AD/A-00 571

THE EFFECT OF ADDITIVES ON LITHIUMCYCLING IN PRC'PYLENE CARBONATE

R. David Rauh, et al

EIC, Incorporated

Prepared for:

Office of Naval ResearchAdvanced Research Projects Agency

February 1975

DISTRIBUTED BY:

National Technical Information ServiceU. S. DEPARTMENT OF COMMERCE

UNCLASSIFIED

SECURITY CLASSIFICATION OF THIS PAGE (Mm,. Dae. EntoeMJ__________________

REPOT DCUMNTATON AGEREAD INSTRUCTIONSREPOT DINIANTAION AGEBEFORE COMPLETING FORMIREPORT NUMDER a OTACCESSOM NO. 3. RECIPIENT'S CATALOG NUMBER

TECHNICAL REPORT NC. 01 'D'!7/4. TITYLE (an utoiubu) S. TYPE Off REPORT & PERIOD COVERED

Final ReportTHE EFFECT OF ADDITIVES ON LITHIUM CYCLING 01 Jan 1974 to 31 Jan 1975IN PROPYLENE CARBONATE S. PERFORMING ORG. REPORT NUMBER

____ ___ ___ ___ ___ ___ ____ ___ ___ ___ ___ ___ ___C-404

7. AUTH~OR(*) S. CONTRACT OR GRANT NUMBER(s)

R. David Rauh and S. Barry Brummer N00014-74-C-0205

AREA 8 WORK UNIT NUMBERS

55 Chapel Street, Newton, Ma. 02158 NR 359-571

It. CONTROLLING OFFICE NAME AND ADDRESS 12. REPORT DATE

Of fice of Naval Research FEBRUARY 1975Department of the We, -' IS. NUMBER OF PAGES

Arlingtca, Virginia 22217 2)____ /___14. MONITORING AIrENCY NAME 8 ADDRESS(If diffeent Itm Controlling Office) 1S. SECURIT' CLASS. (of this report)

UNCLASSIFIED

ISo. DECL ASSI FICATION/ DOWNGRADINGSCHEDULE

16. DISTRIBUTION STATEMENT (of this Report)

Approved for Public Release; Distribution Unlimited.

17 DISTRASBUTISON STATEMENT (of the obstract entlered In Block 20. it different from Report)

18 SUPPLEMENTARY NOTESReproduced by

NATIONAL TECHNICALINFORMATION SERVICE ra~ UJC OCAG

Sprrgfild, VA. 22151

19, KEY WORDS (Continue on 00ter&@ side if neee85my and tdeewttly F bilock nurbor)

Lithium Corrosion, Lithium Electrode, Lithium Plating, Lithium SecondaryBatteries, Nitromethane, Passivation of Lithium, Propylene Carbonate,Sulfur Dioxide.

20 ABSTRACT (Continue on reverse side It nrcosaW'7 arnd idgflfilj by block mmnbe)

Deposition and discharge of lithium on an inert (nickel) substratein IM LiClO4/propylene carbonate (PC) has been studied as a functionof plating and stripping currents, cycle number, and electrolyte composi-tion. Water (< 1200 ppm), nitromethane, and SO2 were found to enhancethe cycling efficiency, which was relatively low in the dry electrolyte

DD M,3 1473 90DITION Of I NOV SS It OBSOLETE UCASFE

JSERCURITY CLASSIFICATION OF THIS PAGE (Weme Dope Entered)

Mew!

UNCLASSIFIED

SKMINSTY CLASUFIATION OF TH9S PAOEEbh DO@ bbi.O

20. Abstract (Cont.)

without additives, In all cases, marked deterioration in efficiencywas noted after 10 to 20 repeated cycles. Electrodeposited lithiumwas found to corrode initially (over the first 10,000 seconds ofits life) at arn average rate of 30-40 vA/cm2 , independent of theadditive within experimental error. The corrosion rate of thedeposit decreases with time, and reaches a limiting value of5-10 ijA/cm 2 . This rate is also not improved by presence of addi-tives.

A mechanism of loss of efficiency during lithium cycling andalso on open circuit stand has been formulated: Granules of lithiumare encapsulated by an insulating film of propylene carbonate-lithium reaction products, thus severing electrical contact withthe substrate. The additives have been postulated to form Li+ -

conductive films on the deposit, allowing the deposition onto thegranules and their growth into larger granules, and retardinginsulation. On open circuit stand, the long-term equilibriumfavors the formation of an insoluble insulating (Li-PC) film overthe partially soluble, Li+-conductive film. It is suggested thatsuch behavior limits the usefulness of propylene carbonate insecondary lithium batteries.

UNCLASSIFIED

gSCU1ITV CLAWICAION OF THI PAOL(Wm Da9

u-p-I

Disclaimers

The views and conclusions contained in this document arethose of the authors and should not be interpreted as

necessarily representing the official policies, eitherexpressed or implied, of the Advanced Research ProjectsAgency, Office or Naval Research or the U.S. Government.

The citation of trade names and names of manufacturers in

this report is not to be construed as official governmentendorsement or approval of commercial products or servicesreferenced herein.

?IJ

1-

It

II

THE EFFECT OF ADDITIVES ON LITHIUM CYCLING*IN PROPYLENE CARBONATE

ABSTRACT

Deposition and discharge of lithium on an inert (nickel)substrate in 1M LiCl04/propylene carbonate (PC) has been studied asa function of plating and stripping currents, cycle number, andelectrclyte composition. Water ( <1200 ppm), nitromethane, and SO2

fZ were found to enhance the cycling efficiency, which was relativelylow in the dry electrolyte without additives. In all cases, markeddeterioration in efficiency was noted after 10 to 20 repeated cycles.

Electrodeposited lithium was found to corrode initially (over thefirst 10,000 seconds of its life) at an average rate of 30-40 pA/cm2 ,

independent of the additive within experimental error. The corrosionrate of the deposit decreases with time, and reaches a limiting valueof 5-10 PA/cm2. This rate is also not improved by presence of addi-tives.

A mechanism of loss of efficiency during lithium cycling andalso on open circuit stand has been formulated: Granules of lithiumare encapsulated by an insulating film of propylene carbonate-lithium

reaction products, thus severing electrical contact with the substrate.The additi-.es have been postulated to form Li+-conductive films on thedeposit, allowing the deposition onto the granules and their growthinto larger granules, and retarding insulation. On open circuit stand,the long-term equilibrium favors the formation of an insoluble insulat-ing (Li-PC) film over the partially soluble, Li+-conductive film.It is suggested that such behavior limits the usefulness of propylenecarbonate in secondary lithium batteries.

INTRODUCTION

The efficient electroplating and stripping of lithium in non-aqueous media is essential to the development of ambient temperature,secondary lithium batteries. Lithium has been reported "stable" ina number of solvents, although the criterion for stability has oftenbeen the failure to observe visually reaction between bulk lithiumand the hot liquid (1). One of the most common media for use inlithium batteries has been propylene carbonate (PC). Yet, recentevidence exists both for the non-Nernstian behavior of the Li (Hg)/Li+ (PC) couple (2), and for the corrosion of electroplated lithiumin PC (3). These results are actually in accordance with the earlierwork of Butler, Cogley and Synnott (4) in which filming was observedthrough a gradual lowering of the exchange current of freshly cut

1

lithium in contact with PC. Such films can protect bulk lithium fromf.-.ther solvent decomposition, thus explaining the discrepancy between

results obtained for bulk lithium and for high surface area electro-plated lithium or for lithium amalgam.

Theoretically, at least, the problem of corrosion of electro-plated lithium can be alleviated by controlling the filming observedby Butler et al. so that the deposited lithium becomes passivated.Such passivation is readily observed, though not thoroughly understood,for bulk lithium in SOC12 (5,6,7) or in S02 plus numerous diluents (8).A further advantage for the use of this phenomenon in batteries isthat the films, though impermeable to corrosive solvents and depolar-izers, are conductive to Li+. Thus, they allow discharge of primarycells based upon Li/SOC12, and presumably discharge and recharge in

S02/nonaqueous solvent systems (8).

If the passivation of lithium following electrodeposition froma nonaqueous electrolyte can be achieved, secondary lithium batteriescontaining soluble cathode materials, with their implications of highdischarge rates, may be considered. In such a system, a film-formingmaterial (the "precursor"), such as S02, would be introduced. Thisprecursor would form a thin film on the lithium, protecting the metalfrom further attack, while still allowing its plating and discharge.Hence, it is important to assess the effect of possible precursors

on the properties of electrodeposited lithium.

The purpose of the present study is to observe the effects ofthree additives on the cycling efficiency, cycle life and corrosion

of lithium in PC: S02 , nitrometbane (NM), and H20. S02 and H20 areknown to have the capability of forming passivating films on lithium.NM has been used as an additive to PC in lithium batteries (9),apparently to facilitate mass transport, and its effect upon corrosionand efficiency is assessed here primarily to determine its practical

usefulness as an additive to secondary cells.

EXPERIMENTAL

1. Solvents and Solutions. PC (Matheson, Coleman and Bell,Reagent Grade), was light yellow in color, as received. Karl Fischeranalysis showed less than 30 ppm H20 in a freshly opened bottle,although PC did tend to pick up water from the atmosphere if nottightly sealed. Distillation was carried out in a Perkin-Elmer 251spinning band column at 1 mm pressure and a 2:1 reflux ratio. Thecolorless middle 60% fraction was collected at 60*C. After back-filling with argon, the receiver was transferred to a dry atmosphereargon-filled glove box (Vacuum-Atmospheres).

Solutions were prepared 1M in either LAICl4 (Foote MineralCompany) or, more often, LiCl04 (Alfa, anhydrous). It was necessary

2

to add the LiAICI4 slowly to avoid a high heat of solution and atoadd the L~~ 4 slto n

yellowing of the solvent. This yellowing was more easily avoidedif the salt were added to PC containing lithium dispersion. AnhydrousLiClOt. -s determined by Karl Fischer analysis to contain 1.5% H20by weignt. In order to reduce the H2 0 content of such solutions belowthe Karl Fischer detection limit of about 10 ppm, the solutions were

dried for three, 24 hour periods with fresh partions of Linde 5A, 600mesh molecular sieve (Alfa) which had been activated for 12 hours at400°C, and subsequently stored in a vacuum.

Stock solutions containing S02 (Matheson, anhydrous) wereprepared by directly bubbling the gas into the solvent. The S02 con-

centration was determined by weight. The solutions were then driedover activated molecular Linde 3A, 600 mesh sieve.

PC/SO 2 solutions which had been allowed to stand in the dry boxover molecular sieves for at least five days assumed a golden color.If the S02 were allowed to evaporate, the color disappeared. Becausethe origin of this effect is unknown, solutions of S02 in PC wereused soon after preparation. NM (Matheson, Coleman, Bell, spectro-quality) was treated with lithium dispersion before use.

2. Cell Design and Construction. The cell used most frequentlyin the lithium deposition and stripping studies is illustr.ted in

Fig. 1. The working electrode was constructed from 0.5 inch wide5 mil nickel ribbon to which a nickel lead had been soldered at oneend, not in contact with the solution. The plating area, firstcleansed by sanding and rinsing with methanol, was masked off withTeflon. One convenient method for doing this was to mask with Teflontape, then to press the electrode at 5,000 psi, thus bonding theTeflon to the substrate. The counter electrode was made from 2" x .015"lithium metal ribbon (Foote Mineral Company) pressed onto expandednickel screen at 1,000 psi. The reference capillary was constructedfrom a pasteur pipette whose tip was packed with pieces of prebakedfiberglass paper. The vial had a polyethylene cap which snapped overthe electrode leads, and which had a hole in the center to hold thereference electrode compartment. The latter contained lithium and aIM solution of Li+ in PC.

Another type of cell, illustrated in Fig. 2, was used for view-ing lithium deposition with a microscope. The working electrode wasa 1/4 inch diameter nickel rod (Research Organic/Research InorganicChemical Corp., 99.9%) encased in Teflon, except for the end whichcomprised the plating surface. The counter electrode was a lithiumring, situated at the bottom of the cell. As the cell had to be usedoutside the glove box, leads to the outside were made air tight usingTeflon adaptors and Teflon tape.

3

SERUM CAPLA

, ,REFERENCE LEAD

.,_______ G LASS TUBE

POLYETHYLENE "SNAP CAP"

WORKING ELECTRODE COUNTER ELECTRODE LEADLEAD

! -IM Li IN PC

TEFLON COATED NICKEL-- Li COUNTER ELECTRODE WITH

WEXPANDED Ni SUPPORTN i S URFACE- PRU

POROUS FIBERGLASS PAPER

3 DRAM CYLINDRICAL-,,- VIAL TEFLON CYLINDER, SLOTTED

TO HOLD ELECTRODES

Fig. 1: Cell for lithium cycling experiments.

4

WORKING ELECTRODE

TEFLON ADAPTORS

REFERENCE LEAD

TEFLON PLUG

1/2" NICKEL ROD~COUNTER LEAD

TEFLON PLUG REFERENCE CAPILLARY

TEFLON TUBING

LITHIUM RINGCOUNTER ELECTRODE

VIEWING DIRECTION

Fig. 2: Cell for viewing lithium deposition.

3. Electrochemical Measurements. Lithium plating and strippingwere carried out galvanostatically using a Princeton Applied ResearchModel 373 potentiostat-galvanostat. The potential between the workingand reference electrodes (IM Li/Li+) was monitored on an EAI 1130 X-Yreco..aer. Th2 potential time profile for cyclic voltammetry wasgenerated using a Wavetek Model. 133 low frequency trigger generator.All potentials are reported vs. IM Li/Li+.

RESULTS

1. The Lithium Electrode - Efficiency Measurements. A fewstudies have appeared in the literature concerning the efficiency ofplating and stripping lithium in nonaqueous solutions of Li+ salts(3,10-15). As is evident from Table 1, reports of the efficiencyof the lithium plating-stripping cycle at room temperature in PCvary considerably. It appears that the overall efficiency may be afunction of a number of parameters, especially solvent, solventpolarity, supporting electrolyte, substrate, plating charge, platingand stripping current densities, and electrode orientation. Becausethe overall efficiency of lithium cycling is a multi-determinedvariable, the results of past studies cannot be comparatively inter-preted. The one thing that is clear is that there are substantialdifficulties in cycling lithium efficiently.

In the present work, lithium was plated (p) and stripped (s)galvanostatically. The overall efficiency of the process is

i t _QS

Eff is- s [i)ipt Qp p

The open circuit potential of the nickel electrode versus the Li/Li+reference was independent of solvent. In PC, methyl acetate, methylchloroformate and DMSO, the OCP was +2.4V. The potential-determiningreaction was not elucidated.

When the current was turned on, the potential of the electrodefell immediately to a cathodic maximum overshoot, then rose withina few seconds to a cathodic plateau. In a cell like that in Fig. 1,the overshoot was typically 500 mV and the plateau was 200 mV, at5 mA/cm2 , uncorrected for iR (25-50Q). Reversing the current to stripthe deposit, no overshoot was noted, even after waiting several hoursafter plating. The plateau stripping (anodic) potential was approxi-mately equal to the plateau plating (cathodic) potential, although atendency toward higher potentials was often noted as a function ofstripping time. At the stripping end point, the potential increasedvery rapidly. A cutoff potential of +I.OV was assumed for efficiencycalculations. The current was turned off at this point to avoid anychance of dissolution of the nickel substrate. An example of aplating-stripping chronopotentiogram is shown in Fig. 3.

6

CD 0 4 (nC' %D Cn-

0 00wi 411

0 1i 4~ r4

.0 0O41 4) 0.4I 0) W W 0. to 4 o 1 0C.) 1-4 W V4 0 -H

4.1 w .1.4, 4 cc

90 >,. 0 4-1 41A4.1 60 1)4 2 r-4 w -H.cc b V 4 v4 .Ln U0 -,. ba r

to >. U. Q) 194 0~~ 0.H- oC

ci 4. 0-. .- 4 U 44) 4-1 4 0 404 co W u- w 5P.4

0

414 00 0%~ 04 0V V-41,W- % -.0 isI T 1

0 40

En- -4C1

41 0

0

0)0)

ci 4 [l4014 va 0 D0 0 e

0- 0 0'0* 0

-o OD. -0 0 CID -4 % .-4 iS 4- Ij 0 r40 -4 -4 0, A- f

I;.j C

4 cz [Z 0 Z m

-4-

L) 4.1- -v-0 0

1-4 -C40~~' 4. v- O 0W -H "4s

U))

co4C-4 .4.

-0.6

0

2

1 0. 10

+06 ______________________________________2__r_____________

306 9 21T-M +0sec3

A'0

Fig. 3: Lithium plating and stripping chronopotentiograms,

on a nickel substrate, in propylene carbonate,

1M LiAlCl4 . Plating current (ip) - 3 mA/cm2

stripping current (is).1. No additives; Efficiency = 40% after

20 cycles.

2. 1% S02 added to cell immediately after

stripping of 1. After five minutes,plating curve 2 was recc-ded, followed

by immediate stripping. Efficiency = 84%.

3. 1% S02 added; deposit stripped after 1.5

hours; Efficiency =45%.

8

Lithium was first plated and stripped from 1M Li:/PC solutionscontaining i.ss than 10 ppm H20, and the efficiency was observed as afunction of plating and stripping current density and cycle number.The charge was held constant.

Efficiencies in these dry solutions were generally quite low,38 to 50%, depetiling on plating and stripping current density, for acharge of 1 coul/cm2 . As noted previously by Selim and Bro (3),lithium remains on the working electrode after the electrochemicalend point has been reached. This phenomenon was observed under 50Xand 10OX magnification.

A speci:l cell (Fig. 2) was constructed for viewing lithiumdeposition on a nickel surface using a Unitron Series N Metallograph.Lithium was first plated at 1 mA/cm2 for a IM LiAlCI4 solution. Thedeposit appeared dull brown to gray, and was homogeneously distributedon the electrode surface. The deposit was then removed galvanostati-cally. After the dissolution was apparently complete electrochemically,it was obvious that an even coating of lithium remained on the nickelsurface. Repeated plating ai.1 stripping gave rise to pitted areas onthe lithium surface. With each cycle, the deposit became more dendritic,with many areas of very loose deposits. Dendrites were also formedmore readily at higher plating and stripping currents. In addition,the dendrites were difficult or impossible to remove electrochemically,even using low current densities or potentiostatically.

For these small charges, the efficiency showed a moderate increasewith increasing plating current density, up to 20 mA/cm2 . Except atvery low stripping current densities, where the efficiency was adverselyaffected, the efficiency was relatively insensitive to is . Theseresults are summarized in Fig. 4.

Efficiency as a function of cycle number was compared for solu-tions LM in LiAICI4 t.nd IM in LiCI04 (ip = is = 2.5 mA/cm2 , Q = 0.5coul/cm2 ). The efficiency over the first few cycles was generallyhigher for this charge than for 1 coul cm2 reported above. In addition,the Li cycling process was more efficient for LiAlCl4 than for LiCl04 -based electrolytes. Both systems showed a decline in efficiency aftera few cycles. The results are shown in Fig. 5. Examination of theworking electrodes following their effective "failure" revealed a greylithium slurry on their surfaces, indicating that most plating siteswere being blocked by insulated deposits. The efficiency characteristicon the first cycles could -. L.ger.erated by wiping clean or replacingthe working electrode. Cyclic voltammetry of these "cycled" solutionsrevealed no new electroactive impurities betweea +0.8 and +4.5V vs.Li/Li+ on Pt, and out to O.OV on nickel.

2. Cor-osion of Electrodeposited Lithium. Two methods were usedto measure tha rate of corrosion of lithium which had been electro-

9

14.,

f~ra

4 4 1"

4 -4 S~

'4-4 w- -

z I 4 I V

W

0 0

I Z 0.41 I'm44

0 0V 44 C* 4 4J34J

V.4 ,-,4

.44 Id-4r

SI 0

40

I~ U4C ) cj n Ln-. 1, -4 .-4

06

10

CIO

V4-C6rA 4

.n 4 ..

IA 6

i461

41 41 tic 4

-H WAiO

---- _ 0 0 -4

10 0 W

Ideposited onto nickel. In the first, or "standing corrosion" method,a certain charge of lithium was plated onto th. nickel, and the opencircuit potential of the electrode versus a Li/Li+ reference wasmonitor3d as a function of time. The potential remained at zerountil the bare substrate became exposed, at which point the potentialrapidly became anodic. The corrosion current was calculated as

ic = Qo. [2]t

where Qo is the amount of lithium (per cm2) which can be removed att - 0 and t is the time, in seconds, for all the plated lithium tocorrode away.

The "standing corrosion" experiment yields only an average valueof the corrosion current, as ic can change with time. The timedeperdence of ic was measured in the following way. Four cells wereplated in series. At varying intervals, the lithium was stripped fromone of the working electrodes. The cells were permuted, plated again,and again stripped at the same intervals. This procedure was carriedGut two or more times, until each cell had been stripped after eachof four waiting intervals. The amount of lithium which could beremoved was averaged for each waiting time period, and was graphedas a function of time.

Standing corrosion experiments. Data on the corrosion ofI coul/cm2 of lithium plated from PC/IM Li+ solutions as a functionof H20 concentration and supporting electrolyte are summarized inthe last column in Table 2. These data reveal an average corrosionrate (ic [open circuit]) of electrodeposited lithium in the presenceof 1M LiCl04 in dry PC to be 10-35 pA/cm2 . Values of ic obtained inthis wnainr were somewhat irreproducible, within these limits. Thisirreproducibility is probably due to morphology variations resultingfrom microscopic differences in the substrate surface, plating currentdensity, and a sensitivity of ic to the presence of certain impuritiesremovable by distillation, but which may reform thermally or viahydrolysis.

Water concentration affected the efficiency of cycling moregreatly than the corrosion rate (vide infra). The corrosion rateobserved in dry solutions was actually somewhat higher than for solu-tions containing 1200 ppm H20. Going from 103 to 104 ppm H2 0 resultedin a fourfold increase in ic and a threefold decrease in cyclingefficiency.

LiA1CI 4 gave a consistently higher corrosion rate than well-

dried solutions of LiCI04. It was not determined whether theinferiority of the LiAlCl4 was due to an impurity in the salt, or to

12

I

HCl, an hydrolysis produced of LiAICI4 . The salt is prepared usingexcess LiCl, so attack by lithium by AIC13 should be ruled out.

The baseline efficiency of lithium cycling was often lower aftera standing corrosion experiment than before. This is probably due tothe formation of residual insulating black deposits on the substrate,

resulting in an inhomogeneous plating surface.

It should be stressed that lithium ribbon remains shiny andunpitted indefinitely ir LM LiCI04 or IM LiA1CI4 solutions of PC.Only with such "dense" lithium may the formation of surface filmspassivate against extensive corrosion. The high corrosion rate ofelectrodeposited lithium is probably due to the diffuse nature of thedeposit in this solvent, giving rise to a high surface area. Lithiumin this form is highly susceptible to attack by solvent and impurities.

Insulating reaction products can also isolate granules of deposit,making them difficult to remove electrochemically.

Corrosion-time experiments. Fig. 6 shows the variation of icwith time, .s measured by the method outlined above. Corrosion ofelectrodeposited lithium was always highest immediately followingplating, and decreased with time. The initial and limiting values ofic measured by this method are also included in Table 2. Solutionscontaining LiAlCI4 were more corrosive than those containing LiCI04,verifying the "standing corrosion" results. It should be noted thatin none of these experiments was the electroplated lithium found tobe completely passivated gainst attack: All the plated lithium woulddisappear after enough timi.

3. Additive Effects.

H20. The efficiency of lithium cycling in IM LiCIO4 wasdetermined as a function of H2O concentration. The results arereportee in Fig. 7. Solutions containing up to 1,200 ppm H20 actuallyshow improved lithium cycling efficiency compared to the dry solutions.At higher H20 concentrations, the efficiency drops and gassing can bedetected.

S02. Fig. 3, curve 2, illustrates the potential-time curve forgalvanos-atic lithium platiag and stripping for a cell IM in LiAlCl4which had been cycled 20 times. After curve 1 was reported, enoughS02 was added to the cell to make i's concentration 1% by weight.Curve 2 sh--:s that this increased the cycling efficiency dramatically,from 40 to 84%. Plating and stripping also occurred at a lowerpotential in the S02 containing solutions, reflecting perhaps a highersolution conductivity.

Similar results were obtained for IM LiCl04, 1% SO2 solutions.Thorough drying was important hefore the addition of S02, because

13

0

('4

0 en~

C) 4V14 41 '

C: 0 4 u40r. -. I

00 ~Uso 10

Uu

U0 00 V).C 0$4

U A.-id* 40

g00

U U

(U u V) 'JO IN 00 c 0

cc a)o 44 cl( - 11 1 i c4 1-4 0

-H4I) -'a4

d) (Vc

-H -4 1cm C) -AI" C

It. $4 I LlI I I I I I I

00i

ca)

-J

~0 :3

0 u Q 4 1

I -(n 0 cJ-n U-,a a) 0 -, -

u wLr4 V) r4C' f J L n Ln L f

-40-4'4 -4 -4 -a)2 UQJ 0) mT

.4 0 2-. 0 ,0 0 0 r-L) L

a cc 00' .

w- 04u

-4 0

-1 f 0 0q. (n-4 -wP -.. w 04 0 .40 0 p.t - 9w *C

0 U u ~ ~ 0 0

4' -4d..4a '

V) 1-4~ - -44. co r-4 ) 1- -4 r- 1-

15

(H 20] M/I

i0-3 10-2 5 x 10-2

100-

80-

00

I LIU

1 200 '400 '600 800 1000 1200

[H 2 0] ppm

Fig. 7: Effect of H20 On the efficiency of cyclinglithium on nickel in propylene carbonate,IM AC104; ip - is M 5 zA/cm2 ; Qp - 1 coul/cM2 .

16

II

residual water would react with SO2 to yield H2 S03 , which would rapidlyattack lithium. A solution of sieve-dried PC, IM "anhydrous" LiCl04,and 1% S02 (obtained from a sieve-dried 30% SO2 in propylene carbonatestock solution) gave only 33% lithium cycling efficiency (Q - 1 coul/cm 2 ).After drying, the efficiency was increased to 91%.

An experiment to determine the optimum SO2 concentration wascarried out. The results, however, were complicated by variations inthe dependence of effici-ncy versus cycle with SO2 concentration. Inthe concentration region of 0.01 to O.05M S02, results were extremelyirreproducible, with efficiencies ranging from 50 to 75% on the firstcycle of 1 coul/cm2 and dropping to below 50% on subsequent cycles.Above O.IM S02, the first cycle was consistently above 80%. Thus,no regular curve of efficiency versus S02 concentration could beconstructed.

The cycling efficiency of PC/S0 2 solutions was further investi-gated as a function of cycle number for a constant charge.The efficiency decreased rapidly over the first 5 to 10 cycles, butmore rapidly for a charge of 1 cjul/cm2 than for a charge of 0.33coul/cm 2 . These results are reported in Fig. 8. The decline inefficiency could be remedied somewhat by waiting between cycles. Thiseffect gave rise to the spiked appearance of the graphs in Fig. 8at the points where a waiting period between cycles was observed.This spiking effect was not observed for the PC/H20 system over theH2 0 concentration range studied.

The effect of additive concentrdtion on the cycling efficiencyis thus somewhat different for NM, S02 and H20. The smooth curves ofefficiency vs. [NM] (Fig. 9), and the diminished spiking in theefficiency vs. cycle curve (except for the later cycles) (Fig. 10)indicate that NM is not consumed, or consumed very little, duringelectroplating. The same appears true for H20 at concLntrations belowabout 1,200 ppm. At higher H20 concentrations, gassing occurs duringplating. Cyclic voltammetry of PC/LiCIO4 solutions, on platinum ornickel, shows no new peaks between +0.8 and +4.5V vs. Li/Li+ when NMis added. Water reduction appears at +1.5V vs. Li/Li+, b-.t thereduction rate at low concentrations is considerably lower than thediffusion limit (17). Hence, with added H20, there is always enoughpresent to interact with the surface of as-plated lithium. Conversely,the spiking observed in the efficiency vs. cycle curve for S02indicates the consumption of SO2 during plating, and the necessityfor waiting between cycles to replenish the S02 at the electrodesurface. Cyclic voltammetry indicates that S02 reduces on platinumat +2.5V vs. Li/Li+.

Fig. 6 shows corrosion as a function of time for a total platingcharge of 2 coul/cm 2 in solutions IM in LiCIO 4 with and without SO2 .

17

100-

zW5U- 0-

5 9 (3 (7 21CYCLE

Fig. 8: Repeated cycling of lithium on nickel in propylene carbonate,1M LiClO4 with and without S02 , k")) IM S02, ip -is -5 mA/cm 2,Qp - 0.33 coul/cm2, (---) wait 1000 sec; (0) 0.65M S02,

=p is - 10 mA/cm2, Qp - 1 coul/cm2, with 1000 sec wait betweencycles except (---) wait 24 hours; (Ai) 0.65M S02, ip -is10 mA/cm2, Qp - 1 coul/cm2 , (- -- ) wait 1000 sec between cycles;(A) no S02', ca. 10 ppm H20, ip, is - 10 mA/cm2, Qp - 1 coul/cm 2.

18

0 _ _ _ _ _ ____ o0I0 I ' I I ' I ' ,

Q

CJ

Ii

0.6)0 1

0

S4

0.4 0

0 0e

0 "

ooo

44 4

oD to . 0

(0/6 AO3DI4

19 0

o .r4

0 0 0 0~

19

In this solvent, the major effect of S02 is upon the efficiency, andnot the corrosion curr:nt. At the limiting final rate of 5 pA/cm2 ,a loss of about 0.5 coul/cm 2-day would be expected for electroplatedlithium, independent of added S02. A similar experiment substitLtingLiA1CI 4 for LiCl04 gave the same high corrosion rate with SO2 aswithout, within experimental error 'Table 2 and Fig. 6).

Fig. 3 shows the chronopotentiogram for stripping lithium whichhad been allowed to stand in contact with the PC, 1% S02 electrolytefor 1.5 hours. Here, stripping takes place at a higber potential andwith a more anodic slope than if it were commenced immediately after

* plating. This behavior indicates the formation of a resistive filmon the deposit in the course of the long term corrosion process.However, it is apparent that if S02 reacts with lithium to form apassivating film, the film is ineffective in protecting the electro-deposit from this long term slow attack by the electrolyte. Conceiv-ably water or an acidic species, even in low concentrations, areresponsible for much of the corrosion, as they are small enoughmolecules or ions to penetrate these passivating films. From diffu-sion theory, a concentration of the order of 5 ppm corrosive species(e.g., H2 0) could account for corrosion currents of 10 iA/cm

2 .

Nitromethane. Nitromethane (NM), like H20 and S02, increasesthe efficiency of lithium cycling, even at low concentrations. Aconcentration of 10- 3M NM was sufficient to raise the efficiency of

cycling lithium on a clean nickel surface from 42 to 58%. Higherconcentrationq of NM caused further increases in efficiency, asreported in Fig. 9.

The results of repeated cycling of NM/PC solutions are shown in

Fig. 10. Deterioration of efficiency of cycling a 1 coul/cm2 chargebegins at the eighth cycle for 0.1M NM solutions, and at approximatelythe fifteenth cycle for 3.7M NM solutions. Thereafter, the 0.1Msolution showed a steeper drop of efficiency with cycle than the 3.7Msolution. If significant time were allowed to elapse between cycles,some retardation of the efficiency decline was obtained for thesesolutions. However, this spiking effect was not nearly as marked asfor SO2 .

Although the cycling efficiency could be improved by adding NM,the corrosion current was the same as for pure PC (35 viA/cm 2 over8,000 seconds, for 1.5 coul/cm 2 plated), if the NM were treated for24 hours over lithium dispersion before use. Initial studies, usinguntreated spectroquality NM, gave a corrosion rate of 112 1A/cm2 fora 0. 11M solution over the same time period. These results are plottedin Fig. 11.

Similar to the case of S02 , partially corroded lithium depositsin NM-containing electrolytes show more anodic stripping potentials

20

Fr40.

(00

wd -4

0.00--0

M0 cc0

C: U'

flyw~.41

0~ 'Jo

o o

94

14

co 0 IT i-i

V16)0 A N3314

21

0lo& ,

U

01 0 .J

OD41 4.1

01.

0 0 cu O

"0(00

-4 >

I ~-404KJ a

04C~

0 / In L4

Ici0

41 rWI4 c

4 0 0DN

6 6 0

22

than deposits removed immediately after plating. As the chronopoten-tiogram in Fig. 12 indicates, the stripping curve after partial

corrosion has a small anodic overshoot, not noted ini the case of S02.Such behavior indicates the presence of a resistive film, as withS02, but in this case the film is (partially) destroyed during thestripping cycle.

DISCUSSION

The relative inefficiency of cycling lithium on an inertsubstrate in PC containing ler' than 10 ppm H20 appears due to insula-tion of the deposit as a rsult of its reaction with the solvent, orwith solvent impuriLies. Tnis phenomenon has been well documentedby Selim and Bro (3), who noted that the true corrosion of lithiumdeposited in PC on brass, as determined by chemical analysis, wasmuch slower than that observed electrochemically: true corrosion =0.7 1A/cm 2 ; electrochemical corrosion = 10 pA/cm 2 .

Microscopic examination showL that corrosion reactions oflithium in PC ultimately destroy the homogeneity of the substrate,which results in an effective decrease in the plating surface area.

As the deposit becomes more dendritic with each successive cycle,it appears that all the lithium is plating out in a few small areason the corrosion-product-covered surface. Hence, the cycling effi-ciency decreases rather rapidly with cycle number.

In order to formulate a detailed mechanism of the electrochemicalloss of deposited lithium, the following experimental results must betaken into account:

1. The dipolar additives studied in this work - H20, S02 andNM - increase the overall efficiency of lithium cycling in PC. Thiseffect is noticeable even at low concentration of these additives.

2. The initial loss of efficiency coincides with negligibleloss of ]lihium from the electrode, as determined by chemical analysis(3), and is a result of encapsulation and insulation of 20 to 30% ofthe electrodeposit. This "short term" process is vcry rapid, andevidently occurs during electroplating.

3. A slower "long term" corrosion process takes place at opencircuit stand, which results in further insulation at an initial rateof 43 viA/cm 2 in the present system. Evidently, the "true" corrosionrate of the plated lithium, as measured by Selim and Bro (3) bychemical analysis, is less than 1 -A/cm2 in PC, 1M LiCI04.

4. Within experimental error, the "long term" corrosion rate ofelectrodeposited lithi'm in PC/lM LiCI0 4 iP little changed by theaddition of NM, S02 or H2 0 ( <1200 ppm) (Table 2).

23

pp(12 0co 00

-T 2 00

0 -a

.0

0 0L W .0

0~04 1 --T

2i UO4w

o d c

o 1---

- 0) Q 0)co .

q 0

0 -H' E3 0

+ + Z- P6 -4

4JJ 0

(S&IOA) !-I -SA -lVlN3l0d r

-r-4

24

5. Both the "long term" and "true" (3) corrosion rates decreaseewith time. However, total passivation (ic = 0) was not observed for

this system.

6. The plateau plating and stripping potentials are 'latter inthe presence of relatively high additive concentrations, as can beseen from Figs. 3 and 10. Without additive (Fig. 3), the platingand stripping curves are likely to show an anodic slope with time.

7. No anodic overshoots were noted on stripping in PC/lM LiCl04,

even after a substantial open circuit stand. At relatively high S02

or NM concentrations, the plateau stripping potentials increased by

about 100 mV if the deposit were allowed to corrode partially (Figs. 3

and 10). In NM, an anodic overshoot can be detected upon strippingsuch deposits (Fig. 10).

From these results, it is evident that as-plated lithium is

isolated from the substrate through reaction with solvent species.This isolation, however, results in little actual loss of lithium, butrather is the effect of a coatiig or reaction products on granulesof the metal. Such reaction with the electrolyte appears fastestduring nucleation and initial deposit growth phases. This "short term"corrosion process is the major reason for the overall inefficiency incycling lithium. It is partially suppressed by additives. Visualcomparison of lithium deposits grown in the presence and absence ofadditives reveals greater coherence in the former case. The flatterplating and strtpping chronopotentiograms obtained in the presence ofadditives also indicate a mor: homogeneous deposit.

It is consistent with these result6 that the additives partiallysuppress the reaction leading to encapsulation of lithium with insulat-ing products. It is known that polar siditives can act as a surfaceadsorbent on the plating substrate, leading to a more homogeneousplating surface, and thus a smoother deposit (18). Surface adsorbentscan also lower the reactivity of the substrate to solvent reduction,thus preventing somewhat che development of resistive areas on thesubstrate surface due to the products. This would also result in asmoother morphology. However, subbtrate effects cannot entirelya~count for the termination of growth, and isolation of lithiumgranules during plating, or for the low efficiency of cycling smallamounts of lithium, in the absence of additives. For example, Selimand Bro's (3) long term corrosion rate, I0 iA/cm2 , was similar to

ours, although they use much thicker deposits and a disferent substrate.

The passivation of lithium by S02, to form a Li+-conductive film, iswell known (8), and i is reasonable to postulate that the additivesform Li+-conductiie films on the metal during electrodeposition. Suchfilms, if they rejected PC from the local environment, would permit

25

the growth of larger lithium granules through retardation of thereaction with the solvent which leads to encapsulation.

The similarity of the "long term" corrosion rates for solutionswith and without additives indicates a finite rate of displacement ofthe conductive additive-induced film by the insulating, and by

inference insoluble, electrolyte-induced film. The stripping chrono-potentiograms of partially corroded deposits do show the buildup offilms in the presence of relatively high con-entrations of additives.

However, these films seem as ineffective in -vg-term protection ofthe deposit as the accumulated lithium-solvent -lating reaction

products. The observed reduction of ic with tim, L independent ofadditive, and probably represents both this buildup , reactionproducts, which limits the rate of diffusion into the iL erior ofthe deposit, as well as the greater adherence of some portions of thedeposit than others.

it may thus be concluded that for the PC system describea,evidence is found for the formation of additive-induced passivatingfilms on open circuit stand, but these films are not effective intotally preventing corrosion of the lithium deposit by the solution.

Lithium whicn is strippable after partial corrosion is covered withthe additive-based film, as stripping chronopotentiograms have shown.Partial solu: ility of these conductive films allows their permeationby solvent molecules, ultimately giving rise to reaction of lithiumwith the solvent. If the interaction between the conductive film

and the solvent were of a more phobic nature, total passivation ofthe deposit would have a greater chance of occurring.

It was pointed out by Selim and Bro (3) that solvents with ahigh dielectric constant have some reducible center of positive charge,and hence would be expected to react with lithium. This is especiallytrue of the freshly nucleated material. Hence, solvents for ambienttemperature, secondary lithium batteries should be sought which formeither Li+-conductive, passivating films on tne metal, or which, in

combination with certain additives, favor the establishment of aLi+-conducti-*e *--uly passivating film. PC, even Lhough unreactivetoward lithium on a massive scale, fulfills neither of th..se criteria.

REFERENCES

1. R. Jasinski, Adv. Electrochem. Electrochem. Eng., 8, 253 (1971).

2. M. Salomon, J. Phys. Chem., 73, 3299 (1969); J. Electroanal.Chem., 26, 319 (1970).

3. R. Selim and P. Bro, J. Electrochem. Soc., 121, 1457 (1974).

26

4. J. N. Butler, D. R. Cogley and J. C. Synnott, J. Phys. Chem., 13,4026 (1969).

5. W. K. Behl, J. A. Christopoulos, M. Ramirez and S. Gilman, ECOMReport 4101, April 1973.

6. W. K. Behl, J. A. Christopoulos, M. RamirLz and S. Gilman, J.Electrochem. Soc., 120, 1619 (1973).

7. J. J. Auborn, K. W. French, S. I. Lieberman, V. K. Shah and

A. Heller, J. Electrochem. Soc., 120, 1613 (1973).

8. D. L. Maricle and J. P. Mohns, Fr. Demandd, 2,014,160 (1970).

9. J. Chilton, W. Conner and R. Holsinger, Contract AF 33(615)1195,Report No. 2, July 1974.

10. D. Semones and J. McCallum, Proc. 24th Power Sources Conf. (1970),p. 16.

11. M. Nicholson, J. Electrochem. Soc., 121, 734 (1974).

12. J. Gabano, G. Gerbier and J. Laurent, Proc. 23rd Power SourcesConf. (1969), p. 80.

13. M. P.ao and K. Hill, Abstracts or the Meeting of the Electro-

chemical Society, Buffalo, New York (1965), p. 40.

14. R. Jasinski, Final Report, Contract NOW 64-0653-F, July 1965.

15. A. N. Dey, J. Electrochem. Soc., 118, 1547 (1971).

16. J. Chilton, W. Conner and R. Holsinger, Ibid. Report No. 1.

17. B. Burrows and S. Kirkland, J. Electrochem. Soc., 115, 1164 (1968).

18. J. Bockris and A. Despid, in Physical Chemistry: An AdvancedTreatise, ed. by H. Eyring et al. (New York: Academic Press, 1970),pp. 611-730.

27