review of the synergies between computational . (willbeinsertedbytheeditor) review of the...
Post on 22-Mar-2019
Embed Size (px)
Journal of Materials Science manuscript No.(will be inserted by the editor)
Review of the Synergies Between Computational Modelingand Experimental Characterization of Materials Across LengthScales
Rmi Dingreville Richard A. Karnesky Guillaume Puel Jean-HubertSchmitt
Received: 11 September 2015 / Accepted: date
Abstract With the increasing interplay between ex-perimental and computational approaches at multiplelength scales, new research directions are emerging inmaterials science and computational mechanics. Suchcooperative interactions find many applications in thedevelopment, characterization and design of complexmaterial systems. This manuscript provides a broadand comprehensive overview of recent trends where pre-dictive modeling capabilities are developed in conjunc-tion with experiments and advanced characterizationto gain a greater insight into structure-properties re-lationships and study various physical phenomena andmechanisms. The focus of this review is on the intersec-tions of multiscale materials experiments and modelingrelevant to the materials mechanics community. After ageneral discussion on the perspective from various com-munities, the article focuses on the latest experimentaland theoretical opportunities. Emphasis is given to therole of experiments in multiscale models, including in-sights into how computations can be used as discoverytools for materials engineering, rather than to simplysupport experimental work. This is illustrated by exam-ples from several application areas on structural mate-rials. This manuscript ends with a discussion on someproblems and open scientific questions that are being
R. DingrevilleSandia National Laboratories, Albuquerque, NM 87185, USAE-mail: email@example.com
R. KarneskySandia National Laboratories, Livermore, CA 94550, USAE-mail: firstname.lastname@example.org
G. Puel, J.-H. SchmittLaboratoire Mcanique des Sols, Structures et Matriaux(MSSMat), CNRS UMR 8579, Centrale/Suplec, UniversitParis-Saclay, 92290 Chtenay Malabry, France
explored in order to advance this relatively new field ofresearch.
Keywords Coupling experiments and modeling Multiscale modeling Integrated computationalmaterials Engineering (ICME) 3-D microstructurecharacterization Characterization methods
Recent advances in theoretical and numerical meth-ods, coupled with an increase of available high perfor-mance computing resources, have led to the develop-ment of large-scale simulations in both Materials Sci-ence and Mechanical Engineering, with the goal of bet-ter characterization and optimization of materials per-formance. These advanced computational capabilitiesextend over a large span of length scalesranging fromthe atomistic to the continuum scaleand serve as in-vestigative tools to get a greater insight into structure-properties relationships. As these predictive modelingcapabilities become more comprehensive and quanti-tative, a comparable level of details from experimentsand characterization tools is now not only necessaryto validate these multiscale models, but also to mo-tivate further theoretical and computational advances.The present complexity of the multiscale approach leadsalso to the need for new algorithms to bridge lengthscales and for model reduction. Strong coupling be-tween experimental data and numerical simulation isrequired for predicting macroscopic behavior with afine description of local fields (i.e. mechanical, compo-sitional, electrical, etc.). An illustration of such syner-gies is the coupling of experimental tools such as Trans-mission Electron Microscopy (TEM) and synchrotron-based X-ray microscopy that enable the collection ofmicrostructural statistical data comparable to the out-
2 Rmi Dingreville et al.
put of three-dimensional (3-D) Crystal Plasticity FiniteElement Method (CPFEM)-based simulations [1, 2].
As illustrated in Fig. 1, based on a citation reportextracted from ISI Web of Knowledge, there is an in-creasing trend of articles in the fields of both Materi-als Science and Mechanics to report on both multiscalemodeling and experiments. In Fig. 1a, we consider thosethat denote a topic area of Coupling Experiments andModeling. A similar trend is observed for articles con-taining Multiscale Modeling in the title for researchareas limited to the same fields of Materials Scienceand Mechanics. Both progressions in the number ofjournal articles published highlight the increasing cross-pollinations between the modeling community and theexperimental community. While the term coupling isfar too inclusive to extract any real correlation betweenthe emergence of advanced characterization methodolo-gies and growth of research into Coupling Experimentsand Modeling, it is interesting to put in perspectivethis increase of publications with respect to the succes-sive introduction of novel experimental characterizationmethodologies such as Scanning Electron Microscopy(SEM) , Atom Probe Tomography (APT) , SEM-based Electron Back Scattered Diffraction (EBSD) and 3-D high energy synchrotron X-ray approaches .As such, the increasing use of various individual exper-imental techniques within modeling paradigms is illus-trated in Fig. 1b.
Just what is this emerging field? After all, many mul-tiscale modeling strategies discussed in the mechan-ics and Materials Science communities couple modelingand experiments intrinsically, i.e., they involve mod-els of different physics at multiple scales and exper-imentally validate those models at the correspondinglength/time scales. But because of the interdisciplinarynature of the fields, a novel branch in Materials Sci-ence that relies on the synergy between experimen-tal and computational approaches at multiple length-scales is opening up new frontiers and research direc-tions at the crossroads of both traditional Computa-tional and Experimental Materials Science, in additionto the emerging field of Integrated Computational Ma-terials Engineering (ICME). Such cooperative interac-tions find many applications in the development, thecharacterization and the design of complex materialsystems. Yet more possibilities are waiting to be ex-plored and many new questions of physical, numerical,analytical or characterization nature have materialized.These things all make this emerging field an extremelyfruitful and promising area of research.
Almost all problems in Materials Science are multi-scale in nature: fundamental small scale physical mech-anisms and processes occurring at the atomic lengthscale and femtosecond time scale have a profound im-pact on how materials perform at larger spatial andtime scales. Despite this inherent multiscale charac-ter, effective macroscopic models and measurements areusedoften with satisfactory accuracythat accountintrinsically for the effects of the underlying microscopicprocesses. For example, a conventional tensile test di-rectly measures the ultimate tensile strength, the maxi-mum elongation, the reduction in area, and determinesa phenomenological uniaxial stress-strain curve. Suchan experiment can be used effectively to calibrate aHollomon-type power law relation between the stressand the amount of plastic strain regardless of the fun-damental microstructural mechanisms responsible forthe macroscopic response.
However, the above class of macroscopic models andexperiments have some limitations. The first obviouslimitation is the complete neglect of microscopic mech-anisms that are sometimes of interest. In our tensiletest example, we may want to know how changes in mi-crostructure would lead to a change in the mechanicalresponse. It is often of interest to know and charac-terize the microstructural defects such as dislocations,martensite formation, and mechanical twinning that areat the origin of plastic deformation, not just the macro-scopic flow. Such information requires modeling and ex-perimental tools that can track such defects. These in-clude, for example, CPFEM or discrete dislocation dy-namics (DDD) on the modeling side and TEM, X-raytomography or SEM-based EBSD on the experimentalside. Another limitation is associated with the heuris-tic nature of the macroscopic models and experimen-tal measurements. These effective models often have anempirical foundation limiting their predictive capabili-ties, while the interpretation of experimental data maybe challenging in the case of complex material systemssubjected to the intricate interplays of thermodynam-ics, transport (diffusion and phase transformations) andnon-linear behavior. As we will describe later, one ofthe advantages of detailed multiscale models is theirability to switch which physical mechanisms are andare not active in the material to decouple and exam-ine the importance of different mechanisms and helpin the interpretation of experimental data. Finally, an-other limitation is that it is difficult to gain insight intothe margins and uncertainty due not only to test-to-test variation, but also the natural stochasticity at thelower length scales of engineering materials that makesan exact knowledge of the underlying microstructureand active mechanisms impossible.
Modeling and Experimental Characterization Across Length Scales 3