STATEGIES FOR MULTISCALE MODELING AND SIMULATION OFORGANIC MATERIALS: Polymers and biopolymers.W. A. Goddard, III, T. Cagin, M. Blanco, N. Vaidehi, S. Dasgupta, W. Floriano,M. Belmares, J. Kua, G. Zamanakos, S. Kashihara, M. Iotov, G. GaoMaterials and Process Simulation Center, 139-74Division of Chemistry and Chemical EngineeringCalifornia Institute of TechnologyPasadena, CA 91125, USAAbstractAdvances in theory and methods making it practical to consider fully first principles (denovo) predictions of structures, properties and processes for organic materials. Howeverdespite the progress there remains enormous challenges in bridging the vast range ofdistances and time scales between de novo atomistic simulations and the quantitativecontinuum models for the macroscopic systems essential in industrial design andoperations. Recent advances relevant to such developments include: Quantum Chemistryincluding continuum solvation and Force Field Embedding, De novo Force Fields todescribe phase transitions , Molecular Dynamics including continuum solvent, Nonequilibrium MD for rheology and thermal conductivity and Mesoscale simulations.To provide some flavor for the opportunities we will illustrate some of the progress andchallenges by summarizing some recent developments in methods and their applicationsto polymers and biopolymers. Four different topics will be covered. I) Hierarchicalmodeling approach applied to modeling olfactory receptors, 2) Stabilization of leucinezipper coils by indroduction of trifluoroleucine, 3) Modeling response of polymerssensors for electronic nose, and 4) Diffusion of gases in amorphous polymers.1. IntroductionIn order to develop new materials and compositions with designed new properties, it is essentialthat these properties be predicted before preparation, processing, and experimental characterization.Despite the tremendous advances made in the modeling of the structural, thermal, mechanical and transportproperties of materials at the macroscopic level (finite element analysis of complicated structures) thereremains tremendous uncertainty about how to predict many critical properties related to performance. Thefundamental problem here is that these properties depend on the atomic level interactions and chemistry(eg. making and breaking of bonds) dealing with the electronic and atomic level description at the level ofnanometers and picoseconds, while the materials designer needs answers from macroscopic modeling(finite element paradigm) of components having scales of cm and milliseconds or larger. To dramaticallyadvance the ability to design useful high performance materials, it is essential that we insert the chemistryinto the mesoscopic and macroscopic (finite element) modeling.The difficulties in doing this are illustrated in Figure 1, where we see that vast length and timescales separate the Quantum Mechanics (QM) from the macroscopic world of engineering design.Tremendous advances have been made recently in first principles QM predictions of chemical reactions,but the state of the art can handle accurately reactions with only ~50 atoms. There is no practical approachto carrying out a QM calculation on the initiation and propagation of a crack through a stabilized zirconiaceramic. Despite this difficulty, the computations MUST be based on accurate first-principles quantummechanics if we are to predict the properties of new materials.Figure 1. Multiscale modeling hieararchyOur strategy for accomplishing this objective is to develop an overlapping array of successivelycoarser modeling techniques where at each plateau (a range of length and time scales), the parameters ofthe coarse description are based on the first principles based parameters of the immediately finerdescription, as illustrated in Figure 1. Thus based on accurate QM calculations we find a Force Field (FF)including charges, force constants, polarization, van der Waals interactions etc that accurately reproducesthe QM. With the FF, the dynamics is described with Newton's equations [Molecular Dynamics (MD)],instead of the Schrodinger Equation.The MD level allows one to predict the structures and properties for systems ~ 105 times largerthan for QM, allowing direct simulations for the properties of many interesting systems. This leads tomany results relevant and useful in materials design, however, many critical problems in materials designrequire time and length scales for too large for practical MD.Thus we need to develop methods treating the mesoscale in between the atomic length and timescales of MD and the macroscopic length and time scales (microns to mm and µsec to sec) of FiniteElement Analysis (FEA). This linking through the mesoscale in which we can describe microstructure isprobably the greatest challenge to developing reliable first principles methods for practical materials designapplications.Only by establishing this connection from microscale to mesoscale will it be possible to build firstprinciples methods for describing the properties of new materials and composites (the domain of materialsscience and engineering) in terms of fundamental principles of physics and chemistry. Thus, forfundamental prediction is to play a direct role in materials innovation and design, it is essential to bridgethe micro-meso gap The problem here is that the methods of coarsening the description from atomistic tomesoscale or mesoscale to continuum is not so obvious as it was in going from electrons to atoms. Forexample, the strategy for polymers seems quite different than for metals, which seem different fromceramics or semiconductors.Given the concepts, it is necessary to carry out calculations for realistic time scales fast enough tobe useful in design. This requires developing software tools useful by design engineers, by incorporatingthe methods and results of the QM to MD to mesoscale simulations. To accomplish the goals ofdeveloping methods for accurate calculations of materials and properties, we focus on (i) implementationsthat make full use of modern highly parallel computers and (ii) building in knowledge based heuristicmethods of accessing this information automatically so that designers can focus on the macroscopic issueswithout concern for the details of theory and simulation. At this point, we expect a revolution in materialsdesign applications where the use of first principles multiscale simulations allows an every increasingamount of the design to be done on