I would not do it with Gaussian. Gaussian is primarily made for bond breaking/forming in not-so-large molecules. Also, dispersion interactions are not the strong point in Gaussian. Optimization of nanotubes structures take some long time.
If I were you, I would go for molecular dynamics routines which are mainly used for proteins. Specifically, I would try NAMD o LAMMPS. Then you can 'draw' the output in VMD. All those are free software.