
In this paper we will advocate for an (infinite-dimensional) geometric interpretation of (1.1)
which combines observations on the vacuum structure of gauge theory [1–4] with recent geometric
results on soft limits in the sigma model [5,6]. In a word, (1.1) is true because the amplitude
AK1¨¨¨Kn;IJ,ab
n`2is actually a section of a bundle over an infinite-dimensional space of vacua Mlabeled
by flat (trivial) G´connections Capxqon the celestial sphere. Insertions of soft gluons in the S-
matrix enact parallel transport about this space, and (1.1) simply computes the holonomy around an
infinitesimal closed curve, also known as the Riemann curvature RpX, Y qZ“ r∇X,∇YsZ´∇rX,Y sZ
„lim
qIpxqÑ0ω, lim
qJpyqÑ0ω1ȷAK1¨¨¨Kn;IJ,ab
n`2“R˜δ
δĂ
CI
apxq
,δ
δĂ
CJ
bpyq¸AK1¨¨¨Kn
n.(1.2)
In this formula, the (shadow-transformed) flat connections r
Capxqare local coordinates on M,δ
δĂ
CI
apxq
is an associated vector field, and the fact that the antisymmetric double-soft limit does not vanish
simply means that the vacuum manifold for Yang-Mills theory in asymptotically flat space is curved.
In other words, there is no “right definition” for multi-soft limits in non-abelian gauge theory, just
as there is no unique parallel transport between two points in a curved space. The result is path
dependent, and (1.1) is just an infinite-dimensional manifestation of that geometric fact. Note that
in this language, the vanishing of the antisymmetric double-soft limit in gravity is a statement
about flatness of the space of supertranslation vacua, and a similar statement holds for abelian
gauge theory.
The claim that there exists an infinite-dimensional space of vacua for gauge theory in asymp-
totically flat space is not new. In fact, this observation [2], and its analog in gravity [1], was the
genesis of the “celestial CFT” program which seeks to construct a boundary dual for asymptotically
flat gravity. The existence of these vacua is a subtle boundary effect which only holds for gauge
theories in the Coulomb phase on non-compact spaces, but the interpretation of (1.1) advanced in
this paper makes it clear that the moduli space of vacua is detectable using standard perturbative
calculations: the formula (1.1) can be derived using standard Feynman rules with no reference to
asymptotic symmetries, boundary conditions, or gauge transformations with non-compact support.
Perturbative S-matrix calculations in Yang-Mills theory (which are by definition performed in the
Coulomb phase, with long-range gauge fields) know about the infinite-dimensional space of vacua.
In fact they know more: they can calculate its curvature.
To put these statements in context, note that the antisymmetric double-soft limit has historically
been a discovery mechanism for hidden symmetries of the S-matrix (see for instance [7] for the case
of E7p7qin N“8supergravity in four dimensions). The gauge theory examples discussed in this
work are really just infinite-dimensional examples of this phenomena. Indeed, these statements
are all simpler and more familiar when the moduli space of vacua Mis finite-dimensional. In this
case, the vacuum is determined by boundary conditions (vacuum expectation values, or vevs) at
spatial infinity (i0) for gapless local fields xΦIyi0“vI, and the dynamics of the long-wavelength
fluctuations about these vevs is described by a sigma model with target space M. The single soft
insertion of a moduli scalar defines an operator whose matrix element is a derivative on the moduli
3