The standard method for computing scattering amplitudes in quantum field theory is to draw Feynman diagrams: pictorial representations of the ways particles can interact, each contributing a term to the perturbative expansion of the S-matrix. For electrodynamics and the electroweak theory this is manageable. For quantum chromodynamics at high loop order it becomes impractical. For quantum gravity it is catastrophically worse: the number of distinct diagrams at L loops grows factorially, and the series is not even Borel summable. Something is fundamentally wrong with the approach, not just technically difficult.
The loop amplitudes paper presents an alternative that avoids diagrams entirely. Every multi-loop scattering amplitude can be computed from a single operation: taking the shadow discontinuity of a tree-level amplitude with more external legs.
What a Shadow Pole Is
In the celestial OPE, two operators ÕΔ1 and ÕΔ2 fuse when their worldlines approach each other. The fusion has a pole when Δ1 + Δ2 = 2: the locus where a shadow pair is created at the collision point. This shadow pole encodes, in boundary language, the exchange of a virtual particle in the bulk. The residue at the pole is the tree-level three-point amplitude. The discontinuity across the pole, defined as the difference between the boundary value from above and below, is the one-loop contribution.
At L loops the pattern extends: the L-loop n-point integrand is the L-fold shadow discontinuity of a tree amplitude with 2L additional external legs. The recursion terminates because tree amplitudes are known in closed form from the Penrose–Ward construction.
The Optical Theorem at the Boundary
This is not a coincidence or a clever trick. It is the boundary expression of the optical theorem. The optical theorem states that the imaginary part of a forward scattering amplitude equals the total cross-section. In Lorentzian signature, imaginary part and discontinuity are related by Im = Disc. In the celestial CFT, imaginary part means the action of the shadow transform, since shadow symmetry is time reversal T, and T acting on a Lorentzian amplitude picks out its imaginary part through the antiunitary phase.
So the shadow discontinuity mechanism is literally the optical theorem, expressed in the boundary language of the celestial CFT. What Feynman diagrams compute by summing over all internal particle paths is what shadow discontinuities compute by a single analytic operation on a higher-point tree amplitude. The physical content is identical; the computational complexity is incomparable.
Quantifying the Speedup
For four-graviton scattering at one loop, the conventional Feynman diagram approach requires evaluating on the order of 105 diagrams after gauge fixing, ghost inclusion, and simplification. The shadow discontinuity method requires a single discontinuity of the six-point tree amplitude, which is given by a product of three MHV factors from the Parke–Taylor formula. The speedup at one loop is approximately a factor of 106. At two loops the conventional count exceeds 108 diagrams; the shadow method requires a two-fold discontinuity of an eight-point tree. The advantage grows without bound.
The speedup is structural, not computational. It does not depend on smarter computers or better algorithms. It is a consequence of choosing the natural language for the problem. Feynman diagrams are the natural language of perturbation theory around a free field in Minkowski space. Shadow discontinuities are the natural language of the celestial CFT. Quantum gravity lives on the celestial boundary, and problems stated in their natural language tend to be tractable.
Every loop is a boundary traversal. Every virtual particle is a shadow. The optical theorem, expressed in the language of the celestial CFT, is the statement that unitarity at the boundary produces the loop structure in the bulk. One discontinuity, one loop. The correspondence is exact.