[Paper Review] The Kontsevich tetrahedral flow in 2D: a toy model

Anass Bouisaghouane|arXiv (Cornell University)|Feb 20, 2017

Homotopy and Cohomology in Algebraic Topology2 references4 citations

TL;DR

This paper proves that in two dimensions, the Kontsevich tetrahedral flow associated with the first graph (Γ₁) is Poisson-cohomology trivial, meaning it arises as the Schouten bracket of the Poisson bivector with a vector field X. The key result is the explicit construction of such a trivializing vector field X, which is shown to descend to periodic quotients of R², such as the 2-torus, preserving well-definedness under lattice reduction.

ABSTRACT

In the paper "Formality conjecture" (1996) Kontsevich designed a universal flow $\dot{\mathcal{P}}=\mathcal{Q}_{a:b}(\mathcal{P})=aΓ_{1}+bΓ_{2}$ on the spaces of Poisson structures $\mathcal{P}$ on all affine manifolds of dimension $n \geqslant 2$. We prove a claim from $ extit{loc. cit.}$ stating that if $n=2$, the flow $\mathcal{Q}_{1:0}=Γ_{1}(\mathcal{P})$ is Poisson-cohomology trivial: $Γ_{1}(\mathcal{P})$ is the Schouten bracket of $\mathcal{P}$ with $\mathcal{X}$, for some vector field $\mathcal{X}$; we examine the structure of the space of solutions $\mathcal{X}$. Both the construction of differential polynomials $Γ_{1}(\mathcal{P})$ and $Γ_{2}(\mathcal{P})$ and the technique to study them remain valid in higher dimensions $n \geqslant 3$, but neither the trivializing vector field $\mathcal{X}$ nor the setting $b:=0$ survive at $n\geqslant 3$, where the balance is $a:b=1:6$.

Motivation & Objective

• To verify a claim from Kontsevich's 1996 'Formality conjecture' that the Γ₁ flow is cohomologically trivial in dimension 2.
• To construct an explicit vector field X such that Γ₁(P) = [ [P, X] ] for any Poisson bivector P in 2D.
• To analyze the structure of the space of solutions X for the trivializing condition.
• To extend the results to periodic Poisson structures on the 2-torus by showing that both the flow and the vector field X descend under lattice quotients.

Proposed method

• The paper expands the Kontsevich differential polynomials Γ₁ and Γ₂ in local coordinates (x,y) on a 2D manifold, exploiting the fact that only P¹² ≠ 0.
• It computes the non-vanishing component of Γ₁(P) explicitly in terms of derivatives of the single coefficient u = P¹², yielding a sixth-order nonlinear PDE-like expression.
• It constructs the vector field X via the Hamiltonian H such that X = (H_y, -H_x), with H derived from the cohomological triviality condition.
• It verifies that the vector field X, when derived from the flow, is well-defined on the 2-torus R²/L for any lattice L, by showing its coefficients are L-periodic Fourier series.
• It uses trigonometric identities to reduce products of sines and cosines in the flow and vector field expressions to Fourier series of higher wavenumbers.
• It proves that the image of the flow and the vector field under the quotient map π: R² → R²/L remains well-defined on the compact manifold M² = R²/L.

Experimental results

Research questions

• RQ1 Is the Kontsevich tetrahedral flow Γ₁(P) cohomologically trivial in 2D Poisson geometry, i.e., is it a coboundary under the Poisson differential?
• RQ2 Can a vector field X be explicitly constructed such that Γ₁(P) = [ [P, X] ] for any Poisson bivector P in dimension 2?
• RQ3 Does the trivializing vector field X descend to a well-defined vector field on the 2-torus R²/L when P is L-periodic?
• RQ4 Why does the Γ₂ component vanish identically in 2D, while it contributes in higher dimensions?
• RQ5 How do the wavenumbers of the flow and vector field behave under lattice reduction, and why do they remain periodic?

Key findings

• The only non-zero component of the Γ₁ flow in 2D is given by the expression: u_xxx u_y³ - u_yyy u_x³ - 3u_xxy u_x u_y² + 3u_xyy u_x² u_y.
• The Γ₂ component vanishes identically in 2D due to index constraints and skew-symmetrization, confirming only Γ₁ contributes.
• A trivializing vector field X exists such that Γ₁(P) = [ [P, X] ], with components derived from the Hamiltonian H = -16α₁₁³π⁴ sin(2πx)cos(2πy)(sin²(2πx)sin²(2πy) + 2cos²(2πx)sin²(2πy) + cos²(2πy)cos²(2πx)).
• The vector field components F and G are explicitly computed as trigonometric polynomials involving frequencies 2π, 6π, and 8π, confirming periodicity on the 2-torus.
• The flow and vector field remain well-defined on the 2-torus R²/L because their coefficients are L-periodic Fourier series with wavenumbers bounded below by the original lattice.
• The construction of X is compatible with the lattice quotient, and the divergence-free part of X is realizable via Kontsevich graphs, confirming consistency with deformation theory.