OpenAI "Single-minus graviton tree amplitudes are nonzero" preprint announcement

TL;DR

OpenAI released a preprint titled Single-minus graviton tree amplitudes are nonzero, showing that a class of graviton scattering amplitudes previously thought to vanish can be non‑zero in a half‑collinear kinematic regime, and that the derivation was largely assisted by the AI model GPT‑5.2 Pro.

Introduction – New non‑zero graviton amplitudes

The paper demonstrates that single‑minus graviton tree amplitudes—configurations with one negative‑helicity graviton and all others positive—do not vanish when particle momenta satisfy a half‑collinear alignment. This overturns standard textbook arguments that predict a zero result for generic momenta and provides explicit distributional formulas for the amplitudes.

Physical significance of single‑minus amplitudes

Single-minus amplitudes encode a minimal helicity violation pattern. Their existence reveals hidden structure in quantum gravity, specifically an infinite‑dimensional “w‑(1+∞)” symmetry first identified by Penrose. Realizing this symmetry at the level of graviton scattering suggests a pathway toward a quantum description of Einstein’s theory.

Technical core – Derivation and formulas

The authors derived the amplitudes using:

  1. Recursion relations that build n‑particle amplitudes from lower‑point building blocks.
  2. Symmetry constraints that enforce the w‑(1+∞) algebra.
  3. The directed matrix‑tree theorem, a combinatorial tool introduced by GPT‑5.2 Pro, which streamlined the construction of the graviton amplitudes.

The resulting expressions are distributional objects supported on the half‑collinear region of momentum space. Analytic checks confirm that the formulas reduce correctly to known limits (e.g., soft‑graviton theorems) and respect all required gauge invariances.

AI‑assisted discovery – Role of GPT‑5.2 Pro

After the earlier gluon result was fed to GPT‑5.2 Pro, the model generated the gravitational analog using the directed matrix‑tree theorem and drafted an initial manuscript. Human authors then verified the derivation analytically, ensured consistency with Penrose’s symmetry, and prepared the final write‑up. The project illustrates a shift in research workflow: hypothesis generation was rapid, while the bulk of effort shifted to verification and exposition.

"For this project, much of the time elapsed from the previous gluon result was spent confirming derivations, checking consistency, and preparing formal write‑ups rather than generating initial conjectures." – OpenAI blog post

Verification and consistency checks

The derived amplitudes were:

  • Analytically verified against known tree‑level limits.
  • Cross‑checked with the infinite‑dimensional symmetry constraints.
  • Confirmed to be compatible with the half‑collinear kinematic regime, where the standard vanishing argument fails.

These checks ensure that the AI‑generated results meet conventional standards of mathematical rigor.

Broader implications for quantum gravity research

The discovery that single‑minus graviton amplitudes can be non‑zero expands the known landscape of scattering processes in quantum gravity. It provides a concrete example of how hidden symmetries manifest at the amplitude level and offers a new testing ground for conjectures about the quantum structure of spacetime.

Future directions

OpenAI indicates ongoing work to extend these methods to other helicity configurations and higher‑loop orders. The broader agenda is to assess how AI‑assisted reasoning can accelerate theoretical physics while preserving the community’s standards for verification.

References

Sources