The Setting
Sidorenko’s conjecture asks whether for every bipartite graph and every graph , where is the homomorphism density of in and is the edge density of . Here is the bipartite graph between 13 points and 22 faces, each face a triple of points, that joins each point to the faces containing it. For a function on the symmetric matrices over a finite field, the homomorphism density of in the weighted graph on these matrices with weight between and is the mean, over a matrix at each point and a matrix at each face , of the product of the weights over the edges of .
The points are numbered and the faces . A fixed table lists the points of each face in its slots , , , and a corner is a pair of a face and a slot, at the point .✦ A second table holds the sides , and of each face in slots , , .✦ The sides run over 33 point pairs, each of them a side of exactly two faces, and a pair is written with its left point smaller than its right point .
Proposition 4.7 takes such means with a kernel of prescribed sign at each of a chosen set of corners, the active corners, and compares them, as the finite field of odd prime order grows, with a model that sees the 13 point matrices only through the quadratic characters of the determinants of their differences along the 33 sides. The comparison is split along the event that all 33 differences are nonsingular.
Signs and Rank Kernels
Let be a finite field with quadratic character . The space of symmetric matrices of size over is the subspace of the matrices over with .✦ For a symmetric bilinear form on a finite-dimensional space over with radical , the radical quotient form is the nondegenerate form on ✦, and the discriminant sign of is of the determinant of a Gram matrix of , which is when has rank zero.✦ The sign of a matrix is the discriminant sign of the form on .✦
The signed rank layer of rank and sign is the set of matrices in of rank and sign ✦, and its layer mass is the proportion .✦ The rank kernel of rank and sign is a real function on whose value at is determined by whether lies in . Throughout, , every kernel has rank , and .
A configuration is a family of matrices in , one at each point. Means are uniform: averages over all configurations and over all .✦
An activity pattern assigns to each face a set of active slots, and assigns an integer to each corner. The face local mean of at a configuration is (1)✦ and the configuration product is , over the 22 faces.✦
For and integers and , the pair polynomial is (2)✦ The face polynomial of a face at is the product of one pair polynomial for each pair of active slots of , (3) and the configuration polynomial is .✦
A configuration is fully transverse when for each of the 33 pairs ; is the event that is fully transverse, and its complement.✦ The pair sign of at a pair with is , which is when the difference is invertible and otherwise.✦ The actual signs of form the vector with when the pair sign at is and otherwise.✦
For a real number and a sign vector , the Boolean configuration is (4) where is the side of through and ; the product over the pairs in is the Boolean face polynomial of . The model at a configuration is .
Transverse and Singular Parts
The transverse difference is the mean of the indicator of times the difference of the configuration product and the configuration polynomial: (5) The matrix singular tail and the Boolean singular part are the means of the configuration product and of the model off : (6)✦ (7)✦
Each of these depends on . A limit as is taken through the odd primes, the primes , with .✦
The Singular Tail Threshold
A profile assigns a natural number to each face and to each pair ; it is geometric in dimension when, for every face , , for each side of , and the profile sizes , with the side of in slot , satisfy .✦
A fixed table gives each face a three-bit label, and the separation count of a pair is the number of bits at which the labels of the two faces containing differ.✦ With the triangular function , the profile cost of in dimension is (8)✦ where is the face natural cost of .
A tail tuple consists of natural numbers , , and triples , of natural numbers✦; its tail sizes are ✦, its tail heights ✦, and its tail offset is .✦ Take a geometric profile in dimension and any set of faces, whose members are called the full faces. A global tail choice picks at each full face a common parameter , an integer with and , and feasible base parameters : integers between and such that, with , for every and every in , (9)✦
The global tail tuple of at a full face is the planted tail tuple , and at any other face the lower tail tuple .✦ The tail gain of a full face is the planted gain of its tuple, (10)✦ with , and the tail gain of any other face is , where is the largest tail height of its tuple. The global tail gain is the sum of the 22 tail gains.✦
The singular tail threshold is an integer, at least , for which the following global tail dichotomy holds.✦ Let , let be a geometric profile in dimension , and let be a global tail choice for any set of full faces. Then (11)✦
The Claim
Theorem 1. Let be a natural number, an activity pattern and a family of integers on the corners. Over every finite field , (12)✦ Suppose now that every is or , and let through the odd primes, with . Then the Boolean singular part tends to , (13)✦ if , the transverse difference tends to , (14)✦ and if , the mean of the configuration product and the mean of the model have the same limit, (15)✦
In the paper, the result reads: for every fixed even , every activity pattern and all incidence signs, (16) uniformly over patterns and signs, where the paper’s is the dimension , is the set of active corners, is the set of points of at its active corners, is the sign at the corner of at the point , is the matrix at the face , and is a family of signs on the point pairs.
The library states the result for fixed , and , as limits through the odd primes, and compares the mean of the configuration product with the mean of the model at the actual signs of the configuration rather than with a mean over .
What Is Checked
The claim is stated over the rank kernels of rank on , the face local means and face polynomials of an activity pattern, the Boolean configuration at the actual signs , the event of fully transverse configurations, uniform means over configurations, limits as through the odd primes, and the singular tail threshold .