1268030030masao dai

Upload: santhosh-lingappa

Post on 03-Jun-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 1268030030masao dai

    1/30

    Challenges in Multi Scale

    Modeling-Our Experience in the Polymer Modeling Project-

    Masao DoiDepartment of Applied Physics

    University of Tokyo

  • 8/12/2019 1268030030masao dai

    2/30

    Outline

    What is multi-scale modeling

    The polymer modeling project Objective

    The difficult part Our strategy Some outcomes

    Conclusion

  • 8/12/2019 1268030030masao dai

    3/30

    What is Multi Scale Modeling

  • 8/12/2019 1268030030masao dai

    4/30

    What is multiscale modeling? A modeling which uses two or more models

    for different scale

    by concurrent calculation e.g. Car-Pallinello methodby separate calculation

    e.g. conventional method

    Multiscale calculation can be done

    The core is to make consistent models bypassing information among differentsimulation programs.

  • 8/12/2019 1268030030masao dai

    5/30

    Multiscale modeling is difficultHow to let different programs share information

    Some kind of translation is needed

    How to gurantee the correctness of the

    translation.Some uncertainities are unavoidable

    How to ensure sustanable development

    Some uncertainities are unavoidable

  • 8/12/2019 1268030030masao dai

    6/30

    The Polymer Modeling Project

  • 8/12/2019 1268030030masao dai

    7/30

    The Polymer Modeling Project

    length cmtime sec

    length nmtime psec

    Seamless Zooming

    Bridging micro and macroMETI project from1998 to 2002

  • 8/12/2019 1268030030masao dai

    8/30

    Mesoscale Simulation Engines

    sec

    msec

    sec

    nsec

    psec

    fsec

    fm pm nm m mm m

    Micro hydrodynami

    Interfacial dynamics

    Molecular dynamicsReptation dynamics

  • 8/12/2019 1268030030masao dai

    9/30

    Molecular Dynamics

    Aoyagi,SawaSolve the equation ofmotion of particles:

    i2i

    2 Udtd

    m rr

    =

  • 8/12/2019 1268030030masao dai

    10/30

    Interfacial Dynamics

    Honda,KawakatsuSove the Edwards equation for

    polymer conformation Structure of polymers near the interface Phase separation, Microphases,

    Micellization

    ( ) ( ) ( )r r r ,nQV6

    b,nQ

    t2

    2

    =

    )]n,r (Q[V)r (V =

  • 8/12/2019 1268030030masao dai

    11/30

    Micro HydrodynamicsYamaue, Taniguchi, Sasaki

    Fluid engine and solid (gel)engine.

    Solve continuum equations Stokes eq. for velocity and

    pressure Elastic eq. for displacement Diffusion eq. for solute and ions.

    Poisson eq. for electric field

  • 8/12/2019 1268030030masao dai

    12/30

    How to Bridge Simulation Engines

    sec

    msec

    sec

    nsec

    psec

    fsec

    fm pm nm m mm m

    S i m u l a t

    i o n P l a t f

    o r m

  • 8/12/2019 1268030030masao dai

    13/30

    Difficulty in the design of the

    simulation platformIntegration of engines requires a

    common expression for physical data,but it is very difficult to set up such dataformat. Different engines use different physical

    models, and need diffirent data.

    The mesoscale engines are evolving, andthe platform has to accomodate theengines which will be developed in future .

    Integration of engines was very very difficult.

  • 8/12/2019 1268030030masao dai

    14/30

    Open is the key

    Multiscale modeling is a grand challenge.Let us try to make soft wares which willgrow and expand after our project ends.

    The system should be able to grow on its own.We should not force our way of zooming: thebasic part must be independent of it.

  • 8/12/2019 1268030030masao dai

    15/30

    Our decision for the platform design

    We leave the task of defining the datastructures to researchers.

    We determine a rule to express the datastructure.

    The platform provides a service for engineswhich state the data structure following thisrule.

    We dont impose zooming.

    We dont define words.

  • 8/12/2019 1268030030masao dai

    16/30

    UDF (User Definable Format)

    UDF consists of two parts:

    the data definition part the data part.

    The definition part defines the type, unitand the other attributes for each item in

    the data.The data part gives the value for eachitem.

  • 8/12/2019 1268030030masao dai

    17/30

    begin{def}oscillator: {

    mass : float [kg]spring_constant : float [N/m]friction_constant : float [N*s/m] }

    initial_condition:{ position: float [m]velocity: float [m/s] }

    end{def}begin{data}

    oscillator:{0.05, 0.01, 0.0001}

    initial_condition:{-0.3, 1.0}end{data}

    An example of UDF

    Definition of

    data structure

    Data

  • 8/12/2019 1268030030masao dai

    18/30

    Unit Name Value

    Script (Python)

    Results

    Simulation platform of Octa system

    n ma on y py on program

  • 8/12/2019 1268030030masao dai

    19/30

    n=size($atom[])while n>0:n=n-1sphere($atom[n].position,0)if n>0:cylinder($atom[n].position, $atom[n-1].position, 2)

    n ma on y py on program

  • 8/12/2019 1268030030masao dai

    20/30

    PASTA

    COGNAC

    SUSHI

    MUFFIN

    PASTA

    Open Computational Tool for Advanced Materials

    O penC omputational T ool forAdvanced material technology

    Octa system

    Open to public athttp:octa.jp in 2002

    Has been updatedby voluntary group

  • 8/12/2019 1268030030masao dai

    21/30

    An example: study of ABAtriblock copolymers

    Symmetric Triblock Polymer AxByAxxB AxA y

    LamellaCylinder Sphere

    What structure is formed for the tri-block copolymers.How does the bridging affects the mechanical properties.

  • 8/12/2019 1268030030masao dai

    22/30

    Results of zooming in

  • 8/12/2019 1268030030masao dai

    23/30

    Aoyagi et al. Fig.6

    Equilibrium structure

    A5B73A5

    PS2600 IP20000 PS2600

    347 polymers8 unit cell of bcc

    Deformation of domains

  • 8/12/2019 1268030030masao dai

    24/30

    Deformation of domains

  • 8/12/2019 1268030030masao dai

    25/30

    Aoyagi et al. Fig.7

    Stress- strain curve for spherical

    structure

  • 8/12/2019 1268030030masao dai

    26/30

    3D TEM Octa

    Fusion with experimental system

  • 8/12/2019 1268030030masao dai

    27/30

    Data of 3D TEMGeneral Chains

    MD EnginesDensity Biased

    Monte Carlo

    Virtual experiments

  • 8/12/2019 1268030030masao dai

    28/30

    Virtual experiments

    Similar things can be don

    MechanicalThermalTransportOptical

  • 8/12/2019 1268030030masao dai

    29/30

    Conclusion

  • 8/12/2019 1268030030masao dai

    30/30

    Summing up our experience

    We realized that collaboration of engines, andcollaboration of human being are more importantthan programmed zooming

    We decided not to pursue automatic zooming.We focused on constructing a platform on whichvarious simulation programs can collaborate witheach other.This gave flexibility and expandability to Octa

    system.ulti-scale modeling is a grand grand chanlleng