Wiki
Wiki

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

Updated

Griffin 2013 minimal pancyclicity

../

claim_1: Bondy's bounds on the minimum size of a pancyclic graph, stated by Bondy without proof according to Griffin, who proves the lower bound from Shi's bound on the number of cycles of a Hamiltonian graph with k chords.

conjecture_1: The paper conjectures that the least number of edges of a pancyclic graph strictly increases from n to n + 1 vertices, for every n at least 3.

corollary_1: The number of cycles in a Hamiltonian graph with k chords is at most 2^{k+1} - 1, the count from Shi's theorem on which the paper's lower bound for m(n) and its exhaustive search rest.

corollary_2: For n > 6, a minimal pancyclic graph on n vertices with an arc of length at least (n-1)/2 gives m(n-1) < m(n), and one with an arc of length at least (n+2)/3 gives m(n-1) <= m(n); the first is a special case of Conjecture 1.

proposition_2: The least number of edges of a pancyclic graph grows by at most two from n to n + 1 vertices, for every n at least 3.

table_1: The exact values of m(n), the least number of edges of a pancyclic graph on n vertices, for n up to 37, from an exhaustive search over Hamiltonian graphs with few chords and a five-chord construction; the values agree with George, Marr and Wallis.

theorem_3: Quotes Rautenbach and Stella's upper bound on the maximum number M(k) of cycles in a Hamiltonian graph with k chords and concludes that m(n) is at least n + C, for C the largest integer k at which the bound is below n - 2.

theorem_4: For a pancyclic graph on more than six vertices, contracting one edge of an arc of length at least (n-1)/2 with no chord joining its ends, or of length at least (n+2)/3 with such a chord, leaves a pancyclic graph.


Sean Griffin, Minimal Pancyclicity. arXiv:1312.0274 (2013). The site's reference key Gr13.

Edition read. The copy read for this card is arXiv:1312.0274v1 (1 December 2013; dated September 6, 2013 on its first page; 6 pages), the only arXiv version, with no journal reference on arXiv and no journal record found by a Crossref bibliographic query on 2026-09-18: a preprint. A text-layer PDF, all of whose pages (pp. 1--6) were read on the rendered page images. Source: https://arxiv.org/abs/1312.0274. The arXiv record names arXiv's non-exclusive distribution license (arXiv:1312.0274), every other right reserved.

Read status: claims checked for Table 1 (p. 2), Corollary 1 (p. 2), Claim 1 with its lower-bound proof (p. 2), Theorem 3 as a statement (p. 3), Proposition 2 with its proof (p. 3), Conjecture 1 (p. 3), Theorem 4 (p. 4) and Corollary 2 (p. 5), read clause by clause on the page images; the exhaustive search behind Table 1 was not rerun, the five-chord construction (Figure 1) was not checked, and the proof of Theorem 4 was read for structure only. Propositions 1, 3 and 4 and Corollary 3 (pp. 3--5) were read on the page images but are not paged. Problem 1016 consumes Table 1 and Claim 1, paged at table_1 and claim_1, and names Theorem 3, Proposition 2 and Conjecture 1.

Writing m(n) for the minimum number of edges of a pancyclic graph on n vertices (one having a cycle of every length from 3 to n), the paper computes m(n) exactly for all n at most 37, combining an exhaustive computer search over Hamiltonian graphs with at most four chords, and with five chords on at most 31 vertices (p. 2; the abstract says graphs with up to 29 vertices), with a new explicit five-chord construction, captioned "Construction with 23 to 37 vertices" (Figure 1, p. 1; Table 1). The search is made feasible by Shi's theorem (Theorem 2) that a given chord set admits at most two cycles using exactly those chords, giving at most 2^{k+1}-1 cycles for k chords (Corollary 1), which together with the search rules out pancyclic graphs with four chords on 25 or more vertices. Section 1.1 states Bondy's bounds n + log_2(n-1) - 1 <= m(n) <= n + log_2 n + H(n) + O(1) (Claim 1), says that Bondy states them without proof, and proves the lower bound from Corollary 1. Section 1.3 studies the growth of m(n): Proposition 2 shows m(n+1) <= m(n)+2, and Conjecture 1 conjectures that m(n) < m(n+1) for all n >= 3; Theorem 4 with Corollary 2 proves this in special cases, for instance when some minimal pancyclic graph on n vertices has a long arc along its Hamiltonian cycle, and Proposition 4 with Corollary 3 gives a conditional alternative for an arc of length n/2 - 1. Table 1, the lower bound and the monotonicity conjecture are the paper's content for problem 1016 on the minimum size of a pancyclic graph.

Bears on. #1016: in the site's notation h(n) = m(n) - n. Table 1 gives h(n) for n <= 37, and Claim 1 gives the proved lower bound h(n) >= log_2(n-1) - 1 (the site's "A proof of the above lower bound is provided by Griffin") and records Bondy's unproved upper bound log_2 n + H(n) + O(1), the site's log_* n form. Corollary 1 is the cycle count behind that lower bound, and Theorem 3 gives an implicit lower bound on h(n) from Rautenbach and Stella's sharper count, still of the form log_2 n + O(1). Proposition 2 gives h(n+1) <= h(n) + 1 for n >= 3, Conjecture 1 conjectures that h is nondecreasing, and Corollary 2, from Theorem 4, proves h(n-1) <= h(n) when some minimal pancyclic graph on n > 6 vertices has an arc of length at least (n-1)/2; these concern the step-by-step growth of h, not the asymptotic question. A preprint.

Results paged.

  • Table 1 (p. 2): exact values of m(n) for 3 <= n <= 37, with the chord number k, agreeing with the preprint of George, Marr and Wallis.
  • Corollary 1 (p. 2): a Hamiltonian graph with k chords has at most 2^{k+1} - 1 cycles, from Shi's Theorem 2.
  • Claim 1 (Bondy; p. 2): n + log_2(n-1) - 1 <= m(n) <= n + log_2(n) + H(n) + O(1), where H(n) is the smallest integer such that log_2 applied H(n) times to n gives a value below 2; stated by Bondy without proof according to the paper, which proves the lower bound from Corollary 1.
  • Theorem 3 (Rautenbach and Stella; p. 3): an upper bound on the maximum number M(k) of cycles of a Hamiltonian graph with k chords, giving m(n) >= n + C for C the largest integer k at which that bound is less than n - 2 (the print says "the expression in Theorem 1").
  • Proposition 2 (p. 3): m(n+1) <= m(n) + 2 for all n >= 3.
  • Conjecture 1 (p. 3): m(n) < m(n+1) for all n >= 3.
  • Theorem 4 (p. 4): for a pancyclic graph on n > 6 vertices, contracting one edge of an arc A gives a pancyclic graph when no chord joins the ends of A and |A| >= (n-1)/2, or when a chord joins them and |A| >= (n+2)/3.
  • Corollary 2 (p. 5): for n > 6, m(n-1) < m(n) if some minimal pancyclic graph on n vertices has an arc of length at least (n-1)/2, and m(n-1) <= m(n) if it has one of length at least (n+2)/3.

Not paged: the five-chord construction (Figure 1, p. 1), which the paper says has cycles of lengths 3 to 19 and n - 17 to n on n = 21 + x vertices, and which Table 1 uses for 25 <= n <= 37; Proposition 1 (p. 3), a bound on the maximum degree of a minimal pancyclic graph; Proposition 3 (p. 4), a condition under which a pancyclic graph is not minimal; and Proposition 4 with Corollary 3 (p. 5), on arcs of length n/2 - 1.

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