Thermal boundary resistance at graphene-pentacene interface explored by a data-intensive approach

Area of Science:

• Materials Science
• Computational Physics
• Artificial Intelligence

Background:

• Interfacial thermal transport is critical for thermal management in graphene-pentacene organic electronics.
• Machine learning (ML) techniques, particularly artificial neural networks (ANNs), excel at modeling complex correlations in large datasets for material property predictions.
• Accurate prediction of thermal boundary resistance (TBR) is essential for designing efficient organic electronic devices.

Purpose of the Study:

• To comprehensively investigate the TBR between graphene and pentacene using classical molecular dynamics simulations and ML.
• To develop and optimize an ANN model for predicting TBR in graphene-pentacene systems.
• To provide guidelines for the thermal design of graphene-pentacene based electronic devices.

Main Methods:

• Classical molecular dynamics simulations were employed to calculate TBR values.
• Artificial neural network (ANN) models were trained to predict TBR based on simulation data.
• Hyperparameter tuning, including network layers and neuron count, was performed to optimize ANN performance.

Main Results:

• TBR values at 300 K were determined along the a, b, and c directions of pentacene: 5.19 ± 0.18 × 10⁻⁸, 3.66 ± 0.36 × 10⁻⁸, and 5.03 ± 0.14 × 10⁻⁸ m²K/W, respectively.
• An optimal two-layer ANN with 40 neurons per layer achieved a normalized mean square error loss of 7.04 × 10⁻⁴.
• The study identified key hyperparameters for accurate TBR prediction using ANNs.

Conclusions:

• The combination of molecular dynamics and ML provides an effective approach for investigating interfacial thermal transport.
• The developed ANN model accurately predicts TBR in graphene-pentacene interfaces.
• These findings offer valuable insights for the thermal design and development of advanced organic electronic devices.