The parameters of many-body potentials for fcc and hcp transition metals, based on the second-moment approximation of a tight-binding Hamiltonian, have been systematically evaluated. The potential scheme, cast in analytical form, allows us to reproduce correctly the thermal behavior of transition metals making use of a small set of adjustable parameters. The large cutoff, which extends the range of the interactions up to the fifth-neighbor distance, ensures good quantitative agreement with the experimental data up to temperatures close to the melting point. The ability of the potentials to describe real systems has been checked by calculating point-defect properties, lattice dynamics, and finite-temperature behavior, and by comparing the results with other potential schemes. Application of this scheme to bcc transition metals has proved unsuccessful. Examples of derivation of many-body potentials for a few transition-metal alloys with cubic structure are also reported.

Thermal conductivity ($\kappa$) of isolated carbon nanotubes (CNTs) is higher than the $\kappa$ of diamond; however, in this Letter we show that the $\kappa$ of a packed bed of three-dimensional random networks of single and multiwall CNTs is smaller than that of thermally insulating amorphous polymers. The thermoelectric power ($\Sigma$) of the random network of CNTs was also measured. The $\Sigma$ of a single wall nanotube network is very similar to that of isolated nanotubes and, in contrast with $\kappa$, $\Sigma$ shows a strong dependence on the tube diameter.

We use molecular dynamics simulations to study the thermal transport properties of a range of poor to good thermal conductors by a method in which two portions are delimited and heated at two different temperatures before the approach-to-equilibrium in the whole structure is monitored. The numerical results are compared to the corresponding solution of the heat equation. Based on this comparison, the observed exponential decay of the temperature difference is interpreted and used to extract the thermal conductivity of homogeneous materials. The method is first applied to bulk silicon and an excellent agreement with previous calculations is obtained. Finally, we predict the thermal conductivity of germanium and α-quartz.

By isolating the process of dislocation emission from a crack tip under an applied tensile stress, we extract from a molecular dynamics simulation the atomic-level displacement and stress fields on the activated slip plane before and after the nucleation event. The stress-displacement relations so obtained provide a direct link with recent continuum descriptions of brittle versus ductile behavior in crack propagation. Crack-tip shielding by the emitted dislocations is demonstrated, as is the role of surface steps in dislocation nucleation and crack-tip blunting.

We investigate the physical meaning of the intrinsic crack resistance in the Griffith theory of brittle fracture by means of atomic-scale simulations. By taking cubic SiC as a typical brittle material, we show that the widely accepted identification of intrinsic crack resistance with the free surface energy underestimates the energy-release rate. The strain dependence of the Young modulus and surface energy, as well as allowance for lattice trapping, improve the estimate of the crack resistance. In the smallest scale limit, crack resistance can be fitted by an empirical elastoplastic model.

Stretching experiments on single molecules of arbitrary length opened the way for studying the statistical mechanics of small systems. In many cases in which the thermodynamic limit is not satisfied, different macroscopic boundary conditions, corresponding to different statistical mechanics ensembles, yield different force-displacement curves. We formulate analytical expressions and develop Monte Carlo simulations to quantitatively evaluate the difference between the Helmholtz and the Gibbs ensembles for a wide range of polymer models of biological relevance. We consider generalizations of the freely jointed chain and of the worm-like chain models with extensible bonds. In all cases we show that the convergence to the thermodynamic limit upon increasing contour length is described by a suitable power law and a specific scaling exponent, characteristic of each model.