Google AI aids cosmic string radiation solution

Michael P. Brenner, Vincent Cohen-Addad and David P. Woodruff described an AI-assisted solution to a long-standing theoretical-physics calculation in a paper dated March 6, 2026. The paper, posted on arXiv and listed by Google Research, says a neuro-symbolic system combining Google’s Gemini Deep Think model, tree search and automated numerical feedback derived exact analytical solutions for the power spectrum of gravitational radiation emitted by cosmic strings. The result centers on a mathematical object the authors call a “core integral” for arbitrary loop geometries. (arxiv.org) In the paper, the authors say earlier AI-assisted work had produced only partial asymptotic solutions, while the new system generated a broader set of exact methods and an asymptotic result that matches numerical calculations. ### What exactly did the researchers say the AI solved? The March 6 paper says the system solved an open analytical problem tied to the gravitational-radiation power spectrum of cosmic strings. (arxiv.org) Cosmic strings are hypothetical defects in spacetime studied in high-energy physics and cosmology, and the paper focuses on radiation from string loops at different harmonics. The authors write that the agent evaluated the integral \(I(N,\alpha)\), which governs the power emitted by a Garfinkle-Vachaspati string, for arbitrary loop geometries. (arxiv.org) They say singularities in the integrand made standard numerical integration unstable and made conventional analytical expansion difficult. ### How did Google’s system approach the problem? Google Research’s publication page says the method combined the Gemini Deep Think large language model with a systematic tree-search framework and automated numerical feedback. (arxiv.org) The paper presents that setup as a neuro-symbolic system, meaning it paired a neural model’s pattern-finding and proposal generation with explicit symbolic mathematical reasoning and checks. The authors say they included prompts, search constraints and intermittent feedback loops to document how the system worked. They write that the approach was meant to show not only the mathematical result but also the procedure that led to it. ### What was new in the math? The paper says the agent found six different analytical methods. The authors describe the “most elegant” one as an expansion of the kernel in Gegenbauer polynomials, which they say absorbs the singularities in a natural way. (research.google) The same paper says those methods produced a large-\(N\) asymptotic result that agrees with numerical results and connects to continuous Feynman parameterization in quantum field theory. That connection is the authors’ characterization in the paper, and they present the derivations in the preprint. (arxiv.org) ### Why were cosmic strings part of the setup now? The arXiv paper says interest in cosmic strings as sources of gravitational radiation has increased following pulsar-timing-array observations of a stochastic background. (arxiv.org) The authors cite that renewed interest as part of the reason the calculation matters in current theoretical work. The paper does not claim a direct observational discovery of cosmic strings. It reports a mathematical and computational advance in a model used to calculate the radiation spectrum those objects would emit if they exist. (arxiv.org) ### Who is behind the work, and where can readers check it? The author line lists Michael P. Brenner, Vincent Cohen-Addad and David P. Woodruff. The affiliations in the arXiv version include Google Research, Harvard University’s School of Engineering and Applied Sciences, and Carnegie Mellon University’s School of Computer Science. (arxiv.org) Google Research hosts the publication page, and arXiv carries the preprint under the title “Solving an Open Problem in Theoretical Physics using AI-Assisted Discovery.” As of mid-May 2026, those two postings are the primary public records for the result and its methods. (arxiv.org) (research.google)