Outer-mass double box

The content on this page was written by AI under human supervision.

The planar two-loop double box with six perimeter lines of a common mass, a massless central rung and massless legs, which is finite in four dimensions and, up to an algebraic prefactor, equal to a single weight-four polylogarithm. Caron-Huot and Henn computed it in 2014, and it is recomputed here without using their answer.

The integral

Planar two-loop double box: a wide rectangle split into two squares by a thin black vertical central rung; the six perimeter segments are drawn as red double lines and labeled m in red; four black external legs labeled p1 to p4 carry arrows, the two on the left (bottom and top) entering and the two on the right (top and bottom) leaving; a gray vertical double arrow labeled s stands beside the left pair of legs and a gray horizontal double arrow labeled t runs above the top edge
Red double lines are the six perimeter propagators of common mass $m$; the thin black central rung and the four external legs $p_1,\dots,p_4$ are massless. Gray arrows mark the two channels, $s=(p_1+p_2)^2$ across the left pair of legs and $t=(p_2+p_3)^2$ across the top pair. The maximal cut of these seven lines is algebraic, which is why the answer is a polylogarithm.

The outer-mass double box is a planar two-loop four-point diagram: two boxes sharing a central rung, with all six propagators around the perimeter carrying a common mass $m$ and the central rung and the four external legs massless, $p_i^2=0$. It is a scalar Feynman integral of the standard form

$$I_G \;=\; \int \prod_{\ell=1}^{L} d^d l_\ell \;\prod_{e\in G}\frac{1}{q_e^2-m_e^2}\,,\qquad d=4-2\varepsilon,$$

where $l_1,\dots,l_L$ are the loop momenta, the product runs over the internal lines $e$ of the graph $G$, $q_e$ is the momentum flowing through line $e$ and $m_e$ its mass; here there are $L=2$ loops, six of the seven masses equal $m$ and the rung is massless. The integral depends on the two Mandelstam invariants through the dimensionless ratios

$$u=\frac{4m^2}{-s},\qquad v=\frac{4m^2}{-t},\qquad s=(p_1+p_2)^2,\quad t=(p_2+p_3)^2,$$

both positive in the Euclidean region $s,t\lt0$. The integral is finite, so it is evaluated directly in $d=4$ with no expansion in $\varepsilon$. Write $I_5$ for the integral with every propagator present once, in $d=4$ at $m=1$ with the measure normalized as in Caron-Huot and Henn, and $g_{10}(u,v)=-\tfrac18\,s^2t\,\sqrt{1+u}\,\sqrt{1+u+v}\;I_5$ for its normalized form at Euclidean $u,v\gt0$, keeping the labels these two functions carry in the literature cited below. The prefactor is fixed by the leading singularitythe algebraic prefactor read off the maximal cut; dividing by it leaves a pure function of uniform transcendental weight of the integral.

At a glance

Caron-Huot and Henn computed this integral in 2014 by solving a differential equation for it directly in four dimensions. Besides rational functions of $u$ and $v$, the alphabet contains only three square roots, $\beta_u=\sqrt{1+u}$, $\beta_v=\sqrt{1+v}$ and $\beta_{uv}=\sqrt{1+u+v}$. The substitution $u=(1-w^2)(1-z^2)/(w-z)^2$, $v=4wz/(w-z)^2$ rationalizes all three at once and maps the Euclidean region to $0\lt z\lt w\lt1$. The twelve-letter alphabet, the maximal cuts and the symbolthe tensor of logarithms that encodes the iterated-integral structure of a polylogarithmic function; two functions with the same symbol differ at most by lower-weight functions times constants also follow from the Landau equations of the graph alone, and the symbol can be recovered by regression from high-precision numerical samples of the integral. The computation here combines the two ideas.

The alphabet is rebuilt here from a Landau singularity analysis of the diagram, which gives six rational letters ($u$, $v$, $1{+}u$, $1{+}v$, $u{+}v$, $1{+}u{+}v$) and a list of candidate algebraic letters (those that go to their inverse when the sign of one square root is flipped); a Regge-limit argument removes the candidates containing $\sqrt u$ or $\sqrt v$ and leaves exactly the twelve letters. The general weight-four symbol in the twelve letters has too many unknown coefficients to determine directly from numerical values of the integral, and three conditions first reduce the space to a basis of 161 weight-four symbols. The first is integrabilitythe condition that a tensor of logarithms $\mathcal{S}$ is the symbol of an actual function, written $d\mathcal{S}\wedge d\mathcal{S}=0$ of the symbol, the condition introduced by Goncharov, Spradlin, Vergu and Volovich in 2010. The second is a first-entry condition that lets only the $s$- and $t$-channel thresholds and pseudo-thresholds open a branch cut. The third is a definite parity under the sign flips of the three square roots, odd in $\beta_u$ and $\beta_{uv}$ and even in $\beta_v$. In the Landau bootstrap the maximal cuts select the answer in this space; here an integer-relation fita search, by PSLQ or lattice reduction, for exact rational coefficients that match high-precision numerical values of the 161 coefficients to high-precision values of the integral on a Euclidean grid selects it instead.

Closed form

In four dimensions the normalized integral is the leading singularity times one element of the constrained basis:

$$g_{10} = -\tfrac18\,s^2 t\,\beta_u\beta_{uv}\, I_5 = \mathrm{LS}\cdot G_{52},\qquad \mathrm{LS}=\tfrac18\sqrt{s(s-4m^2)}\,\sqrt{st\,(st-4m^2(s+t))},$$

where $\mathrm{LS}$ is the leading singularity written in $s$, $t$ and $m$. The fit assigns a nonzero coefficient to exactly one basis element, $G_{52}$, and that coefficient is $1$ (the subscripts on $G$, and on the letters $L_6$ and $L_9$ below, follow the numbering in section 4.5 of Polylogarithmic, elliptic, and Calabi–Yau Feynman integrals from a hybrid bootstrap). The symbol of $G_{52}$, written $\mathcal S(G_{52})$, coincides term by term with the symbol published by Caron-Huot and Henn and later rederived from the Landau equations. It has eighteen terms over the twelve-letter alphabet

$$\Big\{\,u,\ v,\ 1{+}u,\ 1{+}v,\ u{+}v,\ 1{+}u{+}v,\ \tfrac{\beta_u-1}{\beta_u+1},\ \tfrac{\beta_v-1}{\beta_v+1},\ \tfrac{\beta_{uv}-1}{\beta_{uv}+1},\ \tfrac{\beta_{uv}-\beta_u}{\beta_{uv}+\beta_u},\ \tfrac{\beta_{uv}-\beta_v}{\beta_{uv}+\beta_v},\ \tfrac{\beta_{uv}-\beta_u\beta_v}{\beta_{uv}+\beta_u\beta_v}\,\Big\},$$

where $\beta_v$ enters only through three of the six algebraic letters, and each letter is fixed up to a constant factor, which drops out of the symbol. Writing $L_6=(\beta_u-1)/(\beta_u+1)$ and $L_9=(\beta_{uv}-\beta_u)/(\beta_{uv}+\beta_u)$, the symbol begins

$$\mathcal S(G_{52}) = -L_6{\otimes}u{\otimes}L_6{\otimes}L_9 + L_6{\otimes}(1{+}u){\otimes}L_6{\otimes}L_9 - L_6{\otimes}u{\otimes}L_9{\otimes}L_6 + L_6{\otimes}(1{+}u){\otimes}L_9{\otimes}L_6 + \cdots$$

(fourteen more terms); the full eighteen-term symbol is written out in section 4.5 of the same paper.

The function $G_{52}$ follows from its symbol by iterated integration in $w$ and $z$ from the line $w=z$, which is the limit $u,v\to\infty$ where $g_{10}$ vanishes. Every letter allowed in the first entry equals one on that line, so the integration converges and introduces no constant beyond the symbol. On the physical sheet $g_{10}$ is singular only at the normal thresholds $s=4m^2$ and $t=4m^2$ and as $s,t\to\infty$; on other sheets it is also singular at $s=0$, $t=0$, $s+t=0$ and on the pseudo-threshold $st=4m^2(s+t)$.

Checks

Checks
pointclosed formindependent valuedigits
$(u,v)=(16/3,\,10/3)$$0.04393211443\ldots$$0.04393211443\ldots$38
$(u,v)=(10/3,\,16/3)$$0.05332108844\ldots$$0.05332108844\ldots$40
$(u,v)=(23/4,\,9/4)$$0.05552972348\ldots$$0.05552972348\ldots$38
$(u,v)=(5/2,\,4)$$0.09129263017\ldots$$0.09129263017\ldots$38
$(u,v)=(10/3,\,5/2)$$0.09313854862\ldots$$0.09313854862\ldots$38
$(u,v)=(11/2,\,7/3)$$0.05657968990\ldots$$0.05657968990\ldots$38
$(u,v)=(4,\,9/2)$$0.04860778225\ldots$$0.04860778225\ldots$38
$(u,v)=(4,\,4)$ (published value)$0.05303862978\ldots$$0.05303862978\ldots$28

The independent values are auxiliary-mass-flow evaluations (AMFlow) of the same integral at Euclidean points used neither in a fit nor to fix a boundary value, together with the value at $u=v=4$ published by Caron-Huot and Henn in 2014. The closed form agrees with them to the number of leading digits in the last column.

Evaluator

Evaluator

Python 3 with mpmath; the zip also contains the script, which reads its data file from the unzipped bundle folder, so run it there. A full run takes a few seconds on a laptop.

Tools
toolrole
SOFIALandau singularity analysis of the diagram, giving the rational letters and the candidate algebraic letters
Ansatzerbuilds the weight-four words over the twelve letters and reduces them by integrability, first entries and square-root parity
PSLQinteger-relation and lattice-reduction fit of the basis coefficients to the numerical values
AMFlowhigh-precision values of the integral for the fit, and the independent numerical evaluations in the Checks table

Same family

The massless planar double box is the same graph with all seven lines massless. It depends on the single ratio $x=t/s$, and its Laurent expansion in $\varepsilon$, poles and finite part, is a sum of harmonic polylogarithms of $x$. The central-mass double box is the same graph with the mass $m$ on the central rung alone and all six perimeter lines massless.

The papers

References

S. Caron-Huot and J. M. Henn, Iterative structure of finite loop integrals, JHEP 06 (2014) 114 [arXiv:1404.2922]the original computation of this integral and the value at $u=v=4$ in the Checks table
H. S. Hannesdóttir, A. J. McLeod, M. D. Schwartz and C. Vergu, Applications of the Landau bootstrap, Phys. Rev. D 111 (2025) 085003 [arXiv:2410.02424]the twelve-letter alphabet, the maximal cuts and the symbol from the Landau equations; the Regge-limit argument used at the alphabet step
O. Barrera, A. Dersy, R. Husain, M. D. Schwartz and X. Zhang, Analytic regression of Feynman integrals from high-precision numerical sampling, JHEP 01 (2026) 014 [arXiv:2507.17815]the same symbol recovered by a lattice fit to numerical samples; the fitting method used here
A. B. Goncharov, M. Spradlin, C. Vergu and A. Volovich, Classical polylogarithms for amplitudes and Wilson loops, Phys. Rev. Lett. 105 (2010) 151605 [arXiv:1006.5703]the symbol of a polylogarithm and its integrability condition
M. Correia, M. Giroux and S. Mizera, SOFIA: Singularities of Feynman integrals automatized [arXiv:2503.16601]the package used here for the singularity analysis of the graph, from which the letters are read off
X. Liu and Y.-Q. Ma, AMFlow: a Mathematica package for Feynman integrals computation via auxiliary mass flow [arXiv:2201.11669]the auxiliary-mass-flow method behind the numerical values used in the fit and in the Checks table

← back to Feynman diagrams