Thermal Runaway SeverityReduction Assessment& Implementation:On LREBA & LPGTByEric Darcy/NASA-JSCFor2015 Space Power WorkshopManhattan Beach, CA12-14 May 2015

Team and Contents Battery Test Team (Subteam to the NESC Led TR SeverityReduction Effort)– Eric Darcy, Test Requestor– Christina Deoja, Laura Baldwin, Dereck Lenoir and Mike Salinas, TestDirectors– Jim Rogers, Bob Bohot, John Weintritt, and Minh Tran, Battery designand assembly– Mark Schaefbauer, Soft Goods Design and Assembly– Tony Parish, Henry Bravo, Bill Holton, and Pete Sanchez, Test Agenda––––––––––Background on LREBA, LPGTSingle cell TR trigger method selected and whyVerification of subscale mitigation measuresFull scale LREBA with those measures leads to catastrophic hazardConsequence of cell TR ejecta products for TR propagationFull scale LREBA with adjacent cells protected from cell vent pathLREBA with severity reduction measures soft goods enclosureFull scale LPGT baseline “CDR” design leads to catastrophic hazardFull scale LPGT with severity reduction measures worksSummary conclusions – No propagation with no flames/sparks ispossible with minimal mass/volume penalty2

Background - Li-ion Rechargeable EVA BatteryAssembly (LREBA)1Current Design Baseline – April 2014 9P cell banks with cell glued in picket fence array ¼” Ni-201 tabs interconnecting cells Cell vents oriented towards edge of housing tray Cell banks wrapped in 1/8” thick Nomex felt Only vents on enclosure are for pressure equalization9P-5S Array of Samsung 2.6Ah 18650 cells to power thespacesuit helmet lights and camera and glove heaters3

LPGT “CDR” Baseline DesignBaseline Design Features (April 2014) 10S array for discharge into Tool 2P-5S array for charging Cells wrapped in PVC shrink wrappedand Nomex paper tube sleeve Cell glued together in contact withepoxy ¼” wide Ni-201 tabs (0.005” thk) forcell interconnects No battery vent4

Selected Bottom Patch Heaters For Triggering TR Two small (3/4”x3/4”) patch heaters located on the bottom ofcylindrical can– Nichrome wire glued to Mica paper– Adhered to bare can by cement bases adhesive Each has 6” of Nichrome wire for a total of 12” per pair– Pair can be powered by up to 90W Main benefit of design – more relevant cell internal short– Deliver high heat flux away from seals, PTC, and CID located in cell header– leaves an axial bond line undisturbed for gluing cell together in one plane– More likely to result in coincident cell venting and TR runaway5

Cell TR Response vs Heat Power38W15W60090W 30W45WTemp, degC50030W40060W300Temp vs time profiles of the TR eventSamsung ICR18650-26F withbottom patch heaters at various powersSide probe TC200100800100012001400time, s TR output heat fairly independent of heater input power High power preferred to reduce risk of biasing hot adjacent cells6

Higher W triggers with Lower Wh Input7Plot courtesy of Bruce Drolen/BoeingLower Energy, Wh, input into the heater presents lower risk of biasing adjacent cells

LREBA 9P Bank Test – Baseline Design8 Picket fence 9P bank withcells in axial contact and withepoxy bond line between cells– End cell trigger with 45W– Open air environment Full cascade of cell TRpropagation in about 10minutes

First Round of Mitigation Measures Ensure cell-cell spacing 1-2mm withFR4/G10 capture plates– Reduce thermal conduction from cell tocell Integrate fusible links into Ni-201 busplates on positive only– Isolate cell with internal shorts fromparallel cells– 15A open current– Reduce thermal conduction viaelectrical connection Include radiation barrier betweencells in 2mm spacing designTest under inert gas– Reduce chaos associated with burningcell ejecta (electrolyte & solids) Results– No TR propagation in all 4 testsconducted in inert gas Radiation barriers helped slightlyBut spacing between cells found mostsignificant–Picket fence design propagated in inertgas– In open air, propagation was likely dueto flammable ejecta impinging onadjacent cells9

1st Full Scale LREBA Test T 5:07 min - FirstCell TR T 16:36 min - Firstflames outsidehousing T 30:28 min – Duringfull TR Propagation T 34:00 min – Final TR10

Bank 1 experienced a sustained short immediatelyafter TR of trigger cell11

1stFull Scale Battery Test – Total PropagationEnd cell in corner of dogleg was triggered.All 45 cells went into TR over 29 minutes.4.5 minutes from trigger cell TR to adjacent cell TRFlames exited housing after 5th cell driven into TR 11minutes into the testVented ejecta bypassed fusible links and createdshort pathsTriggerCell12

Cell TR Ejecta Assessment Trigger cell next to 0% SoC cell without bus bars and with bank inside LREBAenclosure to assess if TR ejecta can electrically bridge between cellsCell 2 appears to have been thermally overstressed causing its sealing gasket tomelt–––– 13This cell definitely vented, albeit without going into TRInternal gas temperatures in inside LREBA enclosure exceeded 230 CAll other cells, #3-9 have healthy OCVsCell TR ejecta found to be electrically conductiveEither way, it indicates that re-design features should– Manage the cell TR ejecta to prevent collateral damage– Reduce trip current of fusible links, but more importantly– Fusible links are more effective on negative terminal, away from vent product path

Major Contributors to Propagation14 Tests, our analysis, and other research identifiedthree key contributors– Cell-to-cell heat transfer Cell-to-cell conduction via contact, through structure,bonding, and electrical interconnects Cell-to-cell thermal radiation found to be a contributor, butnot leading– Electrical shorts causing adjacent parallel cell heating– Violent release of high temperature gas/liquids/solids(TR ejecta) with exothermic reactions Bottled up in an enclosure with no place to go Test indicates that we must prevent the firstpropagation of TR, otherwise there is little hopestopping the runaway train

15Next Full Scale LREBA Test ConfigurationCell Ejecta Exhaust Piped Top Macor (machinable glassceramic) Matching exhaust ports inhousing for pipes Mica paper wrapped on cellcans Fusible bus bars on bothpositives and negatives Same 15A trip9P bank inside LREBA housing with exhaust holes

More Photos of Mitigation FeaturesMachinable glass ceramic (Macor )Fusible (15A) bus plates connected onboth terminalsMica paper as radiation barriers and toelectrically isolate cell cans 2-8Heater placed on end cells 1 & 916

Next Full Scale Test - Pre Test PhotosOne active 9P bank in dogleg with endcell trigger heaters powered at 90W4 dummy banks uncharged to take upvolume inside enclosureAl foil covering housing ejecta holes tolimit air circulation and prevent FODentering17

18No TR PropagationHalf of heater fails open in first second, heater runs at 45W, nevertheless, TR reached in 72s. Bottom of triggercell reaches 543 C, while mid and top get to 319-344 C. Cell 2 maxes out on all 3 TCs at 100 C.

Trigger Cell Positive Fusible Link OpensAt video time 13m:18s19

20Cell VentingAt video time 13m:19s

21Trigger Cell TRAt video time 13m:20s

Post Test Photos Bank voltage at 4.07V Isolated Cell 2 voltage measured at2.5V (blown positive fuse) on 8/27 and1.1V on 8/29 Internal soft short suspected Megaohms measured between cell 4-9cans and housing Negative fuse on trigger cell also foundblown22

23Cell Mica Paper Wrap Heat Affected ZonesAdjacent cells 8 and 2 showing significant heat affected zones Burn marks indicate cell 8 was more impacted than cell 2 bottom nearheater: Suggest that our heater edges may be too close to the adjacent cells Moving to a single 45W heater (1”x0.5”) placed on bottom side of triggercell opposite the adjacent cell Burn marks on top of the mica paper similar on both cells: Indicates some bypass of ejecta between the cell and G10 capture plate The epoxy must be melting and need to go to a higher temp epoxy

Lesson Learned and Next Design Iteration24 Redesign LREBA parts– Locate thinner, taller heater to side opposite ofcell from adjacent cell– Add mica half cylinder to the trigger cell to protectadjacent cell– Use high temp epoxy to bond cells to captureplates– Plug all housing holes with Al foil tape– Route all TCs away from trigger cell Next run – same as previous except triggercell 1 and add soft goods bag

Next Run – Pre testSoft goods bag added to dog leg of the battery25

Trigger Cell Vents Smoke/Electrolyte26

1s later, Instantaneous fireball withsparks27

After 2s, Flaring28

Flaring lasts for 3s, then small flame for 15s20s from smoke vent to flame out29

Overall Plot30Steady OCV - No soft shortsTR of trigger occurs 310s after heater on. Trigger cell max temp range from 474-631C.Cell 2 max temp range from 147-317C. Cell 3 max is 80C 600s after heater on.Trigger pipe max is 1146C.

Run 57 – Post TestTrigger cell ejecta burn hole31

Post Test – TMG Bag Soft goods bag (rip stop nylon, 7 layers of aluminized mylar, and kevlarreinforced fabric) was quickly perforated by cell TR flare/flame Need to reinforce it with higher temperature metal foils (ex, Ni) Need to dissipate the heat of the flame/flare with a porous media32

ESLI Carbon Fibercore Torch Test Lightweight tiny carbonfibers glued to Al foils– Very surface area of very highthermal conductivity material– Sample tested was ¼” thick Blow torch flame did notpenetrate through sample– Even after 10 secondapplication33

34Full Bag DesignBeta Beta cloth (Teflon reinforced fiberglass)Ni Nickel 201 alloy (annealed) 0.001” ips of ESLI a/Ni/Beta/Ni/BetaBeta/Ni/Beta/Ni/BetaOpening for TCs

Run 58 Pre TestWith Carbon Fibercore (CFC)35

Run 58 – Overall PlotSteady OCV - No soft shortsTrigger fuse blows at 168s, TR in 169s, heater left on for 176s,max temp on trigger cell pipe 1224C36

Snap Shots – Initial Vent of Smoke37

1s later38

2s later39

3s later40

41Post PicturesTrigger cell around cornerBottom edge of batteryCarbon Fibercore (CFC)Large bag flapSmall bag flap

Post PicturesFlame arrestingCarbon fibercore (CFC)42

Run 59 – Without the CFCCell TR ejecta burns right through 2 layers of Ni foil (0.001”)43

44Focus on TRTrigger pipe reaches 1146C, trigger cell skin max T range 534-722 C,Cell2 range 144-163 C, Cell 3 reached 64 C. Note that it takes 800s to cool trigger cell from peak to 100 C

45Al Heat Spreader (run 60-61)Top and bottom heat spreaders connects every other cell thermally

46Runs 60 – 61 – No sparks, no fire exit bag Bag internal layering reinforced with 4 layers of Ni foil opposing cellexhaust ports Bag overlap layering added at corners to prevent exiting sparks Heat spreader conducts heat to enclosure and reduces maxtemperature and duration of trigger cell

Plot of run 60 – Heat Spreader & CFC47Steady OCV - No soft shortsHeater power bumped up from 45 to 55W just prior to TR, which occurs 10.6 minutes after heater turn on. Much longer to drive TR.Trigger cell max temp range is 294-408 C, Cell 2 is 104-122 C, and cell 3 reaches 74 C. Cooler Ts with heat spreader except for cell 3.The heat spreader reaches 173 and 94 C near the trigger and cell 3, respectively.

Run 61 – Heat Spreader, no CFC48Steady OCV - No soft shortsHeater set at 50W and on for 329s. TR occurs in 320s. Internal short circuit occurs 147s after heater on, possibly venting. ThenTR occurs 57s later. Max Ts on trigger cell range is 555-686C, cell 8 is 110-115C, and cell 7 reaches 76C. Note that it takes 449sfor max temps to cool from peak to 100C. Heat spreader does not keep trigger cell as cool, but does protect adjacent cell.

Run 61 – No CFC TR ejecta burns through first Ni layerand damages second layer 3rd and 4th Ni layers are undamaged Ni melts at 1455 C Adjacent cells retained OCV 4V DPA of adjacent cells from runs 60 &61 indicated no heat effected zoneson jellyroll plastic wrap49

Recap of Mitigation Measures for LREBA Control the conduction paths– Ensure cells are space out 1mm G10 capture plates– Decrease conduction of cellinterconnects Fusible links– Increase conduction to the enclosure Heat spreaders and gap padsLimit shorting paths– Fusible links in the negative cellinterconnects– Mica paper sleeves on cell cans Control the TR ejecta path to protectadjacent cells– Seal cell positives to capture plates withhigh temperature adhesive to preventbypass of hot gases– Protect materials in ejecta path withceramic pipes and exhaust ports Limit the flare/fire/sparts exiting thebattery enclosure– Flame arresting screen to cool the hotgases leaving the battery exhaust ports50

LPGT “CDR” Baseline TestFull thermal runaway propagation to all 9 other cells over about 10 minutes with release of sparks51

Snapshot – more sparks at 9:56More sparks occur at 10:20, at 12:37 the supporting tile cracks, and smolderingsmoke is intense for another 5 min.52

LPGT with Severity Reduction Features53

54LPGT Assembly PhotosBrick Details Cell wrapped in 0.004” mica with ¼” overlap Inserted into Al heat spreader block with allcells same orientation Block leave ½” bottoms of cell can exposedfor heaters and TCs Trigger cells: E2 and E10 Adjacent cell TCs stuck through micawindows TCs on trigger cells placed on negative tabs Tab/wire/fuse assemblies routed throughhalf cylinder cut outs and center hole ofblock left open

55LPGT Assembly PhotosAssembly Details Cell vents point towards dischargeconnector Most of cell ejecta routed through centerhole of heat spreader towards baffle ofbracket 2 orthogonal layers of 1/16” CFC and 2 thinstainless screens per vent port Housing vent port geometry changed toprovide structural support to the screens Only one 0.02” gap pad fit – placed betweenblock and lid

56Run 13 Pre-Test Photos Lid/housing seated firm with four 6-32screws Discharge connector opening filledwith wires and sealed with caulk Vent ports, lid, and housinginstrumented with TCs

LPGT Run 13 – Close upE2E3E4E10E1E6E5E7E8Positive endE9Unfortunately, we lost TCs of adj cell 3 and cell 6.From hottest to colder: E2, inside near disch connector, E1, E10, E8, Trigger side vent port57

Video Snapshot – No sparksA split second after trigger cell TR, most of the vent cloud exits the vent ports, therest exits the housing to lid joint – All adjacent cell OCVs unaffected by test58

Lessons Learned To Date59 Design must prevent first TR propagation from initial failed cell: Entire battery gets hotter with each subsequent cell TR event Limiting cell-to-cell thermal conduction appears to work: Spacing out the cells 1mm is very beneficial Parallel cell bussing can provide significant in-rush currents into failed cell,which gets them hot: Individually fusing parallel cells is effective 18 LREBA and 7 LPGT full scale tests with no propagation Last 8 LREBA runs with interstitial materials reduced adjacent cell max temp, resulted noOCV decline even with 20% higher energy density cell design Soft goods bag needs reinforcements Additional Ni foil layers help, but flame arresting carbon fibercore found to be more effective Managing the vent/ejecta path is critical: Combustion of expelled electrolyte must be directed away from adjacentcells with path sealed good high temperature materials & joints Cell TR ejecta can bridge to adjacent cells and cause cascading shorts(suggests need for interstitial material between cells to protect cell cans) Cell TR flame/flare attenuation with SS screens and carbon fibercoreprotected by baffle and tortuous vent path works Subscale test results can be misleading and no replacement for full scale testverifications

Acknowledgements Chris Iannello/NASANESC for the vision thatTR severity reductionwas possible withoutbig performanceimpacts and securingfunding for the effort– Fiery 1st full scale testresult– By end of Oct, we’veachieved 8 runs with nopropagation Last 2 runs with no sparks,flames, or flare leaving theenclosure20 June 201430 Oct 201460

Acknowledgements (cont.) TR Severity Reduction Team– Paul Shack, Assessment Lead– Chris Iannello, NESC Technical Fellow forElectrical Power, and Deputy, Rob Button– Steve Rickman, NESC Technical Fellow forPassive Thermal– Eric Darcy, Test Lead for EVA Batteries, NASAJSC– Sam Russell, Mike Fowler, Judy Jeevarajan,Craig Clark, John Weintritt, Christina Deoja andStacie Cox/NASA-JSC– Bob Christie, Tom Miller, Penni Dalton/NASAGRC– Dan Doughty, Bruce Drolen, Ralph White, GaryBayles, and Jim Womack/NESC Consultants61