Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Statement

Notation as on the definitions page.

Theorem 3 (p. 362, quoted). "Assume G. C. H. (generalized continuum hypothesis). Let α\alpha be infinite, α≥β\alpha\ge\beta, α>γ\alpha>\gamma, β>δ≥2\beta>\delta\ge2 then [α,β]↛[γ,δ][\alpha,\beta]\not\to[\gamma,\delta]."

The paper calls this, in view of Section 2, a best possible negative result that settles all the problems concerning the vertex-decomposition symbol (p. 362).

Corollary 2 (p. 362). Assume GCH, α≥ω\alpha\ge\omega, α≥β\alpha\ge\beta and β>δ≥2\beta>\delta\ge2. Then there is a graph G\mathcal G with α(G)=α\alpha(\mathcal G)=\alpha and β(G)≤β\beta(\mathcal G)\le\beta such that for every γ<α\gamma<\alpha and every vertex-decomposition Gξ\mathcal G_\xi, ξ<γ\xi<\gamma, of type γ\gamma some member has β(Gξ)>δ\beta(\mathcal G_\xi)>\delta. It follows from Theorem 3 and 2.8 (p. 362).

Corollary 3 (p. 363). For all finite γ\gamma and δ\delta with γ≥2\gamma\ge2, δ≥2\delta\ge2 there is αγ,δ<ω\alpha_{\gamma,\delta}<\omega with [αγ,δ,δ+1]↛[γ,δ][\alpha_{\gamma,\delta},\delta+1]\not\to[\gamma,\delta]. The paper says this is implied by the case α=ω\alpha=\omega of Theorem 3, was proved earlier by Erdős and Rogers (its reference [6]) with a good estimate for αγ,δ\alpha_{\gamma,\delta}, and returns to it in Section 4.

Source. P. Erdős and A. Hajnal, On decomposition of graphs, Acta Math. Acad. Sci. Hungar. 18 (1967), 359--377, doi:10.1007/BF02280296; the edition read is named on the source card.

Read depth. Claims checked: Theorem 3, Corollaries 2 and 3, 2.8 and the proof of Theorem 3 (pp. 362--363) were read clause by clause on the page images. The proofs of Theorems 1 and 2, which it uses, were followed in outline only. Nothing here is independently reviewed.

Proof pointer

Pp. 362--363. Since β>δ\beta>\delta, monotonicity reduces Theorem 3 to [α,δ+]↛[γ,δ][\alpha,\delta^+]\not\to[\gamma,\delta] for α≥δ+\alpha\ge\delta^+, α>γ\alpha>\gamma. For regular α\alpha, GCH gives αδ=α\alpha^\delta=\alpha and the claim follows from Theorem 1 (finite δ\delta, where δ+=δ+1\delta^+=\delta+1) and Theorem 2 (infinite δ\delta). For singular α\alpha, α>δ+\alpha>\delta^+ and there is a regular α′\alpha' with max⁡(γ,δ+)≤α′<α\max(\gamma,\delta^+)\le\alpha'<\alpha; the relation for α′\alpha' transfers to α\alpha by monotonicity.

Dependencies

Theorem 1, Theorem 2 and 2.8 of the same paper; GCH as a hypothesis.

Bears on

None directly among the problems this corpus records. Corollary 3 is the vertex-decomposition input to Pósa's edge-decomposition result recorded on the display (1) page.