Wiki
Wiki

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

Updated

Chen 1994 clique partitions split graphs

../

conjecture_p28: Chen, Erdős and Ordman's conjecture that the edges of every split graph on n vertices can be partitioned into at most n^2/6 + n/6 cliques, stated after they found no split graph needing more, with their Example 2 showing that deleting connecting edges can raise the number of cliques needed.

example_1: The split graph with n/3 clique vertices joined completely to 2n/3 independent vertices needs exactly n^2/6 + n/6 cliques to partition its edges when 6 divides n; the paper builds the partition and cites earlier work for its minimality.

theorem_1: Chen, Erdős and Ordman's bound that the edges of every split graph on n vertices can be partitioned into at most (3/16)n^2 + O(n) cliques, assembled from five bounds by the fraction r of vertices in the clique (Lemmas 1 to 5), with the same bound for threshold graphs (Corollary 1).

theorem_2: Chen, Erdős and Ordman's theorem that the complement of a clique, K_n with the edges of a K_m removed (an m-vertex independent set joined completely to an (n − m)-clique), has a clique partition into at most n^2/6 + O(n) cliques.


G.-T. Chen, P. Erdős and E. T. Ordman, Clique partitions of split graphs, in: Y. Alavi, D. R. Lick and J. Liu (eds.), Combinatorics, Graph Theory, Algorithms and Applications (Beijing, 1--5 June 1993), World Scientific, Singapore, 1994, pp. 21--30 (ISBN 981-02-1855-9). The work was done at Memphis State University (footnote, p. 21).

The copy read for this card is an image-only scan (EPSON Scan, no text layer) of six landscape PDF pages, each holding two printed pages: PDF p. 1 holds pp. 21--22 under a handwritten note giving the volume, its editors, publisher and conference; PDF pp. 2--5 hold pp. 23--24, 25--26, 27--28 and 29--30; PDF p. 6 holds the volume's title page and imprint page. Everything below was read on the page images. The title and authors on p. 21 identify the work as this paper, the #81 problem page's [CEO94], and not the chordal-graph paper of Erdős, Ordman and Zalcstein, [EOZ93] on that page. Provenance: the copy was downloaded in September 2026; the download URL was not recorded. 430,411 bytes. The scan prints "Copyright © 1994 by World Scientific Publishing Co. Pte. Ltd. All rights reserved." on the volume's imprint page (PDF p. 6, read on the page image), going on to say that the book or parts of it may not be reproduced in any form or by any means without written permission from the publisher, every other right reserved.

Read status: claims checked for Lemmas 1--5, Theorem 1, Corollary 1 and Theorem 2 (p. 23), Example 1 (p. 22), and the conjecture and Example 2 (p. 28), read on the page images; the proofs (pp. 24--27) and the argument for Example 2 (pp. 28--29) were read for their structure only and not checked step by step. No problem page states a result from this source yet.

Contents

  • Definitions (p. 21): a clique partition of GG is a set of cliques containing each edge exactly once, and cp⁡(G)\operatorname{cp}(G) is its least size; a graph is split if its vertices divide into a clique AA and an independent set BB. Every threshold graph is split and every split graph is chordal (p. 22).
  • Context (p. 22): from [9], Erdős, Ordman and Zalcstein's chordal-graph paper, a chordal graph GnG_n has cp⁡(Gn)≤n2(1/4−c)\operatorname{cp}(G_n)\le n^2(1/4-c) for some c>0c>0, the value of cc unknown, and cp⁡\operatorname{cp} can exceed n2/6n^2/6 by at least O(n)O(n). Example 1 (p. 22): cp⁡(Kn−Kˉ2n/3)=n2/6+n/6\operatorname{cp}(K_n-\bar K_{2n/3})=n^2/6+n/6 when 6∣n6\mid n, where Kn−KˉmK_n-\bar K_m is the split graph with an mm-vertex independent set joined completely to an (n−m)(n-m)-clique.
  • Lemmas 1--5 (p. 23), for a split graph GnG_n with rnrn vertices in the clique and (1−r)n(1-r)n in the independent set: cp⁡(Gn)≤(r−32r2)n2+O(n)≤n2/6+O(n)\operatorname{cp}(G_n)\le(r-\tfrac32r^2)n^2+O(n)\le n^2/6+O(n) for 0≤r≤1/30\le r\le1/3; ≤34(r−r2)n2+O(n)≤316n2+O(n)\le\tfrac34(r-r^2)n^2+O(n)\le\tfrac3{16}n^2+O(n) for 1/3≤r≤1/21/3\le r\le1/2 and for 1/2≤r≤2/31/2\le r\le2/3; ≤(r2−38r2)n2+O(n)≤n2/6+O(n)\le(\tfrac r2-\tfrac38r^2)n^2+O(n)\le n^2/6+O(n) for 2/3≤r≤4/52/3\le r\le4/5; ≤(r−r2)n2+O(n)≤425n2+O(n)\le(r-r^2)n^2+O(n)\le\tfrac4{25}n^2+O(n) for 4/5≤r≤14/5\le r\le1 (printed "425n2/6+O(n)\tfrac4{25}n^2/6+O(n)", a misprint: r−r2≤425r-r^2\le\tfrac4{25} on that range, and §4.1 (p. 27) recalls, for r=4/5r=4/5, "a covering by about 425n2\tfrac4{25}n^2 cliques").
  • Theorem 1 (p. 23): for all split graphs GnG_n, cp⁡(Gn)≤316n2+O(n)\operatorname{cp}(G_n)\le\tfrac3{16}n^2+O(n). Corollary 1: the same bound for threshold graphs, improving [9]. Theorem 2 (p. 23): a graph of the form Kn−KˉmK_n-\bar K_m has clique partition number at most n2/6+O(n)n^2/6+O(n).
  • Section 4, remarks (pp. 27--30): the large-clique case and its relation to resolvable block designs (4.1); the case r=1/2r=1/2 with a quarter of the connecting edges missing, where Example 2 (p. 28) shows that deleting connecting edges from Kn−Kˉn/2K_n-\bar K_{n/2} can force at least (18+1128)n2(\tfrac18+\tfrac1{128})n^2 cliques although about n2/8n^2/8 suffice for the full graph; and the authors' conjecture (p. 28) that n2/6+n/6n^2/6+n/6 cliques always suffice for split graphs, no example requiring more being known to them.

Compiled scope

All printed pages were viewed on the page images; the statements above were read on pp. 21--23 and 27--30, and the proofs were read for their structure only. Nothing here is independently reviewed.

Bears on. #81, which asks whether every chordal graph on nn vertices has a clique partition into n2/6+O(n)n^2/6+O(n) cliques: split graphs are chordal (p. 22), so the paper treats a subclass of the problem's graphs. For split graphs Theorem 1 gives 316n2+O(n)\tfrac3{16}n^2+O(n), a larger constant than the problem asks for; Theorem 2 gives n2/6+O(n)n^2/6+O(n) for the split graphs Kn−KˉmK_n-\bar K_m with every connecting edge present; Example 1 shows that n2/6+n/6n^2/6+n/6 cliques can be needed, so the constant 16\tfrac16 cannot be lowered; and the p. 28 conjecture asks for n2/6+n/6n^2/6+n/6 for every split graph. The paper also restates (p. 22) the chordal bound n2(1/4−c)n^2(1/4-c) of the problem page's [EOZ93].

Results.

  • Theorem 1, p. 23: every split graph GnG_n has cp⁡(Gn)≤316n2+O(n)\operatorname{cp}(G_n)\le\tfrac3{16}n^2+O(n), from Lemmas 1--5 (p. 23), with Corollary 1 (p. 23) for threshold graphs.
  • Theorem 2, p. 23: cp⁡(Kn−Kˉm)≤n2/6+O(n)\operatorname{cp}(K_n-\bar K_m)\le n^2/6+O(n).
  • Example 1, p. 22: cp⁡(Kn−Kˉ2n/3)=n2/6+n/6\operatorname{cp}(K_n-\bar K_{2n/3})=n^2/6+n/6 when 6∣n6\mid n, the minimality cited to the paper's references [7] and [14].
  • Conjecture, p. 28: n2/6+n/6n^2/6+n/6 cliques always suffice for split graphs; posed, not proved, with Example 2 (p. 28).

No file of this source is held: no license on record permits its redistribution, and the card cites the edition it names above.