(220a) Investigating Ergodicity-Broken Rotational Dynamics of SWCNT in Hexagonally Packed Colloidal Pores Via Machine Learning

The diffusion of stiff rods in porous media is not completely understood. Single-molecule imaging of single-walled carbon nanotubes (SWCNTs) in agarose clarified stiff rod diffusion in gels and biological tissues (low volume fraction of flexible obstacles). However, diffusion in porous media with a high volume fraction of hard obstacles—such as packed columns and porous rock—is still incomplete. For example, the impact of the structure of the porous media (pore size, packing type) on stiff rod diffusion is poorly understood. We fabricated porous media from randomly (r-) and hexagonally (h-) packed micron-size colloids to generate sub-micron diameter pores with varying structure and connectivity. We studied the diffusion of SWCNTs (~2 to 10 µm length, ~14 µm persistence length) within the varying pore structures via near-infrared single-molecule microscopy to track SWCNT location, orientation, and curvature. Via automatic image recognition tools, we generated trajectories from SWCNT motion videos, and analyzed the trajectories via machine learning and statistical tools. We find that SWCNT rotation in h-packed colloid pores conforms to a weak ergodicity broken continuous time random walk (CTRW). In h-packed colloidal pores, diffusing SWCNTs face pores with large orientation differences. To enter these pores, SWCNTs must bend significantly, overcoming a large energy barrier, leading to non-convergent averaged waiting times for pore transitions. R-packed colloidal pores have contiguous pores with smaller relative orientational changes; hence, SWCNTs can transition between pores with smaller bending. In these media, we find that slowed SWCNT rotation under highly bent configurations is rare. This shorter, convergent pore transition time leads to diffusion patterns resembling fractional Brownian motion. These findings will help understanding semiflexible sensor dispersion in rock-like porous media.