![Page 1: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/1.jpg)
SMALL SUBGRAPHS OFRANDOM GRAPHS
Andrzej Rucinski
Adam Mickiewicz University, Poznan, Poland,
Emory University
MAA 2005, Atlanta – p. 1/49
![Page 2: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/2.jpg)
Subgraph counts
A copy of a graph G in another graph F is a (weak)subgraph G′ of F isomorphic to G.
Given graph G with v = vG vertices and e = eG
edges, letXG(n, p) = XG = X
be the number of copies of G in G(n, p).
The expectation:
EX = N(n, G)pe =
(
n
v
)
v!
aut(G)pe
MAA 2005, Atlanta – p. 2/49
![Page 3: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/3.jpg)
Subgraph counts
A copy of a graph G in another graph F is a (weak)subgraph G′ of F isomorphic to G.
Given graph G with v = vG vertices and e = eG
edges, letXG(n, p) = XG = X
be the number of copies of G in G(n, p).
The expectation:
EX = N(n, G)pe =
(
n
v
)
v!
aut(G)pe
MAA 2005, Atlanta – p. 2/49
![Page 4: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/4.jpg)
Subgraph counts
A copy of a graph G in another graph F is a (weak)subgraph G′ of F isomorphic to G.
Given graph G with v = vG vertices and e = eG
edges, letXG(n, p) = XG = X
be the number of copies of G in G(n, p).
The expectation:
EX = N(n, G)pe =
(
n
v
)
v!
aut(G)pe
MAA 2005, Atlanta – p. 2/49
![Page 5: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/5.jpg)
1st moment method
By Markov’s inequality
p n−v/e ⇒ P(X > 0) ≤ EX → 0
Note that
p n−v/e ⇔ EX →∞
Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 3/49
![Page 6: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/6.jpg)
1st moment method
By Markov’s inequality
p n−v/e ⇒ P(X > 0) ≤ EX → 0
Note that
p n−v/e ⇔ EX →∞
Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 3/49
![Page 7: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/7.jpg)
1st moment method
By Markov’s inequality
p n−v/e ⇒ P(X > 0) ≤ EX → 0
Note that
p n−v/e ⇔ EX →∞
Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 3/49
![Page 8: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/8.jpg)
1st moment method
By Markov’s inequality
p n−v/e ⇒ P(X > 0) ≤ EX → 0
Note that
p n−v/e ⇔ EX →∞Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 3/49
![Page 9: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/9.jpg)
Example - the diamond
G = D, the diamond, that is D = K4 −K2.
EXD = 6
(
n
4
)
p5 → 0 ⇐ p = o(n−4/5)
Let p n−4/5, D1, . . . , D6(n4)
be all copies of D in Kn,
Ii = 1 if G(n, p) ⊃ Di and 0 otherwise. Write i ∼ j ifE(Di) ∩ E(Dj) 6= ∅.
MAA 2005, Atlanta – p. 4/49
![Page 10: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/10.jpg)
Example - the diamond
G = D, the diamond, that is D = K4 −K2.
EXD = 6
(
n
4
)
p5 → 0 ⇐ p = o(n−4/5)
Let p n−4/5, D1, . . . , D6(n4)
be all copies of D in Kn,
Ii = 1 if G(n, p) ⊃ Di and 0 otherwise. Write i ∼ j ifE(Di) ∩ E(Dj) 6= ∅.
MAA 2005, Atlanta – p. 4/49
![Page 11: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/11.jpg)
Example - the diamond
G = D, the diamond, that is D = K4 −K2.
EXD = 6
(
n
4
)
p5 → 0
⇐ p = o(n−4/5)
Let p n−4/5, D1, . . . , D6(n4)
be all copies of D in Kn,
Ii = 1 if G(n, p) ⊃ Di and 0 otherwise. Write i ∼ j ifE(Di) ∩ E(Dj) 6= ∅.
MAA 2005, Atlanta – p. 4/49
![Page 12: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/12.jpg)
Example - the diamond
G = D, the diamond, that is D = K4 −K2.
EXD = 6
(
n
4
)
p5 → 0 ⇐ p = o(n−4/5)
Let p n−4/5, D1, . . . , D6(n4)
be all copies of D in Kn,
Ii = 1 if G(n, p) ⊃ Di and 0 otherwise. Write i ∼ j ifE(Di) ∩ E(Dj) 6= ∅.
MAA 2005, Atlanta – p. 4/49
![Page 13: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/13.jpg)
Example - the diamond
G = D, the diamond, that is D = K4 −K2.
EXD = 6
(
n
4
)
p5 → 0 ⇐ p = o(n−4/5)
Let p n−4/5, D1, . . . , D6(n4)
be all copies of D in Kn,
Ii = 1 if G(n, p) ⊃ Di and 0 otherwise. Write i ∼ j ifE(Di) ∩ E(Dj) 6= ∅.
MAA 2005, Atlanta – p. 4/49
![Page 14: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/14.jpg)
Example - the diamond
G = D, the diamond, that is D = K4 −K2.
EXD = 6
(
n
4
)
p5 → 0 ⇐ p = o(n−4/5)
Let p n−4/5, D1, . . . , D6(n4)
be all copies of D in Kn,
Ii = 1 if G(n, p) ⊃ Di and 0 otherwise.
Write i ∼ j ifE(Di) ∩ E(Dj) 6= ∅.
MAA 2005, Atlanta – p. 4/49
![Page 15: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/15.jpg)
Example - the diamond
G = D, the diamond, that is D = K4 −K2.
EXD = 6
(
n
4
)
p5 → 0 ⇐ p = o(n−4/5)
Let p n−4/5, D1, . . . , D6(n4)
be all copies of D in Kn,
Ii = 1 if G(n, p) ⊃ Di and 0 otherwise. Write i ∼ j ifE(Di) ∩ E(Dj) 6= ∅.
MAA 2005, Atlanta – p. 4/49
![Page 16: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/16.jpg)
The variance
Then
var(XD) =
var
(
∑
i
Ii
)
=
∑
i
∑
j∼i
cov(Ii, Ij)≤∑
i
∑
j∼i
E(IiIj)≤
O(n6p9 + n5p8 + n5p7 + n4p6 + n4p5)
MAA 2005, Atlanta – p. 5/49
![Page 17: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/17.jpg)
The variance
Then
var(XD) = var
(
∑
i
Ii
)
=
∑
i
∑
j∼i
cov(Ii, Ij)
≤∑
i
∑
j∼i
E(IiIj)≤
O(n6p9 + n5p8 + n5p7 + n4p6 + n4p5)
MAA 2005, Atlanta – p. 5/49
![Page 18: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/18.jpg)
The variance
Then
var(XD) = var
(
∑
i
Ii
)
=
∑
i
∑
j∼i
cov(Ii, Ij) ≤∑
i
∑
j∼i
E(IiIj)
≤
O(n6p9 + n5p8 + n5p7 + n4p6 + n4p5)
MAA 2005, Atlanta – p. 5/49
![Page 19: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/19.jpg)
The variance
Then
var(XD) = var
(
∑
i
Ii
)
=
∑
i
∑
j∼i
cov(Ii, Ij) ≤∑
i
∑
j∼i
E(IiIj) ≤
O(n6p9 + n5p8 + n5p7 + n4p6 + n4p5)
MAA 2005, Atlanta – p. 5/49
![Page 20: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/20.jpg)
The variance
Then
var(XD) = var
(
∑
i
Ii
)
=
∑
i
∑
j∼i
cov(Ii, Ij) ≤∑
i
∑
j∼i
E(IiIj) ≤
O(n6p9 + n5p8 + n5p7 + n4p6 + n4p5)
MAA 2005, Atlanta – p. 5/49
![Page 21: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/21.jpg)
2nd moment method
Hence,
P(XD = 0) ≤ var(XD)
(EXD)2≤
O
(
1
n2p+
1
n3p2+
1
n3p3+
1
n4p4+
1
n4p5
)
= o(1)
provided p n−4/5. Indeed, e.g.,
n2p = (np1/2)2 ≥ (np5/4)2 →∞.
Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 6/49
![Page 22: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/22.jpg)
2nd moment method
Hence,
P(XD = 0) ≤ var(XD)
(EXD)2≤
O
(
1
n2p+
1
n3p2+
1
n3p3+
1
n4p4+
1
n4p5
)
= o(1)
provided p n−4/5. Indeed, e.g.,
n2p = (np1/2)2 ≥ (np5/4)2 →∞.
Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 6/49
![Page 23: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/23.jpg)
2nd moment method
Hence,
P(XD = 0) ≤ var(XD)
(EXD)2≤
O
(
1
n2p+
1
n3p2+
1
n3p3+
1
n4p4+
1
n4p5
)
= o(1)
provided p n−4/5.
Indeed, e.g.,
n2p = (np1/2)2 ≥ (np5/4)2 →∞.
Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 6/49
![Page 24: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/24.jpg)
2nd moment method
Hence,
P(XD = 0) ≤ var(XD)
(EXD)2≤
O
(
1
n2p+
1
n3p2+
1
n3p3+
1
n4p4+
1
n4p5
)
= o(1)
provided p n−4/5. Indeed, e.g.,
n2p = (np1/2)2 ≥ (np5/4)2 →∞.
Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 6/49
![Page 25: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/25.jpg)
2nd moment method
Hence,
P(XD = 0) ≤ var(XD)
(EXD)2≤
O
(
1
n2p+
1
n3p2+
1
n3p3+
1
n4p4+
1
n4p5
)
= o(1)
provided p n−4/5. Indeed, e.g.,
n2p = (np1/2)2 ≥ (np5/4)2 →∞.
Is it true that P(X > 0) → 1 if EX →∞ ???
MAA 2005, Atlanta – p. 6/49
![Page 26: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/26.jpg)
Counterexample
“Conjecture”: p0(there is a copy of G) = n−vG/eG
IS FALSE!!!
n−5/6 p = n−9/11 n−4/5
P(XK > 0) ≤ P(XD > 0) = o(1)
MAA 2005, Atlanta – p. 7/49
![Page 27: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/27.jpg)
Counterexample
“Conjecture”: p0(there is a copy of G) = n−vG/eG
IS FALSE!!!
n−5/6 p = n−9/11 n−4/5
P(XK > 0) ≤ P(XD > 0) = o(1)
MAA 2005, Atlanta – p. 7/49
![Page 28: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/28.jpg)
Counterexample
“Conjecture”: p0(there is a copy of G) = n−vG/eG
IS FALSE!!!Counterexample: H = K, the kite,
n−5/6 p = n−9/11 n−4/5
P(XK > 0) ≤ P(XD > 0) = o(1)
MAA 2005, Atlanta – p. 7/49
![Page 29: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/29.jpg)
Counterexample
“Conjecture”: p0(there is a copy of G) = n−vG/eG
IS FALSE!!!Counterexample: H = K, the kite,
n−5/6 p = n−9/11 n−4/5
P(XK > 0) ≤ P(XD > 0) = o(1)
MAA 2005, Atlanta – p. 7/49
![Page 30: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/30.jpg)
Counterexample
“Conjecture”: p0(there is a copy of G) = n−vG/eG
IS FALSE!!!Counterexample: H = K, the kite,
n−5/6 p = n−9/11 n−4/5
P(XK > 0) ≤ P(XD > 0) = o(1)
MAA 2005, Atlanta – p. 7/49
![Page 31: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/31.jpg)
Threshold - general case
In general, let
dH =eH
vHand mG = max
H⊆GdH .
MAA 2005, Atlanta – p. 8/49
![Page 32: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/32.jpg)
Threshold - general case
In general, let
dH =eH
vHand mG = max
H⊆GdH .
Theorem (Bollobás, 1981) For every graph G witheG > 0,
p0(there is a copy of G) = n−1/mG,
MAA 2005, Atlanta – p. 8/49
![Page 33: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/33.jpg)
Threshold - general case
In general, let
dH =eH
vHand mG = max
H⊆GdH .
Theorem (Bollobás, 1981) For every graph G witheG > 0,
p0(there is a copy of G) = n−1/mG,
that is,
P(XG > 0) =
0 if p n−1/mG
1 if p n−1/mG
MAA 2005, Atlanta – p. 8/49
![Page 34: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/34.jpg)
Proof
Proof: Let npmG → 0 and let H ⊆ G be such thatdH = mG. Then
P(XG > 0) ≤ P(XH > 0) ≤ EXH
=O(nvHpeH) = (npdH)vH = o(1). Let npmG →∞. Then,for every H ⊆ G,
nvHpeH = (npdH)vH →∞
MAA 2005, Atlanta – p. 9/49
![Page 35: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/35.jpg)
Proof
Proof: Let npmG → 0 and let H ⊆ G be such thatdH = mG. Then
P(XG > 0) ≤ P(XH > 0) ≤ EXH
= O(nvHpeH) = (npdH)vH = o(1).
Let npmG →∞. Then, for every H ⊆ G,
nvHpeH = (npdH)vH →∞
MAA 2005, Atlanta – p. 9/49
![Page 36: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/36.jpg)
Proof
Proof: Let npmG → 0 and let H ⊆ G be such thatdH = mG. Then
P(XG > 0) ≤ P(XH > 0) ≤ EXH
= O(nvHpeH) = (npdH)vH = o(1).
Let npmG →∞. Then, for every H ⊆ G,
nvHpeH = (npdH)vH →∞
MAA 2005, Atlanta – p. 9/49
![Page 37: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/37.jpg)
Proof
Proof: Let npmG → 0 and let H ⊆ G be such thatdH = mG. Then
P(XG > 0) ≤ P(XH > 0) ≤ EXH
= O(nvHpeH) = (npdH)vH = o(1).
Let npmG →∞.
Then, for every H ⊆ G,
nvHpeH = (npdH)vH →∞
MAA 2005, Atlanta – p. 9/49
![Page 38: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/38.jpg)
Proof
Proof: Let npmG → 0 and let H ⊆ G be such thatdH = mG. Then
P(XG > 0) ≤ P(XH > 0) ≤ EXH
= O(nvHpeH) = (npdH)vH = o(1).
Let npmG →∞. Then, for every H ⊆ G,
nvHpeH = (npdH)vH →∞
MAA 2005, Atlanta – p. 9/49
![Page 39: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/39.jpg)
Proof
Proof: Let npmG → 0 and let H ⊆ G be such thatdH = mG. Then
P(XG > 0) ≤ P(XH > 0) ≤ EXH
= O(nvHpeH) = (npdH)vH = o(1).
Let npmG →∞. Then, for every H ⊆ G,
nvHpeH = (npdH)vH →∞and
P(XG = 0) ≤ var(XG)
(EXG)2= O
(
∑
H⊆G
1
nvHpeH
)
= o(1).
MAA 2005, Atlanta – p. 9/49
![Page 40: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/40.jpg)
At the threshold
p = Θ(n−1/mG), or npmG → c > 0.
Theorem (Bollobás (81), Karonski, Rucinski (83)) IfG is strictly balanced, that is, for all H ⊂ G we havedH < dG, and npmG → c > 0then XG has asymptoticallyPoisson distribution with expectationλ = cv/aut(G),that is, for every i ≥ 0 we have
limn→∞
P(XG = i) = e−λλi
i!.
MAA 2005, Atlanta – p. 10/49
![Page 41: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/41.jpg)
At the threshold
p = Θ(n−1/mG), or npmG → c > 0.
Theorem (Bollobás (81), Karonski, Rucinski (83)) IfG is strictly balanced, that is, for all H ⊂ G we havedH < dG,
and npmG → c > 0then XG has asymptoticallyPoisson distribution with expectationλ = cv/aut(G),that is, for every i ≥ 0 we have
limn→∞
P(XG = i) = e−λλi
i!.
MAA 2005, Atlanta – p. 10/49
![Page 42: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/42.jpg)
At the threshold
p = Θ(n−1/mG), or npmG → c > 0.
Theorem (Bollobás (81), Karonski, Rucinski (83)) IfG is strictly balanced, that is, for all H ⊂ G we havedH < dG, and npmG → c > 0
then XG has asymptoticallyPoisson distribution with expectationλ = cv/aut(G),that is, for every i ≥ 0 we have
limn→∞
P(XG = i) = e−λλi
i!.
MAA 2005, Atlanta – p. 10/49
![Page 43: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/43.jpg)
At the threshold
p = Θ(n−1/mG), or npmG → c > 0.
Theorem (Bollobás (81), Karonski, Rucinski (83)) IfG is strictly balanced, that is, for all H ⊂ G we havedH < dG, and npmG → c > 0 then XG hasasymptotically Poisson distribution with expectationλ = cv/aut(G),
that is, for every i ≥ 0 we have
limn→∞
P(XG = i) = e−λλi
i!.
MAA 2005, Atlanta – p. 10/49
![Page 44: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/44.jpg)
At the threshold
p = Θ(n−1/mG), or npmG → c > 0.
Theorem (Bollobás (81), Karonski, Rucinski (83)) IfG is strictly balanced, that is, for all H ⊂ G we havedH < dG, and npmG → c > 0 then XG hasasymptotically Poisson distribution with expectationλ = cv/aut(G), that is, for every i ≥ 0 we have
limn→∞
P(XG = i) = e−λλi
i!.
MAA 2005, Atlanta – p. 10/49
![Page 45: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/45.jpg)
The method of moments
If for every k ≥ 1
E(Xn)k = EXn(Xn − 1) · · · (Xn − k + 1) → λk
then Xn has asymptotically Poisson distribution withexpectation λ.
Note: (XG)k counts ordered k-tuples of distinct copies ofG in G(n, p).
MAA 2005, Atlanta – p. 11/49
![Page 46: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/46.jpg)
The method of moments
If for every k ≥ 1
E(Xn)k = EXn(Xn − 1) · · · (Xn − k + 1) → λk
then Xn has asymptotically Poisson distribution withexpectation λ.
Note: (XG)k counts ordered k-tuples of distinct copies ofG in G(n, p).
MAA 2005, Atlanta – p. 11/49
![Page 47: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/47.jpg)
The method of moments
If for every k ≥ 1
E(Xn)k = EXn(Xn − 1) · · · (Xn − k + 1) → λk
then Xn has asymptotically Poisson distribution withexpectation λ.
Note: (XG)k counts ordered k-tuples of distinct copies ofG in G(n, p).
MAA 2005, Atlanta – p. 11/49
![Page 48: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/48.jpg)
Proof for triangles
Set G = K3, the triangle, for convenience.
Let T1, T2, . . .be all triangles in Kn and I1, I2, . . . the correspondingindicators. Then
(XK3)k =
∑
(i1,...,ik)
Ii1 · · · Iik
and
E(XK3)k =
∑
(i1,...,ik)
E(Ii1 · · · Iik) = E′k + E
′′k
where the sum splits over disjoint and not disjointk-tuples.
MAA 2005, Atlanta – p. 12/49
![Page 49: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/49.jpg)
Proof for triangles
Set G = K3, the triangle, for convenience. Let T1, T2, . . .be all triangles in Kn and I1, I2, . . . the correspondingindicators.
Then
(XK3)k =
∑
(i1,...,ik)
Ii1 · · · Iik
and
E(XK3)k =
∑
(i1,...,ik)
E(Ii1 · · · Iik) = E′k + E
′′k
where the sum splits over disjoint and not disjointk-tuples.
MAA 2005, Atlanta – p. 12/49
![Page 50: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/50.jpg)
Proof for triangles
Set G = K3, the triangle, for convenience. Let T1, T2, . . .be all triangles in Kn and I1, I2, . . . the correspondingindicators. Then
(XK3)k =
∑
(i1,...,ik)
Ii1 · · · Iik
and
E(XK3)k =
∑
(i1,...,ik)
E(Ii1 · · · Iik) = E′k + E
′′k
where the sum splits over disjoint and not disjointk-tuples.
MAA 2005, Atlanta – p. 12/49
![Page 51: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/51.jpg)
Proof for triangles
Set G = K3, the triangle, for convenience. Let T1, T2, . . .be all triangles in Kn and I1, I2, . . . the correspondingindicators. Then
(XK3)k =
∑
(i1,...,ik)
Ii1 · · · Iik
and
E(XK3)k =
∑
(i1,...,ik)
E(Ii1 · · · Iik) = E′k + E
′′k
where the sum splits over disjoint and not disjointk-tuples.
MAA 2005, Atlanta – p. 12/49
![Page 52: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/52.jpg)
Proof for triangles
Set G = K3, the triangle, for convenience. Let T1, T2, . . .be all triangles in Kn and I1, I2, . . . the correspondingindicators. Then
(XK3)k =
∑
(i1,...,ik)
Ii1 · · · Iik
and
E(XK3)k =
∑
(i1,...,ik)
E(Ii1 · · · Iik) = E′k + E
′′k
where the sum splits over disjoint and not disjointk-tuples.
MAA 2005, Atlanta – p. 12/49
![Page 53: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/53.jpg)
Proof for triangles – cont.
E′k =
(
n
3, . . . , 3, n− 3k
)
p3k ∼(
1
6n3p3
)k
∼ (EXK3)k
Let F be a union of k not all disjoint triangles. TheneF > vF .
E′′k = O
(
∑
F
nvF peF
)
= O
(
∑
F
(np)vF peF−vF
)
= O(p)
By monotonicity assume that p = o(1). Then
E(XK3)k = E
′k + E
′′k ∼ (EXK3
)k + O(p) → λk
MAA 2005, Atlanta – p. 13/49
![Page 54: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/54.jpg)
Proof for triangles – cont.
E′k =
(
n
3, . . . , 3, n− 3k
)
p3k ∼(
1
6n3p3
)k
∼ (EXK3)k
Let F be a union of k not all disjoint triangles. TheneF > vF .
E′′k = O
(
∑
F
nvF peF
)
= O
(
∑
F
(np)vF peF−vF
)
= O(p)
By monotonicity assume that p = o(1). Then
E(XK3)k = E
′k + E
′′k ∼ (EXK3
)k + O(p) → λk
MAA 2005, Atlanta – p. 13/49
![Page 55: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/55.jpg)
Proof for triangles – cont.
E′k =
(
n
3, . . . , 3, n− 3k
)
p3k ∼(
1
6n3p3
)k
∼ (EXK3)k
Let F be a union of k not all disjoint triangles. TheneF > vF .
E′′k = O
(
∑
F
nvF peF
)
= O
(
∑
F
(np)vF peF−vF
)
= O(p)
By monotonicity assume that p = o(1). Then
E(XK3)k = E
′k + E
′′k ∼ (EXK3
)k + O(p) → λk
MAA 2005, Atlanta – p. 13/49
![Page 56: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/56.jpg)
Proof for triangles – cont.
E′k =
(
n
3, . . . , 3, n− 3k
)
p3k ∼(
1
6n3p3
)k
∼ (EXK3)k
Let F be a union of k not all disjoint triangles. TheneF > vF .
E′′k = O
(
∑
F
nvF peF
)
= O
(
∑
F
(np)vF peF−vF
)
= O(p)
By monotonicity assume that p = o(1).
Then
E(XK3)k = E
′k + E
′′k ∼ (EXK3
)k + O(p) → λk
MAA 2005, Atlanta – p. 13/49
![Page 57: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/57.jpg)
Proof for triangles – cont.
E′k =
(
n
3, . . . , 3, n− 3k
)
p3k ∼(
1
6n3p3
)k
∼ (EXK3)k
Let F be a union of k not all disjoint triangles. TheneF > vF .
E′′k = O
(
∑
F
nvF peF
)
= O
(
∑
F
(np)vF peF−vF
)
= O(p)
By monotonicity assume that p = o(1). Then
E(XK3)k = E
′k + E
′′k ∼ (EXK3
)k + O(p) → λk
MAA 2005, Atlanta – p. 13/49
![Page 58: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/58.jpg)
Beyond the threshold
npmG →∞
Question 1: Asymptotic distribution of XG
Question 2: The rate of decay of P(XG = 0)
Theorem (Rucinski (1988)) For every graph G witheG > 0,
XG − EXG√
var(XG)→ N (0, 1) as n →∞
if and only if npmG →∞ and n2(1− p) →∞.Proof: By the method of moments (details omitted).
MAA 2005, Atlanta – p. 14/49
![Page 59: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/59.jpg)
Beyond the threshold
npmG →∞Question 1: Asymptotic distribution of XG
Question 2: The rate of decay of P(XG = 0)
Theorem (Rucinski (1988)) For every graph G witheG > 0,
XG − EXG√
var(XG)→ N (0, 1) as n →∞
if and only if npmG →∞ and n2(1− p) →∞.Proof: By the method of moments (details omitted).
MAA 2005, Atlanta – p. 14/49
![Page 60: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/60.jpg)
Beyond the threshold
npmG →∞Question 1: Asymptotic distribution of XG
Question 2: The rate of decay of P(XG = 0)
Theorem (Rucinski (1988)) For every graph G witheG > 0,
XG − EXG√
var(XG)→ N (0, 1) as n →∞
if and only if npmG →∞ and n2(1− p) →∞.Proof: By the method of moments (details omitted).
MAA 2005, Atlanta – p. 14/49
![Page 61: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/61.jpg)
Beyond the threshold
npmG →∞Question 1: Asymptotic distribution of XG
Question 2: The rate of decay of P(XG = 0)
Theorem (Rucinski (1988)) For every graph G witheG > 0,
XG − EXG√
var(XG)→ N (0, 1) as n →∞
if and only if npmG →∞ and n2(1− p) →∞.
Proof: By the method of moments (details omitted).
MAA 2005, Atlanta – p. 14/49
![Page 62: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/62.jpg)
Beyond the threshold
npmG →∞Question 1: Asymptotic distribution of XG
Question 2: The rate of decay of P(XG = 0)
Theorem (Rucinski (1988)) For every graph G witheG > 0,
XG − EXG√
var(XG)→ N (0, 1) as n →∞
if and only if npmG →∞ and n2(1− p) →∞.Proof: By the method of moments (details omitted).
MAA 2005, Atlanta – p. 14/49
![Page 63: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/63.jpg)
FKG-inequality
P(XG = 0) ≥ maxH⊆G
P(XH = 0)
By FKG
P(XH = 0) ≥N(n,H)∏
i=1
P(Ii = 0) = (1− peH)N(n,H)
Finally, with ΨH = nvHpeH and p = p(n) < c < 1,
P(XG = 0) ≥ maxH⊆G
exp
−EXH
1− p
= exp
−Θ
(
minH⊆G
ΨH
)
MAA 2005, Atlanta – p. 15/49
![Page 64: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/64.jpg)
FKG-inequality
P(XG = 0) ≥ maxH⊆G
P(XH = 0)
By FKG
P(XH = 0) ≥N(n,H)∏
i=1
P(Ii = 0) = (1− peH)N(n,H)
Finally, with ΨH = nvHpeH and p = p(n) < c < 1,
P(XG = 0) ≥ maxH⊆G
exp
−EXH
1− p
= exp
−Θ
(
minH⊆G
ΨH
)
MAA 2005, Atlanta – p. 15/49
![Page 65: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/65.jpg)
FKG-inequality
P(XG = 0) ≥ maxH⊆G
P(XH = 0)
By FKG
P(XH = 0) ≥N(n,H)∏
i=1
P(Ii = 0) = (1− peH)N(n,H)
Finally, with ΨH = nvHpeH and p = p(n) < c < 1,
P(XG = 0) ≥ maxH⊆G
exp
−EXH
1− p
= exp
−Θ
(
minH⊆G
ΨH
)
MAA 2005, Atlanta – p. 15/49
![Page 66: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/66.jpg)
FKG-inequality
P(XG = 0) ≥ maxH⊆G
P(XH = 0)
By FKG
P(XH = 0) ≥N(n,H)∏
i=1
P(Ii = 0) = (1− peH)N(n,H)
Finally, with ΨH = nvHpeH and p = p(n) < c < 1,
P(XG = 0) ≥ maxH⊆G
exp
−EXH
1− p
= exp
−Θ
(
minH⊆G
ΨH
)
MAA 2005, Atlanta – p. 15/49
![Page 67: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/67.jpg)
Random subsets
Γ - finite set, Γp - a random binomial subset of Γ (eachelement included independently with probability p),
S - family of subsets of Γ,for each A ∈ S , IA is theindicator of A in Γp,
X =∑
A∈SIA
By FKG, P(X = 0) ≥ exp−EX/(1− p).
Example. Γ =(
[n]2
)
, Γp = G(n, p), S – all copies of G inKn, X = XG.
MAA 2005, Atlanta – p. 16/49
![Page 68: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/68.jpg)
Random subsets
Γ - finite set, Γp - a random binomial subset of Γ (eachelement included independently with probability p),S - family of subsets of Γ,
for each A ∈ S , IA is theindicator of A in Γp,
X =∑
A∈SIA
By FKG, P(X = 0) ≥ exp−EX/(1− p).
Example. Γ =(
[n]2
)
, Γp = G(n, p), S – all copies of G inKn, X = XG.
MAA 2005, Atlanta – p. 16/49
![Page 69: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/69.jpg)
Random subsets
Γ - finite set, Γp - a random binomial subset of Γ (eachelement included independently with probability p),S - family of subsets of Γ, for each A ∈ S , IA is theindicator of A in Γp,
X =∑
A∈SIA
By FKG, P(X = 0) ≥ exp−EX/(1− p).
Example. Γ =(
[n]2
)
, Γp = G(n, p), S – all copies of G inKn, X = XG.
MAA 2005, Atlanta – p. 16/49
![Page 70: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/70.jpg)
Random subsets
Γ - finite set, Γp - a random binomial subset of Γ (eachelement included independently with probability p),S - family of subsets of Γ, for each A ∈ S , IA is theindicator of A in Γp,
X =∑
A∈SIA
By FKG, P(X = 0) ≥ exp−EX/(1− p).
Example. Γ =(
[n]2
)
, Γp = G(n, p), S – all copies of G inKn, X = XG.
MAA 2005, Atlanta – p. 16/49
![Page 71: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/71.jpg)
Random subsets
Γ - finite set, Γp - a random binomial subset of Γ (eachelement included independently with probability p),S - family of subsets of Γ, for each A ∈ S , IA is theindicator of A in Γp,
X =∑
A∈SIA
By FKG, P(X = 0) ≥ exp−EX/(1− p).
Example. Γ =(
[n]2
)
, Γp = G(n, p), S – all copies of G inKn, X = XG.
MAA 2005, Atlanta – p. 16/49
![Page 72: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/72.jpg)
Random subsets
Γ - finite set, Γp - a random binomial subset of Γ (eachelement included independently with probability p),S - family of subsets of Γ, for each A ∈ S , IA is theindicator of A in Γp,
X =∑
A∈SIA
By FKG, P(X = 0) ≥ exp−EX/(1− p).
Example. Γ =(
[n]2
)
, Γp = G(n, p), S – all copies of G inKn, X = XG.
MAA 2005, Atlanta – p. 16/49
![Page 73: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/73.jpg)
The Janson inequality
λ = EX, ∆ =∑ ∑
A∩B 6=∅E(IAIB)
φ(x) = (1 + x) log(1 + x)− x, x ≥ −1
Theorem (Janson,1990) For all 0 ≤ t ≤ λ
P(X ≤ λ− t) ≤ exp
−φ(−t/λ)λ2
∆
≤ exp
− t2
2∆
Proof: by Laplace transforms, FKG and Jenseninequalities (omitted).Note: for disjoint A’s, IA’s are independent and we getthe (lower tail) Chernoff bound.
MAA 2005, Atlanta – p. 17/49
![Page 74: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/74.jpg)
The Janson inequality
λ = EX, ∆ =∑ ∑
A∩B 6=∅E(IAIB)
φ(x) = (1 + x) log(1 + x)− x, x ≥ −1
Theorem (Janson,1990) For all 0 ≤ t ≤ λ
P(X ≤ λ− t) ≤ exp
−φ(−t/λ)λ2
∆
≤ exp
− t2
2∆
Proof: by Laplace transforms, FKG and Jenseninequalities (omitted).Note: for disjoint A’s, IA’s are independent and we getthe (lower tail) Chernoff bound.
MAA 2005, Atlanta – p. 17/49
![Page 75: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/75.jpg)
The Janson inequality
λ = EX, ∆ =∑ ∑
A∩B 6=∅E(IAIB)
φ(x) = (1 + x) log(1 + x)− x, x ≥ −1
Theorem (Janson,1990) For all 0 ≤ t ≤ λ
P(X ≤ λ− t) ≤ exp
−φ(−t/λ)λ2
∆
≤ exp
− t2
2∆
Proof: by Laplace transforms, FKG and Jenseninequalities (omitted).Note: for disjoint A’s, IA’s are independent and we getthe (lower tail) Chernoff bound.
MAA 2005, Atlanta – p. 17/49
![Page 76: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/76.jpg)
The Janson inequality
λ = EX, ∆ =∑ ∑
A∩B 6=∅E(IAIB)
φ(x) = (1 + x) log(1 + x)− x, x ≥ −1
Theorem (Janson,1990) For all 0 ≤ t ≤ λ
P(X ≤ λ− t) ≤ exp
−φ(−t/λ)λ2
∆
≤ exp
− t2
2∆
Proof: by Laplace transforms, FKG and Jenseninequalities (omitted).
Note: for disjoint A’s, IA’s are independent and we getthe (lower tail) Chernoff bound.
MAA 2005, Atlanta – p. 17/49
![Page 77: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/77.jpg)
The Janson inequality
λ = EX, ∆ =∑ ∑
A∩B 6=∅E(IAIB)
φ(x) = (1 + x) log(1 + x)− x, x ≥ −1
Theorem (Janson,1990) For all 0 ≤ t ≤ λ
P(X ≤ λ− t) ≤ exp
−φ(−t/λ)λ2
∆
≤ exp
− t2
2∆
Proof: by Laplace transforms, FKG and Jenseninequalities (omitted).Note: for disjoint A’s, IA’s are independent and we getthe (lower tail) Chernoff bound. MAA 2005, Atlanta – p. 17/49
![Page 78: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/78.jpg)
The rate of decay of P(X = 0)
Corollary (Janson, Łuczak, Rucinski, 1990)
P(X = 0) ≤ exp
−λ2
∆
Proof: (Weaker version – Boppana, Spencer, 1989): let
∆ =1
2
∑
A6=B
∑
A∩B 6=∅E(IAIB),
thus ∆ = λ + 2∆.
MAA 2005, Atlanta – p. 18/49
![Page 79: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/79.jpg)
The rate of decay of P(X = 0)
Corollary (Janson, Łuczak, Rucinski, 1990)
P(X = 0) ≤ exp
−λ2
∆
Proof: (Weaker version – Boppana, Spencer, 1989):
let
∆ =1
2
∑
A6=B
∑
A∩B 6=∅E(IAIB),
thus ∆ = λ + 2∆.
MAA 2005, Atlanta – p. 18/49
![Page 80: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/80.jpg)
The rate of decay of P(X = 0)
Corollary (Janson, Łuczak, Rucinski, 1990)
P(X = 0) ≤ exp
−λ2
∆
Proof: (Weaker version – Boppana, Spencer, 1989): let
∆ =1
2
∑
A6=B
∑
A∩B 6=∅E(IAIB),
thus ∆ = λ + 2∆.
MAA 2005, Atlanta – p. 18/49
![Page 81: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/81.jpg)
The rate of decay of P(X = 0)
Corollary (Janson, Łuczak, Rucinski, 1990)
P(X = 0) ≤ exp
−λ2
∆
Proof: (Weaker version – Boppana, Spencer, 1989): let
∆ =1
2
∑
A6=B
∑
A∩B 6=∅E(IAIB),
thus ∆ = λ + 2∆.
MAA 2005, Atlanta – p. 18/49
![Page 82: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/82.jpg)
Proof of weaker version
First show: P(X = 0) ≤ exp −λ + ∆.
Enumerate S = A1, . . . , Ak, denoteBi = Γp ⊇ Ai.Then
λ =k∑
i=1
P(Bi) and ∆ =1
2
∑
i∼j
∑
i6=j
P(Bi ∩Bj)
By the chain formula
P(X = 0) = P
(
k⋂
i=1
Bi
)
=k∏
i=1
P
(
Bi |i−1⋂
j=1
Bj
)
MAA 2005, Atlanta – p. 19/49
![Page 83: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/83.jpg)
Proof of weaker version
First show: P(X = 0) ≤ exp −λ + ∆.
Enumerate S = A1, . . . , Ak, denoteBi = Γp ⊇ Ai.
Then
λ =k∑
i=1
P(Bi) and ∆ =1
2
∑
i∼j
∑
i6=j
P(Bi ∩Bj)
By the chain formula
P(X = 0) = P
(
k⋂
i=1
Bi
)
=k∏
i=1
P
(
Bi |i−1⋂
j=1
Bj
)
MAA 2005, Atlanta – p. 19/49
![Page 84: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/84.jpg)
Proof of weaker version
First show: P(X = 0) ≤ exp −λ + ∆.
Enumerate S = A1, . . . , Ak, denote Bi = Γp ⊇ Ai.Then
λ =k∑
i=1
P(Bi) and ∆ =1
2
∑
i∼j
∑
i6=j
P(Bi ∩Bj)
By the chain formula
P(X = 0) = P
(
k⋂
i=1
Bi
)
=k∏
i=1
P
(
Bi |i−1⋂
j=1
Bj
)
MAA 2005, Atlanta – p. 19/49
![Page 85: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/85.jpg)
Proof of weaker version
First show: P(X = 0) ≤ exp −λ + ∆.
Enumerate S = A1, . . . , Ak, denote Bi = Γp ⊇ Ai.Then
λ =k∑
i=1
P(Bi) and ∆ =1
2
∑
i∼j
∑
i6=j
P(Bi ∩Bj)
By the chain formula
P(X = 0) = P
(
k⋂
i=1
Bi
)
=k∏
i=1
P
(
Bi |i−1⋂
j=1
Bj
)
MAA 2005, Atlanta – p. 19/49
![Page 86: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/86.jpg)
Probability calculus
Notation: For i 6= j, i ∼ j if Ai ∩ Aj 6= ∅, that is, if Bi
and Bj are dependent.
P
(
Bi |i−1⋂
j=1
Bj
)
≥ P
Bi ∩⋂
j∼i
Bj |⋂
j 6∼i
Bj
≥
P(
Bi |⋂
j 6∼i Bj
)
−P
(
Bi ∩⋃
j∼i Bj |⋂
j 6∼i Bj
)
≥P(Bi)−
∑
j∼i P(Bi ∩Bj) by FKG
MAA 2005, Atlanta – p. 20/49
![Page 87: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/87.jpg)
Probability calculus
Notation: For i 6= j, i ∼ j if Ai ∩ Aj 6= ∅, that is, if Bi
and Bj are dependent.
P
(
Bi |i−1⋂
j=1
Bj
)
≥ P
Bi ∩⋂
j∼i
Bj |⋂
j 6∼i
Bj
≥
P
Bi |⋂
j 6∼i
Bj
−P
Bi ∩⋃
j∼i
Bj |⋂
j 6∼i
Bj
≥
P(Bi)−∑
j∼i P(Bi ∩Bj) by FKG
MAA 2005, Atlanta – p. 20/49
![Page 88: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/88.jpg)
Probability calculus
Notation: For i 6= j, i ∼ j if Ai ∩ Aj 6= ∅, that is, if Bi
and Bj are dependent.
P
(
Bi |i−1⋂
j=1
Bj
)
≥ P
Bi ∩⋂
j∼i
Bj |⋂
j 6∼i
Bj
≥
P
Bi |⋂
j 6∼i
Bj
−P
Bi ∩⋃
j∼i
Bj |⋂
j 6∼i
Bj
≥
P(Bi)−∑
j∼i
P(Bi ∩Bj) by FKG
MAA 2005, Atlanta – p. 20/49
![Page 89: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/89.jpg)
Probability calculus
Notation: For i 6= j, i ∼ j if Ai ∩ Aj 6= ∅, that is, if Bi
and Bj are dependent.
P
(
Bi |i−1⋂
j=1
Bj
)
≥ P
Bi ∩⋂
j∼i
Bj |⋂
j 6∼i
Bj
≥
P
Bi |⋂
j 6∼i
Bj
−P
Bi ∩⋃
j∼i
Bj |⋂
j 6∼i
Bj
≥
P(Bi)−∑
j∼i
P(Bi ∩Bj) by FKG
MAA 2005, Atlanta – p. 20/49
![Page 90: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/90.jpg)
Putting together
P
(
k⋂
i=1
Bi
)
≤k∏
i=1
(
1−P(Bi) +∑
j∼i,j<i
P(Bi ∩Bj)
)
≤
exp−λ + ∆ ≤ exp
− λ2
2(λ + 2∆)
provided λ ≥ 2∆. Otherwise ...
Note that above is true for any subset of indices from [k].
MAA 2005, Atlanta – p. 21/49
![Page 91: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/91.jpg)
Putting together
P
(
k⋂
i=1
Bi
)
≤k∏
i=1
(
1−P(Bi) +∑
j∼i,j<i
P(Bi ∩Bj)
)
≤
exp−λ + ∆ ≤ exp
− λ2
2(λ + 2∆)
provided λ ≥ 2∆.
Otherwise ...
Note that above is true for any subset of indices from [k].
MAA 2005, Atlanta – p. 21/49
![Page 92: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/92.jpg)
Putting together
P
(
k⋂
i=1
Bi
)
≤k∏
i=1
(
1−P(Bi) +∑
j∼i,j<i
P(Bi ∩Bj)
)
≤
exp−λ + ∆ ≤ exp
− λ2
2(λ + 2∆)
provided λ ≥ 2∆. Otherwise ...
Note that above is true for any subset of indices from [k].
MAA 2005, Atlanta – p. 21/49
![Page 93: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/93.jpg)
Putting together
P
(
k⋂
i=1
Bi
)
≤k∏
i=1
(
1−P(Bi) +∑
j∼i,j<i
P(Bi ∩Bj)
)
≤
exp−λ + ∆ ≤ exp
− λ2
2(λ + 2∆)
provided λ ≥ 2∆. Otherwise ...
Note that above is true for any subset of indices from [k].
MAA 2005, Atlanta – p. 21/49
![Page 94: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/94.jpg)
The probabilistic method
Set
q =λ
2∆
and R = [k]q.
Let Ii = 1 if i ∈ R and Ii = 0otherwise.Let
Y = − lnP
(
⋂
i∈R
Bi
)
and
Z =k∑
i=1
P(Bi)Ii −1
2
∑
i∼j
∑
i6=j
P(Bi ∩Bj)IiIj
MAA 2005, Atlanta – p. 22/49
![Page 95: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/95.jpg)
The probabilistic method
Set
q =λ
2∆
and R = [k]q. Let Ii = 1 if i ∈ R and Ii = 0otherwise.
Let
Y = − lnP
(
⋂
i∈R
Bi
)
and
Z =k∑
i=1
P(Bi)Ii −1
2
∑
i∼j
∑
i6=j
P(Bi ∩Bj)IiIj
MAA 2005, Atlanta – p. 22/49
![Page 96: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/96.jpg)
The probabilistic method
Set
q =λ
2∆
and R = [k]q. Let Ii = 1 if i ∈ R and Ii = 0 otherwise.Let
Y = − lnP
(
⋂
i∈R
Bi
)
and
Z =k∑
i=1
P(Bi)Ii −1
2
∑
i∼j
∑
i6=j
P(Bi ∩Bj)IiIj
MAA 2005, Atlanta – p. 22/49
![Page 97: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/97.jpg)
The probabilistic method
Set
q =λ
2∆
and R = [k]q. Let Ii = 1 if i ∈ R and Ii = 0 otherwise.Let
Y = − lnP
(
⋂
i∈R
Bi
)
and
Z =k∑
i=1
P(Bi)Ii −1
2
∑
i∼j
∑
i6=j
P(Bi ∩Bj)IiIj
MAA 2005, Atlanta – p. 22/49
![Page 98: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/98.jpg)
The probabilistic method – cont.
Thus
Y ≥ Z and EY ≥ EZ
= λq −∆q2 =λ2
4∆
So, there is S ⊆ [k] such that
Y (S) = − lnP
(
⋂
i∈S
Bi
)
≥ λ2
4∆and
P
(
k⋂
i=1
Bi
)
≤ exp
− λ2
4∆
≤ exp
− λ2
2(λ + 2∆)
MAA 2005, Atlanta – p. 23/49
![Page 99: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/99.jpg)
The probabilistic method – cont.
Thus
Y ≥ Z and EY ≥ EZ = λq −∆q2 =λ2
4∆
So, there is S ⊆ [k] such that
Y (S) = − lnP
(
⋂
i∈S
Bi
)
≥ λ2
4∆and
P
(
k⋂
i=1
Bi
)
≤ exp
− λ2
4∆
≤ exp
− λ2
2(λ + 2∆)
MAA 2005, Atlanta – p. 23/49
![Page 100: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/100.jpg)
The probabilistic method – cont.
Thus
Y ≥ Z and EY ≥ EZ = λq −∆q2 =λ2
4∆
So, there is S ⊆ [k] such that
Y (S) = − lnP
(
⋂
i∈S
Bi
)
≥ λ2
4∆
and
P
(
k⋂
i=1
Bi
)
≤ exp
− λ2
4∆
≤ exp
− λ2
2(λ + 2∆)
MAA 2005, Atlanta – p. 23/49
![Page 101: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/101.jpg)
The probabilistic method – cont.
Thus
Y ≥ Z and EY ≥ EZ = λq −∆q2 =λ2
4∆
So, there is S ⊆ [k] such that
Y (S) = − lnP
(
⋂
i∈S
Bi
)
≥ λ2
4∆and
P
(
k⋂
i=1
Bi
)
≤ exp
− λ2
4∆
≤ exp
− λ2
2(λ + 2∆)
MAA 2005, Atlanta – p. 23/49
![Page 102: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/102.jpg)
The probabilistic method – cont.
Thus
Y ≥ Z and EY ≥ EZ = λq −∆q2 =λ2
4∆
So, there is S ⊆ [k] such that
Y (S) = − lnP
(
⋂
i∈S
Bi
)
≥ λ2
4∆and
P
(
k⋂
i=1
Bi
)
≤ exp
− λ2
4∆
≤ exp
− λ2
2(λ + 2∆)
MAA 2005, Atlanta – p. 23/49
![Page 103: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/103.jpg)
Back to subgraphs
∆ =∑
H⊆G
∑
Gi∩Gj=H
p2eG−eH
= Θ
(
(EXG)2
minH ΨH
)
so
P(XG = 0) = exp
−Θ
(
minH⊆G
ΨH
)
(recall ΨH = nvHpeH )
MAA 2005, Atlanta – p. 24/49
![Page 104: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/104.jpg)
Back to subgraphs
∆ =∑
H⊆G
∑
Gi∩Gj=H
p2eG−eH = Θ
(
(EXG)2
minH ΨH
)
so
P(XG = 0) = exp
−Θ
(
minH⊆G
ΨH
)
(recall ΨH = nvHpeH )
MAA 2005, Atlanta – p. 24/49
![Page 105: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/105.jpg)
Back to subgraphs
∆ =∑
H⊆G
∑
Gi∩Gj=H
p2eG−eH = Θ
(
(EXG)2
minH ΨH
)
so
P(XG = 0) = exp
−Θ
(
minH⊆G
ΨH
)
(recall ΨH = nvHpeH )
MAA 2005, Atlanta – p. 24/49
![Page 106: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/106.jpg)
Back to subgraphs
∆ =∑
H⊆G
∑
Gi∩Gj=H
p2eG−eH = Θ
(
(EXG)2
minH ΨH
)
so
P(XG = 0) = exp
−Θ
(
minH⊆G
ΨH
)
(recall ΨH = nvHpeH )
MAA 2005, Atlanta – p. 24/49
![Page 107: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/107.jpg)
Two milestones
minH
ΨH = Ω(n) iff p = Ω(
n−1/m(1)G
)
,
wherem
(1)G = max
H⊆G
eH
vH − 1
minH
ΨH = Ω(n2p) iff p = Ω(
n−1/m(2)G
)
,
where
m(2)G = max
H⊆G
eH − 1
vH − 2
MAA 2005, Atlanta – p. 25/49
![Page 108: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/108.jpg)
Two milestones
minH
ΨH = Ω(n) iff p = Ω(
n−1/m(1)G
)
,
wherem
(1)G = max
H⊆G
eH
vH − 1
minH
ΨH = Ω(n2p) iff p = Ω(
n−1/m(2)G
)
,
where
m(2)G = max
H⊆G
eH − 1
vH − 2
MAA 2005, Atlanta – p. 25/49
![Page 109: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/109.jpg)
Two milestones
minH
ΨH = Ω(n) iff p = Ω(
n−1/m(1)G
)
,
wherem
(1)G = max
H⊆G
eH
vH − 1
minH
ΨH = Ω(n2p) iff p = Ω(
n−1/m(2)G
)
,
where
m(2)G = max
H⊆G
eH − 1
vH − 2
MAA 2005, Atlanta – p. 25/49
![Page 110: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/110.jpg)
Two milestones
minH
ΨH = Ω(n) iff p = Ω(
n−1/m(1)G
)
,
wherem
(1)G = max
H⊆G
eH
vH − 1
minH
ΨH = Ω(n2p) iff p = Ω(
n−1/m(2)G
)
,
where
m(2)G = max
H⊆G
eH − 1
vH − 2
MAA 2005, Atlanta – p. 25/49
![Page 111: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/111.jpg)
Almost perfect G-factors
Set G = K3, m(1)K3
= 3/2.
Proposition Fix ε > 0 and let p ≥ Cεn−2/3.Then, a.a.s.,
all but at most εn vertices of G(n, p) can be covered byvertex-disjoint triangles.Proof: The probability of opposite event can bebounded by
P(there is an εn-subset with no triangle) ≤(
n
εn
)
P(G(εn, p) 6⊃ K3) < 2nP(XK3
(εn, p) = 0) ≤
2ne−Θ(n) → 0
MAA 2005, Atlanta – p. 26/49
![Page 112: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/112.jpg)
Almost perfect G-factors
Set G = K3, m(1)K3
= 3/2.
Proposition Fix ε > 0 and let p ≥ Cεn−2/3.
Then, a.a.s.,all but at most εn vertices of G(n, p) can be covered byvertex-disjoint triangles.Proof: The probability of opposite event can bebounded by
P(there is an εn-subset with no triangle) ≤(
n
εn
)
P(G(εn, p) 6⊃ K3) < 2nP(XK3
(εn, p) = 0) ≤
2ne−Θ(n) → 0
MAA 2005, Atlanta – p. 26/49
![Page 113: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/113.jpg)
Almost perfect G-factors
Set G = K3, m(1)K3
= 3/2.
Proposition Fix ε > 0 and let p ≥ Cεn−2/3. Then,
a.a.s., all but at most εn vertices of G(n, p) can becovered by vertex-disjoint triangles.
Proof: The probability of opposite event can bebounded by
P(there is an εn-subset with no triangle) ≤(
n
εn
)
P(G(εn, p) 6⊃ K3) < 2nP(XK3
(εn, p) = 0) ≤
2ne−Θ(n) → 0
MAA 2005, Atlanta – p. 26/49
![Page 114: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/114.jpg)
Almost perfect G-factors
Set G = K3, m(1)K3
= 3/2.
Proposition Fix ε > 0 and let p ≥ Cεn−2/3. Then,
a.a.s., all but at most εn vertices of G(n, p) can becovered by vertex-disjoint triangles.Proof: The probability of opposite event can bebounded by
P(there is an εn-subset with no triangle) ≤(
n
εn
)
P(G(εn, p) 6⊃ K3) < 2nP(XK3
(εn, p) = 0) ≤
2ne−Θ(n) → 0
MAA 2005, Atlanta – p. 26/49
![Page 115: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/115.jpg)
Almost perfect G-factors
Set G = K3, m(1)K3
= 3/2.
Proposition Fix ε > 0 and let p ≥ Cεn−2/3. Then,
a.a.s., all but at most εn vertices of G(n, p) can becovered by vertex-disjoint triangles.Proof: The probability of opposite event can bebounded by
P(there is an εn-subset with no triangle) ≤
(
n
εn
)
P(G(εn, p) 6⊃ K3) < 2nP(XK3
(εn, p) = 0) ≤
2ne−Θ(n) → 0
MAA 2005, Atlanta – p. 26/49
![Page 116: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/116.jpg)
Almost perfect G-factors
Set G = K3, m(1)K3
= 3/2.
Proposition Fix ε > 0 and let p ≥ Cεn−2/3. Then,
a.a.s., all but at most εn vertices of G(n, p) can becovered by vertex-disjoint triangles.Proof: The probability of opposite event can bebounded by
P(there is an εn-subset with no triangle) ≤(
n
εn
)
P(G(εn, p) 6⊃ K3) < 2nP(XK3
(εn, p) = 0) ≤
2ne−Θ(n) → 0
MAA 2005, Atlanta – p. 26/49
![Page 117: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/117.jpg)
Almost perfect G-factors
Set G = K3, m(1)K3
= 3/2.
Proposition Fix ε > 0 and let p ≥ Cεn−2/3. Then,
a.a.s., all but at most εn vertices of G(n, p) can becovered by vertex-disjoint triangles.Proof: The probability of opposite event can bebounded by
P(there is an εn-subset with no triangle) ≤(
n
εn
)
P(G(εn, p) 6⊃ K3) < 2nP(XK3
(εn, p) = 0) ≤
2ne−Θ(n) → 0
MAA 2005, Atlanta – p. 26/49
![Page 118: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/118.jpg)
Open problem
Find the threshold for the existence of a perfecttriangle-factor in G(n, p).
Conjecture The threshold is
p0 = n−2/3 log1/3 n
Krivelevich: p0 ≤ n−3/5 (Ω(n) copies of the diamond).
Kim: even better
Alon-Yuster, Rucinski: p0 = n−2/3 is the threshold for aperfect (K4 −K1,2)-factor.
MAA 2005, Atlanta – p. 27/49
![Page 119: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/119.jpg)
Open problem
Find the threshold for the existence of a perfecttriangle-factor in G(n, p).
Conjecture The threshold is
p0 = n−2/3 log1/3 n
Krivelevich: p0 ≤ n−3/5 (Ω(n) copies of the diamond).
Kim: even better
Alon-Yuster, Rucinski: p0 = n−2/3 is the threshold for aperfect (K4 −K1,2)-factor.
MAA 2005, Atlanta – p. 27/49
![Page 120: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/120.jpg)
Open problem
Find the threshold for the existence of a perfecttriangle-factor in G(n, p).
Conjecture The threshold is
p0 = n−2/3 log1/3 n
Krivelevich: p0 ≤ n−3/5 (Ω(n) copies of the diamond).
Kim: even better
Alon-Yuster, Rucinski: p0 = n−2/3 is the threshold for aperfect (K4 −K1,2)-factor.
MAA 2005, Atlanta – p. 27/49
![Page 121: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/121.jpg)
Open problem
Find the threshold for the existence of a perfecttriangle-factor in G(n, p).
Conjecture The threshold is
p0 = n−2/3 log1/3 n
Krivelevich: p0 ≤ n−3/5 (Ω(n) copies of the diamond).
Kim: even better
Alon-Yuster, Rucinski: p0 = n−2/3 is the threshold for aperfect (K4 −K1,2)-factor.
MAA 2005, Atlanta – p. 27/49
![Page 122: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/122.jpg)
Open problem
Find the threshold for the existence of a perfecttriangle-factor in G(n, p).
Conjecture The threshold is
p0 = n−2/3 log1/3 n
Krivelevich: p0 ≤ n−3/5 (Ω(n) copies of the diamond).
Kim: even better
Alon-Yuster, Rucinski: p0 = n−2/3 is the threshold for aperfect (K4 −K1,2)-factor.
MAA 2005, Atlanta – p. 27/49
![Page 123: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/123.jpg)
Vertex-partition properties
Arrow notation: F → (G)vr means that every r-coloring
of the vertices of F results in a monochromatic copy ofG.
Theorem (Łuczak, Rucinski, Voigt (1992)) For every
r ≥ 2, p0 = n−1/m(1)G is the threshold for the property
G(n, p) → (G)vr .
Proof: (the easy part)
P (G(n, p) 6→ (G)vr) ≤
(
n
dn/re
)
P (G(dn/re, p) 6⊃ G) .
Note: this threshold is sharp (Friedgut, Krivelevich,1999)
MAA 2005, Atlanta – p. 28/49
![Page 124: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/124.jpg)
Vertex-partition properties
Arrow notation: F → (G)vr means that every r-coloring
of the vertices of F results in a monochromatic copy ofG.Theorem (Łuczak, Rucinski, Voigt (1992)) For every
r ≥ 2, p0 = n−1/m(1)G is the threshold for the property
G(n, p) → (G)vr .
Proof: (the easy part)
P (G(n, p) 6→ (G)vr) ≤
(
n
dn/re
)
P (G(dn/re, p) 6⊃ G) .
Note: this threshold is sharp (Friedgut, Krivelevich,1999)
MAA 2005, Atlanta – p. 28/49
![Page 125: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/125.jpg)
Vertex-partition properties
Arrow notation: F → (G)vr means that every r-coloring
of the vertices of F results in a monochromatic copy ofG.Theorem (Łuczak, Rucinski, Voigt (1992)) For every
r ≥ 2, p0 = n−1/m(1)G is the threshold for the property
G(n, p) → (G)vr .
Proof: (the easy part)
P (G(n, p) 6→ (G)vr) ≤
(
n
dn/re
)
P (G(dn/re, p) 6⊃ G) .
Note: this threshold is sharp (Friedgut, Krivelevich,1999)
MAA 2005, Atlanta – p. 28/49
![Page 126: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/126.jpg)
Vertex-partition properties
Arrow notation: F → (G)vr means that every r-coloring
of the vertices of F results in a monochromatic copy ofG.Theorem (Łuczak, Rucinski, Voigt (1992)) For every
r ≥ 2, p0 = n−1/m(1)G is the threshold for the property
G(n, p) → (G)vr .
Proof: (the easy part)
P (G(n, p) 6→ (G)vr) ≤
(
n
dn/re
)
P (G(dn/re, p) 6⊃ G) .
Note: this threshold is sharp (Friedgut, Krivelevich,1999)
MAA 2005, Atlanta – p. 28/49
![Page 127: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/127.jpg)
Vertex-partition properties
Arrow notation: F → (G)vr means that every r-coloring
of the vertices of F results in a monochromatic copy ofG.Theorem (Łuczak, Rucinski, Voigt (1992)) For every
r ≥ 2, p0 = n−1/m(1)G is the threshold for the property
G(n, p) → (G)vr .
Proof: (the easy part)
P (G(n, p) 6→ (G)vr) ≤
(
n
dn/re
)
P (G(dn/re, p) 6⊃ G) .
Note: this threshold is sharp (Friedgut, Krivelevich,1999)
MAA 2005, Atlanta – p. 28/49
![Page 128: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/128.jpg)
A question of Erdos
Pigeon-hole: K5 → (K3)v2.
Easy: K1 + C27 → (K3)
v2. Note: K1 + C2
7 6⊃ K5.
Erdos: Does there exist an F such that F 6⊃ K4 andF → (K3)
v2?
YES!!! (Erdos, Rogers (1962) and Folkman (1970)).Set p = Cn−2/3, so thatP (G(n, p) → (K3)
v2) = 1− o(1).
We know: P(XK4= 0) ∼ e−λ ∼ c0 > 0
Switching to the model G(n, M): about c0
( (n2)
(C/2)n4/3
)
graphs with n vertices and M = (C/2)n4/3 edges aresuch.
MAA 2005, Atlanta – p. 29/49
![Page 129: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/129.jpg)
A question of Erdos
Pigeon-hole: K5 → (K3)v2.
Easy: K1 + C27 → (K3)
v2.
Note: K1 + C27 6⊃ K5.
Erdos: Does there exist an F such that F 6⊃ K4 andF → (K3)
v2?
YES!!! (Erdos, Rogers (1962) and Folkman (1970)).Set p = Cn−2/3, so thatP (G(n, p) → (K3)
v2) = 1− o(1).
We know: P(XK4= 0) ∼ e−λ ∼ c0 > 0
Switching to the model G(n, M): about c0
( (n2)
(C/2)n4/3
)
graphs with n vertices and M = (C/2)n4/3 edges aresuch.
MAA 2005, Atlanta – p. 29/49
![Page 130: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/130.jpg)
A question of Erdos
Pigeon-hole: K5 → (K3)v2.
Easy: K1 + C27 → (K3)
v2. Note: K1 + C2
7 6⊃ K5.
Erdos: Does there exist an F such that F 6⊃ K4 andF → (K3)
v2?
YES!!! (Erdos, Rogers (1962) and Folkman (1970)).Set p = Cn−2/3, so thatP (G(n, p) → (K3)
v2) = 1− o(1).
We know: P(XK4= 0) ∼ e−λ ∼ c0 > 0
Switching to the model G(n, M): about c0
( (n2)
(C/2)n4/3
)
graphs with n vertices and M = (C/2)n4/3 edges aresuch.
MAA 2005, Atlanta – p. 29/49
![Page 131: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/131.jpg)
A question of Erdos
Pigeon-hole: K5 → (K3)v2.
Easy: K1 + C27 → (K3)
v2. Note: K1 + C2
7 6⊃ K5.
Erdos: Does there exist an F such that F 6⊃ K4 andF → (K3)
v2?
YES!!! (Erdos, Rogers (1962) and Folkman (1970)).Set p = Cn−2/3, so thatP (G(n, p) → (K3)
v2) = 1− o(1).
We know: P(XK4= 0) ∼ e−λ ∼ c0 > 0
Switching to the model G(n, M): about c0
( (n2)
(C/2)n4/3
)
graphs with n vertices and M = (C/2)n4/3 edges aresuch.
MAA 2005, Atlanta – p. 29/49
![Page 132: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/132.jpg)
A question of Erdos
Pigeon-hole: K5 → (K3)v2.
Easy: K1 + C27 → (K3)
v2. Note: K1 + C2
7 6⊃ K5.
Erdos: Does there exist an F such that F 6⊃ K4 andF → (K3)
v2?
YES!!! (Erdos, Rogers (1962) and Folkman (1970)).
Set p = Cn−2/3, so thatP (G(n, p) → (K3)
v2) = 1− o(1).
We know: P(XK4= 0) ∼ e−λ ∼ c0 > 0
Switching to the model G(n, M): about c0
( (n2)
(C/2)n4/3
)
graphs with n vertices and M = (C/2)n4/3 edges aresuch.
MAA 2005, Atlanta – p. 29/49
![Page 133: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/133.jpg)
A question of Erdos
Pigeon-hole: K5 → (K3)v2.
Easy: K1 + C27 → (K3)
v2. Note: K1 + C2
7 6⊃ K5.
Erdos: Does there exist an F such that F 6⊃ K4 andF → (K3)
v2?
YES!!! (Erdos, Rogers (1962) and Folkman (1970)).Set p = Cn−2/3, so thatP (G(n, p) → (K3)
v2) = 1− o(1).
We know: P(XK4= 0) ∼ e−λ ∼ c0 > 0
Switching to the model G(n, M): about c0
( (n2)
(C/2)n4/3
)
graphs with n vertices and M = (C/2)n4/3 edges aresuch.
MAA 2005, Atlanta – p. 29/49
![Page 134: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/134.jpg)
A question of Erdos
Pigeon-hole: K5 → (K3)v2.
Easy: K1 + C27 → (K3)
v2. Note: K1 + C2
7 6⊃ K5.
Erdos: Does there exist an F such that F 6⊃ K4 andF → (K3)
v2?
YES!!! (Erdos, Rogers (1962) and Folkman (1970)).Set p = Cn−2/3, so thatP (G(n, p) → (K3)
v2) = 1− o(1).
We know: P(XK4= 0) ∼ e−λ ∼ c0 > 0
Switching to the model G(n, M): about c0
( (n2)
(C/2)n4/3
)
graphs with n vertices and M = (C/2)n4/3 edges aresuch.
MAA 2005, Atlanta – p. 29/49
![Page 135: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/135.jpg)
A question of Erdos
Pigeon-hole: K5 → (K3)v2.
Easy: K1 + C27 → (K3)
v2. Note: K1 + C2
7 6⊃ K5.
Erdos: Does there exist an F such that F 6⊃ K4 andF → (K3)
v2?
YES!!! (Erdos, Rogers (1962) and Folkman (1970)).Set p = Cn−2/3, so thatP (G(n, p) → (K3)
v2) = 1− o(1).
We know: P(XK4= 0) ∼ e−λ ∼ c0 > 0
Switching to the model G(n, M): about c0
( (n2)
(C/2)n4/3
)
graphs with n vertices and M = (C/2)n4/3 edges aresuch. MAA 2005, Atlanta – p. 29/49
![Page 136: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/136.jpg)
Ramsey properties
F → (G)er means that every r-coloring of the edges of F
results in a monochromatic copy of G.
Theorem (Rödl, Rucinski (1995)) For every r ≥ 2,
p0 = n−1/m(2)G is the threshold for the property
G(n, p) → (G)er.
Proof: see J. AMS (1995)
Theorem (Friedgut,Rödl, Rucinski,Tetali (2005)) Theproperty G(n, p) → (K3)
e2 has a sharp threshold.
Proof: 99 pages long proof omitted.
MAA 2005, Atlanta – p. 30/49
![Page 137: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/137.jpg)
Ramsey properties
F → (G)er means that every r-coloring of the edges of F
results in a monochromatic copy of G.
Theorem (Rödl, Rucinski (1995)) For every r ≥ 2,
p0 = n−1/m(2)G is the threshold for the property
G(n, p) → (G)er.
Proof: see J. AMS (1995)
Theorem (Friedgut,Rödl, Rucinski,Tetali (2005)) Theproperty G(n, p) → (K3)
e2 has a sharp threshold.
Proof: 99 pages long proof omitted.
MAA 2005, Atlanta – p. 30/49
![Page 138: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/138.jpg)
Ramsey properties
F → (G)er means that every r-coloring of the edges of F
results in a monochromatic copy of G.
Theorem (Rödl, Rucinski (1995)) For every r ≥ 2,
p0 = n−1/m(2)G is the threshold for the property
G(n, p) → (G)er.
Proof: see J. AMS (1995)
Theorem (Friedgut,Rödl, Rucinski,Tetali (2005)) Theproperty G(n, p) → (K3)
e2 has a sharp threshold.
Proof: 99 pages long proof omitted.
MAA 2005, Atlanta – p. 30/49
![Page 139: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/139.jpg)
Ramsey properties
F → (G)er means that every r-coloring of the edges of F
results in a monochromatic copy of G.
Theorem (Rödl, Rucinski (1995)) For every r ≥ 2,
p0 = n−1/m(2)G is the threshold for the property
G(n, p) → (G)er.
Proof: see J. AMS (1995)
Theorem (Friedgut,Rödl, Rucinski,Tetali (2005)) Theproperty G(n, p) → (K3)
e2 has a sharp threshold.
Proof: 99 pages long proof omitted.
MAA 2005, Atlanta – p. 30/49
![Page 140: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/140.jpg)
Ramsey properties
F → (G)er means that every r-coloring of the edges of F
results in a monochromatic copy of G.
Theorem (Rödl, Rucinski (1995)) For every r ≥ 2,
p0 = n−1/m(2)G is the threshold for the property
G(n, p) → (G)er.
Proof: see J. AMS (1995)
Theorem (Friedgut,Rödl, Rucinski,Tetali (2005)) Theproperty G(n, p) → (K3)
e2 has a sharp threshold.
Proof: 99 pages long proof omitted.
MAA 2005, Atlanta – p. 30/49
![Page 141: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/141.jpg)
Turán properties
ex(n, G) = maxeF : G 6⊆ F ⊆ Kn
=
(
1− 1
χ(G)− 1+ o(1)
)(
n
2
)
(Turán, Erdos, Stone, Simonovits 1941-1966)
If p n−1/m(2)G then minH ΨH n2pand
for some H ⊆ G, a.a.s XH n2p.It is then possible to destroy all copies of G by deletingo(n2p) edges– the Turán property does not hold forG(n, p) in this case.
MAA 2005, Atlanta – p. 31/49
![Page 142: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/142.jpg)
Turán properties
ex(n, G) = maxeF : G 6⊆ F ⊆ Kn
=
(
1− 1
χ(G)− 1+ o(1)
)(
n
2
)
(Turán, Erdos, Stone, Simonovits 1941-1966)
If p n−1/m(2)G then minH ΨH n2pand
for some H ⊆ G, a.a.s XH n2p.It is then possible to destroy all copies of G by deletingo(n2p) edges– the Turán property does not hold forG(n, p) in this case.
MAA 2005, Atlanta – p. 31/49
![Page 143: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/143.jpg)
Turán properties
ex(n, G) = maxeF : G 6⊆ F ⊆ Kn
=
(
1− 1
χ(G)− 1+ o(1)
)(
n
2
)
(Turán, Erdos, Stone, Simonovits 1941-1966)
If p n−1/m(2)G then minH ΨH n2pand
for some H ⊆ G, a.a.s XH n2p.It is then possible to destroy all copies of G by deletingo(n2p) edges– the Turán property does not hold forG(n, p) in this case.
MAA 2005, Atlanta – p. 31/49
![Page 144: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/144.jpg)
Turán properties
ex(n, G) = maxeF : G 6⊆ F ⊆ Kn
=
(
1− 1
χ(G)− 1+ o(1)
)(
n
2
)
(Turán, Erdos, Stone, Simonovits 1941-1966)
If p n−1/m(2)G then minH ΨH n2p
andfor some H ⊆ G, a.a.s XH n2p.It is then possible to destroy all copies of G by deletingo(n2p) edges– the Turán property does not hold forG(n, p) in this case.
MAA 2005, Atlanta – p. 31/49
![Page 145: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/145.jpg)
Turán properties
ex(n, G) = maxeF : G 6⊆ F ⊆ Kn
=
(
1− 1
χ(G)− 1+ o(1)
)(
n
2
)
(Turán, Erdos, Stone, Simonovits 1941-1966)
If p n−1/m(2)G then minH ΨH n2p and
for some H ⊆ G, a.a.s XH n2p.
It is then possible to destroy all copies of G by deletingo(n2p) edges– the Turán property does not hold forG(n, p) in this case.
MAA 2005, Atlanta – p. 31/49
![Page 146: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/146.jpg)
Turán properties
ex(n, G) = maxeF : G 6⊆ F ⊆ Kn
=
(
1− 1
χ(G)− 1+ o(1)
)(
n
2
)
(Turán, Erdos, Stone, Simonovits 1941-1966)
If p n−1/m(2)G then minH ΨH n2p and
for some H ⊆ G, a.a.s XH n2p.It is then possible to destroy all copies of G by deletingo(n2p) edges
– the Turán property does not hold forG(n, p) in this case.
MAA 2005, Atlanta – p. 31/49
![Page 147: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/147.jpg)
Turán properties
ex(n, G) = maxeF : G 6⊆ F ⊆ Kn
=
(
1− 1
χ(G)− 1+ o(1)
)(
n
2
)
(Turán, Erdos, Stone, Simonovits 1941-1966)
If p n−1/m(2)G then minH ΨH n2p and
for some H ⊆ G, a.a.s XH n2p.It is then possible to destroy all copies of G by deletingo(n2p) edges – the Turán property does not hold forG(n, p) in this case.
MAA 2005, Atlanta – p. 31/49
![Page 148: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/148.jpg)
Turán properties for G(n, p)
Conjecture For every η > 0 there is C > 0 such that if
p ≥ Cn−1/m(2)G then a.a.s. every subgraph of G(n, p) with
at least(
1− 1
χ(G)− 1+ η
)(
n
2
)
p
edges contains a copy of G.
True for G = K3, K4, K5, K6 and for all cycles G = Ck.(Frankl, Füredi, Gerke, Haxell, Kohayakawa, Kreuter,Łuczak, Rödl, Sabo, Schacht, Steger, Taraz, Vu, ...)
MAA 2005, Atlanta – p. 32/49
![Page 149: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/149.jpg)
Turán properties for G(n, p)
Conjecture For every η > 0 there is C > 0 such that if
p ≥ Cn−1/m(2)G then a.a.s. every subgraph of G(n, p) with
at least(
1− 1
χ(G)− 1+ η
)(
n
2
)
p
edges contains a copy of G.
True for G = K3, K4, K5, K6 and for all cycles G = Ck.
(Frankl, Füredi, Gerke, Haxell, Kohayakawa, Kreuter,Łuczak, Rödl, Sabo, Schacht, Steger, Taraz, Vu, ...)
MAA 2005, Atlanta – p. 32/49
![Page 150: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/150.jpg)
Turán properties for G(n, p)
Conjecture For every η > 0 there is C > 0 such that if
p ≥ Cn−1/m(2)G then a.a.s. every subgraph of G(n, p) with
at least(
1− 1
χ(G)− 1+ η
)(
n
2
)
p
edges contains a copy of G.
True for G = K3, K4, K5, K6 and for all cycles G = Ck.(Frankl, Füredi, Gerke, Haxell, Kohayakawa, Kreuter,Łuczak, Rödl, Sabo, Schacht, Steger, Taraz, Vu, ...)
MAA 2005, Atlanta – p. 32/49
![Page 151: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/151.jpg)
Upper tails - early results (Vu 2001)
P(XG ≥ tEXG)
For all balanced graphs G
P(XG ≥ tEXG) ≤ exp
−cεΨ1/(vG−1)G
For all graphs G
P(XG ≥ tEXG) ≥ exp
−CεΨ1/α∗GG log(1/p)
,
where α∗G is the fractional independence number of G.These bounds are far apart (they essentially match eachother only for stars K1,k).
MAA 2005, Atlanta – p. 33/49
![Page 152: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/152.jpg)
Upper tails - early results (Vu 2001)
P(XG ≥ tEXG)
For all balanced graphs G
P(XG ≥ tEXG) ≤ exp
−cεΨ1/(vG−1)G
For all graphs G
P(XG ≥ tEXG) ≥ exp
−CεΨ1/α∗GG log(1/p)
,
where α∗G is the fractional independence number of G.These bounds are far apart (they essentially match eachother only for stars K1,k).
MAA 2005, Atlanta – p. 33/49
![Page 153: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/153.jpg)
Upper tails - early results (Vu 2001)
P(XG ≥ tEXG)
For all balanced graphs G
P(XG ≥ tEXG) ≤ exp
−cεΨ1/(vG−1)G
For all graphs G
P(XG ≥ tEXG) ≥ exp
−CεΨ1/α∗GG log(1/p)
,
where α∗G is the fractional independence number of G.These bounds are far apart (they essentially match eachother only for stars K1,k).
MAA 2005, Atlanta – p. 33/49
![Page 154: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/154.jpg)
Upper tails - early results (Vu 2001)
P(XG ≥ tEXG)
For all balanced graphs G
P(XG ≥ tEXG) ≤ exp
−cεΨ1/(vG−1)G
For all graphs G
P(XG ≥ tEXG) ≥ exp
−CεΨ1/α∗GG log(1/p)
,
where α∗G is the fractional independence number of G.
These bounds are far apart (they essentially match eachother only for stars K1,k).
MAA 2005, Atlanta – p. 33/49
![Page 155: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/155.jpg)
Upper tails - early results (Vu 2001)
P(XG ≥ tEXG)
For all balanced graphs G
P(XG ≥ tEXG) ≤ exp
−cεΨ1/(vG−1)G
For all graphs G
P(XG ≥ tEXG) ≥ exp
−CεΨ1/α∗GG log(1/p)
,
where α∗G is the fractional independence number of G.These bounds are far apart (they essentially match eachother only for stars K1,k).
MAA 2005, Atlanta – p. 33/49
![Page 156: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/156.jpg)
Fractional independence number
α∗G is the largest value of∑
v αv over all weightingsαv ∈ [0, 1] of V (G) satisfying:
αv + αw ≤ 1 for all vw ∈ E(G)
Properties (assume eG > 0):
for regular G, α∗G = vG/2
for bipartite G, α∗G = αG
for all G,
1 ≤ 1
2vG ≤ α∗G ≤ vG −
eG
∆G≤ vG − 1
MAA 2005, Atlanta – p. 34/49
![Page 157: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/157.jpg)
Fractional independence number
α∗G is the largest value of∑
v αv over all weightingsαv ∈ [0, 1] of V (G) satisfying:
αv + αw ≤ 1 for all vw ∈ E(G)
Properties (assume eG > 0):
for regular G, α∗G = vG/2
for bipartite G, α∗G = αG
for all G,
1 ≤ 1
2vG ≤ α∗G ≤ vG −
eG
∆G≤ vG − 1
MAA 2005, Atlanta – p. 34/49
![Page 158: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/158.jpg)
Fractional independence number
α∗G is the largest value of∑
v αv over all weightingsαv ∈ [0, 1] of V (G) satisfying:
αv + αw ≤ 1 for all vw ∈ E(G)
Properties (assume eG > 0):
for regular G, α∗G = vG/2
for bipartite G, α∗G = αG
for all G,
1 ≤ 1
2vG ≤ α∗G ≤ vG −
eG
∆G≤ vG − 1
MAA 2005, Atlanta – p. 34/49
![Page 159: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/159.jpg)
Fractional independence number
α∗G is the largest value of∑
v αv over all weightingsαv ∈ [0, 1] of V (G) satisfying:
αv + αw ≤ 1 for all vw ∈ E(G)
Properties (assume eG > 0):
for regular G, α∗G = vG/2
for bipartite G, α∗G = αG
for all G,
1 ≤ 1
2vG ≤ α∗G ≤ vG −
eG
∆G≤ vG − 1
MAA 2005, Atlanta – p. 34/49
![Page 160: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/160.jpg)
Fractional independence number
α∗G is the largest value of∑
v αv over all weightingsαv ∈ [0, 1] of V (G) satisfying:
αv + αw ≤ 1 for all vw ∈ E(G)
Properties (assume eG > 0):
for regular G, α∗G = vG/2
for bipartite G, α∗G = αG
for all G,
1 ≤ 1
2vG ≤ α∗G ≤ vG −
eG
∆G≤ vG − 1
MAA 2005, Atlanta – p. 34/49
![Page 161: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/161.jpg)
Toward a general, tight upper tail
N(F, G) – the number of copies of G in F
(so,N(Kn, G) = N(n, G)).N(n, m, G) – the maximum of N(F, G) over all graphsF with vF ≤ n and eF ≤ m.
RecallΨH = nvHpeH
MAA 2005, Atlanta – p. 35/49
![Page 162: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/162.jpg)
Toward a general, tight upper tail
N(F, G) – the number of copies of G in F (so,N(Kn, G) = N(n, G)).
N(n, m, G) – the maximum of N(F, G) over all graphsF with vF ≤ n and eF ≤ m.
RecallΨH = nvHpeH
MAA 2005, Atlanta – p. 35/49
![Page 163: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/163.jpg)
Toward a general, tight upper tail
N(F, G) – the number of copies of G in F (so,N(Kn, G) = N(n, G)).N(n, m, G) – the maximum of N(F, G) over all graphsF with vF ≤ n and eF ≤ m.
RecallΨH = nvHpeH
MAA 2005, Atlanta – p. 35/49
![Page 164: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/164.jpg)
Toward a general, tight upper tail
N(F, G) – the number of copies of G in F (so,N(Kn, G) = N(n, G)).N(n, m, G) – the maximum of N(F, G) over all graphsF with vF ≤ n and eF ≤ m.
M ∗G = M ∗
G(n, p) :=
maxm ≤(
n2
)
: ∀H ⊆ G N(n, m, H) ≤ ΨH p ≥ n−2,
1 p < n−2.
RecallΨH = nvHpeH
MAA 2005, Atlanta – p. 35/49
![Page 165: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/165.jpg)
Toward a general, tight upper tail
N(F, G) – the number of copies of G in F (so,N(Kn, G) = N(n, G)).N(n, m, G) – the maximum of N(F, G) over all graphsF with vF ≤ n and eF ≤ m.
M ∗G = M ∗
G(n, p) :=
maxm ≤(
n2
)
: ∀H ⊆ G N(n, m, H) ≤ ΨH p ≥ n−2,
1 p < n−2.
RecallΨH = nvHpeH
MAA 2005, Atlanta – p. 35/49
![Page 166: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/166.jpg)
Upper tail – new result
Theorem (Janson, Oleszkiewicz, Rucinski (2004))For every graph G and for every t > 1 there existconstants c(t, G) > 0 and C(t, G) > 0 such that for alln ≥ vG and p ∈ (0, 1)
P(XG ≥ tEXG) ≤ exp −c(t, G)M ∗G(n, p) ,
and, provided tEXG ≤ N(n, G),
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p).
If tEXG > N(n, G), the probability is trivially 0.If tEXG ≤ N(n, G) then tpeG ≤ 1, so p ≤ t−1/eG < 1.
MAA 2005, Atlanta – p. 36/49
![Page 167: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/167.jpg)
Upper tail – new result
Theorem (Janson, Oleszkiewicz, Rucinski (2004))For every graph G and for every t > 1 there existconstants c(t, G) > 0 and C(t, G) > 0 such that for alln ≥ vG and p ∈ (0, 1)
P(XG ≥ tEXG) ≤ exp −c(t, G)M ∗G(n, p) ,
and, provided tEXG ≤ N(n, G),
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p).
If tEXG > N(n, G), the probability is trivially 0.If tEXG ≤ N(n, G) then tpeG ≤ 1, so p ≤ t−1/eG < 1.
MAA 2005, Atlanta – p. 36/49
![Page 168: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/168.jpg)
Upper tail – new result
Theorem (Janson, Oleszkiewicz, Rucinski (2004))For every graph G and for every t > 1 there existconstants c(t, G) > 0 and C(t, G) > 0 such that for alln ≥ vG and p ∈ (0, 1)
P(XG ≥ tEXG) ≤ exp −c(t, G)M ∗G(n, p) ,
and, provided tEXG ≤ N(n, G),
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p).
If tEXG > N(n, G), the probability is trivially 0.If tEXG ≤ N(n, G) then tpeG ≤ 1, so p ≤ t−1/eG < 1.
MAA 2005, Atlanta – p. 36/49
![Page 169: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/169.jpg)
Upper tail – new result
Theorem (Janson, Oleszkiewicz, Rucinski (2004))For every graph G and for every t > 1 there existconstants c(t, G) > 0 and C(t, G) > 0 such that for alln ≥ vG and p ∈ (0, 1)
P(XG ≥ tEXG) ≤ exp −c(t, G)M ∗G(n, p) ,
and, provided tEXG ≤ N(n, G),
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p).
If tEXG > N(n, G), the probability is trivially 0.If tEXG ≤ N(n, G) then tpeG ≤ 1, so p ≤ t−1/eG < 1.
MAA 2005, Atlanta – p. 36/49
![Page 170: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/170.jpg)
Upper tail – new result
Theorem (Janson, Oleszkiewicz, Rucinski (2004))For every graph G and for every t > 1 there existconstants c(t, G) > 0 and C(t, G) > 0 such that for alln ≥ vG and p ∈ (0, 1)
P(XG ≥ tEXG) ≤ exp −c(t, G)M ∗G(n, p) ,
and, provided tEXG ≤ N(n, G),
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p).
If tEXG > N(n, G), the probability is trivially 0.
If tEXG ≤ N(n, G) then tpeG ≤ 1, so p ≤ t−1/eG < 1.
MAA 2005, Atlanta – p. 36/49
![Page 171: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/171.jpg)
Upper tail – new result
Theorem (Janson, Oleszkiewicz, Rucinski (2004))For every graph G and for every t > 1 there existconstants c(t, G) > 0 and C(t, G) > 0 such that for alln ≥ vG and p ∈ (0, 1)
P(XG ≥ tEXG) ≤ exp −c(t, G)M ∗G(n, p) ,
and, provided tEXG ≤ N(n, G),
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p).
If tEXG > N(n, G), the probability is trivially 0.If tEXG ≤ N(n, G) then tpeG ≤ 1, so p ≤ t−1/eG < 1.
MAA 2005, Atlanta – p. 36/49
![Page 172: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/172.jpg)
N(n, m, H) explicitly
Theorem For every graph H without isolated vertices,and for all m ≥ eH and n ≥ vH , we have,
N(n, m, H) =
Θ(mα∗H) if m ≤ n (Alon 1981)Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤
(
n2
)
Θ(nvH) if m ≥(
n2
)
(trivial).
Theorem For every graph G and n ≥ vG we have
M ∗G(n, p) =
Θ(1) if p ≤ n−1/mG
Θ(
minH⊆G Ψ1/α∗HH
)
if n−1/mG ≤ p ≤ n−1/∆G
Θ(n2p∆G) if p ≥ n−1/∆G.
MAA 2005, Atlanta – p. 37/49
![Page 173: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/173.jpg)
N(n, m, H) explicitly
Theorem For every graph H without isolated vertices,and for all m ≥ eH and n ≥ vH , we have,
N(n, m, H) =
Θ(mα∗H) if m ≤ n (Alon 1981)
Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤(
n2
)
Θ(nvH) if m ≥(
n2
)
(trivial).
Theorem For every graph G and n ≥ vG we have
M ∗G(n, p) =
Θ(1) if p ≤ n−1/mG
Θ(
minH⊆G Ψ1/α∗HH
)
if n−1/mG ≤ p ≤ n−1/∆G
Θ(n2p∆G) if p ≥ n−1/∆G.
MAA 2005, Atlanta – p. 37/49
![Page 174: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/174.jpg)
N(n, m, H) explicitly
Theorem For every graph H without isolated vertices,and for all m ≥ eH and n ≥ vH , we have,
N(n, m, H) =
Θ(mα∗H) if m ≤ n (Alon 1981)Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤
(
n2
)
Θ(nvH) if m ≥(
n2
)
(trivial).
Theorem For every graph G and n ≥ vG we have
M ∗G(n, p) =
Θ(1) if p ≤ n−1/mG
Θ(
minH⊆G Ψ1/α∗HH
)
if n−1/mG ≤ p ≤ n−1/∆G
Θ(n2p∆G) if p ≥ n−1/∆G.
MAA 2005, Atlanta – p. 37/49
![Page 175: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/175.jpg)
N(n, m, H) explicitly
Theorem For every graph H without isolated vertices,and for all m ≥ eH and n ≥ vH , we have,
N(n, m, H) =
Θ(mα∗H) if m ≤ n (Alon 1981)Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤
(
n2
)
Θ(nvH) if m ≥(
n2
)
(trivial).
Theorem For every graph G and n ≥ vG we have
M ∗G(n, p) =
Θ(1) if p ≤ n−1/mG
Θ(
minH⊆G Ψ1/α∗HH
)
if n−1/mG ≤ p ≤ n−1/∆G
Θ(n2p∆G) if p ≥ n−1/∆G.
MAA 2005, Atlanta – p. 37/49
![Page 176: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/176.jpg)
N(n, m, H) explicitly
Theorem For every graph H without isolated vertices,and for all m ≥ eH and n ≥ vH , we have,
N(n, m, H) =
Θ(mα∗H) if m ≤ n (Alon 1981)Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤
(
n2
)
Θ(nvH) if m ≥(
n2
)
(trivial).
Theorem For every graph G and n ≥ vG we have
M ∗G(n, p) =
Θ(1) if p ≤ n−1/mG
Θ(
minH⊆G Ψ1/α∗HH
)
if n−1/mG ≤ p ≤ n−1/∆G
Θ(n2p∆G) if p ≥ n−1/∆G.
MAA 2005, Atlanta – p. 37/49
![Page 177: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/177.jpg)
N(n, m, H) explicitly
Theorem For every graph H without isolated vertices,and for all m ≥ eH and n ≥ vH , we have,
N(n, m, H) =
Θ(mα∗H) if m ≤ n (Alon 1981)Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤
(
n2
)
Θ(nvH) if m ≥(
n2
)
(trivial).
Theorem For every graph G and n ≥ vG we have
M ∗G(n, p) =
Θ(1) if p ≤ n−1/mG
Θ(
minH⊆G Ψ1/α∗HH
)
if n−1/mG ≤ p ≤ n−1/∆G
Θ(n2p∆G) if p ≥ n−1/∆G.
MAA 2005, Atlanta – p. 37/49
![Page 178: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/178.jpg)
N(n, m, H) explicitly
Theorem For every graph H without isolated vertices,and for all m ≥ eH and n ≥ vH , we have,
N(n, m, H) =
Θ(mα∗H) if m ≤ n (Alon 1981)Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤
(
n2
)
Θ(nvH) if m ≥(
n2
)
(trivial).
Theorem For every graph G and n ≥ vG we have
M ∗G(n, p) =
Θ(1) if p ≤ n−1/mG
Θ(
minH⊆G Ψ1/α∗HH
)
if n−1/mG ≤ p ≤ n−1/∆G
Θ(n2p∆G) if p ≥ n−1/∆G.
MAA 2005, Atlanta – p. 37/49
![Page 179: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/179.jpg)
Special cases: regular graphs, stars
Corollary If G is a k-regular graph, thenM ∗
G = Θ(n2pk) for all p ≥ n−1/mG = n−2/k .
Corollary Let G be the k-armed star K1,k, with k ≥ 1,and assume p ≥ n−1/mG = n−1−1/k. Then
M ∗G =
Θ(n1+1/kp) if p ≤ n−1/k,
Θ(n2pk) if p ≥ n−1/k.
MAA 2005, Atlanta – p. 38/49
![Page 180: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/180.jpg)
Special cases: regular graphs, stars
Corollary If G is a k-regular graph, thenM ∗
G = Θ(n2pk) for all p ≥ n−1/mG = n−2/k .
Corollary Let G be the k-armed star K1,k, with k ≥ 1,and assume p ≥ n−1/mG = n−1−1/k. Then
M ∗G =
Θ(n1+1/kp) if p ≤ n−1/k,
Θ(n2pk) if p ≥ n−1/k.
MAA 2005, Atlanta – p. 38/49
![Page 181: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/181.jpg)
Special cases: paths
Corollary Let Pk be the path on k vertices and assumep ≥ n−1/mPk = n−1−1/(k−1). Then, if k ≥ 3 is odd,
M ∗Pk
=
Θ(
n2 kk+1p2k−1
k+1
)
if p ≤ n−1/2,
Θ(
n2p2)
if p ≥ n−1/2,
and, if k ≥ 4 is even,
M ∗Pk
=
Θ(
n2p2k−1k
)
if p ≤ n−1,
Θ(
n2k−1k p2k−2
k
)
if n−1 ≤ p ≤ n−1/2,
Θ(
n2p2)
if p ≥ n−1/2.
MAA 2005, Atlanta – p. 39/49
![Page 182: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/182.jpg)
Special cases: paths
Corollary Let Pk be the path on k vertices and assumep ≥ n−1/mPk = n−1−1/(k−1). Then, if k ≥ 3 is odd,
M ∗Pk
=
Θ(
n2 kk+1p2k−1
k+1
)
if p ≤ n−1/2,
Θ(
n2p2)
if p ≥ n−1/2,
and, if k ≥ 4 is even,
M ∗Pk
=
Θ(
n2p2k−1k
)
if p ≤ n−1,
Θ(
n2k−1k p2k−2
k
)
if n−1 ≤ p ≤ n−1/2,
Θ(
n2p2)
if p ≥ n−1/2.
MAA 2005, Atlanta – p. 39/49
![Page 183: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/183.jpg)
Graphs with many phases
Let T k be the tree obtained by taking k stars K1,i,i = 1, . . . k, and tying them up by merging one pendantvertex from each star into one vertex.
Proposition For every k ≥ 2, the graph T k describedabove has k + 1 phases for the upper tail.
MAA 2005, Atlanta – p. 40/49
![Page 184: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/184.jpg)
Graphs with many phases
Let T k be the tree obtained by taking k stars K1,i,i = 1, . . . k, and tying them up by merging one pendantvertex from each star into one vertex.
Proposition For every k ≥ 2, the graph T k describedabove has k + 1 phases for the upper tail.
MAA 2005, Atlanta – p. 40/49
![Page 185: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/185.jpg)
Graphs with many phases
Let T k be the tree obtained by taking k stars K1,i,i = 1, . . . k, and tying them up by merging one pendantvertex from each star into one vertex.
Proposition For every k ≥ 2, the graph T k describedabove has k + 1 phases for the upper tail.
MAA 2005, Atlanta – p. 40/49
![Page 186: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/186.jpg)
Idea of proof : lower bound
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p)
By the definition of M ∗G there is H ⊆ G:
N(n, M ∗G+1, H) > ΨH ⇒ N(n, CtM
∗G, H) > tΨH
For simplicity, say, H = G, and take m = CtM∗G.Then
∃F ⊆ Kn, eF ≤ m : N(F, G) > tΨG > tEXG
Finally,
P(XG ≥ tEXG) ≥ P (G(n, p) ⊇ F ) = peF ≥ pm.
MAA 2005, Atlanta – p. 41/49
![Page 187: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/187.jpg)
Idea of proof : lower bound
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p)
By the definition of M ∗G there is H ⊆ G:
N(n, M ∗G+1, H) > ΨH ⇒ N(n, CtM
∗G, H) > tΨH
For simplicity, say, H = G, and take m = CtM∗G.Then
∃F ⊆ Kn, eF ≤ m : N(F, G) > tΨG > tEXG
Finally,
P(XG ≥ tEXG) ≥ P (G(n, p) ⊇ F ) = peF ≥ pm.
MAA 2005, Atlanta – p. 41/49
![Page 188: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/188.jpg)
Idea of proof : lower bound
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p)
By the definition of M ∗G there is H ⊆ G:
N(n, M ∗G+1, H) > ΨH ⇒ N(n, CtM
∗G, H) > tΨH
For simplicity, say, H = G, and take m = CtM∗G.
Then
∃F ⊆ Kn, eF ≤ m : N(F, G) > tΨG > tEXG
Finally,
P(XG ≥ tEXG) ≥ P (G(n, p) ⊇ F ) = peF ≥ pm.
MAA 2005, Atlanta – p. 41/49
![Page 189: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/189.jpg)
Idea of proof : lower bound
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p)
By the definition of M ∗G there is H ⊆ G:
N(n, M ∗G+1, H) > ΨH ⇒ N(n, CtM
∗G, H) > tΨH
For simplicity, say, H = G, and take m = CtM∗G. Then
∃F ⊆ Kn, eF ≤ m : N(F, G) > tΨG > tEXG
Finally,
P(XG ≥ tEXG) ≥ P (G(n, p) ⊇ F ) = peF ≥ pm.
MAA 2005, Atlanta – p. 41/49
![Page 190: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/190.jpg)
Idea of proof : lower bound
P(XG ≥ tEXG) ≥ pC(t,G)M∗
G(n,p)
By the definition of M ∗G there is H ⊆ G:
N(n, M ∗G+1, H) > ΨH ⇒ N(n, CtM
∗G, H) > tΨH
For simplicity, say, H = G, and take m = CtM∗G. Then
∃F ⊆ Kn, eF ≤ m : N(F, G) > tΨG > tEXG
Finally,
P(XG ≥ tEXG) ≥ P (G(n, p) ⊇ F ) = peF ≥ pm.
MAA 2005, Atlanta – p. 41/49
![Page 191: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/191.jpg)
Idea of proof : upper bound
By Markov’s inequality, with λG = EXG, for everym ≥ 1
P(XG ≥ tλG) = P(XmG ≥ tmλm
G) ≤ E(XmG )
tmλmG
For suitable choice of c′, with m = c′M ∗G,
E(XmG ) ≤ λm
Gtm/2,
so
P(XG ≥ tλG) ≤ t−m/2 = exp−(m/2) log t = exp−cM ∗G,
where c = (c′/2) log t.
MAA 2005, Atlanta – p. 42/49
![Page 192: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/192.jpg)
Idea of proof : upper bound
By Markov’s inequality, with λG = EXG, for everym ≥ 1
P(XG ≥ tλG) = P(XmG ≥ tmλm
G) ≤ E(XmG )
tmλmG
For suitable choice of c′, with m = c′M ∗G,
E(XmG ) ≤ λm
Gtm/2,
so
P(XG ≥ tλG) ≤ t−m/2 = exp−(m/2) log t = exp−cM ∗G,
where c = (c′/2) log t.
MAA 2005, Atlanta – p. 42/49
![Page 193: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/193.jpg)
Idea of proof : upper bound
By Markov’s inequality, with λG = EXG, for everym ≥ 1
P(XG ≥ tλG) = P(XmG ≥ tmλm
G) ≤ E(XmG )
tmλmG
For suitable choice of c′, with m = c′M ∗G,
E(XmG ) ≤ λm
Gtm/2,
so
P(XG ≥ tλG) ≤ t−m/2 = exp−(m/2) log t = exp−cM ∗G,
where c = (c′/2) log t.MAA 2005, Atlanta – p. 42/49
![Page 194: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/194.jpg)
The mth moment
We will show by induction on m that
E(XmG ) ≤ λm
G
(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)m−1
This is trivially true for m = 1. Assume true for m− 1.
Let G1, · · · , GN(n,G) be all copies of G in G(n, p) and letIi be the indicator of presence of Gi in G(n, p).Form ≥ 2,
E(XmG ) =
∑
i1,...,im
E(Ii1 · · · Iim) =∑
i1,...,im
pe(Gi1∪···∪Gim)
MAA 2005, Atlanta – p. 43/49
![Page 195: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/195.jpg)
The mth moment
We will show by induction on m that
E(XmG ) ≤ λm
G
(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)m−1
This is trivially true for m = 1.
Assume true for m− 1.
Let G1, · · · , GN(n,G) be all copies of G in G(n, p) and letIi be the indicator of presence of Gi in G(n, p).Form ≥ 2,
E(XmG ) =
∑
i1,...,im
E(Ii1 · · · Iim) =∑
i1,...,im
pe(Gi1∪···∪Gim)
MAA 2005, Atlanta – p. 43/49
![Page 196: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/196.jpg)
The mth moment
We will show by induction on m that
E(XmG ) ≤ λm
G
(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)m−1
This is trivially true for m = 1. Assume true for m− 1.
Let G1, · · · , GN(n,G) be all copies of G in G(n, p) and letIi be the indicator of presence of Gi in G(n, p).Form ≥ 2,
E(XmG ) =
∑
i1,...,im
E(Ii1 · · · Iim) =∑
i1,...,im
pe(Gi1∪···∪Gim)
MAA 2005, Atlanta – p. 43/49
![Page 197: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/197.jpg)
The mth moment
We will show by induction on m that
E(XmG ) ≤ λm
G
(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)m−1
This is trivially true for m = 1. Assume true for m− 1.
Let G1, · · · , GN(n,G) be all copies of G in G(n, p) and letIi be the indicator of presence of Gi in G(n, p).
Form ≥ 2,
E(XmG ) =
∑
i1,...,im
E(Ii1 · · · Iim) =∑
i1,...,im
pe(Gi1∪···∪Gim)
MAA 2005, Atlanta – p. 43/49
![Page 198: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/198.jpg)
The mth moment
We will show by induction on m that
E(XmG ) ≤ λm
G
(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)m−1
This is trivially true for m = 1. Assume true for m− 1.
Let G1, · · · , GN(n,G) be all copies of G in G(n, p) and letIi be the indicator of presence of Gi in G(n, p). Form ≥ 2,
E(XmG ) =
∑
i1,...,im
E(Ii1 · · · Iim) =∑
i1,...,im
pe(Gi1∪···∪Gim)
MAA 2005, Atlanta – p. 43/49
![Page 199: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/199.jpg)
Induction step
Set F = F (i1, . . . , im−1) = Gi1 ∪ · · · ∪Gim−1.
∑
i1,...,im
pe(Gi1∪···∪Gim) =
∑
i1,...,im−1
pe(F )∑
im
peG−e(F∩Gim)
≤∑
i1,...,im−1pe(F )
(
N(n, G)peG +∑
H⊆G
∑
Gi∩F∼=H peG−eH)
≤∑
i1,...,im−1pe(F )
(
λG +∑
H⊆G N(n, (m− 1)eG, H)ΨG
ΨH
)
≤ E(Xm−1G ) · λG
(
1 + 2vG!∑
H⊆GN(n,(m−1)eG,H)
ΨH
)
.
MAA 2005, Atlanta – p. 44/49
![Page 200: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/200.jpg)
Induction step
Set F = F (i1, . . . , im−1) = Gi1 ∪ · · · ∪Gim−1.
∑
i1,...,im
pe(Gi1∪···∪Gim) =
∑
i1,...,im−1
pe(F )∑
im
peG−e(F∩Gim)
≤∑
i1,...,im−1
pe(F )
(
N(n, G)peG +∑
H⊆G
∑
Gi∩F∼=H
peG−eH
)
≤∑
i1,...,im−1pe(F )
(
λG +∑
H⊆G N(n, (m− 1)eG, H)ΨG
ΨH
)
≤ E(Xm−1G ) · λG
(
1 + 2vG!∑
H⊆GN(n,(m−1)eG,H)
ΨH
)
.
MAA 2005, Atlanta – p. 44/49
![Page 201: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/201.jpg)
Induction step
Set F = F (i1, . . . , im−1) = Gi1 ∪ · · · ∪Gim−1.
∑
i1,...,im
pe(Gi1∪···∪Gim) =
∑
i1,...,im−1
pe(F )∑
im
peG−e(F∩Gim)
≤∑
i1,...,im−1
pe(F )
(
N(n, G)peG +∑
H⊆G
∑
Gi∩F∼=H
peG−eH
)
≤∑
i1,...,im−1
pe(F )
(
λG +∑
H⊆G
N(n, (m− 1)eG, H)ΨG
ΨH
)
≤ E(Xm−1G ) · λG
(
1 + 2vG!∑
H⊆GN(n,(m−1)eG,H)
ΨH
)
.
MAA 2005, Atlanta – p. 44/49
![Page 202: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/202.jpg)
Induction step
Set F = F (i1, . . . , im−1) = Gi1 ∪ · · · ∪Gim−1.
∑
i1,...,im
pe(Gi1∪···∪Gim) =
∑
i1,...,im−1
pe(F )∑
im
peG−e(F∩Gim)
≤∑
i1,...,im−1
pe(F )
(
N(n, G)peG +∑
H⊆G
∑
Gi∩F∼=H
peG−eH
)
≤∑
i1,...,im−1
pe(F )
(
λG +∑
H⊆G
N(n, (m− 1)eG, H)ΨG
ΨH
)
≤ E(Xm−1G ) · λG
(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)
.
MAA 2005, Atlanta – p. 44/49
![Page 203: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/203.jpg)
Induction step
Set F = F (i1, . . . , im−1) = Gi1 ∪ · · · ∪Gim−1.
∑
i1,...,im
pe(Gi1∪···∪Gim) =
∑
i1,...,im−1
pe(F )∑
im
peG−e(F∩Gim)
≤∑
i1,...,im−1
pe(F )
(
N(n, G)peG +∑
H⊆G
∑
Gi∩F∼=H
peG−eH
)
≤∑
i1,...,im−1
pe(F )
(
λG +∑
H⊆G
N(n, (m− 1)eG, H)ΨG
ΨH
)
≤ E(Xm−1G ) · λG
(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)
.
MAA 2005, Atlanta – p. 44/49
![Page 204: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/204.jpg)
Bounding the mth moment
With m = c′M ∗G,
N(n, (m− 1)eG, H) ≤ c′′N(n, M ∗G, H) < c′′ΨH
and for c′ = c′(G, t) small enough(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)
≤√
t
which proves that
E(XmG ) ≤ λm
Gtm/2.
MAA 2005, Atlanta – p. 45/49
![Page 205: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/205.jpg)
Bounding the mth moment
With m = c′M ∗G,
N(n, (m− 1)eG, H) ≤ c′′N(n, M ∗G, H) < c′′ΨH
and for c′ = c′(G, t) small enough(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)
≤√
t
which proves that
E(XmG ) ≤ λm
Gtm/2.
MAA 2005, Atlanta – p. 45/49
![Page 206: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/206.jpg)
Bounding the mth moment
With m = c′M ∗G,
N(n, (m− 1)eG, H) ≤ c′′N(n, M ∗G, H) < c′′ΨH
and for c′ = c′(G, t) small enough(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)
≤√
t
which proves that
E(XmG ) ≤ λm
Gtm/2.
MAA 2005, Atlanta – p. 45/49
![Page 207: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/207.jpg)
Bounding the mth moment
With m = c′M ∗G,
N(n, (m− 1)eG, H) ≤ c′′N(n, M ∗G, H) < c′′ΨH
and for c′ = c′(G, t) small enough(
1 + 2vG!∑
H⊆G
N(n, (m− 1)eG, H)
ΨH
)
≤√
t
which proves that
E(XmG ) ≤ λm
Gtm/2.
MAA 2005, Atlanta – p. 45/49
![Page 208: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/208.jpg)
Estimating N(n, m, H)
To prove:
N(n, m, H) = Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤(
n
2
)
Consider LP: max∑
v∈V xv
given
0 ≤ xv ≤ log n and ∀vw ∈ E : xv + xw ≤ log m.
Let γ be the value of an optimal solution (xv).
MAA 2005, Atlanta – p. 46/49
![Page 209: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/209.jpg)
Estimating N(n, m, H)
To prove:
N(n, m, H) = Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤(
n
2
)
Consider LP: max∑
v∈V xv
given
0 ≤ xv ≤ log n and ∀vw ∈ E : xv + xw ≤ log m.
Let γ be the value of an optimal solution (xv).
MAA 2005, Atlanta – p. 46/49
![Page 210: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/210.jpg)
Estimating N(n, m, H)
To prove:
N(n, m, H) = Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤(
n
2
)
Consider LP: max∑
v∈V xv
given
0 ≤ xv ≤ log n and ∀vw ∈ E : xv + xw ≤ log m.
Let γ be the value of an optimal solution (xv).
MAA 2005, Atlanta – p. 46/49
![Page 211: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/211.jpg)
Estimating N(n, m, H)
To prove:
N(n, m, H) = Θ(mvH−α∗Hn2α∗H−vH) if n ≤ m ≤(
n
2
)
Consider LP: max∑
v∈V xv
given
0 ≤ xv ≤ log n and ∀vw ∈ E : xv + xw ≤ log m.
Let γ be the value of an optimal solution (xv).
MAA 2005, Atlanta – p. 46/49
![Page 212: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/212.jpg)
Computing γ
We have xv ≥ log m− log n.
Write:
xv = log m− log n + (2 log n− log m)αv
where 0 ≤ αv ≤ 1, and ∀vw ∈ E: αv + αw ≤ 1.Then
γ =∑
v
xv = (log m−log n)vH+(2 log n−log m)∑
v
αv
so
eγ =(m
n
)vH
(
n2
m
)α∗H
MAA 2005, Atlanta – p. 47/49
![Page 213: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/213.jpg)
Computing γ
We have xv ≥ log m− log n. Write:
xv = log m− log n + (2 log n− log m)αv
where 0 ≤ αv ≤ 1, and ∀vw ∈ E: αv + αw ≤ 1.Then
γ =∑
v
xv = (log m−log n)vH+(2 log n−log m)∑
v
αv
so
eγ =(m
n
)vH
(
n2
m
)α∗H
MAA 2005, Atlanta – p. 47/49
![Page 214: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/214.jpg)
Computing γ
We have xv ≥ log m− log n. Write:
xv = log m− log n + (2 log n− log m)αv
where 0 ≤ αv ≤ 1, and ∀vw ∈ E: αv + αw ≤ 1.
Then
γ =∑
v
xv = (log m−log n)vH+(2 log n−log m)∑
v
αv
so
eγ =(m
n
)vH
(
n2
m
)α∗H
MAA 2005, Atlanta – p. 47/49
![Page 215: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/215.jpg)
Computing γ
We have xv ≥ log m− log n. Write:
xv = log m− log n + (2 log n− log m)αv
where 0 ≤ αv ≤ 1, and ∀vw ∈ E: αv + αw ≤ 1. Then
γ =∑
v
xv = (log m−log n)vH+(2 log n−log m)∑
v
αv
so
eγ =(m
n
)vH
(
n2
m
)α∗H
MAA 2005, Atlanta – p. 47/49
![Page 216: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/216.jpg)
Computing γ
We have xv ≥ log m− log n. Write:
xv = log m− log n + (2 log n− log m)αv
where 0 ≤ αv ≤ 1, and ∀vw ∈ E: αv + αw ≤ 1. Then
γ =∑
v
xv = (log m−log n)vH+(2 log n−log m)∑
v
αv
so
eγ =(m
n
)vH
(
n2
m
)α∗H
MAA 2005, Atlanta – p. 47/49
![Page 217: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/217.jpg)
Relating γ to N(n, m, H)
Proposition N(n, m, H) = Θ (eγ)
Proof: (only from below)based on an optimal solution(xv), construct F rich in copies of H .How?Blow up H , replacing each v by nv = exv/vH verticesand each vw ∈ E by K(nv, nw).ThenvF =
∑
v nv ≤ n and
eF =∑
vw∈E
nvnw ≤∑
vw∈E
m/v2H < m
ButN(F, H) ≥
∏
v
nv = cvHeγ.
MAA 2005, Atlanta – p. 48/49
![Page 218: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/218.jpg)
Relating γ to N(n, m, H)
Proposition N(n, m, H) = Θ (eγ)Proof: (only from below)
based on an optimal solution(xv), construct F rich in copies of H .How?Blow up H , replacing each v by nv = exv/vH verticesand each vw ∈ E by K(nv, nw).ThenvF =
∑
v nv ≤ n and
eF =∑
vw∈E
nvnw ≤∑
vw∈E
m/v2H < m
ButN(F, H) ≥
∏
v
nv = cvHeγ.
MAA 2005, Atlanta – p. 48/49
![Page 219: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/219.jpg)
Relating γ to N(n, m, H)
Proposition N(n, m, H) = Θ (eγ)Proof: (only from below) based on an optimal solution(xv), construct F rich in copies of H .
How?Blow up H , replacing each v by nv = exv/vH verticesand each vw ∈ E by K(nv, nw).ThenvF =
∑
v nv ≤ n and
eF =∑
vw∈E
nvnw ≤∑
vw∈E
m/v2H < m
ButN(F, H) ≥
∏
v
nv = cvHeγ.
MAA 2005, Atlanta – p. 48/49
![Page 220: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/220.jpg)
Relating γ to N(n, m, H)
Proposition N(n, m, H) = Θ (eγ)Proof: (only from below) based on an optimal solution(xv), construct F rich in copies of H . How?
Blow up H , replacing each v by nv = exv/vH verticesand each vw ∈ E by K(nv, nw).ThenvF =
∑
v nv ≤ n and
eF =∑
vw∈E
nvnw ≤∑
vw∈E
m/v2H < m
ButN(F, H) ≥
∏
v
nv = cvHeγ.
MAA 2005, Atlanta – p. 48/49
![Page 221: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/221.jpg)
Relating γ to N(n, m, H)
Proposition N(n, m, H) = Θ (eγ)Proof: (only from below) based on an optimal solution(xv), construct F rich in copies of H . How?Blow up H , replacing each v by nv = exv/vH verticesand each vw ∈ E by K(nv, nw).
ThenvF =
∑
v nv ≤ n and
eF =∑
vw∈E
nvnw ≤∑
vw∈E
m/v2H < m
ButN(F, H) ≥
∏
v
nv = cvHeγ.
MAA 2005, Atlanta – p. 48/49
![Page 222: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/222.jpg)
Relating γ to N(n, m, H)
Proposition N(n, m, H) = Θ (eγ)Proof: (only from below) based on an optimal solution(xv), construct F rich in copies of H . How?Blow up H , replacing each v by nv = exv/vH verticesand each vw ∈ E by K(nv, nw). ThenvF =
∑
v nv ≤ n
and
eF =∑
vw∈E
nvnw ≤∑
vw∈E
m/v2H < m
ButN(F, H) ≥
∏
v
nv = cvHeγ.
MAA 2005, Atlanta – p. 48/49
![Page 223: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/223.jpg)
Relating γ to N(n, m, H)
Proposition N(n, m, H) = Θ (eγ)Proof: (only from below) based on an optimal solution(xv), construct F rich in copies of H . How?Blow up H , replacing each v by nv = exv/vH verticesand each vw ∈ E by K(nv, nw). ThenvF =
∑
v nv ≤ n and
eF =∑
vw∈E
nvnw ≤∑
vw∈E
m/v2H < m
ButN(F, H) ≥
∏
v
nv = cvHeγ.
MAA 2005, Atlanta – p. 48/49
![Page 224: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/224.jpg)
Relating γ to N(n, m, H)
Proposition N(n, m, H) = Θ (eγ)Proof: (only from below) based on an optimal solution(xv), construct F rich in copies of H . How?Blow up H , replacing each v by nv = exv/vH verticesand each vw ∈ E by K(nv, nw). ThenvF =
∑
v nv ≤ n and
eF =∑
vw∈E
nvnw ≤∑
vw∈E
m/v2H < m
ButN(F, H) ≥
∏
v
nv = cvHeγ.
MAA 2005, Atlanta – p. 48/49
![Page 225: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/225.jpg)
Open problem
Determine the order of magnitude for
− log P(XG ≥ tEXG)
It is between Θ(M ∗G) and Θ(M ∗
G log(1/p)).For G = K2, it is Θ(M ∗
G) (Chernoff).For G = K4 and n−2/3 log1/6 n p ≤ n−1/2−ε, there isan upper bound
P(XG ≥ 2EXG) ≤ exp
−cM ∗G(n, p) log1/2 n
,
by the deletion method (Janson, Rucinski (2004)).Thus, neither end is sharp!
MAA 2005, Atlanta – p. 49/49
![Page 226: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/226.jpg)
Open problem
Determine the order of magnitude for
− log P(XG ≥ tEXG)
It is between Θ(M ∗G) and Θ(M ∗
G log(1/p)).
For G = K2, it is Θ(M ∗G) (Chernoff).
For G = K4 and n−2/3 log1/6 n p ≤ n−1/2−ε, there isan upper bound
P(XG ≥ 2EXG) ≤ exp
−cM ∗G(n, p) log1/2 n
,
by the deletion method (Janson, Rucinski (2004)).Thus, neither end is sharp!
MAA 2005, Atlanta – p. 49/49
![Page 227: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/227.jpg)
Open problem
Determine the order of magnitude for
− log P(XG ≥ tEXG)
It is between Θ(M ∗G) and Θ(M ∗
G log(1/p)).For G = K2, it is Θ(M ∗
G) (Chernoff).
For G = K4 and n−2/3 log1/6 n p ≤ n−1/2−ε, there isan upper bound
P(XG ≥ 2EXG) ≤ exp
−cM ∗G(n, p) log1/2 n
,
by the deletion method (Janson, Rucinski (2004)).Thus, neither end is sharp!
MAA 2005, Atlanta – p. 49/49
![Page 228: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/228.jpg)
Open problem
Determine the order of magnitude for
− log P(XG ≥ tEXG)
It is between Θ(M ∗G) and Θ(M ∗
G log(1/p)).For G = K2, it is Θ(M ∗
G) (Chernoff).For G = K4 and n−2/3 log1/6 n p ≤ n−1/2−ε, there isan upper bound
P(XG ≥ 2EXG) ≤ exp
−cM ∗G(n, p) log1/2 n
,
by the deletion method (Janson, Rucinski (2004)).Thus, neither end is sharp!
MAA 2005, Atlanta – p. 49/49
![Page 229: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/229.jpg)
Open problem
Determine the order of magnitude for
− log P(XG ≥ tEXG)
It is between Θ(M ∗G) and Θ(M ∗
G log(1/p)).For G = K2, it is Θ(M ∗
G) (Chernoff).For G = K4 and n−2/3 log1/6 n p ≤ n−1/2−ε, there isan upper bound
P(XG ≥ 2EXG) ≤ exp
−cM ∗G(n, p) log1/2 n
,
by the deletion method (Janson, Rucinski (2004)).
Thus, neither end is sharp!
MAA 2005, Atlanta – p. 49/49
![Page 230: SMALL SUBGRAPHS OF RANDOM GRAPHS - CMUaf1p/MAA2005/L4.pdfMAA 2005, Atlanta – p. 2/49 Subgraph counts A copy of a graph G in another graph F is a (weak) subgraph G0 of F isomorphic](https://reader033.vdocuments.mx/reader033/viewer/2022053108/6077fccf0774da647d42f108/html5/thumbnails/230.jpg)
Open problem
Determine the order of magnitude for
− log P(XG ≥ tEXG)
It is between Θ(M ∗G) and Θ(M ∗
G log(1/p)).For G = K2, it is Θ(M ∗
G) (Chernoff).For G = K4 and n−2/3 log1/6 n p ≤ n−1/2−ε, there isan upper bound
P(XG ≥ 2EXG) ≤ exp
−cM ∗G(n, p) log1/2 n
,
by the deletion method (Janson, Rucinski (2004)).Thus, neither end is sharp!
MAA 2005, Atlanta – p. 49/49