Electrically Rechargeable Metal-air Batteries Compared to Advanced Lithium-ion Batteries Jeff Dahn NSERC/3M Canada Industrial Research Chair Depts. of Physics and Chemistry Dalhousie University, Canada 1 There are no fundamental scientific obstacles to creating batteries with ten times the energy content - for a given weight - of the best current batteries. From the program of the Almaden Institute But should this really be the focus of our efforts? 2 Hymotion 5kWh LiFePO4-based Li-ion pack About 30 km allelectric range It’s a pretty big battery pack! 3 “Lithium-ion cells have high energy density” Where has it gone? System Cell V A-hr A123 26650 3.3 LiFePO4 Panasonic 18670 1.2 Ni-MH E-One 26700 3.8 Moli LiMn2O4 2.3 Volume Mass Wh/L (mL) (g) 35.5 70 215 3.2 17.0 60 225 2.9 37.2 101 232 Panasonic 18650 3.7 LiCoO2 2.55 16.5 46.5 570 4 Material Crystallographic density (g/cm3) Average potential (V) (vs Li) Reversible specific capacity (Ah/g) LiCoO2 5.05 3.9 0.15 (to 4.2 V) 2.95 Li[Ni1/3Mn1/3Co1/3]O2 4.77 3.7 0.163 (to 4.3 V) 2.87 Called NMC Volumetric energy density (Wh/cm3) Li1.1Mn1.9O4 4.18 4.1 0.120 2.06 LiFePO4 3.60 3.44 0.160 1.98 LiFePO4 will lead to the lowest energy density of any of the popular positive electrode choices! LiMn2O4 is not much better. Why are they being selected? 5 “In our opinion, the cathode materials can be ranked from most safe to least safe in the following order LiFePO4, Li[Ni1/3Mn1/3Co1/3]O2, Li1+xMn2-xO4, LiCoO2, Li[Ni0.7Co0.2Ti0.05Mg0.05]O2, LiNi0.8Co0.2O2, LiNiO2.” D.D. MacNeil et al. - Journal of Power Sources 108, 8–14, (2002). Adding Al to Li[Ni1/3Mn1/3Co1/3-zAlz]O2 improves the safety even more! See next slide. 6 dT/dt (ºC/min) 100 10 LiMn2O4 charged to 4.3 V 1 0.1 100 10 NMC charged to 4.0 V 1 0.1 100 10 NMC charged to 4.1 V 1 0.1 100 10 NMC charged to 4.2 V 1 0.1 100 10 NMC charged to 4.3 V 1 0.1 100 10 z = 0.1 charged to 4.3 V 1 0.1 0.01 100 200 Temperature (ºC) b a z = 0.1 is a very good compromise between capacity and safety. Safer than LiMn2O4!!!!! 300 7 Where do we stand now? Material Density (g/cm3) Average potential (V) Reversible specific capacity (Ah/g) Volumetric energy density (Wh/cm3) LiCoO2 5.05 3.9 0.15 (to 4.2 V) 2.95 Li[Ni1/3Mn1/3Co1/3]O2 4.77 3.7 0.163 (to 4.3 V) 2.87 Li1.1Mn1.9O4 4.18 4.1 0.120 2.06 LiFePO4 3.60 3.44 0.160 1.98 Li[Ni1/3Mn1/3Co0.233Al0.1]O2 4.60 3.75 0.140 (to 4.3 V) 2.42 8 Summary: Li[Ni1/3Mn1/3Co0.233Al0.1]O2 stores almost as much energy as LiCoO2. It appears to be safer than Li1.1Mn1.9O4 and rivals LiFePO4. Lots of scientists are working on SAFE materials that are better than LiFePO4. I am confident that energy density of computer cells can be realized in SAFE Li-ion cells for vehicles. TESLA USES SUCH CELLS NOW. Therefore – LiCoO2/graphite numbers are valid for comparison to new technologies. 9 Air-fueled Battery Could Last Up to 10 Times Longer by Engineering and Physical Sciences Research Council Swindon, UK [RenewableEnergyWorld.com] A new type of air-fueled battery could give up to ten times the energy storage of designs currently available. "The key is to use oxygen in the air as a re-agent, rather than carry the necessary chemicals around inside the battery." -- Peter Bruce, Chemistry Professor, University of St Andrews This step-change in capacity could pave the way for a new generation of electric cars, mobile phones and laptops. The research work, funded by the Engineering and Physical Sciences Research Council (EPSRC), is being led by researchers at the University of St Andrews with partners at Strathclyde and Newcastle. 10 Lithium–Air Battery SJ Visco, E Nimon, LC De Jonghe, PolyPlus Battery Company, Berkeley, CA, USA, and Lawrence Berkeley National Laboratory, Berkeley, USA Elsevier B.V. All rights reserved. Introduction The large free energy for the reaction of lithium with oxygen has attracted the interest of battery researchers for decades. At a nominal potential of about 3V, the theoretical specific energy for a lithium/air battery is over 5000 Wh kg-1 for the reaction forming LiOH (Li + 1/4O2 + 1/2 H2O ↔ LiOH) and 11,000 Wh kg-1 for the reaction forming Li2O2 (2 Li + O2 ↔ Li2O2) or for the reaction of lithium with dissolved oxygen in seawater, rivaling the energy density for hydrocarbon fuel cells and far exceeding Liion battery chemistry that has a theoretical specific energy of about 400 Wh kg-1. 11 Where does 11,000 Wh/kg come from? Start with Li - (6.9 g/mole) Assume oxygen is “free” and we don’t need to worry about its mass. Spec. E. = 3 V x 96500 C/mol 3600 C/Ah 0.0069 kg/mol Spec. E. = 11,500 Wh/kg OK – we understand the assumptions. 12 But surely this is unfair and we need to count the mass of the formed Li2O2, especially if we’re talking about a rechargeable battery. Li Li2O2 Vol. E. = 3500 Wh/kg (assuming only Li2O2) 13 Zn-air is fundamentally different to Li-air because ZnO14 forms in the same space as the Zn occupied. Zn-air and Li-air are fundamentally different. Zn-air is a “one compartment cell” 830 Wh/kg Li Li2O2 “3500” Wh/kg LiCoO2 ZnO Zn graphite Li-air is a “two compartment cell” 15 390 Wh/kg Hard to compare based on mass, since don’t know what mass the space for the Li2O2 and the peripherals will have. Instead, compare based on volume: [N.B. – Volume matters for a vehicle and for energy storage applications. This was well pointed out by Ted Miller yesterday] Assumptions: LiCoO2 – 140 mAh/g, 5.05 g/cc, 3.8V Li-ion cell Li2O2 – 1165 mAh/g, 2.3 g/cc, 3.0 V Li-air cell Li – 3860 mAh/g, 0.51 g/cc Graphite – 350 mAh/g, 2.2 g/cc Use Feynmanstyle approach! 16 LiCoO2 separator Bulk Li-ion graphite Li2O2 separator Li Bulk Li-Air Assume no porosity required Assume no porosity required Assume perfect, zero thickness separator Assume perfect, zero thickness separator Assume 3.0 V Assume 3.8 V Assume Li thickness = x Assume graphite thickness = x Li2O2 needs to be = .74x LiCoO2 needs to be 1.15x 17 LiCoO2 separator Bulk Li-ion graphite Li2O2 separator Li Bulk Li-Air Assume no porosity required Assume no porosity required Assume perfect, zero thickness separator Assume perfect, zero thickness separator Assume 3.0 V Assume 3.8 V Get 3400 Wh/L Get 1450 Wh/L 18 But Li metal does not cycle efficiently. Need to have excess lithium. This amount is debatable but we will pick a three-fold excess. K. Brandt Solid State Ionics 69 (1994) 173-183. “Through a combination of these measures, cells of the Canadian company Moli Energy achieved in the late 80's lithium cycling efficiencies of over 99% [ 19 ] translating into a life cycle of about 300 cycles with a lithium excess according to Eq. (2) of R= 3.” This means 4 times what is really used. 19 LiCoO2 separator Bulk Li-ion graphite Li2O2 separator Li Bulk Li-Air Assume no porosity required Assume no porosity required Assume perfect, zero thickness separator Assume perfect, zero thickness separator Assume 3.0 V, Assume 3.8 V Assume 3-fold excess Li Get 1450 Wh/L Get 1254 Wh/L 20 5000 2000 1000 Li2O2 3000 separator 4000 Li Density (Wh/L) Volumetric Energy 6000 0 0 1 2 3 4 Lithium excess Assume no porosity required Assume perfect, zero thickness separator Assume 3.0 V, ASSUME Li2O2 is zero thickness (This accounts for possiblity of Li2O). Assume n-fold excess Li 21 0 1 2 3 Li2S separator Li Volumetric Energy Density (Wh/L) 4000 3500 3000 2500 2000 1500 1000 500 0 4 Lithium excess Assume no porosity required Assume perfect, zero thickness separator Assume 2.0 V, ASSUME Li2S is zero thickness. This slide is for Linda Assume n-fold excess Li 22 Therefore, the Li-air and Li-S batteries are all about Li. In order for these batteries to be volumetrically useful, one needs to basically have a very small lithium excess. This means learning how to strip and plate lithium effectively which was studied intensely in the 70’s, 80’s and early 90’s. Maybe we’re smarter now. 23 Li/MnO2 rechargeable cell, Moli Energy about 1989 [Is this ancient history, irrelevant to the present day? Am I a dinosaur?] 24 - Dec. 1988 - 2 Million Li/MoS2 cells in the field (NEC laptop and NTT Cell phone) - Spring 1989 - Li/MnO2 cell ready to go - Spring 1989 - Safety incidents in the field - Summer 1989 - We understand the problems - Summer 1989 - Complete recall of all phone packs and Moli goes into receivership. - Spring 1990 - NEC and Mitsui buy Moli. 25 The Safety Problem (Cells with Li metal anodes) Li is not thermodynamically stable in non-aqueous electrolytes 2 Li + EC → Li2CO3 + ethylene -69 kcal/mole Li H2 + 1/2 O2 → H2O -56 kcal/mole As cells cycle, the surface area of the metallic Li increases without limit because the plated Li deposit is not compact [This is Elton’s Gray Layer]. After about 25 to 50 cycles under certain conditions, cells cannot withstand temperatures above 120oC. The passivation mechanism breaks down and thermal runaway occurs26 Safety of metallic Li gets worse and worse with cycle number. Even though cells can pass oven test at the beginning of life, mossy cycled Li is very reactive. This is why the whole industry switched to Liion in 19891990. 27 U. von Sacken et al. / Journal of Power Sources 54 (1995) 240-245 Nevertheless, NEC and Mitsui insist on the "Confirmation Test“ when they purchase Moli in 1990. 500,000 cells made, each individually x-rayed for manufacturing defects. 50,000 cell phone battery packs built and cycled under low rate conditions. (Over 5000 chargers built for the task.) After 1.5 years of assembly and testing many serious failures. NEC and Mitsui decide to abandon Li metal cells FOREVER. Even Hydro Quebec eventually abandoned Li metal cells with polymer electrolyte. 28 LiCoO2 separator Bulk Li-ion graphite Li2O2 separator graphite Bulk Li-Air with graphite negative Assume no porosity required Assume no porosity required Assume perfect, zero thickness separator Assume perfect, zero thickness separator Assume 3.0 V Assume 3.8 V Assume 60 micron graphite Assume 60 micron graphite Li2O2 needs to be 19 microns LiCoO2 needs to be 69 microns29 LiCoO2 separator Bulk Li-ion graphite Li2O2 separator graphite Bulk Li-Air with graphite negative Total stack height = 79 micron Total stack height = 129 micron Voltage = 3 V Voltage = 3.8 V Same capacity since same graphite Same capacity since same graphite 1880 Wh/L 1450 Wh/L 18650 cells get 600 Wh/L now. 30 Comment: Rechargeable Li-air and rechargable Li-S will be very very challenging. Even if graphite (or Si) is used there are a whole host of problems. [Protecting the negative, Li excess, air handling, etc.] Don’t bet the farm. Simple Li-ion today – without advanced electrodes, can be as volumetrically efficient as Li-air in my opinion. With advanced electrodes, it can be better. 31 Zn-air is fundamentally different to Li-air because ZnO32 forms in the same space as the Zn occupied. Bulk Zn-air Assume: Zn + ½ O2 ↔ ZnO 1.2 V delivered Perfect separator (zero thickness) ZnO Perfect catalyst (zero thickness) No current collectors No porosity Vol. Energy Density = 0.661 Ah/g * 5606 g/L * 1.2 V = 4400 Wh/L Even the tiny Duracell DA13 gets 1756 Wh/L ! 33 This looks pretty good for a typical Li-ion computer cell. But it’s not a Liion cell. It is a NiZn cell, with a Zn electrode. T. C. Adler, F. R. McLarnon, and E. J. Cairns, J. Electrochem. Soc. 140, 293 (1993). These 1.35 Ah cells used a novel electrolyte designed to give better stability to the Zn electrode. Doesn’t this prove that the Zn electrode can be cycled?? 34 Zinc-nickel oxide battery.--The Zn/KOH/NiOOH cell is based on dissolution-precipitation reactions at the Zn electrode: Zn + 4 OH- ↔ Zn(OH)42- + 2eZn(OH)42- ↔ ZnO + 2 OH- + H20 This is the same reaction as happens in the Zn-Air cell. “Within the limits of our experimental parameters and experimental reproducibility and error, the optimum composition is 3.2-4.5 M KOH, 1.8 M each KF and K2CO3, and 0.5 M LiF in suspension, saturated with ZnO.” - From a 1998 paper by the Cairns group. 35 Potential (V) 2.5 2 1.5 4 mA discharge and charge Hysteresis 1 0.5 0 0 40 80 120 Time (hr) 160 200 Discharge and charge of a commercial #675 size (nonrechargeable) Zn-air cell. 4 mA current. Cell weight = 1.8 grams. Note that the cell is somewhat rechargeable 36 How Zn-Air Works (Briefly) The discharge reactions are as follows. • At the Zn electrode: Zn + 4OH − ← ⎯→ Zn(OH )4(aq ) + 2e − −2 Zn(OH )4(aq ) ← ⎯→ ZnO( s ) + 2OH − + H 2O −2 •At the Carbon electrode: O2 + H 2O + 2e − ↔ HO2−(aq ) + OH − Need Catalyst HO2−(aq ) ↔ 12 O2 + OH − Need Catalyst 37 Cyclic Voltammetry Experiments 1 2H2O O2+4e-+4H+ I (mA/cm2) 0 0.8 V -1 -2 Pt catalyst in acid O2+4e-+4H+ 2H2O Scan in Ar Scan in O2 -3 -4 -0.4 0 0.4 0.8 1.2 E (V vs. RHE) 1.6 2 The hysteresis in the cell arises mostly at the air electrode side. The oxygen evolution reaction and the oxygen reduction reaction are separated by 0.7 to 0.8 V due to activation barriers. 38 Let’s talk about efficiencies: Zn-air – Charge 2.0 V, Discharge 1.2 V 60% Li-air – Charge 3.8 V, Discharge 3.0 V 79% [I have given the Li/air guys the benefit of the doubt here as well] Li-ion – No true hysteresis – 95% under real conditions 39 What really matters for automotive and energy storage? Volumetric Energy Density – We discussed this. Efficiency – cost of a charge, don’t want to waste renewable energy. Can be addressed by Univ. and Govt. labs thru new materials. Cost of batteries – can this be addressed by universities and govt. labs? Safety – hard to address by universities and govt. labs. Cycle life – at the automotive and energy storage level > 3000 cycles, need real cells - so hard for 40 universities and govt. labs. (OR IS IT?) Automotive and energy storage cells require excellent cycle life (> 3000 cycles) There have been numerous papers and patents published on electrolyte additives, positive electrode material coatings, new electrode materials, etc. that lead to better cycle life. e.g. SEI modifiers on the negative electrode side – vinylene carbonate, vinyl acetate e.g. H2O and HF scavengers to reduce transition metal dissolution on positive electrode e.g. AlF3 and other coatings on NMC to improve cycle life to high potential, etc. etc. etc. 41 The cycle life of rechargeable cells is not infinity because small fractions of cell components are consumed during each cycle. The amount of these components consumed can be measured using the “coulombic efficiency” (CE): CE = Qc/Qd = [Charge in]/[Charge out] It is of utmost importance to be able to measure the coulombic efficiency accurately, but traditional battery charger systems cannot. 10,000 cycles need at least 99.99% CE e.g. 0.999910000 = 0.367 42 Typical Literature Data Look at the scatter! 43 High Precision Charger (HPC) Used to measure the columbic efficiency of different electrodes and electrolyte additives Delivers current with an accuracy of 0.03% Apparatus Keithley 220 current source Keithley 2750 digital multimeter Omega temperature controller Temperature controlled box Custom labview software program The HPC has many impressive characteristics! 44 LiN2/3Mn1/3O2 Cycled on the High Precission Charger 1.04 LiNi2/3Mn1/3O2 Cycled on a traditional battery charger 1.04 1.03 1.03 1.02 1.02 Qc/Qd Qc/Qd Small noise Large noise 1.01 Large shift from 1.00 1 0.99 10 20 30 Cycle # 40 50 0.03% charge accuracy over 1 year, better per cycle 1.01 1 >1% charge inaccuracy 0 No shift from 1.00 60 0.99 0 10 20 Cycle # 30 40 Our high precision charger is the best in the world, but still not what we ultimately need. 45 Moli Sony 1000 Capacity (mAh) 3000 800 2000 600 400 1000 200 Capacity (mAh) 0 2420 Qdn+1/Qc Moli LiCoO2/graphite 0 840 2410 820 compared to 2400 800 2390 2380 1 780 1 0.999 0.999 0.998 0.998 0.997 0.997 0.996 0.996 0.995 0 4 8 12 16 20 0 4 8 Cycle Number 0.995 12 16 20 Sony Nexelion LiCoO2-NMC/Sn-TM-C All cycled at C/24 rates 46 1 Qd n+1/Qc 0.999 0.998 0.997 0.996 0.995 Moli Graphite 0 5 10 15 20 25 0 5 Cycle Number Sony Co30Sn30C40 10 15 20 25 Li-ion cells compared to corresponding ½ cells 47 Example: NMC positive electrode material NMC Li-ion Cells cycled to > 4.2 V sometimes show short life. Graph below predicts this. Potential solution: additives improve the CE of NMC charged to 4.6 V AS4015 & AS3151 Series NMC positive electrodes AS4023 Series - Varying Silica Additive (SiO2) 1.02 0% SiO2 1% SiO2 4% SiO2 1.015 1.02 3.0 - 3.8V 3.0 - 4.0V 3.0 - 4.2V 3.0 - 4.4V 3.0 - 4.6V 1.012 Qc/Qd Qc/Qd 1.016 1.008 1.005 1.004 1 1.01 0 10 20 30 Cycle # 40 50 1 0 5 10 15 20 Cycle # 25 30 35 48 40-channel highprecision charger operational at Dalhousie (taken August 22) You can’t do the science without the tools We need systems 10x better than this!! IBM (and Si Valley) can surely help here! 49 Concluding remarks 1. There is a lot do. Lithium-ion is pretty hard to beat volumetrically AND it will improve incrementally over time as well (See Talk by Mike Thackeray) 2. University (and other) researchers need to “get relevant” to help with automotive batteries. 3. Focus on what’s important and learn from the past. Do not use metallic lithium in rechargeable batteries (my STRONG opinion). Don’t bet the farm on air and S-systems. 4. Hold more events like this. 50 Specfic Energy - Wh/kg Taken from the Sion Power website August 24, 2009 - Not more volumetrically efficient than Li-ion. What about 51 safety? Why no products after decades? GM’s Chevy Volt will be a plug-in hybrid using Li-ion batteries. The batteries considered were graphite/LiFePO4 Li-ion batteries made by A123 Co. 52 or graphite/LiMn2O4 batteries made by LG Chem.