Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
2015 LAMMPS Users’ Workshop
Pressure Transferable Coarse-Grained Potentials for Polymers under Isothermal and Shock
Compressions
Vipin Agrawal and Jay OswaldArizona State University
http://compmech.lab.asu.edu
2015 LAMMPS Users’ Workshop
Finite element analysis
Overall goals
Advance fundamental understanding of polymer stress response from materials-by-design perspective under extreme loading conditions
Improve thermodynamically-consistent, coarse-grained molecular dynamics for polymer materials
Representability: predict more than just structural properties
Transferability: predict properties across a range of thermodynamic states
Molecular dynamics
Coarse graining
2015 LAMMPS Users’ Workshop
Outline
• Transferability problem for CG potentials of soft materials: the CG potential is only accurate at the thermodynamic state where it was calibrated.
– Temperature transferability
– Pressure transferability
• IBI-EAM and many body CG potential
– Pressure optimization algorithm
• Model verification and validation
– Bulk modulus
– Shock hugoniot response
2015 LAMMPS Users’ Workshop
Coarse-Grained mapping
Groups of atoms represented by super atoms
All-atom molecule
Coarse-Grained molecule
Coarse-Grained mapping
What coarse-grained potential will reproduce the same thermodynamics of the MD system?
2015 LAMMPS Users’ Workshop
Transferability and representability• Representability: does the CG model accurately reproduce
thermodynamic properties?
• Transferability: is the CG potential accurate at a different thermodynamic state than where it was calibrated?
Polyethylene
Equation of state of polyethylene at T=300K computed from CG models derived by iterative Boltzmann inversion.
For CG models, naïve application of IBI-
derived potentials are overly soft in
volumetric compression.
2015 LAMMPS Users’ Workshop
Transferability of IBI polyethylene potential
Calibration pressure
• Probability of finding a particle at distance r from a reference point
( )( ) exp
B
u rg r A
k T
U(r) – potential mean force
2015 LAMMPS Users’ Workshop
Many body Coarse-Grained potential
1( ) ( )
2i i j ij ij ij
j i j i
E F r r
2
( ) 1ij
c
rr
r
Energy of atom i
Embedding energy function
local densityfunction
PairPotential
1D density profiles
2D density profile (1 atom)
Multi-atom density profile
Arbitrary density function
( Electron densityFunction )
2015 LAMMPS Users’ Workshop
Calibration of the IBI-EAM Potential
Pressure (MPa)
Density(g/cc)
-50.7 0.74857
-10.1 0.78123
0.1 0.78417
1.0 0.78585
10.0 0.79213
100 0.82962
1000 0.97115
2000 1.04353
5000 1.16120
10000 1.27479
Calibrated state points
Pair Potentials
Angle Potentials
( ) ln( ( ))BV r k T g r
( )( ) ln
sin( )B
PV k T
Bond Potentials
2
( )( ) lnB
P lV l k T
l
2015 LAMMPS Users’ Workshop
1
3i i i i i
i
p mV
v v r f
ke Fp pp p
1
3ke i i i
i
p mV
v v
'1ˆ ( )
3i ij ij ij
i j i
p rV
r r
' '1ˆ( ) ( )
3F i i ij ij ij
i j j i
p F r rV
r r
Calibration of Embedding Energy
Virial pressure equation
Decomposition of pressure
Kinetic pressure
Pairwise pressure
Embedded pressure
2015 LAMMPS Users’ Workshop
Pressure optimization algorithm
2
0
( ) ( ) ( )4cr
i ij i
j
Nr g r r r dr
V
1
nk
k
k
F a
( 1)'
1
nk T
k
k
F ka
C P
1×m[ ]kaC( 1)
n m
kAk
P
1
T T
C P P P Δp
Embedding energy coefficients are solved by least-squares minimization
The mean density at a particle can be computed from the radial density function g(r)
Taylor series expansion of the embedding energy
Derivative of embedding energy and corresponding vector form
Taylor polynomial derivatives Taylor polynomial coefficients
2015 LAMMPS Users’ Workshop
Calibrated IBI-EAM model for polyethylene
Pressure vs Density Optimized Embedded Function
2015 LAMMPS Users’ Workshop
Calibrated IBI-EAM model for polyethylene
Optimized Angle Potentials Optimized Pair Potentials
2015 LAMMPS Users’ Workshop
Transferability of IBI-EAM model for PE
Pair and angle distributions are well-reproduced over a wide pressure range!
2015 LAMMPS Users’ Workshop
MODEL VALIDATION AND VERIFICATION
2015 LAMMPS Users’ Workshop
Isothermal Bulk modulus comparison for PE
0
( )1 ln 1
V P PC
V B
( )T
T
Pk P V
V
2( )
B
T
k T Vk P
V
Tait EOS
Isothermal Bulk Modulus
Heydemann, P. L. M., and J. C. Houck. "Bulk modulus and density of polyethylene to 30 kbar." Journal of Polymer Science Part A‐2: Polymer Physics 10.9 (1972)
( ) 1 ln 1T
P B Pk P C
B C
2015 LAMMPS Users’ Workshop
Hugoniot us-up Calculations for PE
1/2
0 0
0
1
ln 1
su P
c k PC
k C
1/2
0 0 0
0
1ln 1
ln 1
pu P PC
c k k CPC
k C
Marsh, Stanley P. LASL shock Hugoniot data. Vol. 5. Univ of California Press, 1980
0
1/2
0 0( / )c
1/2
0
0 0
1
1 /s
P Pu
V V
01 /pu V V
Sound Speed
Jump Conditions
“pseudo” us-up plane
2015 LAMMPS Users’ Workshop
Shock Hugoniot Curve for PE
Marsh, Stanley P. LASL shock Hugoniot data. Vol. 5. Univ of California Press, 1980.
Experimental PE density – 0.916 g/cc MD PE density = 0.81 g/cc
2015 LAMMPS Users’ Workshop
Mie-Gruneisen EOS
/
/ dT
V
V V
dP dTdP
V dE dE
Gruneisen EOSγ is Gruneisen parameter
v
V V
E SdE TdS PdV C T
T T
V T
S STdS T dT T dV
T V
V T
P S
T V
v
PTdS C dT T dV
T
2015 LAMMPS Users’ Workshop
Temperature Rise Along the Hugoniot
0
( )( )H
v
f vT v T e e e dv
C
dvv
v v H
H H
dE dTC T C P
dv dv v
0
1( ) ( )
2H H
dE dPv v f v
dv dv
( )H
v
dT f vT P
dv v C
From Hugoniot conditions
Forbes, Jerry W. Shock wave compression of condensed matter: a primer. Springer, 2013
2015 LAMMPS Users’ Workshop
∂P/∂T and ∂E/∂T for PE
• ∂P/∂T and ∂E/∂T are significantly different between MD and CG• The ratio of these two are not very high• For most metals, ϒ/v assume to be constant while calibrated ratio from MD
simulations increases with volumetric compression
2015 LAMMPS Users’ Workshop
Summary of IBI-EAM shock study
IBI-EAM PE potential Matches mechanical properties reasonably well Temperature rise along the Hugoniot shows deviations