Abstract
Machine learning has proven to be technically flexible in recent years, which allows it to be successfully implemented in problems in various areas of knowledge. Carbon nanotubes have been studied to describe their properties or predict possible material responses during their synthesis process or in different conditions and environments. In this review, we analyze the machine learning approaches used in modeling physical properties in carbon nanotubes. This work reveals a remarkable match between the amount of experimental data, the number of parameters, and the algorithms used to model uncontrolled physical properties exhibited by carbon nanotubes. The importance of artificial neural networks, support vector machines, decision trees, random forests, and K-Nearest Neighbors is highlighted, mainly in analyzing these nanostructures. The evaluation of mechanical, thermal, electrical, and electronic properties of carbon nanotubes has been reported. Regarding the thermal, electrical, and electronic properties, it is still necessary to complement the molecular dynamics and density functional theory results, respectively, with machine learning. Mechanical properties present an open line of research regarding vibrational properties, where chiral geometric parameters are used to study the vibrational response of carbon nanotubes; therefore, more accurate estimates are required to predict these frequencies. There is conclusive evidence that there is a relationship between detecting defects in carbon nanotubes and the number of iterations required to describe thermionic and vibrational properties using machine learning. An understanding of the vibratory behavior in these nanomaterials would be helpful in the development of nanosensors. Finally, using simulation models and machine learning would help reduce cost and experimentation time studying these properties.
NSTL
Elsevier