Multiscale, multiphase numerical modeling is used to explain the mechanisms of effective control of chirality distributions of single-walled carbon nanotubes in direct plasma growth and suggest effective approaches to further improvement. The model includes an unprecedented combination of the plasma sheath, ion/radical transport, species creation/loss, plasma-surface interaction, heat transfer, surface/bulk diffusion, graphene layer nucleation, and bending/lift-off modules. It is shown that the constructive interplay between the plasma and the Gibbs-Thomson effect can lead to the effective nucleation and lift-off of small graphene layers on small metal catalyst nanoparticles. As a result, much thinner nanotubes with narrower chirality distributions can nucleate at much lower process temperatures and pressures compared to thermal CVD. This approach is validated by a host of experimental results, substantially reduces the amounts of energy and atomic matter required for the nanotube growth, and can be extended to other nanoscale structures and materials systems, thereby nearing the ultimate goal of energy- and matter-efficient nanotechnology.