OpenAI preprint reports nonzero “single-minus” gluon amplitudes, with GPT‑5.2 assisting parts of the derivation
A new OpenAI preprint claims certain single-minus gluon amplitudes are nonzero, highlighting an AI-assisted derivation workflow.
OpenAI and academic collaborators have released an arXiv preprint claiming that a well-known ‘zero’ result in gluon scattering amplitudes is not as universal as textbooks suggest.
## What’s new
The paper focuses on *tree-level* scattering amplitudes where **one gluon has negative helicity** and the remaining **n−1 gluons have positive helicity** (often referred to as a “single-minus” configuration). Standard arguments treat this amplitude as zero for *generic* momenta.
The authors identify a **half-collinear slice of momentum space** where those assumptions break down. In that regime, the amplitude **does not vanish**, and the paper derives explicit expressions.
## Where AI fit in
According to the authors:
- Human derivations produced complex expressions for small n (up to n=6).
- **GPT‑5.2 Pro** simplified those expressions and proposed a pattern for all n.
- A scaffolded internal GPT‑5.2 run reportedly produced a formal proof and checks against known constraints (recursion relations and soft theorems).
## Why it matters
Even niche-looking amplitude identities often unlock broader structure in quantum field theory. If the result holds up to peer review, it could:
- reopen work on degenerate kinematic regimes that were previously ignored,
- inform related calculations for **gravitons**, and
- serve as a concrete template for *AI-assisted* mathematical physics workflows (conjecture → simplification → proof → independent checks).
## What to watch next
The paper is currently a preprint and the authors invite community feedback while submitting for publication.
**Source:** OpenAI (link below), with the full preprint on arXiv: https://arxiv.org/abs/2602.12176
Source: OpenAI