Wiki
Wiki

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

Updated


Statement. Let P,QP,Q be finite planar sets with $2\leq m=|P|\leq n=|Q|$, and use the line families of Proposition 27. Every regulus RR satisfies

∣L∩R∣≤max⁡{4m,  4D(P,Q)+2m}.(1)|L\cap R|\leq\max\{4m,\;4D(P,Q)+2m\}. \tag{1}

Consequently, for each fixed c>0c>0, either D(P,Q)≥cmnD(P,Q)\geq c\sqrt{mn} or every regulus contains at most (4c+4)mn(4c+4)\sqrt{mn} lines of LL.

Source. Mathialagan, published 2021 PDF, pp. 12, 23, Lemma 26. The source states the consequence for sufficiently large cc. The explicit estimate (1) supplies its required constant and fills in the counts for the second ruling and the second color.

Proof. There are 2m2m fixed-endpoint families LpiL_p^i. If each supplies at most two lines in RR, then ∣L∩R∣≤4m|L\cap R|\leq4m. Otherwise one supplies three distinct lines. Reflection across z=0z=0 interchanges the colors and preserves LL, so suppose they are $\ell_{p,a_1},\ell_{p,a_2}, \ell_{p,a_3}$ with p∈Pp\in P and distinct ai∈Qa_i\in Q.

By Proposition 27 these three lines are projectively disjoint. The projective ruling classification in Proposition 36 puts them in the same ruling and says they determine the unique quadratic surface RR. The three endpoints are either collinear, or lie on a unique circle of positive radius. The latter follows by intersecting two perpendicular bisectors; their directions are independent for a noncollinear triple.

In the circle case, Proposition 40 says that all lines in RR have one of the two forms

A={ℓp,a:a∈C(q,r)},B={ℓb,q:b∈C(p,r)}.A=\{\ell_{p,a}:a\in C(q,r)\},\qquad B=\{\ell_{b,q}:b\in C(p,r)\}.

Uniqueness of ordered endpoints in Proposition 27 gives the following bounds, including the membership conditions which can make a count zero.

Ruling and colorNecessary endpoint conditionsBound
A∩L1A\cap L^1p∈P, a∈Q∩C(q,r)p\in P,\ a\in Q\cap C(q,r)2D(P,Q)2D(P,Q)
A∩L2A\cap L^2p∈Q, a∈P∩C(q,r)p\in Q,\ a\in P\cap C(q,r)mm
B∩L1B\cap L^1q∈Q, b∈P∩C(p,r)q\in Q,\ b\in P\cap C(p,r)mm
B∩L2B\cap L^2q∈P, b∈Q∩C(p,r)q\in P,\ b\in Q\cap C(p,r)2D(P,Q)2D(P,Q)

The 2D(P,Q)2D(P,Q) bounds are precisely the circle case of Lemma 34 with A=PA=P. Summing is an upper bound even when a line has both colors. The two rulings themselves have no common line. This yields ∣L∩R∣≤4D(P,Q)+2m|L\cap R|\leq4D(P,Q)+2m.

In the collinear case Proposition 42 gives one ruling {ℓp,a:a∈λ}\{\ell_{p,a}:a\in\lambda\} and a wholly horizontal opposite ruling. No line of LL is horizontal. The first ruling contributes at most ∣Q∩λ∣≤2D(P,Q)|Q\cap\lambda|\leq2D(P,Q) lines of color 1 and at most mm of color 2, so this case is bounded by 2D(P,Q)+m2D(P,Q)+m, which is sufficient for (1).

Finally m≤mnm\leq\sqrt{mn}. If D(P,Q)<cmnD(P,Q)<c\sqrt{mn}, each term in the maximum (1) is at most (4c+4)mn(4c+4)\sqrt{mn}. This proves the alternative.

Thresholds and dependencies. The source begins with at most four lines per family and otherwise selects three same-ruling lines from five. Its Propositions 40 and 42 refer to the externally cited Lemma 39, whose printed threshold is seven. This proof does not infer a seven-line hypothesis from five lines. Instead Proposition 36 proves the ruling classification and uniqueness directly; the three fixed-endpoint lines are already projectively disjoint and hence in the same ruling. Propositions 40 and 42 prove the complete rulings without Lemma 39. Only the line/circle specialization of Lemma 34 is needed.

Verification scope. Verified within the independently reviewed Theorem 3 chain, retained in the final review; the strengthened bookkeeping estimate, both-color/both-ruling counts and geometric dependencies belong to the living Theorem 3 record. This is a compilation-supplied elaboration of the published argument, not an author-issued correction.

Bears on. Problem 661.