An Efficient Descriptor for Tubular LDPE Melt Strength via Physics-Guided Learning and Diffusion-Based Inverse Design

Melt strength (MS) is a key parameter governing the processability and application of low-density polyethylene (LDPE), dictated by a complex interplay of molecular weight distribution (MWD), long-chain branching (LCB), and high-molecular-weight content—features poorly captured by conventional metrics like weight average molecular weight (Mw). This paper addresses the challenge by developing a physics-guided machine learning framework that constructs a list of molecular descriptors directly from full-spectrum MWD data using modern multi-detector high-temperature gel permeation chromatography (MD-GPC), explicitly aligned with the polymer entanglement fundamentals underlying melt strength. The descriptor enables accurate prediction and interpretable disentanglement of topological contributions. The paper further introduced a diffusion-based generative model for inverse design: given a target melt strength, the model proposed candidate molecular architectures and validated them through internal consistency with the learned structure–property relationships, offering a rational and data-driven strategy for the precise engineering of LDPE molecular structures, thus providing considerable flexibility for downstream engineering decisions. Practitioners may further refine or screen these candidates in light of real-world factors—including polymerization processing constraints, cost considerations for resin selection and blending, and manufacturing operation windows—thereby tailoring LDPE molecular structures/products to specific application requirements with high precision.