The choice of a battery for an application is dictated by various performance, cost and usability metrics. Common performance metrics include voltage, power density, energy density and shelf life. Often times you will have to trade one performance metric against another. Chemistry obviously plays a critical role in defining most of these metrics. In part 1 of this blog post we discussed the role of chemistry in primary batteries, while in this article we discuss secondary batteries.
Secondary Battery Chemistry
Let’s start with the oldest (157 years and counting) technology still in active use: the lead-acid battery. This 2.0 V chemistry (Pb⇒ PbO2) has an unusual feature: when fully discharged, both electrodes are the same compound.
Pb (cathode) + PbO2 (anode) + 2H2SO4 ⇒ 2PbSO4 + 2H2O
The buildup of PbSO4 on the electrodes is the limiting factor in PbH+ batteries: recharging will bring little joy, due to the insolubility of large PbSO4 crystals in dilute acid. Repeated discharge of a PbH+ cell limits operation to at most 3 cycles. Common use of PbH+ technology is, of course, automotive starter batteries, and typical usage is 40% global share), divided into 3 cathode types with different intercalation mechanisms (intercalation = Li insertion). This property is related to the crystal structure of cathode materials, depicted in the figure below.
Figure 1. Schematics of Li-ion cathode materials’ structures, showing a) the 2-dimensional access of Li+ planes in LiMO2 (M = Co, Ni, Mn); b) 3-dimension Li+ channels in cubic spinel (LiMn2O4); and c) unidirectional Li+ tunnels in LiFePO4’s olivine framework. Li is represented by spheres, MO6 octahedra and PO4 tetrahedra by triangles. M. Doeff, Encyclopedia of Sustainability Science and Technology (2011), Robert A. Meyers, Ed.
Layered LiMO2 species represent the past and future of LIBs. Initially, M = Ni, but this proved unsafe and susceptible to overheating when charged, due to the presence of unavoidable diffusion-retarding NiO impurities. This less-than-successful material was quickly replaced with LiCoO2 25 years ago, which is still the common intercalant in small cells for portable devices. Co and Ni are very much alike (electro)chemically, and research proved that as little as 15% Co replacing Ni in LiNiO2 prevented NiO contaminants and mitigated thermal excursions while boosting energy output >20%. Unfortunately, heat produced by overcharging was still too energetic for safety, until a few years ago, when LiNi0.85Co0.10Al0.05O2 appeared in commercial cells (including Tesla’s!). The Al holds the oxygen planes in place even after ~70% of the Li has been extracted during charge, and prevents O2 loss.
Consider what happens in an overcharged LIB with Li metal oxide cathode. As the average metal oxidation state gets above +3.5 (i.e., Li0.5MO2), the structure becomes less stable and more willing to give up some O2 (highly reactive nascent, ie, atomic oxygen), reverting to stable MO. Overcharging also requires more energy, which is manifested by cell heating. Further, LIB electrolytes comprise LiPF6 dissolved in organic solvents, and the entire system is in a closed cell. In summary, you have heat, oxygen, a fuel source, and a sealed container, which is one way to describe a grenade!
Researchers found that LiMO2 materials had a reduced exotherm with oxygen evolution when M was a combination of Ni, Co, and Mn, that is, molecular breakdown generated less heat than Li(Ni,Co)O2 compounds. These trimetallic species, LiNiaCobMncO2 (labelled NCM; a+b+c=1), incorporate Mn+4 as the stabilizing cation, as it does not participate in the usual redox reactions. Why aren’t all three cations in a +3 oxidation state? Because at LiMO2 preparation temperatures of 750-900 0C the oxidizing energy of Ni+3 is strong enough to force an equivalent amount of Mn+3 to Mn+4, and of course, the Ni is reduced to Ni+2. Most of the energy in NCM cells is derived from the redox reaction Ni+2 ↔ Ni+4, and while some originates from Co+3 ↔ Co+4, cobalt’s primary role is to facilitate Li+ diffusion within the NCM particles.
When a=b=c, this NCM material generates about the same energy as LiCoO2 (ca 575mWh/g), but is safer and uses only 1/3 as much expensive Co. The Co content can be reduced without impacting performance: in fact, LiNi0.45Co0.10Mn0.45O2 takes advantage of the greater energy of the Ni couple, and yields ca 700mWh/g. Asymmetric NCMs – [Ni] > [Mn] – have performances roughly equivalent to LCO, but are more prone to NiO impurities and thus have shorter cycle life.
The next generation of NCM compounds, now in advanced development, is non-stoichiometric with respect to Li (also known as Li-rich compounds, LR-NMC) and features higher voltage (4.65V). The Mn gets involved in cell redox reactions, and coupled with high Ni content, LR-NCM materials can produce >1000mWh/g, depending on the composition, when cycled over a 2.5-4.7 V range. Unfortunately, the extra Li causes “slippage” into a more stable spinel structure, which lowers the voltage curve and causes >13% energy loss in only 20 cycles, with a continuing decline. When this problem is solved, electric vehicles with 250+ mile ranges will become commonplace.
The cubic structure (b, above) is most commonly associated with LiMn2O4 (aka spinel), another 4V cathode material that, at best, produces about 460 mWh/g. While more stable when charged than LiCoO2 – and a lot cheaper – spinel’s energy is derived from the Mn+3 ↔ Mn+4 couple, and there’s only one Mn+3 per molecule. Further, acid, a ubiquitous impurity in LiPF6 electrolytes, catalyzes Mn+3 disproportionation, producing Mn+2 and Mn+4. The former dissolves into the electrolyte and diffuses to the anode, where it combines with reduced electrolyte anions and forms insulating salts, blocking Li+ transport into and out of the electrode. The quadrivalent ion precipitates (as MnO2) within the spinel lattice, disrupting the intercalation channels. Performance fade can exceed 1%/cycle, not a viable situation.
The solution was to replace up to 5% of the Mn+3 with a non-oxidizable cation – Al+3 is the favorite, because it doesn’t distort the crystal lattice – that mitigates Mn+3 loss and extends working life to >500 cycles. The side effect reduces energy output to ca 400 mWh/g, which restricts spinel’s usage in many high-energy applications. Excess Li, as much as 15%, has the same effect: longer life but lower energy as Li content climbs. Finally, Ni+2 provided some benefit, but it wasn’t until all the Mn+3 was replaced that performance fade was controlled. The result, LiNi0.5Mn1.5O4, is a 4.6V spinel (Ni+2 ↔ Ni+4) whose output is ca 550 mWh/g. There is a premium on purity: if the Ni:Mn ratio is not exactly 1:3, side reactions between high oxidation state metal oxides and the electrolyte truncate working life.
The last cathode material, LiFePO4 or LFP, is an exception to oxide intercalants, with an olivine structure. Note that in schematic c, above, this structure is considerably more crowded than the oxide lattices. The added weight of the phosphate groups (versus oxide) and the low voltage of the Fe+2 ↔ Fe+3 couple (3.35V) lead to only 525 mWh/g output, not competitive with LiMO2 cathode materials. However, the phosphate groups prevent oxygen release from delithiated LFP, making this by far the safest cathode intercalant, a very attractive feature for hybrid vehicles and portable tools. LFP is also the longest-lived of its contemporaries, commonly exceeding 2000 cycles with deep-discharge usage and >10,000 cycles of pulse operation.
Let’s compare the three cathode structures for power capability: that is, how do they perform at high discharge rates? Li channels in cubic spinels parallel all three axes, LiMO2, including NCM, species have Li x,y planes, and olivines have Li conduits only in the y direction. Not surprisingly, spinels have the best power capability, since Li can enter the particles from any direction. This explains why spinel cells are most commonly found in e-bikes and portable tools that need acceleration/pulsed output. LiMO2 compounds have intermediate-to-poor rate performance, due to the restrictive 2-dimensional particle entry points. It’s no surprise that these materials are seen in high energy/long run time applications, from portable communication devices to electric vehicles. Finally, LFP is the fussiest cathode material, and is not a viable material even at 1-hour charge/discharge cycles. Why, then, is LFP so successful in high rate environments? Because during preparation, the LFP is coated with a very conductive amorphous carbon skin, only a few nm thick, which facilitates Li diffusion into the micron-sized particles. While LiMO2 species function poorly above 2C (0.5 hour) discharge, spinels are good to 5C, and there are reports of LFP providing workable energy at 30C (2 minute) rate.
Lastly, a few words about Li-ion anode materials. The figure below models battery activity with layered cathodes and graphite anodes, until recently the only Li-storage species in LIBs. Nature has conveniently arranged for graphene layers in graphite to be 3.4Å apart, just enough for Li to reversibly squeeze in. This is so easy that the potential for this storage is about 0.05 V: remember, Vcell = Vcathode – Vanode. Also, note the 6-membered carbon rings: each Li atom is centered on the ring, half way between layers. The valence electron on Li becomes delocalized into the ring, explaining the yellow color of Li-infused graphite. This electron shift greatly reduces the reactivity of Li: on exposure to water, LiC6 bubbles slowly, while Li metal sparks and ignites the hydrogen by-product.
Figure 2. Structures of layered LiCoO2 cathode and graphite anode in common LIB. From http://tymkrs.tumblr.com/post/7846476684/lithium-ion-battery-how-does-it-work.
In extreme rate-capable cells designed for power applications, Li4Ti5O12 (LTO) might be the anode. This spinel structure is stable for thousands of cycles and capable of 100C (36 second) discharge. The downside to LTO is its Li storage potential of 1.35V, which reduces LFP/LTO cell output to 2V and 300mAh/g: even matched with LiNi0.5Mn1.5O4 or LR-NCM, cell potentials are 3.2V, while gravimetric energy densities are ~440 and 800mAh/g, respectively.
Silicon is an emerging anode for LIBs, with roughly 10X the storage capacity of graphite. This is an alloying anode, not an intercalant, and the Li storage potential is