UUITP–34/19

All-order α0 -expansion of one-loop open-string integrals

Carlos R. Mafraa and Oliver Schlottererb

a
STAG Research Centre and Mathematical Sciences,
University of Southampton, Highfield, Southampton SO17 1BJ, UK and
b
Department of Physics and Astronomy, Uppsala University, 75108 Uppsala, Sweden.

We present a new method to evaluate the α0 -expansion of genus-one integrals over open-string
punctures and unravel the structure of the elliptic multiple zeta values in its coefficients. This is
done by obtaining a simple differential equation of Knizhnik–Zamolodchikov–Bernard-type satisfied
by generating functions of such integrals, and solving it via Picard iteration. The initial condition
involves the generating functions at the cusp τ → i∞ and can be reduced to genus-zero integrals.

INTRODUCTION location of the vertex operator associated with the j th

arXiv:1908.09848v1 [hep-th] 26 Aug 2019

external string state is parameterized by the coordinates

Elliptic analogues of polylogarithms [1, 2] and multiple
zj = uj τ +vj with uj , vj ∈ (0, 1), where τ is the modulus
zeta values [3] have become a driving force in higher-order
with Im τ > 0, see figure 1, and we define zij ≡ zi − zj .
computations of scattering amplitudes in quantum field
By suitable involutions of the torus [13], one ob-
theories and string theories. The study of their differen-
tains the surfaces describing the scattering of open-string
tial equations and their connections with modular forms
states, the cylinder and the Möbius strip. The two
turned into a vibrant research area at the interface of
boundaries of the cylinder will be parameterized by the
particle phenomenology, string theory and number the-
A-cycle zj ∈ (0, 1) and its displacement zj ∈ τ2 +(0, 1) by
ory. In the same way as a variety of Feynman integrals
half a B-cycle, i.e. uj ∈{0, 21 } and dzj =dvj . See figure 2.
has been recently expressed in terms of elliptic polyloga-
rithms and iterated integrals of modular forms [4, 5], the zr+3 ... Im(z)
• •
low-energy expansion of one-loop open-string amplitudes . τ• || • τ +1
• zr+2 • ..
introduces elliptic multiple zeta values (eMZVs) [6–8].
• •
So far, the appearance of eMZVs in one-loop open- zr+1 zn
zr+1 zr+2 . . . zn τ
string amplitudes arose from direct integration over τ
• • • • • • 2 +1
2
the punctures on a genus-one worldsheet of cylinder or z3 ...
• •
Möbius-strip topology. Although there is no conceptual .

|
• z2 • .. Re(z)
bottleneck in extending the techniques of [6–8] to arbi- • • • • • || . •. . • •
trary multiplicities and orders in the inverse string ten- z1 zr z1 =0 z2 z3 zr 1
sion α0 , in this letter we will present a new method to
evaluate these genus-one integrals which is related to el- FIG. 2: The cylinder parameterization.
liptic associators [9] and Tsunogai’s derivations dual to
Eisenstein series [10]. The results are given by eMZVs in The massless n-point one-loop amplitudes of the open
their minimal form [3, 11] and reveal elegant structures in superstring give rise to integrals of the form (z1 = 0) [6]
the α0 -expansions. More details will be given in a longer Z Y n  n
X 
companion paper [12]. (k ) (k )
dzj fi1 j11 fi2 j22 · · · exp sij G(zij , τ ) , (1)
C(∗) j=2 i<j

OPEN-STRING INTEGRALS AT GENUS ONE with differing integration domains C(∗) for the cylinder
and the Möbius strips. For planar cylinders, we set ∗ →
One-loop string amplitudes are described by correlation
1, 2, . . . , n and parametrize the domain as
functions of vertex operators in a conformal field the-
ory over a genus-one Riemann surface, the torus. The C(1, 2, . . . , n) = {zj=2,...,n ∈ R, 0<z2 < . . . <zn <1} , (2)
see figure 2 and [12] for the non-planar analogue with ∗ →
r+1,...,n (k)
1,2,3,...,r . Furthermore, in the integrand of (1), fij ≡
(k)
f (zij , τ ) denote the Laurent coefficients of the doubly-
periodic Kronecker–Eisenstein series defined by [2, 14]
 Im z  θ10 (0, τ )θ1 (z + η, τ )
Ω(z, η, τ ) = exp 2πiη , (3)
Im τ θ1 (z, τ )θ1 (η, τ )
∞
X
C
FIG. 1: We parameterize the torus through the lattice Z+τ Z Ω(z, η, τ ) = η k−1 f (k) (z, τ ) . (4)
with identifications z ∼
= z+1 ∼
= z+τ along the A- and B-cycle. k=0
 2

The simplest examples of the coefficient functions are where the (n−1)! × (n−1)! matrix Dη~τ is a differential
Im z
f (0) (z, τ ) = 1 and f (1) (z, τ ) = ∂z log θ1 (z, τ ) + 2πi Im τ, operator w.r.t. ηj . Its detailed form will be exemplified
(k≥2) in the next section and follows from the properties (5)
and higher f Pn (z, τ ) do not have

any poles in z.
Finally, exp i<j sij G(zij , τ ) in (1) is the Koba–
of
R the Green function, the vanishing of boundary terms
Nielsen factor written in terms of dimensionless Man- dvj ∂vj (. . .) and the mixed heat equation (u, v ∈ R)
delstam invariants sij = −2α0 ki · kj and Green functions
2πi∂τ Ω(uτ +v, η, τ ) = ∂v ∂η Ω(uτ +v, η, τ ) . (9)
G(z, τ ) subject to the universal differential equation
∂vi G(zij , τ ) = −f (1) (zij , τ ) (5) Most importantly, the form of Dη~τ (B|C) does not depend
on the planar or non-planar integration cycle A, and its
2πi∂τ G(zij , τ ) = −f (2) (zij , τ ) − 2ζ2 , entries are linear in the dimensionless Mandelstam in-
where ∂vP
i
is the derivative along the cylinder boundary, variants sij and therefore in α0 .
∞
and ζn = k=1 k1n with n≥2 denote Riemann zeta values. Hence, the α0 -expansion of the genus-one integrals Zη~τ
A. Generating functions: Instead of handling the follows from the solution of (8) via Picard iteration,
α0 -expansion of the individual integrals (1) as in the X∞ 
1 k τ
Z Z τ1 Z τk−1
τ
method of [6–8], we will evaluate the following generating Zη~ (A|1, B) = dτ1 dτ2 . . . dτk
2πi i∞ i∞ i∞
function of integrals (with η23...n = η2 +η3 + . . . +ηn ) k=0
X
n
Z Y n
X  × (Dη~τk · . . . · Dη~τ2 · Dη~τ1 )(B|C) Zη~i∞ (A|1, C) (10)
Zη~τ (∗|1, 2, . . . , n) = dzj exp sij G(zij , τ ) (6) C∈Sn−1
j=2 i<j
C(∗)
with matrix products Dη~τk · . . . · Dη~τ2 Dη~τ1 . As an initial
× Ω(z12 , η23...n , τ )Ω(z23 , η3...n , τ ) . . . Ω(zn−1,n , ηn , τ ) . value, the degeneration Zη~i∞ at the cusp τ →i∞ will be
(k ) (k )
The integrands fi1 j11 fi2 j22 · · · in (1) relevant to n-point expressed in terms of disk integrals with two additional
open-superstring amplitudes have k1 +k2 + . . . = n−4 and punctures from the pinching of the A-cycle in figure 1.
reside at the order of ηj−3 of (6). Moreover, (n≥8)- As will be detailed in [12], the entire τ -dependence of
point integrands additionally involve holomorphic Eisen- Dη~τ is carried by Weierstrass functions (with G0 = −1)
stein series G`≥4 (τ ) = −f (`) (0, τ ) [6] multiplying (1) at ∞
G0 X
k1 +k2 + . . . = n−4−` as seen at the ηj−3−` -order of (6). ℘(η, τ ) = − + (k−1)η k−2 Gk (τ ) . (11)
Although the cylinder contribution to one-loop open- η2
k=4
string amplitudes is localized at purely imaginary τ as
This allows us to decompose
drawn in figure 2, we will define and evaluate the in-
tegrals (6) for generic τ in the upper half plane with ∞
X
Re τ 6= 0. In view of the parental torus, Zη~τ (1, 2, . . . , n|·) Dη~τ = (1−k)Gk (τ )rη~ (k ) , (12)
r+1,...,n k=0
and Zη~τ ( 1,2,3,...,r |·) will be referred to as planar and non-
planar A-cycle integrals, respectively. where rη~ (k ) are (n−1)!×(n−1)! matrices whose entries
Möbius-strip integrals can be reconstructed by spe- are independent of τ , rational functions of ηj , linear in
cializing planar A-cycle integrals to Re τ = 21 , and the sij and may involve second derivatives ∂ηi ∂ηj . Note that
cancellation of tadpole divergences from one-loop open- rη~ (2 ) = 0 and rη~ (2p−1 ) = 0 ∀ p ∈ N by (11).
superstring amplitudes can be analyzed as in [15].
C. The main result: Based on (12), the open-string
The A-cycle integrand (6) at n points involves n−1
integrals (10) can be expressed in terms of iterated Eisen-
factors of the Kronecker–Eisenstein series (4) at different
stein integrals
arguments. The second entry Zη~τ (∗|A) specifies permu-
Z i∞ 0
tations A = a1 a2 . . . an ∈ Sn of these arguments, and dτ
Ω(. . .) at different zaj , ηaj are related by the Fay identity γ(k1 , k2 , . . . , kr |τ ) = Gkr (τ 0 ) γ(k1 , . . . , kr−1 |τ 0 )
τ 2πi
Ω(z1 , η1 , τ )Ω(z2 , η2 , τ ) = Ω(z1 , η1 +η2 , τ )Ω(z2 −z1 , η2 , τ ) (13)
subject to γ(∅|τ ) = 1 and tangential-base-point regu-
+ Ω(z2 , η1 +η2 , τ )Ω(z1 −z2 , η1 , τ ) . (7) τ
Pn larization [16], e.g. γ(0|τ ) = 2πi . As the main result
Repeated use of (7) and imposing η1 = − j=2 ηj only of this work, we can therefore bring the open-string α0 -
leaves (n−1)! independent permutations of the integrand expansion into the following elegant form:
in (6), and we will use a basis of Zη~τ (∗|1, B) with permu- ∞
tations B ∈ Sn−1 acting on 2, 3, . . . , n.
X X
Zη~τ (A|1, B) = γ(k1 , k2 , . . . , kr |τ ) (14)
B. The differential equation: As will be derived in r=0 k1 ,k2 ,...,kr
=0,4,6,8,...
[12], the τ -derivatives of (6) can be written as r
Y X
× (kj −1) rη~ (kr . . . k2 k1 )B C Zη~i∞ (A|1, C) ,
X
2πi∂τ Zη~τ (A|1, B) = Dη~τ (B|C)Zη~τ (A|1, C) , (8)
j=1 C∈Sn−1
C∈Sn−1
 3
Pp
where rη~ (kr . . . k2 k1 )≡rη~ (kr ) · · · rη~ (k2 )rη~ (k1 ). Since that mix under τ -derivatives (s12...p ≡ 1≤i<j sij ),
each order in α0 is expressible in terms of eMZVs [6–8], 
the rη~ (k ) should be matrix representations of Tsuno- 2πi∂τ Zητ2 ,η3 (∗|1, 2, 3) = − 2ζ2 s123 (19)
gai’s derivations k dual to Eisenstein series [10]. In 1 2  1 2 
particular, (12) brings the differential equation (8) of + s12 2 ∂η2 − ℘(η2 +η3 , τ ) + s13 2 ∂η3 − ℘(η3 , τ )
Zη~τ into the same form as that of the elliptic Knizhnik– 
+ s23 12 (∂η2 −∂η3 )2 − ℘(η3 , τ ) Zητ2 ,η3 (∗|1, 2, 3)

Zamolodchikov–Bernard associator [9], where the deriva-
+ s13 ℘(η2 +η3 , τ ) − ℘(η3 , τ ) Zητ2 ,η3 (∗|1, 3, 2) .
 
tions k act on its non-commutative arguments.
The decomposition of eMZVs into iterated Eisenstein
The resulting matrix entries of the 2 × 2 differential op-
integrals automatically incorporates all their relations
erator in (8) read
over the rational numbers [11]. Moreover, the deriva-
tion of (14) does not rely on any relation among the Dητ2 ,η3 (2, 3|2, 3) = −2ζ2 s123 + s12 21 ∂η22 −℘(η2 +η3 , τ )
 
Mandelstam invariants. The n-point results of this work
+ s23 12 (∂η2 −∂η3 )2 −℘(η3 , τ ) + s13 21 ∂η23 −℘(η3 , τ )
   
are valid for 21 n(n−1) independent sij , and one can still
impose momentum conservation when applying the α0 - Dητ2 ,η3 (2, 3|3, 2) = s13 ℘(η2 +η3 , τ )−℘(η3 , τ ) ,
 
(20)
expansion of Zη~τ to string amplitudes.
and the first row is always sufficient to generate the re-
maining entries via permutations of sij and ηj , e.g.
EXAMPLES FOR DIFFERENTIAL OPERATORS s12 ↔s13
Dητ2 ,η3 (3, 2|3, 2) = Dητ2 ,η3 (2, 3|2, 3) η ↔η (21)
2 3
In this section, we present (n<4)-point examples of τ τ
s12 ↔s13
the matrix-valued differential operators Dη~τ in (8), and Dη2 ,η3 (3, 2|2, 3) = Dη2 ,η3 (2, 3|3, 2) η2 ↔η3 .
the four-point case is relegated to the appendix. All-
One can read off the 2 × 2 matrix representations of the
multiplicity expressions as well as detailed derivations of
derivations (k 6= 2),
the differential equations can be found in [12].
 1 1
A. Two points allow for a single planar and non- rη2 ,η3 (k ) = δk,0 2ζ2 s123 − s23 (∂η2 −∂η3 )2 − s12 ∂η22
planar A-cycle integral (6) each, 2 2
1 2

k−2

s12 −s13

− s13 ∂η3 12×2 + η23 −s12 s13 (22)
Z 1 2 
Zητ2 (1, 2|1, 2) = dv2 Ω(v12 , η2 , τ ) es12 G(v12 ,τ ) (15)
  
0
+ η2k−2 s012 s12 +s
0
23
+ η3k−2 s13 +s
0
23 s13
0 ,
Z 1 τ
Zητ2 dv2 Ω(v12 + τ2 , η2 , τ ) es12 G(v12 + 2 ,τ ) .
2 |1, 2


1 = where [rη2 ,η3 (k1 ≥4 ), rη2 ,η3 (k2 ≥4 )] no longer vanish indi-
0 vidually, and relations in the derivation algebra [10, 11,
Their τ -derivatives resulting from (5), (9) and integration 17] hold non-trivially.
by parts w.r.t. v2 take the universal form
1  EXAMPLES FOR INITIAL VALUES
2πi∂τ Zητ2 (∗|1, 2) = s12 ∂η22 −℘(η2 , τ )−2ζ2 Zητ2 (∗|1, 2) , This section is dedicated to the degeneration of A-cycle
2
(16) integrals (6) at the cusp τ → i∞ which enters the α0 -
expansion (14) as an initial value.
so one can read off the scalar differential operator in (8)
A. Generalities: The behaviour of A-cycle integrals
and the resulting representation of the derivations,
at the cusp is most conveniently studied in the variables
1 
Dητ2 (2|2) = s12 ∂η22 − ℘(η2 , τ ) − 2ζ2 , (17) dσj σi +σj
2 σj = e2πizj , dzj = , Gij = 2πi , (23)
1 2πiσj σi −σj
1 
rη2 (0 ) = s12 2 +2ζ2 − ∂η22 , rη2 (k≥4 ) = s12 η2k−2 .
η2 2 where the planar Green function and Kronecker–
Eisenstein series degenerate to (σji ≡ σj −σi )
Note that various combinations of iterated Eisenstein in-
tegrals drop out from the two-point instance of (14) since lim Ω(vij , η, τ ) = π cot(πη) + Gij (24)
τ →i∞
commutators [rη2 (k1 ), rη2 (k2 )] with k1 , k2 ≥ 4 vanish. 1 1
lim G(vij , τ ) = log(σi ) + log(σj ) − log(σji ) .
B. Three points give rise to A-cycle integrals τ →i∞ 2 2
Z Their non-planar analogues take an even simpler form,
Zητ2 ,η3 (∗|1, 2, 3) = dz2 dz3 Ω(z12 , η2 +η3 , τ ) (18) π
C(∗) lim Ω(vij + τ2 , η, τ ) = , lim G(vij + τ2 , τ ) = 0 .
τ →i∞ sin(πη) τ →i∞
× Ω(z23 , η3 , τ )es12 G(z12 ,τ )+s13 G(z13 ,τ )+s23 G(z23 ,τ ) (25)
 4

Since string-theory applications of (14) involve the coef- where

ficients w.r.t. ηj , we will need the expansions 1 h π  π 
I tree (1, a2 , a3 |1) = − 2 sin (s1a2 +s23 ) sin s1a3
∞ 2π 2 2
1 X
−2 ζ2k η 2k−1

π cot(πη) = (26) × Z tree (+, a2 , a3 , 1, −|+, 2, 3, −, 1)
η
k=1  i
π 1
∞
X 2k−1
2 −1 + Z tree (+, a2 , a3 , 1, −|+, 3, 2, −, 1) + (2 ↔ 3)
= + ζ2k η 2k−1 .
sin(πη) η 22k−2 1 h π  π 
k=1 I tree (1, a2 , a3 |G23 ) = sin (s1a2 +s23 ) cos s1a3
2π 2 2
As will be detailed in [12], the σj -integration in n-point

tree
× Z (+, a2 , a3 , 1, −|+, 2, 3, −, 1) (33)
Zη~i∞ lines up with explicitly known combinations of N =  i
(n+2)-point disk integrals [18] − Z tree (+, a2 , a3 , 1, −|+, 3, 2, −, 1) + (2 ↔ 3) .
Z
tree
Z (a1 , a2 , . . . , aN |1, 2, . . . , N ) = Their leading low-energy orders read [12]
−∞<σa1 <σa2 <...<σaN <∞
1 ζ2 2
QN −sij + (s12 +s213 +s223 )+O(α03 ) (34)
I tree (1, 2, 3|1) =
dσ1 dσ2 . . . dσN i<j |σij | 2 8
. (27) 1 ζ2
vol SL2 (R) σ12 σ23 . . . σN −1,N σN,1 tree
I (1, 2, 3|G23 ) = + (s12 +s13 +s23 )2 +O(α02 )
s23 4s23
The two extra punctures n+1 → + and n+2 → − are
associated with Mandelstam invariants and exemplify that integrals over k factors of Gij in (23)
n n may have up to k kinematic poles.
1 X X
sj+ = sj− = − sij , s+,− = sij . (28) Non-planar three-point initial values in turn boil down
2
1≤i6=j 1≤i<j to four-point disk integrals with α0 -expansions in (30),
The α0 -expansion of (27) and therefore Zη~i∞ involves mul-
π 2 cot(πη23 ) Γ(1 − s12 )
Zηi∞ 3


tiple zeta values (MZVs) which can be systematically 2 ,η3 1,2 |1, 2, 3 = 2 (35)
sin(πη3 ) Γ(1 − s12 )

generated from the all-multiplicity methods of [19, 20]. 2

B. Two points: Planar initial values at two points π2 Γ(1 − s12 )

Zηi∞ 3


2 ,η3 1,2 |1, 3, 2 =  .
sin(πη23 ) sin(πη3 ) Γ(1 − s12 ) 2

descend from four-point tree-level integrals, 2
 πs 
12
Zηi∞ (1, 2|1, 2) = π cot(πη2 ) 2i sin
2
2
Z 1
dσ2 s12 /2 CONCLUSIONS AND FURTHER DIRECTIONS
× σ2 (1 − σ2 )−s12 (29)
0 2πiσ 2 In this letter we presented a method to expand a gen-
Γ(1 − s12 ) erating series of genus-one integrals (6) relevant to one-
= π cot(πη2 )  2 .
Γ(1 − s212 ) loop open-string amplitudes. At each order in the inverse
string tension α0 , our main result (14) pinpoints the ac-
The factor of 2i sin( πs212 ) and similar trigonometric func- companying eMZVs in their minimal and canonical rep-
tions below stem from contour deformations detailed in resentation via iterated Eisenstein integrals.
[12]. The gamma functions with standard α0 -expansion Genus-zero integrals relevant to open-string tree am-
X ∞  plitudes obey Knizhnik–Zamolodchikov equations with
Γ(1 − s12 ) ζk 1−k k
2 = exp (1 − 2 )s12 (30) a characteristic linear factor of α0 on their right-hand
Γ(1 − s212 ) k

k=2 side [19]. This structure is analogous to the ε-form
1 1 19 4 2 of differential equations among Feynman integrals with
= 1+ s212 ζ2 + s312 ζ3 + s ζ +O(α05 )
4 4 160 12 2 dimensional-regularization parameter ε [5, 21], suggest-
do not appear in the non-planar counterpart of (29) ing a correspondence between α0 and ε. By the linear-
π ity of the differential operators Dη~τ in sij = −2α0 ki · kj ,
Zηi∞


2 |1, 2 =
1 . (31) the Knizhnik–Zamolodchikov–Bernard-type equation (8)
2
sin(πη2 )
also becomes linear in α0 . So our results generalize this
C. Three points: Degenerate A-cycle integrals at intriguing correspondence to genus one and provide the
three points introduce five-point disk integrals, string-theory analogue of the ε-form for differential equa-
tions of elliptic Feynman integrals [5].
Zηi∞
2 ,η3
(1, a2 , a3 |1, 2, 3) (32) The generating functions Zη~τ are expected to comprise
 s 13

any moduli-space integral in massless one-loop ampli-
= π 2 cot(πη23 ) cot(πη3 ) + I tree (1, a2 , a3 |1)
s123 tudes of open bosonic strings and superstrings upon ex-
 s23 
pansion in ηj . Accordingly, they are proposed to gener-
+ π cot(πη23 ) + cot(πη3 ) I tree (1, a2 , a3 |G23 ) ,
s12 alize the universal disk-integrals (27) that appear in the
 5

double-copy representation of string tree-level amplitudes A. Differential equation: The 6 × 6 differential op-
[18, 22]. Hence, the study of the genus-one integrals Zη~τ erator Dη~τ = Dητ2 ,η3 ,η4 in (8) is determined by
is an essential step towards universal double-copy struc- 4 4
tures in one-loop amplitudes of different string theories X s1j X sij
Dη~τ (2, 3, 4|2, 3, 4) = ∂η2j + (∂ηi −∂ηj )2
that generalize those of the superstring [23]. 2 2
j=2 2≤i<j
The generating functions Zη~τ can be adapted to a
closed-string context, encoding the integrals over torus − s12 ℘(η234 , τ ) − (s13 +s23 )℘(η34 , τ )
punctures in one-loop amplitudes of type-II, heterotic − (s14 +s24 +s34 )℘(η4 , τ ) − 2ζ2 s1234 (36)
and closed bosonic string theories. Closed-string ana- τ

Dη~ (2, 3, 4|2, 4, 3) = (s14 +s24 ) ℘(η34 , τ ) − ℘(η4 , τ )

logues of Zη~τ will be shown [24] to obey similar differen-
Dη~τ (2, 3, 4|3, 2, 4) = s13 ℘(η234 , τ ) − ℘(η34 , τ )
 
tial equations and to shed new light on the properties of
Dη~τ (2, 3, 4|3, 4, 2) = s13 ℘(η234 , τ ) − ℘(η34 , τ )
 
modular graph forms [25] including their relation with
open-string amplitudes [26]. Dη~τ (2, 3, 4|4, 2, 3) = s14 ℘(η34 , τ ) − ℘(η4 , τ )
 
Moreover, the method of this work to infer moduli-
Dη~τ (2, 3, 4|4, 3, 2) = s14 ℘(η34 , τ ) − ℘(η234 , τ )
 
space integrals from differential equations should be ap-
plicable at higher loops. In the same way as disk integrals with ηij...p = ηi +ηj + . . . +ηp . The corresponding matrix
were used as the initial value for our one-loop results, representations of the derivations (k 6= 2)
higher-genus integrals in string amplitudes are expected 4 4
X X
to obey differential equations w.r.t. complex-structure k−2
rη~ (k ) = η234 rη~ (e234 ) + k−2
ηij rη~ (eij ) + ηjk−2 rη~ (ej )
moduli such that their separating and non-separating de- 2≤i<j j=2
generations set the initial conditions. It would be inter-  4
X sij X 4
s1j 2 
esting to explore a differential-equation approach of this + δk,0 2ζ2 s1234 − (∂ηi −∂ηj )2 − ∂ηj 16×6 (37)
2 j=2
2
type to the higher-genus modular graph functions of [27]. 2≤i<j

In summary, our new approach to one-loop open- can be assembled from (S123,4 ≡ s14 +s24 +s34 )
string amplitudes via differential equations connects with
s12 0 −s13 −s13 0 s14
 
state-of-the-art techniques in particle phenomenology  0 s12 0 s13 −s14 −s14 
and provides explicit matrix representations of profound 
−s12 −s12 s13 0 s14 0

rη~ (e234 ) =  (38)
 
number-theoretic structures. As will be elaborated in  0 s12 0 s13 −s14 −s14


[12], our results manifest important formal properties  −s12 −s12 s13 0 s14 0 
of string amplitudes such as uniform transcendentality, s12 0 −s13 −s13 0 s14
coaction formulae and the dropout of twisted eMZVs 
s13 +s23 −s14 −s24 s13 s13 −s14 −s14

from non-planar open-string amplitudes.  −s13 −s23 s14 +s24 −s13 −s13 s14 s14 
0 0 0 0 0 0
 
Acknowledgements: We are grateful to Johannes Broedel, rη~ (e34 ) = 
 
0 0 0 0 0 0

 
Jan Gerken, Axel Kleinschmidt, Nils Matthes and Fed-  0 0 0 0 0 0 
erico Zerbini for inspiring discussions and collaboration 0 0 0 0 0 0
on related topics. Moreover, Claude Duhr, Hermann 
S123,4 s14 +s24 0 0 s14 0

Nicolai, Albert Schwarz and in particular Sebastian Miz-  0 0 0 0 0 0 
era are thanked for valuable discussions, and we are 
0 0 S123,4 s14 +s34 0 s14 

rη~ (e4 ) = 

grateful to Sebastian Mizera for helpful comments on 0 0 0 0 0 0 


a draft. We would like to thank the organizers of the
 0 0 0 0 0 0 
programme “Modular forms, periods and scattering am- 0 0 0 0 0 0
plitudes” at the ETH Institute for Theoretical Studies in and relabellings.
Zürich for providing a stimulating atmosphere and finan- B. Initial values: The four-point integrals in massless
cial support. CRM is supported by a University Research one-loop string amplitudes descend from orders of Zη~τ
Fellowship from the Royal Society. OS is grateful to the with odd homogeneity degree in ηj . Since the derivations
organizers of the workshop “Automorphic Structures in (37) do not mix odd and even functions of ηj , we only
String Theory” at the Simons Center in Stony Brook and spell out the odd part of the planar initial value
those of the workshop “String Theory from a Worldsheet odd
Zη~i∞ (∗|1, 2, 3, 4) planar ∗ = π cot(πη34 )I tree (∗|G12 G34 )
Perspective” at the GGI in Florence for setting up inspir-
ing meetings. OS is supported by the European Research π cot(πη4 )  
+ s34 I tree (∗|G12 G34 ) − s14 I tree (∗|G14 G23 )
Council under ERC-STG-804286 UNISCAMP. s123
π cot(πη234 )  
+ s12 I tree (∗|G12 G34 ) − s14 I tree (∗|G14 G23 )
s
APPENDIX: FOUR-POINT EXAMPLES  234 s13 s24  tree
+ 6ζ2 π cot(πη4 ) + π cot(πη234 ) I (∗|1)
This appendix provides further details on the expansion s123 s234
(14) of four-point A-cycle integrals (6). + π 3 cot(πη234 ) cot(πη34 ) cot(πη4 )I tree (∗|1) . (39)
 6

Similar to (34), I tree denote combinations of six-point [7] J. Broedel, N. Matthes, G. Richter and O. Schlotterer,
disk integrals (27) which no longer depend on ηj , see J. Phys. A51 (2018) 285401, [1704.03449].
section 5.5 of [12] for further details. [8] J. Broedel and O. Schlotterer, in Proceedings, KMPB
Conference: Elliptic Integrals, Elliptic Functions and
Non-planar four-point initial values reduce to four- and
Modular Forms in Quantum Field Theory, (2019) 133.
five-point disk integrals, e.g. [9] D. Calaque, B. Enriquez and P. Etingof, in Algebra, arith-
metic, and geometry: in honor of Yu. I. Manin. Vol. I,
3,4 π 3 cot(πη234 ) cot(πη4 ) vol. 269 of Progr. Math. (2009) 165. • B. Enriquez, Se-
Zη~i∞


1,2 |1, 2, 3, 4 =
lecta Math. (N.S.) 20 (2014) 491.
sin(πη34 )
Γ(1 − s12 )Γ(1 − s34 ) [10] H. Tsunogai, Publ. Res. Inst. Math. Sci. 31 (1995) 113.
× 2 (40) [11] J. Broedel, N. Matthes and O. Schlotterer, J. Phys. A49
Γ(1 − s212 ) Γ(1 − s234 ) (2016) 155203, [1507.02254].
π [12] C. Mafra and O. Schlotterer, [1908.xxxxx].
Zη~i∞ 2,3,4
Zηi∞


1 |1, 2, 3, 4 = sin(πη 2 ,η3
(2, 3, 4|2, 3, 4) , [13] J. Polchinski, Cambridge University Press (2007).
234 )
[14] L. Kronecker, Mathematische Werke IV (1881) 313.
[15] M. B. Green and J. H. Schwarz, Phys. Lett. 151B (1985)
see (32) for Zηi∞
2 ,η3
(2, 3, 4|2, 3, 4). By extracting the order 21.
−1 −1 −1
of η234 η34 η4 from (14), we have checked (36), (39) and [16] F. Brown, 1407.5167. • R. Hain, 1309.0580.
(40) to reproduce the α0 -expansions of [6, 7] to the or- [17] A. Pollack, Undergrad. thesis, Duke University (2009).
ders of α02 and α03 in the planar and non-planar sectors, [18] J. Broedel, O. Schlotterer and S. Stieberger, Fortsch.
respectively. Phys. 61 (2013) 812, [1304.7267].
[19] J. Broedel, O. Schlotterer, S. Stieberger and T. Tera-
soma, Phys. Rev. D89 (2014) 066014, [1304.7304].
[20] C. R. Mafra and O. Schlotterer, JHEP 01 (2017) 031,
[1609.07078].
[1] A. Levin, Compositio Mathematica 106 (1997) 267. [21] J. M. Henn, Phys. Rev. Lett. 110 (2013) 251601,
[2] F. Brown and A. Levin, 1110.6917. [1304.1806].
[3] B. Enriquez, Bull. Soc. Math. France 144 (2016) 395, [22] C. R. Mafra, O. Schlotterer and S. Stieberger, Nucl.
[1301.3042]. Phys. B873 (2013) 419, [1106.2645]. • T. Azevedo,
[4] S. Bloch and P. Vanhove, J. Number Theory 148 (2015) M. Chiodaroli, H. Johansson and O. Schlotterer, JHEP
328, [1309.5865]. • S. Bloch, M. Kerr and P. Vanhove, 10 (2018) 012, [1803.05452].
Compos. Math. 151 (2015) no.12, 2329, [1406.2664]. • [23] C. R. Mafra and O. Schlotterer, Phys. Rev. Lett.
L. Adams and S. Weinzierl, Commun. Num. Theor. Phys. 121 (2018), 011601, [1711.09104]. • C. R. Mafra
12 (2018) 193, [1704.08895]. • J. Ablinger, J. Blümlein, and O. Schlotterer, [1812.10969], [1812.10970],
A. De Freitas, M. van Hoeij, E. Imamoglu, C. G. Raab, [1812.10971].
C. S. Radu and C. Schneider, J. Math. Phys. 59 (2018), [24] J. E. Gerken, A. Kleinschmidt and O. Schlotterer, work
062305 [1706.01299]. • E. Remiddi and L. Tancredi, in progress.
Nucl. Phys. B925 (2017) 212, [1709.03622]. • J. L. [25] M. B. Green, J. G. Russo and P. Vanhove, JHEP 02
Bourjaily, A. J. McLeod, M. Spradlin, M. von Hippel (2008) 020, [0801.0322]. • E. D’Hoker, M. B. Green,
and M. Wilhelm, Phys. Rev. Lett. 120 (2018) 121603, Ö. Gürdogan and P. Vanhove, Commun. Num. Theor.
[1712.02785]. • J. Broedel, C. Duhr, F. Dulat and Phys. 11 (2017) 165, [1512.06779]. • E. D’Hoker
L. Tancredi, JHEP 05 (2018) 093, [1712.07089]. • and M. B. Green, J. Number Theory 189 (2018) 25,
J. Broedel, C. Duhr, F. Dulat and L. Tancredi, Phys. [1603.00839].
Rev. D97 (2018) 116009, [1712.07095]. • J. Broedel, [26] F. Zerbini, Commun. Num. Theor. Phys. 10 (2016)
C. Duhr, F. Dulat, B. Penante and L. Tancredi, JHEP 703, [1512.05689]. • J. Broedel, O. Schlotterer and
08 (2018) 014, [1803.10256]. • L. Adams, E. Chaubey F. Zerbini, JHEP 01 (2019) 155, [1803.00527]. • J. E.
and S. Weinzierl, Phys. Rev. Lett. 121 (2018) 142001, Gerken, A. Kleinschmidt and O. Schlotterer, JHEP 01
[1804.11144]. • L. Adams, E. Chaubey and S. Weinzierl, (2019) 052, [1811.02548]. • D. Zagier and F. Zerbini,
JHEP 1810 (2018) 206, [1806.04981]. • J. Broedel, 1906.12339.
C. Duhr, F. Dulat, B. Penante and L. Tancredi, in KMPB [27] E. D’Hoker, M. B. Green and B. Pioline, Commun. Math.
Conference: Elliptic Integrals, Elliptic Functions and Phys. 366 (2019) no.3, 927, [1712.06135]. • E. D’Hoker,
Modular Forms in Quantum Field Theory (2019) 107, M. B. Green and B. Pioline, Commun. Num. Theor.
[1807.00842]. • J. Blümlein, A. De Freitas, M. Van Hoeij, Phys. 13 (2019) 351, [1806.02691].
E. Imamoglu, P. Marquard and C. Schneider, PoS
LL2018 (2018) 017, [1807.05287]. • J. Broedel, C. Duhr,
F. Dulat, B. Penante and L. Tancredi, JHEP 1905
(2019) 120, [1902.09971]. • C. Bogner, S. Müller-Stach
and S. Weinzierl, [1907.01251]. • J. Broedel, C. Duhr,
F. Dulat, R. Marzucca, B. Penante and L. Tancredi,
[1907.03787].
[5] L. Adams and S. Weinzierl, Phys. Lett. B781 (2018) 270,
[1802.05020].
[6] J. Broedel, C. R. Mafra, N. Matthes and O. Schlotterer,
JHEP 07 (2015) 112, [1412.5535].