Tailoring Three-Point Functions and Integrability

TL;DR

The paper develops an integrability-based framework to compute tree-level planar three-point structure constants in $ ext{N}=4$ SYM by representing the three single-trace operators as Bethe eigenstates of SU(2) spin chains. It introduces a cutting–flipping–sewing procedure: each closed chain is split into open subchains, one subchain is flipped to a bra, and overlaps (scalar products) of Bethe states are evaluated to assemble the three-point function. The results yield explicit, normalized expressions for generic non-extremal correlators in the SU(2) sector, including cases with BPS descendants, and reveal a clean scaling (thermodynamic) limit described by finite-gap, algebraic-curve data. The work connects weak-coupling structure constants to classical string theory via a finite-gap picture and outlines clear paths for extending to higher-rank sectors and loop corrections, highlighting open problems and potential string-field-theory interpretations.

Abstract

We use Integrability techniques to compute structure constants in N=4 SYM to leading order. Three closed spin chains, which represent the single trace gauge-invariant operators in N=4 SYM, are cut into six open chains which are then sewed back together into some nice pants, the three-point function. The algebraic and coordinate Bethe ansatz tools necessary for this task are reviewed. Finally, we discuss the classical limit of our results, anticipating some predictions for quasi-classical string correlators in terms of algebraic curves.

Figures (6)

• Figure 1: (a) The planar tree-level contraction of three single trace operators in the double line notation. The diagram has a pair of pants topology (sphere with three punctures). (b) In the spin chain picture, each of the three single trace operators corresponds to a state on a closed chain. The closed chains are cut into right and left open chains where the external states are represented. The three states are sewed together into the three-point function by overlapping the wave functions on each right chain with the wave function on the left subchain of the next operator.
• Figure 2: Since $i^{-1}R(0)=\mathbb{P}$, $i^{-L}\hat{T}(0)$ is the unit shift operator to the right, see the upper right corner of the figure. By definition of inverse $i^{L}\hat{T}^{-1}(0)$ is the unit shift operator to the left. When computing $\hat{T}'(0)$ the derivative will act on one of the $R$'s at position $k$, hence the sum in the last line. $R'(0)=\mathbb{I}$ which leads to the last line. We see that $\hat{T}'(0)$ is a sum of terms which are almost a total shift of one unit to the right except for a small "impurity" at site $k$. Therefore, when multiplying by $\hat{T}^{-1}(0)$, we almost get the identity operator acting on the full Hilbert space. The impurity simply leads to a permutation of acting on sites $k$ and $k-1$. Hence (\ref{['HABA2']}) leads to (\ref{['Hsu2']}).
• Figure 3: The R-matrix is a very special operator. It is designed to satisfy the Yang-Baxter equation depicted at the top. This equation is arguably the most important equation in quantum integrability. In particular it implies the $LLR=RLL$ type relation represented at the bottom. To prove this relation we simply move one of the vertical lines from the left group to the right region using Yang-Baxter and repeat this procedure until all the vertical lines are to the right of the R-matrix. From this simple equation all the algebra relations (table \ref{['FZalg']}) of the monodromy matrix elements $\mathcal{A},\mathcal{B},\mathcal{C},\mathcal{D}$ follow trivially as explained in the main text. Finally multiplying this equation by $R^{-1}$ and taking the trace over the tensor product of the two auxiliary spaces $0_1$ and $0_2$ we derive (\ref{['ttcom']}).
• Figure 4: (a) Three-point function of $SU(2)$ operators at tree level. All contractions are such that R-charge is preserved. This is the simplest non-trivial configuration which is not extremal. Note that the number of excitations on each chain is subject to the condition $N_1= N_2+N_3$. Also, if we denote by $l_{ij}$ the number of propagators between operators $i$ and $j$, we have $l_{12}=L_1-N_3$, $l_{13}=N_3$ and $l_{23}=L_3-N_3$. (b) We show the partitions of the excitations in ${\mathcal{O}}_1$. Note that we only need to use the operation of cutting for this operator, as a part of its excitations $(\alpha)$ are contracted with those of ${\mathcal{O}}_2$ and another part $(\bar{\alpha})$ with those of ${\mathcal{O}}_3$.
• Figure 5: In the classical limit the Bethe roots distribute themselves along disjoint cuts in the complex plane. The resolvent (\ref{['res']}) has square root cuts at the position of the Bethe roots in the continuum limit and defines a Riemann surface. The final result for the structure constants (\ref{['chapeau']}) is given by some integrals in this surface. The integral involving the quasi-momenta $q(v)$ is over the $A$-cycles of the surface denoted in green in the figure. The integral involving the density $\rho(u)$ is taken over the cuts. $n=1,2$ are the mode numbers in this real data example.