The use of virtual structure libraries for computational screening to identify lead systems for further investigation has become a standard approach in drug discovery. Transferring this paradigm to challenges in material science is a recent possibility due to advances in the speed of computational resources and the efficiency and stability of materials modeling packages. This makes it possible for individual calculation steps to be executed in sequence comprising a high-throughput quantum chemistry workflow, in which material systems of varying structure and composition are analyzed in an automated fashion with the results collected in a growing data record. This record can then be sorted and mined to identify lead candidates and establish critical structure-property limits within a given chemical design space. To-date, only a small number of studies have been reported in which quantum chemical calculations are used in a high-throughput fashion to compute properties and screen for optimal materials solutions. However, with time, high-throughput computational screening will become central to advanced materials research. In this presentation, the use of high-throughput quantum chemistry to analyze and screen a materials structure library is demonstrated for Li-Ion battery additives based on ethylene carbonate (EC).

1. High-throughput Quantum
Chemistry and Virtual Screening
for Lithium Ion Battery Electrolytes
George Fitzgeralda
Mathew D. Hallsa
Ken Tasakib
a Accelrys, Inc
b Mitsubishi Chemical, Inc

2. Introduction
• Computational approach to designing new materials is well-established
– Polymers
– Catalysts
– Active Pharmaceutical Ingredients
– Semiconductors
– Nanotubes
• Computational approaches:
– Save time by screening compounds rapidly in silico and forwarding only the best
leads for experimental screening
– Deliver information that you can’t get from experiment
– Provide fundamental insight at the atomic level
• Automation provides
– A means to screen more leads faster
– A method to make sense of your results
• This presentation demonstrates this approach for lithium ion battery electrolytes
© 2008 Accelrys, Inc. 2

3. Acknowledgements
• Support from Accelrys and Mitsubishi Chemical is gratefully
acknowledged
• Computational resources were provided by Hewlett-Packard
© 2008 Accelrys, Inc. 3

4. Lithium Ion Batteries and SEI Film Formation
• The electrolyte typically consists of one or more lithium salts dissolved in
an aprotic solvent with at least one additional functional additive
• Additives are included in electrolyte formulations to increase the
dielectric strength and enhance electrode stability by facilitating the
formation of the solid/electrolyte interface (SEI) layer
© 2008 Accelrys, Inc. 4

5. Lithium Ion Batteries and SEI Film Formation
1 e- decomposition
scheme
2 e- decomposition
scheme
• Initiation step leading to anode SEI formation is electron transfer to the
SEI forming species resulting in a concerted or multi-step decomposition
reaction producing the passivating SEI layer at the graphite-electrolyte
interface
• Important requirements for electrolyte additives selected to facilitate
good SEI formation are:
– higher reduction potential than the base solvent
– maximal reactivity for a given chemical design space
– large dipole moment for interaction with Li
© 2008 Accelrys, Inc. 5

6. In Silico Materials Analysis and Optimization
Select alloy, Structural
Develop
ceramic, optimization Property
structural Lead system
dielectric or dynamics prediction
model
material, etc simulations
Change constituent atoms,
substitute additive, etc
• Requires user intervention at each step
– allows for user error setting parameters
– time between compute steps for user action
• Manually extract properties from output and compute derived properties
• Manually make comparisons with data for other systems
This is a labor-intensive process !
© 2008 Accelrys, Inc. 6

7. Virtual Screening
• Virtual screening is the cornerstone of in
silico drug discovery
• Allows researchers to effectively screen
drug design space to identify most
promising structures
– reduces the size of a chemical library to
be screened experimentally: O(106) to
O(10)
• Improves the likelihood of finding
interesting structures
– systematic screening
– screen possible design space before
synthesized
• Saves time and money
– computational evaluation is faster and
much less expensive than experimental
testing
High-throughput virtual screening will revolutionize the
discovery and optimization of materials systems
© 2008 Accelrys, Inc. 7

8. Materials Discovery and Optimization using Virtual
Screening
Chemical Virtual Automated
Motif Library QC
Design Enumeration Calculation
Identification Virtual
of optimum Materials
leads Library /
Database
Experimental
Analysis
screening
© 2008 Accelrys, Inc. 8

9. Anode SEI Additive Structure Library
X X Z Z X
X X X X
X X X X X
R4
O
O X
Z
X X
X
R3
X Z
O R2
X
X X
R1 Z
X z1
X X
X = F or H
• Cyclic carbonates, related to ethylene carbonate (EC), are often used as
anode SEI additives for use with graphite anodes
• To explore the effect of alkylation or fluorination on EC-based additive
properties an R-Group based enumeration scheme was used to generate a
EC-based additive structure library (7381 stereochemically unique
structures)
© 2008 Accelrys, Inc. 9

10. Anode SEI Additive Descriptors
• Increased reduction potential correlates
with a lower LUMO energy value or a
higher vertical electron affinity (EAv)
• Measure of stability or reactivity is the
chemical hardness of a system (η) ELUMO, EAv
• Larger dipole moment leads to stronger
dipole-cation inteactions (μ)
• Work by Chung et al, has shown that
the PM3 semiempirical Hamiltonian is
effective in computing properties μ
related to electrolyte components
performance G.-G. Chung, H.-J. Kim, S.-I. Yu, S.-H. Jan, J.-W. Choi and M.-
H. Kim, J. Electrochem. Soc. 147, 4291 (2000).
© 2008 Accelrys, Inc. 10

11. LUMO & Electron Affinity
LUMO
• Lower LUMO energy reflects the ease of
electron transfer at the anode surface which
is the activation step leading to reductive
decomposition
• Relative LUMO energy plots with respect to
the EC LUMO shows that LUMO variability
across the library is 3.7 eV, with the lower
limit in the ca -3.4 eV range
• A better indicator of high reduction potential Electron Affinity
is a larger electron affinity. Additives are
selected with higher reduction potential than
base solvent
• Relative EA plots with respect to the EC EA
shows that EA variability across the library is
4.14 eV, with the upper limit in the ca +4.1 eV
range
© 2008 Accelrys, Inc. 11

12. Dipole Moment & Hardness
Dipole Moment
• Larger dipole moment leads to stronger
nonbonded interactions with the Li-cation
• Relative dipole moment plots with respect to
the EC dipole shows that dipole variability
across the library is 7.5 Debye, with the upper
limit in the ca 2.81 D range
• Lower chemical hardness indicates increased
Chemical Hardness
reactivity and lower stability which is
desirable for SEI film formation
• Relative hardness plots with respect to the
EC hardness shows that hardness variability
across the library is 1.7 eV, with the lower
limit in the ca -1.6 eV range
© 2008 Accelrys, Inc. 12

13. Anode SEI Additive Results
• Optimal materials must satisfy a number of objectives
• Multi-objective solutions represent a trade-off between objectives
• One approach is to adopt the “Pareto-optimal” solution
– Set of solutions such that is not possible to improve one property without
making any other property worse
– This case:
• Minimize the chemical hardness
• Maximize the dipole moment and electron affinity
© 2008 Accelrys, Inc. 13

14. 3D View of Pareto Surface
© 2008 Accelrys, Inc. 14

15. Anode SEI Additive Pareto Optimal Candidate
• Screening the EC-based additive
library gives structure 1573 as a
typical Pareto-optimal solution
↑(EA and μ) and ↓η
© 2008 Accelrys, Inc. 15

16. Method Validation
• Butyl sultone (BS) has been reported to be a highly effective graphite anode SEI forming additive in
electrolyte formulations involving propylene carbonate (PC) as a co-solvent
• The use of BS as an electrolyte additive overcame performance issues seen with the application of PC as
a co-solvent, such as the loss of discharge capacity and decrease of cycling stability
• BS is predicted to be more easily reduced than PC, having an electron affinity 1.27 eV larger than that
computed for PC. The chemical hardness of BS is 0.60 eV lower than that of PC, suggesting it would be
more reactive, facilitating SEI formation
• Fluoroethylene carbonate (FEC) has also been used experimentally as an SEI additive in electrolytes for
lithium ion batteries
• Using an ethylene carbonate (EC) containing solvent, the addition of FEC at 10% and 30% levels, shifted
the onset of SEI formation to higher potentials by +0.25 eV and +0.38 eV, compared to 0.4 eV for the base
solvent vs. Li/Li+
• FEC is predicted to has a higher electron affinity (+0.42 eV) and a lower chemical hardness (−0.13 eV)
than EC, suggesting superior SEI forming behaviour
© 2008 Accelrys, Inc. 16

17. Summary
• The use of high-throughput quantum chemistry to analyze and
screen a materials structure library representing a well defined
chemical design space has been applied to fluoro- and alkyl
derivatized ethylene carbonate (EC)
• The effect of fluorination leads to a maximum electron affinity
across the library of 4.13 eV, compared to alkylation leading to a
maximum value of only 0.43 eV, relative to EC
• The results presented here introduce a new and powerful
approach for exploring the property limits of structural motifs for
lithium battery electrolyte additives. High-throughput
computational screening has the potential to dramatically reduce
the time and effort for evaluating new synthetic directions for
anode SEI formation additives
• This work has appeared in print as Journal of Power Sources 195
(2010) 1472–1478.
© 2008 Accelrys, Inc. 17

18. Related Work
• A similar approach has been applied to other materials:
– OLEDS
– Fuel cell catalysts (ECS Transactions, 25 (2009) 1335-1344)
– Polyolefin catalysts (poster, 21st North American Catalyst Society meeting)
• Methods are being developed to improve the optimization process for searching
these very large libraries
– QSARs
– Neural Networks
– Evolutionary Algorithms
© 2008 Accelrys, Inc. 18