---
title: Bolotin's Polynomial Integrability Conjecture
url: https://www.emergentmind.com/topics/bolotin-s-polynomial-integrability-conjecture
type: topic
---

# Bolotin's Polynomial Integrability Conjecture

Searching arXiv for the cited papers and closely related work to ground the article.
arXiv search query: Bolotin polynomial integrability conjecture billiards constant curvature 1706.04030
Bolotin’s Polynomial Integrability Conjecture, in its billiard-theoretic form, is the algebraic version of the classical Birkhoff Conjecture. It asserts that if a convex bounded planar billiard with smooth boundary is polynomially integrable, then its boundary is a conic; for bounded convex planar billiards this means an ellipse. In "On polynomially integrable Birkhoff billiards on surfaces of constant curvature" [1706.04030], this conjecture is resolved and extended from the Euclidean plane to the sphere and the Lobachevsky plane: polynomial integrability is characterized by boundaries that are unions of confocal conical arcs and certain admissible geodesic segments.

## 1. Conjecture and polynomial integrability

The basic notion is the following. Let $\Sigma$ be a two-dimensional surface with a Riemannian metric, and let $\Omega\subset \Sigma$ be a domain with piecewise smooth boundary. The billiard in $\Omega$ is said to be polynomially integrable if its flow has a first integral on $T\Sigma|_\Omega$ that is a polynomial in the velocity $P$ and whose restriction to the hypersurface $\{|P|=1\}$ is non-constant [1706.04030].

In the planar case, Bolotin’s Polynomial Birkhoff Conjecture is the statement that a convex bounded planar billiard with smooth boundary is polynomially integrable only when its boundary is a conic. For bounded planar billiards with connected $C^2$ boundary, the theorem proved in [1706.04030] yields the sharper conclusion that every such billiard is an ellipse. The paper formulates the constant-curvature version as: if a billiard in $\Sigma$ with a $C^2$-smooth connected boundary is polynomially integrable and the boundary is not contained in a geodesic, then the boundary is a conic, or a connected component of a conic.

This polynomial notion is closely related to analytic integrability. The paper notes that analytic integrability implies polynomial integrability, since each homogeneous part in $P$ of an analytic integral is itself a first integral. It also recalls Bolotin’s result that on simply connected complete surfaces of constant curvature with smooth connected boundary, polynomial integrability is equivalent to the existence of a polynomial integral near the unit tangent bundle to the boundary; moreover, a polynomial-in-$P$ integral is globally analytic on $T\Sigma$ [1706.04030].

The significance of the conjecture is its role as a rigid algebraic analogue of the classical Birkhoff Conjecture. The latter remains open in full generality, whereas the polynomial version admits a complete classification in constant curvature.

## 2. Constant-curvature formulation and confocality

The global classification in [1706.04030] is stated for simply connected complete two-dimensional surfaces of constant curvature realized in $\mathbb{R}^3$ with a quadratic form $\langle A x,x\rangle$.

| Surface $\Sigma$ | Model in $\mathbb{R}^3$ | Matrix $A$ |
|---|---|---|
| Euclidean plane | $\{x_3=1\}$ | $\operatorname{diag}(1,1,0)$ |
| Unit sphere | $\{x_1^2+x_2^2+x_3^2=1\}$ | $\operatorname{Id}$ |
| Hyperbolic plane | $\{x_1^2+x_2^2-x_3^2=-1,\ x_3>0\}$ | $\operatorname{diag}(1,1,-1)$ |

The metric on $\Sigma$ is induced by $\langle A x,x\rangle$ on tangent planes. Geodesics are intersections of $\Sigma$ with two-dimensional vector subspaces of $\mathbb{R}^3$, and conics on $\Sigma$ are intersections with quadrics $\{\langle Cx,x\rangle=0\}$, where $C$ is a real symmetric $3\times 3$ matrix [1706.04030].

The organizing structure for the classification is a confocal pencil. Given a real symmetric matrix $B$ not proportional to $A$, one considers
$$
\Gamma_\lambda=\Sigma\cap \{\langle B_\lambda x,x\rangle=0\},\qquad B_\lambda=(B-\lambda A)^{-1}.
$$
At singular values of $\lambda$, the definition is modified using the kernel of $B-\lambda A$, producing geodesics $\Gamma_\lambda$ or $\Gamma_\lambda(H)$ when the kernel has dimension one or two. The associated admissible geodesics are precisely those singular members, together with additional geodesics in the special tensor case
$$
B=A\,a\otimes b+b\otimes A a \pmod A,\qquad \langle a,b\rangle=0,
$$
described explicitly in the theorem statement [1706.04030].

A billiard with countably piecewise smooth boundary is called countably confocal when its regular boundary consists of arcs of conics from one confocal pencil, possibly together with segments of admissible geodesics. The paper’s main global theorem states that if a billiard in $\Sigma$ with countably piecewise $C^2$-smooth boundary is polynomially integrable and its regular boundary contains at least one non-geodesic arc, then the billiard is countably confocal [1706.04030].

The converse is due to Bolotin. Every countably confocal billiard is polynomially integrable: it has a non-trivial first integral that is linear, quadratic, or a degree $4$ polynomial in the velocity components, nonconstant on the unit velocity hypersurface. If all geodesic pieces lie in the singular members $\Gamma_\lambda$ or $\Gamma_\lambda(H)$, the integral can be chosen of degree at most $2$; the irreducible degree-$4$ case occurs exactly for confocal pencils of the special types described in Definition 1.20, when the boundary contains a segment of the corresponding admissible geodesic [1706.04030].

## 3. Duality, momentum variables, and the algebraic reduction

A central step in the proof is to pass from the boundary curve to its projective dual in momentum variables. Bolotin’s theorem states that for every polynomially integrable billiard with countably piecewise $C^2$-smooth boundary, a polynomial integral non-constant on $\{|P|=1\}$ can be chosen as a homogeneous polynomial $\Psi(M)$ of even degree in the components of the moment vector
$$
M=[r,P]=(x_2P_3-x_3P_2,\ -x_1P_3+x_3P_1,\ x_1P_2-x_2P_1),
$$
and every $C^2$-smooth boundary arc with non-zero geodesic curvature lies in an algebraic curve [1706.04030].

The projective setting uses the tautological projection $\pi:\mathbb{R}^3\setminus\{0\}\to \mathbb{RP}^2$ and the polarity induced by the quadratic form $\langle A\cdot,\cdot\rangle$. The polar of a point $[x]$ is the line
$$
\{[y]:x^\top A y=0\}.
$$
For a $C^2$-smooth curve $\alpha\subset \Sigma$ with nonzero geodesic curvature, the $\Sigma$-dual curve $\alpha^*\subset \mathbb{RP}^2$ is obtained by projecting $\alpha$, taking the tangent lines to $\pi(\alpha)$, and sending each tangent line to its $A$-polar point. In the momentum-plane language, $\alpha^*$ consists of the points dual to the tangent lines of $\pi(\alpha)$ [1706.04030].

The relevant projective quadric is the absolute conic
$$
\mathcal{A}=\{[M]\in \mathbb{CP}^2:\langle AM,M\rangle=0\}.
$$
For $\Psi$ of even degree $2n$, Bialy and Mironov consider the rational function
$$
G(M)=\frac{\Psi(M)}{\langle AM,M\rangle^n}.
$$
They show that for each point $B$ on the dual curve $\alpha^*$, the restriction of $G$ to the tangent line $T_B\alpha^*$ is invariant under the unique projective involution fixing $B$ that preserves each line through $B$ and interchanges its intersection points with $\mathcal{A}$. This is the $\mathcal{A}$-angular symmetry centered at $B$ [1706.04030].

The reduction obtained by Bialy and Mironov is decisive. The complex Zariski closure of $\alpha^*$ is algebraic; each non-linear irreducible component $\gamma$ generates a rationally integrable $\mathcal{A}$-angular billiard; and all singular points and inflection points of $\gamma$, if any, lie in $\mathcal{A}$ [1706.04030]. The billiard problem is therefore transformed into a rigid algebraic problem about projective curves with constrained local geometry relative to the absolute conic.

## 4. The conic conclusion and resolution of the conjecture

The new ingredient supplied in [1706.04030] is the algebraic theorem that completes the reduction. It states: let $\mathcal{A}\subset \mathbb{CP}^2$ be a conic, either smooth or a union of two lines. Every irreducible algebraic curve $\gamma\subset \mathbb{CP}^2$ different from a line and from $\mathcal{A}$ that generates a rationally integrable $\mathcal{A}$-angular billiard is a conic.

This is the conic conclusion. It implies that every non-linear irreducible component of the dual curve is a conic, so the original boundary arc is itself a conic arc. Bolotin’s earlier theorem then forces the entire billiard to be confocal whenever the boundary contains a non-geodesic conical arc. Combining these results yields the complete constant-curvature classification: polynomial integrability holds exactly for billiards whose boundaries are unions of confocal conical arcs and admissible geodesic segments [1706.04030].

The proof of the conic conclusion combines local and global algebraic geometry. Bialy and Mironov’s Hessian identity is written, for a defining equation $f(x,y)=0$ of $\gamma$, in the form
$$
g^3(x,y)H(f)(x,y)=H(gf)(x,y)=c\,Q(x,y)^{3m-3},\qquad c\neq 0,
$$
where $Q(x,y)=\langle AM,M\rangle$ in an affine chart and
$$
H(f)=f_{xx}f_y^2-2f_{xy}f_xf_y+f_{yy}f_x^2.
$$
The paper analyzes this identity near points of $\gamma\cap \mathcal{A}$ using Puiseux asymptotics. It proves that branches transverse to $\mathcal{A}$ at regular points are regular and quadratic, branches tangent to $\mathcal{A}$ are quadratic, and in the union-of-two-lines case branches transversal to both lines are quadratic [1706.04030].

These local constraints are then combined with global intersection-theoretic relations, including the Hessian intersection formula
$$
3d(d-2)=\sum_{C\in \gamma} h(\gamma,C),
$$
together with Plücker-type relations and Hironaka’s genus bound
$$
\sum \delta \le \frac{(d-1)(d-2)}{2}.
$$
The resulting inequalities force the degree $d$ of $\gamma$ to be $2$. Hence $\gamma$ is a conic [1706.04030].

In the planar bounded case, the corollary is exactly the desired resolution: every bounded polynomially integrable planar billiard with a $C^2$-smooth connected boundary is an ellipse.

## 5. Examples, complexification, and relation to the classical Birkhoff problem

The basic model example is the planar disk billiard. In the disk centered at the origin in $\mathbb{R}^2_{(x_1,x_2)}$, the quantity
$$
x_1P_2-P_1x_2
$$
is linear in $P$ and is a first integral under elastic reflections. More generally, for any conic boundary on $\Sigma$, Bolotin’s proposition yields a quadratic first integral in $P$; and in the special admissible-geodesic configurations associated with Definition 1.20, the minimal degree may be $4$ [1706.04030].

The confocal description is explicit on $S^2$ and $H^2$. On the sphere, conics are intersections with quadratic cones $\{\langle Cx,x\rangle=0\}$, confocal families are given by $B_\lambda=(B-\lambda A)^{-1}$, and admissible geodesics arise at singular values of $\lambda$ through $\Gamma_\lambda=\Sigma\cap K_\lambda^\perp$ and, in the special tensor case, through the geodesics $\{\langle r,a\rangle=0\}$ and $\{\langle r,Ab\rangle=0\}$ when nonempty. The hyperbolic case is analogous with $A=\operatorname{diag}(1,1,-1)$ [1706.04030].

The paper also develops a complex version. In complexified $\mathbb{R}^3$ with $A\in\{\operatorname{diag}(1,1,0),\operatorname{diag}(1,1,\pm1)\}$ and $\Sigma$ equal to either $\mathbb{C}^2$ or $\Sigma_\pm=\{\langle Ax,x\rangle=\pm1\}$, a complex billiard is a collection of holomorphic curves $\Gamma_t\subset \Sigma$ that are not isotropic lines. Polynomial integrability is defined by a polynomial $\Phi(r,P)$ in $P$ satisfying invariance under geodesic translation and under the nontrivial $\langle\cdot,\cdot\rangle$-orthogonal involution fixing the tangent line at reflection points. The main complex theorem states that every polynomially integrable complex billiard containing at least one non-geodesic curve is confocal, with a homogeneous integral $\Psi(M)$ of degree at most $4$, quadratic in $M$ except in the special admissible-geodesic cases where the minimal degree is $4$ [1706.04030].

The conceptual relation to the classical Birkhoff Conjecture is precise. The classical conjecture says that among strictly convex bounded planar billiards with smooth boundary, the only caustic-integrable ones are ellipses; this remains open in full generality. The polynomial assumption is much stronger and leads to an algebraic classification through duality, angular symmetry, and projective algebraic geometry. A plausible implication is that polynomial integrability isolates exactly the rigid, confocal examples that classical theory regards as model integrable billiards.

## 6. Related uses of the term and broader mathematical context

The expression “Bolotin’s Polynomial Integrability Conjecture” also appears in convex geometry, where it refers to a different problem: polynomiality of parallel section functions of convex bodies. For a convex body $K\subset \mathbb{R}^n$, one considers
$$
A_{K,\xi}(t)=V_{n-1}(K\cap H_{t,\xi}),\qquad H_{t,\xi}=\{x\in \mathbb{R}^n:\langle x,\xi\rangle=t\}.
$$
In that setting, the conjecture asserts that if $n$ is even there are no polynomially integrable convex bodies, while if $n$ is odd the only ones are ellipsoids; centered ellipsoids indeed have polynomial parallel section functions of degree $n-1$. The paper "An extension of polynomial integrability to dual quermassintegrals" [1803.00199] states that this conjecture has been resolved under a $C^\infty$ boundary assumption by Koldobsky, Merkurjev, and Yaskin, and then extends the characterization to dual quermassintegrals without a smoothness assumption: for odd index $m$ no convex body has a polynomial $m$th dual parallel section function, and for even $m$ the only such bodies are ellipsoids.

A further, distinct usage appears in the literature on planar polynomial vector fields. "Differential Galois Theory and non-Integrability of Planar Polynomial Vector Fields" [1707.04446] does not explicitly cite Bolotin, but it discusses a commonly accepted formulation according to which, for a generic planar polynomial differential system of degree $d\ge 2$, there is no nonconstant polynomial first integral, equivalently the polynomially integrable locus is thin in parameter space. The paper develops differential-Galois obstructions to rational, hence polynomial, first integrals via variational equations, Liouville’s theorem, and the Risch equation, and presents parameter families where non-abelian Galois groups force non-integrability [1707.04446].

These usages are mathematically different but structurally related. In each case, polynomial integrability imposes strong algebraic constraints and leads either to a rigid classification or to generic nonexistence results. For billiards, the classification is complete on surfaces of constant curvature: the polynomially integrable examples are exactly the confocal ones, and in the bounded planar smooth connected case they are precisely ellipses [1706.04030].

Source: https://www.emergentmind.com/topics/bolotin-s-polynomial-integrability-conjecture