Center for Industrial Mathematics and Statistics » REU 2023 – Terahertz Time Domain Spectroscopy for Characterizing Properties of Carbon Nanotube Yarns

Students: Natalie Frey, Dashiell Lipsey, and Anthony Miller

Overview

Industrial Sponsor: Huntsman Advanced Materials

Industrial Liaison: Robert Casoni

Faculty Advisors: Dr. Randy Paffenroth, Mathematical Sciences, WPI and Dr. Lyubov Titova, Physics, WPI

MIRALON® yarns are composed of aligned single and multiple CNT bundles. Unlike conventional carbon fibers, the CNT bundles within the yarns transfer load by van der Waals forces, not covalent bonds, resulting in yarns that are laterally flexible yet stiff and strong, fatigue and impact resistant, and highly thermally and electrically conductive in the axial direction. For quality control and for establishing the correlations between alignment of CNTs within the yarn and its tenacity and other properties, it is imperative to develop a rapid, repeatable, and non-destructive approach to quantify CNT alignment. As CNTs are highly conductive along their length, aligned CNTs absorb (and reflect) THz radiation that is polarized along the CNT direction, and transmit THz radiation polarized perpendicular to the CNT direction. Studying THz transmission as a function of yarn orientation relative to the THz pulse polarization provides a measure of CNT alignment. In addition, modern machine learning (ML) techniques have demonstrated state-of-the-art performance in many data oriented problems and this REU focused on several ML problems related to MIRALON® yarns including:

1. Determining any meaningful relationships between the Tex Measurement System (TMS) and in-line measurements,

2. determining any meaningful relationships between the TMS, in-line measurements as well as any relationships between the standard test data (tex, tenacity and resistivity) and determining their “predictive power”,

3. quantifying the signal versus noise in the measurements and determining any recommendations to reduce measurement noise (such as splines, weighted moving averages, wavelets or others), and

4. determining any meaningful patterns in the TMS and in-line measurement data both within-lot and lot-to-lot.