You are on page 1of 6

MPP-2023-266

Coulomb Branch Amplitudes from a Deformed Amplituhedron Geometry

Nima Arkani-Hamed
School of Natural Sciences, Institute for Advanced Study, Princeton, NJ 08540, USA

Wojciech Flieger, Johannes M. Henn, and Anders Schreiber


Max-Planck-Institut für Physik, Werner-Heisenberg-Institut, D-80805 München, Germany

Jaroslav Trnka
Center for Quantum Mathematics and Physics (QMAP),
Department of Physics, University of California, Davis, CA 95616, USA and
Institute for Particle and Nuclear Physics, Charles University in Prague, Czech Republic

The Amplituhedron provides, via geometric means, the all-loop integrand of scattering amplitudes
arXiv:2311.10814v1 [hep-th] 17 Nov 2023

in maximally supersymmetric Yang-Mills theory. Unfortunately, dimensional regularization, used


conventionally for integration, breaks the beautiful geometric picture. This motivates us to propose
a ‘deformed’ Amplituhedron. Focusing on the four-particle amplitude, we introduce two deformation
parameters, which can be interpreted as particle masses. We provide evidence that the mass pattern
corresponds to a specific choice of vacuum expectation values on the Coulomb branch. The deformed
amplitude is infrared finite, making the answer well-defined in four dimensions. Leveraging four-
dimensional integration techniques based on differential equations, we compute the amplitude up to
two loops. In the limit where the deformation parameters are taken to zero, we recover the known
Bern-Dixon-Smirnov amplitude. In the limit where only one deformation parameter is taken to zero,
we find a connection to the angle-dependent cusp anomalous dimension.

INTRODUCTION geometric origin of the integrand will be important


for determining the final answer after integrating over
Yang-Mills theory has been the foundation of the Stan- Minkowski space. Unfortunately, the fact that massless
dard Model of Particle Physics for over five decades, de- particles have infrared divergences requires the use of a
scribing the fundamental interactions of elementary par- regulator. The standard method, dimensional regular-
ticles. Despite its importance, the theory remains in- ization, breaks the beautiful geometric picture. This mo-
completely understood, and making precision predictions tivates us to propose a ‘deformed’ Amplituhedron. Fo-
for comparison with experiment is notoriously difficult. cusing on the four-particle amplitude, in this paper we
However, the study of scattering amplitudes has revealed introduce certain shifts in the kinematics. The latter
many surprising properties, hidden symmetries and new are somewhat reminiscent of, but different from, BCFW
conceptual approaches that hint at a different formula- shifts [21]. As a result, we obtain a deformed amplitude
tion of Yang-Mills theory, where Lagrangians and Feyn- that depends on two deformation parameters and that is
man diagrams do not play a central role. well-defined in four dimensions. In a nutshell, the input
is the four-particle scattering kinematics; after working
In the ’t Hooft limit of a large number of colors, only
out the form and performing the integrations, the output
planar Feynman diagrams contribute. The latter may be
is a finite amplitude, at a given loop order. Our formula
seen as approximating a two-dimensional surface, hinting
exactly matches certain Coulomb branch amplitudes.
at a geometric formulation of the theory. The maximally
supersymmetric Yang-Mills theory (sYM) is an ideal test-
ing ground for such novel ideas. In the AdS/CFT corre-
FOUR-POINT DEFORMED AMPLITUHEDRON
spondence, this picture is made more concrete: the scat-
tering amplitudes at strong coupling are described by the
volume of certain minimal surfaces [1]. The definition of the four-point Amplituhedron ge-
How does such a geometric picture emerge from pertur- ometry consists of two components: four external mo-
bation theory? The Amplituhedron [2, 3] provides a novel mentum twistors Z1 , Z2 , Z3 , Z4 and L lines {(AB)i }.
geometric framework for defining the all-loop integrand The former satisfy the positivity condition ⟨1234⟩ ≡
of the planar sYM amplitudes. The latter is obtained det(Z1 Z2 Z3 Z4 )>0, and can be considered as fixed points
as the canonical differential form on a certain geometric in the momentum twistor space. The Amplituhedron is
space, defined by a set of inequalities that involve the then a configuration of lines {(AB)i } satisfying the fol-
particle kinematics. Important work by mathematicians lowing inequalities:
and physicists has been devoted to understanding, in de-
tail, the geometry of the space [4–9], [10–20]. ⟨(AB)i 12⟩>0, ⟨(AB)i 23⟩>0, ⟨(AB)i 34⟩>0,
(1)
We expect that this improved understanding of the ⟨(AB)i 14⟩>0 , ⟨(AB)i 13⟩<0, ⟨(AB)i 24⟩<0 ,
2

collinear configurations are removed as boundaries. As


a consequence, the number of boundaries of the high-
est codimension (leading singularities) is only two rather
than six: only the configurations with (AB)i = 13 and
(AB)i = 24 remain, while all other soft-collinear con-
figurations (AB)i = k k+1 are not present. At one-
loop, the undeformed Amplituhedron is a positive Grass-
mannian G+ (2, 4) with a number of boundaries of var-
ious co-dimensions, (4, 10, 12, 6), and Euler characteris-
tic χ = +1. For example, the codimension-two bound-
aries correspond to four double cuts ⟨AB12⟩=⟨AB23⟩=0
(and cyclic), each having two solutions – either AB pass-
ing Z2 or being in the plane (Z1 Z2 Z3 ), and two double
cuts ⟨AB12⟩=⟨AB34⟩=0 and ⟨AB23⟩=⟨AB14⟩=0, total-
ing 10 codimension-two boundaries. In the deformed case
any double cut, ⟨AB X̂i ⟩ = ⟨AB X̂j ⟩ = 0, has only one so-
FIG. 1. Illustration of the undeformed Amplituhedron ge- lution, hence we have only six codimension-two bound-
ometry (first quadrant, in light blue), for η = η̃ = 0, and of aries. The number of boundaries of various co-dimensions
the deformed geometry (wedge, in red) for η = 14 , η̃ = 21 . is now (4, 6, 4, 2), which demonstrates the simplicity of
the geometry. The Euler characteristic is again χ = +1.
with the last two inequalities following from the sign flip
condition [3]. Additional positivity conditions are im-
posed between each pair of lines (AB)i and (AB)j ,
TWO-LOOP INTEGRAND
⟨(AB)i (AB)j ⟩ > 0 . (2)

The external momentum twistors form four lines X1 ≡ At two-loops, the geometric space is more involved and
Z1 Z2 , X2 ≡ Z2 Z3 , X3 ≡ Z3 Z4 , X4 ≡ Z1 Z4 , which corre- the direct triangulation, though possible, is more diffi-
spond to massless propagators, by virtue of ⟨Xi Xi ⟩ = 0. cult. Therefore here we determine the form by making
Adjacent pairs of lines intersect, the following ansatz,
⟨X1 X2 ⟩ = ⟨X2 X3 ⟩ = ⟨X3 X4 ⟩ = ⟨X4 X1 ⟩ = 0 , (3)
dµAB dµCD · N (AB, CD, Zi , ρ, ρ̃, η, η̃)
To obtain the deformed Amplituhedron, one shifts the Ω(2) =  ,
⟨AB X̂1 ⟩⟨AB X̂2 ⟩⟨AB X̂3 ⟩⟨AB X̂4 ⟩⟨ABCD⟩
external kinematics Xi → X bi . This is done while pre-
⟨CDX̂1 ⟩⟨CDX̂2 ⟩⟨CDX̂3 ⟩⟨CDX̂4 ⟩
serving (3) but allowing ⟨X bi ⟩ =
bi X ̸ 0. We define (9)
b1 = X1 − ρX3 , b3 = X3 − ρ̃X1 , which is based on the known denominator structure and
X X (4)
impose consistency conditions on the numerator. One
b2 = X2 − ηX4 ,
X b4 = X4 − η̃X2 .
X (5) of these conditions is the absence of “non-planar” cuts,
which violate the ⟨ABCD⟩ > 0 condition. For example,
We now replace Xi → X bi in the first four inequalities we can cut following propagators,
of (1). Other inequalities remain unchanged. Using the
following parametrization of the (AB)i lines,
⟨AB X̂1 ⟩ = ⟨AB X̂2 ⟩ = ⟨CDX̂1 ⟩ = ⟨CDX̂2 ⟩ = 0 , (10)
ZAi = Z1 + xi Z2 − yi Z4 , ZBi = Z3 + zi Z2 + wi Z4 , (6)

the original conditions (1), imply xi , yi , zi , wi > 0, while which fixes y1 = ρz1 , y2 = ρz2 , w1 = ηx1 , w2 = ηx2 .
instead in the shifted case we have (cf. Fig. 1), Then evaluating ⟨ABCD⟩ in this parametrization gives

yi > ρzi > 0, zi > ρ̃yi > 0 , (7)


⟨ABCD⟩ = −ρ(z1 − z2 )2 − η(x1 − x2 )2 . (11)
wi > ηxi > 0, xi > η̃wi > 0 , (8)

where we further impose conditions on shift parameters This expression is manifestly negative, which is in tension
ρ, ρ̃, η, η̃ > 0, such that ρρ̃ < 1 and η η̃ < 1. with the ⟨ABCD⟩ > 0 condition. Hence this cut is un-
The boundary structure of the deformed Amplituhe- physical and the numerator must vanish when evaluated
dron happens to be much simpler than of the origi- on it. Imposing all such conditions the ansatz reduces
nal space. Thanks to the deformation parameters, all to the sum of deformed double-box integrals (in analogy
3

with the non-deformed case), m3


p2 p3
2
(a) dµAB dµCD · ⟨X̂1 X̂3 ⟩ ⟨X̂2 X̂4 ⟩
Ωdb = −  , (12)
⟨AB X̂1 ⟩⟨AB X̂2 ⟩⟨AB X̂3 ⟩⟨ABCD⟩ m2 m4
⟨CDX̂3 ⟩⟨CDX̂4 ⟩⟨CDX̂1 ⟩
dµAB dµCD · ⟨X̂1 X̂3 ⟩⟨X̂2 X̂4 ⟩2 p1 p4
(b) m1
Ωdb = −  . (13)
⟨AB X̂2 ⟩⟨AB X̂3 ⟩⟨AB X̂4 ⟩⟨ABCD⟩
⟨CDX̂4 ⟩⟨CDX̂1 ⟩⟨CDX̂2 ⟩ FIG. 2. One-loop box integral in the deformed Ampli-
tuhedron kinematics. The on-shell conditions read p2i =
and Ω(2) is a symmetrized sum in AB ↔ CD, m2i + m2i+1 , with m5 ≡ m1 .
(a) (b)
Ω(2) = n(2a) Ωdb + n(2b) Ωdb + (AB ↔ CD) (14)
Unlike in the massless case there is no cancellation of spu- COMPARISON TO COULOMB-BRANCH
(a) (b)
rious cuts between Ωdb and Ωdb , and their coefficients AMPLITUDES
(2a) (2b)
n and n are unrelated.
Interestingly, our deformation leads to massive parti-
cles, without the need of introducing an infinity twistor.
AMPLITUDES FROM DEFORMED
AMPLITUHEDRON Translating the kinematics of the deformed Amplituhe-
dron from momentum twistor space to ordinary momen-
tum space, we find that the amplitude corresponds to
The Amplituhedron defines the planar L-loop inte-
scattering of particles, as shown in Figure 2. The masses
grand Ω(L) at all orders in perturbation theory. More
are simple functions of the deformation parameters, and
concretely, working with the amplitude, normalized by
vanish when the latter are sent to zero. The dual confor-
its tree-level contribution, we have
mal invariants are related to the kinematics as follows,
M = 1 + g 2 M (1) + g 4 M (2) + O(g 6 ) , (15)
with g = 2 2
gYM Nc /(16π 2 ), and (−s + m21 + m23 )2 (−t + m22 + m24 )2
u= , v= , (20)
Z 4m21 m23 4m22 m24
M (L) = Ω(L) . (16)
where s = (p1 + p2 )2 , t = (p2 + p3 )2 . Thanks to the
(L)
The L-loop integrand Ω is the form with logarithmic masses, the amplitude is infrared finite. This finiteness
singularities on the boundaries of the deformed Ampli- property is built into our geometric setup: it can be seen
tuhedron space. as a consequence of the absence of collinear and soft-
The one-loop form associated to the deformed Ampli- collinear regions in the boundary structure of the de-
tuhedron is found to be formed Amplituhedron.
dµAB ⟨X̂1 X̂3 ⟩⟨X̂2 X̂4 ⟩ Our setup is reminiscent of the AdS-inspired Coulomb
Ω(1) = −n(1) , (17) branch amplitudes [23, 24], where one considers a Higgs
⟨AB X̂1 ⟩⟨AB X̂2 ⟩⟨AB X̂3 ⟩⟨AB X̂4 ⟩
mechanism U (4 + N ) → U (4) × U (N ). This is done by
where dµAB = ⟨ABd2 A⟩⟨ABd2 B⟩ is measure in momen- choosing vacuum expectation values in the U (4) part as
tum twistor space, normalized such that [22] follows,
Z
dµAB 1 1
=  . (18) 1
⟨ABQ⟩ 4 Γ(4) 1 ⟨QQ⟩ 2 ⟨Φ⟩ = diag(m1⃗n1 , m2⃗n2 , m3⃗n3 , m4⃗n4 ) , (21)
2 g
The normalization n(1) is not determined by the Ampli-
tuhedron, and it is related to the value of the leading where ⃗ni are SO(6) vectors. In [23], all ⃗ni were chosen
singularity – in the undeformed case the latter is equal to be equal, which lead to a certain mass configuration.
to 1. Note that in general, n(1) is a function of the de- Instead, we find that we can match the mass configu-
formation parameters ρ, ρ̃, η, η̃. ration above by choosing ⃗n1 = −⃗n3 , ⃗n2 = −⃗n4 , with
Thanks to dual conformal symmetry, the integrated ⃗n1 · ⃗n2 = 0.[25] We conjecture that the deformed Ampli-
amplitude M depends on two invariants only, which are tuhedron describes these Coulomb branch amplitudes. If
this conjecture is true, then we can use unitarity of the
⟨X̂1 X̂3 ⟩2 ⟨X̂2 X̂4 ⟩2 Coulomb branch amplitudes to determine the relative co-
u= , v= , (19) efficients of the positive geometries. Indeed, performing
X̂12 X̂32 X̂22 X̂42 a one-loop computation as in Appendix B of [23], we find
which in terms of deformation parameters are given by that n(1) = 1, and likewise one may fix n(2a) = n(2b) = 1
(1 + ρρ̃)2 /(4ρρ̃) and (1 + η η̃)2 /(4η η̃), respectively. from generalized unitarity.
4

INTEGRATED ONE- AND TWO-LOOP and


AMPLITUDES 1
J3 (z) = log3 (z) + log (z) Li2 (1 − z)
4
For carrying out the integration, we find it convenient
  (30)
1
to parametrize the kinematics as follows, − 2Li3 (1 − z) − 2Li3 1 − .
z
ρ = ρ̃ = −x , η = η̃ = −y . (22) Q and J3 are real-valued for positive arguments. As in
the famous result for the six-gluon two-loop remainder
with 0 < x < 1, 0 < y < 1. This corresponds to a sub- function in sYM [28], only classical polylogarithms are
set of the Euclidean kinematic region, where the answer needed. In contrast, the two-loop four-particle amplitude
is real-valued [26]. Note that this is different from the in the (slightly different) Coulomb-branch setup of [23]
Amplituhedron region. The cross-ratios of eq. (19) then is more complicated [27] .
take the form We also note that the amplitude up to two loops has
1

1
2
1

1
2 the following ten-letter symbol alphabet,
u= x+ , v= y+ . (23)
4 x 4 y A ={x, 1 ± x, y, 1 ± y, x ± y, 1 ± xy} . (31)
Starting from eqs. (16) and (17), one may obtain a pa- It is interesting to study the behavior of M near singular
rameter integral representation for M (1) by introducing kinematic configurations, i.e. where one or more terms
Feynman parameters, and then using the basic identity in eq. (31) vanish. We now present exact formulas for
(18), with the result the leading behavior in two such limits.

(1 − x)(1 − y) ∞ d4 α
Z
M (1) (x, y) = − × (24) EXACT RESULTS IN KINEMATIC LIMITS
(1 + x)(1 + y) 0 GL(1)
(1 − x2 ) (1 − y 2 )
. High-energy limit. This corresponds to s, t → ∞, or
[(α1 x + α3 )(α1 + α3 x) + (α2 y + α4 )(α2 + α4 y)]2
equivalently, x, y → 0, keeping x/y = t/s fixed. This
Carrying out the integration yields a surprisingly simple limit where the deformation parameters are taken to zero
answer, allows us to connect to the massless, infrared-divergent
amplitude. We conjecture that the following exact for-
(1 − x)(1 − y)
M (1) (x, y) = −2 log(x) log(y) . (25) mula holds in the high-energy limit,
(1 + x)(1 + y)
x,y→0 1
At two loops, we have log M = − Γcusp (g) log x log y
2 (32)
M (2) (x, y) = Idb (x, y) + Idb (y, x) , (26) + Γcollinear (g) (log x + log y) + C(g) ,
which is analogous to formulas in dimensional regular-
where ization [29, 30] and on the Coulomb branch [23]. Here
Γcusp = 4g 2 − 8ζ2 g 4 + O(g 6 ) is the light-like cusp anoma-
Z Z
(a) (b)
Idb (x, y) = Ωdb , Idb (y, x) = Ωdb . (27)
lous dimension, which is famously known from integrabil-
ity [31]. Comparing to our perturbative results, we find
Leveraging four-dimensional differential equation tech- that (32) holds with the collinear anomalous dimension
niques [27] we find that, remarkably, the answer for the Γcollinear (g) = −4ζ3 g 4 + O(g 6 ) and C(g) = −3/10π 4 g 4 +
double box can be written in just a few lines, O(g 6 ). The latter two are both regularization-scheme
(1 − x)2 (1 − y) h dependent [32].
Idb (x, y) = J3 (x2 ) log(y 2 ) Regge limit t → ∞. This corresponds to y → 0, keep-
(1 + x)2 (1 + y)
 2 ing x fixed. Thanks to dual conformal symmetry, this
2 1 y 1 2 2
i
− Q(y ) + Q + Q(x y ) , (28) may be equivalently viewed as the small mass m1 → 0
2 x2 2 limit. Therefore we expect this limit to be governed by
with eikonal physics [33], and in particular by the anomalous
dimension of a cusped Wilson loop, which to two loops
3
Q(z) = 3Li4 (z) − 3 log(z)Li3 (z) + log2 (z)Li2 (z) is given by [34]
2
1 3π 4 π2 Γcusp (ϕ, θ; g) =g 2 ξ(−2 log x)+ (33)
+ log3 (z) log(1 − z) + + log2 (z) ( "
2 10 4 4 2
+ g ξ log x π + log x + ξ 4Li3 x2
4 2 2
 
3 √ 
+ log4 (z) + log2 (z) Li2 (1 − z) + 4π 2 Li2 − z 3
16   #)
1 4 2
2
log(x) − log3 (x) − π 2 log(x) − 4ζ3

− log (z) Li3 1 − − log (z) Li3 (1 − z) , (29) − 4Li2 x .
z 3 3
5

Here the Euclidean cusp angle ϕ is related to x via x = Our novel starting point also gives completely new
eiϕ , and the second parameter θ parametrizes the Wilson ways of studying physical limits, and to potentially ob-
loop’s coupling to scalars, via the combination tain exact results. Thanks to the simpler geometric con-
figurations in the limit, obtaining the canonical form
cos θ − cos ϕ 1 + x2 − 2x cos θ and integrating it should be easier. One particularly
ξ= = . (34)
i sin ϕ 1 − x2 interesting example is the ϕ → 0 limit of the angle-
dependent cusp anomalous dimension, whose first term
The choice of SO(6) vectors we made below eq. (21) Γcusp (ϕ, θ = 0; g) = −ϕ2 B(g) + O(ϕ4 ) is the exactly-
suggests that here θ = 0, which implies ξ = (1 − x)/(1 + known Bremsstrahlung function [36].
x). Indeed, we find that the leading terms in the Regge
limit are given by
ACKNOWLEDGMENTS
iϕ y→0 Γcusp (ϕ;0;g) 0
M (x = e , y) = r(ϕ, 0; g) y + O(y ) , (35)
We thank Simon Caron-Huot, Diego Correa and
where r(ϕ, 0; g) being the finite part in the Regge limit. Martı́n Lagares for discussions. This research received
It is interesting to note that by making a more general funding from the European Research Council (ERC) un-
choice of SO(6) vectors in eq. (21), such that ⃗n1 · ⃗n3 = der the European Union’s Horizon 2020 research and in-
− cos θ, we can match the full ξ dependence. novation programme (grant agreement No 725110), Novel
structures in scattering amplitudes, GAČR 21-26574S,
DOE grants No. SC0009999 and No. SC0009988, and
SUMMARY AND OUTLOOK the funds of the University of California.

We generalized the four-particle Amplituhedron geom-


etry of planar sYM such that the amplitude M (x, y) is
infrared finite and depends on two dual conformal param-
[1] L. F. Alday and J. M. Maldacena, JHEP 06, 064 (2007),
eters x, y. The finiteness is due to massive propagators. arXiv:0705.0303 [hep-th].
The mass configuration is different from the Coulomb [2] N. Arkani-Hamed and J. Trnka, JHEP 10, 030 (2014),
branch setup of [23], but we found that a slightly differ- arXiv:1312.2007 [hep-th].
ent pattern of vacuum expectation values matches our [3] N. Arkani-Hamed, H. Thomas, and J. Trnka, JHEP 01,
geometric setup. It remains to be proven that the de- 016 (2018), arXiv:1704.05069 [hep-th].
formed Amplituhedron not only matches the kinematics, [4] S. N. Karp and L. K. Williams, Int. Math. Res. Not. 5,
but gives exactly those Coulomb branch amplitudes. 1401 (2019), arXiv:1608.08288 [math.CO].
[5] S. N. Karp, L. K. Williams, and Y. X. Zhang, Ann.
In the original Amplituhedron, fixing the integrand in- Inst. H. Poincare D Comb. Phys. Interact. 7, 303 (2020),
volved the assumption of normalizing the canonical forms arXiv:1708.09525 [math.CO].
to have unit (constant) leading singularities on certain [6] T. Lukowski, M. Parisi, and L. K. Williams, (2020),
boundaries. Instead, the Coulomb branch amplitudes in arXiv:2002.06164 [math.CO].
general have kinematic-dependent leading singularities. [7] M. Parisi, M. Sherman-Bennett, and L. Williams,
This means that a generalization is needed of how we (2021), arXiv:2104.08254 [math.CO].
[8] C. Even-Zohar, T. Lakrec, and R. J. Tessler, (2021),
think about canonical forms. In this Letter, we fixed
arXiv:2112.02703 [math-ph].
those normalizations by comparing against generalized [9] X. Blot and J.-R. Li, (2022), arXiv:2206.03435
cuts of Coulomb branch amplitudes. One may ask – are [math.CO].
there other ways of fixing the answer that do not refer to [10] N. Arkani-Hamed and J. Trnka, JHEP 12, 182 (2014),
field theory at all? arXiv:1312.7878 [hep-th].
Given the two-loop integrand obtained in this way, we [11] S. Franco, D. Galloni, A. Mariotti, and J. Trnka, JHEP
computed the analytic result for M (x, y) up to two loops, 03, 128 (2015), arXiv:1408.3410 [hep-th].
[12] N. Arkani-Hamed, A. Hodges, and J. Trnka, JHEP 08,
and conjectured exact formulas both in the high-energy
030 (2015), arXiv:1412.8478 [hep-th].
as well as the Regge limit. Given the relatively simple [13] N. Arkani-Hamed, C. Langer, A. Yelleshpur Srikant,
symbol alphabet that we uncovered, it may be possible and J. Trnka, Phys. Rev. Lett. 122, 051601 (2019),
to bootstrap higher-loop results, given sufficient physical arXiv:1810.08208 [hep-th].
input from limiting configurations and analytic proper- [14] E. Herrmann, C. Langer, J. Trnka, and M. Zheng, JHEP
ties (for a review see [35]). 01, 035 (2021), arXiv:2009.05607 [hep-th].
[15] G. Dian, P. Heslop, and A. Stewart, (2022),
We expect that the new setup will lead to substantial
arXiv:2207.12464 [hep-th].
progress in making the connection between geometry and [16] E. Herrmann and J. Trnka, J. Phys. A 55, 443008 (2022),
the integrated functions, benefiting from recent work in arXiv:2203.13018 [hep-th].
mathematics [8]. This could lead to a geometry-based [17] N. Arkani-Hamed, J. Henn, and J. Trnka, JHEP 03, 108
differential equations method, with broader applications. (2022), arXiv:2112.06956 [hep-th].
6

[18] D. Damgaard, L. Ferro, T. Lukowski, and M. Parisi, arXiv:1006.5703 [hep-th].


JHEP 08, 042 (2019), arXiv:1905.04216 [hep-th]. [29] Z. Bern, L. J. Dixon, and V. A. Smirnov, Phys. Rev. D
[19] L. Ferro and T. Lukowski, JHEP 05, 183 (2023), 72, 085001 (2005), arXiv:hep-th/0505205.
arXiv:2210.01127 [hep-th]. [30] J. M. Drummond, J. Henn, G. P. Korchemsky,
[20] S. He, Y.-t. Huang, and C.-K. Kuo, (2023), and E. Sokatchev, Nucl. Phys. B 826, 337 (2010),
arXiv:2306.00951 [hep-th]. arXiv:0712.1223 [hep-th].
[21] R. Britto, F. Cachazo, B. Feng, and E. Witten, Phys. [31] N. Beisert, B. Eden, and M. Staudacher, J. Stat. Mech.
Rev. Lett. 94, 181602 (2005), arXiv:hep-th/0501052. 0701, P01021 (2007), arXiv:hep-th/0610251.
[22] N. Arkani-Hamed, J. L. Bourjaily, F. Cachazo, and [32] J. M. Henn, S. Moch, and S. G. Naculich, JHEP 12, 024
J. Trnka, JHEP 06, 125 (2012), arXiv:1012.6032 [hep- (2011), arXiv:1109.5057 [hep-th].
th]. [33] J. M. Henn, S. G. Naculich, H. J. Schnitzer, and
[23] L. F. Alday, J. M. Henn, J. Plefka, and T. Schuster, M. Spradlin, JHEP 04, 038 (2010), arXiv:1001.1358 [hep-
JHEP 01, 077 (2010), arXiv:0908.0684 [hep-th]. th].
[24] D. Correa, J. Henn, J. Maldacena, and A. Sever, JHEP [34] N. Drukker and V. Forini, JHEP 06, 131 (2011),
05, 098 (2012), arXiv:1203.1019 [hep-th]. arXiv:1105.5144 [hep-th].
[25] We note that it is also possible to switch on further an- [35] S. Caron-Huot, L. J. Dixon, J. M. Drummond, F. Dulat,
gles, such that ⃗n1 · ⃗n3 = − cos θ13 and ⃗n2 · ⃗n4 = − cos θ24 . J. Foster, O. Gürdoğan, M. von Hippel, A. J. McLeod,
[26] Note that although the cross-ratios are seemingly in- and G. Papathanasiou, PoS CORFU2019, 003 (2020),
dependent e.g. under x ↔ −x, this transformation re- arXiv:2005.06735 [hep-th].
quires analytic continuation. See e.g. Duplancic, Nizic, [36] D. Correa, J. Henn, J. Maldacena, and A. Sever, JHEP
Eur.Phys.J.C 24 (2002) for a related discussion. 06, 048 (2012), arXiv:1202.4455 [hep-th].
[27] S. Caron-Huot and J. M. Henn, JHEP 06, 114 (2014),
arXiv:1404.2922 [hep-th].
[28] A. B. Goncharov, M. Spradlin, C. Vergu, and
A. Volovich, Phys. Rev. Lett. 105, 151605 (2010),

You might also like