How batteries work Chapter 12 Negative electrodes

Chapter 12

Negative electrode materials: graphite, silicon, lithium metal and anode-free cells

What could replace graphite, and why has it not happened yet?

Where we are. In chapter 11 we built real electrodes out of particles, carbon, binder and pores, and our ion made the round trip through them. The round trip works; now let us push it further. We begin with the negative electrode, the side that holds the lithium when your cell is charged: first graphite, which made the first lithium-ion cell possible, then the materials that might replace it, silicon, conversion materials and lithium metal itself, and the ions that might one day replace lithium. Then we cross to the positive electrode: the oxides in today’s cells, why nickel-rich NMC trades stability for capacity, what comes next, and the organic molecules that do without transition metals. Along the way we answer the question above.

Figure 12.1 Three panels. At the top, bars of the theoretical capacity of classes of negative electrode material from a 2026 review (nanostructured carbons, metal oxides, nitrides, sulfides and phosphides) and of alloying elements (tin, indium, germanium, phosphorus, silicon) on the same capacity axis. In the middle, a map of negative electrode materials: potential against lithium from 0 to 2 volts against specific capacity on a log scale from 100 to 5000 milliampere-hours per gram, with graphite, titanium and niobium oxides, silicon, tin, germanium, antimony, iron, manganese and molybdenum compounds and lithium metal. Beside it, bars of the specific energy of the chosen material paired with a positive electrode, against graphite, drawn as horizontal bars at the bottom. Intercalation hosts are drawn as circles, alloys as diamonds, conversion materials as squares and lithium metal as a metallic disc.
intercalation (circle)alloy (diamond)conversion (square)lithium metal
The map of negative electrodes, and what extra capacity really buys. Try this: compare silicon and lithium metal; then try Li4Ti5O12 and see what its voltage costs.
SourceClasses, theoretical capacities and working potentials after Rahman et al. 2025, Tables 1 and 2 and section 2 (graphite 372 mAh/g at 0.15 V; Li4Ti5O12 175 at about 1.55 V; TiNb2O7 about 350 and TiO2 330, niobium oxides at about 1 to 2 V and TiO2 at 1.5 V; Si 3579, Sn 994, Ge 1624, Sb 660; Fe2O3 1007 and MnO2 1233 at 1.0 to 1.5 V; MoS2 670 at 1.1 to 1.8 V; FeP about 900 at 0.5 to 1.0 V; lithium 3860 at 0 V). Alloy potentials, 0.2 to 0.8 V, after Goodenough and Park 2013; dots mark the middle of each range. The energy of the pair, (Vpos − Vneg) × QposQneg/(Qpos + Qneg), counts the two active materials only, against a positive electrode of 200 mAh/g (about what NMC811 gives, Marie and Gifford 2023) at an illustrative 3.8 V: this page’s own working. Top strip: theoretical capacities after Faruque et al. 2026, Table 6 (nanostructured carbons up to 1750 mAh/g; metal oxides 375 to 1170, nitrides 400 to 1300, sulfides 447 to 1230, phosphides 700 to 1800; Sn 960, In 1012, Ge 1384, P 2596, Si 4200 mAh/g). For silicon, germanium and tin the review’s values differ from those used on the map (3579, 1624 and 994).

Let us start with what the negative electrode has to do. It stores the lithium while the cell is charged and gives it back on discharge. It was the negative electrode that made the first lithium-ion cell possible: graphite held lithium reversibly and stayed stable in carbonate electrolytes, where lithium metal did not.1 Graphite’s 372 mAh/g now limits how much energy a cell can hold, and the search for something better has produced four families of material, named for how they store lithium: intercalation hosts, alloys, conversion materials and lithium metal.1

What would a better material have to offer? A good negative electrode needs a high reversible capacity, a low potential, fast kinetics, low cost, stability against the electrolyte, a stable SEI and a light environmental footprint.1 On the first two counts lithium metal and silicon stand out; both still face serious problems at the level of the material, the electrode and the cell.1

Worked example. Ten times the capacity, but not ten times the energy. Take a positive electrode like NMC811, which gives about 200 mAh/g,2 at an illustrative average of 3.8 V. The same charge passes through both electrodes, so a gram of the pair passes Q = QposQneg/(Qpos + Qneg). graphite: Q=200×372200+372=130 mAh/g, at 3.8 − 0.15 = 3.65 V: 475 Wh/kg lithium metal: Q=200×3860200+3860=190 mAh/g, at 3.8 V: 723 Wh/kg A factor of 1.5, because the positive electrode now limits the cell. The gain from lithium metal over today’s lithium-ion cells is put at up to a factor of two.1

A negative electrode should sit low and store a lot, but past a point the positive electrode sets the limit.

A 2026 review gives the theoretical capacities by class (the strip at the top of figure 12.1): nanostructured carbons such as nanotubes, graphene and porous carbons up to 1750 mAh/g; metal oxides 375 to 1170, nitrides 400 to 1300, sulfides 447 to 1230 and phosphides 700 to 1800, all by conversion; and, among alloying elements, tin 960, indium 1012, germanium 1384, phosphorus 2596 and silicon 4200 mAh/g.3 Its value for silicon is higher than the 3579 mAh/g on our map, a reminder to check which lithiated end-state a quoted capacity assumes. The review also notes that Li4Ti5O12 cycles well because it neither expands nor contracts as lithium goes in and out, at the price of its relatively high potential, and that composites of carbon nanotubes with silicon or metal oxides have shown over 1000 mAh/g with good retention over hundreds of cycles.3

Figure 12.2 The potential of the graphite surface against its state of charge, at rest (dashed) and during charging (solid), with the region below 0 volts shaded where lithium plates. Sliders set the charging rate and temperature. Beside it, bars of the margin above 0 volts for graphite near full charge, a niobium oxide and lithium titanate, with the charging pull drawn as a dashed line.
graphite at rest (illustrative)while charginglithium plates below 0 V
Fast charging: how close can the negative electrode come to 0 V? Try this: at 25 °C find the rate at which plating starts; then cool the cell to 0 °C and find it again.
SourcePlating when lithium is intercalated faster than it is supplied, the low lithiation potential of graphite, more plating in the cold, and titanate and niobium oxides at 1 to 2 V avoiding plating and SEI formation, after Rahman et al. 2025, sections 2(i), 3 and 5. The graphite curve and the pull, 15 mV per C at 25 °C rising with an activation energy of 40 kJ/mol, are this page’s illustrative model, not data; the niobium bar uses the lower end of their 1 to 2 V.

Charging fast

Plating, which we just met in the cold, is also a risk when we charge quickly. Electric cars, drones, electric aircraft and grid storage all want faster charging; industry aims to reach 80 % of the charge in under 15 minutes without losing cycle life or safety.1 On the negative electrode three things go wrong. Lithium can arrive faster than it can enter the graphite and plate as metal, a risk made worse by graphite’s low potential; the particles are stressed and crack; and the heat destabilizes the SEI.1

The remedies run from material to cell: smaller particles and porous structures; thin, porous electrodes of low tortuosity; electrolytes with a LiFSI salt or an FEC additive, which improve the SEI and reduce plating; and, at cell level, the right ratio of negative to positive capacity (the N/P ratio), the charging protocol and cooling.1 Silicon-based negative electrodes are being developed for fast charging too: one company reports silicon nanowires charged from 0 to 80 % in under 6 minutes, another columnar silicon grown on copper charged to 80 % in 7 minutes at 1350 Wh/L.1

The other route is to move away from 0 V. Lithium titanate, Li4Ti5O12, works at about 1.55 V, where little SEI forms, and changes volume by only about 0.2 %, a “zero-strain” material: it is safe, long-lived and fast, but stores only 175 mAh/g, conducts poorly and can give off gas (H2, CO2, CO) by decomposing the electrolyte on its surface, a mechanism not yet well understood.1 Niobium oxides such as Nb2O5, TiNb2O7 and Nb14W3O44 also work at 1 to 2 V, store more (about 200 to 350 mAh/g), change little in volume and charge very fast; they are already being made at industrial scale for high-power cells.1

Worked example. What the safe voltage costs. Keep the same positive electrode, like NMC811, which gives about 200 mAh/g,2 at an illustrative average of 3.8 V: graphite: 3.65 V × 130 mAh/g = 475 Wh/kg of the pair Li4Ti5O12: Q=200×175375=93 mAh/g, at 3.8 − 1.55 = 2.25 V: 210 Wh/kg Less than half the energy, in exchange for a volt and a half of margin against plating. That is why titanate serves where power and life matter more than range.
Figure 12.3 Three particles breathing as lithium goes in and out: lithium titanate almost unchanged, graphite about 10 percent larger, silicon nearly four times its volume. Each has an SEI rim; on silicon the rim cracks on every swelling and grows thicker, and the particle cracks, unless it is given room to swell. Below, a bar shows the capacity of a graphite and silicon blend for the chosen silicon fraction.
titanategraphitesiliconSEIcarbon framework: room to swell
Silicon breathes, and its SEI pays for it. Try this: watch the silicon for three cycles, then give it room to swell; slide the silicon fraction from 5 to 20 wt%.
SourceVolume changes (about 0.2 % for Li4Ti5O12, about 10 % for graphite, about 280 % for silicon), cracking of the SEI and its regrowth, pulverization and delamination, the mitigations, and commercial blends with 5 to 10 wt% silicon, after Rahman et al. 2025, sections 2(ii), 4 and 5 and Figs. 4e and 5c–d; silicon’s 3579 mAh/g from their Table 2. Particle radii follow the cube root of the volume; when the cracks appear is schematic. The blend capacity is a simple mass-weighted mixture: this page’s working.

Silicon: ten times the capacity, four times the volume

Titanate trades energy for margin. Silicon makes the opposite trade. It stores 3579 mAh/g, close to ten times graphite, at a low potential, and it is cheap. Its problem is volume: it swells by about 280 % when fully lithiated. The particles pulverize, lose contact with each other, and the electrode can come away from its current collector; the SEI breaks on every swelling and forms again on the fresh surface, using up lithium and electrolyte, so the first-cycle efficiency is low and the capacity fades fast. Silicon also conducts electrons poorly.1

Nanoparticles, porous and hollow spheres, nanowires, yolk–shell and pomegranate-like structures all help, though small particles cost volumetric energy. Binders that hold the contact, perhaps self-healing polymers, could be the breakthrough. Today’s best cells already use graphite with 5 to 10 wt% silicon; more than 20 % is the industry’s next step, trading energy against cycle life.1 The market for silicon negative electrodes, about $279 million in 2023, is projected to reach $15 billion by 2035.1

Worked example. What 280 % does to a particle, and what a little silicon does for a blend. volume after = 1 + 2.8 = 3.8 times; diameter after = 3.81/3 = 1.56 times: a 100 nm particle grows to about 156 nm and back on every cycle 8 wt% silicon in graphite: 0.92×372+0.08×3579=629 mAh/g, 1.7 times graphite The coating, the binder and the SEI on each silicon particle all have to follow that breathing.
Figure 12.4 A metal oxide particle that turns into metal grains embedded in lithium oxide as lithium goes in and reforms as it comes out; beside it, the voltage for lithium in and lithium out against lithium content, separated by a hysteresis set with a slider, with the area between them shaded and graphite's flat line for scale.
metal oxide (Fe2O3)metal grainsLi2Olithium inlithium outenergy lost per cycle
Conversion: high capacity, paid for with a volt of hysteresis. Try this: halve the hysteresis and read how much of the energy comes back.
SourceThe conversion reaction, the capacities of Fe2O3, MnO2 and MoS2, hysteresis of about 1 V from the interconversion of several solid phases, and the drawbacks, after Rahman et al. 2025, sections 2(iii) and 4, Table 1 and Fig. 5e–f; two to three times the capacity of carbon after Tarascon and Armand 2001. The voltage curves are schematic around 1.25 V; the round-trip efficiency against a 3.8 V positive electrode, (3.8 − V − h/2)/(3.8 − V + h/2), is this page’s estimate.

The third mechanism goes further still. A conversion electrode is not a host at all. Lithium reduces the compound to metal grains embedded in a lithium compound, for an oxide MxOy + 2yLi+ + 2ye− ⇌ yLi2O + xM, and on the way back the compound forms again. Oxides, sulfides, phosphides and fluorides work this way, with high capacities: about 1000 mAh/g for Fe2O3, 1230 for MnO2, 670 for MoS2 and 900 for FeP.1 Nano-sized transition-metal oxides were shown to store two to three times the capacity of carbon in 2000.4

The drawbacks are four: the particles swell and pulverize; they conduct poorly (the Li2O formed is an insulator); the SEI is unstable; and the voltage hysteresis is about 1 V, because several solid phases must form and dissolve and oxygen and metal atoms must rearrange, not just lithium.1 On top of that, the first cycle loses about 30 %, and the wasted energy is the main obstacle.5 Nano-sized particles shorten the paths and spread the stress, but increase the surface that reacts with the electrolyte.1

The same chemistry also appears on the other side of the cell. Transition-metal fluorides work the same way as positive electrodes, with theoretical capacities of 500 to 1500 mAh/g, and sulfur, at 2.4 V, is also a conversion electrode.4,5

Figure 12.5 A lithium-metal electrode cycling against a positive electrode. Each plating leaves the surface rougher; at higher current density the deposit grows as whiskers from its tips; pieces that break off on stripping collect as dead lithium; the deposit grows toward the separator. With a ceramic separator most of the deposit stops, but filaments pass along grain boundaries.
lithium metaldead lithiumSEIseparatorceramic separatorpositive electrode
Lithium metal: dendrites and dead lithium. Try this: play ten cycles at 1 mA/cm2, then reset the page and play at 4 mA/cm2; then switch to the ceramic separator.
SourceDendrites, the space-charge, tip and defect explanations, dead lithium, its thickening from about 2 µm to about 100 µm, SEI on fresh surfaces, short circuits, the Young’s modulus of lithium (about 4.9 GPa) and dendrites through the grain boundaries of ceramic electrolytes, after Rahman et al. 2025, sections 6 and 7.1 and Fig. 10; cycle life in 2004 after Winter and Brodd 2004, section 2.5. The growth follows a random rule weighted toward tips, drawn only as “rough, whisker-like”; the current density scale and cycle counter are not predictions.

Lithium metal

We now reach the far corner of the map. Lithium metal is the most attractive negative electrode of all: 3860 mAh/g, the lowest potential, and nothing in it but active material. It was studied long before graphite, and kept out of commercial cells by dendrites.1 In 2004 lithium-metal cells reached about 100 to 150 cycles, against the 300 a commercial cell needed.6 Its problems are dendrites, an unstable SEI, low Coulombic efficiency, dead lithium, volume changes and pulverization, and reaction with the electrolyte.1

Why does lithium plate back so unevenly? Several models explain the dendrites. In the space-charge model, when plating outruns ion conduction at high current, the Li+ concentration near the surface falls to zero and dendritic growth begins. Cracks in the SEI, tips and surface defects concentrate the electric field, and lithium grows fastest there. On stripping, pieces break off the dendrites and lose contact: dead lithium, which in one study thickened from about 2 µm to about 100 µm within a few dozen cycles, lowering the efficiency and raising the impedance. Whiskers that pierce the separator short the cell, with a risk of fire.1

Most of these problems start at the interface, not in the bulk metal, which is why coatings, artificial SEI layers and new electrolytes are where the work is.1 The interface is studied with surface spectroscopy (XPS, soft X-ray absorption), optical and electron microscopy, cryo-electron microscopy, which preserves fragile interphase compounds, and synchrotron X-ray diffraction and pair-distribution analysis, which identified the much-debated LiH in the interphase on lithium metal.1

Lithium metal stores the most, but it plates back unevenly, and that, not the chemistry of storage, is what limits it.

Figure 12.6 Capacity against cycle number up to 500 cycles for a lithium-metal cell with no excess lithium (anode-free), one times excess and three times excess, for a Coulombic efficiency set with a slider; a dashed line at 80 percent marks the end of life.
anode-free1× excess3× excess
How many cycles? Coulombic efficiency against lithium in reserve. Try this: find the efficiency at which the anode-free cell lasts 100 cycles; then the one at which it lasts 500.
SourceAnode-free cells plating the positive electrode’s lithium onto the current collector, with no excess to make up for losses, after Rahman et al. 2025, sections 7 and 7.2; the excess needed for N cycles, R = N(1 − E), after Brandt 1994. The curves, capacity unchanged until the excess is spent and then falling by CE per cycle, are this page’s model.

If every cycle loses a little lithium, how much lithium should the cell carry in reserve? An anode-free cell carries none: it is assembled with no lithium on the negative side at all, only a copper current collector. On the first charge the lithium in the positive electrode is plated onto the copper, and on discharge it goes back. That is the highest energy a lithium-metal cell can reach, and the simplest to make; but with no lithium in reserve, every loss shows directly as lost capacity.1

Worked example. Why early lithium-metal cells carried far more lithium than they used. Define the lithium cycling efficiency E from the charge stripped each cycle and the excess of lithium used up; Brandt’s relation then gives the relative excess needed for N cycles, R = N(1 − E). LiClO4 in propylene carbonate, E ≈ 60 %: for just 100 cycles, R = 100 × 0.40 = an excess of 40 times the lithium cycled Moli Energy, late 1980s, E > 99 %: about 300 cycles, but only by carrying three times the lithium: R = 300 × 0.01 = an excess of 3 times and for an anode-free cell to keep 80 % for 1000 cycles: E1000≥0.8, so E≥0.81/1000=99.978 % The efficiency, not the capacity, decides whether lithium metal can be used.7
Figure 12.7 An anode-free cell: a copper current collector, electrolyte, separator and positive electrode, with lithium plating onto the copper and back. Steps show three fixes: a graphene-oxide film on the copper, an additive in the electrolyte and a porous copper collector, each with the results reported.
copperplated lithiumartificial SEI (graphene oxide)additivelithium in the positive electrode
Making anode-free cells work: three places to intervene. Try this: step through the four callouts and, for each fix, find the number that shows it worked.
SourceAdvantages, challenges and strategies (Table 4); graphene oxide on copper (Tadesse et al.), KNO3 (Sahalie et al.), dual-salt LiDFOB/LiBF4 (Weber et al.), localized high-concentration electrolytes, fluorinated solvents and a porous copper collector (Yun et al.), with the numbers each reported, after Rahman et al. 2025, sections 7.2 and 7.3 and Fig. 11. Drawn schematically; the reported numbers come from different cells and tests and should not be compared with each other.

The efficiency, then, is what we must raise, and three kinds of fix are being tried. An artificial SEI, made beforehand on the copper or formed in the cell, guides the lithium to deposit evenly. Electrolyte engineering: additives that build a better SEI, two-salt electrolytes, localized high-concentration electrolytes that mix a dissolving solvent with a diluent that does not dissolve the salt, and fluorinated solvents that are stable against both lithium and nickel-rich positive electrodes. And modified current collectors, such as three-dimensional porous copper that spreads the current and gives the lithium room.1

Lithium metal was long thought too reactive to cycle in a liquid electrolyte. The new electrolytes have changed that: they now allow promising cycling with little excess lithium and heavily loaded positive electrodes.1

References for this chapter

Numbers in the text refer to this list. All references of the course, with the chapters that use each one.

  1. 1 M. M. Rahman, U. Nisar, A. Abouimrane, I. Belharouak, R. Amin, "Valuation of Anode Materials for High-Performance Lithium Batteries: From Graphite to Lithium Metal and Beyond", Electrochemical Energy Reviews 8, 2025, 14. doi:10.1007/s41918-025-00249-w
  2. 2 J.-J. Marie, S. Gifford, "Developments in lithium-ion battery cathodes", Faraday Insights, Issue 18 (The Faraday Institution), 2023, September 2023. link
  3. 3 O. Faruque, M. J. Isalm, O. Faruq, M. M. Alam, "A comprehensive review of past and present developments of Li-ion batteries", Discover Sustainability 7, 2026, 881. doi:10.1007/s43621-026-02920-8
  4. 4 J.-M. Tarascon, M. Armand, "Issues and challenges facing rechargeable lithium batteries", Nature 414, 2001, 359-367. doi:10.1038/35104644
  5. 5 D. Shanmukaraj, P. Ranque, H. Ben Youcef, T. Rojo, P. Poizot, S. Grugeon, S. Laruelle, D. Guyomard, "Review—Towards Efficient Energy Storage Materials: Lithium Intercalation/Organic Electrodes to Polymer Electrolytes—A Road Map (Tribute to Michel Armand)", Journal of The Electrochemical Society 167, 2020, 070530. doi:10.1149/1945-7111/ab787a
  6. 6 M. Winter, R. J. Brodd, "What Are Batteries, Fuel Cells, and Supercapacitors?", Chemical Reviews 104, 2004, 4245-4269. doi:10.1021/cr020730k
  7. 7 K. Brandt, "Historical development of secondary lithium batteries", Solid State Ionics 69, 1994, 173-183. doi:10.1016/0167-2738(94)90408-1

Cite this page

Using this page in teaching, a thesis or an article? Cite it as follows; the date is the last update of the course.

Shyma Sajeevan, A. (2026). Negative electrode materials: graphite, silicon, lithium metal and anode-free cells. How batteries work. https://asajeevan.github.io/batteries/negative-electrodes/

↑