Nanotechnology and Lower Scale Phenomena


USACM Technical Thrust Areas

 

Committee: Nanotechnology and Lower Scale Phenomena

Chair: Jaroslaw Knap, U.S. Army Research Laboratory
Vice-Chair: Amartya Banerjee, University of California Los Angeles
Members-at-Large: Nikhil Chandra Admal, University of Illinois Urbana-Champaign,
Susanta Ghosh, Michigan Technological University

 

Webinar Series 

No webinars currently scheduled.


Past Webinars

November 27, 2024; 2pm-3pm CST

Speaker: Prof. Gábor Csányi, University of Cambridge

Title: Foundational models for materials chemistry

Abstract: A new computational task has been defined and solved over the past 15 years for extended material systems: the analytic fitting of the Born-Oppenheimer potential energy surface as a function of nuclear coordinates. The resulting potentials  ("force fields") are reactive, many-body, with evaluation costs that are currently on the order of 0.1-10 ms/atom/cpu core (or about 1-10ms on a powerful GPU), and reach accuracies of a few meV/atom when trained specifically for a given system using iterative or active learning methods. The latest and most successful architectures leverage many-body symmetric descriptions of local geometry and equivariant message passing networks.  Perhaps the most surprising recent result is the stability of models trained on very diverse training sets across the whole periodic table. Our recent discovery is that the MACE-MP-0 model that was trained on just ~150,000 real and hypothetical small inorganic crystals (90% of training set < 70 atoms), is capable of stable molecular dynamics at ambient conditions on any system tested so far - this includes crystals, liquids, surfaces, clusters, molecules, and combinations of all of these. The astounding generalisation performance of such foundation models open the possibility to creating a universally applicable interatomic potential with useful accuracy for materials (especially when fine-tuned with a little bit of domain-specific data), and democratise quantum-accurate large scale molecular simulations by lowering the barrier to entry into the field. Similarly, in the domain of organic chemistry, training just on small molecules and small clusters allows accurate simulation of condensed phase systems, and first principles prediction of quantities such as hydration free energies for the first time.

October 30, 2024; 2pm-3pm CDT

Speaker: Prof. Taher Saif, University of Illinois Urbana-Champaign

Discussant: Prof. Sulin Zhang, The Pennsylvania State University

Title: Living machines

Abstract: Since the discovery of genes, there is a considerable body of knowledge on engineering living cells. It is now possible to envision biohybrid machines and robots with living cells and scaffolds. These machines may self assemble and emerge from complex interactions between the cells and the scaffolds at various hierarchical levels. We will discuss one of such machines - a biohybrid swimmer - that emerges from interactions between muscle cells and neurons. These interactions are mechanical in nature, they are highly non-linear and dynamic. They give rise to long distance cooperation between cells resulting in a phase transition, and the emergence of self-assembled machines. While such machines demonstrate the first milestone achieved in this new field of living robots and materials with unprecedented opportunities, they also highlight the current limitations and gaps in the field. Closing these fundamental gaps will not only pave the way to more complex engineered living systems, but will also provide new insight on biological processes and the life itself. A few key challenges and unanswered questions will be discussed.

Bio: Dr Taher Saif is the Edward William and Jane Marr Gutgsell Professor in the Department of Mechanical Science and Engineering at the University of Illinois at Urbana-Champaign. His current research includes tumor micro environment, mechanics of neurons and cardiac cells, and development of biological machines. His research involves exploration of the underlying mechanism of cell-cell and cell-scaffold interactions, as well as the biophysical processes by which cells remodel their microenvironment. He served as the research lead for biohybrid machines group in the NSF Science and Technology Center, EBICS. He is the recipient of 2020 Engineering Science Medal from the Society of Engineering Science, and the 2018 Warner T. Koiter Medal from American Society of Mechanical Engineers. He has been elected to the National Academy of Engineering of USA in 2024.

September 25, 2024; 2pm-3pm CST

Speaker: Prof. Michele Ceriotti, EPFL

Title: More than physics, more than data: Integrated machine-learning models for materials

Abstract: Machine-learning techniques are often applied to perform "end-to-end" predictions, that is to make a black-box estimate of a property of interest using only a coarse description of the corresponding inputs. In contrast, atomic-scale modeling of matter is most useful when it allows one to gather a mechanistic insight into the microscopic processes that underlie the behavior of molecules and materials.  In this talk I will provide an overview of the progress that has been made combining these two philosophies, using data-driven techniques to build surrogate models of the quantum mechanical behavior of atoms, enabling "bottom-up" simulations that reveal the behavior of matter in realistic conditions with uncompromising accuracy.  I will discuss two ways by which physical-chemical ideas can be integrated into a machine-learning framework.  One way involves using physical priors, such as smoothness or symmetry of the structure-property relations, to inform the mathematical structure of a generic ML approximation. The other entails a deeper level of integration, in which explicit physics-based models and approximations are built into the model architecture.  I will discuss several examples of the application of these ideas, from the calculation of electronic excitations to the design of solid-state electrolyte materials for batteries and high-entropy alloys for catalysis, emphasizing both the accuracy and the interpretability that can be achieved with a hybrid modeling approach, and providing an overview of the exciting research directions that are made available by these new modeling tools. 

Bio: I received my Ph.D. in Physics from ETH Zürich, and spent three years in Oxford as a Junior Research Fellow at Merton College. Since 2013 I have led the laboratory for Computational Science and Modeling, in the institute of Materials at EPFL, focusing on method development for atomistic materials modeling, bridging quantum mechanics, statistical mechanics and machine learning.  If you're curious about my grants and prizes, Google is your friend. If you're curious about stuff I actually care about, I am very proud to have contributed to the development of several open-source software packages, including http://ipi-code.org and http://chemiscope.org, and to have served the atomistic modeling community as an associate editor of the Journal of Chemical Physics, as a moderator of the physics.chem-ph section of the arXiv, and as an editorial board member of Physical Review Materials.

February 28, 2024

Speaker: Gideon Simpson, Drexel University
Discussant: Petr Plechac, University of Delaware

TitleLearning Dynamics with Random Fourier Features

Abstract: A common challenge in modelling, at all scales, is how to “learn” an evolution law from time series data, allowing for prediction.  Classical approaches have involved mathematical modeling with differential equations.  This has been extended to the neural ODEs and PDEs, which incorporate data into fitting generalized “right hand sides” to the evolution equations with neural nets fo various architectures.  Here, we make use of recently developed adaptive random Fourier feature methods to learn near optimal approximations in Fourier space, representing our evolution map, directly, in terms of trigonometric activation functions within a shallow neural network.  This talk will present the method, its theoretical basis, our progress in learning dynamics, and highlight outstanding questions and future directions of activity.  This is joint work with P. Plechac and J. Knap, along with graduate students.

January 31, 2024

Speaker: Surya R. Kalidindi, Georgia Institute of Technology
Discussant: Raymundo Arroyave, Texas A&M University

TitleAccelerated development of materials using high-throughput strategies and AI/ML

Abstract: The dramatic acceleration of the materials innovation cycles is contingent on the development and implementation of high throughput strategies in both experimentation and physics-based simulations, and their seamless integration using the emergent AI/ML (artificial intelligence/machine learning) toolsets. This talk presents recent advances made in the presenter’s research group, including: (i) a novel information gain-driven Bayesian ML framework that identifies the next best step in materials innovation (i.e., the next experiment and/or physics-based simulation to be performed) that maximizes the expected information gain towards a specified target (e.g., optimized combination of material properties, refinement of a material constitutive response), (ii) computationally efficient versatile material structure analyses and statistical quantification tools, (iii) formulation of reduced-order process-structure-property models that enable comprehensive inverse solutions needed in materials design (e.g., identifying specific compositions and/or process histories that will produce a desired combination of material properties), and (iv) high throughput experimental protocols for multi-resolution (spatial resolutions in the range of 50 nm to 500 microns) mechanical characterization of heterogeneous materials in small volumes (e.g., individual constituents in composite material samples, thin coatings or layers in a multilayered sample). These recent advances will be illustrated with case studies.

Bio: Surya Kalidindi is a Regents Professor and Rae S. and Frank H. Neely Chair Professor in the George W. Woodruff School of Mechanical Engineering with joint appointments in the School of Computational Science and Engineering and the School of Materials Science and Engineering at Georgia Institute of Technology, Georgia, USA. Surya’s research efforts have made seminal contributions to the fields of crystal plasticity, microstructure design, high-throughput mechanical characterization and materials informatics. Surya has been elected a Fellow of ASM International, TMS, and ASME. He has also been recognized with the Alexander von Humboldt Research Award, the Vannever Bush Faculty Fellow, and the Khan International Award. He has authored/co-authored 2 books, 8 book chapters, 2 edited volumes, and over 300 archival journal articles. His research currently has a h-index of 91 (as per Google Scholar). Most recently, he has co-founded the new venture-funded start-up Multiscale Technologies, Inc., which offers a commercial SaaS platform for Al/ML driven accelerated materials innovation.

December 6, 2023

Speaker: Timofey Frolov, Lawrence Livermore National Laboratory
Discussant: Fadi Abdeljawad, Lehigh University

Title: Atomistic simulations of grain boundary structure and behavior (not recorded at the request of the speaker)

Abstract: Grain boundaries greatly influence many properties of engineering materials. Accurate prediction of their structure and possible transitions using atomistic modeling are important for strategies that aim to improve properties of materials. Recent years have seen a rapid growth of evidence suggesting that materials interfaces are capable of first-order structural transformations in which the interface properties undergo discontinuous changes. Experiments have linked these transitions to abnormal grain growth in ceramics, activated sintering and liquid metal embrittlement and raised a number of fundamental questions concerning the atomic structures and kinetic properties of these interface phases. The first part of the talk I will describe the state-of-the-art modeling tools we use to predict grain boundary phases and new methodologies to model grain boundary phase transformations in metallic systems.

In the second part of the talk, I will focus on our recent massively parallel atomisticsimulations of bicrystal deformation performed on scale approaching microns using the cross-scale method.  The scale of these simulations allows to model dynamic evolution of dislocation networks in the presence of grain boundaries with atomic resolution to reveal new details about the nature of their interaction. This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.

Bio: Timofey Frolov is a staff scientist at Lawrence Livermore National Laboratory, his research expertise is in thermodynamics and computational modeling of materials. He uses state-of–the-art computational tools to model materials interfaces and interface related phenomena and investigate their impact on materials properties. He received his PhD from George Mason University in 2012 where he worked on developing a thermodynamic theory of solid-solid and solid-liquid interfaces that included effects of non-hydrostatic stresses. After the PhD, he did a postdoc as a Miller Fellow at UC Berkeley and subsequently a postdoc at Lawrence Livermore National Lab, where he primarily worked on grain boundary phase transitions in metallic systems. He is a recipient of a prestigious Mercator Fellowship from the German Science Foundation. This award funds a collaboration with the microscopy group at Max Plank Institute in Germany to couple his modeling efforts with direct experimental observations of GB transitions in alloys. In 2022 he received DOE Early Career Award to work on tungsten alloys for fusion energy applications.

September 27, 2023

Speaker: Kaushik Bhattacharya
Discussant: Miguel Bessa, Brown University

TitleData-driven constitutive relations: Multiscale modeling and experimental inference

The talk addresses the challenge of computing complex phenomena at the scale of applications.  In addition to the universal laws (balance of mass, momenta etc.), these phenomena require a constitutive (closure) relation that describes the behavior of the medium at the scale of applications.    Such behavior can be nonlinear, nonlocal, anisotropic, history dependent etc., and thus impossible to characterize to the desired level by the classical approach of postulating a parametrized relation and fitting the parameters to selected experiments.  The talk describes two broad approaches to using data-driven methods to overcome this challenge.  The first approach is multiscale modeling where one recognizes that the effective behavior at the scale of applications is determined by physics at multiple length and time scales: electronic, atomistic, domains, defects etc.  The data-driven constitutive relation is obtained as a neural approximation is trained using data generated by repeated solution of the small scale problem.  The second approach seeks to infer it from automated experiments that are not amenable to easy inversion.  The talk will describe these approaches, challenges they raise and strategies to overcome them.  The ideas will be illustrated with applications from materials science and geology.

August 30, 2023

Speaker: Krishna Garikipati
Discussants: Harley T. Johnson, University of Illinois Urbana-Champaign and Shailendra P. Joshi, University of Houston

TitleA free energy-based framework for scale bridging in crystalline solids--with some use of machine learning methods

The free energy plays a fundamental role in theories of phase transformations and microstructural evolution in crystalline solids. It encodes the thermodynamic coupling between mechanics and chemistry within continuum descriptions of non-equilibrium materials phenomena. In mechano-chemically interacting materials systems, consideration of compositions, order parameters and strains results in a high-dimensional free energy density function. Since its origins lie in the electronic structure, a rigorous representation of the free energy presents a framework for scale bridging in solids. In this study we have been exploring such a framework, while developing practical machine learning methods to contend with high dimensionality and efficient sampling. We have developed integrable deep neural networks (IDNNs) that are trained to free energy derivative data generated by statistical mechanics simulations. The latter are based on cluster Hamiltonians, themselves trained on density functional theory calculations. The IDNNs can be analytically integrated to recover a free energy density function. We combine the IDNNs with active learning workflows for well-distributed sampling of the free energy derivative data in high-dimensional input spaces. This enables scale bridging between first-principles statistical mechanics and continuum phase field models. As prototypical material systems we focus on applications in Ni-Al alloys and in the battery cathode material: Li$_x$ CoO$_2$.

October 26, 2022

Speaker: Stefanie Reese, RWTH Aachen University
Discussant: Celia Reina, University of Pennsylvania

Title: Data-Driven Mechanics for Elastic and Inelastic Problems Including Uncertainty Quantification
(not recorded at presenter's request)

The data-driven computing paradigm for mechanical systems as proposed in [1] is further extended. Its main principle, to define the solution of boundary value problems explicitly in the material data has several advantages. While generally accepted physical principles such as conservation laws and thermodynamic constraints are enforced, any functional modelling of the material data is avoided. The open-ended process of modelling and calibration as the material data set grows is thus circumvented. Model uncertainties regarding the choice of functions, proper usage within its range of validity and loss of information are thus minimized. One main challenge is the extension to inelastic material behaviours, also included in the talk. The approach of a previous work [2] is extended by material-independent thermodynamic constraints, involving the Helmholtz free energy as first principal data, which may be obtained by increasingly accurate low-scale simulations. Further important aspects are the data search [3] as well as the uncertainty quantification [4]. The talk will be concluded by structural examples.

[1] Kirchdoerfer, T., & Ortiz, M. (2016). Data-driven computational mechanics. Computer Methods in Applied Mechanics and Engineering, 304, 81-101.
[2] Eggersmann, R., Kirchdoerfer, T., Reese, S., Stainier, L., & Ortiz, M. (2019). Model-free data-driven inelasticity. Computer Methods in Applied Mechanics and Engineering, 350, 81-99.
[3] Eggersmann, R., Stainier, L., Ortiz, M., & Reese, S. (2021). Efficient data structures for model-free data-driven computational mechanics. Computer Methods in Applied Mechanics and Engineering, 382, 113855.
[4] Prume, E., Reese, S., Ortiz, M. (2022). Model-free data-driven inference in computational mechanics, arXiv.org, arXiv:2207.06419 [cs.CE]

October 12, 2022

Speaker: Steve WaiChing Sun, Columbia University
Discussant: Amartya Banerjee, University of California, Los Angeles

TitleGeometric Learning for Discovering Complex Material Behaviors of Microstructures

Plasticity models often include a scalar-valued yield function to implicitly represent the boundary between elastic and plastic material states. However, a surface can also be represented explicitly by a manifold of which the tangential space is Euclidean. In this work, we introduce the concept of yielding manifold. This yielding manifold is reconstructed by stitching local multi-resolution patches together. This treatment enables us to construct a highly complex and precise yield envelope by breaking it down into multiple coordinate charts. The global atlas is then built to enforce the consistency of the machine learning generated yield surface. In contrast to data from smooth manifolds, data collected from sensors and numerical simulations are often discrete. As such, we propose to use graph embedding to create latent space inferred from 3D microstructural data represented via weighted graphs. Vectors of this latent space are used as both geometrical descriptors and internal variables of which a decoder can be used to convert predictions in the latent space back as a weight graph. This approach establishes a direct connection between nonlinear dimensional reduced simulations with machine learning constitutive models. Potential applications of geometric reinforcement learning for inverse problems and the design of experiments in the limited data regime will also be discussed. 

August 31, 2022

Speaker: Michael Shields, Johns Hopkins University
Discussant: Anter El-Azab, Purdue University

TitleThe Intersection of Machine Learning & Uncertainty Quantification in Physics-Based Modeling for Materials Systems

Machine Learning (ML) and Uncertainty Quantification (UQ) have gained widespread popularity in the scientific community, to such an extent that ML/UQ seems to appear in some capacity in nearly all modern scientific investigations. This is particularly true in physics-based modeling where machine learning algorithms are being specially designed to adhere to physical principles, such as the popular physics-informed neural networks (PINNs). In this talk, we will discuss the relationship between these two important research areas in the context of physics-based modeling, with an emphasis on simulating materials systems. We will specifically discuss how the two areas complement one another to enhance modeling capability by making the critical distinction between UQ for ML and ML for UQ. In the former case, we will argue that modern ML methods require UQ as an integral component and show recent advances in the learning of uncertainty-aware Bayesian Neural Networks. In the latter case, we will argue that UQ can be viewed as an exercise in ML and will show how modern ML methods (from Hamiltonian Neural Networks to manifold learning) enable UQ in physical systems while, in many cases, remaining constrained by the underlying physics. Applications to materials modeling will be shown ranging from equation of state modeling for warm dense matter to hierarchical multi-scale models for structural mechanics.

July 27, 2022

Speaker: Michael Falk, Johns Hopkins University
Discussant: Anter El-Azab, Purdue University

TitleQuesting for Structural Predictors of Plastic and Failure Response in Glasses

For decades materials scientists, mechanicians and physicists have searched for structural predictors for plastic flow and failure in glasses and other amorphous materials. Due to the lack of any crystalline order in these materials, disorder rules the day on many scales. This makes quantifying the microstructures of amorphous materials difficult. Such quantification is necessary for building predictive theories that can guide materials design and the development of processing to improve mechanical properties. Here I will consider some particular case studies from my own work with collaborators: the use of machine learning to fit a constitutive model to molecular dynamics data, the use of computer modeling to quantify of the local yield surface in two- and three-dimensions, and an attempt to use an equation-free method to harvest simulation data for the quantification of plastic constitutive response in a 3D binary glass. Reflections on these efforts will be discussed in order to consider the prospects for harnessing machine learning for the development of physically interpretable structural characterization and materials response theories.

June 29, 2022

Speaker: Markus Hütter, Eindhoven University of Technology
Discussant: Joerg Rottler, University of British Columbia

Title: Molecular Approach to Plasticity in Polymer Glasses: A Journey

A major benefit of multiscale modeling is that it helps to shed light on constitutive assumptions in macroscopic approaches to mechanics [1]; the stress tensor and the rate of plastic deformation are of particular interest. Nonequilibrium statistical mechanics is a powerful technique in this field, and it has been applied to study the plastic deformation of solids [2,3]. In this presentation, the focus is on multiscale modeling of solids in the glassy state. Structural glasses are particularly interesting (and challenging) for two reasons: they can age in the course of time, and the microscopic carriers of plastic deformation have not been identified to date (in contrast to the well-known dislocations for crystalline materials). Molecular simulations will be used to study glassy materials, in particular polymers. The goal is to establish a fine level of description that is suitable for a subsequent coarse-graining step to the macroscopic continuum level.
When studying polymer glasses on a molecular level, the structural rearrangements related to physical ageing and plastic deformation are rare events, in comparison to the rapid and continuously ongoing molecular vibrations. For studying in an efficient manner these rare events while still keeping molecular detail, a procedure has been proposed that scans the space of molecular configurations specifically for local minima and transitions between them [4]. A key ingredient in this procedure is the calculation of the free energy under appropriate mechanical boundary conditions [5]. In this contribution, we present results of the rare-event sampling for atactic polystyrene - as a prototypical example - below its glass-transition temperature. In the absence of deformation, we obtain rate constants for the minimum-to-minimum transitions extended over 30 orders of magnitude, with well-defined peaks at the time scales corresponding to the subglass relaxations of polystyrene [6]. As deformation is imposed, we observe that the transition states go through an instability and eventually collapse abruptly onto one of the connected local minima; furthermore, we present results about the transition rates as a function of deformation [7]. It will be discussed how these observations will eventually relate to the rate of macroscopic plastic deformation.
Acknowledgment: Part of this work is support by the Dutch Polymer Institute (DPI), projects no. 745ft14 and 820.
REFERENCES
[1] E. van der Giessen, et al., Modelling Simul. Mater. Sci. Eng. 28 (2020) 043001 (61pp).
[2] M. Hütter, T.A. Tervoort, Adv. Appl. Mech. 42 (2008) 253-317.
[3] M. Kooiman, M. Hütter, M.G.D. Geers, J. Mech. Phys. Solids 90 (2016) 77-90.
[4] G.C. Boulougouris, D.N. Theodorou, J. Chem. Phys., 127(8) (2007) 084903.
[5] G.G. Vogiatzis, et al., Comput. Phys. Commun., 249 (2020) 107008.
[6] G.G. Vogiatzis, L.C.A. van Breemen, M. Hütter, J. Phys. Chem. B, 125(26) (2021) 7273-7289.
[7] G.G. Vogiatzis, L.C.A. van Breemen, M. Hütter, J. Phys. Chem. B (submitted, 2022).

May 25, 2022

Speaker: Alejandro Strachan, Purdue University; 

Discussant: Arun Mannodi Kanakkithod, Purdue University

Title: Enhancing Physics-Based Modeling and Extracting Physical Insight from

Data Using Machine Learning 

 

The synergy between principles-based modeling and data science is playing an increasingly important role in materials science and engineering. In addition, there is significant interest in using machine learning (ML) tools to extract physical laws as well as symmetries and associated invariants from data. I will discuss recent progress in my group on machine learning applied to multiscale modeling and in the development of interpretable models that balance accuracy with parsimony.
ML for multiscale modeling. ML interatomic potentials have shown near quantum mechanical accuracy at a fraction of the cost, enabling large-scale atomistic simulations. We developed an iterative approach to address the challenge of generating training data and address the stochastic nature of NN training. I will demonstrate the approach with the development of neural network reactive force fields (NNRF) for energetic materials over a wide range of temperatures and pressures and the phase change material GeSbTe of interest in electronics. The next step of the multiscale ladder is coarse-graining atomistic simulations, and I will show how dimensionality reduction techniques like non-negative matrix factorization and autoencoders can extract physically interpretable collective variables from MD simulations. Specifically, we developed a reduced-order chemical kinetics model with three components and two reactions starting from a 280-dimensional space.
Making workflows and data FAIR. Finally, I will also describe recent developments in nanoHUB, an open cyberinfrastructure for cloud scientific computing, towards making simulation workflows and their data findable, accessible, interoperable, and reusable (FAIR). We introduce Sim2Ls (pronounced sim tools) and the Sim2L Python library that allow developers to create and share end-to-end computational workflows with well-defined and verified inputs and outputs. The Sim2L library makes Sim2Ls, their requirements, and their services discoverable, verifies inputs and outputs, and automatically stores results in a globally accessible simulation cache and results database.

1)Yoo P, Sakano M, Desai S, Islam MM, Liao P, Strachan A. Neural network reactive force field for C, H, N, and O systems. npj Computational Materials. 2021 Jan 22;7(1):1-0.
2)Sakano MN, Hamed A, Kober EM, Grilli N, Hamilton BW, Islam MM, Koslowski M, Strachan A. Unsupervised learning-based multiscale model of thermochemistry in 1, 3, 5-Trinitro-1, 3, 5-triazinane (RDX). The Journal of Physical Chemistry A. 2020 Oct 28;124(44):9141-55.
3)Hunt M, Clark S, Mejia D, Desai S, Strachan A. Sim2Ls: FAIR simulation workflows and data. Plos one. 2022 Mar 10;17(3):e0264492.