Proved theorem Unrefereed

Lonely runner: an exact five-speed family

No dated manuscript catalogued; standalone verification archive linked above. On this site since · last revised · Verify & co-author
Kind
Theorem with a written proof
Status
Proof held by a falsifier agent, with exact values checked by computer; unrefereed
Prior work
Kravitz's Lemma 9.4 gives it once 5s exceeds 1,372; no source read states it for small s
On this page
  1. The statement
  2. Check it yourself
  3. Worked examples
  4. Status and who checked it
  5. Prior work
  6. Further reading

The statement

Write ∥x∥\|x\| for the distance from a real number xx to the nearest integer. For distinct positive integer speeds v=(v1,…,vk)v = (v_1,\allowbreak \ldots,\allowbreak v_k), the loneliness gap is

δ(v)=max⁡tmin⁡i∥t vi∥.\delta(v) = \max_t \min_i \|t\,v_i\|.

The lonely runner conjecture says δ(v)≥1/(k+1)\delta(v) \ge 1/(k + 1) for every such vv. A tuple lies in the near-tight band when 1/(k+1)≤δ(v)<1/k1/(k + 1) \le \delta(v) < 1/k; at k=5k = 5 the band is 1/6≤δ<1/51/6 \le \delta < 1/5.

Theorem. For every s≥2s \ge 2,

δ(1,3,4,7,5s)=s5s+1.\delta(1, 3, 4, 7, 5s) = \frac{s}{5s + 1}.

For s≥3s \ge 3 it is attained only at t=s/(5s+1)t = s/(5s + 1) and 1−t1 - t; at s=2s = 2 also at 5/115/11 and 6/116/11. At s=1s = 1 the formula fails: δ(1,3,4,5,7)=2/11\delta(1,\allowbreak 3,\allowbreak 4,\allowbreak 5,\allowbreak 7) = 2/11, still in the band.

Check it yourself

The time that attains it. Put θ=s/(5s+1)\theta = s/(5s + 1) and t=θt = \theta. Then ∥t∥=θ\|t\| = \theta and 5st=s−θ5st = s - \theta, so ∥5st∥=θ\|5st\| = \theta, while

∥3t∥=2s+15s+1,∥4t∥=s+15s+1,∥7t∥=2s−15s+1,\|3t\| = \frac{2s + 1}{5s + 1}, \quad \|4t\| = \frac{s + 1}{5s + 1}, \quad \|7t\| = \frac{2s - 1}{5s + 1},

each at least θ\theta. At s=2s = 2, t=2/11t = 2/11 puts the five runners at distances 2/112/11, 5/115/11, 3/113/11, 3/113/11 and 2/112/11 from the start.

Why no time does better. For θ≥2/11\theta \ge 2/11 the four slow speeds 1, 3, 4, 7 all stay at least θ\theta from the start only on two short pieces of [0,1/2][0,\allowbreak 1/2]: [θ,(1−θ)/4][\theta,\allowbreak (1 - \theta)/4] and [(1+θ)/3,(3−θ)/7][(1 + \theta)/3,\allowbreak (3 - \theta)/7] (plus the point 5/115/11 when θ=2/11\theta = 2/11). On the earlier piece the fast runner 5s5s falls from distance θ\theta to 0 and then rises only to s/(4(5s+1))<θs/(4(5s + 1)) < \theta; the later piece lies within 1/(15(5s+1))1/(15(5s + 1)) of 2/52/5, where 5st5st is the integer 2s2s, so there the fast runner is within s/(3(5s+1))<θs/(3(5s + 1)) < \theta of the start.

Worked examples

  • The family beside it. Kravitz's Theorem 5.1 gives δ(1,2,3,4,5s)=s/(5s+1)\delta(1,\allowbreak 2,\allowbreak 3,\allowbreak 4,\allowbreak 5s) = s/(5s + 1) for every ss. The two lines start from the two tight four-speed tuples, (1, 2, 3, 4) and (1, 3, 4, 7).
  • The census that prompted it. Of the 44 band five-speed tuples with speeds up to 80, 32 lie on the two lines, at s=1s = 1 to 16 on each, and 12 do not, for example (1, 2, 3, 5, 8) and (3, 5, 8, 11, 13).
  • What it gives. With further arguments that rest on the list of tight four-speed tuples, all but finitely many band five-speed tuples are (1,2,3,4,5s)(1,\allowbreak 2,\allowbreak 3,\allowbreak 4,\allowbreak 5s) or (1,3,4,7,5s)(1,\allowbreak 3,\allowbreak 4,\allowbreak 7,\allowbreak 5s) up to order; that corollary is conditional and was not reviewed as a claim of its own.

Status and who checked it

  • Proved by hand, unconditionally; no condition is attached to this statement.
  • Who checked it. A falsifier agent checked the algebra by hand and the set of allowed times exactly, and its own exact computation of δ\delta and the attaining times agrees up to s=40s = 40 ("Section 2's proof is sound"). A second script agrees for s=2s = 2 to 30. No second reviewer graded it. Unrefereed.

Prior work

Credited. The other line, (1,2,3,4,5s)(1,\allowbreak 2,\allowbreak 3,\allowbreak 4,\allowbreak 5s), is Kravitz's Theorem 5.1. Kravitz's Lemma 9.4, applied to the tight tuple (1, 3, 4, 7), gives the second line once 5s>4⋅735s > 4 \cdot 7^3 = 1,372; his §10 remarks that (1, 3, 4, 7) satisfies the condition his method needs. So the claim here is the range s≥2s \ge 2 below that bound.

Searches run. Only the sources already held for the project; no web search was made for this statement. No source in them states the second line for small ss.

Still owed. A logged search of the literature for this formula.

Further reading