[Paper Review] The degree distribution in bipartite planar maps: applications to the Ising model

Mireille Bousquet‐Mélou, Gilles Schaeffer|ArXiv.org|Nov 4, 2002

Stochastic processes and statistical mechanicsMathematics2 references58 citations

TL;DR

This paper presents a combinatorial, bijective approach to solving the Ising and hard particle models on random planar lattices by characterizing the generating function of bipartite planar maps according to black and white vertex degrees. Using a tree-map correspondence and algebraic manipulation, it proves that the solutions for bounded-degree maps are algebraic, recovering and extending results previously obtained via matrix integrals with greater mathematical clarity and combinatorial insight.

ABSTRACT

We characterize the generating function of bipartite planar maps counted according to the degree distribution of their black and white vertices. This result is applied to the solution of the hard particle and Ising models on random planar lattices. We thus recover and extend some results previously obtained by means of matrix integrals. Proofs are purely combinatorial and rely on the idea that planar maps are conjugacy classes of trees. In particular, these trees explain why the solutions of the Ising and hard particle models on maps of bounded degree are always algebraic.

Motivation & Objective

• To provide a purely combinatorial derivation of the generating functions for the Ising and hard particle models on random planar lattices.
• To characterize the generating function of bipartite planar maps by the degree distribution of their black and white vertices.
• To explain the algebraic nature of the solutions for bounded-degree maps through a bijective correspondence with trees.
• To replace matrix integral techniques—common in physics literature but lacking combinatorial justification—with a rigorous, elementary combinatorial framework.

Proposed method

• The authors use the bijection between planar maps and conjugacy classes of trees to encode the degree distribution of bipartite maps.
• They define blossom trees with vertices of degree 2 and m, which encode maps with only degree-2 and degree-m vertices.
• The generating function for these trees is derived via a parametrization involving a series P satisfying a cubic algebraic equation.
• The Ising model generating function is expressed as an integral of derivatives of map-generating functions, which are then rewritten in terms of the tree parameter P.
• The integrand is shown to be a rational function of P, allowing the integral to be evaluated as a rational function of P and its boundary values.
• The final expression for the Ising generating function is algebraic, as it is a rational function of P and P at z=0, both of which are algebraic.

Experimental results

Research questions

• RQ1 Why are the solutions of the Ising and hard particle models on maps of bounded degree always algebraic?
• RQ2 How can the degree distribution of black and white vertices in bipartite planar maps be systematically encoded in a generating function?
• RQ3 Can the algebraic structure of the Ising model solutions be explained combinatorially, without relying on matrix integrals?
• RQ4 What is the precise connection between the generating function of maps and the generating function of the underlying tree structures in the bijection?
• RQ5 How can the Ising model on tetravalent maps be solved combinatorially, and why does the solution remain algebraic?

Key findings

• The generating function for m-valent bipartite maps rooted at a white vertex is expressed as $ I(X,Y,u) = A - \binom{m-1}{2}A^2 + \frac{mx}{2}\int_0^v \frac{\partial}{\partial x}(\bar{M}(x,y,z) + \bar{M}(y,x,z)) \frac{dz}{z} $, where $ A $ satisfies $ A = xy(1 + (m-1)A)^{m-1} $.
• For tetravalent maps, the Ising generating function $ I(X,Y,u) $ is algebraic of degree 7, and can be expressed rationally in terms of a series $ P $ satisfying $ P = 1 + 3xyP^3 + v^2 \frac{P(1+3xP)(1+3yP)}{(1-9xyP^2)^2} $.
• The series $ P(x,y,0) $, which counts blossom trees with only degree-4 vertices, is related to $ A $ by $ P(x,y,0) = 1 + 3A $, and this connection simplifies the final expression.
• The integral in the Ising generating function is reduced to a rational function of $ P $, ensuring the algebraicity of the final solution.
• The solution for the Ising model on tetravalent maps matches the parametrization previously obtained via matrix integrals, confirming consistency while providing a combinatorial derivation.
• The leading terms in the vertex expansion of $ I(tX,tY,u) $ are $ t(2X) + t^2(9X^2 + XY(8u^2 + u^4)) + O(t^3) $, consistent with explicit enumeration of small maps.