Title: Theoretical HighPerformance Approach to the Study of Mechanisms of Lithium Transport in Polymer Elec
1Theoretical High-Performance Approach to the
Study of Mechanisms of Lithium Transport in
Polymer Electrolytes
- Yuhua Duan
- Department of Chemical Engineering and Materials
Science - University of Minnesota
2Outline
- Introduction to Polymer Electrolytes
- Theoretical Model Simulation Methods
- Force Field Parameterization
- Build Amorphous PEO Simulation System
- Ion Pairing in Amorphous PEO
- Li Transport in Amorphous PEO
3Battery
Chemical energy
Electrical energy
4Battery Structure
Cathode
Anode
separator
electrolytes
5Lithium ion battery
- Li metal as anode has high energy density and
could be used to build high energy battery.
- Problem Li burns in water.
- To solve this problem, Use solid materials as
electrolytes.
6Lithium Ion Polymer Battery
- Conducting electrolytes practical quantities of
very ionic lithium salts (LiClO4, LiBF4, LiPF6,
LiCF3SO3, etc) dissolved in polymer. No
traditional separator needed. - Anode Li insert into graphite LiC6
- Li?Lie-
- Cathode non-aqueous LiCoO2 , TiS2, LiMn2O4
- xe-xLiTiS2?LixTiS2 or xe-xLiLiMn2O4?Li
1xMn2O2
1mm
7Solid Electrolytes
- Ionically conducting solid materials display many
advantages over their liquid counterparts - Virtually unlimited shelf life
- Wide operating temperature range
- High energy density and voltage density
- No gassing, corrosion or leakage
- Leakproof design and safety
8Polymer Electrolytes
- Polymer Electrolyte is a new type of solid state
electrolytes. It is already used in battery, but
such battery can not use for driving automobile - Properties of polymer Electrolytes
- True solid crystal and amorphous
- Local relaxations provide liquid-like degrees of
freedom - Compare with solid oxide electrolyte, polymer
electrolyte is easy to make different shape
9What Material can be Polymer Electrolytes?
- As electrochemical point of view, electrolyte
should satisfy - Conductivity 10-2 10-3 S/cm
- Electrochemical stability at least as wide as
the voltage window defined by electrode reactions - Compatibility chemically and electrochemically
compatible with electrode materials - Thermal and Mechanical stability
- Availability easy to obtain raw materials at low
cost - Except for the first criterion, Polyethylene
oxide(PEO) and related polymers are good
candidates
10PEO Related Polymer Electrolytes
- Drawback ionic conductivity is of the order of
100 to 1000 times lower than other kinds of
materials.
- This drawback could be compensated some by some
factors, but still not meet the required level - Form thin films of large surface area giving high
power levels (gt100W/dm3). - Raise the temperature.
- It could be improved by investigating mechanism
of conductivity. Thats why this field is very
important and useful.
11Polyethylene Oxide (PEO)
- Structure formula --CH2CH2O--n,
- Tmp66?C, Tg?-60?C, soluble in H2O, CHCl3
- Chain-size from experimental synthesis is very
long. Always get amorphous structure at room
temperature. - NMR experimental established that Li moves
through the amorphous portions of the polymer.
12Computational Goals and Approach
- To understand Li transport and ion pairing
mechanisms in amorphous PEO in order to guide
synthetic chemists in the search for better
materials. - To treat the non-thermo-equilibrium system and
face the long time scale simulations - Our approach is to build molecular dynamics (MD)
model of the amorphous regions of the PEO for
simulation by ab initio parameterization of the
force field - Studying system PEO with Li ClO4-
13Molecular Dynamics(MD) Method
- MD method
- Newtons Motion equation d2ri(t)/dt2Fi(r)/mi
- Force calculation Fi(r)-dV(ri)/dri
- Particle motionVerlet algorithm
- ri(th)2ri(t)-ri(t-h)h2Firi(t)/mi
- v(t)(r(th)-r(t-h))/2h
- (NPE), (NVE) periodic boundary condition applied
- The force field of the model system
- VVintraVinter
- Coding Serial and Parallel with MPI.
14Intra-molecular Potential for PEO and ClO4-
Force-field Parameter Fitting
- Calculate the vibrational frequencies for
isolated dimethyl ether(DME) and ClO4- with ab
initio method (Gaussian94 98 package). - Using the analytical formula to fit the
frequencies by changing the parameters.
15Fitting Procedure
- With the analytical potential form and trial
parameters, to form the 3Nx3N Hessian matrix
which is the second partial derivatives of the
potential V with respect to displacement of the
atoms. - Diagonalize this matrix.
- A set of 3N eigenvectorsnormal modes
- 3N eigenvalues? frequencies square
- Compare with ab initio results, adjust the
parameters, until these two results are very
close to each other.
16Geometry and vibrational frequencies of DME
17Bond length and frequencies for ClO4-
18Inter-molecular Force Field
- PEO chainPEO Chain
- United model is used for CH2 CH3 groups
- Most force parameters are taken from other work
- Li or ClO4- PEO Li or ClO4- Li or ClO4-
- Approach the ions to PEO with different possible
paths, calculate potential energy curve - Fit to an analytical form.
19Example C of PEO with Cl of ClO4-
16 approach configurations
ClO4-
DME
20PEOClO4- Parameterization
21Fitted Inter-molecular Potential Parameters
22Building Simulation Systems
- The system is in a meta-stable non-equilibrium
amorphous state at room temperature. We can not
get structure from crystal. - We use a computational polymerization method to
simulate the polymer configuration starting from
Dimethyl Ether(CH3OCH3) liquid - Imitate the experimental synthesis process
- Compare results with neutron scattering
experimental results to test our model - Adjust our model by a control parameter in our
algorithms until close to experimental results.
23Polymerization Procedure
- Step 1 Put certain of DME into system, pick a
CH3 center from the end of a chain at random - Step 2 Explore a sphere(gtCH2-CH2 bond length)
around this center. If no chain end found within
the sphere, go back to step 1. - Step 3 Change the inter-molecular force to
intra-molecular. Change mass of CH3 to CH2 - Step 4 Run MD to quasi-equilibrium for giving
annealing time ta to produce amorphous
structures, ta is our control parameter to get
characteristically different structure - Go back to step 1 until no more bonds can be
formed. Count and list the chains formed - Compare with experimental to adjust ta.
24PEO Simulation Models
- Using 216 DME to build this system
- of Chains 23
- Longest chain size is 29
- shortest is 2
- Total atoms 1080
- 1728 DME
- of Chains 165
- Longest Chain-size 46
- Total atoms 8640
25Structure of PEO Model
- Here g(r) is weighted sum all of gij(r ) together
- The results from our model(JCP,115(2001)3957)
agree with the experimental results well? our
model is reasonable
26Ion Pairing in the Polymer
- Ion pairing can significantly reduce the lithium
ion conductivity. Even if the Li-anion pair is
mobile, no net charge may be conducted - Experimentally the degree of pairing depends on
the identity of the anion but no principles guide
the choice of what anion - To explore if Li and ClO4- are paired in PEO
- Get the modelPut LiClO4 pair into our PEO Model
randomly, run MD to local equilibrium state
27LiClO4 pairing in amorphous PEO
- Potential of Mean Force Calculations
- Wmf-kBT ln gLi-Cl(r)
- Problem only can get gLi-Cl(r) around local
equilibrium, the sample region can not reach the
short distance between Li and Cl. - For one pair case, we can fix the LiCl
separation, directly calculation the Wmf
-----expensive way since extensive sampling is
needed.
28Radial Distribution Function of PEO(LiClO4)
- gLi-Cl(r) two peak?two bound states
- gLi-O(r) coordinated O around Li is about 6
29Potential of Mean Force PEO(LiClO4)
Calculated from g(r) T280K
Direct calculated
- Simulate at 5 different temperatures
- For g(r) calculation, we take average about 15
different configurations, for direct calculation,
we take average more than 50 different
configurations - Two bound states are found from both calculations
- No obvious entropy contribution in these two
bound states
30Radial Distribution Function of PEO(LiClO4)5
- gLi-Cl(r) Compare with 1 pair case, the first
peak around 3.5Å is very small - gLi-O(r) the coordinated O around Li is about
6, at least one from ClO4- - Each ClO4- has 2 Li around it, form chain-like
structure Li--ClO4---Li
31Potential of Mean Force PEO(LiClO4)5 by g(r)
Calculations
32Ion-pairing Conclusions(JCP,111(1999)3302)
- Two bound-states of LiClO4 in amorphous PEO
- From g(r) of PEO(LiClO4), Li has 6 Oxygen
neighbors. In PEO(LiCl4)5, each ClO4- has about 2
Li near it - Li partial paired during transport, this could
be one of the reason for the low conductivity
since the net current is reduced during pairing - To deal with this problem, we need to investigate
other anions in our system to help
experimentalists choose right cathode materials
33Li Transport in Amorphous PEO
- The Mechanism of Li conductivity in PEO at room
temperature is not fully understood - Li move mainly through the amorphous region
- Li conductivity doesnt arise from a simple
process of statistically independent Li hops
through a static polymer matrix, but dynamics of
polymer matrix are essential to the transport - Li bound to O of polymer very strong, Li
hopping events have a high barrier and expected
to be rare - Up to 100ps MD, no Li hopping found
- Some attempts could find such hopping, but off
the battery technology interest - Raising above melting point, studying short chain
system, reducing interactions.
34Li Transport in Amorphous PEO
- We want to catch these rare events efficiently
since MD simulation time is very long.
- Voter(PRB, 57(1998)R13985) Parallel Replica
method - Infrequent-event system can be exploited in a
different way to develop an efficient parallel
approach to the dynamics - For a system in which successive transitions are
uncorrelated, running a number of independent
trajectories in parallel gives the exact
dynamical evolution from state to state.
35Time Distribution Between Rare Events
- We use this method to investigate the nature of
Li move in the different conformation
- According to the assumption of Voters method,
if this method is applicable, the of events vs.
simulation time is exponential.
36Simulation Scheme
- Replica Initiate N copies of the simulation box
same position, but different initial velocity
distributions - Do ordinary MD for M steps(in our case, M1000)
- quench of N copies relax to local equilibrium
at T0 K (our time-step is 0.42fs) - Determine sum of changes(?) of all Li position
- If ? lt ?0, (?0 is fixed critical value, 1.5Å),
continue MD and quench - If one of ? gt ?0, an event found. Run this
sample at finite temperature for a relaxation
time. Then replica this sample to find another
event
37Simulation Scheme
Start
No
Replica N sets of Data
Parallel run MD, quench
Meet the criteria
Stop all Jobs Collect events
yes
Calculate more?
Relaxation before replica
Yes
Stop
No
38Parallel Performance
Normal Parallel MD
Parallel Replica
- Speedup goes down if more than 20 processors are
used, due to heavy data transfer and
communication among processors.
- The speedup is almost linear with the number of
processor since only a broadcast communication
are involved among the processors.
39Snap-shot Movie of Li Moving
40Coordination Oxygen around Li
The of O around Li within the radius of 2.4Å
for 102 events
From this we can learn The O exchange events is
more reasonable for Li transport. Without O
exchange, Li could move back along with polymer
rearrangement.
41Li Diffusion Constant
Experimental D (2.20.3)x10-14 m2/s We get D1
(2.250.31)x10-13 m2/s include all events
D2 (4.733.01)x10-14 m2/s only O
exchange events
42Li Diffusion vs Torsion barrier of PEO
- Reduce the torsion force in PEO chain, the Li
diffusion constant will be increased. Torsion
barrier domain the chain relaxation - This could help synthetic chemist to choose
better polymer which torsion barrier is lower
than PEO
43Conclusion for Li Transport (JCP,122(2005)4702)
- This is the first time the parallel replica
method has been used into polymer system - The Li Motions associated with diffusion
occurred with the order of ps separated by long
rearrangement in the order of ns - Results shows the conductivity could be increased
by decreasing the torsion barrier. This will help
synthetic chemists to choose right polymer
electrolyte materials.
44Further Work
- To investigate relations between torsion barrier
and conductivity will get more information about
Li transport - To study the act of anions during Li transport
by introducing other kinds of anions (such as
triflate, TFSI, etc.) - Improve the algorithm to speedup the simulation
by treating low-frequency modes differently
45My Related Publications
- Y. Duan, J. W. Halley, L. Curtiss and P. Redfern,
Mechanisms of Lithium Transport in Amorphous
Polyethylence Oxide, J. Chem. Phys.,
122(2005)054702 - J. W. Halley, Y. Duan, K. Lidke, A. Wynveen and
M. Zhuang, Multiscale Modeling of Many Body
Systems, Condensed Matter Theories Vol. 17
Superconductivity, Superfluidity and Quantum Hall
systems, M. P. Das and F. Green ed., Page
257-278, Nova Science Pub Inc, New York, 2003 - J. W. Halley, Y. Duan, Role of Atomic Level
Simulation in Development of Batteries, J. Power
Sources, 110(2002)383-388 - J. W. Halley, Y. Duan, B. Nielsen, L. Curtiss and
P. Redfern, Simulation of Polyethylene Oxide
Improved Structure Using Better Models for
Hydrogen and Flexible Walls, J. Chem. Phys.,
115(2001)3957-3966 - J. W. Halley, Y. Duan, Mechanisms of Lithium
Conductance in PEO from Molecular Simulation,
Proceedings of Electrochemistry Society Vol.
2000-36, A. Landgrebe, R. J. Klingler ed., page
317-325, 2000 - J. W. Halley, P. Schelling and Y. Duan,
Simulation Methods for Chemically Specific
Modeling of Electrochemical Interfaces,
Electrochimica Acta, 46(2000)239-245 - J. W. Halley, Y. Duan, Simulation of Battery
Components and Interfaces on the Atomic Scale
Ion Pairing as an Example of What We Can Learn,
J. Power Sources, 89(2000)139-142 - J. W. Halley, Y. Duan, B. Nielsen, L. Curtiss and
A. G. Baboul, Lithium perclorate ion pairing
transport in a model of amorphous polyethylene
oxide, J. Chem. Phys., 11(1999)3302-3308
46Acknowledgements
- Co-workers
- J. Woods Halley, Bin Lin, B. Nielson (UMN)
- L.A. Curtiss, A. Baboul, P. Redfern, M.-L.
Saboungi (ANL)
- Supported by
- Department of Energy(DOE)
- Minnesota Supercomputing Institute(MSI)
47Thank You!