Journal of The Electrochemical Society, 161 (8) E3009-E3014 (2014) E30090013-4651/2014/161(8)/E3009/6/$31.00 © The Electrochemical Society

JES FOCUS ISSUE ON MATHEMATICAL MODELING OF ELECTROCHEMICAL SYSTEMS AT MULTIPLE SCALES

ReaxFF Reactive Force Field Simulations on the Influence of Teflonon Electrolyte Decomposition during Li/SWCNT Anode Dischargein Lithium-Sulfur BatteriesMd Mahbubul Islam,a Vyacheslav S. Bryantsev,b,∗ and Adri C. T. van Duina,z

aDepartment of Mechanical and Nuclear Engineering, The Pennsylvania State University, University Park,Pennsylvania 16802, USAbLiox Power, Inc., Pasadena, California 91106,USA

Lithium-sulfur batteries are amongst the most appealing choices for the next generation large-scale energy storage applications.However, these batteries still suffer several formidable performance degradation issues that impede its commercialization. Thelithium negative electrode yields high anodic capacity, but it causes dendrite formation and raises safety concerns. Furthermore, thehigh reactivity of lithium is accountable for electrolyte decomposition. To investigate these issues and possible countermeasures, weused ReaxFF reactive molecular dynamics simulations to elucidate anode-electrolyte interfacial chemistry and utilized an ex-situanode surface treatment with Teflon coating. In this study, we employed Li/SWCNT (single-wall carbon nanotube) composite anodeinstead of lithium metal and tetra(ethylene glycol) dimethyl ether (TEGDME) as electrolyte. We find that at lithium rich environmentat the anode-electrolyte interface, electrolyte dissociates and generates ethylene gas as a major reaction product, while utilization ofTeflon layer suppresses the lithium reactivity and reduces electrolyte decomposition. Lithium discharge from the negative electrodeis an exothermic event that creates local hot spots at the interfacial region and expedites electrolyte dissociation reaction kinetics.Usage of Teflon dampens initial heat flow and effectively reduces lithium reactivity with the electrolyte.© 2014 The Electrochemical Society. [DOI: 10.1149/2.005408jes] All rights reserved.

Manuscript submitted January 29, 2014; revised manuscript received February 18, 2014. Published February 25, 2014. This wasPaper 566 from the San Francisco, California, Meeting of the Society, October 27–November 1, 2013. This paper is part of the JESFocus Issue on Mathematical Modeling of Electrochemical Systems at Multiple Scales.

Lithium-sulfur batteries has received profound attention in therechargeable energy storage technology due to its tremendouslyhigh energy density, environmental friendliness, abundance, and low-cost.1,2 Although Li-ion batteries are now ubiquitous in consumerelectronics, their practical capacity is still inadequate for the electricvehicle (EV), and large scale renewable energy storage systems.3 Li-Sredox-couples are deemed the most promising to meet future energystorage demands, since theoretical capacities of Li and S are 3860 and1672 mAhg−1, respectively.2

In spite of all these advantages, commercialization of Li-S batter-ies is plagued owing to severe shortcomings in prolonged life-cycleretention. High electrochemical potential of sulfur cathode entails Lianode for practically realizable energy density.4 The use of metal-lic lithium anode is still problematic due to its low cycling efficiency,deleterious dendrite formation, and safety concerns.5 During charging,lithium ion reduced at the anode surface and deposit as metallic phase.Thus, dendrite like structure grows and further penetrates through theseparator, leading to short circuit and thermal runaway.6 Moreover,dissolved polysulfide anions migrate through the electrolyte by well-known “shuttle mechanism” and react at the anode surface to forminsoluble sulfides (Li2S and Li2S2) and these sulfides develop a pas-sivation layer onto bare lithium surface. This insulating layer hinderslithium access to the anode, increases internal cell resistance resultingin poor rate-capability.7,8

Furthermore, the chemical potential of the lithium is higher than thelowest unoccupied molecular orbital (LUMO) of commonly used or-ganic liquid electrolytes.9 During charging, freshly deposited lithiumon anode surface is extremely reactive and can react with most of theorganic electrolytes. Therefore, electrolytes can be reduced at the Lianode, hence the reaction products may form a solid electrolyte inter-phase (SEI) in between electrolyte and anode.5 SEI allows transporta-tion of Li-ions but suppresses electron transfer and thus it preventsfurther anodic reactions.5 However, such a stable SEI formation is notcommonly observed in most available ether type electrolytes. There-fore, to improve the performance of Li-anode, both ex-situ and in-situtreatments are reported in the literature. One ex-situ approach to abate

∗Electrochemical Society Active Member.zE-mail: acv13@psu.edu

the associated problems with the lithium metal anode is to introduce aLi-ion penetrable passivation layer in between electrolytes and anodesurface to isolate highly reactive Li-anode from electrolyte solventsand polysulfides. Li-ion conducting passivation layer using Li3N,10

Li2CO3,11 LiPON,12 and UV cured polymerization13 have been previ-ously studied. These protective thin films inhibit lithium reactivity andcorrosion by polysufides; however, their fabrication process is costly.14

Recently, in-situ protective layer formation during cell operation us-ing LiNO3 additives in electrolyte has been investigated.14,15 LiNO3

exhibits promising performance in protecting Li-anode via formationof a SEI layer, however, it may reduce at the carbonaceous surface ofthe cathode. These irreversible reduction products affect reversibilityand capacity of Li-S batteries. Moreover, continuous growth of passi-vation film with the consumption of LiNO3 has detrimental effect onLi-anode performance.15

Li-S battery electrolytes should comply with the requirementof high ionic conductivity, low viscosity, good polysulfide solu-bility, electrochemical stability with lithium metal electrodes, andsafety.16 Ether based electrolytes show highest polysulfides solubil-ity and expedite electrochemical activity of sulfur reactions.17 Al-though organic carbonate solvents, such as ethylene carbonate (EC),propylene carbonate (PC), di-methyl carbonate (DMC) are preva-lent in commercial Li-ion batteries, these are not used in Li-S bat-teries due to their reactivity with the soluble polysulfides. Myriadsof experimental studies have been conducted on the electrochem-ical properties of various electrolytes in Li-S batteries, such astetra(ethylene glycol) dimethyl ether (TEGDME)1,18–24, poly(ethyleneglycol) dimethyl ether(PEGDME),25 tetrahydrofuran(THF),20 1,2-dimethoxyethane (DME),26 1,3-dioxolane(DOXL).21,27,28 Especially,TEGDME has been reported as a good liquid electrolyte because ofhigh discharge capacity at room temperature.19,20,29,30

However, to address the issues attributed to the current Li-S bat-tery technologies, more profound understanding of the electrode-electrolyte interface, electrolyte dissociation chemistry, and morphol-ogy changes of electrodes during cell operation are required. Atomicand molecular level insights can effectively enlighten the underlyingintricate chemistry and the limitations in the current experimental ap-proaches. In this study, we used ReaxFF reactive molecular dynamicssimulations to elucidate anode-electrolyte interfacial chemistry and

  ecsdl.org/site/terms_use address. Redistribution subject to ECS license or copyright; see 130.203.115.129Downloaded on 2014-02-25 to IP

E3010 Journal of The Electrochemical Society, 161 (8) E3009-E3014 (2014)

the role of protective layer on anode surface in electrolyte dissociation.Previously, ReaxFF method was successfully employed for study-ing electrolyte chemistry in Li-ion batteries. Using ReaxFF, Bedrovet al.31 investigated reactions of singly reduced EC and S.-P Kimet al.32 studied SEI formation in Li-ion batteries.

Although Li-metal anode is most prevalent in experimental Li-S batteries, in this study, in lieu of lithium metal anode, we usedLi/SWCNT composite in order to preclude dendrite formation. Incomparison with the Li intercalated graphite anode (LiC6, correspondsto the capacity of 372 mAhg−1), recent experiments reported higherlithium capacity in Li/SWCNT anode (Li1.6C6, corresponds to thecapacity of 600 mAhg−1).33 Capacity reduction due to the usage ofLi/SWCNT composite material instead of lithium metal might bean acceptable trade-off among improved cell-life, performance, andsafety. It is known that the performance limiting parameter in state-of-the-art lithium batteries is the lower capacity of the cathode materials.

In addition, to control lithium reactivity, we employed a polyte-trafluoroethylene (PTFE) or Teflon monolayer at the anode surface.Previously, Chang et al.34 studied Teflon coated lithium anode toprotect it from lithium dendrite formation. We investigate if the pres-ence of Teflon can significantly scale down lithium reactivity with theTEGDME resulting in less electrolyte decomposition. We also reportelectrolyte dissociation pathways upon interaction with the lithiumion.

ReaxFF Background

ReaxFF is a general bond order35,36 (BO) based empirical forcefield method which allows bond breaking and formation during simu-lations. In ReaxFF, forces on each atom are derived from the followingenergy expression

Esystem = Ebond + Eover + Eunder + Elp + Eval + Etor + EvdWaals + ECoulomb

where partial energy contributions include bond, over-coordinationpenalty and under-coordination stability, lone pair, valence, and tor-sion, non-bonded interactions van der Waals, and Coulomb energies,respectively.

BOs are calculated from the interatomic distances, which act asa key component for all the bonded interactions and it is updated inevery iteration. Since, all the connectivity dependent interactions, i.e.valence and torsion energy are BO dependent, their energy contri-bution diminishes upon bond breaking. In non-reactive force fields,non-bonded interactions i.e. van der Waals and Coulomb are usuallycalculated between atom pairs that are not involve in a bond or notsharing a valence angle. However, in reactive environments atomicconnectivity changes during simulation and it is awkward to setupsuch exclusion rule, therefore, in ReaxFF non-bonded interactions arecalculated between all the atom pairs irrespective of their connectiv-ity. Any excessive short distance non-bonded interactions are circum-vented by using a shielding term in the van der Waals and Coulombenergy expressions. To eliminate any discontinuity in the non-bondedinteraction energies, a seventh order taper function is employed.37

A geometry dependent charge calculation scheme, ElectronegativityEqualization Method (EEM)38 is used for charge calculation. For amore detailed description of the ReaxFF method, see van Duin et al.,39

Chenoweth et al.,40 and Russo Jr. et al.41

Force Field Development

We began our force field development for describing C/H/O/S/Li/Finteractions by merging previously published parameters. C/H/O/Sand Li parameters were adopted from Castro-Marcano et al.,42 andBedrov et al.,31 respectively, and fluorine parameters were devel-oped based on the training set as described by Paupitz et al.43 Inthis work, we performed additional force field training against quan-tum chemistry (QC) calculations for describing bond dissociation,equation of state for Li-F interaction, and Li-binding energies onelectrolyte molecules. A successive one-parameter parabolic extrap-

Figure 1. Comparison of the ReaxFF and QC data for (a) Li-F bond dissocia-tion in LiF, and (b) LiF simple cubic crystal equation of state. Purple and limerepresent lithium and fluorine atoms, respectively.

olation method44 was used for fitting ReaxFF parameters against QCdata.

Non-periodic QC calculations presented in this work were car-ried out in the Jaguar 7.5 program45 using density functionaltheory (DFT) with the B3LYP46,47 hybrid functional and the 6–311++G** basis set. SeqQuest code (version 2.61j)48 was em-ployed for the periodic calculation using a Gaussian-based lin-ear combination of atomic orbital method and double -ξ pluspolarization49 with Perdew−Becke−Ernzerhof (PBE) generalizedgradient approximation.50 For more details regarding these DFT cal-culations, please refer to Ref. 51.

To parameterize ReaxFF bond energy data, we carried out QCcalculations on Li-F bond dissociation in a LiF molecule. Figure 1ashows the comparison of the ReaxFF and QC results. In this case,we constructed ground state geometries through full geometry opti-mization. In order to obtain dissociation profiles, bond restraint wasapplied in the atom pair of interest during geometry relaxation. Bonddistances were varied from 1.45 to 9 Å. ReaxFF nicely reproducesfull dissociation profile and in particular, near the equilibrium regionit is in excellent agreement with the QC values.

Equation of state calculation was performed on LiF simple cubiccrystal to obtain energy-volume relationships. Both compression andexpansion with respect to the equilibrium volume were applied on thegeometry and the QC energies were calculated at different volumes. Infitting procedure, the energies from the ReaxFF and QC optimizationof each corresponding volumes are compared. Energy response withthe change in volume as predicted by the ReaxFF and QC is shown in

  ecsdl.org/site/terms_use address. Redistribution subject to ECS license or copyright; see 130.203.115.129Downloaded on 2014-02-25 to IP

Journal of The Electrochemical Society, 161 (8) E3009-E3014 (2014) E3011

Table I. Binding energies of Li and electrolyte molecules at theirdifferent charged states.

System �EReaxFF (kcal/mol) �EQC (kcal/mol)

Li- 4DMA −82.412 −79.77Li+-4DMA+ −115.446 −116.36Li-4DMTFA −97.23 −95.53

Figure 1b. We find that the ReaxFF method gives a proper descriptionof the overall energy landscape in comparison with the QC method.

Furthermore, force field parameters were also trained for theLi-binding energies in N,N dimethylacetamide (DMA) and N,N-dimethyltrifluoroacetamide (DMTFA) electrolytes at their differentcharged states. Table I represents the data obtained from the ReaxFFand QC method. In first case, neutral lithium atom binds with a clus-ter comprises of four neutral DMA molecules, and in second case,Li cation binding energy was calculated for the positively chargedDMA cluster. Neutral Li binding energy with the neutral cluster offour DMTFA molecules was also calculated. Binding energies arecalculated by subtracting individual energies of Li-atom and elec-trolyte cluster from the Li-bound electrolyte composition energies. Inall these cases, ReaxFF energies are in excellent agreement with theQC data that demonstrates the capability of the ReaxFF method todescribe Li-electrolyte interactions.

Simulation Methodology

We utilized our developed ReaxFF reactive force field to set upfull lithium-sulfur battery simulations. The Li/SWCNT anode wascoupled with the α-sulfur cathode material and TEGDME was usedas electrolyte. The anode was prepared by adding lithium atoms in(5,5) chirality armchair configuration SWCNT at a ratio close to LiC6.Nanotube strands were placed facing their open ends toward the elec-trolytes. To prepare the electrolyte geometry, 54 TEGDME moleculeswere randomly placed in a large simulation box. An NPT (isother-mal, isobaric) simulation was performed to compress the volume ofthe simulation cell to obtain room temperature density of the elec-trolyte as well as to get the periodic box dimension so that it matcheswith the periodic cell dimension. This simulation was carried out at300 K temperature and at 100 MPa pressure. A Berendsen ther-mostat and barostat52 were employed to regulate the temperature

and pressure of the system with damping constants of 100 fs and5000 fs, respectively, and MD time step was 0.2 fs.

For the first set of simulations, we placed Li/SWCNT anode,TEGDME electrolyte and α-sulfur in a simulation cell with periodicboundary condition in two directions (i.e., y and z) and the directionperpendicular to the electrode surface is set as non-periodic boundary.A fixed graphene wall is used in the non-periodic dimension parallelto the electrodes to prevent escaping of lithium and other generatedspecies during MD simulations. The wall reflects back the generatedspecies to the simulation cell. The cathode was comprised of α-sulfurwith a dimension of (1 × 3 × 1 ao

3). Simulation snapshot is shown inFigure 2a.

In the second set of calculations, in order to reduce lithium reactiv-ity, a monolayer consists of short-chain Teflon was utilized as anodesurface coating. Total number of atoms in the system with or withoutTeflon was 3684 and 3484, respectively.

For both systems, the simulation procedure began with the relax-ation of the entire system using a low-temperature MD simulation.Next, 300 K MD simulations were performed in the NVT (isovol-ume, isothermal) ensemble using a temperature damping constants of100 fs and a MD time step of 0.20 fs.

However, in this manuscript, we are focusing only on the anode-electrolyte interface and electrolyte dissociation chemistry while leav-ing lithium-sulfur interaction for a future publication.53

Results and Discussion

Li-SWCNT/TEGDME/Sulfur simulation.— During anodic dis-charge NVT-MD simulations at 300 K, lithium atoms release from theSWCNT with positive charge (> 0.5e) and interact with the TEGDMEelectrolytes. It was observed that release of Li from the SWCNT isa highly exothermic event. The exothermic heat flow causes localheating and increases anode-electrolyte interface temperature sub-stantially. This high temperature enhances reaction kinetics of thelithium interaction with the electrolyte resulting in dissociation of theTEGDME molecules. Major reactive events take place within first fewpicoseconds of the simulation, when electrolyte dissociation gener-ates various species. Once the highly energetic lithium atoms transfertheir energy to the neighboring electrolyte molecules they become lessreactive. As the dynamics proceeds, local hot spots dissipate heat tothe surrounding regions and gradually overall system temperature sta-bilizes to 300 K, and no significant reactive events are observed at

Figure 2. (a) simulation snapshot at 10 ps; cyan, red, gray, purple represent Carbon, Oxygen, Hydrogen, Lithium atoms, respectively; Two dimensional temperaturedistributions (b), (c), and (d) at 1, 100, and 300 ps, respectively. Temperature in the color bar is in K. (e) System temperature profile with simulation time, and(f) Surface plot showing Li-concentration as function of both cell length and time.

  ecsdl.org/site/terms_use address. Redistribution subject to ECS license or copyright; see 130.203.115.129Downloaded on 2014-02-25 to IP

E3012 Journal of The Electrochemical Society, 161 (8) E3009-E3014 (2014)

Figure 3. (a) TEGDME dissociation pathways; cyan, red, gray, purple represent Carbon, Oxygen, Hydrogen, Lithium atoms, respectively (a) onset of dissociationvia cleavage of C-O bond at site A (b) C-O bond breaking at site B and formation of Li-O bond (c) CH3OLi generates and C-O bond cleavage at site C release ofC2H4 (d) C-O bond cleavage at site D, (e) Li2O and another C2H4 formation and (f) potential energy profile during dissociation reactions.

300 K equilibrated temperature regimes. Figure 2b–2d representstemperature distribution in the simulation cell at different time steps.These contour plots are constructed by taking an average of the atomictemperature in 10 × 10 Å grids (in x and y direction) over 1 ps duration.These figures displays high temperature regions at anode-electrolyteinterface during initial stages of lithium discharge from the SWCNT.However, as system equilibration proceeds, local hot spots disappearand temperature distribution at 300 ps exhibits more uniform temper-ature in the simulation cell. It was observed that no noticeable localhot-spots were present after 50 ps. Figure 2e displays system temper-ature profile with time during the simulation. We note that the peaktemperature in the simulation, averaged over the full simulation box,is ca. 420 K because of the fast electrolyte dissociation reactions atabout 1 ps. This system temperature is comparable with the experi-mentally measured temperature range of 443–573 K due to the violentreactions between lithium and electrolyte in Li-ion batteries.54

We investigated the dissociation pathways of a TEGDME moleculeupon interaction with the lithium ions. To study these reaction path-ways, we randomly placed one TEGDME molecule and six lithiumatoms in a simulation cell and run NVT-MD simulation at 300 K for1 ps. TEGDME dissociation reactions with Li are exothermic and re-actions occurred very rapidly within about 0.75 ps. One of the reactionpathways are shown in Figure 3a–3e. Dissociation initiated throughthe cleavage of C-O bond as lithium atom attacks site A and subse-quent formation of Li-O bond was observed. Lithium interaction atsite B breaks C-O bond to form CH3OLi. The system interacts withother Li atom and an energetically favorable pathway induces C-Obond breaking at site C and subsequently releases C2H4. Again, Liatom interacts with the Li-O bond at site D produces Li2O and anotherC2H4 molecule. Figure 3f shows the potential energy profile duringTEGDME dissociation reactions.

However, the type of reaction products generated in this single-TEGDME simulation is contingent on the number of lithium atomsthat are interacting with the TEGDME molecule. Cleavage of twoconsecutive C-O bonds through lithium interaction releases C2H4 gasas a reaction species. CH3 radical formation was observed through

the terminal C-O bond dissociation, however, we did not notice for-mation of any ethane gas via combination of two methyl radicals.Rather, CH3 radical occasionally reacted with the lithium atom andforms LiCH3. Near the anode surface, at high lithium concentrationregion, TEGDME molecule undergoes multiple lithium induced C-O bond breaking reactions, resulting in higher number density oftrapped ethylene gas. Electrolyte dissociation further from the anodesurface, where Li concentration is lower, typically produces higherorder alkoxy lithium (R-O-Li). In this simulation, LixO (x = 2,3,4)was not observed as a stable species rather it survived only about fewhundred femtoseconds, and sometimes combined with the availableCH3 radical to form lithium alkoxide. However, in contrast with theterminal C-O bond, intermediate C-O bond cleavages were detectedat higher frequency resulting in the large number of C2H4 formedcompared to lithium alkoxides.

Figure 4 shows the statistics of the major reaction products, C2H4

and TEGDME during the MD simulation. Other species with lowerconcentration than the major reaction products, and which survivedfor only a short residence time, were excluded from this analysis. Asshown in this figure, TEGDME dissociation and ethylene gas forma-tion reactive events are observed until approximately 50 ps. Theseresults are correlated with our previous observation, which shows nosignificant local heating occurred at about 50 ps or subsequent timesteps.

The lithium ion density in our simulations is higher than the prac-tical lithium ion density during charging and discharging. In exper-imental studies, typical Li ion concentration in the electrolyte is ca.1–1.2 M,1,21 whereas in our simulation, the interfacial lithium ion con-centration was about 10 M, which is accountable for de-stabilizing theelectrolyte and fast dissociation chemistry. Li-concentration distribu-tion as function of both cell length and time is shown in Figure 2f. Inthis surface plot, first peak indicates the lithium concentration in theSWCNT anode at the beginning of the simulation. The spatial varia-tion in the lithium concentration is contingent on the lithium diffusionthrough the electrolyte, and after initial release of the lithium atomsfrom the anode, no significant concentration variation is observed

  ecsdl.org/site/terms_use address. Redistribution subject to ECS license or copyright; see 130.203.115.129Downloaded on 2014-02-25 to IP

Journal of The Electrochemical Society, 161 (8) E3009-E3014 (2014) E3013

Figure 4. Evolution of two major species C10H22O5 and C2H4 during 300 KNVT-MD simulation (without Teflon).

with the time. Our overall observations on electrolyte chemistry are(i) high energetic lithium released from the Li/SWCNT anode causeslocal heating and thus expedite TEGDME decomposition chemistry,(ii) higher lithium concentration facilitate cleavage of multiple C-Obond leading to formation of more ethylene gas. However, for theprolonged life cycle of Li-S batteries, it is essential to inhibit elec-trolyte loss via dissociation. Furthermore, gas formation during bat-tery cycling leads to the internal pressure buildup and potential safetyconcerns. To approach this issue, we discuss an ex-situ anodic surfacetreatment approach in the next section.

Li-SWCNT/Teflon/TEGDME/Sulfur simulation.— In the previoussection, we discussed high lithium reactivity induced electrolyte de-struction. To reduce lithium reactivity, improve electrolyte stabil-ity, and concomitantly allowing Li-ion diffusion through the anode-electrolyte interfaces, a porous Teflon coating was used. Teflon is afluor-carbon polymeric compound with high chemical stability and

Table II. Statistics of TEGDME electrolyte dissociation and C2H4formation at 300 ps simulation.

Number of remaining Number ofSimulation C10H22O5 (out of 54) generated C2H4

Without Teflon 47 21With Teflon 53 2

melting point. A mono-layer of Teflon is chosen for this simulationbecause multiple layer of Teflon will reduce the lithium diffusivity,and within the given MD time scale, lithium-electrolyte interactionmay not be achievable. In this 300 K NVT-MD anodic dischargesimulation, highly energetic lithium atoms that are released from theanode by an exothermic process, at first interact with the Teflon coat-ing. Then Li dissipates its energy through transferring heat to theTeflon, and then relatively non-reactive diffusion of lower energy Lioccurs through the electrolyte. Teflon effectively withstands high heatflow at the anode-electrolyte interfacial region and reduces subsequentelectrolyte decomposition reactions. Figure 5a shows the simulationsnapshot and 5b–5d displays two-dimensional contour plots for thetemperature distribution in the simulation cell. These contour plotswere drawn following the same procedure as described in the previ-ous section. Interfacial temperature distribution in this case is analo-gous to the non-Teflon case and the system temperature profile withtime as shown in Figure 2e and 5e has an identical high temperaturepeak during the simulation that indicates system undergoes similarlevel of exothermic process. The high temperature region in the con-tour plots described in Figure 2b and 5b for without and with Tefloncase, respectively, occurs at the identical spatial region, however, inwith-Teflon case this region is occupied by the Teflon coating ratherthan the electrolyte. Therefore, lower temperature is observed pastthe Teflon layer that yields improved electrolyte stability. Distribu-tion of Li-concentration in both spatial direction and time is shown inFigure 5f. Usage of Teflon layer demonstrates a notable improvementin electrolyte stabilization, which can be represented by the numberof electrolyte molecules dissociated during this simulation. As shownin Table II, number of dissociated TEGDME and generated C2H4

molecules significantly decreases with the usage of Teflon coating.We also studied lithium-Teflon interaction chemistry. C-F bond en-ergy in isolated Teflon is 128 kcal/mol,55 while in the lithium richenvironment, we calculated C-F bond energy as about 30 kcal/mol.

Figure 5. (a) simulation snapshot at 10 ps; cyan, red, gray, purple represent Carbon, Oxygen, Hydrogen, Lithium atoms, respectively; Two dimensional temperaturedistributions (b), (c), and (d) at 1, 100, and 300 ps, respectively. Temperature in the color bar is in K. (e) System temperature profile with simulation time, and(f) Surface plot showing Li-concentration as function of both cell length and time.

  ecsdl.org/site/terms_use address. Redistribution subject to ECS license or copyright; see 130.203.115.129Downloaded on 2014-02-25 to IP

E3014 Journal of The Electrochemical Society, 161 (8) E3009-E3014 (2014)

Therefore, in presence of local hotspots, we observed a few C-F bondbreaking in Teflon with subsequent formation of LiF. Moreover, per-forated Teflon coating can also prevent precipitation of insoluble andinsulating polysulfides, such as Li2S2, and Li2S on the anode surface,while maintaining sufficient lithium diffusion through the coating.

Conclusions

In summary, we performed ReaxFF reactive molecular dynam-ics simulations to investigate the anode-electrolyte interfacial chem-istry of Li/SWCNT anode and TEGDME electrolyte. We developed aReaxFF interatomic potential through parameterization against a set ofQC data for describing electrode-electrolyte interfacial chemistry. Inour anodic discharge simulations, lithium ions release from the anodeby an exothermic reactive process and their interaction with the elec-trolyte causes decomposition of a number of electrolyte molecules.We observed that lithium concentration in the electrolyte plays animportant role in the reaction products distribution and electrolyteconsumption. At high lithium concentration region near the anodesurface ethylene gas is found as a major reaction product. Electrolytedecomposition and subsequent gas formation has a detrimental effecton battery performance. Therefore, to circumvent electrolyte dissoci-ation, we incorporated an ex-situ anode surface treatment with porousTeflon coating. Utilization of the Teflon coating dampens high reactiv-ity of the anode discharged lithium ions and thus minimizes electrolytedestruction resulting in improved electrolyte stability.

Finally, the large scale reactive molecular dynamics (RMD) sim-ulations performed using ReaxFF is a step toward the fundamentalunderstanding of the electrode-electrolyte interfacial chemistry at theatomic or molecular level. RMD simulations can assist to comple-ment the analysis of experimental results on interfacial phenomenonand to explore SEI formation mechanism using mixture of differentelectrolyte solvents and additives for developing high performanceLi-S batteries.

Acknowledgments

This work was supported by a grant from the U.S. Army ResearchLaboratory through the Collaborative Research Alliance (CRA)for Multi Scale Multidisciplinary Modeling of Electronic Materials(MSME).

References

1. X. Ji, K. T. Lee, and L. F. Nazar, Nat. Mater., 8, 500 (2009).2. M.-K. Song, E. J. Cairns, and Y. Zhang, Nanoscale, 5, 2186 (2013).3. P. G. Bruce, S. A. Freunberger, L. J. Hardwick, and J.-M. Tarascon, Nat. Mater., 11,

19 (2012).4. J. B. Goodenough and Y. Kim, J. Power Sources, 196, 6688 (2011).5. D. Aurbach, E. Pollak, R. Elazari, G. Salitra, C. S. Kelley, and J. Affinito, J. Elec-

trochem. Soc., 156, A694 (2009).6. M. Barghamadi, A. Kapoor, and C. Wen, J. Electrochem. Soc., 160, A1256

(2013).7. C. Liang, N. J. Dudney, and J. Y. Howe, Chem. Mater., 21, 4724 (2009).8. M. S. Whittingham, Chem. Rev., 104, 4271 (2004).9. J. B. Goodenough and K.-S. Park, J. Am. Chem. Soc., 135, 1167 (2013).

10. H. Wang, W.-D. Zhang, Z.-Q. Deng, and M.-C. Chen, Solid State Ion., 180, 212(2009).

11. T. Osaka, T. Momma, Y. Matsumoto, and Y. Uchida, J. Electrochem. Soc., 144, 1709(1997).

12. N. J. Dudney, J. Power Sources, 89, 176 (2000).13. Y. M. Lee, N.-S. Choi, J. H. Park, and J.-K. Park, J. Power Sources, 119–121, 964

(2003).14. X. Liang, Z. Wen, Y. Liu, M. Wu, J. Jin, H. Zhang, and X. Wu, J. Power Sources,

196, 9839 (2011).15. S. S. Zhang, Electrochimica Acta, 70, 344 (2012).16. J.-W. Choi, J.-K. Kim, G. Cheruvally, J.-H. Ahn, H.-J. Ahn, and K.-W. Kim, Elec-

trochimica Acta, 52, 2075 (2007).17. Y.-X. Yin, S. Xin, Y.-G. Guo, and L.-J. Wan, Angew. Chem. Int. Ed., n/a–n/a (2013).18. J. Hassoun, J. Kim, D.-J. Lee, H.-G. Jung, S.-M. Lee, Y.-K. Sun, and B. Scrosati, J.

Power Sources, 202, 308 (2012).19. H. S. Ryu, H. J. Ahn, K. W. Kim, J. H. Ahn, K. K. Cho, and T. H. Nam, Electrochim-

ica Acta, 52, 1563 (2006).20. H. S. Ryu, H. J. Ahn, K. W. Kim, J. H. Ahn, J. Y. Lee, and E. J. Cairns, J. Power

Sources, 140, 365 (2005).21. D.-R. Chang, S.-H. Lee, S.-W. Kim, and H.-T. Kim, J. Power Sources, 112, 452

(2002).22. G. Zheng, Y. Yang, J. J. Cha, S. S. Hong, and Y. Cui, Nano Lett., 11, 4462 (2011).23. R. Elazari, G. Salitra, Y. Talyosef, J. Grinblat, C. Scordilis-Kelley, A. Xiao, J. Affinito,

and D. Aurbach, J. Electrochem. Soc., 157, A1131 (2010).24. H. Wang, Y. Yang, Y. Liang, J. T. Robinson, Y. Li, A. Jackson, Y. Cui, and H. Dai,

Nano Lett., 11, 2644 (2011).25. J. Shim, K. A. Striebel, and E. J. Cairns, J. Electrochem. Soc., 149, A1321 (2002).26. W. Wang, Y. Wang, Y. Huang, C. Huang, Z. Yu, H. Zhang, A. Wang, and K. Yuan, J.

Appl. Electrochem., 40, 321 (2010).27. S. Kim, Y. Jung, and S.-J. Park, J. Power Sources, 152, 272 (2005).28. E. Peled, Y. Sternberg, A. Gorenshtein, and Y. Lavi, J. Electrochem. Soc., 136, 1621

(1989).29. H.-S. Ryu, H.-J. Ahn, K.-W. Kim, J.-H. Ahn, K.-K. Cho, T.-H. Nam, J.-U. Kim, and

G.-B. Cho, J. Power Sources, 163, 201 (2006).30. S.-E. Cheon, K.-S. Ko, J.-H. Cho, S.-W. Kim, E.-Y. Chin, and H.-T. Kim, J. Elec-

trochem. Soc., 150, A800 (2003).31. D. Bedrov, G. D. Smith, and A. C. van Duin, J. Phys. Chem. A, 116, 2978 (2012).32. S.-P. Kim, A. C. T. van Duin, and V. B. Shenoy, J. Power Sources, 196, 8590 (2011).33. B. Gao, A. Kleinhammes, X. P. Tang, C. Bower, L. Fleming, Y. Wu, and O. Zhou,

Chem. Phys. Lett., 307, 153 (1999).34. O.-K. Chang, J. C. Hall, J. Phillips, and L. F. Sylvester, (1989).35. J. Tersoff, Phys. Rev. Lett., 61, 2879 (1988).36. D. W. Brenner, Phys. Rev. B, 42, 9458 (1990).37. A. C. Van Duin, A. Strachan, S. Stewman, Q. Zhang, X. Xu, and W. A. Goddard, J.

Phys. Chem. A, 107, 3803 (2003).38. W. J. Mortier, S. K. Ghosh, and S. Shankar, J. Am. Chem. Soc., 108, 4315 (1986).39. A. C. Van Duin, S. Dasgupta, F. Lorant, and W. A. Goddard, J. Phys. Chem. A, 105,

9396 (2001).40. K. Chenoweth, A. C. T. van Duin, and W. A. Goddard, J. Phys. Chem. A, 112, 1040

(2008).41. M. F. Russo Jr. and A. C. T. van Duin, Nucl. Instrum. Methods Phys. Res. Sect. B

Beam Interact. Mater. At., 269, 1549 (2011).42. F. Castro-Marcano, A. M. Kamat, M. F. Russo Jr, A. C. van Duin, and J. P. Mathews,

Combust. Flame, 159, 1272 (2012).43. R. Paupitz, P. a. S. Autreto, S. B. Legoas, S. G. Srinivasan, A. C. T. van Duin, and

D. S. Galvao, Nanotechnology, 24, 035706 (2013).44. A. C. T. van Duin, J. M. A. Baas, and B. van de Graaf, J. Chem. Soc. Faraday Trans.,

90, 2881 (1994).45. Jaguar, version 7.5, Schrodinger, LLC: New York, (2008).46. A. D. Becke, Phys. Rev. A, 38, 3098 (1988).47. C. Lee, W. Yang, and R. G. Parr, Phys. Rev. B, 37, 785 (1988).48. P. A. Schultz, Phys. Rev. Lett., 84, 1942 (2000).49. A. E. Mattsson, P. A. Schultz, M. P. Desjarlais, T. R. Mattsson, and K. Leung, Model.

Simul. Mater. Sci. Eng., 13, R1 (2005).50. J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett., 77, 3865 (1996).51. V. S. Bryantsev, V. Giordani, W. Walker, J. Uddin, I. Lee, A. C. T. van Duin,

G. V. Chase, and D. Addison, J. Phys. Chem. C, 117, 11977 (2013).52. H. J. C. Berendsen, J. P. M. Postma, W. F. van Gunsteren, A. DiNola, and J. R. Haak,

J. Chem. Phys., 81, 3684 (1984).53. M. M. Islam, A. Ostadhossein, and A. C. T. van Duin, in preparation.54. H. Yang, H. Bang, K. Amine, and J. Prakash, J. Electrochem. Soc., 152, A73 (2005).55. Y. Sato and M. Murahara, J. Adhes. Sci. Technol., 18, 1545 (2004).

  ecsdl.org/site/terms_use address. Redistribution subject to ECS license or copyright; see 130.203.115.129Downloaded on 2014-02-25 to IP