The Nordhaus-Gaddum upper-bound conjecture for the Cheeger constant
Let be a graph, let be its complement, and let denote the Cheeger constant of . Write for the star graph on vertices, and let denote the join of the complete graph on three vertices and the edgeless graph on vertices.
Cheeger-constant Nordhaus-Gaddum conjecture. If , then
The conjecture is motivated by computational and structural comparisons among graph families, but the supplied text gives no resolution or status evidence beyond presenting it as a conjecture.
References
Primary source
Mark Kempton, Xavier Zaitzeff and Sibi Muthuprakash, “Nordhaus-Gaddum upper bounds for graph connectivity parameters”, arXiv:2606.12751 (2026).
Additional references
4 papers in this index state this conjecture (2018–2026). The statement above is taken from the most recent of them; the others are arXiv:2206.03723, arXiv:1808.05576, arXiv:1807.06436.
Progress summary
A reader-submitted calculation claims the conjecture is false on seven vertices, but no independent verification was found.
Kempton, Zaitzeff, and Muthuprakash state this as Conjecture 1: excluding the star, the larger Cheeger constant of a graph and its complement should not exceed that of the proposed extremal graph. Their paper presents computational and structural motivation but no resolution.
Community submission (unverified), August 23, 2026
A submitted calculation claims a counterexample at seven vertices within the generalized-star family, taking . It reports an exhaustive cut-type computation yielding and asserts that this exceeds the conjectured benchmark . The argument is not independently verified.
Current status (as of August 2026): The conjecture has an unverified claimed counterexample at ; no independently confirmed proof or disproof is recorded.
Sources
Solutions 1
CounterexampleThis solution needs a summarySee full solution
A seven-vertex generalized-star counterexample
Conjecture 1 of Kempton, Zaitzeff, and Muthuprakash, Nordhaus–Gaddum upper bounds for graph connectivity parameters, asserts that every graph satisfies
where
The assertion already fails within the generalized-star family discussed in the paper.
Cut formula for generalized stars
Let
and write and for its clique and independent-set parts. For any vertex subset , put
The clique vertices have degree , and the independent-set vertices have degree . Consequently,
The cut ratio is unchanged when is replaced by . It therefore suffices to list one representative from each complementary pair of nonempty proper cut types.
The counterexample
Take
Formula (3) gives the following complete list, where denotes the smaller of the two volumes:
Hence
On the other hand, the proposed seven-vertex extremal graph is
Its complete list of complementary cut types is
Therefore
The graph is neither a star nor a complete graph. Its complement is
which is disconnected, so . Combining (4) and (5),
Thus the conjectured bound is false. In particular, the proposed extremizer is not maximal even among seven-vertex generalized stars. The disconnected complement is permitted by the conjecture as stated; indeed, the proposed extremizer itself has disconnected complement.