Laminar Boundary Layer Mass Transport‐Guided Machine Learning for Precision Density Regulation of Horizontal Carbon Nanotube Arrays

ABSTRACT

Horizontally aligned carbon nanotube (HACNT) arrays are essential building blocks for next‐generation carbon electronics. However, precise control over the density of HACNT remains a formidable bottleneck due to the complex, nonlinear coupling of multi‐physical parameters. To address this challenge, we propose a machine learning (ML) workflow that incorporates boundary layer fluid dynamics (BLFD). Conventional high‐dimensional growth parameters are condensed into a concise feature space dominated by two physics‐informed descriptors: mass transfer coefficient ( K _m ) and residence time ( t ). This kinetics‐informed feature set significantly improves HACNT density prediction, increasing the coefficient of determination (R²) from 0.68 to 0.84.  Consequently, 4200 virtual screenings map a global kinetic phase diagram, revealing an extremely narrow growth window for high‐density HACNT that demands sufficient yet moderate carbon precursor supply combined with appropriate t . Subsequent validation experiments confirm the reliability of BLFD‐informed ML model in HACNT density prediction and identifies the failure mechanisms, and demonstrates the capability to guide targeted growth of array with 120 tubes/µm. By uncovering the underlying kinetic mechanisms rather than merely fitting data, this BLFD‐ML fusion framework would contribute to the HACNT synthetic paradigm shifting from empirical trial‐and‐error to a rational design process governed by fluid dynamics and reaction kinetics.