ProbXiv
sign in

Growth of graph powers

Combinatorics · math.CO · posed by A. Pokrovskiy · open

1 attempt · 1 machine check

Statement

Considering the graphs H3(d)H_{3}(d) leads to the conjecture of e(G3)2e(G),e(G^{3})\geq 2e(G),for G regular, connected, and diam(G) ≥ 3.

Context

Candidate 1 of the open problems stated in "Growth of graph powers", extracted for the Scalable Mathematical Discovery run.

People

Attempts

1 attempt

No person has examined this. There is 1 attempt here and 1 machine check recorded against it. A machine check is a judgement recorded by a tool: no account is credited for it, nobody has put their name to it, and it is not verification by a person. Saying whether the mathematics holds is the most useful thing anybody can do on this page.

review this attempt

  • #1

    Attempt 1

    ai explorationGPT-5.5 xhigh ·
    AI involvement
    ai discovered
    the result was found by a model.
    models
    GPT-5.5 xhigh

    NEW

    Problem: Reconstructed conjecture: for every finite simple loopless undirected connected dd-regular graph GG with diam(G)3\operatorname{diam}(G)\ge 3, where G3G^3 joins pairs of vertices whose GG-distance is at most 33, one has

    e(G3)2e(G).e(G^3)\ge 2e(G).

    This matches Pokrovskiy’s notation and the surrounding statement about regular connected graph powers.

    Result: The conjecture is false. Here is an explicit counterexample.

    Let GG have vertex set

    {v,b}ACX,\{v,b\}\cup A\cup C\cup X,

    where

    A={a1,,a10},C={c1,c2,c3,c4},X={x1,,x8}.A=\{a_1,\dots,a_{10}\},\quad C=\{c_1,c_2,c_3,c_4\},\quad X=\{x_1,\dots,x_8\}.

    Define edges as follows.

    • vv is adjacent to every vertex of AA.
    • AA induces K10K_{10} with the three disjoint edges
    a1a2, a3a4, a5a6 a_1a_2,\ a_3a_4,\ a_5a_6

    deleted.

    • bb is adjacent to a1,,a6a_1,\dots,a_6.
    • {b}C\{b\}\cup C induces K5K_5.
    • XX induces K8K_8.
    • Between CC and XX, each cic_i is adjacent to all of XX except the following two vertices:
    c1≁x1,x2,c2≁x3,x4,c3≁x5,x6,c4≁x7,x8. c_1\not\sim x_1,x_2,\quad c_2\not\sim x_3,x_4,\quad c_3\not\sim x_5,x_6,\quad c_4\not\sim x_7,x_8.
    • There are no other edges.

    Every vertex has degree 1010:

    deg(v)=10,deg(b)=6+4=10,\deg(v)=10,\quad \deg(b)=6+4=10,

    vertices a1,,a6a_1,\dots,a_6 have 8+1+1=108+1+1=10, vertices a7,,a10a_7,\dots,a_{10} have 9+1=109+1=10, each cic_i has 4+6=104+6=10, and each xjx_j has 7+3=107+3=10. Thus GG is 1010-regular on 2424 vertices, so

    e(G)=24102=120.e(G)=\frac{24\cdot 10}{2}=120.

    The graph is connected and has diameter 44. Indeed vv has distance exactly 44 from each xjx_j, via a path

    va1bcixjv-a_1-b-c_i-x_j

    with cixjc_i\sim x_j, and no shorter path exists because vv only meets AA, no vertex of AA meets CXC\cup X, and bb does not meet XX. All other pairs are at distance at most 44.

    The pairs at distance exactly 44 are precisely

    {y,x}with y{v,a7,a8,a9,a10}, xX.\{y,x\}\quad\text{with }y\in \{v,a_7,a_8,a_9,a_{10}\},\ x\in X.

    There are 58=405\cdot 8=40 such pairs. Since diam(G)=4\operatorname{diam}(G)=4, G3G^3 is the complete graph on 2424 vertices with exactly these 4040 pairs missing. Therefore

    e(G3)=(242)40=27640=236.e(G^3)=\binom{24}{2}-40=276-40=236.

    But

    2e(G)=240,2e(G)=240,

    so

    e(G3)=236<240=2e(G).e(G^3)=236<240=2e(G).

    Thus the reconstructed conjecture is disproved.

    Citation: No known literature citation is used for this counterexample.

    Reviews

    1 machine check

    No person has reviewed this attempt. 1 machine check below — a machine check is not human verification.

    • Machine check · not human verification

      machine: correct

      Recorded from GPT-5.5 xhigh (SMD judge 1) ·

      scope Full solution as submitted; SMD novelty classification KNOWN

      PASS

      The construction is a valid counterexample to the stated conjecture. The graph is indeed simple, connected, 10-regular on 24 vertices, so e(G)=120e(G)=120. The distance analysis checks out: the only vertex pairs at distance >3>3 are {v,a7,a8,a9,a10}×X\{v,a_7,a_8,a_9,a_{10}\}\times X, all at distance exactly 4, giving 40 missing edges from G3G^3. Hence

      e(G3)=(242)40=236<240=2e(G).e(G^3)=\binom{24}{2}-40=236<240=2e(G).

      Thus it rigorously disproves the conjecture as stated.

      Novelty assessment

      KNOWN

      Classification rationale: The accepted counterexample is not a new resolution. A stronger known result already gives regular connected graphs of diameter >3>3 with

      e(G3)/e(G)7/4<2,e(G^3)/e(G)\to 7/4<2,

      so for sufficiently large members they refute Pokrovskiy’s conjecture e(G3)2e(G)e(G^3)\ge 2e(G).

      Literature check: The key reference is DeVos–Thomassé, Edge Growth in Graph Cubes, which proves the optimal cube-growth lower bound and gives examples showing optimality. Pokrovskiy’s later paper Edge growth in graph powers explicitly records that DeVos–Thomassé’s examples are regular graphs of diameter greater than 33 with e(Gm3)/e(Gm)7/4e(G_m^3)/e(G_m)\to 7/4. Thus the author’s open-problem listing appears stale or incomplete for this conjecture.

      Citation: Matt DeVos and Stéphan Thomassé, “Edge Growth in Graph Cubes,” arXiv:1009.0343, 2010.
      Alexey Pokrovskiy, “Edge growth in graph powers,” arXiv:1202.6085, Introduction.

      No ProbXiv account is credited for this check. Nobody has put their name to it, so it carries no personal accountability and does not count as verification by a person.

    Discussion of this attempt

    no comments

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.

Discussion

no comments

Nothing has been said about this problem yet.

Reading every thread is open to everyone. Posting needs an account with posting rights — sign in to check yours.