Wiki
Wiki

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

Updated


Statement. For p,q∈R2p,q\in\mathbb R^2 and r>0r>0, there is a regulus whose two affine rulings are exactly

A={ℓp,a:a∈C(q,r)},B={ℓb,q:b∈C(p,r)},(1)A=\{\ell_{p,a}:a\in C(q,r)\},\qquad B=\{\ell_{b,q}:b\in C(p,r)\}, \tag{1}

where C(c,r)C(c,r) is the circle with center cc and radius rr.

Source. Mathialagan, published 2021 PDF, p. 21, Proposition 40, introduced on p. 20 with its Figure 3. The proof below preserves the source's two circle families, makes the omitted-line limits explicit and proves completeness without importing its Lemma 39.

Existence. Choose distinct a1,a2,a3∈C(q,r)a_1,a_2,a_3\in C(q,r). The lines ℓp,ai\ell_{p,a_i} are pairwise projectively disjoint by Proposition 27, and determine the quadric RR of Proposition 36. For b∈C(p,r)b\in C(p,r), the segments (p,b)(p,b) and (ai,q)(a_i,q) have common positive length rr. Their unique proper motion sends pp to aia_i and bb to qq. It is a translation exactly when ai−p=q−ba_i-p=q-b, namely when b=bi=p+q−aib=b_i=p+q-a_i. Otherwise it is a nonidentity rotation and ℓb,q\ell_{b,q} meets ℓp,ai\ell_{p,a_i}.

If bb differs from the three bib_i, the three intersection points are distinct because the original lines are disjoint. The degree-two restriction argument in Proposition 36 puts ℓb,q\ell_{b,q} entirely on RR. This also holds at an exceptional bib_i: take a sequence of nonexceptional points of the circle tending to bib_i and substitute the parametrization of ℓb,q\ell_{b,q} into a defining quadratic of RR. Each coefficient of the resulting polynomial in zz varies continuously with bb and vanishes along the sequence. All coefficients vanish at bib_i as well.

Thus every line of BB lies on RR. They are pairwise projectively disjoint, so lie in one ruling, opposite to the original triple. For any a∈C(q,r)a\in C(q,r), the same motion argument shows ℓp,a\ell_{p,a} meets all lines of BB except the single translation case b=p+q−ab=p+q-a. Three of those intersections already put ℓp,a\ell_{p,a} on RR, in the other ruling. This proves containment of both families (1).

No other nonhorizontal lines. Consider a nonhorizontal line ℓx,y\ell_{x,y} in the ruling of AA. By Corollary 37 it intersects all but at most one member ℓb,q\ell_{b,q} of BB. At each intersection the corresponding rotation sends xx to yy and bb to qq, so

∣x−b∣=∣y−q∣.|x-b|=|y-q|.

This holds for infinitely many b∈C(p,r)b\in C(p,r). If x≠px\ne p, two circles with distinct centers intersect in at most two points, by the elementary argument in Lemma 34. Hence x=px=p, and then ∣y−q∣=r|y-q|=r. The line is a member of AA.

Similarly a nonhorizontal ℓx,y\ell_{x,y} in the ruling of BB intersects infinitely many ℓp,a\ell_{p,a}, forcing ∣x−p∣=∣y−a∣|x-p|=|y-a| for infinitely many a∈C(q,r)a\in C(q,r). Hence y=qy=q and ∣x−p∣=r|x-p|=r, so it belongs to BB.

No horizontal lines. On a horizontal spatial line the angle is fixed and its centers oo run along an affine line. Writing its rotations as go(t)=Rt+(I−R)og_o(t)=Rt+(I-R)o, the map o↦go(p)o\mapsto g_o(p) is an invertible affine map, since R≠IR\ne I. Thus its images form a planar line, which meets C(q,r)C(q,r) in at most two points. A horizontal line therefore cannot intersect infinitely many members of AA. By Corollary 37 it cannot belong to the opposite ruling. Replacing go(p)g_o(p) by go−1(q)=R−1q+(I−R−1)og_o^{-1}(q)=R^{-1}q+(I-R^{-1})o gives the same conclusion for intersections with BB, excluding a horizontal line in the first ruling. This completes the identification of both affine rulings.

Dependencies and source qualifications. This proof uses Propositions 20, 27, 36, Corollary 37, and the circle-intersection calculation of Lemma 34. It corrects the endpoint letters in the final paragraph of the source proof. The external seven-line premise printed as Lemma 39 is not used: the two infinite families and their completeness are proved directly.

Verification scope. Verified within the independently reviewed Theorem 3 chain, retained in the final review; the complete circle classification and its applications in Lemma 26 belong to the living Theorem 3 record.

Bears on. Problem 661.