RESEARCH INTERESTS

My research expertise is in computational mechanics, materials modeling & design, and machine learning. During my master's study at NTU, I worked on problems in the area of solid mechanics, structural dynamics, and signal processing. During my Ph.D. study at MIT, I focused on understanding the mechanics of deformation and failure of materials at the atomic level. More specifically, I investigated the structure-property relationships of various nanomaterials and biomaterials using atomistic modeling tools including molecular dynamics (MD) and density functional theory (DFT). In 2016, I was deeply inspired by the success of Google DeepMind's AlphaGo program and started to develop machine learning approaches for solving computational mechanics problems. I have published 5 peer-reviewed papers on materials discovery and design using machine learning. I envision that rapid advances in machine learning and artificial intelligence will help solve inverse design problems as well as reduce the computational cost of multiscale modeling and overcome its time-scale and length-scale barriers.

 

I greatly appreciate the financial support that I have received for my study and research. During my Ph.D. study at MIT, I received financial support via a fellowship from the Taiwanese Government and funding from CRP Henri Tudor in the framework of the BioNanotechnology project. During my postdoctoral training at MIT, I received research funding from the Office of Naval Research (ONR) and the Multidisciplinary University Research Initiative (MURI). During my postdoctoral training at UC Berkeley, I am fully funded by the Department of Mechanical Engineering at UC Berkeley.

January 22, 2020

Abstract: In recent years, machine learning (ML) techniques are seen to be promising tools to discover and design novel materials. However, the lack of robust inverse design approaches to identify promising candidate materials without exploring the entire design space causes a fundamental bottleneck. A general‐purpose inverse design approach is presented using generative inverse design networks. This ML‐based inverse design approach uses backpropagation to calculate the analytical gradients of an objective function with respect to design variables. This inverse design approach is capable of overcoming local minima traps by using backpropagation to provide rapid calculations of gradient information and running millions of optimizations with different initial values. Furthermore, an active l...

January 22, 2020

Abstract: Machine learning (ML) has been perceived as a promising tool for the design and discovery of novel materials for a broad range of applications. In this prospective paper, we summarize recent progress in the applications of ML to composite materials modeling and design. An overview of how different types of ML algorithms can be applied to accelerate composite research is presented. This framework is envisioned to revolutionize approaches to design and optimize composites for the next generation of materials with unprecedented properties.

Full paper: CT Chen and GX Gu, Machine learning for composite materials, MRS Communications, 2019, DOI: 10.1557/mrc.2019.32

January 22, 2020

Abstract: Nature assembles a range of biological composites with remarkable mechanical properties despite being composed of relatively weak polymeric and ceramic components. However, the architectures of biomaterials cannot be considered as optimal designs for engineering applications since biomaterials are constantly evolving for multiple functions beyond carrying external loading. Here, it is aimed to develop an intelligent approach to design superior composites from scratch—starting from constituent materials. A systematic computational investigation of the effect of constituent materials (assumed to be perfectly brittle) on the behavior of composites using an integrated approach combining finite element method, molecular dynamics, and machine learning (ML) is reported. It is demonstrat...

May 2, 2019

Abstract: We report a comprehensive ab initio structural investigation of more than 43000 probable molecular structures of polydopamine (PDA) and eumelanin in various oxidation states. With the aid of a computational approach including a brute-force algorithmic generation of chemical isomers and density functional theory, all probable oxidized 5,6-dihydroxyindole (DHI) oligomers, ranging from dimers to tetramers, have been systematically generated and evaluated. We identify a set of the most stable molecular structures of PDA and eumelanin which represent the chemically diverse nature of these materials. Results show that more planar molecular structures have a tendency to be more stable. We also observe that, in some cases, forming cyclic molecular structures could reduce the energy of a...

May 2, 2019

Abstract: Biomimicry, adapting and implementing nature's designs provides an adequate first-order solution to achieving superior mechanical properties. However, the design space is too vast even using biomimetic designs as prototypes for optimization. Here, we propose a new approach to design hierarchical materials using machine learning, trained with a database of hundreds of thousands of structures from finite element analysis, together with a self-learning algorithm for discovering high-performing materials where inferior designs are phased out for superior candidates. Results show that our approach can create microstructural patterns that lead to tougher and stronger materials, which are validated through additive manufacturing and testing. We further show that machine learning can be...

November 14, 2017

Abstract: Composites are widely used to create tunable materials to achieve superior mechanical properties. Brittle materials fail catastrophically in the presence of cracks. Incorporating softer constituents into brittle materials can alleviate stress concentration, leading to tougher and stronger composites. However, searching for the optimal designs of composites is extremely challenging due to the astronomical number of possible material and geometry combinations. Here, we apply machine learning to a composite system and demonstrate its capacity to accurately and efficiently predict mechanical properties including toughness and strength. The method we used incorporates machine learning techniques to generate optimal designs with orders of magnitude better than the mean properties of th...

October 7, 2017

Abstract: Graphene and other two-dimensional materials have unique physical and chemical properties of broad relevance. It has been suggested that the transformation of these atomically planar materials to three-dimensional (3D) geometries by bending, wrinkling, or folding could significantly alter their properties and lead to novel structures and devices with compact form factors, but strategies to enable this shape change remain limited. We report a benign thermally responsive method to fold and unfold monolayer graphene into predesigned, ordered 3D structures. The methodology involves the surface functionalization of monolayer graphene using ultrathin noncovalently bonded mussel-inspired polydopamine and thermoresponsive poly(N-isopropylacrylamide) brushes. The functionalized graphene i...

May 10, 2017

Abstract: Inspired by the hierarchical structure of nacre and the robust adhesive ability of mussel threads, graphene oxide–polydopamine (GO–PDA) nanocomposites are designed and synthesized to achieve enhanced mechanical properties and to provide additional functionalities. Here we report a joint experimental/computational investigation of GO–PDA nanocomposites, proposing a probable chemical reduction mechanism of PDA to convert GO to reduced GO (rGO), which helps increase the electrical conductivity. The most stable chemical connection between PDA and GO is also proposed. Our artificial nacre-like GO–PDA nanocomposites are shown to have higher tensile strength and toughness compared to natural nacre. The pulling tests conducted by molecular dynamics simulations, which are supported by our...

November 3, 2016

Abstract: A set of computational methods that contains a brute-force algorithmic generation of chemical isomers, molecular dynamics (MD) simulations, and density functional theory (DFT) calculations is reported and applied to investigate nearly 3000 probable molecular structures of polydopamine (PDA) and eumelanin. All probable early-polymerized 5,6-dihydroxyindole (DHI) oligomers, ranging from dimers to tetramers, have been systematically analyzed to find the most stable geometry connections as well as to propose a set of molecular models that represents the chemically diverse nature of PDA and eumelanin. Our results indicate that more planar oligomers have a tendency to be more stable. This finding is in good agreement with recent experimental observations, which suggested that PDA and e...

May 23, 2014

Abstract: Eumelanin is a ubiquitous biological pigment, and the origin of its broadband absorption spectrum has long been a topic of scientific debate. Here, we report a first-principles computational investigation to explain its broadband absorption feature. These computations are complemented by experimental results showing a broadening of the absorption spectra of dopamine solutions upon their oxidation. We consider a variety of eumelanin molecular structures supported by experiments or theoretical studies, and calculate the absorption spectra with proper account of the excitonic couplings based on the Frenkel exciton model. The interplay of geometric order and disorder of eumelanin aggregate structures broadens the absorption spectrum and gives rise to a relative enhancement of absorpt...

Please reload

© 2017 by Chun-Teh Chen

  • LinkedIn Social Icon