Abstract
Carbon nanotubes (CNTs) have gained a significant amount of research interest for use in thermal management applications. Measurements have shown an exceptionally high intrinsic thermal conductivity, k, of individual CNTs. However, when the CNTs are formed into structures or incorporated into a polymer matrix, the collective thermal conductivity through the CNT network or nanocomposite material is found to be orders of magnitude lower than the conductivity of the constituent CNTs. Two major factors may be responsible for the low conductivity of the CNT structures: the reduction of the intrinsic conductivity of individual CNTs due to inter-tube interactions, and low thermal conductance at CNT-CNT contacts that may be sensitive to the surrounding environment.
To investigate the effects of local structure on the intrinsic conductivity of individual CNTs and the thermal conductance across CNT-CNT interfaces, non-equilibrium molecular dynamics (MD) simulations are performed for different parameters of CNT junctions (e.g., contact area, contact angle, etc.) and local structural environments characteristic of CNT network materials (e.g., CNT length, distance to nearest neighboring junction, etc.). The results of MD simulations suggest that (1) contrary to the widespread notion of strongly reduced conductivity of individual CNTs in bundles as compared to the conductivity of isolated CNTs, the van der Waals interactions between defect-free well-aligned CNTs in a bundle have negligible effect on the intrinsic conductivity of the CNTs, (2) the effect of neighboring junctions on the conductance at CNT-CNT junctions is weak and only present when the junction separation distance is within the range of direct van der Waals interactions, (3) the conductance through the overlap region between neighboring parallel CNTs is linearly dependent on the length of the overlap for CNTs and CNT-CNT overlaps longer than several tens of nanometers, and (4) the linear dependence of conductance on the overlap area is found to break down for non-parallel configurations with small contact areas.
In addition, the effect of variation in parameters of MD simulations (e.g., interatomic potential, thermal bath region length, etc.) on the prediction of CNT thermal properties is systematically investigated. Discrepancy in non-equilibrium MD predictions of CNT conductivity found in the literature is shown to be partially caused by erroneous assumptions about how conductivity is affected by the location and length of the regions where the thermal flux is imparted. A more accurate convention for defining the configuration of the computational cell is presented. Based on the results of the MD simulations, a general description of the conductance at CNT-CNT contacts is designed for an arbitrary configuration. Agreement in predicted values of CNT-CNT conductance between this general model and MD simulations is demonstrated. Overall, the results presented here help elucidate the sensitivity of the intrinsic CNT conductivity and CNT-CNT conductance to structural parameters which are relevant for the optimization of the thermal performance of CNT materials.
