Ellipse Theorem: Geometric & Analytic Insights

• Ellipse Theorem is a collection of results where ellipses emerge as rigid loci defined by distance-squared functions, tangent conic properties, and numerical range boundaries.
• It spans triangle and quadrilateral geometries, using constructs like the Steiner in-ellipse and Newton line to illustrate precise tangency and center properties.
• The theory bridges classical geometry and modern analysis by linking integrable billiard dynamics, operator theory, and inverse spectral methods to establish rigidity and uniqueness.

In contemporary mathematics, the expression “Ellipse Theorem” is used for several distinct theorem families rather than for a single canonical statement. Across Euclidean geometry, algebraic geometry, billiard dynamics, operator theory, and inverse spectral theory, the common theme is that an ellipse emerges as a rigid locus, a tangent conic, a boundary-generating curve, or a uniquely determined object. Representative formulations include a distance-squared characterization of ellipses by the sides of a triangle (Abboud, 2014), center and focus theorems for conics associated with quadrilaterals (Kaldybayev, 2022, Horwitz, 7 May 2026), billiard and triangle-center loci that are themselves ellipses (Garcia, 2016, Helman et al., 2020), the elliptical range theorem for $2\times 2$ matrices (Paparella et al., 2018), and rigidity results asserting that sufficiently small-eccentricity ellipses are spectrally or dynamically determined (Hezari et al., 2019, Huang et al., 2017). This suggests that the term is context-dependent rather than standardized.

1. Triangle-based locus and definition theorems

A prominent geometric usage arises from the reformulation of Viviani’s theorem by means of a distance-sum function $V(P)$, the sum of the perpendicular distances from a point $P$ to the sides of a polygon. For triangles and convex polygons, $V$ is constant along parallel line segments; in a triangle, $V$ is constant on the whole triangle iff the triangle is equilateral (Abboud, 2014). In the same framework, the paper introduces the “geometric creatures” Canghareeb and Sanghareeb. Canghareeb is constrained by a constant sum of distances to the sides of a triangle, so its locus is typically a line segment. Sanghareeb is constrained by a constant sum of squares of distances to the sides or their extensions, and its locus is an ellipse (Abboud, 2014).

The resulting theorem is explicit: if $d_1,d_2,d_3$ are the perpendicular distances from a point $P$ to the sidelines of a triangle, then the locus

$d_1^2+d_2^2+d_3^2=k$

is an ellipse, and conversely every ellipse can be realized in this way for a suitable triangle (Abboud, 2014). In coordinates

$$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$

the locus is

$$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$

The same paper isolates the circle as the special case: the locus $V(P)$0 is a circle iff the triangle is equilateral (Abboud, 2014). This yields a line-based characterization of the ellipse parallel to, but distinct from, the classical two-focus definition.

A second triangle theorem of a different type is Marden’s theorem. If $V(P)$1 are non-collinear and

$V(P)$2

then the unique ellipse inscribed in the triangle $V(P)$3 and tangent at the side midpoints—the Steiner in-ellipse—has foci equal to the zeros of $V(P)$4 (Rohe, 2015). Thus the derivative roots of a cubic are identified with the foci of a canonical inscribed ellipse. In the equilateral case the two foci coalesce, and the Steiner in-ellipse becomes the incircle (Rohe, 2015).

2. Quadrilateral tangency, focus, and center theorems

For quadrilaterals, one major theorem family concerns the Newton line. A generalized Newton quadrilateral theorem states that if a non-parallelogram quadrilateral $V(P)$5 has an ellipse or hyperbola tangent to its four extended sides, then the center of that conic lies on the Newton line, the line through the midpoints of the diagonals (Kaldybayev, 2022). The converse is equally strong: every point on that line, except the midpoint of each diagonal and, in the nontrapezoid case, the third diagonal midpoint, is the center of some tangent ellipse or hyperbola (Kaldybayev, 2022). The circle version is classical; the extension to ellipses and hyperbolas identifies the Newton line as the full center locus for centered tangent conics.

A second quadrilateral theorem answers Besant’s problem. Let $V(P)$6 be a cyclic convex quadrilateral, let $V(P)$7 denote the point equidistant from its four vertices, and let $V(P)$8 denote the intersection point of its diagonals. If an ellipse inscribed in $V(P)$9 has one focus at $P$0, then the other focus is $P$1; conversely, if one focus is $P$2, the other is $P$3 (Horwitz, 7 May 2026). The same work proves more: such an ellipse exists iff $P$4 is orthodiagonal, and in that case the ellipse center is $P$5 (Horwitz, 7 May 2026). In the trapezoid case, the theorem refines further: $P$6 is a Besant quadrilateral iff it is isosceles, and the corresponding Besant ellipse is the maximal-area inscribed ellipse (Horwitz, 7 May 2026).

Midpoint tangency imposes another rigidity pattern. For a convex quadrilateral that is not a parallelogram, there is no inscribed ellipse tangent at the midpoints of three sides (Horwitz, 2017). Tangency at the midpoints of two sides is already exceptional: it occurs iff the quadrilateral is either a trapezoid or a midpoint diagonal quadrilateral, meaning that the diagonal intersection is the midpoint of at least one diagonal (Horwitz, 2017). In a non-parallelogram trapezoid there is a unique such ellipse, and it is also the unique maximal-area inscribed ellipse; in a nontrapezoid midpoint diagonal quadrilateral there are exactly two such ellipses (Horwitz, 2017).

A different quadrilateral theorem counts inscribed ellipses through a prescribed point. If $P$7 lies in the interior of a convex quadrilateral and is off both diagonals, then exactly two inscribed ellipses pass through $P$8; if $P$9 lies on one diagonal but is not the diagonal intersection, exactly one does; if $V$0 is the diagonal intersection, none does (Horwitz, 2015). If $V$1 lies on a side but is not a vertex, exactly one inscribed ellipse passes through it, necessarily tangent there (Horwitz, 2015). This yields a sharp classification by incidence with the diagonals.

3. Elliptic billiards and moving-triangle loci

In elliptic billiards, the locus of triangle centers often becomes an ellipse. For the standard elliptic billiard

$V$2

the centers of the inscribed circles of the triangular $V$3-periodic billiard orbits form another ellipse (Garcia, 2016). The paper computes the canonical equation explicitly: $V$4 shows that the locus has constant affine curvature, and identifies its foci as

$V$5

It also proves that this incenter ellipse is similar to the confocal Poncelet caustic associated with the $V$6-periodic family (Garcia, 2016).

A related locus theorem concerns triangles inscribed in a fixed ellipse with two vertices fixed and the third moving along the boundary. Let $V$7 denote the barycenter and $V$8 the orthocenter of the moving triangle, and consider any fixed affine combination

$V$9

As the moving vertex traverses the ellipse, the locus of $V$0 is an ellipse for every fixed $V$1 (Helman et al., 2020). The result includes the classical special cases: the barycenter locus is a translate of a homothetic copy of the original ellipse, while the orthocenter locus is also an ellipse, with explicit semiaxes and center (Helman et al., 2020).

The same work establishes two line-sweeping invariants. For fixed vertices $V$2, the centers $V$3 of the locus ellipses move on a line as $V$4 varies; for families of parallel fixed chords $V$5, the locus ellipses translate rigidly along a second line through the ellipse center (Helman et al., 2020). It also derives envelope theorems: for fixed $V$6, the centers $V$7 as $V$8 varies trace an axis-parallel ellipse, and in the orthocenter case the outer envelope is an affine image of Pascal’s limaçon (Helman et al., 2020).

4. Algebraic and operator-theoretic formulations

A substantial algebraic theorem identifies the geometry of an ellipse directly from the coefficients of its quadratic equation

$V$9

For a nondegenerate ellipse, the invariants

$d_1,d_2,d_3$0

satisfy $d_1,d_2,d_3$1, $d_1,d_2,d_3$2, and $d_1,d_2,d_3$3, while the center is

$d_1,d_2,d_3$4

From these data one obtains explicit formulas for the semi-axis lengths, the foci, and the rotation angle, thereby refining the ambiguous classical relation $d_1,d_2,d_3$5 to a branch-correct formula for $d_1,d_2,d_3$6 itself (Horwitz, 2017). The result amounts to a coefficient-to-geometry dictionary for ellipses.

In operator theory, the elliptical range theorem asserts that the numerical range of a $d_1,d_2,d_3$7 complex matrix is an elliptical disk (Paparella et al., 2018). If $d_1,d_2,d_3$8 has eigenvalues $d_1,d_2,d_3$9, then its numerical range $P$0 is centered at

$P$1

has foci $P$2, and has minor axis length

$P$3

Using Kippenhahn’s theorem, the proof identifies a projective dual conic whose real affine part is the boundary-generating ellipse; convexity then yields the filled ellipse as the numerical range (Paparella et al., 2018).

Marden’s theorem belongs here as well because it translates a cubic polynomial into a canonical inscribed ellipse. The zeros of $P$4 for

$P$5

are precisely the foci of the Steiner in-ellipse of the triangle with vertices $P$6 (Rohe, 2015). The theorem is a prototypical algebra-geometry correspondence: derivative roots become ellipse foci.

5. Rigidity, integrability, and inverse spectral theory

A different “ellipse theorem” appears in billiard rigidity. A local version of the Birkhoff conjecture states that a sufficiently small perturbation of an ellipse of sufficiently small eccentricity that is integrable near the boundary must itself be an ellipse (Huang et al., 2017). The precise hypothesis is $P$7-rational integrability: the existence of integrable rational caustics of rotation number $P$8 for all $P$9. The paper proves this for $d_1^2+d_2^2+d_3^2=k$0, and for $d_1^2+d_2^2+d_3^2=k$1 under explicit matrix nondegeneracy conditions (Huang et al., 2017). One crucial ingredient is a high-order eccentricity expansion of action-angle coordinates, which yields linear relations among Fourier modes of the boundary deformation.

Inverse spectral theory supplies an even more rigid formulation. There exists $d_1^2+d_2^2+d_3^2=k$2 such that any ellipse of eccentricity less than $d_1^2+d_2^2+d_3^2=k$3 is uniquely determined, up to Euclidean isometry, by its Dirichlet or Neumann Laplace spectrum among all smooth domains (Hezari et al., 2019). The same work proves that, for nearly circular domains, the lengths of periodic billiard trajectories shorter than the perimeter belong to the singular support of the wave trace, and it deduces that an isospectral competitor to a nearly circular ellipse must itself be nearly circular, rationally integrable, hence an ellipse (Hezari et al., 2019).

Taken together, these results tie ellipse theorems to two rigidity paradigms. The dynamical theorem uses rational caustics and near-boundary integrability (Huang et al., 2017); the spectral theorem uses wave-trace singularities and length-spectrum separation (Hezari et al., 2019). This suggests a deep interaction among ellipses, integrable billiards, and inverse problems.

6. Analytic inequalities, combinatorial products, and generalized inversions

The expression also appears in more specialized analytic and combinatorial settings. For the standard ellipse with semiaxes $d_1^2+d_2^2+d_3^2=k$4, the exact perimeter is

$d_1^2+d_2^2+d_3^2=k$5

and an explicit rigorous upper bound sharper than the elementary estimate $d_1^2+d_2^2+d_3^2=k$6 is derived directly from the elliptic-integral representation (Pain, 2022). The bound is exact both in the circle case $d_1^2+d_2^2+d_3^2=k$7 and in the degenerate limit $d_1^2+d_2^2+d_3^2=k$8 (Pain, 2022).

A classical chord-product theorem due to Price provides another ellipse theorem. Stretch the unit circle horizontally by $d_1^2+d_2^2+d_3^2=k$9 and vertically by $$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$0, so the marked points on the ellipse are $$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$1, where $$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$2. Then the product of the $$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$3 chord lengths from the base point $$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$4 to the remaining marked points is

$$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$5

The proof connects the geometry of ellipse chords to generalized Lucas polynomials, generalized Fibonacci polynomials, Newton identities, and Cardano’s solution of the cubic (Blum-Smith et al., 2018).

The theory of inversion in an ellipse generalizes circular inversion by replacing a fixed inversion radius with the distance from the center to the ellipse in a given ray direction. For

$$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$6

the inverse of $$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$7 is

$$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$8

Under this map, a line not through the center inverts to an ellipse through the center, homothetic ellipses are preserved as a family, and the elliptic Pappus Chain Theorem takes the form

$$A(0,a),\qquad B(-b,0),\qquad C(c,0),\qquad a,b,c>0,$$9

for the center height $$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$0 and semi-minor axis $$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$1 of the $$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$2th ellipse in the chain (Ramírez, 2013).

Not all ellipse theorems in the literature are fully proved. For the normalized ellipse

$$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$3

the Ptolemy constant satisfies the rigorous bounds

$$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$4

and there is nonrigorous evidence that the lower bound is exact, attained by the four axis points $$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$5, $$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$6, $$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$7, and $$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$8 (Finch, 2016). The paper explicitly labels the equality claim as unproved. This is a useful caution against a common misconception: not every statement presented informally as an “ellipse theorem” has the same proof status.

A final applied-geometric example is the ladder-ellipse problem. If a fixed-length ladder has its ends on the positive coordinate axes and is tangent to a first-quadrant ellipse tangent to the axes at $$\frac{(ax+cy-ac)^2}{a^2+c^2}+\frac{(ax-by+ab)^2}{a^2+b^2}+y^2=k.$$9 and $V(P)$00, then there is a critical length $V(P)$01: one admissible ladder position when $V(P)$02, two when $V(P)$03, and none when $V(P)$04; the governing equation is equivalent to a quartic polynomial (Horwitz, 2015). Here the ellipse theorem takes the form of a bifurcation statement for tangent configurations.

Across these examples, the phrase “Ellipse Theorem” designates not one theorem but a cluster of structural facts in which the ellipse plays the role of a distinguished rigid object: a locus defined by distances, a tangent conic encoded by quadrilateral geometry, a trajectory center set in billiards, a numerical-range boundary, or a uniquely audible or dynamically integrable shape.