Investigation of the energetic and structural features of physical adsorption of molecular hydrogen on carbon nanotubes (CNTs) was carried out within the framework of the semiempirical quantum-mechanical method with the use of the original NDDO/sp-spd package in the regime of parallel computation (Original Package, 2009). Various variants of hydrogen adsorption relative to nanotubes were considered, namely, the positions of hydrogen molecules at different distances from the nanotube walls in the direction perpendicular to the central axis and along it. Hydrogen adsorption inside and outside open carbon nanotubes in different positions relative to the surface of nanotubes was considered. Both single-wall and double-wall carbon nanotubes were considered in the work. It is shown for the former that the energy of physical adsorption of a hydrogen molecule inside nanotubes is higher than the energy of adsorption on the outside of the tubes. It turned out for the latter that the adsorption of hydrogen between the layers of the tubes is energetically more beneficial. In the case of single-layer carbon nanotubes, the dependence of the hydrogen adsorption energy on the diameter of the tube and its chirality was considered. The "armchair" and "zig-zag" types of nanotubes were used. Calculations showed that carbon nanotubes of different chirality but similar in diameter virtually identically adsorb hydrogen, but with increase in the diameter of a nanotube hydrogen adsorption becomes less advantageous energetically. Moreover, adsorption of hydrogen in nanotubes containing nanoinclusions of some transition metals was considered. These metals can naturally and partially fill nanotubes in the course of their catalytic synthesis where they are used as catalysts. In particular, it is shown that nanoclusters of iron and cobalt inside single-layer and multilayer carbon nanotubes activate them increasing considerably the energy of hydrogen adsorption inside nanotubes modified by the metals. The obtained results of modeling agree well with the well-known experimental data (Dillon et al., 1997; Li et al., 2001; Ye et al., 1999; Liu et al., 1999). © 2010 by Begell House, Inc.