Electrochimica Acta 50 (2004) 235–242

Interfacial behavior of polymer electrolytes

John B. Kerr∗, Yong Bong Han, Gao Liu, Craig Reeder,Jiangbing Xie, Xiaoguang Sun

Lawrence Berkeley National Laboratory, MS 62R0203, 1 Cyclotron Road, Berkeley, CA 94602, USA

Received 2 June 2003; received in revised form 9 December 2003; accepted 28 January 2004Available online 7 August 2004

Abstract

Evidence is presented concerning the effect of surfaces on the segmental motion of PEO-based polymer electrolytes in lithium batteries.For dry systems with no moisture the effect of surfaces of nanoparticle fillers is to inhibit the segmental motion and to reduce the lithium iontransport. These effects also occur at the surfaces in composite electrodes that contain considerable quantities of carbon black nanoparticlesfor electronic connection. The problem of reduced polymer mobility is compounded by the generation of salt concentration gradients withint ross-linking.C to loss ofc tly impactt urrent wast portion oft hat are veryr ced mobilityp e less thani©

K

1

eibvofcpti

pos-quiteces.parti-umid-ed toell as

talwthideof

ainlyen-

0d

he composite electrode. Highly concentrated polymer electrolytes have reduced transport properties due to the increased ionic combined with the interfacial interactions this leads to the generation of low mobility electrolyte layers within the electrode andapacity and power capability. It is shown that even with planar lithium metal electrodes the concentration gradients can significanhe interfacial impedance. The interfacial impedance of lithium/PEO–LiTFSI cells varies depending upon the time elapsed since curned off after polarization. The behavior is consistent with relaxation of the salt concentration gradients and indicates that ahe interfacial impedance usually attributed to the SEI layer is due to concentrated salt solutions next to the electrode surfaces tesistive. These resistive layers may undergo actual phase changes in a non-uniform manner and the possible role of the reduolymer layers in dendrite initiation and growth is also explored. It is concluded that PEO and ethylene oxide-based polymers ar

deal with respect to this interfacial behavior.2004 Elsevier Ltd. All rights reserved.

eywords: Interfacial behavior; Lithium batteries; Polyelectrolytes; Polymer electrolytes

. Introduction

The mechanisms of ion transport in dry and gel polymerlectrolytes and polyelectrolytes has been the subject of much

nterest due to their use as separators in rechargeable lithiumatteries and fuel cells. While the simplest approach is to in-estigate the motion through the bulk of the membrane, muchf the critical ion transport occurs close to the electrode sur-

aces, which are frequently high surface area composites withomplex morphologies. Since its known that the mobility ofolymers is considerably altered in the presence of nanopar-

icles[1,2] and surfaces[3], it is important to investigate theon motion in the regions close to the electrodes. Battery sys-

∗ Corresponding author. Tel.: +1 510 486 6279; fax: +1 510 486 4995.E-mail address:jbkerr@lbl.gov (J.B. Kerr).

tem modeling of gel and dry polymer systems[4–7] showsthat much of the performance loss occurs in the comite structures where the ion transport properties can bedifferent from the bulk due to the influence of the surfaThe recent intense interest in the use of ceramic nanocle filler material to alter the transport properties of lithisalts in polymer electrolytes[8–19] has generated a conserable body of useful data in this respect that can be usunderstand the behavior of composite electrodes as wmembranes.

The transport of lithium ions to the surface of lithium meelectrodes is one of a number of critical factors in the groof dendrites, which can lead to coulombic inefficiency, sreactions, short circuiting of the cell and the formationmossy lithium, a serious safety issue. Dendrites certgrow if the current density exceeds the limiting current d

013-4686/$ – see front matter © 2004 Elsevier Ltd. All rights reserved.oi:10.1016/j.electacta.2004.01.089

236 J.B. Kerr et al. / Electrochimica Acta 50 (2004) 235–242

sity [20–22], which is determined by the salt concentration,the cell geometry and the transport properties of the elec-trolyte. In the case of a binary salt polymer electrolyte, thetransport properties are the conductivity (σ), the salt diffusioncoefficient (Ds) and the transference number (t+0). The trans-port properties are generally measured as bulk properties andtherefore any changes in these properties that occur as a re-sult of the presence of the electrode interfaces can induce thegrowth of dendrites at unexpectedly low current densities.It has long been observed that dendrite growth also occursat current densities well below the limiting current[23,24]and the factors that influence the initiation[25,26]and prop-agation[27–29]of dendrite growth are a multifaceted prob-lem that includes transport properties, interfacial propertiesand mechanical properties of the electrolytes in addition tothe properties and possible non-uniformity of the solid elec-trolyte interphase (SEI)[30] on the lithium metal. Hence theunderstanding of the behavior of polymer electrolytes at in-terfaces is of considerable importance to the efficiency andsafe operation of lithium metal batteries and some initial ob-servations are presented in this paper.

2. Experimental

wasp riedu -d formDPa zer,a Ma-t wt.%f tionp -N80w andw detaile uma x.L( mo odw d asc lack( teel.P nicr .T andh GPCa anda

ace( bics ina

(neutral, 5.8 nm i.d. particles size) from Aldrich were driedunder 160◦C vacuum over P2O5 for 2 days before stored indry box. Acetonitrile and glyme were obtained from Burdick& Jackson with highest purity (>99.9%) and lowest watercontent (<0.001%) available. Prior to use, they were dried inthe “solvent” glove box by percolation through a column ofsuper-activated alumina (ICN).

Two glove boxes were used in the experiments. One is de-fined as the “solvent” glove box, which was used for solvent-involved operations such as dissolving, film casting, etc. Theother is defined as the “dry” glove box, which was scrupu-lously free of organic solvent vapors. It was used for the stor-age of dry chemicals and samples, cell assembly, etc. Bothglove boxes were filled with helium with H2O and O2 con-centrations <1 ppm. Unless otherwise mentioned, the dry boxstands for “dry” glove box.

The polymers, salts, and fillers (measured in weight per-cent of lab-purified PEO) were weighed in the dry box andtransferred in capped bottles to the “solvent” glove box. Theywere dissolved in dried acetonitrile and the solutions werecast on Teflon plates. The solvent was evaporated and theresulting films were fully dried in the vacuum anti-chamberof the “solvent” glove box at room temperature. The filmswere then transferred in sealed containers to the “dry” glovebox. The films were assembled in Swagelok cells in the dryb withk e di-m kinge olledp allowm n outo -th andc s thee d andt Thec ments.C aledf

nse( l in-t mea-s rwisem least1 ts be-l abo-r lysis( na-l s. Ar ely,n hem le top mo-s ted at1 re

PPO (Parel) was a gift from Zeon chemical andurified by soxhlet extraction with methanol and then dnder vacuum over P2O5 in a drying pistol prior to introuction to the glove box. PEO samples were obtainedow Chemical. PEO (Polyox WSR-301,Mw = 5 × 106 andolyox WSR-N80,Mw = 2 × 105) was recrystallized fromcetone to remove butylhydroxytoluene (BHT) stabilind was defined as lab-purified PEO. From the product

erial Safety Data Sheet, these polymers contain up to 4umed silica and calcium salt residues. The recrystallizarocess does not remove these additives. Polyox WSRith all additives removed is defined as ultra-pure PEOas prepared according to a method that is described inlsewhere[31]. All PEO samples were dried under vacut 50◦C over P2O5 overnight before storing in the dry boiTFSI (LiN(SO2CF3)2) a gift from 3M, LiTf (LiCF3SO3)Aldrich) and LiClO4 (Alfa Aesar) were dried under vacuuver P2O5 in a drying pistol at 160◦C and transferred try box without exposure to air. V6O13 (KM VANOX-13)as a gift from Kerr–McGee Corporation and prepareomposite electrodes with PEO–LiTFSI and acetylene bChevron) using a solvent slurry coating on stainless solytrimethylene oxide (PTMO) was prepared by catio

ing-opening polymerization of oxetane with BF3-etheratehe PTMO was fractionated three times using THFexane mixture solvent and dried under high vacuum.nalysis of the pure PTMO polymer using THF eluentPLGel Mixed B column gaveMw = 65,000; PDI = 5.7.Fumed silica from Degussa with a hydrophilic surf

Aerosil 200, 12 nm i.d. particle size) and a hydrophourface (Aerosil R805, 12 nm i.d. particle size) and alum

ox with two stainless-steel electrodes. A spacer ringnown thickness and central area was used to control thensions of polymer electrolytes between the two bloclectrodes. The Swagelok cell is designed to exert contrressure on the cell sandwich and completely sealed toeasurement outside of the dry box. The cells were takef the dry box and cured at∼110◦C overnight in a convec

ion oven with temperature control within±0.5◦C. Underigh temperature and pressure, the polymers melt, flowompletely fill the central hole of the spacer. To assesffect of exposure to atmosphere, the cells were opene

he sample allowed to rest in ambient air for up to 72 h.ells were then reassembled and resealed for measureontrol cells were opened in the dry box and then rese

or comparison.A Solartron SI 1254 four-channel frequency respo

65 kHz to 0.01 Hz) analyzer and a 1286 electrochemicaerface were used to measure the ac conductivity. Theurements were carried out in cooling scan unless otheentioned. The cells were equilibrated in the oven for ath before the conductivity was measured. Measuremen

ow room temperature were taken by placing cells in a latory refrigerator. The dynamic mechanical thermal anaDMTA) was conducted on a Rheometric RSAII Solids Ayzer in compression mode with a parallel plate apparatuefrigerant air drier was used to supply dry air; alternativitrogen evaporated from a liquid N2 dewar was used as teasurement atmosphere. However, it was not possibrepare the samples for DMTA without exposure to atphere. To compensate for this the samples were hea10◦C for 30 min prior to drive off any absorbed moistu

J.B. Kerr et al. / Electrochimica Acta 50 (2004) 235–242 237

before quenching to the measurement temperature. Thermalmeasurements were performed on a Perkin–Elmer DSC-7apparatus equipped with a nitrogen glove box. The furnacewas purged with helium in measurement. The samples weresealed hermetically in aluminum sample containers in thedry box and then transferred to DSC glove box. A hole waspunctured on the sample container to expose samples to purg-ing gas and remove vapors in heating for certain non-driedsamples.

3. Results and discussion

Fig. 1 shows two schemes that represent possible effectsof surfaces on polymer dynamics. The first (a) postulates thatthe interaction of the polymer with the solid surface leadsto an immobile layer of polymer next to the surface plus alayer of polymer that is restricted in its motion in additionto polymer that is sufficiently far from the surfaces that it isunaffected by the presence of the filler. When the particles aresufficiently small and the concentration of particles are opti-mum, a secondtg may be observed in dynamic mechanicalthermal analysis at temperatures up to 100◦C higher than thebulk tg [1,32]. The presence of the two different relaxations inthe presence of nanoparticles has been confirmed by neutrons blet ly-m f thep ee rfacei iono topi yt ly-m ta ft nsi-

F otionb poly-m obilityp onep

Fig. 2. Dynamic mechanical thermal analysis of PPO with added fumedsilica (A200). tanδ against temperature, recorded at 10 Hz and at 10◦C/min.The legend denotes the % by weight of fumed silica added to the polymer.

tion [1] and progressively diminishes as proportion of fillernanoparticles is increased (% by weight in legend ofFig. 2),demonstrating the complete immobilization of a portion ofthe polymer due to the presence of particle surfaces. Thestorage and loss moduli (not shown) of the polymer increaseat temperatures above thetg upon addition of fumed silicaindicating the improvement in mechanical properties that isconsistent with a decrease in the polymer motion. This de-crease in motion is due to strong interactions between thepolymer chains and the particle surfaces. Weston and Steele[37] reported on the improvements in mechanical strength ofPEO electrolytes upon addition of the filler. They also notedthat the conductivity of the filled polymer electrolytes de-creased upon addition of filler, a result that is consistent withthe immobilization of polymer motion if the mechanism ofion conduction is primarily linked to the segmental motion ofthe polymer[38,39]. Weston and Steele[40] also noted thatthe behavior of PEO electrolytes was very dependent uponthe preparation conditions and recent results obtained in thislaboratory[31] have demonstrated the critical importance ofthe presence of water and other impurities in the determina-tion of transport properties.

It has been shown in this laboratory that the addi-tion of nanoparticulate ceramic fillers such as fumed sil-ica or alumina to polymer electrolytes such as PEO–LiX( − −bP e int hani-c andd -t carew f hy-d

cattering[33]. The second scheme (b), which is applicao highMw polymers, shows the interaction of the long poer chains with many particles and leads to restriction oolymer motion with an increase intg [2]. In either case thffect is dependent upon the strength of the polymer–su

nteraction.Fig. 2 shows the effect on the glass transitf addition of fumed silica (Degussa A200, hydrophilic)olypropylene oxide, as measured by DMTA. The tanδ curve

s the ratio of the storage modulus (E′ in this case) divided bhe loss modulus (E′′) and is sensitive to changes in the poer state such as the glass transition[34–36]. The heighnd area of the peak in the tanδ curve is an indication o

he proportion of the polymer that is undergoing the tra

ig. 1. Schematic representations of inhibition of polymer segmental my filler particle surfaces. (a) Black shading represents immobilizeder on the filler particle surface, gray shading represents restricted molymer layers[1]. (b) Long polymer chains interact with more thanarticle leading to reduced mobility through out the polymer[2].

X = triflate (CF3SO3 ), perchlorate (ClO4 ) andis(trifluoromethylsulfonyl)imide (TFSI (CF3SO2)2N−))),PO–LiTFSI and PEMO–LiTFSI leads to an increas

he glass transition temperature, increases in the mecal moduli in the rubbery state, decreased conductivityecreased salt diffusion coefficients[31]. These observa

ions were made under conditions where considerableas taken to exclude moisture. For example, addition orophilic fumed silica (A200) to PEO–LiClO4 leads to a

238 J.B. Kerr et al. / Electrochimica Acta 50 (2004) 235–242

20◦C increase intg and a large decrease in the glass transi-tion tanδ peak size as well as an order of magnitude decreasein the conductivity. Similar observations are made with dryPEO–LiTFSI and PEO–LiTf. This clearly demonstrates therelationship between the polymer segmental motion and thelithium ion transport properties in dry electrolytes. Modifica-tion of the surface of the filler material by covalent binding ofhydrophobic octyl groups as in the case of R805 fumed silicamodifies the effect of the surface–polymer interaction andleads to less inhibition of both ion transport and segmentalmotion. The presence of moisture and other impurities also al-ters the surface of filler additives such as silica or carbon blackand leads to a change in the strength of the polymer–surfaceinteraction that controls the polymer mobility next to the sur-face. Since the strength of the polymer–surface interactionhas been shown to play a major role in the alteration of poly-mer dynamics[1], control over the surface is an importantparameter. Since electrode interfaces also involve polymerchains next to surfaces, it is likely that similar effects willalso control the polymer motion at the surfaces of electrodessuch as carbon, metal oxides and lithium metal.

Fig. 3 shows a schematic of a typical lithiummetal–polymer battery with a composite cathode comprisedof particles of an intercalation compound (30–50%), carbonblack (10%) added to make electronic connection and thep andp It iso lytec hodep arti-c esa sur-f tiono 20a PPOs e-p en iti ithint the

F lyte-c s (ca.1

Fig. 4. Dependence of glass transition temperature (tg) of PEMO, PPO andPTMO on salt concentration (shown at the ratio of oxygen concentration tolithium ion concentration). Neat polymer with no salt added is plotted as100:1 for convenience.

those measured in the bulk membrane separator due to theconfinement effect of the surfaces on polymer chain motion.

The bulk transport properties of PEO and PPO electrolyteswith LiTFSI and LiTf have been reported[41–43]and gener-ally the conductivities and salt diffusion coefficients decreaseas the salt concentration increases above oxygen–lithium ra-tios of about 8:1. This behavior can be understood in viewof the dependence of the glass transition temperature (tg)on the concentration of salt. Examples are shown inFig. 4for the amorphous polymers, PEMO (oxymethylene-linkedPEG400), PPO and PTMO. Thetg is inversely related to thesegmental motion of the polymer, which decreases as thetgincreases. As can be seen from the figure, thetg for the PPOand PEMO polymers increases rapidly as the salt concen-tration increases above 10:1 oxygen:Li ratio. The decreasein polymer mobility with salt concentration is attributed toionic cross-linking due to the binding of the lithium ions tothe polymer chains. The decrease in mobility apparently out-weighs the increase in charge carrier concentration for thesepolymers resulting in a decrease in transport properties. Thecombination of high salt concentrations with surfaces thatalso inhibit polymer motion and further increase thetg im-plies that close to surfaces the physical state of the polymerelectrolyte may change to a much less mobile and even aglassy form and the ion transport properties will be quite dif-f hus,t pos-i quitev

i gea gradi-

olymer electrolyte which binds the particles togetherrovides the ionic connection to the cathode surfaces.bvious that a large proportion of the polymer electrohains are in close proximity to the surfaces of the catarticles and the carbon black. Although the cathode ples are typically 1–10�m in diameter the carbon particlre usually 20–100 nm in size and present a very large

ace area for the polymer to interact with. The proporf filler particles in the polymer is equivalent to betweennd 50%, which is similar to the amounts added to thehown inFig. 2. If the lithium ion transport properties are dendent upon segmental motion of the polymer chains th

s reasonable to suppose that the transport properties whe composite electrode will be somewhat different from

ig. 3. Schematic representation of a lithium metal–polymer electroomposite electrode showing active intercalation compound particle–10�m) and carbon black particles (20–40 nm).

erent from those measured in the bulk of the separator. The transport properties adjacent to the surfaces of comte and lithium metal electrodes may be expected to beariable.

System modeling[4] of lithium polymer[7] and lithiumon polymer batteries[5,6] shows that during the discharnd charge of these batteries, substantial concentration

J.B. Kerr et al. / Electrochimica Acta 50 (2004) 235–242 239

ents develop within the composite electrodes and adjacent tothe lithium electrode. Depletion effects are known to initiatedendrite growth[22] at lithium metal and to limit the usablecapacity of the composite electrode[44]. The effect of highsalt concentrations next to surfaces has mostly been consid-ered in terms of salt precipitation[45], which represents achange in physical state. However, from the considerationof the behavior oftg with salt concentration and inhibitionof chain segmental motion next to surfaces outlined above,a simpler state change to the glassy form could also occurbefore precipitation. In either case, an immobile and poorlyconducting layer is formed next to the electrode that caninvolve a large proportion of the electrolyte in a compositeelectrode if the particle sizes are small enough. Once formed,these immobile layers will relax quite slowly leading to abuild up with cycling of the resistive layer and this will lead tolosses in capacity utilization due to an increase in the cathodeimpedance. An example of this effect is given inFig. 5whichshows the capacity fading behavior of a Li/PEO–LiTFSI(8:1)/V6O13 cell cycled at 0.1 mA/cm2 at 85◦C (Fig. 5a) andthe development of the impedance spectrum during cycling(Fig. 5b). In this particular case, the fading is exaggerated

Fci

due to effects involved with obtaining the impedance spectra,which appear to have caused some damage to the cell. Thatthis fading is mostly due to electrolyte transport effects maybe demonstrated by improvements in the capacity fade ratebrought about by variation of the electrode thickness and byreplacement of the electrolyte with other polymers contain-ing different salt concentrations and which have improvedtransport and interfacial properties. Increased temperaturewas also observed to improve the fading problem, whichis also consistent with better transport properties. The caseof the high capacity V6O13 appears to be different from thelower capacity LixMnO2 where depletion effects dominate[45,46] but both cases demonstrate the significant effectthat poor electrolyte transport properties have upon capacityutilization and fading. Clearly phase changes that are inducedby concentration gradients next to the electrodes will dependon the transport properties of the electrolyte, the currentdensity used and the duration of the discharge or chargein addition to the material properties of the polymer (tg).Changes in the composite electrode formulation (differentloadings of carbon, active material, particle sizes, modifica-tion of the surfaces[47]) result in better performance and,

ig. 5. (a) Cycling capacity of Li/P(EO)30–LiTFSI/V6O13 with cycle number. Dathode = 30% by weight (1.26 mg), 10% weight carbon black, 2% Brij, 58%

n (a). PEO is lab purified.

ischarged and charged at 0.1 mA/cm2 at 85◦C. Weight of active material inP(EO)30–LiTFSI. (b) Impedance spectra taken during cycling of the cell shown

240 J.B. Kerr et al. / Electrochimica Acta 50 (2004) 235–242

Fig. 6. Polarization behavior of a Li/P(EO)30–LiTFSI/Li cell at 85◦C. Sixsuccessive polarizations in same direction at 0.2 mA/cm2 for 2 C each.Impedance spectra (seeFig. 7) and 1 h polarization between polarizations.PEO is lab purified.

indeed, the capacity retention of VxOy cathodes have beenconsiderably improved by these means.

Although these interfacial effects appear to be intuitivelyreasonable for composite electrodes with very high surfaceareas, it is less obvious that they would greatly affect pla-nar electrodes such as lithium metal if the current density iskept well below the limiting current[48]. However,Figs. 6–8demonstrate that even at a planar lithium electrode the interfa-cial effects can be considerable.Fig. 6shows the polarizationbehavior of a symmetrical lithium/P(EO)30–LiTFSI/lithiumcell at 85◦C polarized six times at 0.2 mA/cm2 for a charge of2.75 C/cm2 in the same direction. In this case the PEO poly-

F0 ated.(

Fig. 8. Polarization behavior of Li/P(EO)30/Li cells cycled symmetrically at0.2 mA/cm2 at 85◦C for 2 C per cycle. One-hour rest between polarizations.Cell thickness (a) 38�m and (b) 265�m.

mer has aMw of 4 × 106 and contains 4% fumed silica andcalcium salts (lab purified, seeSection 2). The polymer elec-trolyte membrane was very thick in this case (500�m) witha low limiting current (ca. 0.25 mA/cm2). After the polariza-tion was terminated, the impedance spectrum of the cell wasrecorded beginning 30 s after polarization was terminated.The cell was then allowed to relax for 1 h following record-ing of the impedance spectrum when a second impedancespectrum was recorded. Polarization was then repeated. Theimpedance spectra recorded immediately after polarizationand after the 1 h relaxation are shown inFig. 7. One cansee that size of the semi-circle measured 30 s (Fig. 7a) af-ter the current is terminated changes with cycling in a quiteunpredictable fashion but tends to return to the same valueafter the 1 h relaxation (Fig. 7b). This indicates that little ir-reversible chemical reaction occurs during polarization andat least some portion of the interfacial impedance is due tosalt concentration gradients at the electrodes that relax withtime. This short-term behavior is not the same as SEI growthdue to long-term corrosion[49] and is consistent with obser-vation of little SEI growth during full cell cycling[45]. Thecontinued increase in the polarization of the cell observedin Fig. 6 indicates that even a 1 h rest is insufficient to fullyrelax the concentration gradients in this cell, which is a verythick cell designed to exaggerate the effects of the transportp

etryp

ig. 7. Impedance spectra recorded on cell shown inFig. 6 from 65 kHz to.1 Hz after polarizations. (a) Recorded 30 s after polarization terminb) Recorded 1 h after the impedance spectra in (a).

roperties.Fig. 8 demonstrates the difference that the cell geom

lays in the development of concentration gradients.Fig. 8a

J.B. Kerr et al. / Electrochimica Acta 50 (2004) 235–242 241

shows the cycling of a symmetrical Li/P(EO)30–LiTFSI/Licell with a separator thickness of 38�m and a steady-statelimiting current of 2.5 mA/cm2 while Fig. 8b shows a cellwith separator thickness of 265�m and a limiting current of0.36 mA/cm2. The cells are cycled at 0.2 mA/cm2 at 85◦Cfor 2 C per cycle with a 1 h rest between polarizations. In thecase of the thin cell the polarization is only 25 mV duringthe first few cycles but this grows on cycling until at cycle21 the voltage does not reach a steady value but rises quitesteeply and the cell short circuits due to growth of a dendrite.With the thick cell, the polarization is considerably larger oncycle 1 (200 mV), increases upon cycling and rapidly fails toreach a steady value. By cycle 16, there is a large polarization,which appears to be close to transition time behavior. Someroughening change of the electrode surface then occurred thatlead to a change in the electrode area or separator thickness.This reduced the polarization to allow the cell to continue cy-cling. The cell finally short-circuits in cycle 52. The thickerseparator with better mechanical properties apparently pre-vented a short circuit due to dendrite growth for a consid-erable time but the electrode shape changes to reduce thepolarization.

Another series of cells was built with higher salt concen-trations (8:1) and cycled under the same conditions. Althoughthese cells had steady state limiting currents nearly four timesh so d tol triesa on-t wthb hasec ext tot highs dientc mere iumm f thec notb ng tov itiont ofa elec-t ate andt te ofg re-fl theni y tod de.A theo ry toe driteg ilityd ought riteg

4. Conclusion

Polymer chain motion in polymer electrolytes is affectedby the presence of surfaces. In dry polymer electrolytes thisusually leads to decreases in lithium ion transport and poorerperformance of the cell. The generation of salt concentra-tion gradients due to binary salts with inadequate transportproperties leads to increased concentrations of salt next tothe electrodes and this compounds the effect of the surfaceon the polymer motion. Both capacity and power capabilityare impacted by these effects. This implies that polyelec-trolyte single-ion conductors will perform better since theydo not generate concentration gradients. The question thenis whether the polyelectrolytes remain mobile at the surfaceand do not self-assemble into a non-conducting layer on theelectrode surface. The results shown here also imply thatpolymers based on ethylene oxide units are not optimum forinterfacial behavior.Fig. 4 shows that the polymer PTMOshows a much smaller dependence oftg on salt concentra-tion. This confirms some recently reported results on comb-branch polymers containing TMO groups[50]. The resultsdescribed in this paper indicate that much improved behav-ior should be expected with TMO-containing polymers, andpolyelectrolytes. The results of experiments on these materi-als will be reported in forthcoming publications.

A

forE ree-d ento

R

067.bove

95)

lingd-

lish-

con,

wer

urces

998)

, J.

[ .K.

[ nics

igher than the cells inFig. 8they all rapidly grew dendriten the first polarization cycle. The cell polarization droppe

ow values indicative of changes in the electrode geomelthough the cells did not completely short out. This is c

rary to expectation if salt depletion drives the dendrite grout could be consistent with a concentration-dependent phange at the anode. The growth of a less mobile layer nhe anode resulting from the cross-linking effects of thealt concentration generated by the concentration graombined with the strong interactions between the polylectrolyte and the constituents of the SEI layer on the lithetal will result in a steady increase in the impedance o

ell. It is postulated that the growth of such a layer maye uniform across the electrode surface thereby leadiariations in the secondary current distribution, a condhat favors the initiation of dendrite growth. The growth

less mobile and more resistive layer at the counterrode will reduce the limiting current of the cell below thxpected from the bulk electrolyte transport propertieshe cell geometry. Since it is also possible that the rarowth of such a layer will not be uniform and may wellect non-uniformities already present on the electrode,t would not be unreasonable to expect non-uniformitevelop in the current distribution at the plating electrolthough a decrease of the local limiting current belowperating current could occur, it is apparently not necessaxceed the limiting current to initiate and propagate denrowth[28] and the changes that occur in polymer mobue to concentration effects at surfaces may well be en

o reduce the critical current density to initiate dendrowth.

cknowledgement

This work was supported by the Assistant Secretarynergy Efficiency and Renewable Energy, Office of FomCAR and Vehicle Technologies of the U.S. Departmf Energy under Contract No. DE-AC03-76SF00098.

eferences

[1] G. Tsagaropoulos, A. Eisenberg, Macromolecules 28 (1995) 6[2] A. Zhu, S.S. Sternstein, Nanofiller–Polymer Interactions at and a

the Glass Transition Temperature, MRS, Boston, MA, 2000.[3] W.E. Wallace, J.H. Vanzanten, W.L. Wu, Phys. Rev. A 52 (19

R3329.[4] K.E. Thomas, R.M. Darling, J. Newman, Mathematical mode

of lithium batteries, in: W.A.V. Schalkwijk, B. Scrosati (Eds.), Avances in Lithium-ion Batteries, Kluwer Academic/Plenum Pubers, New York, NY, 2002, p. 345.

[5] M. Doyle, J. Newman, A.S. Gozdz, C.N. Schmutz, J.M. TarasJ. Electrochem. Soc. 143 (1996) 1890.

[6] P. Arora, M. Doyle, A.S. Gozdz, R.E. White, J. Newman, J. PoSources 88 (2000) 219.

[7] K.E. Thomas, S.E. Sloop, J.B. Kerr, J. Newman, J. Power So89 (2000) 132.

[8] F. Croce, G.B. Appetecchi, L. Persi, B. Scrosati, Nature V394 (1456.

[9] L. Persi, F. Croce, B. Scrosati, E. Plichta, M.A. HendricksonElectrochem. Soc. V149 (2002) A212.

10] H.J. Walls, J. Zhou, J.A. Yerian, P.S. Fedkiw, S.A. Khan, MStowe, G.L. Baker, J. Power Sources V89 (2000) 156.

11] A.S. Best, A. Ferry, D.R. MacFarlane, M. Forsyth, Solid State Io126 (1999) 269.

242 J.B. Kerr et al. / Electrochimica Acta 50 (2004) 235–242

[12] A.S. Best, J. Adebahr, P. Jacobsson, D.R. MacFarlane, M. Forsyth,Macromolecules 34 (2001) 4549.

[13] J. Adebahr, A.S. Best, N. Byrne, P. Jacobsson, D.R. MacFarlane, M.Forsyth, Phys. Chem. Chem. Phys. 5 (2003) 720.

[14] M. Forsyth, D.R. MacFarlane, A. Best, J. Adebahr, P. Jacobsson,A.J. Hill, Solid State Ionics 147 (2002) 203.

[15] B. Kumar, L.G. Scanlon, R.J. Spry, J. Power Sources 96 (2001) 337.[16] B. Kumar, S.J. Rodrigues, J. Electrochem. Soc. V148 (2001) A1336.[17] B. Kumar, L. Scanlon, R. Marsh, R. Mason, R. Higgins, R. Baldwin,

Electrochim. Acta V46 (2001) 1515.[18] Y. Aihara, G.B. Appetecchi, B. Scrosati, J. Electrochem. Soc. 149

(2002) A849.[19] Y. Aihara, G.B. Appetecchi, B. Scrosati, K. Hayamizu, Phys. Chem.

Chem. Phys. 4 (2002) 3443.[20] M. Rosso, T. Gobron, C. Brissot, J.N. Chazalviel, S. Lascaud, J.

Power Sources 97–98 (2001) 804.[21] C. Brissot, M. Rosso, J.N. Chazalviel, S. Lascaud, J. Power Sources

94 (2001) 212.[22] O. Buriez, Y.B. Han, J. Hou, J.B. Kerr, J. Qiao, S.E. Sloop, M.M.

Tian, S.G. Wang, J. Power Sources 89 (2000) 149.[23] J.L. Barton, J.O.M. Bockris, Proc. R. Soc. London, Ser. A 268 (1962)

485.[24] M. Dolle, L. Sannier, B. Beaudoin, M. Trentin, J.M. Tarascon, Elec-

trochem. Solid-State Lett. 5 (2002) A286.[25] L.G. Sundstrom, F.H. Bark, Electrochim. Acta 40 (1995) 599.[26] J.-I. Yamaki, S.-I. Tobishima, K. Hayashi, K. Saito, Y. Nemoto, M.

Arakawa, J. Power Sources 74 (1998) 219.[27] J.O. Besenhard, J. Guertler, P. Komenda, M. Josowicz, Proc. Elec-

trochem. Soc. 88–86 (1988) 618.[28] C. Monroe, J. Newman, J. Electrochem. Soc. 150 (2003) 1377.[ 01)

[[ pub-

[32] G. Tsagaropoulos, A. Eisenberg, Macromolecules 28 (1995)396.

[33] S. Gagliardi, V. Arrighi, R. Ferguson, M.T.F. Telling, Physica B 301(2001) 110.

[34] E.A. Turi,Thermal Characterization of Polymeric Materials, vol. 1,Academic Press, San Diego, CA, 1997.

[35] J.D. Ferry, Viscoelastic Properties of Polymers, third ed., Wiley, NewYork, NY, 1980, p. 75.

[36] J.M.G. Cowie, Polymer: Chemistry and Physics of Modern Materi-als, second ed., Blackie, London, 1991.

[37] J.E. Weston, B.C.H. Steele, Solid State Ionics 7 (1982).[38] M.A. Ratner, in: J.R. MacCallum, C.A. Vincent (Eds.), Polymer

Electrolyte Reviews, vol. 1, Elsevier Applied Science, London, NewYork, 1987, p. 173.

[39] M.A. Ratner, P. Johansson, D.F. Shriver, MRS Bull. 25 (2000)31.

[40] J.E. Weston, B.C.H. Steele, Solid State Ionics 7 (1982) 81.[41] M.M. Doeff, L. Edman, S.E. Sloop, J. Kerr, L.C. De Jonghe, J.

Power Sources 89 (2000) 227.[42] L. Edman, M.M. Doeff, A. Ferry, J. Kerr, L.C. De Jonghe, J. Phys.

Chem. B 104 (2000) 3476.[43] J.B. Kerr, S.E. Sloop, G. Liu, Y.B. Han, J. Hou, S. Wang, J. Power

Sources 110 (2002) 389.[44] M. Doyle, T.F. Fuller, J. Newman, Electrochim. Acta 39 (2073).[45] L. Edman, M.M. Doeff, Solid State Ionics 158 (2003) 177.[46] M.M. Doeff, A. Ferry, Y.P. Ma, L. Ding, L.C. De Jonghe, J. Elec-

trochem. Soc. 144 (1997) L 20.[47] L.F. Nazar, G. Goward, F. Leroux, M. Duncan, H. Huang, T. Kerr,

J. Gaubicher, Int. J. Inorg. Mater. 3 (2001) 191.[48] J. Newman, Electrochemical Systems, Prentice-Hall, Englewood

[ a 44

[ 48

29] T. Tatsuma, M. Taguchi, N. Oyama, Electrochim. Acta 46 (201201.

30] E. Peled, J. Electrochem. Soc. 126 (1979) 2047.31] J. Xie, R. Duan, J.B. Kerr, J. Electrochem. Soc., submitted for

lication.

Cliffs, NJ, 1991.49] B. Laik, L. Legrand, A. Chausse, R. Messina, Electrochim. Act

(1998) 773.50] J.B. Kerr, G. Liu, L.A. Curtiss, P.C. Redfern, Electrochim. Acta

(2003) 2305.