← Latest papers
⚛️ phenomenology

SecRed: A program for sector-wise IBP reduction of Feynman integrals

SecRed is a memory-efficient C++ program that performs sector-wise integration-by-parts reduction of multi-loop Feynman integrals using specialized algorithms for parallel processing and finite-field reconstruction, enabling the solution of large equation systems that exceed single-node memory limits.

Original author: Ruichang Niu, Chao Zeng, Bin Gong, Jianxiong Wang

Published 2026-11-04
📖 5 min read🧠 Deep dive

Original author: Ruichang Niu, Chao Zeng, Bin Gong, Jianxiong Wang

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the author. For technical accuracy, refer to the original paper. Read full disclaimer

To understand the work of the researchers behind the new software SecRed, one must first step into the world of high-energy physics, where scientists attempt to predict the outcomes of particle collisions with extreme precision. When particles smash together at facilities like the Large Hadron Collider, they do not simply bounce off one another; they transform, creating a cascade of new particles that must be calculated to understand the fundamental forces of nature. These calculations rely on mathematical objects known as Feynman integrals, which represent the sum of all possible ways a process can happen. As physicists push for higher precision, they must account for increasingly complex interactions, involving multiple loops of virtual particles and many different paths. The sheer number of these mathematical terms grows explosively, creating a massive system of equations that must be simplified. The goal is to reduce this overwhelming complexity down to a small, manageable set of fundamental building blocks, called master integrals, which can then be solved. Without this reduction, the calculations required to match experimental data would be impossible to complete.

The challenge lies in the sheer scale of the problem. For complex processes involving four loops of interaction, the number of raw equations generated to describe the system can reach into the tens of billions. Traditional methods attempt to solve this entire system at once, but the memory required to hold such a massive collection of numbers exceeds the capacity of even the most powerful single computer. It is like trying to solve a puzzle where the pieces are scattered across a continent; you cannot hold the whole picture in your mind at once. The researchers, led by Ruichang Niu at the Institute of High Energy Physics in China, developed a new program called SecRed to navigate this obstacle. Instead of trying to hold the entire universe of equations in memory simultaneously, SecRed breaks the problem down into smaller, independent sections. It processes these sections one by one, or in parallel groups, ensuring that the computer never has to store more information than it can physically handle.

The core innovation of SecRed is its ability to work sector by sector. In the language of these calculations, a "sector" is a specific grouping of integrals that share certain structural features. SecRed tackles each sector individually, determining which equations are truly necessary and which can be discarded without losing the final answer. It uses a smart, adaptive strategy to decide how many equations to generate for each section. If a section is too large, the program can trim the excess, keeping only the most critical parts while remembering to check later if it cut too deep. This self-correcting approach ensures that the final result remains accurate even as the program aggressively prunes the data to fit within memory limits. Once a sector is processed, the program extracts only the essential, independent equations and passes them down to the next stage, much like a relay race where each runner passes only the baton, not the entire race history.

A significant portion of the computational cost comes from the repetitive nature of the math. The program must solve the same system of equations many times, using different numerical values to reconstruct the final answer. SecRed optimizes this by recording the order in which it eliminates variables just once. It then replays this exact sequence for every new calculation, skipping the time-consuming step of figuring out the order again. This allows the program to focus purely on the arithmetic, speeding up the process significantly. Furthermore, the program includes a memory-aware scheduler that acts as a traffic controller. It monitors the computer's available memory in real time, deciding which tasks to start and which to wait on, ensuring the system never crashes from running out of space. This allows the program to run on a single machine with a fixed memory budget, a feat that was previously impossible for such large problems.

The researchers tested SecRed on two difficult examples: a two-loop five-point family of integrals and a four-loop four-point family. In the first case, the program completed the reduction in just over a minute, using less than seven gigabytes of memory, while other existing programs took significantly longer or required more resources. In the second, more extreme case involving four loops, the total number of equations was so vast that a standard approach would have failed immediately, unable to even assemble the full system in memory. SecRed, however, successfully completed the reduction, finishing in about three and a half hours while staying within a memory limit of roughly 425 gigabytes. The program produced a set of master integrals that matched the results of other independent methods, confirming its accuracy.

The findings demonstrate that by organizing the calculation into smaller, manageable pieces and optimizing the order of operations, it is possible to solve problems that were previously considered too large for a single computer. The program does not rely on manual tuning or human intervention to adjust its settings; it adapts automatically to the specific structure of the problem. This capability is crucial for the future of particle physics, where calculations are becoming increasingly complex. The software is already being used to support calculations for the production of specific particles, such as the J/psi meson, which are vital for testing the Standard Model of particle physics. By making these massive reductions feasible on standard hardware, SecRed opens the door to deeper and more precise investigations into the fundamental laws of the universe, allowing physicists to explore questions that were previously out of reach.

Drowning in papers in your field?

Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.

Try Digest →