Alphabeta Math
Session-authored (Fable 5 assisted)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

11 results · all verified · 7 also independently AI-judged
Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs. The 4 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

The Erdős–Hajnal Property and Homogeneous Sets

1 · Prerequisites

2 · Summary

The published clique and stable-set numbers of a finite graph, the notions of a hereditary class and of an H-free class, the complement of a graph and of a class, and the facts that complementation preserves hereditary classes and exchanges cliques with stable sets are the setting for what follows; induced embeddings compose. Finite Ramsey theory supplies the binomial bound on Ramsey numbers; the Erdős–Rényi random graph, the probability that a fixed pattern occurs, and the first-moment method supply the probabilistic counterpart; and the published theory of real powers and logarithms, including change of base and the fact that a logarithm grows more slowly than every positive power, is what makes logn and nϵ comparable.

A homogeneous set is a clique or a stable set, and hom(G) is the larger of the two numbers; a hereditary class has the Erdős–Hajnal property when some positive exponent ϵ forces hom(G)V(G)ϵ throughout the class. The page proves that the admissible exponents are downward closed, that every nonempty n-vertex graph satisfies hom(G)12log2n, that Kt-free classes have the property and every P3-free graph satisfies hom(G)V(G), and that for n16 some n-vertex graph has hom(G)<3log2n, so the class of all graphs does not have the property. It then establishes complement invariance, passage to hereditary subclasses, monotonicity under induced pattern containment, the property for every graph on at most three vertices, and the equivalence of the single-pattern and finite-family formulations, closing with a remark stating the conjecture.

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

Homogeneous vertex sets and the homogeneous number hom(G)=max{ω(G),α(G)}

Definition

Let G be a finite graph. A vertex set SV(G) is homogeneous if it is a clique or a stable set in G (Cliques, stable sets, the clique number ω(G) and stability number α(G)). The homogeneous number of G is hom(G):=max{ω(G),α(G)}.

For the null graph, the published conventions give ω(G)=α(G)=0, and hence hom(G)=0.

DefinitionDefinition: Literature-sourcedProof: Not applicableaudited 2026-08-16Open item page →

The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class

Definition

Let C be a hereditary class of finite graphs (Hereditary graph classes). A real number ϵ>0 is an Erdős–Hajnal constant for C if every nonempty GC satisfies hom(G)V(G)ϵ, where the homogeneous number is that of Homogeneous vertex sets and the homogeneous number hom(G)=max{ω(G),α(G)} and the power is that of Real powers for positive bases, with the zero-base positive-exponent convention. The class C has the Erdős–Hajnal property if it has an Erdős–Hajnal constant.

For a finite graph H, we say that H has the Erdős–Hajnal property when the hereditary class of H-free graphs has it. The same terminology applies to a finite family F through its class of F-free graphs.

LemmaStatement: AI-adaptedProof: AI-generatedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

Every smaller positive exponent is again an Erdős–Hajnal constant

Statement

Let C be a hereditary graph class. If ϵ is an Erdős–Hajnal constant for C and 0<δϵ, then δ is also an Erdős–Hajnal constant for C.

Facts & Assumptions

Given: A hereditary class C, an Erdős–Hajnal constant ϵ for it, and a real δ with 0<δϵ.

[L1]

A positive real c is an Erdős–Hajnal constant for C exactly when every nonempty GC satisfies hom(G)V(G)c, with ac=exp(cloga) for a>0 (The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class).

[L3]

The exponential function is strictly increasing on R (The exponential function is strictly increasing).

Proof

technique · direct
1.1

Let GC be nonempty and put n=V(G)1.

given
1.2

If n=1, then nδ=nϵ=1, so the required inequality follows from the one for ϵ.

L1algebra
1.3

If n>1, then logn>0 by [L2], and hence δlognϵlogn.

givenL2algebra
2.1

In the case n>1, [L3] and the real-power convention in [L1] give nδ=exp(δlogn)exp(ϵlogn)=nϵ.

step 1.3L1L3
3.1

In both cases, hom(G)nϵnδ; since G was arbitrary, δ is an Erdős–Hajnal constant for C.

step 1.2step 2.1L1
TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-16Open item page →

Every nonempty n-vertex graph satisfies hom(G)12log2n

Statement

Every nonempty finite graph G of order n satisfies hom(G)12log2n.

Facts & Assumptions

Given: A nonempty finite graph G with n=V(G).

[L1]

For every graph F, hom(F)=max{ω(F),α(F)} (Homogeneous vertex sets and the homogeneous number hom(G)=max{ω(G),α(G)}).

[L2]

For positive natural numbers s,t, every graph on at least (s+t2s1) vertices has an s-vertex clique or a t-vertex stable set (Finite graph Ramsey theorem: (s+t2s1)(s,t)2 for all positive s,t).

[L3]

The number (mr) counts the r-element subsets of an m-element set (The set [A]k of k-element subsets and the binomial coefficient (nk):=[n]k).

[L4]

For b>0 with b1 and x>0, logbx:=logx/logb (The logarithm to a positive base other than one).

[L5]

log:(0,)R is strictly increasing, log(xy)=logx+logy for x,y>0, and log1=0 (Order, continuity, range, and the product, quotient, and reciprocal laws for the natural logarithm).

Proof

technique · direct
1.1

Put k=12log2n+1. Then k is a positive integer, k>12log2n, and 2k2log2n.

givenalgebra
2.1

By [L5], log2>log1=0, so 2k2log2n=logn/log2 gives (2k2)log2logn. Applying log(xy)=logx+logy to the 2k2 factors of 22k2 gives log(22k2)=(2k2)log2logn, and log is strictly increasing, so 22k2n.

step 1.1L4L5algebra
3.1

The (k1)-subsets of a (2k2)-set form part of its power set, and binary membership choices give the power set 22k2 elements, so (2k2k1)22k2n.

step 2.1L3algebra
4.1

Apply [L2] with s=t=k: G has a clique or stable set of order at least k.

step 3.1L2
5.1

Therefore hom(G)k>12log2n by [L1], which proves the stated weak inequality; when n=1, this reads 10 and the same argument has k=1.

step 1.1step 4.1L1
TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-16Open item page →

For every t1, the class of Kt-free graphs has the Erdős–Hajnal property

Statement

For every positive integer t, the hereditary class of Kt-free finite graphs has the Erdős–Hajnal property.

Facts & Assumptions

Given: A positive integer t and the class Ct of Kt-free finite graphs.

[L1]

For a graph G, hom(G)=max{ω(G),α(G)} (Homogeneous vertex sets and the homogeneous number hom(G)=max{ω(G),α(G)}).

[L2]

A hereditary class has the Erdős–Hajnal property when some ϵ>0 satisfies hom(G)V(G)ϵ for every nonempty member G (The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class).

[L3]

A graph is Kt-free when it has no induced copy of Kt (H-free and F-free graphs under the induced-subgraph convention), and the class of graphs free of any fixed family is hereditary (Every class defined by forbidden induced subgraphs is hereditary).

[L4]

The graph Kt has every pair of its t vertices as an edge, while an empty graph has no edges (Empty and complete graphs, complete bipartite graphs, and the convention that Pn and Cn have n vertices).

[L5]

For positive a,b, every graph on at least (a+b2a1) vertices contains an a-clique or a b-vertex stable set (Finite graph Ramsey theorem: (s+t2s1)(s,t)2 for all positive s,t).

[L6]

The binomial coefficient (mr) counts the r-subsets of an m-set (The set [A]k of k-element subsets and the binomial coefficient (nk):=[n]k).

[L7]

For x>0 and real u, xu=exp(ulogx) (Real powers for positive bases, with the zero-base positive-exponent convention).

[L8]

The logarithm is strictly increasing, maps 1 to 0, and obeys the product and quotient laws (Order, continuity, range, and the product, quotient, and reciprocal laws for the natural logarithm).

[L9]

The exponential is strictly increasing (The exponential function is strictly increasing).

Proof

technique · direct
1.1

By [L3], Ct is hereditary. If t=1, it has no nonempty member, so any positive exponent works in [L2]; if t=2, every member is empty by [L3] and [L4], so hom(G)=V(G) and exponent 1 works.

L1L2L3L4
1.2

Assume t3. For all sufficiently large integers n, the integer s=12n1/(t1) satisfies st2, sn1/t, and s1; these assertions follow from [L7], [L8], and [L9] because n1/(t1)/n1/t=n1/(t(t1)) tends to infinity.

givenL7L8L9algebra
2.1

For such n, t+s22s, and hence (t+s2t1)(t+s2)t1(2s)t1n; the first inequality counts ordered choices containing every (t1)-subset.

step 1.2L6algebra
3.1

If GCt has sufficiently large order n, [L5] with parameters (t,s) and step 2.1 give a t-clique or an s-vertex stable set; the first is forbidden, so hom(G)α(G)sn1/t.

step 1.2step 2.1L1L3L5
4.1

Choose an integer threshold N2 beyond which step 3.1 applies, and choose 0<ϵ1/t so small that Nϵ2. Such an ϵ exists by [L7], [L8], and [L9].

step 3.1L7L8L9choose
5.1

If GCt has nN, then hom(G)n1/tnϵ; if 2n<N, an edge gives a two-vertex clique and a nonedge gives a two-vertex stable set, so hom(G)2nϵ; and if n=1, both sides equal 1. Thus ϵ is an Erdős–Hajnal constant for Ct.

step 3.1step 4.1L1L2L4algebra
TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

Every P3-free graph G satisfies hom(G)V(G)

Statement

Every P3-free finite graph G satisfies hom(G)V(G). Consequently the hereditary class of P3-free graphs has Erdős–Hajnal constant 1/2.

Facts & Assumptions

Given: A finite P3-free graph G.

[L1]

The homogeneous number is hom(G)=max{ω(G),α(G)} (Homogeneous vertex sets and the homogeneous number hom(G)=max{ω(G),α(G)}).

[L2]

A hereditary class has constant c>0 when every nonempty member F satisfies hom(F)V(F)c (The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class).

[L3]

P3-free means having no induced copy of the three-vertex path, and every fixed-pattern-free class is hereditary (H-free and F-free graphs under the induced-subgraph convention, Every class defined by forbidden induced subgraphs is hereditary).

[L4]

The graph P3 has three vertices and exactly its two consecutive edges (Empty and complete graphs, complete bipartite graphs, and the convention that Pn and Cn have n vertices).

[L5]

Connected vertices are joined by a path, and a component is the induced graph on all vertices reachable from one vertex (Connected graphs and connected components defined by the existence of vertex paths).

[L6]

Component vertex sets are nonempty, pairwise disjoint, cover V(G), and induce connected graphs (The connected components of a graph partition its vertex set and are its maximal connected subgraphs).

[L7]

A path has distinct vertices and consecutive vertices adjacent (Walks, closed walks, trails, paths and cycles, with length equal to the number of traversed edges); the distance of connected vertices is the minimum length of a path joining them (Graph distance within a component, eccentricity, diameter and girth, including the acyclic convention).

[L8]

Every nonnegative real a has a unique nonnegative square root a (Square roots exist: a unique a0 with (a)2=a; the positives are {x2:x0}), and a1/2 agrees with the rational-power square root, including at a=0 (The exponential definition of real powers agrees with the existing rational powers).

Proof

technique · direct
1.1

If G is null, then hom(G)=0=0 by [L1] and [L8]. Assume henceforth that G is nonempty.

givenL1L8
1.2

Every connected component of G is a clique: otherwise two nonadjacent vertices in one component have a shortest path v0,,v with 2; the vertices v0,v1,v2 are distinct, the consecutive pairs are edges, and v0v2 is not an edge because it would shorten the path, so they induce P3, contrary to [L3].

L3L4L5L7
1.3

Let r1 be the number of connected components of G. Choosing one vertex from each of these finitely many nonempty components gives a stable set, since an edge would put its endpoints in one component; hence α(G)r.

L5L6choose
2.1

Let the component orders be c1,,cr. By [L6], r1, each ci1, and V(G)=ici; step 1.2 gives ω(G)maxici.

step 1.1step 1.2L6choose
3.1

Therefore V(G)=icirmaxiciα(G)ω(G)hom(G)2.

step 2.1step 1.3L1algebra
4.1

Both sides are nonnegative, so [L8] and step 3.1 yield hom(G)V(G)=V(G)1/2. Together with [L2] and [L3], this makes 1/2 an Erdős–Hajnal constant for the P3-free class.

step 3.1L2L3L8
TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

For every n16 there is an n-vertex graph with hom(G)<3log2n

Statement

For every integer n16, there exists an n-vertex graph G such that hom(G)<3log2n.

Facts & Assumptions

Given: An integer n16 and k=3log2n.

[L1]

The homogeneous number of a graph is the larger of its clique and stable-set numbers (Homogeneous vertex sets and the homogeneous number hom(G)=max{ω(G),α(G)}).

[L2]

In G(n,1/2), all possible edges are independent Bernoulli variables on the labelled vertex set [n] (The Erdős-Rényi finite random graph G(n,p)).

[L3]

Prescribing r present edges and s absent edges in G(n,p) has probability pr(1p)s (A prescribed set of present and absent edges in G(n,p) has product probability).

[L4]

Expectations of a finite family of random variables add without an independence hypothesis (Expectation is linear for every finite family of random variables, without any independence hypothesis).

[L5]

If a nonnegative integer-valued random variable on a finite probability space has expectation below 1, some outcome makes it 0 (The first-moment method for avoiding or forcing a finite count of bad events).

[L7]

For b>0 with b1 and x>0, logbx:=logx/logb (The logarithm to a positive base other than one).

[L8]

log:(0,)R is strictly increasing, and for x,y>0 satisfies log(xy)=logx+logy, log(1/x)=logx, and log1=0 (Order, continuity, range, and the product, quotient, and reciprocal laws for the natural logarithm).

Proof

technique · constructive
1.1

By [L8], log2>log1=0, so [L7] makes log2=log/log2 strictly increasing with log21=0. Applying log(xy)=logx+logy to the factors of an integer power, and log(1/x)=logx for a negative exponent, gives log2(2m)=m for every integer m and log2(xj)=jlog2x for every x>0 and every positive integer j.

L7L8algebra
2.1

Since n16=24, step 1.1 gives L:=log2n4, and k=3L satisfies k3L12.

givenstep 1.1algebra
3.1

In G(n,1/2) let X count the k-subsets that induce a clique or a stable set. For a fixed k-subset these two disjoint events each prescribe all (k2) pairs, so [L2] and [L3] give probability 22(k2).

step 2.1L2L3L6
4.1

By [L4] and [L6], EX=2(nk)2k(k1)/22nk2k(k1)/2.

step 3.1L4L6algebra
5.1

By step 1.1 the base-two logarithm of that last bound is 1+kLk(k1)/2=1+k(L(k1)/2)1+3L(1L)/2<0, where the first inequality uses k3L and L4, which make L(k1)/2(1L)/2<0. Since log2 is strictly increasing with log21=0, the bound itself is below 1, so EX<1.

step 1.1step 2.1step 4.1algebra
6.1

By [L5], choose an outcome graph G with X=0. It has no homogeneous k-subset, and any homogeneous set of order at least k would contain one, so hom(G)k1<3L=3log2n.

step 3.1step 5.1L1L5constructdischarge-construct
CorollaryStatement: AI-adaptedProof: AI-generatedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

The hereditary class of all finite graphs does not have the Erdős–Hajnal property

Statement

The hereditary class of all finite graphs does not have the Erdős–Hajnal property.

Facts & Assumptions

Given: The class G of all finite graphs.

[L1]

A hereditary class has the Erdős–Hajnal property exactly when some ϵ>0 satisfies hom(G)V(G)ϵ for every nonempty graph in the class (The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class).

[L2]

For every n16, some n-vertex graph G satisfies hom(G)<3log2n (For every n16 there is an n-vertex graph with hom(G)<3log2n).

[L3]

For every ϵ>0, logx/xϵ0 as x+ (The logarithm grows more slowly than every positive real power).

[L4]

For x>0, log2x=logx/log2 (Change of base and inversion of the positive-base real exponential).

Proof

technique · contradiction
1.1

Suppose, for contradiction, that G has an Erdős–Hajnal constant ϵ>0.

assume-contraL1
1.2

By [L3] and [L4], choose an integer n16 so large that 3log2n<nϵ.

L3L4choose
2.1

Choose from [L2] an n-vertex graph G with hom(G)<3log2n<nϵ, contradicting [L1] and step 1.1. Therefore G does not have the Erdős–Hajnal property.

step 1.1step 1.2L1L2discharge-contradiction
PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-16Open item page →

A hereditary class has the Erdős–Hajnal property exactly when its complementary class does, with the same constants

Statement

Let C be a hereditary graph class and let C be its complement class. Then C has the Erdős–Hajnal property if and only if C does. More precisely, the two classes have exactly the same Erdős–Hajnal constants. Consequently a graph H and its complement H have the same Erdős–Hajnal constants.

Facts & Assumptions

Given: A hereditary graph class C.

[L1]

An exponent ϵ>0 is an Erdős–Hajnal constant for a hereditary class when every nonempty member G satisfies hom(G)V(G)ϵ (The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class).

[L2]

The complement class is C={G:GC} (The complement of a graph class).

[L4]

Complementation exchanges cliques and stable sets, so ω(G)=α(G) and α(G)=ω(G) (Complementation swaps cliques with stable sets, so ω(G)=α(G)).

[L5]

A graph G is H-free if and only if G is H-free (G is H-free if and only if G is H-free).

Proof

technique · direct
1.1

By [L3], both classes in the statement are hereditary, and [L4] gives hom(G)=hom(G) for every G.

L3L4
2.1

Let ϵ be a constant for C and let FC be nonempty. Then FC by [L2], while V(F)=V(F) and hom(F)=hom(F) by step 1.1, so [L1] gives hom(F)V(F)ϵ.

step 1.1L1L2
3.1

Thus every constant of C is a constant of C; applying the same argument to C and using G=G gives the reverse inclusion of constant sets.

step 2.1L2
4.1

By [L5], complementation bijects the H-free class with the H-free class, so step 3.1 gives the fixed-pattern consequence.

step 3.1L5
CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-16Open item page →

Every graph on at most three vertices has the Erdős–Hajnal property

Statement

Every finite graph H with V(H)3 has the Erdős–Hajnal property.

Facts & Assumptions

Given: A finite graph H with at most three vertices.

[L2]

For every t1, the class of Kt-free graphs has the Erdős–Hajnal property (For every t1, the class of Kt-free graphs has the Erdős–Hajnal property).

[L3]

Every P3-free graph G satisfies hom(G)V(G), so P3 has the property (Every P3-free graph G satisfies hom(G)V(G)).

[L4]
[L5]

The graphs Kt and P3 have the standard edge sets, and P0 is the null graph (Empty and complete graphs, complete bipartite graphs, and the convention that Pn and Cn have n vertices); complementation replaces the edge set by all missing pairs (Graph isomorphisms, automorphisms and graph complements).

Proof

technique · cases
1.1

[assume-case null] If V(H)=0, every graph contains the unique empty induced embedding of H, so the H-free class has no members and [L1] is vacuously satisfied by every positive exponent.

L1
1.2

[assume-case nonnull] Suppose 1V(H)3. Up to isomorphism and complementation, H is one of K1,K2,K3, or P3: this follows by the edge count for orders at most two, and for order three by separating the cases of zero, one, two, or three edges.

L5algebra
2.1

Each complete case has the property by [L2], the path case has it by [L3], and every complementary case has it by [L4].

step 1.2L2L3L4
3.1

The cases are exhaustive, so every graph on at most three vertices has the Erdős–Hajnal property.

step 1.1step 2.1cases-exhaustive
PropositionStatement: AI-adaptedProof: AI-generatedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

The Erdős–Hajnal property and each of its constants pass to hereditary subclasses

Statement

If DC are hereditary graph classes, then every Erdős–Hajnal constant for C is one for D. In particular, the Erdős–Hajnal property passes from C to D.

Facts & Assumptions

Given: Hereditary graph classes DC and an Erdős–Hajnal constant ϵ for C.

[L1]

The constant condition says that every nonempty G in the class satisfies hom(G)V(G)ϵ (The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class).

Proof

technique · direct
1.1

Every nonempty GD also lies in C, so [L1] gives hom(G)V(G)ϵ.

givenL1
2.1

Thus ϵ is a constant for D; the existence assertion follows by retaining any constant of C.

step 1.1L1
PropositionStatement: AI-adaptedProof: AI-generatedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

If H is an induced subgraph of H and H has the Erdős–Hajnal property, then H has it with every constant of H

Statement

Suppose H has an induced embedding into H. Then every Erdős–Hajnal constant for H is one for H. Consequently, if H has the Erdős–Hajnal property, then so does H.

Facts & Assumptions

Given: Finite graphs H,H and an induced embedding HindH.

[L1]

Every Erdős–Hajnal constant passes from a hereditary class to any hereditary subclass (The Erdős–Hajnal property and each of its constants pass to hereditary subclasses).

[L2]

A graph is F-free when it has no induced embedding of F (H-free and F-free graphs under the induced-subgraph convention).

[L3]

Induced embeddings compose, so induced-subgraph containment is transitive (Induced embeddings compose, and the induced-subgraph relation is transitive up to isomorphism).

[L4]

Every fixed-pattern-free graph class is hereditary (Every class defined by forbidden induced subgraphs is hereditary).

Proof

technique · direct
1.1

If G is H-free, then it is H-free: an induced embedding HindG would compose with the Given embedding to put H inducedly in G.

givenL2L3
2.1

Hence the H-free class is a subclass of the H-free class, and both are hereditary by [L4].

step 1.1L4
3.1

Applying [L1] proves that every constant of H is a constant of H, and therefore proves the property implication.

step 2.1L1
TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

The single-forbidden-graph and finite-nonempty-family formulations of the Erdős–Hajnal conjecture are equivalent

Statement

The following assertions are equivalent:

  1. every finite graph H has the Erdős–Hajnal property;
  2. for every finite nonempty family F of finite graphs, the hereditary class of F-free graphs has the Erdős–Hajnal property.

Facts & Assumptions

Given: The two universally quantified assertions in the Statement.

[L1]

The Erdős–Hajnal property of a graph H is the property of its H-free class (The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class).

[L2]

A graph is F-free exactly when it is H-free for every HF (H-free and F-free graphs under the induced-subgraph convention).

[L3]

The Erdős–Hajnal property passes to hereditary subclasses (The Erdős–Hajnal property and each of its constants pass to hereditary subclasses), and every family-free class is hereditary (Every class defined by forbidden induced subgraphs is hereditary).

Proof

technique · direct
1.1

Assume assertion 1, let F be finite and nonempty, and choose HF. By [L2], every F-free graph is H-free.

givenL2choose
1.2

Conversely, assume assertion 2 and let H be any finite graph. Applying assertion 2 to the finite nonempty family {H} gives the property for the H-free class, which is assertion 1 by [L1] and [L2].

givenL1L2
2.1

The H-free class has the property by assertion 1 and [L1], so its hereditary subclass of F-free graphs has it by [L3]. This proves assertion 2.

step 1.1L1L3
3.1

The two implications prove the equivalence.

step 2.1step 1.2
RemarkRemark: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-16Open item page →

The Erdős–Hajnal conjecture: every fixed forbidden induced graph admits a positive exponent

The Erdős–Hajnal conjecture asserts that every finite graph H has the Erdős–Hajnal property: equivalently, for each H there is an exponent ϵ(H)>0 such that every nonempty H-free graph G satisfies hom(G)V(G)ϵ(H) (The Erdős–Hajnal property and an Erdős–Hajnal constant for a hereditary graph class, H-free and F-free graphs under the induced-subgraph convention).

5 · Examples, counterexamples and false statements

None yet.

Sources