ci with |Ci | ≥ 2, then for any j ̸= i, we have that:
If e(Cj , G − C) ≤ 2|Cj |, then F ′ := F − e(Cj , G − C) + e(Cj , Ci ) is also a chordal graph with
e(F ′ ) ≥ e(F ). We may assume that F is the chordal subgraph with least components in F [C]. So we
have
X X
e(Cj , G − C) > 2 |Cj | = 2[e(C, G − C) − |Ci |] (16)
j̸=i j̸=i

If there is another components besides Ci for example Cp with Cp ≥ 2 then we have e(Ci , G − C) ≥
P
2|Ci |, so i e(Ci , G − C) = e(C, G − C) > 2|C|. If all other components are single vertex, then
suppose |Ci | = s, and k − s single vertices. Every single vertex accidents at least s + 1, and the edges
connects Ci and G − C are at least s + 1.
So e(C, G − C) ≥ (k − s)s = 2 + s + 2, and k − 1 ≥ s ≥ 2, we have that: e(C, G − C) ≥ 2k, which
is a contradiction.
If t = 1, which means that F [C] is connected, then if there exist a edge e ∈ / F, e ∈ C, then suppose
e = uv, then F + e will contains s circle q with length large than 4 and we may assume that q contains
the vertices in G − C. We suppose that F is the chordal graph that e(F ) < 2k − 2 and with the
smallest e(F [C], F [G − C]).
Let A1 is the vertices that connects only with u but not with v in C; A2 is the vertices that only
connects with v but not with u in C; A3 is the vertices connect with both u and v; B is the vertices
connect neither u or v. u′ , v ′ are the vertices adjacent to u and v in G − C(we may assume that u ≁ v ′
and v ≁ u′ ), then p is the shortest path in G − C connects u′ , v ′ . For the chordality, we know that all
the vertices in P are connect with all the vertices in A3 . Suppose that the number of vertices in A1
that connects u′ and v ′ with 2 edges, 1 edge and 0 edge is x2 , x1 , x0 , and |A1 | = x = x2 + x1 + x0 .
The number of vertices in A2 that connects u′ and v ′ with 2 edges, 1 edge and 0 edge is y 2 , y1 , y0 ,
1 1
and |A2 | = y = y 2 + y 1 + y 0 . z = |A3 |. So the a := e(C, G − C) ≥ 2z + 2x2 + 2y 2 + x +y2 +2 , because

4
 1 1
these edges can make a chordal graph. The x +y2 +2 part is because we can choose the vertices only
connects u or v. And the edges that not exist in F of E(C) are at least 2(k − 2 − x − y − z) + (x0 +
x1 )(y 0 + y 1 ) + x + y + 1 := c. The part (x0 + x1 )(y 0 + y 1 ) is because any two vertices in A1 and A2
connects only one vertices in {u, v} can’t connect in F, or it will have a chordal in a large circle.
We assume that b := e(F [C], F [G − C]), we have b − a > c,otherwise we can delete all edges in
b − a and adds all edges in C to make it a k-clique, it still be chordal graph. Then we have
1 1
x2 + y 2 + (x0 + x1 )(y 0 + y 1 ) ≤ x1 + y 1 + x0 + y 0 − 1 (17)
2 2
If x0 + x1 ≥ 1 and y 0 + y 1 ≥ 1, then the above inequality will not established or the equation will
be established(but in this time we can find a chordal graph F’ with the same size but with C in F’).
Therefor x0 + x1 = 0 or y 0 + y 1 = 0 but they can’t all be 0. So we can assume that x0 + x1 = 0
and y 0 + y 1 ≥ 1. For any vertex x in A2 and B, uxe(F ), so when add ux, there exist a circle
without chordal that larger than 3. Besides, e(u, G − C) ≥ |B + |A2 | = y + z, or we can delete all
the edges of E(u, G − C) and add edges of E(u, A2 ) then we find a larger F ′ or the same size with
e(F [C], F [G − C]) > e(F ′ [C], F ′ [G − C]). For any ω i in E(u,
G-C) wecanf indtheshortestpathPi in G − C.
If P i pass though v, for any a ∈ A3 , aω i ∈ E(F ); if P i not pass though v, then there will exist a
edge in E(G − C, C) not exist in all the edges we have known. So we have

e(C, G − C) ≥ 2x2 + 2(k − 2 − x2 − y − z) + 2y + 2z + 2 = 2k − 2 (18)

Contradiction. So we can find a maximum chordal graph that contains C as a clique in it.

5
Theorem 6. If G is a graph with n vertices and m edges and tk (n) + 1 ≤ m ≤ tk+1 (n) without k + 2
clique, H is the maximum chordal subgraph of G and when n is large enough H has a induced subgraph
F which has exactly k−2 2k−3 1
k−1 n cliques with k + 1 vertices and v(F ) ≤ ( 2(k−1) − 2k2 )n + k.

Proof. We assume that F is a minimal chordal subgraph of H with all the k + 1 cliques of H. Since
F is a chordal graph, then it will has a PEO, remark the number of edges every time a vertex put
in the graph according to the PEO. We can denote the number of edges put in F as (a1 , a2 , . . . , an ),
and the bi is the number of j that aj = bi . Then we can proof that b0 ≤ b1 ≤ · · · ≤ bk :
By induction on n, if n = k + 1, then we have a maximal clique, which established. Then if it is
established when n-1, then for any such chordal graph F, we find a vertex v that N (v) is a clique,
and for F is a minimal clique that N (v) = k. So we can delete all the simple vertices adjacent to v,
and we get another chordal graph F’ is also a minimal graph with some k+1 cliques. We can define
b′i the same with it in F ′ , and then b′0 ≤ b′1 ≤ · · · ≤ b′k . Then we will adds the vertices deleted before,
which will adds (c, c + 1, . . . , k) edges. So bc = b′c + 1, . . . , bk = b′k + 1, bi = b′i , i < c. So we have
b0 ≤ b1 ≤ · · · ≤ bk .
Assume that t = b0 +b1 +· · ·+bk−1 , then b0 +b1 +· · ·+dk = n, e(F ) = 0·d0 +1·b1 +2·b2 +· · ·+k·bk .
So e(F ) = 0·d0 +1·b1 +2·b2 +· · ·+k ·bk ≤ (k−1)(k−2) 2 +(t−k +1)(k −1)+(n−t)k = nk − 12 k(k −1)−t.
n 1
If t > k−1 , then e(F ) ≤ n(k − k−1 ) − 12 k(k − 1), which is a contradiction.
So we have can assume that v(F ) = k−2 k+t−1
k−1 n + t, then bk ≥ k−1 , so we have that: e(F ) ≥
k+t−1 (k−1)k
k−2
k−1 nk + k−1 · 2
k−2
= k−1 nk + (k+t−1)k
2 . So we have that kt k(k−1) 1
2 + k−2 ≤ (k − k )n, because when n is
q
large enough, we need to assure that e(F ) is large that Sudakov’s results (k− k1 )n+ (k+1)a −2− k+1


k 2 .
Then we get: t < 21 ( k−1
1
− k12 )n, i.e. v(F ) < ( 2(k−1)
2k−3
− 2k12 )n + k.

Assume that F is the chordal graph in theorem 6 which has k + t + k−2

k−1 n vertices then e(F ) ≥
k−2 k−1 k−2 k−1



k−1 nk + 2 + (t + 1)(k + 1), and e(F ) ≤ k−1 nk + 2 + (t + 1)(k − 1)

Theorem 7. From theorem 6, we can let the vertices set be the vertices of F. Then we can get a lower
bound of the ratio tω (H), where H is the maximal chordal graph of G, and G is the graph defined
k+t+ k−2
k−1
n
in theorem 6. Then tω(H) ≤ k−2
nk+k k+t−1 . This is monotonically increasing with respect to t. So
k−2 2
1
k− 2k − 12 − 23 1 2
tω (H) ≤ 7
2k
1
k2 − 4 k− 4k − 14
:= d(K). We can find that when k ≥ 4, k−1 < d(k) < k, so this is a lower
upper bound of tω (H).

Theorem 8. If G has n vertices and tk (n) + 1 edges, and there exist a partition of V (G) {V1 , . . . , Vk }
such that for any Vi , ∃vi ∈ Vi and for all uj ∈ / Vj , uj ∼ vi , and there exist a Vt and ω1 , ω2 ∈ Vt and
ωi ̸= vt , that ωi adjacent to all vertices not in Vt , then the maximum chordal graph of G has at least
(k − k1 )n + 2 − k+1


2 edges.

References