---
title: Poncelet Triangle Family
url: https://www.emergentmind.com/topics/poncelet-triangle-family
type: topic
---

# Poncelet Triangle Family

Searching arXiv for recent and foundational papers on Poncelet triangle families and related loci/invariants.
A Poncelet triangle family is a one-parameter family of triangles simultaneously inscribed in an outer conic and circumscribed about an inner conic: if one triangle can be drawn with its vertices on the outer conic and its sides tangent to the inner conic, then Poncelet’s porism implies that any point of the outer conic can serve as a vertex of another such triangle. In the literature surveyed here, the topic appears in several closely related forms: classical real porisms for nested ellipses or circles, affine and projective variants, finite-field analogues, and modern studies of loci, invariants, envelopes, and associated constructions such as inversions and parabolas [2106.00715], [1604.00436].

## 1. Definition and foundational configurations

The standard triangular case begins with two smooth conics \(C\) and \(C'\), or in affine coordinates an outer ellipse
\[
\frac{x^2}{a^2}+\frac{y^2}{b^2}=1,
\]
together with an inner conic serving as a caustic. A triangle \(P_1P_2P_3\) is inscribed in the outer conic if each vertex lies on it, and circumscribed about the inner conic if each side is tangent to it. Once one such triangle exists, the family closes for every starting point on the outer conic, producing a continuous one-dimensional family \(T(\lambda)\) or \(T(t)\) [2106.00715], [2106.04521].

Several configurations recur throughout the literature. In the confocal case, the inner ellipse shares foci with the outer ellipse; in the incircle family, the outer conic is an ellipse and the caustic is a concentric circle; in the homothetic pair, the inner ellipse is a homothetic copy of the outer; in the circumcircle–inellipse case, the outer conic is a circle and the inner one an ellipse; and in the Brocard and MacBeath porisms the outer circle is paired with a distinguished inellipse [2106.00715], [2012.11270], [2201.06960].

For concentric, axis-aligned ellipse pairs with semi-axes \((a,b)\) and \((a',b')\), the classical Cayley condition for the existence of a 3-periodic family reduces to
\[
\frac{a'}{a}+\frac{b'}{b}=1.
\]
Specializations include the ellipse–incircle family, where
\[
r=\frac{ab}{a+b},
\]
the circumcircle–inellipse family, where \(R=a+b\), and the homothetic family, where the inner ellipse has semi-axes \((a/2,b/2)\) [2012.11270], [2106.04521].

A finite-field version replaces the real or complex plane by \(\mathbf P^2_{\F_q}\). There, a pair of nonsingular conics \((\mathcal A,\mathcal B)\) satisfies the Poncelet triangle condition if they intersect transversally over \(\overline{\F_q}\) and there exists at least one non-degenerate triangle in \(\mathbf P^2_{\overline{\F_q}}\) inscribed in \(\mathcal A\) and circumscribed about \(\mathcal B\) [1604.00436]. This arithmetic formulation preserves the closure mechanism while changing the statistical and counting questions.

## 2. Algebraic criteria and parametrizations

A central algebraic criterion is Cayley’s condition. If \(\mathcal A,\mathcal B\) correspond to symmetric matrices \(A,B\), one forms
\[
\Delta(t)=\det(tA+B)=d_0+d_1t+d_2t^2+d_3t^3,
\]
and expands
\[
\sqrt{\Delta(t)}=H_0+H_1t+H_2t^2+\cdots.
\]
For triangles, the pair satisfies the Poncelet triangle condition if and only if
\[
H_2=0.
\]
Over \(\overline{\F_q}\), the curve \(u^2=\Delta(t)\) is an elliptic curve \(E\), and closure after three steps is equivalent to a torsion condition \(\tau\in E[3]\), algebraically encoded by \(H_2=0\) [1604.00436].

In the real setting, two parametrization strategies are especially prominent. One is the elementary vertex parametrization
\[
P_1(t)=(a\cos t,\;b\sin t),
\]
with the remaining vertices obtained by tangent construction to the caustic [2201.06960], [2106.04521]. The other is the affine reduction of the outer ellipse to the unit circle, followed by a Blaschke-product description of the three vertices. In that framework, if the inner conic has foci \(f,g\in\mathbb C\), the vertices \(z_1,z_2,z_3\) are roots of
\[
B(z)=\lambda,\qquad |\lambda|=1,
\]
or, in symmetric form,
\[
z_1+z_2+z_3 = f+g + (fg)\lambda,\qquad
z_1z_2+z_2z_3+z_3z_1 = fg + (f+g)\lambda,\qquad
z_1z_2z_3 = \lambda.
\]
This representation makes many center coordinates rational in \(\lambda\), and therefore amenable to elimination and locus computations [2106.00715], [2508.02368].

A recurring analytic normal form for a center locus is
\[
X(\lambda)=u\,\lambda+v\,\lambda^{-1}+w.
\]
When \(u,v,w\in\mathbb C\), the map \(t\mapsto u e^{it}+v e^{-it}+w\) traces an ellipse centered at \(w\), with semiaxes \(|u|+|v|\) and \(\bigl||u|-|v|\bigr|\). This lemma underlies many conicity results for center loci [2106.00715].

## 3. Loci of classical and derived triangle centers

The modern literature on Poncelet triangle families is dominated by locus problems: as the triangle moves, where do classical centers go? A major unifying result states that over any generic pair of nested ellipses, if a triangle center can be written as
\[
X=\alpha X_2+\beta X_3,
\]
then its locus is an ellipse; if the family admits a third stationary center \(X_k\), then any fixed affine combination
\[
X=\alpha X_2+\beta X_3+\gamma X_k
\]
also has an elliptic locus [2106.00715]. Closely related work proves that in a generic nested pair, any center \(X=\alpha X_2+\beta X_3\) traces an exact ellipse, and notes that among the first \(1000\) Kimberling centers, exactly \(226\) lie on the Euler line and are fixed affine combinations of \(X_2,X_3\) [2102.09438].

In the confocal family, it is now standard that the loci of \(X_1,X_2,X_3,X_4\) are ellipses [2201.06960], and the theory of locus ellipticity explains many additional cases. In the confocal pair, \(45\) of the first \(200\) centers \(X_k\) have elliptic loci by linear-combination arguments involving \(X_2\), \(X_3\), and the stationary Mittenpunkt \(X_9\) [2106.00715]. In the incircle family, \(38\) such centers \(X_k\le 200\) are proven elliptic [2106.00715].

Not all loci are conics. Over the confocal pair, the symmedian point \(X_6\) traces a quartic, \(X_{11}\) traces the inner caustic, and \(X_{59}\) yields a 6-cusped sextic [2201.06960]. Over derived triangles, ellipticity can break down: for the orthic triangles of the confocal family, only the circumcenter locus remains a single ellipse, while other center loci become piecewise-elliptic or non-smooth [2106.04521].

Recent work extends conicity beyond the usual triangle itself. The locus of the orthocenter \(X_4\) over a Poncelet family between two nested ellipses is not only a conic, but axis-aligned and homothetic to a \(90^\circ\)-rotated copy of the outer ellipse [2508.02368]. For a fixed point \(P\), the locus of its isogonal conjugate over the moving family is also a conic, although the expected degree was four; it becomes a parabola when \(P\) lies on the quartic envelope of the circumcircle, and a line when \(P\in \mathcal E\) [2508.02368].

A further extension concerns inversive constructions. Over any generic Poncelet triangle family, the circumcenter \(X_3'\) of the inversive triangle \(A'B'C'\) traces a nondegenerate conic \(L_3'\). The type depends on the location of the inversion center \(O\) relative to the region \(R\) swept by the usual circumcircle: ellipse if \(O\notin R\), hyperbola if \(O\in\mathrm{int}(R)\), parabola if \(O\in\partial R\) [2604.26035].

## 4. Invariants and stationary phenomena

Poncelet triangle families support a large collection of metric invariants. The app-oriented synthesis of the subject lists, for different families, conserved quantities such as perimeter \(L\), area \(A\), \(\sum s_i^2\), \(\sum \cos\theta_i\), \(\sum \cot\theta_i\), \(r/R\), Brocard angle \(\omega\), and \(\rho^2=4R^2-(\sum s_i^2)/2\) [2106.04521]. The precise invariant package depends strongly on the family.

For the confocal family, invariants include \(L\), \(r/R\), and \(\sum \cos\theta_i\) [2106.04521]. For the ellipse–incircle family, one has fixed \(R=(a+b)/2\), fixed \(r=ab/(a+b)\), and
\[
\sum_{i=1}^3\cos\theta_i=\frac{a^2+4ab+b^2}{(a+b)^2},
\qquad
\prod_{i=1}^3\sin\frac{\theta_i}{2}=\frac{ab}{2(a+b)^2}.
\]
For the circumcircle–inellipse family,
\[
\prod_{i=1}^3\cos\theta_i=\frac{ab}{2(a+b)^2},
\qquad
\sum_{i=1}^3 s_i^2=4(a+2b)(2a+b).
\]
For the homothetic family,
\[
A=\frac{3\sqrt3}{4}ab,\qquad
\sum_{i=1}^3 s_i^2=\frac92(a^2+b^2),\qquad
\cot\omega=\frac{\tfrac{\sqrt3}{2}(a^2+b^2)}{ab},
\]
so the Brocard angle is constant [2012.11270].

The homothetic pair and the Brocard porism are linked by a variable similarity. In the homothetic pair, every triangle has constant area
\[
A=\frac{3\sqrt3}{2}ab,
\]
constant sum of squared side lengths
\[
\sum s_i^2=\frac92(a^2+b^2),
\]
and therefore constant Brocard angle, with
\[
\cot\omega=\sqrt3\,\frac{a^2+b^2}{2ab}.
\]
In the Brocard porism, the Brocard points coincide with the foci of the inellipse and are stationary, while \(X_3\) and \(X_6\) are fixed [2009.07647].

Stationary centers are a structural organizing principle. Different families are named by a center \(X_k\) that remains fixed: \(X_9\) for confocal, \(X_1\) for incircle, \(X_3\) for circumcircle–inellipse, \(X_2\) for homothetic, \(X_4\) for dual, \(X_6\) for excentral [2201.06960]. More specialized incircle-based families with stationary \(X_1\), \(X_2\), \(X_4\), or \(X_7\) have also been described, together with conserved quantities such as \(\sum\sin(\theta_i/2)\), \(\sum\tan(\theta_i/2)\), and fixed center distances [2512.17440].

Invariant power is another theme. For any concentric ellipse pair supporting a Poncelet 3-periodic family, the power of the common center \(O\) with respect to the moving circumcircle and Euler circle is constant [2102.09438]. More recently, for any generic Poncelet family there exist two fixed points \(P_3,P_5\) such that the power of \(P_3\) to each circumcircle and the power of \(P_5\) to each Euler circle are constant [2604.26035].

## 5. Special families, degeneracies, and envelopes

The ellipse-inscribed, circle-caustic family exhibits especially rich Euclidean behavior. For an outer ellipse
\[
E:\;x^2/a^2+y^2/b^2=1
\]
and an inner circle
\[
K:\;(x-x_c)^2+(y-y_c)^2=r^2,
\]
the radius \(r\) is determined by Cayley’s closure condition, and many classical centers sweep conics. The loci of \(X_2\), \(X_4\), \(X_7\), \(X_5\), \(X_8\), and \(X_{36}\) are given explicitly in terms of \((a,b,x_c,y_c)\); in particular, the locus of \(X_{36}\), the inversive image of \(X_1\) in the moving circumcircle, is always a perfect circle [2409.19464].

A decisive degeneracy occurs when the family contains an equilateral triangle. This happens precisely when the incircle center \(C\) lies on the “equilateral ellipse”
\[
E_{eq}:\;\frac{x^2}{a_{eq}^2}+\frac{y^2}{b_{eq}^2}=1,
\quad
a_{eq}=\frac{a\,c^2}{a^2+3b^2},
\quad
b_{eq}=\frac{b\,c^2}{3a^2+b^2}.
\]
When \(C\in E_{eq}\), the nine-point-center locus collapses from an ellipse to a straight segment, the circular locus of \(X_{36}\) degenerates to a straight line, the Feuerbach point \(X_{11}\) becomes stationary on the incircle, and the isogonal conjugate \(X_{59}\) collapses from a high-degree oval to an ellipse internally tangent to the outer ellipse [2409.19464].

The same equilateral mechanism appears in a different guise for circle-inscribed families. If a Poncelet family \(T_o^*\) inscribed in the unit circle and circumscribing an ellipse \(E_c\) contains an equilateral triangle, then the focus \(X_{110}\) of the Kiepert in-parabola is stationary and equal to
\[
X_{110}=\Bigl(\tfrac1f+\tfrac1g\Bigr)^{-1}\in\mathbb T,
\]
where \(f,g\) are the complex foci of \(E_c\) [2512.16678].

Envelope phenomena are equally prominent. For the incircle family, the envelope of the moving circumcircle is exactly the union of two fixed circles, and the envelope of the radical axis of incircle and circumcircle is a conic whose type depends on the location of the incircle center relative to the circumcenter locus [2409.19464]. In a more general nested-ellipse setting, the envelope of the circumcircle and the envelope of the radical axis contain a conic component if and only if the caustic is a circle; in the circumcircle case, the envelope is the union of two circles [2508.02368].

## 6. Arithmetic, generalized closure, and broader directions

The finite-field analogue recasts the family in arithmetic-geometric terms. If
\[
\Psi=\{\text{transverse pairs }(\mathcal A,\mathcal B)\},
\qquad
\Gamma=\{\text{such pairs satisfying PTC}\},
\]
then, assuming \(\operatorname{ch}(\F_q)\neq2,3\),
\[
\frac{q-16}{q(q+1)}\le\frac{|\Gamma|}{|\Psi|}\le\frac{q+5}{(q-2)(q-3)},
\]
so
\[
\frac{|\Gamma|}{|\Psi|}=\frac1q+O(q^{-2})\qquad(q\to\infty).
\]
Thus, over finite fields, Poncelet-triangle pairs are statistically rare: asymptotically, about one pair in \(q\) satisfies the triangle condition [1604.00436]. The same work conjectures analogous asymptotics \(\sim \tau_n/q\) for \(n\)-gons, with experimental values
\[
\tau_3=1,\;
\tau_4=3,\;
\tau_5=1,\;
\tau_6=4,\;
\tau_7=1,\;
\tau_8=6,\;
\tau_9=2.
\]

A distinct extension is the “general closure case,” where different sides of the triangle may be tangent to different members of the same pencil of conics. In the bic-II family, the incenter still sweeps a circle, while the barycenter becomes a sextic and only one excenter remains circular. In the conf-II family, the incenter is elliptic only in the special confocal case, but two excenters still sweep the same ellipse. In the full three-caustic cases bic-III and conf-III, no tested center locus remains a conic [2108.05430]. This suggests that some low-degree loci are robust under partial loss of symmetry, whereas full multi-caustic freedom destroys conicity.

Another broad generalization replaces the inner circle by a central conic. A triangle inscribed in a fixed circle and circumscribed about a central ellipse or hyperbola exists if and only if
\[
(R^2-d_+^2)(R^2-d_-^2)=4\,\varepsilon\,b^2\,R^2,
\]
with \(\varepsilon=+1\) for ellipses and \(\varepsilon=-1\) for hyperbolas. This is a generalized Chapple–Euler relation, reducing to the classical bicentric formula when the foci coalesce [2603.28822], [2603.00001]. In special cases where the circumcenter coincides with the center or a focus of the inner conic, quantities such as \(\sin^2A+\sin^2B+\sin^2C\), \(\cos A\cos B\cos C\), the orthic inradius, and the polar-circle radius become invariant [2603.28822].

The subject has also developed an experimental and computational branch. A browser-based app supports real-time exploration of over \(400\) premade experiments, automatic conic detection via direct least-squares fitting, and numerical testing of many metric invariants [2106.04521]. This computational culture has influenced adjacent work on “wrought-iron” loci [2201.06960] and on parabolic accessories such as inparabolas and circumparabolas, whose focuses, directrices, and perspectors often sweep lines, circles, ellipses, or parabolas over selected Poncelet families [2110.06356].

Taken together, these developments show that the Poncelet triangle family is not a single rigid object but a framework connecting projective geometry, elliptic curves, affine and inversive constructions, integrable billiards, arithmetic statistics, and computational experimentation. A plausible implication is that the persistent appearance of conic loci, stationary centers, and low-degree envelopes reflects a common algebraic mechanism: once the family is reduced to a one-parameter model with sufficiently symmetric dependence on that parameter, unexpectedly simple geometry becomes visible.

Source: https://www.emergentmind.com/topics/poncelet-triangle-family