A combinatorial formula for Wilson loop expectations on compact surfaces

Imagine you are standing on a complex, multi-holed donut (a surface like a torus or a sphere with handles). On this surface, you have drawn a bunch of tangled strings. These strings can cross over each other, loop around the holes, or even form isolated circles that don't touch anything else.

In the world of theoretical physics, specifically a field called Yang-Mills theory, these strings represent "loops" of force. Physicists want to calculate a specific number for these loops, called a Wilson Loop Expectation. Think of this number as a "score" or a "probability weight" that tells you how likely a certain tangled configuration of strings is to appear in the quantum universe.

For decades, calculating this score was like trying to solve a massive, 4D puzzle using only a calculator. It required complex integrals and heavy machinery.

Thierry Lévy's paper is like discovering a new, almost purely combinatorial (counting-based) recipe to solve this puzzle. Instead of doing heavy calculus, he shows you how to count, arrange, and multiply simple numbers to get the answer.

Here is the breakdown of his "recipe" using everyday analogies:

1. The Map and the Territory

Imagine your tangled strings divide the surface of your donut into separate "rooms" or "faces."

The Strings: These are the edges of a graph.
The Crossings: Where strings cross, they create "vertices" (corners).
The Rooms: The empty spaces between the strings are the "faces."

Lévy's formula says: To find the score of the whole tangled mess, you don't need to look at the whole mess at once. Instead, you look at each room individually and assign it a "label."

2. The Labels (Highest Weights)

Imagine each room gets a special ID card. In math, these are called Highest Weights.

Think of an ID card as a list of numbers (like a barcode).
The rule is: If two rooms share a wall (an edge), their ID cards must be compatible. They can't be just any random numbers; they must follow a specific "handshake" rule. If Room A's ID is 10, 5, 2, Room B's ID might have to be 10, 6, 2. They differ by just one tiny step.

3. The Score Calculation (The Magic Formula)

Once you assign a valid ID card to every room, you calculate a "contribution" for that specific arrangement. The total score is the sum of all possible valid arrangements.

For each arrangement, the score is made of three ingredients:

The Size of the Room (Exponential Factor):
Imagine the room has an area. The bigger the room, the more "weight" it carries. This is calculated using a special number (the Casimir) associated with the room's ID card. It's like a tax based on the room's size.
The Complexity of the Room (Dimensions):
Every ID card corresponds to a specific type of "shape" or "representation." Some shapes are simple (like a line), others are complex (like a 3D cube). The formula multiplies the "size" (dimension) of these shapes.
- Analogy: If a room is a simple square, it contributes a small number. If it's a complex, multi-layered crystal, it contributes a huge number.
The Crossings (The Sine and Cosine Factor):
This is the most unique part of Lévy's discovery. When two strings cross at a vertex, they create a "knot" in the logic.
- Type 1 Crossing: The ID cards on the four sides of the crossing are arranged in a way that is "balanced." The contribution is a Cosine of an angle.
- Type 2 Crossing: The ID cards are arranged differently. The contribution is a Sine of an angle.
- The Angle: This angle isn't measured with a protractor on the paper. It's a "mathematical angle" derived from the difference between the ID card numbers. It's like the angle between two gears meshing together.

4. The "Spin Network" Analogy

The paper uses a concept called Spin Networks. Imagine a spiderweb where every intersection is a tiny machine.

The strings are the wires.
The rooms are the power sources.
The crossings are the gears.
Lévy's formula essentially says: "To get the total energy of the web, you just need to multiply the power of each source by the efficiency of each gear, and then add up all the possible ways you could wire the power sources."

5. Why is this a Big Deal?

It's "Almost Purely Combinatorial": Before this, you needed to integrate over infinite possibilities (calculus). Lévy reduced it to a sum of counting problems (combinatorics). It's like switching from solving a differential equation to just adding up a list of numbers.
It Explains the "Makeenko-Migdal Equations": These are famous rules in physics that describe how the score changes if you stretch or shrink a room. Lévy's formula proves these rules naturally, almost as a side effect. It's like discovering that a new recipe automatically explains why a cake rises perfectly every time.
It's Rational: Even though the formula involves sines and cosines (which often give messy, irrational numbers like $\pi$ or $\sqrt{2}$ ), Lévy proves that when you multiply them all together for a valid configuration, the messy parts cancel out, and you always get a clean, rational number (like $3/4 $or$ 5$).

Summary

Thierry Lévy has found a new way to calculate the "probability score" of tangled strings on a curved surface. Instead of using heavy calculus, he showed that you can:

Divide the surface into rooms.
Assign compatible number-lists (ID cards) to each room.
Multiply the room sizes, the complexity of the ID cards, and some trigonometric "gear ratios" at the crossings.
Add up all the results.

It turns a chaotic quantum physics problem into a structured, countable puzzle, revealing a hidden order in the way these strings interact.

Here is a detailed technical summary of the paper "A Combinatorial Formula for Wilson Loop Expectations on Compact Surfaces" by Thierry Lévy.

1. Problem Statement

The paper addresses the computation of Wilson loop expectations for the 2-dimensional Yang–Mills holonomy process with values in the unitary group $U(N)$ on a compact oriented surface $\Sigma$ (which may have a boundary).

Specifically, the author seeks an explicit, almost purely combinatorial expression for the non-normalized expectation of products of traces of holonomies along an arbitrary family of immersed curves $\ell = (\ell_1, \dots, \ell_m)$ :
$Z(\Sigma) \mathbb{E}\left[ \prod_{j=1}^m \text{Tr}(H_{\ell_j}) \right]$
where $H_{\ell_j}$ is the random holonomy along the loop $\ell_j$ , and $Z(\Sigma)$ is the Yang–Mills partition function. The curves are allowed to have transverse self-intersections (double points) but no triple points. The result must hold for arbitrary boundary conditions (constrained or free).

While previous approaches (e.g., Driver–Sengupta formula) expressed these expectations as integrals over products of unitary groups, Lévy aims to transform this into a discrete sum over combinatorial data (highest weights) involving local geometric factors at intersection points.

2. Methodology

The proof is structured as a seven-step computation that systematically transforms the integral representation of the Wilson loop expectation into a combinatorial sum.

Step 1: Integral Representation (Driver–Sengupta)

The starting point is the Driver–Sengupta formula, which expresses the Wilson loop expectation as an integral over the configuration space $U(N)^E$ (where $E$ is the set of edges of a graph associated with the loop configuration). The integrand consists of:

Traces of holonomies along the loops.
Heat kernels $p_{|F|}$ evaluated on the boundaries of the faces $F$ of the graph.
Boundary condition measures if $\Sigma$ has a boundary.

Step 2: Fourier Expansion (Peter–Weyl)

The heat kernels are expanded into their Fourier series (Peter–Weyl expansion) in terms of Schur functions $s_\lambda$ (characters of irreducible representations of $U(N)$ ).

This converts the integral into an infinite sum over configurations of highest weights $\Lambda: F \to \mathbb{Z}^N_{\downarrow}$ , assigning a highest weight $\lambda_F$ to each face $F$ .
The term associated with each configuration is called the flat contribution $F(\Lambda)$ , which involves an integral of a product of Schur functions.

Step 3: Schur–Weyl Duality

The author utilizes Schur–Weyl duality to relate the unitary group $U(N)$ to the symmetric group $S_n$ .

Schur functions are expressed as traces of projectors $\pi_\lambda$ acting on tensor powers of $\mathbb{C}^N$ .
The integrand is rewritten as the trace of a product of three linear operators acting on a large tensor product space:
1. $I(g)$ : Encoding the gauge configuration $g$ .
2. $\Pi(\Lambda)$ : Encoding the highest weight projectors.
3. $X(\Lambda)$ : A vertex-dependent operator encoding the graph topology.
This reformulation interprets the integrand as a spin network.

Step 4: Integration over Unitary Variables

The integration over $U(N)$ is performed using the Collins–Sniady formula.

This formula evaluates integrals of the form $\int_{U(N)} x^{\otimes n} \otimes (x^\vee)^{\otimes m} dx$ .
The result vanishes unless $n=m$ (a weak balance condition).
When non-zero, the integral is replaced by a sum over permutations in the symmetric group $S_n$ , effectively decoupling the edge variables and reducing the problem to a sum over permutations associated with the edges of the graph.

Step 5: Symmetric Group Modules

The computation shifts entirely to the representation theory of symmetric groups.

The branching rule for symmetric groups is used to analyze how irreducible representations $V^\mu$ of $S_n$ decompose into representations of $S_{n-1}$ .
A crucial condition emerges: the flat contribution vanishes unless the configuration of highest weights is balanced. A configuration is balanced if, for every edge $e$ , the highest weight on the right face extends the highest weight on the left face ( $\lambda_{le} \uparrow \lambda_{re}$ ).

Step 6: Summation over Permutations

The sum over permutations associated with the edges is performed.

Using orthogonality relations of the symmetric group, the sum collapses.
The result is a concise expression involving the dimensions of the representations and a scalar factor $a(v)$ associated with each vertex $v$ of the graph.

Step 7: Computation of the Scalar (Okounkov–Vershik Theory)

The final step computes the scalar $a(v)$ using the Okounkov–Vershik approach to symmetric group representations.

This involves Gelfand–Zetlin bases and Jucys–Murphy elements.
The scalar $a(v)$ is determined by the geometry of the Young diagrams (partitions) at the vertex.
Depending on the relative positions of the highest weights at the four faces surrounding a vertex, the scalar is either a cosine or a sine of an angle $\theta_v$ .
The angle is defined by the "contents" of the boxes added to the Young diagrams: $\cos \theta_v = \frac{1}{c(\lambda_e/\lambda_n) - c(\lambda_n/\lambda_w)}$ .

3. Key Contributions and Results

The Main Theorem (Theorem S.3)

The paper provides a closed-form combinatorial formula for the Wilson loop expectation:
$Z(\Sigma) \mathbb{E}\left[ \prod \text{Tr}(H_{\ell_j}) \right] = \sum_{\Lambda \in \mathcal{B}} e^{-\frac{1}{2N} \sum |F| c_{\lambda_F}} \left( \prod_{F} (d_{\lambda_F})^{e_F} \right) \left( \prod_{v} (d_{\lambda_{nv}} d_{\lambda_{sv}})^{-1/2} \right) \times \left( \prod_{v \in V_1} \cos \theta^\Lambda_v \prod_{v \in V_2} \sin \theta^\Lambda_v \right)$
Where:

$\mathcal{B}$ is the set of balanced configurations of highest weights (vanishing at free boundaries).
$d_\lambda$ is the dimension of the representation (Weyl dimension formula).
$c_\lambda$ is the quadratic Casimir number.
$e_F$ is the Euler characteristic of the face.
$V_1$ and $V_2$ partition the vertices based on whether the highest weights on the North and South faces are equal (Type 1) or distinct (Type 2).
$\theta^\Lambda_v$ is an angle determined by the "contents" of the skew Young diagrams at the vertex.

Key Technical Insights

Rationality of Flat Contributions: Although the formula involves sines and cosines of angles that are generally irrational, the author proves (Proposition 8.1) that the total product for any balanced configuration is a rational number. This is shown by analyzing "level lines" (edges with specific weight differences) and proving that transversal crossings occur in even numbers.
Vanishing Condition: The expectation is zero if the total homology class of the loop configuration does not belong to the subgroup generated by the constrained boundary components (Proposition 8.2).
Makeenko–Migdal Equations: The paper provides a short, new proof of the Makeenko–Migdal equations (differential equations governing the variation of Wilson loops with respect to face areas) as a direct corollary of the main formula (Proposition 8.3).

4. Significance

Combinatorial Rigor: The paper bridges the gap between the analytic definition of the Yang–Mills measure (via integrals) and the combinatorial structures (Young diagrams, symmetric groups) often seen in physics literature (e.g., Witten's work on IRF models). It fills in the computational gaps left by Witten's 1991 paper.
New Proof Techniques: By leveraging the Collins–Sniady formula and Okounkov–Vershik theory, the author provides a self-contained, rigorous derivation that avoids heavy reliance on lattice approximations or large $N$ limits.
Applications: The explicit formula is expected to be useful for studying the large $N$ limit (planar limit) of 2D Yang–Mills theory and for understanding the topological properties of the theory on arbitrary surfaces.
Unification: It unifies the treatment of surfaces with and without boundaries and handles arbitrary boundary conditions within a single framework.

In summary, Lévy's work transforms a complex stochastic integral problem into a tractable combinatorial sum, revealing deep connections between 2D gauge theory, representation theory of symmetric groups, and the geometry of Young diagrams.