Recognition: unknown
On Solving Dual Conformal Integrals in Coulomb-branch Amplitudes and Their Periods
Pith reviewed 2026-05-05 05:00 UTC · model claude-opus-4-7
The pith
An infinite family of all-loop planar four-point integrals is solved recursively, indexed by binary strings without consecutive 1s, with ladders and zigzags as the two extremes.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The authors define an infinite family of all-loop, planar, dual conformal four-point integrals — "binary DCI integrals" — that arise from f-graphs in N=4 super Yang-Mills and can be solved recursively by second-order "boxing" differential equations. Each integral evaluates to a single-valued harmonic polylogarithm labeled by a binary string of 0s and 1s with no consecutive 1s, where the no-consecutive-1s rule is the planarity (extended Steinmann) constraint. Ladder integrals correspond to the string 10…0 and zigzag integrals (from the antiprism f-graphs) to the alternating string 1010…; all other planar binary DCIs sit between these two. The zigzag periods reproduce the known ζ_{2L+1} zigzag
What carries the argument
The "boxing" differential operator on cross-ratios z, z̄ — z̄z∂_z∂_z̄ or (1−z)(1−z̄)∂_z∂_z̄, depending on which external leg is acted on — recursively peels a loop off a planar DCI integral, mapping it to a centrally symmetric word over {0,1} that records the operator sequence. Combined with single-valuedness and parity-oddness, this fixes each integral uniquely; planarity forbids consecutive 1s, giving the extended Steinmann subspace. Periods are computed via graphical functions in the HyperlogProcedures package, with f-graph topology providing identities (P_{n_1,…,n_r} = P_{n_r,…,n_1}) that follow from the freedom to choose any four-cycle as external points.
If this is right
- Every planar four-point DCI integral solvable by iterated boxing carries a unique binary-string label with no consecutive 1s, and the count of such integrals at L loops follows the Fibonacci sequence.
- The antiprism f-graphs — conjecturally the largest-coefficient f-graphs at each even loop order in the four-point correlator — produce exactly the zigzag DCIs with period proportional to ζ_{2L+1}.
- Periods of these binary integrals saturate a basis for single-valued motivic multiple zeta values up to weight 21 (ten loops), giving an explicit physical realization of that basis.
- The reversal identity P_{n_1,…,n_r} = P_{n_r,…,n_1} for binary periods follows from the f-graph being agnostic to which four points are taken external, providing a graphical proof of nontrivial relations among single-valued MZVs.
- Higher-loop ladders, zigzags, and 2×n fishnets all sit inside this same family, with fishnet periods coinciding with zigzag periods via shared f-graph origin.
Where Pith is reading between the lines
- The fact that binary-string periods saturate the single-valued MZV basis up to weight 21 but cannot continue to do so beyond weight 26 suggests a natural extension to 'ternary' Steinmann SVHPLs (allowing all three boxing operators) might restore saturation and deserves a separate count.
- Conjecture 1 (P_{2,3,…,3,2} ∝ ζ_{6m+9}) hints at a hidden depth-reduction mechanism for symmetric strings of this shape, possibly provable by exploiting a symmetry of the f-graph that flips the inner 3-blocks.
- If the bounding inequality is rigorous, ladder and zigzag integrals would serve as universal calibrators for any conjectured all-loop resummation of four-point Coulomb-branch amplitudes.
- The graphical identity P(f^{(L)}) = P′(f^{(L−M)} × f^{(M)}) suggests viewing the period map as a coalgebra on f-graphs, which could connect these results to recently observed antipodal self-dualities for fishnet integrals.
Load-bearing premise
The claimed numerical sandwich — that zigzag periods are the smallest and ladder periods the largest among all binary-string periods at each loop — is observed up to ten loops but not proven, so the all-loop bound rests on extrapolation from finite data.
What would settle it
Compute the period of any binary-string DCI integral at L ≥ 11 loops and check whether it falls outside the closed interval [P_zigzag(L), P_ladder(L)]; a single binary string whose period exceeds the ladder or falls below the zigzag would refute the conjectured bound and the proposed organization of these single-valued MZVs.
read the original abstract
We define and study infinite families of all-loop planar, dual conformal invariant (DCI) integrals, which contribute to four-point Coulomb-branch amplitudes and correlators in ${\cal N}=4$ supersymmetric Yang-Mills theory, by solving ``boxing'' differential equations via \texttt{HyperlogProcedures}~\cite{hyperlogprocedures}; The resulting single-valued harmonic polylogarithmic functions (SVHPL) are nicely labeled by ``binary'' strings of $0$ and $1$ without consecutive $1$'s. These functions are special cases of the so-called generalized ladders studied in~\cite{Drummond:2012bg}, where extended Steinmann relations (no consecutive $1$'s) are imposed due to planarity. Our results can be viewed as ``two-dimensional'' extensions of the well-known ladder integrals to many more infinite families of DCI integrals: the ladders have strings with a single $1$ followed by all $0$'s, and the other extreme, which nicely evaluate to the ``zigzag'' SVHPL functions with alternating $1$'s and $0$'s, are nothing but the four-point DCI integrals from the very special family of anti-prism $f$-graphs (while all other binary DCI integrals lie in between these two extreme cases). We also study periods of these integrals: while their periods are in general complicated single-valued multiple zeta values (SVMZV), the ``zigzag'' DCI integrals from anti-prism gives exactly the famous ``zigzag'' periods proportional to $\zeta_{2L{+}1}$, and empirically it provides a numerical lower-bound for $L$-loop periods of any binary string, with the upper-bound given by that of the ladder (also proportional to $\zeta_{2L{+}1}$). Based on $f$-graphs as a tool for studying these periods, we discuss several interesting facts and observations about these (motivic) SVMZV and relations among them to all loops, and enumerate a basis for them up to $L=10$.
Editorial analysis
A structured set of objections, weighed in public.
Forward citations
Cited by 1 Pith paper
-
Loops and legs: ABJM amplitudes from $f$-graphs
ABJM amplitudes of arbitrary multiplicity and loop order can be reconstructed from squared amplitudes encoded in a permutation-symmetric generating function of planar f-graphs.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.