Conjugate Diameters¶

A conic-indexed symmetric relation that pairs distinct diameter directions when each bisects real chords parallel to the other, equivalently when their quadratic-form pairing is zero.

Version: v1 · 2026-10-25 · History
Domain-specific #: 14372
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomain: Conic Geometry → Mathematics

Core Idea¶

Conjugate diameters names a binary relation on the eligible centre directions of each real ellipse or hyperbola.^[1] It selects a pair when each centre line bisects every existing real chord parallel to the other direction.^[1] In projective terms, each diameter is the polar of the other direction's point at infinity.^[1] Each selected pair has two distinct non-asymptotic directions; the relation is the rule selecting such pairs, not one pair by itself.

The relation can be checked algebraically. Write a real nonsingular central conic, after choosing its centre as origin, as \(q(x)=x^{\mathsf T}Ax=1\), with symmetric invertible \(A\). Let \(B(u,v)=u^{\mathsf T}Av\) and let \(D_C\) be the eligible non-asymptotic directions of the declared conic \(C\). The relation is \(R_C=\{([u],[v])\in D_C^2:[u]\ne[v]\text{ and }B(u,v)=0\}\); the quadratic expansion below proves this coordinate restatement of Lehmer's geometric rule.^[1] Since \(B\) and distinctness are symmetric, \(R_C\) equals its converse. A particular worked pair is an instance of \(R_C\), not the relation as a whole. The conic, not the ordinary Euclidean dot product, determines the pairing. A circle is the familiar perpendicular special case.^[1] The same rule also reaches an ellipse with skew directions and a hyperbola whose partner direction can meet its conjugate hyperbola instead of the original real curve.^[1]^[2]

Structural Signature¶

Sig role-phrases: real central conic and eligible direction domain → first direction → distinct partner direction → symmetric membership rule → real-chord boundary.

Eligible central conic. A real nonsingular ellipse or hyperbola supplies a finite centre, an invertible symmetric form and real chords.^[1] A parabola has no proper conjugate diameter pair of this kind.^[1] An equation with no one-dimensional real locus would make a real-chord test vacuous.
First direction. A non-asymptotic line through the centre sets the parallel-chord family to be bisected.^[1] Once the conic and this direction are fixed, its bilinear-orthogonal partner direction is mathematically determined by the nondegenerate two-dimensional form; choosing the first direction is still necessary to specify the pair.
Partner direction. A distinct centre line cuts the first family's real chords at their midpoints and supplies its own reciprocal family.^[1] An arbitrary second centre line does not qualify.
Recognition relation. On eligible distinct directions, this relation includes exactly the pairs with zero bilinear pairing by the quadratic derivation below.^[1] Its membership is unchanged by rescaling either direction vector, and reversing a member pair yields another member. A diagram's apparent angle is not sufficient outside the circle.
Real-chord boundary. The midpoint claim concerns lines that intersect the real curve at two points. On a hyperbola, a partner direction can reach the conjugate hyperbola rather than the original curve.^[2] An asymptote is a self-conjugate limiting ray rather than a distinct partner in this article's sense.^[1]

For the algebraic equivalence, intersect the conic with \(x=m+tu\). The resulting quadratic is \(q(u)t^2+2B(m,u)t+q(m)-1=0\). When it has two real roots and \(q(u)\ne0\), their mean is \(-B(m,u)/q(u)\); the chord is therefore bisected at \(m\) exactly when \(B(m,u)=0\). Taking a nonzero midpoint \(m=sv\) of an existing two-endpoint chord gives \(sB(u,v)=0\), hence \(B(u,v)=0\); symmetry supplies the converse. This calculation derives Lehmer's geometric rule in the stated coordinate form.^[1]

What It Is Not¶

Every pair of centre lines. Sharing a centre does not force \(B(u,v)=0\).
Euclidean perpendicularity in general. The circle reduces conjugacy to perpendicularity, but a noncircular ellipse or hyperbola can have a conjugate pair with nonzero Euclidean dot product.^[1]
An ellipse or hyperbola by itself. The curve is the carrier; the abstraction is its rule for selecting eligible ordered direction pairs. A single successful pair is a member of that relation.
A reconstruction algorithm. This relation tests directions on a declared conic; reconstructing axes or a whole curve is a separate task and is not part of the membership test.
A universal four-tangent bounding parallelogram. The finite parallelogram formed by tangents at four endpoints on the original ellipse is an ellipse-specific construction.^[1] A hyperbola partner direction need not meet the original real curve.^[1]^[2]

Scope of Application¶

The relation applies to a real ellipse, including the circle, and to a real hyperbola with a finite centre.^[1] It compares directions relative to that conic; changing the conic changes the pairing. Lehmer gives reciprocal bisection and tangent results, then explicitly applies conjugate diameters to a hyperbola.^[1] Besant treats a hyperbola together with its conjugate hyperbola, so both partner lines need not intersect the original real branch.^[2]

The following ellipse and hyperbola consequences are outside the defining two-direction test and require the stated extra assumptions. Historical axis constructions are treated below only as bounded, derived reconstruction tasks; engineering uses and spacetime interpretations remain outside this entry.

Derived geometric consequences.

For a fixed ellipse \(q(x)=x^{\mathsf T}Ax=1\) with positive-definite \(A\), take endpoint vectors \(a,b\) with \(q(a)=q(b)=1\) and \(B(a,b)=0\). Besant states that the tangent parallelograms associated with such conjugate diameter pairs all have the same area.^[2] Directly, the tangents at \(\pm a\) and \(\pm b\) form a parallelogram with vertices \(\pm a\pm b\) and area \(4|\det[a\ b]|=4/\sqrt{\det A}\), equal to the rectangle on the full principal axes. Four unlabeled tangent sides alone do not determine an ellipse: for every \(|t|<1\), the distinct ellipse \((x^2-2txy+y^2)/(1-t^2)=1\) is tangent to the same square \(|x|\leq1,|y|\leq1\). This counterexample concerns reconstruction data, not conjugate-pair membership.

The same normalized endpoint pair gives an ellipse parametrization \(x(t)=a\cos t+b\sin t\): the cross term vanishes, so \(q(x(t))=\cos^2t+\sin^2t=1\), and the independent vectors \(a,b\) span every point on the curve. For a hyperbola \(q(x)=1\) of signature \((1,1)\), choose \(q(a)=1\), \(q(b)=-1\) and \(B(a,b)=0\), with \(b\) on the conjugate hyperbola. Then \(x(t)=a\cosh t+b\sinh t\) traces the branch containing \(a\); \(-x(t)\) traces the other. Endpoint normalization, the sign of \(q(b)\) and branch choice are necessary conditions, not additional defining roles. These formulas follow by direct substitution into the declared quadratic forms.

In orthonormal coordinates for a rectangular hyperbola \(q(x,y)=x^2-y^2\), reflection across the asymptote \(y=x\) sends \((x,y)\) to \((y,x)\). Direct substitution gives \(q(y,x)=-q(x,y)\) and \(B((x,y),(y,x))=xy-yx=0\). Thus reflection exchanges the hyperbola with its conjugate and pairs reflected non-asymptotic directions. This Euclidean reflection consequence depends on the rectangular form; it is not a construction for every hyperbola.

Lehmer gives the general involution separation criterion and identifies conjugate diameters as an involution with hyperbola asymptotes as its double rays.^[1] Combining those results with the absence of real ellipse asymptotes yields this conic-specific deduction: with four distinct directions in projective cyclic order, two different conjugate pairs on an ellipse interleave, whereas on a hyperbola they do not. Identical, reversed or overlapping pairs are outside that four-point separation comparison.

Derived constructions using conjugacy. Axis reconstruction is an inverse task distinct from testing whether two directions belong to \(R_C\). Its input is an ellipse centre and two specified, nonparallel conjugate semidiameter endpoint vectors \(a,b\), including their positions and lengths, with \(q(a)=q(b)=1\) and \(B(a,b)=0\). With \(M=[a\ b]\), conjugacy gives \(M^{\mathsf T}AM=I\), hence \(A=M^{-\mathsf T}M^{-1}\). The principal-axis directions are the eigenvectors of positive-definite \(A\), and its eigenvalues \(\lambda_1,\lambda_2\) give semiaxis lengths \(1/\sqrt{\lambda_1},1/\sqrt{\lambda_2}\). For \(a=(2,0)\), \(b=(1,1)\), this produces \(A=\left(\begin{smallmatrix}1/4&-1/4\\-1/4&5/4\end{smallmatrix}\right)\); both endpoint equations and \(B(a,b)=0\) check directly. This matrix proof is a modern derivation, not a transcription of a classical geometric procedure. Pappus gives an axis construction from conjugate diameters in Book VIII §XVII; Mossbrugger's 1845 construction uses the given diameters' positions and lengths and credits Rytz.^[3]^[4] Four unlabeled tangent sides do not supply the specified pair; the square counterexample above proves their insufficiency. For a circle the semiaxis lengths are fixed but axis orientation is nonunique.

A second derived procedure starts with \(q(x)=B(x,x)=1\) for a real centered hyperbola of signature \((1,1)\), a point \(Q\) on its original real locus, and a nonzero tangent direction \(u\) at \(Q\). Tangency gives \(B(Q,u)=0\). The orthogonal complement of the positive vector \(Q\) is negative, so \(q(u)<0\); set \(E=u/\sqrt{-q(u)}\) on the conjugate hyperbola \(q(E)=-1\). Then \(B(Q,E)=0\), and the tangent at \(E\) is parallel to \(OQ\). In \(x^2-y^2=1\), choosing \(Q=(\cosh t,\sinh t)\) and \(u=(\sinh t,\cosh t)\) verifies \(q(u)=-1\) and \(E=u\) for every real \(t\). The sign of \(E\) does not change the selected partner direction. This coordinate construction is a consequence of the current relation and tangent equation; no unverified ancient wording or original-hyperbola endpoint parallelogram follows from it.

Clarity¶

Start by declaring the conic and its centre. For an ellipse or hyperbola in quadratic form, calculate \(B(u,v)\) for the proposed directions. Zero is the membership test by the quadratic derivation above; for the geometric reading, check midpoint bisection only on lines with two real intersections.^[1] A nonzero result on the same conic is the closest near miss. A parabola, an imaginary real locus and an asymptote-only self-pair fail for different reasons, so name the boundary rather than treating every failure as a bad angle.^[1]

Manages Complexity¶

The conic determines infinitely many chord families, but one bilinear-zero test identifies which two centre directions organize them reciprocally. The test lets an analyst move between an affine drawing, a projective polar statement and a centered equation without requiring the diameters to look perpendicular.^[1] It also stops an ellipse-specific tangent construction from being copied without qualification to a hyperbola.^[2]

Abstract Reasoning¶

Given \(A\) and a non-null direction \(u\), solve the linear equation \(B(u,v)=0\) for a nonzero \(v\). In a nonsingular two-dimensional form, the partner direction is unique up to scale. Check that it is distinct from \(u\); if \(u\) is a null hyperbolic asymptote, its orthogonal complement is its own line and the bounded distinct-pair test stops. Finally confirm that the real chord statement is made only where two endpoints exist. These steps separate the formal polarity from the existence of a particular drawn chord.^[1]

Knowledge Transfer¶

Within conic geometry, the reciprocal test transfers from a closed ellipse to an open hyperbola: a positive-definite form becomes an indefinite one, but \(B(u,v)=0\) and midpoint bisection of existing real chords remain.^[1]^[2] What does not transfer is the claim that both directions cut the same real curve or that four original-curve endpoint tangents make a finite bounding parallelogram.^[1]^[2] Calling a pair of unrelated physical axes “conjugate” without an identified conic form and bilinear test would be analogy, not an instance of this entry.

Examples¶

Canonical: skew pair on an ellipse¶

Take \(x^2/4+y^2=1\), with centre \((0,0)\) and \(A=\operatorname{diag}(1/4,1)\). Set \(u=(1,1/2)\) and \(v=(-2,1)\). Then \(B(u,v)=(-2)/4+(1/2)(1)=0\), while the Euclidean dot product is \(u\cdot v=-3/2\). Both \(q(u)=1/2\) and \(q(v)=2\) are positive, so both centre directions reach real ellipse endpoints. These coordinates are a checked illustrative derivation, not a diagram supplied by Lehmer. Lehmer's geometric rule supplies the source interpretation of the calculation.^[1]

Mapped back: the ellipse is the eligible conic; \(u\) is the first direction; \(v\) is its distinct partner; \(([u],[v])\in R_C\) follows from zero bilinear pairing and gives reciprocal midpoint bisection; both families have real two-endpoint chords. The nonzero Euclidean dot product shows why visual perpendicularity is the wrong general test.

Applied / In Practice: paired directions of a hyperbola¶

Take \(x^2-y^2=1\), with centre \((0,0)\) and \(A=\operatorname{diag}(1,-1)\). Set \(u=(2,1)\) and \(v=(1,2)\). The form gives \(B(u,v)=2-2=0\), while \(q(u)=3>0\) and \(q(v)=-3<0\). The \(u\) centre line reaches the original hyperbola; the \(v\) centre line reaches the real conjugate hyperbola \(y^2-x^2=1\). The existence and midpoint properties of displaced real chords follow from the quadratic expansion; they do not assert chords for every displacement. This is a checked coordinate example, not a quoted historical diagram. Lehmer and Besant supply the hyperbola and conjugate-hyperbola interpretation.^[1]^[2]

Mapped back: the hyperbola is the eligible conic; \(u\) and \(v\) are distinct non-asymptotic directions; \(B(u,v)=0\) places \(([u],[v])\) in this conic's \(R_C\); real parallel chords, wherever they have two endpoints, are bisected by the partner centre line. The negative value of \(q(v)\) explains why looking for both real endpoints on the original hyperbola would reject a valid conjugate pair.

Structural Tensions¶

No inherent two-sided optimization or interpretive conflict is established by the sources for this formal relation. The distinction between Euclidean and conic orthogonality is a correctness test; the distinction between an ellipse drawing and a hyperbola drawing is a scope condition. Neither is presented as a tradeoff to be optimized.

Structural–Framed Character¶

Vocabulary travels across ellipse and hyperbola: conic, centre, paired directions and reciprocal chord midpoint survive while their real-locus geometry changes. Evaluative weight is low because \(B(u,v)=0\) is a mathematical membership test, not a preference for one diameter. Institutional origin is low: Lehmer and Besant document the relation, but neither author or edition makes a pair conjugate by decree. Human-practice dependence is limited to choosing coordinates, real-locus convention and a question worth asking; the result does not depend on the observer. Import versus recognition is bounded: using the relation on a new eligible conic recognizes the same form, whereas applying it to unrelated axes without a conic form imports a metaphor. The portable skeleton is reciprocal pairing with a unique partner under a nondegenerate form; the verified operation still requires conic chords and a finite centre. Its character: structural within conic geometry, domain-framed at the transfer boundary.

Structural Core vs. Domain Accent¶

The candidate portable core is a symmetric membership rule under a nondegenerate form: fixing one eligible direction determines a partner and licenses a reciprocal inference. The domain accent is precise: a real central conic, its polarity, its parallel real chords and their midpoint loci. The live Prime Duality is a close comparison, but its full two-side, structure-translating signature is not established merely by the symmetric midpoint rule. The live Pole and polar entry describes the general conic point-line correspondence. The live Symmetric relation owns the reversal-invariant binary-relation genus that this conic-indexed rule instantiates. This entry does not independently meet the Prime bar because its defining midpoint and diameter tests still require central-conic geometry. A substrate-neutral form-induced reciprocity could be a future-Prime question if unlike nonconic carriers and the same inference license are established; these two conic cases do not make that claim.

This entry is a kind of Symmetric relation.

Duality is relevant because the conic determines a reciprocal partner relation: each eligible direction uniquely picks out the other, and the chord-midpoint statement works in both directions. Its broader formulations concern other carriers and structural translations as well. Symmetric relation is the immediate strict genus: for each fixed conic, the conjugacy relation is homogeneous and reversing a member pair preserves membership; its bilinear-zero and chord-midpoint tests supply the stable differentia. The live Prime Relation is an ancestor through Symmetric relation, so a separate direct edge would repeat the same axis. Pole and polar supplies the projective vocabulary and the ideal-point construction, but its general point-line map is distinct from this conic-indexed relation between centre directions.^[1]

Relationships to Other Abstractions¶

Current abstraction Conjugate Diameters Domain-specific

Parents (1) — outgoing typed relations

Conjugate Diameters is a kind of Symmetric relation Domain-specific

For each eligible real central conic, conjugacy is a symmetric binary relation on its non-asymptotic diameter directions, distinguished by bilinear-zero pairing and reciprocal real-chord bisection.

Condition / exception The article restricts to distinct non-asymptotic directions of real central conics with a nonempty one-dimensional real locus; null asymptote self-pairs are a projective boundary, not members of R_C.

Hierarchy path (1) — routes to 1 parentless root

Conjugate Diameters → is a kind of Symmetric relation → is a kind of Relation

Neighborhood in Abstraction Space¶

Conjugate Diameters sits in a sparse region of the domain-specific corpus (79^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2769 abstractions)

Nearest neighbors

Circular arc — 0.83
Vertex (curve) — 0.83
Nine-Point Conic — 0.83
Ellipse — 0.83
Laguerre Formula — 0.82

Computed from structural-signature embeddings · 2026-10-25

Not to Be Confused With¶

Ellipse names one eligible carrier, not the pair relation. Pole and polar is the broader point-line correspondence, not the exact two-direction chord-midpoint claim. Perpendicular diameters coincide with conjugate diameters only in the circle case.^[1] Hyperbolic asymptotes are self-conjugate limiting rays in Lehmer's involution account, not a distinct two-line pair under this entry's boundary.^[1] An axis-reconstruction procedure uses specified geometric data and has a different output from this membership test.

References¶

[1] Derrick Norman Lehmer. An Elementary Course in Synthetic Projective Geometry. Ginn and Company, 1917, printed pp. 62–64, §§107–115; pp. 85–86, §§143–144; and p. 89, §148. Original geometry textbook; Project Gutenberg full-text transcription. https://www.gutenberg.org/files/17001/17001-h/17001-h.htm registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y ↩z ↩27 ↩28 ↩29 ↩30

[2] William Henry Besant. Conic Sections: Treated Geometrically. George Bell and Sons, 1895, printed p. 73, art. 87, and pp. 108–109, art. 123. Original geometry text; Project Gutenberg re-typeset transcription of Cornell source images. https://gutenberg.org/files/29913/29913-pdf.pdf registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i

[3] Pappus. Collection, Book VIII, §XVII, in Friedrich Hultsch, ed., Pappi Alexandrini Collectionis quae supersunt, vol. 3, 1878, printed p. 1083. Original Greek and Latin edition, consulted at https://www.wilbourhall.org/pdfs/collection/Hultsch_1876_Pappos-3-1_S.pdf. The edition's §XIV begins a different construction; no unqualified proposition-14 locator is used here. registry ↩

[4] Leopold Mossbrugger. Grösstentheils neue Aufgaben aus dem Gebiete der Géométrie descriptive. Zürich: Meyer und Zeller, 1845, printed pp. 123–125. Original text and figure; p. 125 credits Professor Rytz with the ellipse-axis construction from two given conjugate diameters. https://doi.org/10.3931/e-rara-54897 registry ↩