The Flat-Space S-Matrix as Conformal Data on the Celestial Sphere: Mellin Bases, Asymptotic Symmetries, and Soft Theorems as Ward Identities as the Boundary Layer of Reconstructed Spacetime

1 Introduction

1.1 The reconstruction thesis in the flat-space regime

A scattering amplitude in four-dimensional Minkowski space is ordinarily presented as a function of on-shell momenta, computed from a Lagrangian by a Feynman expansion or a recursion relation, and evaluated on plane-wave asymptotic states. In this presentation the spacetime is fixed background data and the amplitude is a derived object. Celestial conformal field theory reverses the order of explanation. It treats the $S$-matrix as a family of correlation functions on the two-sphere at null infinity, and it treats the symmetries of that boundary — the asymptotic Bondi–Metzner–Sachs (BMS) group and its soft-theorem Ward identities — as the invariant data from which the bulk scattering process is to be reconstructed.

This paper develops that reversal as an instance of a general thesis about quantum gravity that we state once and then use as an organizing principle: quantum gravity is not primarily the quantization of material objects in a fixed spacetime; it is the reconstruction of spacetime, matter, causality, and measurement from invariant representation structures. In the flat-space regime the invariant structure is the asymptotic symmetry algebra together with its Ward identities. The soft theorems are those Ward identities. The Mellin transform to boost eigenstates is the change of basis that makes the boundary conformal structure manifest. The celestial correlators are the reconstructed $S$-matrix, now presented as conformal data.

We are careful throughout to separate what is established from what is programmatic. The identification of the four-dimensional Lorentz group with the global conformal group of the celestial sphere is a statement about the action of $\mathrm{SL}(2,\mathbb{C})$ on $\mathbb{CP}^{1}$, and it is exact. The Mellin transform is an invertible change of basis on the principal series. The equivalence between Weinberg’s leading soft theorem and a supertranslation Ward identity is a theorem at tree level. By contrast, the assertion that these correlators are governed by an autonomous, unitarity- and locality-satisfying two-dimensional conformal field theory — a genuine holographic dual with its own definition independent of the bulk amplitudes it must reproduce — is not established. We mark it as such.

1.2 What is standard, what is heuristic, what is speculative

The project within which this paper sits attaches to every representational claim an epistemic warrant drawn from the ordered set $S < H < P$: $S$ for a standard mathematical or mathematical-physics correspondence, $H$ for a strong heuristic physical representation, and $P$ for a speculative ontological extension. Warrants compose by a worst-component-wins rule, so a chain of reasoning inherits the weakest link in it. We use these tags on every labelled result. In the celestial regime the tags fall out roughly as follows.

• [ S ] The Lorentz action $\mathrm{SL}(2,\mathbb{C})$ on the celestial sphere as Möbius transformations; the Mellin transform to conformal primary wavefunctions and the completeness of the principal-series basis; the kinematic fact that null directions in $\mathbb{R}^{1,3}$ are points of $\mathbb{CP}^{1}$; Weinberg’s soft factorization as a statement about amplitudes; the equivalence of the leading soft-graviton theorem with the supertranslation Ward identity at tree level.

• [ S/H ] The celestial operator product expansion read off from collinear limits; the appearance of the Euler beta function in the leading celestial gluon OPE; the identification of conformally soft modes with symmetry currents.

• [ H ] The organization of the soft towers into current algebras and the $w_{1+\infty}$ wedge algebra; the interpretation of asymptotic-symmetry charges as celestial-CFT currents.

• [ H/P ] The existence of a complete, intrinsically defined celestial CFT that serves as a holographic dual of flat-space quantum gravity, with the attendant unitarity, reflection-positivity, and spectral properties of a standard conformal field theory.

We keep the celestial status distinct from the AdS/CFT status, where a large class of examples supply a boundary theory with an independent definition. In flat space no such independent definition is available. The correct comparison is therefore not “celestial holography is like AdS/CFT” but “celestial holography reorganizes known $S$-matrix data into conformal form, and asks whether that form is autonomous.”

1.3 Relation to the surrounding modular framework

This is one entry in a modular library of quantum-gravity regimes, each represented by its own invariant. The amplitude regime represents scattering by positive geometries and their canonical forms; the holographic (AdS) regime represents bulk reconstruction by tensor-network and operator-algebra error correction; the covariant regime represents histories by spin foams. The present regime represents the flat-space $S$-matrix by celestial conformal data, with the soft theorems as the invariant. These regimes are not collapsed into one theory. They overlap partially, and the overlaps are recorded as translations with explicit status, not as identities. 5 records the concrete overlaps with the amplitude regime (double copy, positive geometry) and with the asymptotic-safety regime, using the same $S/H/P$ calculus.

1.4 Organization

2 sets up the celestial sphere, the Lorentz-conformal identification, the Mellin transform and conformal primary wavefunctions, the BMS group, and the celestial OPE. 3 states the main results as labelled propositions and theorems with proofs or proof sketches, each tagged. 4 works two examples: the Mellin transform of a massless four-point contact amplitude, and Weinberg’s soft factor in celestial variables together with the supertranslation charge. 5 records the relation to other regimes with the status calculus. 6 states the limitations and open problems honestly. 7 documents the accompanying Haskell code. 9 concludes. [app:mellin,app:soft] collect computational details.

2 Mathematical framework

2.1 Null infinity, the celestial sphere, and the Lorentz-conformal map

Fix Minkowski space $\mathbb{R}^{1,3}$ with mostly-plus metric $\eta_{\mu\nu}=\mathrm{diag}(-1,+1,+1,+1)$ and coordinates $X^{\mu}$. Massless particles reach the conformal boundary at null infinity $\mathscr{I}= \mathscr{I}^{+}\cup \mathscr{I}^{-}$, each component of which has topology $\mathbb{R}\times S^{2}$: a null line $\mathbb{R}$ (retarded or advanced time) fibered over a two-sphere of null directions. That two-sphere is the celestial sphere. We coordinatize it by a complex stereographic coordinate $z \in \mathbb{C}\cup \{\infty\} = \mathbb{CP}^{1}$.

A future-pointing null momentum is written $$\begin{equation} p^{\mu} = \omega\, q^{\mu}(z,\bar{z}), \qquad q^{\mu}(z,\bar{z}) = \bigl(1+z\bar{z},\; z+\bar{z},\; -\mathrm{i}(z-\bar{z}),\; 1-z\bar{z}\bigr), \end{equation}$$ where $\omega > 0$ is the energy and $q^{\mu}$ is a fixed null reference vector in the direction labelled by $z$. One checks directly that $\eta_{\mu\nu}q^{\mu}q^{\nu} = 0$, so $p^{\mu}$ is null for every $z$, and that $q^{\mu}$ is future-pointing. The map $z \mapsto q^{\mu}(z,\bar{z})/q^{0}$ is a diffeomorphism from $\mathbb{CP}^{1}$ onto the celestial sphere of null directions.

The Lorentz group $\mathrm{SO}^{+}(1,3){}$ acts on null directions, hence on the celestial sphere. Its double cover is $\mathrm{SL}(2,\mathbb{C})$, and the induced action on the stereographic coordinate is the Möbius transformation $$\begin{equation} z \longmapsto \frac{az+b}{cz+d}, \qquad \begin{pmatrix} a & b \\ c & d \end{pmatrix} \in \mathrm{SL}(2,\mathbb{C}), \qquad ad-bc = 1, \end{equation}$$ with $\bar{z}\mapsto (\bar a\bar{z}+ \bar b)/(\bar c\bar{z}+ \bar d)$; we write $z' = (az+b)/(cz+d)$ for the image point. The corresponding Lorentz transformation $\Lambda$ acts on the reference null vector by the standard identity $$\begin{equation} \Lambda^{\mu}{}_{\nu}\,q^{\nu}(z,\bar{z}) = |cz+d|^{2}\,q^{\mu}(z',\bar{z}'), \end{equation}$$ so the null direction is carried to $z'$ and the energy is rescaled by the conformal factor $|cz+d|^{2}$, which records the boost weight. This is the content of the statement that the four-dimensional Lorentz group is the two-dimensional global conformal group.

2.2 The Mellin transform to boost eigenstates

Plane-wave states $\ket{p} = \ket{\omega,z,\bar{z}}$ diagonalize the momentum operator. Boost eigenstates diagonalize instead the boost generator that scales $\omega$. The change of basis is a Mellin transform in the energy.

Applied to a plane wave $e^{\pm\mathrm{i}\omega\,q\cdot X}$ the Mellin transform produces the massless conformal primary wavefunction of Pasterski and Shao.

The evaluation in [eq:cpw] is the standard integral $\int_{0}^{\infty}\mathrm{d}\omega\,\omega^{\Delta-1}\mathrm{e}^{-s\omega} = \Gamma(\Delta) s^{-\Delta}$ with $s = \mp\mathrm{i}\,q\cdot X + \epsilon$, valid for $\Re\Delta> 0$ and continued in $\Delta$. Since $\Delta$ is complex on the principal series, the complex power $s^{-\Delta}$ is defined on the principal branch of the logarithm, $s^{-\Delta}=\mathrm{e}^{-\Delta\log s}$ with $\log s$ real for $s>0$ and $\arg s\in (-\pi,\pi]$; the $\mp\mathrm{i}\epsilon$ prescription keeps $s$ off the branch cut. In this convention [eq:cpw] is the form $\Gamma(\Delta)\,(\epsilon\mp\mathrm{i}\,q\cdot X)^{-\Delta}$ used in the standard references. Because $q\cdot X$ is linear in $X$ and scales as $|cz+d|^{2}$ under [eq:qtransf], the wavefunction picks up the primary transformation law $$\begin{equation} \varphi^{\pm}_{\Delta}\!\left(X; \tfrac{az+b}{cz+d}\right) = |cz+d|^{2\Delta}\,\varphi^{\pm}_{\Delta}(\Lambda^{-1}X; z), \end{equation}$$ which is the defining property of a scalar conformal primary of dimension $\Delta$. The completeness statement is due to Pasterski and Shao.

The proof reduces to Mellin inversion: the map $g(\omega) \mapsto \widetilde{g}(1+\mathrm{i}\lambda)$ is, up to a change of variables $\omega = \mathrm{e}^{t}$, the Fourier transform on $\mathbb{R}$, which is unitary. Completeness of plane waves then transports to completeness of conformal primaries on the critical line. We record this as 5 below.

2.3 Celestial amplitudes as conformal correlators

Let $\mathcal{A}_{n}(\{\omega_{i},z_{i},\bar{z}_{i};J_{i}\})$ be a massless $n$-point $S$-matrix element, with $J_{i}$ the helicities; for a massless particle the spin of the associated celestial primary equals its helicity, so we use the single symbol $J_{i}$ throughout. The celestial amplitude is its Mellin transform in every external energy.

2.4 Asymptotic symmetries: the BMS group

The isometry group of Minkowski space is the Poincaré group. The asymptotic symmetry group of asymptotically flat spacetimes at null infinity is larger. It was found by Bondi, van der Burg and Metzner and by Sachs to be the infinite-dimensional BMS group.

Barnich and Troessaert extended the Lorentz factor to the full local conformal (Virasoro) algebra, giving superrotations, and computed the surface charges. The charge algebra realizes the symmetry with a possible field-dependent central extension satisfying a generalized cocycle condition.

The physically load-bearing feature, shared with every regime in this library, is that the boundary data are organized by a group action with retained automorphism data: the supertranslation and superrotation charges are the generators, and physical amplitudes are constrained to be invariant under them, not merely under the finite Poincaré subgroup.

2.5 Soft theorems as Ward identities

A soft theorem states that an amplitude with one extra low-energy massless particle factorizes, in the limit of vanishing energy of that particle, into a universal soft factor times the amplitude without it. Weinberg established the leading (pole) soft factors for photons and gravitons.

The reinterpretation of these factorization statements as Ward identities of asymptotic symmetries is the central structural result of the regime. We state the two cases and prove the abelian one in detail in 3.

2.6 Celestial operator product expansion and conformally soft currents

Collinear limits of massless amplitudes, where two external momenta become parallel, translate under the Mellin transform into the coincidence limit $z_{1}\to z_{2}$ of two celestial operators. This gives a celestial operator product expansion (OPE).

The beta function follows from Mellin-transforming the universal collinear splitting factor of Yang–Mills amplitudes. Its poles, at $\Delta_{1}=1$ or $\Delta_{2}=1$ and their descendants, mark the conformally soft operators: when $\Delta\to 1$ a gluon primary becomes a dimension-one holomorphic current, and the residues of the OPE reproduce a Kac–Moody-type current algebra. This is the celestial face of the soft theorem: the soft gluon theorem is the Ward identity of the current at $\Delta=1$. For gravitons the analogous tower of conformally soft operators at $\Delta= 1,0,-1,\dots$ organizes, in the positive-helicity self-dual sector, into the wedge subalgebra of $w_{1+\infty}$.

3 Main results

We collect the results as labelled statements. Each carries an $S/H/P$ tag. Proofs are complete where the statement is standard and are sketches, with references to the primary literature, where the result is heuristic.

3.1 The change of basis is faithful and covariant

14 is the precise sense in which the celestial correlators represent the flat-space $S$-matrix: the representation is a faithful, covariant, invertible change of basis. It does not by itself assert that the correlators are governed by an autonomous CFT. That is the content of the $H/P$-labelled program in 6.

3.2 Leading soft photon equals large-gauge Ward identity

3.3 Leading soft graviton equals supertranslation Ward identity

3.4 Celestial soft currents

4 Worked examples

4.1 Mellin transform of a massless four-point contact amplitude

We compute the celestial transform of the simplest massless amplitude, a four-point contact term, to display the distributional support of celestial correlators. Take $$\begin{equation} \mathcal{A}_{4} = \mathrm{i}\lambda_{0}\,(2\pi)^{4}\,\delta^{4}\!\Bigl(\sum_{i=1}^{4} \eta_{i}\,\omega_{i}\,q_{i}^{\mu}\Bigr), \end{equation}$$ with $\eta_{i}=+1$ for outgoing and $-1$ for incoming, $\lambda_{0}$ a coupling. The four null vectors $q_{i}^{\mu}=q^{\mu}(z_{i},\bar{z}_{i})$ are fixed by the directions and the delta function enforces momentum conservation.

Write the four-fold Mellin transform $$\begin{equation} \widetilde{\mathcal{A}}_{4} = \mathrm{i}\lambda_{0}(2\pi)^{4}\prod_{i=1}^{4}\int_{0}^{\infty} \mathrm{d}\omega_{i}\,\omega_{i}^{\Delta_{i}-1}\, \delta^{4}\!\Bigl(\sum_{i}\eta_{i}\omega_{i}q_{i}\Bigr). \end{equation}$$ Represent the delta function as an integral over a bulk point $X^{\mu}$, $\delta^{4}(P)=(2\pi)^{-4}\int\mathrm{d}^{4}X\,\mathrm{e}^{\mathrm{i}P\cdot X}$, and do the $\omega_{i}$ integrals using [eq:cpw]. Each gives a conformal primary $\varphi_{\Delta_{i}}(X;z_{i})=\Gamma(\Delta_{i})(\mathrm{i}\eta_{i}q_{i}\cdot X + \epsilon)^{-\Delta_{i}}$, so $$\begin{equation} \widetilde{\mathcal{A}}_{4} = \mathrm{i}\lambda_{0}\prod_{i=1}^{4}\Gamma(\Delta_{i}) \int\mathrm{d}^{4}X\;\prod_{i=1}^{4} \bigl(\mathrm{i}\eta_{i}\,q_{i}\cdot X + \epsilon\bigr)^{-\Delta_{i}}. \end{equation}$$ The remaining bulk integral is a conformal integral. Its value depends on the $z_{i}$ only through the $\mathrm{SL}(2,\mathbb{C})$-invariant cross-ratio $$\begin{equation} \mathrm{z} = \frac{z_{12}z_{34}}{z_{13}z_{24}}, \qquad \bar{\mathrm{z}} = \frac{\bar{z}_{12}\bar{z}_{34}}{\bar{z}_{13}\bar{z}_{24}}, \end{equation}$$ and, after stripping the covariant weight factors dictated by 7, reduces to a function of $\mathrm{z},\bar{\mathrm{z}}$ supported on the real locus $\mathrm{z}=\bar{\mathrm{z}}$. The physical reason is that four generic null directions in $\mathbb{R}^{1,3}$ cannot conserve momentum: the delta function [eq:contact] forces a relation among the directions, which in cross-ratio variables is $\mathrm{z}=\bar{\mathrm{z}}$. Hence $$\begin{equation} \widetilde{\mathcal{A}}_{4} = \bigl(\text{weight factors in } z_{ij},\bar{z}_{ij}\bigr)\times \delta(\mathrm{z}-\bar{\mathrm{z}})\times f(\mathrm{z};\Delta_{i}), \end{equation}$$ with $f$ a function of the cross-ratio and the dimensions. The full computation, including the constant, is in 10.

4.2 Weinberg’s soft factor in celestial variables and the supertranslation charge

We now translate the leading soft-graviton factor into celestial variables and exhibit its match to the supertranslation charge, illustrating 16. Take the soft graviton to have positive helicity, direction $z$, and polarization tensor $\varepsilon^{+}_{\mu\nu} = \varepsilon^{+}_{\mu}\varepsilon^{+}_{\nu}$ with $$\begin{equation} \varepsilon^{+}_{\mu}(z,\bar{z}) = \frac{1}{\sqrt2}\,\partial_{z}q_{\mu}(z,\bar{z}) = \frac{1}{\sqrt2}\bigl(-\bar{z},\,1,\,-\mathrm{i},\,-\bar{z}\bigr). \end{equation}$$ For a hard massless particle $k$ of energy $\omega_{k}$ and direction $z_{k}$, so that $p_{k}^{\mu}=\omega_{k}q^{\mu}(z_{k},\bar{z}_{k})$, the inner products are $$\begin{equation} q\cdot p_{k} = -2\,\omega_{k}\,|z-z_{k}|^{2}, \qquad \varepsilon^{+}\cdot p_{k} = -\sqrt2\,\omega_{k}\,(\bar{z}-\bar{z}_{k}), \end{equation}$$ derived by direct expansion of [eq:nullparam] and [eq:polvec] in 11. Substituting these into the graviton soft factor gives $$\begin{equation} S^{(0)}_{\text{graviton}} = \frac{\kappa}{2}\sum_{k} \frac{(\varepsilon^{+}\cdot p_{k})^{2}}{q\cdot p_{k}} = -\frac{\kappa}{2}\sum_{k}\omega_{k}\, \frac{\bar{z}-\bar{z}_{k}}{z-z_{k}}. \end{equation}$$ The soft factor in celestial variables is a sum over hard particles of $\omega_{k}(\bar{z}-\bar{z}_{k})/(z-z_{k})$: it has a simple holomorphic pole at $z=z_{k}$ with residue proportional to the energy $\omega_{k}$ of particle $k$.

Now compare with the hard supertranslation charge from the proof of 16. Acting on external states, $\mathcal{Q}^{\text{hard}}_{f}$ gives $\sum_{k}f(z_{k},\bar{z}_{k})\,\omega_{k}$. The Ward identity [eq:gravward] equates this to the soft insertion. Choosing $f$ to be the Green’s function that inverts the operator appearing in the soft-graviton constraint, namely $f_{z}(w,\bar w)$ with $\partial_{\bar w}^{2}f_{z} \propto \delta^{2}(w-z)$ so that $f_{z}(w) \sim (\bar w-\bar z)/(w-z)$ near $w=z$, the soft insertion reproduces exactly the celestial soft factor [eq:softcelestial]. The pole at $z=z_{k}$ with residue $\omega_{k}$ on the soft side matches the evaluation $f_{z}(z_{k})\,\omega_{k}$ on the hard side. This is the promised match: the soft theorem in celestial variables [eq:softcelestial] is the supertranslation Ward identity [eq:gravward], smeared against the appropriate $f$.

5 Relation to other regimes and the status calculus

The celestial regime overlaps three neighbouring regimes in the library. We record the overlaps as translations with explicit $S/H/P$ status. Composition of statuses is worst-component-wins, so a chain of translations inherits its weakest link.

5.1 Amplitudes: double copy and positive geometry

The momentum-space amplitudes that we Mellin-transform are the same objects studied in the amplitude regime. Two relations are worth stating.

5.2 Asymptotic safety

5.3 AdS/CFT: keeping the status distinct

5.4 Status ledger

1 collects the tags. Reading it top to bottom is reading a reconstruction pipeline: each row is a translation, and the pipeline’s overall warrant is the worst entry, $H/P$.

Status ledger for the celestial regime. Warrants compose worst-component-wins; the pipeline that ends in an autonomous dual is $H/P$.

Statement	Warrant
Lorentz $=\mathrm{SL}(2,\mathbb{C})$ acts as global conformal group on the celestial sphere (1)	S
Mellin transform to conformal primaries; completeness on principal series ([prop:completeness,prop:mellininvertible])	S
Celestial amplitude transforms as a 2D conformal correlator ([prop:covariance,thm:faithful])	S
Leading soft photon $=$ large-gauge Ward identity (15)	S
Leading soft graviton $=$ supertranslation Ward identity (16)	S
Subleading soft graviton $=$ superrotation Ward identity (17)	S/H
Celestial OPE from collinear limits; Euler beta coefficient ([def:gluonope,prop:betaope])	S/H
Conformally soft currents (18)	S/H
Soft towers as current algebras; $w_{1+\infty}$ wedge (20)	H
Double copy through the Mellin transform (23)	S/H
Positive-geometry structure of celestial correlators (24)	H
Celestial spectrum $\leftrightarrow$ asymptotic-safety exponents (25)	H/P
Autonomous celestial CFT / complete holographic dual (6)	H/P

6 Limitations and open problems

We state the limitations plainly. Each is a place where the current construction stops short of a complete holographic duality.

6.0.0.1 No independent definition of the dual.

[ H/P ] The celestial CFT is presently defined by the flat-space amplitudes it must reproduce. There is no known intrinsic definition — no Lagrangian, no complete set of axioms with a solved spectrum — from which the amplitudes would follow. This is the sharpest difference from the best-understood AdS/CFT examples and the reason we tag the strong holographic claim $H/P$. It is also the concrete content of open-problem 6 in the library’s queue: encode the assumptions for the scattering-basis transform, the soft sectors, and the asymptotic-symmetry constraints, rather than assert a dual.

6.0.0.2 Distributional correlators and continuous spectrum.

[ S ] As shown in 4.1, translation invariance forces momentum-conserving delta functions whose Mellin transforms are distributions supported where the cross-ratios are real. The operator spectrum is the continuous principal series $\Delta\in1+\mathrm{i}\mathbb{R}$, not a discrete set of primaries. Both features distinguish celestial correlators from standard compact 2D CFT correlators and obstruct a naive state-operator correspondence.

6.0.0.3 Reflection positivity and unitarity.

[ H ] Whether the celestial correlators define a reflection-positive (unitary) 2D CFT is not established. The bulk $S$-matrix is unitary, but unitarity in the celestial basis is not the standard 2D reflection positivity, and the map between the two is not understood.

6.0.0.4 Signature and contours.

[ S/H ] The cleanest holomorphic factorization occurs in $(2,2)$ split signature, where $z$ and $\bar{z}$ are independent real variables. In Lorentzian $(1,3)$ signature $\bar{z}$ is the complex conjugate of $z$ and the correlators are distributions (21). Relating the two requires care with contours and analytic continuation.

6.0.0.5 Loop corrections and the soft algebras.

[ H ] The $w_{1+\infty}$ organization of the soft graviton tower is established in the tree-level self-dual sector. Loop corrections, and the fate of the algebra in the full nonlinear theory, are under study. We do not claim the algebra is exact beyond the sectors where it has been checked.

6.0.0.6 Massive states and celestial primaries.

[ S/H ] Massive external states map to celestial primaries by an integral over the mass hyperboloid rather than a pure Mellin transform, and the treatment of massive particles, bound states, and thresholds in the celestial basis is less complete than the massless case.

7 Accompanying code

The repository includes a small Haskell package that makes the paper’s computational claims checkable. It is base-only (no external dependencies) and compiles with ghc. The modules are:

• Core.hs: the Mellin kernel and a numerical Mellin transform; the Möbius action [eq:mobius] on $\mathbb{CP}^{1}$; the cross-ratio [eq:crossratio]; the weight bookkeeping [eq:weights]; the Euler beta function; and the celestial soft factor [eq:softcelestial].

• Properties.hs: a suite of checkable properties, listed below.

• Main.hs: demonstrations that print the results and run the property suite, exiting non-zero on any failure.

The checkable properties are:

1. Mellin normalization. The Mellin transform of $\mathrm{e}^{-\omega}$ at $\Delta=n$ equals $\Gamma(n)=(n-1)!$, checked numerically against the factorial (2, [eq:cpw]).

2. Möbius group action. Composition of two Möbius maps equals the map of the product matrix, and the cross-ratio [eq:crossratio] is invariant under a common Möbius transformation of the four points (1).

3. Weight bookkeeping. For every $(\Delta,J)$, the weights $(h,\bar{h})=((\Delta+J)/2,(\Delta-J)/2)$ satisfy $h+\bar{h}=\Delta$ and $h-\bar{h}=J$ (7).

4. Euler beta identities. $\mathrm{B}(a,b)=\mathrm{B}(b,a)$ and $\mathrm{B}(a,b)=\Gamma(a)\Gamma(b)/\Gamma(a+b)$, and $\mathrm{B}$ has poles at $a\in\{0,-1,-2,\dots\}$, checked against the conformally soft points (19).

5. Soft-factor homogeneity. The leading soft factor [eq:softcelestial] is homogeneous of degree $-1$ under a common rescaling of the soft direction, and its residue at $z=z_{k}$ is proportional to $\omega_{k}$ (16, [eq:softcelestial]).

6. Momentum conservation. For the configurations used in 4.1, a set of null momenta $\{\eta_{i}\omega_{i}q^{\mu}(z_{i})\}$ satisfies $$\sum_{i}\eta_{i}\omega_{i}\,q^{\mu}(z_{i})=0$$ within numerical tolerance.

8 Discussion

The celestial construction makes precise a limited but genuine sense in which the flat-space $S$-matrix is boundary conformal data. The precise part is the change of basis: 14 shows that Mellin-transforming to boost eigenstates is a faithful, covariant, invertible presentation of the massless $S$-matrix as a 2D-conformally-covariant object. The genuine part is the symmetry content: [thm:softphoton,thm:softgraviton] show that the leading soft theorems, the universal infrared data of the $S$-matrix, are exactly the Ward identities of the asymptotic symmetries acting on the celestial sphere. The soft theorems are the invariant boundary data, and the reconstruction thesis is literally realized: from the boundary current (the soft factor [eq:softcelestial]) and its zero mode (the supertranslation charge) one recovers the bulk soft dynamics.

The limited part is equally important. The celestial correlators are distributional, the spectrum is continuous, reflection positivity is open, and there is no independent definition of the dual. These are not cosmetic. They are the reasons the strong holographic claim sits at $H/P$ and must be kept distinct from AdS/CFT. The honest statement is that celestial CFT is a powerful reorganization of $S$-matrix data by its asymptotic symmetries, which has uncovered real new structure (the celestial OPE, the conformally soft currents, the $w_{1+\infty}$ tower), and which aspires to an autonomous dual without yet possessing one.

Within the modular library this is the intended reading. Each regime supplies a faithful representation of some quantum-gravity data by an invariant, and the regimes are related by translations of explicit status rather than fused into one theory. The celestial regime’s invariant is the soft-theorem/asymptotic-symmetry Ward identity, and its honest warrant ledger is 1.

9 Conclusion

We have presented celestial CFT as a representation of the flat-space $S$-matrix, organized around the thesis that quantum gravity in this regime is the reconstruction of scattering from invariant boundary data. Three standard statements carry the account: Lorentz is the global conformal group of the celestial sphere (1); the Mellin transform is a faithful, invertible change of basis to conformal primaries (14); and the leading soft theorems are the Ward identities of the asymptotic symmetries ([thm:softphoton,thm:softgraviton]). Beyond these we recorded the celestial OPE and its Euler-beta structure, the conformally soft currents, and the $w_{1+\infty}$ tower, at their honest heuristic status, and we worked the Mellin transform of a contact amplitude and the celestial soft factor in full. We stated the limitations without softening them: distributional correlators, continuous spectrum, open reflection positivity, and, above all, the absence of an independent definition of the dual, which keeps the celestial holographic claim at $H/P$ and distinct from AdS/CFT. The accompanying code checks the Mellin normalization, the conformal group action, the weight bookkeeping, the Euler beta structure, and the soft-factor homogeneity.

10 Mellin transform of the contact amplitude

We fill in 4.1. Starting from [eq:mellin4ptX], $$\begin{equation} \widetilde{\mathcal{A}}_{4} = \mathrm{i}\lambda_{0}\prod_{i=1}^{4}\Gamma(\Delta_{i}) \int\mathrm{d}^{4}X\;\prod_{i=1}^{4} \bigl(\mathrm{i}\eta_{i}\,q_{i}\cdot X + \epsilon\bigr)^{-\Delta_{i}}, \end{equation}$$ introduce Schwinger parameters $t_{i}\ge0$ via $(\mathrm{i}\eta_{i}q_{i}\cdot X+\epsilon)^{-\Delta_{i}} = \frac{1}{\Gamma(\Delta_{i})}\int_{0}^{\infty}\mathrm{d}t_{i}\,t_{i}^{\Delta_{i}-1} \mathrm{e}^{-t_{i}(\mathrm{i}\eta_{i}q_{i}\cdot X+\epsilon)}$. Then $$\begin{equation} \widetilde{\mathcal{A}}_{4} = \mathrm{i}\lambda_{0}\int\!\prod_{i}\mathrm{d}t_{i}\,t_{i}^{\Delta_{i}-1} \int\mathrm{d}^{4}X\;\mathrm{e}^{-\mathrm{i}X\cdot\sum_{i}\eta_{i}t_{i}q_{i}-\epsilon\sum_i t_i} = \mathrm{i}\lambda_{0}(2\pi)^{4}\int\!\prod_{i}\mathrm{d}t_{i}\,t_{i}^{\Delta_{i}-1}\, \delta^{4}\!\Bigl(\sum_{i}\eta_{i}t_{i}q_{i}\Bigr), \end{equation}$$ which is the same structure as [eq:mellin4pt] with $t_{i}$ in place of $\omega_{i}$, confirming self-consistency. The delta function constrains the four $t_{i}$ (up to one overall scale) to the locus where the four null vectors sum to zero. Parametrize by solving three of the four components; the remaining integral over the overall scale $s=\sum_{i}t_{i}$ gives, using $\sum_{i}\Delta_{i} = 4$ on the principal series in the balanced case, a factor $\int_{0}^{\infty}\mathrm{d}s\,s^{\sum\Delta_{i}-4-1}$ times a Jacobian $J(z_{i})$. The Jacobian is the modulus of the determinant of the $4\times3$ system reduced to the momentum-conservation surface; a short computation expresses it in the covariant weight factors of 7 times the delta function $\delta(\mathrm{z}-\bar{\mathrm{z}})$ of [eq:distributional]. The upshot is [eq:distributional]: after removing the weights the correlator is a function of the single real cross-ratio $\mathrm{z}$, supported on $\mathrm{z}=\bar{\mathrm{z}}$. The locus $\mathrm{z}=\bar{\mathrm{z}}$ has a clean geometric meaning: a real cross-ratio means the four points $z_{1},\dots,z_{4}$ lie on a common circle on the Riemann sphere, which is the celestial imprint of the four null momenta being linearly dependent (spanning a three-dimensional subspace of $\mathbb{R}^{1,3}$), exactly the condition that momentum conservation among four massless particles imposes. The appearance of a single cross-ratio, rather than the independent $\mathrm{z},\bar{\mathrm{z}}$ of a generic 2D CFT, is the distributional feature discussed in 21.

11 Contraction of the soft factor

We fill in 4.2. With $q^{\mu}=q^{\mu}(z,\bar{z})$ from [eq:nullparam] and $p_{k}^{\mu}=\omega_{k}q^{\mu}(z_{k},\bar{z}_{k})$, compute $$\begin{equation} q\cdot p_{k} = \omega_{k}\,\eta_{\mu\nu}q^{\mu}(z)q^{\nu}(z_{k}) = -2\,\omega_{k}\,|z-z_{k}|^{2}, \end{equation}$$ using $\eta_{\mu\nu}q^{\mu}(z)q^{\nu}(w) = -2|z-w|^{2}$, which follows from [eq:nullparam] by direct expansion. For the positive-helicity polarization [eq:polvec], $$\begin{equation} \varepsilon^{+}\cdot p_{k} = \omega_{k}\,\varepsilon^{+}_{\mu}(z)\,q^{\mu}(z_{k}) = -\sqrt2\,\omega_{k}\,(\bar{z}-\bar{z}_{k}), \end{equation}$$ where the last equality uses $\varepsilon^{+}_{\mu}(z)q^{\mu}(w)=-\sqrt2(\bar{z}-\bar w)$, again by direct expansion of [eq:polvec] and [eq:nullparam]. Therefore $$\begin{equation} \frac{(\varepsilon^{+}\cdot p_{k})^{2}}{q\cdot p_{k}} = \frac{2\,\omega_{k}^{2}\,(\bar{z}-\bar{z}_{k})^{2}} {-2\,\omega_{k}\,(z-z_{k})(\bar{z}-\bar{z}_{k})} = -\,\omega_{k}\,\frac{\bar{z}-\bar{z}_{k}}{z-z_{k}}, \end{equation}$$ and summing over hard particles with the prefactor $\kappa/2$ gives [eq:softcelestial]. The residue at $z=z_{k}$ is $-\tfrac{\kappa}{2}\omega_{k}(\bar{z}-\bar{z}_{k})$, proportional to the energy $\omega_{k}$, as used in 4.2. The homogeneity check in 7 is the statement that under $z-z_{k}\mapsto s(z-z_{k})$, $\bar{z}-\bar{z}_{k}\mapsto\bar s(\bar{z}-\bar{z}_{k})$ the factor scales as $\bar s/s$, of total degree $0$ in modulus and degree $-1$ in the holomorphic variable, matching a weight-$(1,0)$ current insertion.