Excess polymer in single-walled carbon nanotube thin-film transistors: Does it really need to be removed prior to fabrication?

Single-walled carbon nanotubes (SWNTs) are promising nanomaterials for solution-processed organic electronic devices. As-synthesized nanotubes are insoluble and comprised of a mixture of metallic and semiconducting SWNTs (scSWNTs), necessitating dispersal and purification before integration into devices. Refinement of conjugated polymer extraction techniques has facilitated the isolation of highly soluble scSWNTs in a reproducible and scalable manner. The availability of ultrapure scSWNT materials has advanced the production of thin-film transistors (TFTs) that outperform most organic semiconducting small molecules and polymers. However, the realization of commercial TFT applications has not yet been achieved due to the prohibitive time and materials cost associated with purifying scSWNTs. Protocols for supramolecular functionalization of scSWNTs with conjugated polymers typically involve three steps: dispersion, removal of SWNT impurities, and removal of excess polymer through filtration or centrifugation. While the first two steps are facile and scalable, removal of excess polymer is laborious and resource-intensive, but viewed as essential for preparing high-performing TFTs.
We performed the first systematic investigation of how the presence of excess polymer during device fabrication affects the performance of SWNT TFTs. scSWNTs were dispersed with two different poly(fluorene)s, with TFTs fabricated from dispersions with varying ratios of excess sorting polymer. Raman spectroscopy experiments proved that for most of the samples a simple rinsing step was sufficient to remove all unbound polymer from the polymer-sorted scSWNT films. The volume of solvent required for this rinsing step was substantially lower than that required for filtration or centrifugation, and is amenable to many TFT recipes. We found that at higher SWNT concentrations the presence of excess polymer during device fabrication resulted in moderately improved TFT performance, with preeminent devices achieved from the original dispersions (before removal of excess polymer). Analysis of AFM images of polymer-sorted scSWNT films with a novel Machine Learning algorithm showed that improved device performance could be attributed to more even scSWNT films and reduced bundling. Our results have important implications in the development of SWNT inks for commercial printing applications, as the presence of excess polymer could increase the stability of the inks and improve viscosity.