A molecular dynamics based artificial intelligence approach for characterizing thermal transport in nanoscale material

Highlights

• Literature establishes motives behind studying thermal property of CNTs.

• An MD-based-AI approach is proposed for studying thermal property of CNTs.

• Integrated approach comprise of molecular dynamics and genetic programming.

• The approach evolves explicit thermal conductivity model of CNTs.

• The approach shows great potential to predict the thermal conductivity of CNTs.

Abstract

A molecular dynamics (MD)-based-artificial intelligence (AI) simulation approach is proposed to investigate thermal transport of carbon nanotubes (CNTs). In this approach, the effect of size, chirality and vacancy defects on the thermal conductivity of CNTs is first analyzed using MD simulation. The data obtained using the MD simulation is then fed into the paradigm of an AI cluster comprising multi-gene genetic programming, which was specifically designed to formulate the explicit relationship of thermal transport of CNT with respect to system size, chirality and vacancy defect concentration. Performance of the proposed model is evaluated against the actual results. We find that our proposed MD-based-AI model is able to model the phenomenon of thermal conductivity of CNTs very well, which can be then used to complement the analytical solution developed by MD simulation. Based on sensitivity and parametric analysis, it was found that length has most dominating influence on thermal conductivity of CNTs.

Introduction

Research in carbon nanotubes (CNT) has attracted significant interest in material science due to its attractive physical and mechanical properties [1], [2]. The exceptional qualities of CNT has been widely studied and investigated to explore its diverse possible applications in real world. These include applications in electric circuits such as CNT-based integrated circuits (ICs), field effect transistors (FET) and solar cells [3], [4], [5]. In addition, CNT is an ideal candidate for potential applications in biomedical, chemical and industrial processes enabled by the use of new CNT materials [6], [7]. These applications of CNT requires a critical understanding of its thermal properties which are key to design future CNT based nano-electronic devices. In addition, the increasing demand to manufacture CNT based nano-components for electronics industry is one of the major incentives to study the thermal properties of CNT.

Numerous studies have been undertaken to predict the thermal properties of CNT using experimental and computational techniques. Cao and Qu [8] studied the effect of size on thermal transport of long CNTs. They found that the mean phonon path of armchair and zigzag CNTs can be analyzed by employing micrometer long CNTs in computational model. Thomas et al. [9] studied the thermal transport in empty and water filled CNTs. They found that the thermal conductivity of water-filled CNTs decreases due to increase in low-frequency acoustic phonon scattering due to interactions with water molecules. Pop et al. [10] conducted laboratory experiments to determine the thermal properties of CNT at room temperature. They determined the thermal conductance of CNT to be approximately 2.4 nW/K. They also found that the thermal conductivity is nearly 3500 W m⁻¹ K⁻¹ at room temperature for a SWNT of length 2.6 μm and diameter 1.7 nm. Kim et al. [11] studied the thermal conductivity of multi-walled CNT using a micro-fabricated suspended device. They observed that the thermal conductivity of multi-walled CNT is more than 3000 W/K m at room temperature. Fujii et al. [12] reported a novel experimental technique to reliably measure the thermal conductivity of a single carbon nanotube using a suspended sample-attached T-type nano-sensor. It was further found from their studies that the temperature dependence of the thermal conductivity for a CNT with a diameter of 16.1 nm appears to have an asymptote near 320 K. The above mentioned literature studies clearly indicate that the thermal transport of CNTs depends on various factors such as system size, chirality, temperature and defects. Hence, understanding the influence of each factor on the thermal transport of CNT is important for optimizing the thermal properties of CNT. One way of optimizing system properties of nanoscale materials is to form an explicit model formulation which can then be used to extract system input variables for desirable material performance [13], [14], [15], [16], [17].

Theoretical studies based on MD simulation has become more popular to study the thermal transport of CNTs when compared to that of laboratory based experiments. This is due to the reason that MD simulation allows rapid reconstruction of defects, altering of chirality and system size [14]. This is useful to understand the influence of system parameters on the thermal properties of CNT. Hence, MD simulation models can be used as a viable alternative compared to time consuming and expensive experiments for monitoring thermal transport at nanoscale. In addition, MD simulation is capable of generating accurate solutions in predicting mechanical and thermal properties of nanoscale system with minimal cost and high rapidity [18], [19], [20], [21]. However, the MD simulation does not provide information on relationship between the input parameters and the generated output. Artificial intelligence (AI) methods can prove to be a useful tool for predicting the relationship between the generated output and input parameters [22], [23], [24]. Additionally, several novel approaches of soft computing methods have been proposed such as hybridizing differential evolution algorithm with receptor editing property of immune system [25], [26], [27], artificial bee colony algorithm with Taguchi’s method [28], [29], differential algorithm with Taguchi’s method [30], cuckoo search algorithm (CS) [31] and immune algorithm with hill climbing local search algorithm [32], [33] to optimize the performance characteristics of the materials. However, evolutionary computing techniques need to be coupled with materials modeling software techniques such as MD simulation in order to be used to predict system properties in nanoscale materials.

Therefore, there is a need to develop an integrated MD based AI simulation technique for modeling the material properties of nanoscale materials such as CNT. The new integrated approach combines powerful advantages of accuracy and low cost of MD simulation with the explicit model formulation of AI techniques. These methods require input training data which can be obtained from the MD simulations which is based on a specific geometry and temperature. Considering input data, the AI technique can then be able to generate meaningful solutions for the complicated problems. Additionally, among the various available AI techniques, an evolutionary approach, namely, multi-gene genetic programming (MGGP) offers the advantage of a fast and cost-effective explicit formulation of a mathematical model based on multiple variables with no existing analytical models [34], [35]. It is to the best of author’s knowledge that limited or no work exists on the application of AI based MD simulation model on evaluating thermal properties of the nanoscale system. Additionally, the potential future applications of CNT in electronics industry require a thorough understanding and investigation of various input parameters on the thermal conductivity of CNT. Hence, the main purpose of the present study is to investigate the thermal conductivity of CNT. The proposed MD based AI approach is employed to investigate the effect of geometry, chirality and vacancy defects on the thermal conductivity of CNT. The functional expression (model) of thermal conductivity with respect to diameter, length and number of defects is obtained. The performance of the proposed model is evaluated against the actual data obtained from literature.